EVALUATION OF OPERATIONAL PERFORMANCE AND ENVIRONMENTAL IMPACT OF A COMMERCIAL SCALE ANAEROBIC DIGESTER UTILIZING MULTIPLE FEEDSTOCKS By Fahmi Dwilaksono A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Biosystems Engineering—Master of Science 2022 ABSTRACT EVALUATION OF OPERATIONAL PERFORMANCE AND ENVIRONMENTAL IMPACT OF A COMMERCIAL SCALE ANAEROBIC DIGESTER UTILIZING MULTIPLE FEEDSTOCKS By Fahmi Dwilaksono Food waste and livestock manure become some the of major sources that contribute to greenhouse gas (GHG) emissions in the U.S. Utilizing manure and food wastes as biogas feedstocks through the anaerobic digestion (AD) process can improve renewable energy production while reducing the impact of climate change due to GHG emission from untreated organic wastes. This study evaluated the operational performance of Michigan State University’s commercial South Campus Anaerobic Digester (SCAD) as well as the environmental impact during its operation from 2014 to 2020. Evaluation of feedstock supplies quantity and output parameters of SCAD was conducted to understand the operational performance of the digester. A life cycle assessment (LCA) was done to know the environmental impact of SCAD by comparing it to the conventional waste management methods. Technoeconomic analysis was conducted to know the financial feasibility of SCAD as a commercial digester. The result shows that during its operation from 2014 to 2020, SCAD has processed 159,145 metric tons of feedstock from 18 different organic wastes to produce 15,165,156 kWh of electricity for the MSU community. LCA results show that the AD system possesses fewer environmental burdens in both global warming potential (GWP) and water eutrophication potential (WEP) compared to the conventional system. Technoeconomic analysis reveals that SCAD needs 21.5 years to accomplish its payback period, which is considered quite economically competitive. Economic sensitivity analysis shows that electricity becomes the most sensitive parameter to affect the payback period. This thesis is dedicated to my mother, Mrs. Fatmah, and my father, Mr. Muhammad Ali Ismail. Thank you for always believing in me. iii ACKNOWLEDGEMENTS This thesis was conducted at the Department of Biosystems and Agricultural Engineering Michigan State University within nine months of intensive rigorous work. This accomplishment would not have been possible without supervision, professional support, and encouragement of a lot of great individuals. I would like to express my deepest gratitude to Dr. Dana Kirk for giving me relentless positive and constructive supervision to broaden my love in the bioenergy field. Thank you very much for accepting me as who I am and guiding me to grow both personally and professionally while pursuing my master’s degree at MSU, especially during the disruption of COVID-19 pandemic. I am also beyond grateful for Dr. Wei Liao and Dr. Yan Liu for being great committee members and providing great support and guidance during this entire journey. I would grab this chance to deliver my huge gratitude to Fulbright Scholarship, Institute of International Education (IIE), and American Indonesian Education Foundation (AMINEF) for giving me this wonderful opportunity. Not only gaining so much experience from world class institutions, but I am also having the chance to giving back to the community in the United States; to share my stories and backgrounds as a young individual coming thousand miles away from a small island in eastern Indonesia called Sumbawa and to contribute a little more diversity at MSU community in East Lansing. I would like to thank my parents, family, and friends for always giving me tons of love and support, and always being my rock during my ups and downs. Thank you for being on this journey together and making this dream happen. I would like to give a huge thank you to my lab partner and fellow graduate student, María Inés Barrios Arosemena for her support, knowledge, a lot of country, BTS, Latin, and Christmas iv music in the lab, also love for Disney during our graduate studies. Thank you very much to Yousef Abdalaal, my fellow graduate student who has helped me so much with the statistical analysis of this study. I would like to thank Louis Faivor and Dalton Brenke for providing a huge data set for this study, also fun lab tours at South Campus Anaerobic Digester (SCAD) that provide me further knowledge and moral support to accomplish this study. I would also like to thank all Anaerobic Digestion Research and Education Center (ADREC) technicians (Brianna, Caelen, Noah, Mary, and Nico) for their help and support while accomplishing this study and sharing a lot of stories in the lab. Thank you so much for the MSU Graduate School and College of Agriculture and Natural Resource’s Office of Academic and Student Affairs for awarding me CV-19 Emergency Funding. I would also like to thank MSU Asian Studies Center for awarding me Dr. Delia Koo Endowment Awards. Those supports are meaningful to support my life as an international student during COVID-19 pandemic. Finally, I would like to thank my beloved department, the Department of Biosystems and Agricultural Engineering, for giving me one of the best experiences in my life. Thank you for having me during my graduate study and providing a great platform for growing up as a better engineer, and simply as a better human being. v TABLE OF CONTENTS LIST OF TABLES ......................................................................................................................... ix LIST OF FIGURES ..................................................................................................................... xiii KEY TO ABBREVIATIONS ..................................................................................................... xvii CHAPTER 1. INTRODUCTION ................................................................................................... 1 1.1 Problem Statement ................................................................................................................ 1 1.2 Goal and objectives ............................................................................................................... 2 CHAPTER 2. LITERATURE REVIEW ........................................................................................ 3 2.1 General introduction – global drivers ................................................................................... 3 2.1.1 Societal change .............................................................................................................. 3 2.1.2 Climate change............................................................................................................... 5 2.1.3 Food waste diversion ..................................................................................................... 6 2.1.4 Manure management ...................................................................................................... 8 2.1.5 Renewable energy ........................................................................................................ 10 2.1.6 Policy drivers – focus in the U.S. ................................................................................ 12 2.2 Anaerobic digestion ............................................................................................................ 20 2.2.1 Introduction .................................................................................................................. 20 2.2.2 Feedstock ..................................................................................................................... 21 2.2.3 Testing and analysis – Operation data ......................................................................... 24 2.3 Michigan State University South Campus Anaerobic Digester (MSU SCAD).................. 31 2.4 Life Cycle Assessment (LCA) ............................................................................................ 33 2.4.1 What is it? .................................................................................................................... 33 2.4.2 Process ......................................................................................................................... 34 2.4.3 Parameters .................................................................................................................... 35 2.5 Revenue Value .................................................................................................................... 36 CHAPTER 3. MATERIAL AND METHODS ............................................................................. 38 3.1 Manure Feedstock ............................................................................................................... 38 3.2 Food Waste Feedstock ........................................................................................................ 38 3.3 Real-time (commercial system) sampling methods ............................................................ 38 3.3.1 Gas quantity and quality, and electricity production ................................................... 38 3.3.2 Digestate, filtrate, and solids sampling ........................................................................ 39 3.3.3 Temperature sampling ................................................................................................. 40 3.3.4 pH sampling ................................................................................................................. 40 3.4 Laboratory Analysis ............................................................................................................ 40 3.4.1 Total Solids and Volatile Solids .................................................................................. 40 3.4.2 Chemical Oxygen Demand .......................................................................................... 41 3.4.3 Total Suspended Solids / Volatile Suspended Solids .................................................. 41 3.4.4 pH and Electrical Conductivity .................................................................................... 42 3.4.5 Alkalinity and Volatile Fatty Acids ............................................................................. 42 vi 3.4.6 Total Nitrogen .............................................................................................................. 44 3.4.7 Ammonia...................................................................................................................... 44 3.4.8 Total Phosphorus ......................................................................................................... 44 3.4.9 Other laboratory tests ................................................................................................... 45 3.5 Statistical Analysis .............................................................................................................. 45 CHAPTER 4. ANALYSIS OF SCAD OPERATIONAL DATA ................................................. 47 4.1 Feedstock Amount .............................................................................................................. 47 4.1.1 Yearly total................................................................................................................... 47 4.1.2 Monthly Average ......................................................................................................... 53 4.1.3 Feedstock Distribution ................................................................................................. 59 4.1.4 Feedstock Characterization .......................................................................................... 68 4.2 Biogas and Electricity Production ...................................................................................... 71 4.3 SCAD Output Data Distribution ......................................................................................... 72 4.4 SCAD Laboratory Analysis ................................................................................................ 84 4.4.1 Solids............................................................................................................................ 85 4.4.2 pH and Electronic Conductivity................................................................................... 86 4.4.3 Chemicals ..................................................................................................................... 87 4.4.4 Fiber Results ................................................................................................................ 88 4.4.5 Other laboratory results................................................................................................ 89 4.5 Discussions ......................................................................................................................... 89 4.5.1 Correlation between feedstock and biogas production ................................................ 89 4.5.2 The impact of OLR, HRT, and digester pH and temperature on biogas production and methane percentage ............................................................................................................... 93 4.5.3 Digester performance from year to year ...................................................................... 98 CHAPTER 5. LIFE CYCLE ASSESSMENT ............................................................................ 103 5.1. Introduction ...................................................................................................................... 103 5.2 Goal, Scope, and Functional Unit for Life Cycle Assessment ......................................... 103 5.3 Life Cycle Data Inventory ................................................................................................ 105 5.4 Data Quality Evaluation.................................................................................................... 110 5.5 Mass and Energy Balance of the process at different months and years .......................... 111 5.6 Impact Assessment ........................................................................................................... 114 5.6.1 Global Warming Potential (GWP) ............................................................................. 114 5.6.2 Water Eutrophication Potential (WEP) ...................................................................... 116 5.7 Interpretation ..................................................................................................................... 118 5.7.1 Sensitivity Analysis ................................................................................................... 118 5.7.2 Consistency and Completeness Check....................................................................... 123 5.8 Technoeconomic Analysis ................................................................................................ 124 5.8.1 Sensitivity Analysis ................................................................................................... 128 CHAPTER 6. OVERALL CONCLUSION AND RECOMMENDATIONS............................. 130 6.1 SCAD operational performance........................................................................................ 130 6.2 Life Cycle Assessment and Technoeconomic Analysis ................................................... 131 6.3 Further recommendations ................................................................................................. 132 APPENDIX ................................................................................................................................. 134 vii REFERENCES ........................................................................................................................... 222 viii LIST OF TABLES Table 2.1 Manure treatments based on solid content ...................................................................... 9 Table 2.2 Food waste classification .............................................................................................. 13 Table 2.3 RIN code definitions (McPhail et al., 2011) ................................................................. 17 Table 2.4 Percentage composition of CH4 from anaerobic digestion of various manures (Anukam et al., 2019). .................................................................................................................................. 22 Table 2.5 Nutrition content of Manures (Chandra, 2005) ............................................................ 23 Table 2.6 Production of volatile fatty acids by microorganisms (Szacherska et al., 2021; Wainaina et al., 2019) ................................................................................................................... 26 Table 2.7 VFA/TA ratio and its correlation to feedstock supply into the digester ....................... 27 Table 2.8 Effect of Ammonia-N concentrations on digester ........................................................ 28 Table 3.1 List of test methods done by A&L Great Lakes Laboratory ........................................ 45 Table 4.1 Feedstocks in SCAD Manure Pit and Food Pit............................................................. 47 Table 4.2 Total Feedstock in Manure Pit Year 2014-2020 ........................................................... 48 Table 4.3 Total Feedstock in Food Pit Year 2014-2020 ............................................................... 51 Table 4.4 Ratio of Manure Pit and Food Pit ................................................................................. 53 Table 4.5 Average Mass Delivered to Digester Each Month (Manure Pit), 2014-2020 .............. 54 Table 4.6 Average Mass Delivered to Digester Each Month (Food Pit), 2014-2020 ................... 55 Table 4.7 ANOVA Tukey Multiple Comparison results for Filtrate manure pit data distribution ....................................................................................................................................................... 60 Table 4.8 ANOVA Tukey Multiple Comparison results for Dairy Gutter data distribution ........ 61 Table 4.9 ANOVA Tukey Multiple Comparison results for Parlor data distribution .................. 62 Table 4.10 ANOVA Tukey Multiple Comparison results for Total Manure Pit data distribution 63 Table 4.11 ANOVA Tukey Multiple Comparison results for Pulp data distribution ................... 64 ix Table 4.12 ANOVA Tukey Multiple Comparison results for FOG data distribution .................. 65 Table 4.13 ANOVA Tukey Multiple Comparison results for Total Food Pit data distribution ... 66 Table 4.14 ANOVA Tukey Multiple Comparison results for Total Feedstock data distribution. 68 Table 4.15 Characterization of SCAD Feedstock a....................................................................... 70 Table 4.16 Biogas production, methane and H2S average year 2014-2020.................................. 71 Table 4.17 ANOVA Tukey Multiple Comparison results for biogas production distribution ..... 73 Table 4.18 ANOVA Tukey Multiple Comparison results for CH4 content distribution .............. 74 Table 4.19 ANOVA Tukey Multiple Comparison results for H2S content distribution ............... 75 Table 4.20 ANOVA Tukey Multiple Comparison results for electricity production distribution 76 Table 4.21 ANOVA Tukey Multiple Comparison results for effluent TS distribution ................ 77 Table 4.22 ANOVA Tukey Multiple Comparison results for effluent TN distribution ............... 78 Table 4.23 ANOVA Tukey Multiple Comparison results for effluent TP distribution ................ 79 Table 4.24 ANOVA Tukey Multiple Comparison results for effluent pH distribution................ 80 Table 4.25 ANOVA Tukey Multiple Comparison results for effluent VFA distribution ............ 81 Table 4.26 SCAD effluent VFA and Alkalinity ........................................................................... 82 Table 4.27 ANOVA Tukey Multiple Comparison results for digester temperature..................... 83 Table 4.28 Comparison of total solids between influent and effluent a ........................................ 85 Table 4.29 Comparison of volatile solids between influent and effluent a ................................... 86 Table 4.30 Comparison of pH between influent and effluent a .................................................... 87 Table 4.31 Comparison of electronic conductivity between influent and effluent a..................... 87 Table 4.32 Comparison of soluble COD between influent and effluent a .................................... 88 Table 4.33 Fiber laboratory results ............................................................................................... 88 Table 4.34 Coefficient of each parameter ..................................................................................... 91 x Table 4.35 MLR results for feedstock impact on the biogas production ...................................... 92 Table 4.36 ANOVA Type II Results for feedstock correlation with the biogas production ........ 92 Table 4.37 MLR results for feedstock impact on the biogas production (after data treatment) ... 93 Table 4.38 Comparison of OLR in 2018-2020: daily, 7 days average, and 30 average ............... 94 Table 4.39 Comparison of HRT in 2018-2020: daily, 7 days average, and 30 average ............... 96 Table 4.40 MLR results for operational parameter impact on the biogas production .................. 98 Table 4.41 MLR results for operational parameter impact on the methane percentage ............... 98 Table 4.42 Summary of SCAD performance .............................................................................. 102 Table 5.1 Data Quality Evaluation Using the Weidema Method (Weidema et al., 2004).......... 106 Table 5.2 Inventory for the life cycle assessment a ..................................................................... 108 Table 5.3 Data Quality Evaluation Summary for LCI ................................................................ 111 Table 5.4 Energy balance of the anaerobic digestion process a .................................................. 114 Table 5.5 Checklist and Inconsistencies based on Data Quality ................................................ 123 Table 5.6 Completeness check for AD system ........................................................................... 124 Table 5.7 Completeness check for landfills system .................................................................... 124 Table 5.8. Economic performance of the digestion process a ..................................................... 126 Table E.1 Summary of TSS results ............................................................................................. 201 Table E.2 Summary of VSS results ............................................................................................ 202 Table E.3 Summary of ammonia results ..................................................................................... 202 Table E.4 Summary of alkalinity results..................................................................................... 203 Table E.5 Summary of TDS results ............................................................................................ 203 Table F.1 Summary of Solids results .......................................................................................... 204 Table F.2 Summary of TKN results ............................................................................................ 204 xi Table F.3 Summary of Phosphorus results ................................................................................. 205 Table F.4 Summary of Potassium results ................................................................................... 205 Table F.5 Summary of Moisture results ..................................................................................... 206 Table F.6 Summary of Sulphur results ....................................................................................... 206 Table F.7 Summary of Magnesium results ................................................................................. 206 Table F.8 Summary of Calcium results ...................................................................................... 207 Table F.9 Summary of Natrium results....................................................................................... 207 Table F.10 Summary of Aluminum results................................................................................. 207 Table F.11 Summary of Copper results ...................................................................................... 208 Table F.12 Summary of Iron results ........................................................................................... 208 Table F.13 Summary of Manganese results ................................................................................ 208 Table F.14 Summary of Zinc results........................................................................................... 209 Table F.15 Summary of Nitrogen results .................................................................................... 209 Table F.16 Summary of Ammonia results .................................................................................. 209 xii LIST OF FIGURES Figure 2.1 RPS Policies Established in the United States (Barbose, 2019) .................................. 15 Figure 2.2 MSU South Campus Anaerobic Digester (personal documentation) .......................... 32 Figure 3.1. Sampling and measurement maps (Source: MSU Anaerobic Research and Education Center)........................................................................................................................................... 39 Figure 4.1 Comparison of feedstock received by SCAD in manure pit ....................................... 52 Figure 4.2 Comparison of feedstock received by SCAD in food pit ............................................ 52 Figure 4.3 Radar chart for feedstock supplies receiveed by SCAD in 2020 ................................ 56 Figure 4.4 Radar chart for feedstock supplies receiveed by SCAD in 2019 ................................ 56 Figure 4.5 Radar chart for feedstock supplies receiveed by SCAD in 2018 ................................ 57 Figure 4.6 Radar chart for feedstock supplies receiveed by SCAD in 2017 ................................ 57 Figure 4.7 Radar chart for feedstock supplies receiveed by SCAD in 2016 ................................ 58 Figure 4.8 Radar chart for feedstock supplies receiveed by SCAD in 2015 ................................ 58 Figure 4.9 Radar chart for feedstock supplies receiveed by SCAD in 2014 ................................ 59 Figure 4.10 Data distribution of Filtrate manure pit ..................................................................... 60 Figure 4.11 Data distribution of Dairy Gutter .............................................................................. 61 Figure 4.12 Data distribution of Parlor ......................................................................................... 62 Figure 4.13 Data distribution for total manure pit ........................................................................ 63 Figure 4.14 Data distribution for Pulp .......................................................................................... 64 Figure 4.15 Data distribution for FOG ......................................................................................... 65 Figure 4.16 Data distribution for Total Food Pit .......................................................................... 66 Figure 4.17 Data distribution for Total Feedstock ........................................................................ 67 Figure 4.18 Correlation of SCAD biogas and electricity over the year ........................................ 72 xiii Figure 4.19 Data distribution for biogas production ..................................................................... 73 Figure 4.20 Data distribution for CH4 content .............................................................................. 74 Figure 4.21 Data distribution for H2S content .............................................................................. 75 Figure 4.22 Data distribution for electricity production ............................................................... 76 Figure 4.23 Data distribution for effluent TS ............................................................................... 77 Figure 4.24 Data distribution for effluent TN ............................................................................... 78 Figure 4.25 Data distribution for effluent TP ............................................................................... 79 Figure 4.26 Data distribution for effluent pH ............................................................................... 80 Figure 4.27 Data distribution for effluent VFA ............................................................................ 81 Figure 4.28 Data distribution for digester temperature ................................................................. 83 Figure 4.29 Digester temperature average in 2014-2020.............................................................. 84 Figure 4.30 Correlation between food waste and biogas production............................................ 90 Figure 4.31 OLR in 30 days average 2018-2020 .......................................................................... 95 Figure 4.32 HRT in 30 days average 2018-2020 .......................................................................... 96 Figure 5.1 System boundaries of Scenario 1 and 0 ..................................................................... 104 Figure 5.2 Mass balance of the anaerobic digestion process ...................................................... 113 Figure 5.3 GWP contribution analysis of the anaerobic digestion process and land applications ..................................................................................................................................................... 116 Figure 5.4 WEP contribution analysis of the anaerobic digestion process and land applications ..................................................................................................................................................... 118 Figure 5.5 Anaerobic digestion sensitivity analysis for GWP .................................................... 120 Figure 5.6 Land applications sensitivity analysis for GWP ........................................................ 120 Figure 5.7 Anaerobic digestion sensitivity analysis for WEP .................................................... 122 Figure 5.8 Land applications sensitivity analysis for WEP ........................................................ 122 Figure 5.9 Total net cash flow and payback period of SCAD .................................................... 127 xiv Figure. 5.10 Sensitivity analysis chart for SCAD payback period ............................................. 129 Figure C.1 Feedstock distribution for Digestate ......................................................................... 161 Figure C.2 Feedstock distribution for SLS Solids ...................................................................... 163 Figure C.3 Feedstock distribution for Beef................................................................................. 167 Figure C.4 Feedstock distribution for Waste Feed Manure ........................................................ 168 Figure C.5 Feedstock distribution for Poultry ............................................................................ 169 Figure C.6 Feedstock distribution for Swine .............................................................................. 170 Figure C.7 Feedstock distribution for ANS Other ...................................................................... 172 Figure C.8 Feedstock distribution for Filtrate Food Pit .............................................................. 174 Figure C.9 Feedstock distribution for SLS Solids Food ............................................................. 176 Figure C.10 Feedstock distribution for Pineapple ...................................................................... 177 Figure C.11 Feedstock distribution for Waste Feed ................................................................... 180 Figure C.12 Feedstock distribution for Other ............................................................................. 181 Figure C.13 Feedstock distribution for Cart Food ...................................................................... 183 Figure H.1 Correlation plot between the parameters part 1 ........................................................ 214 Figure H.2 Correlation plot between the parameters part 2 ........................................................ 215 Figure H.3 Correlation plot between the operational parameters ............................................... 215 Figure H.4 Correlation plot between the total feedstock and the operational parameters .......... 216 Figure H.5 Correlation plot between the total manure pit and the biogas production ................ 216 Figure H.6 Correlation plot between the total food pit and the biogas production .................... 217 Figure H.7 Correlation plot between the total feedstock and the biogas production .................. 217 Figure H.8 The data distribution of the biogas production from 2014 to 2020 .......................... 218 Figure H.9 Linear model plot of biogas production (residuals vs fitted).................................... 218 xv Figure I.1 Correlation plot between the operational parameters and the dependent variables (biogas production and methane concentration) ......................................................................... 220 Figure I.2 Scatter plots of variables ............................................................................................ 221 xvi KEY TO ABBREVIATIONS AD Anaerobic Digester ANS Animal Science ADREC Anaerobic Digestion Research and Education Center CapEX Capital Expenditure CH4 Methane CO2 Carbon Dioxide COD Chemical Oxygen Demand CSTR Continuous Stirred Tank Reactor DQI Data Quality Inventory EC Electrical Conductivity EPA Environmental Protection Agency FOG Fat, Oil, and Grease FU Functional Unit GWP Global Warming Potential H2S Hydrogen Sulfide LCA Life Cycle Analysis LCFS Low Carbon Fuel Standard LCI Life Cycle Inventory LCIA Life Cycle Impact Assessment MLR Multi Linear Regression MSU Michigan State University xvii N2O Nitrous Oxide NH3 Ammonia OLR Organic Loading Rate OpEX Operational Expenditure REC Renewable Energy Credit RIN Renewable Identification Number RFS Renewable Fuel Standard RPS Renewable Portfolio Standards SCAD South Campus Anaerobic Digester TA Total Alkalinity TN Total Nitrogen TP Total Phosphorus TS Total Solids TSS Total Suspended Solids VFA Volatile Fatty Acids VS Volatile Solids VSS Volatile Suspended Solids WEP Water Eutrophication Potential xviii CHAPTER 1. INTRODUCTION 1.1 Problem Statement To address climate change issues and contribute to consuming more renewable energy, Michigan State University launched South Campus Anaerobic Digester (SCAD), a commercial anaerobic digester that produces biogas and renewable electricity from livestock manure and food waste from student’s hall to provide energy to south campus buildings. This facility operated starting in 2013 and is still running today. Throughout the food system, there is a significant loss of food due to spoilage, processing, and damage. The largest source of food waste in American waste occurs at the household level, where approximately 0.5 lb./d/person of food waste is generated (U.S. Environmental Protection Agency, 2018). Manure management is a significant source of methane (CH4) emissions and a contributor to the carbon footprint of food production. For one of gallon milk, manure management accounts for 24% of the carbon dioxide equivalent (CO2-e) emissions, which are largely related to CH4 emissions from long-term manure storage (Thoma et al., 2013). Animal handling sectors such as farms contribute about 18% of GHG emissions when it is not responsibly managed (Esfandiari et al., 2011). Meanwhile in the United States itself, approximately 14% of ammonia emissions come from livestock manure management which is one component of acid rain (Eckert et al., 2018). Environmental impacts associated with food waste and manure management are important contributors to climate change. One of the solutions to address climate change is to produce and utilize more renewable energy sources, especially waste materials. Given the statement, the food system contributes to vast quantities of organic waste materials, such as livestock manure and food waste. These resources can be utilized to produce biogas and further to be renewable electricity (Hosseini & Wahid, 2014). 1 The SCAD system has not had a comprehensive evaluation of the operational performance or the environmental impact of its existence to date. These operational performances include feedstock supplies (livestock manure and food waste), digester performance, biogas quantity and quality, electricity production, and laboratory analysis of influent and effluent. A life cycle assessment (LCA) needs to be conducted to evaluate SCAD in reducing the negative impact on the environment, such as greenhouse gas (GHG) emissions. Evaluating these parameters will provide valuable information on whether SCAD has been operating in its best condition and provide further recommendations on improving SCAD performance to continue serving the MSU community. 1.2 Goal and objectives The goal of this study is to evaluate the operational performance of Michigan State University’s commercial South Campus Anaerobic Digester (SCAD) as well as the environmental impact during its operation from 2014 to 2020. Additionally, this study aims to achieve these objectives: 1. Evaluate the feedstocks supplies quantity that SCAD received during 2014-2020 2. Evaluate output parameters of SCAD during 2014-2020 3. Determine feedstocks that potentially have the most significant impact on biogas production 4. Compare a life-cycle assessment of SCAD to the conventional waste management method 5. Conduct a technoeconomic analysis to know the financial feasibility of SCAD as a commercial digester 6. Summarize lessons learned and operational experiences 2 CHAPTER 2. LITERATURE REVIEW Chapter 2 provides a series of information related to this study. It starts with changes in the global community over several decades, which has eventually led to climate change. Some of these changes have come from the food system, such as food and manure waste from farming. This waste has disrupted the ecosystem balance. To overcome this situation, a series of policies have been implemented to reduce the rate of climate change. This chapter specifically discusses the efforts that the U.S. government has taken to mitigate the impact of climate change, narrowing down to the promotion and utilization of renewable energy sources. There are various types of renewable energy, including biogas that is produced from the anaerobic digestion (AD) process. The next step of this chapter talks about the AD and several factors that affect its performance. The advantage of this process led Michigan State University (MSU) to launch the South Campus Anaerobic Digester (SCAD), a commercial digester that provides electricity to power buildings on the south campus area. Furthermore, a life cycle assessment (LCA) was conducted to evaluate SCAD’s performance in mitigating environmental burdens. The revenue section concludes this chapter to showcase the profits gained by the digester. 2.1 General introduction – global drivers 2.1.1 Societal change Societal change is an alteration of social structure where social relationships become involved in the process. Relationship in this context refers to interactions, patterns, and processes that involve mutual activities of the various parts of society (Greenwood & Guner, 2008; Sharma, 2007). For example, significant societal change has been experienced by many countries in recent decades as the result of economic restructuring, changes in societal value systems, the spread of 3 media technology, and changes in educational systems or population composition (Weichold & Barber, 2009). Societal changes are the result of many factors. One of the most significant factors in the recent era technological advancement. Technology exercises its vast influence by changing the environment that demands society to adapt by modifying social norms (Greenwood & Guner, 2008). Endless new technological discoveries have created more machines and methods of communication which alter social interactions. For example, society is experiencing change due to the development and invention of electric, steam, and petrol driven machines for food production. This change cannot be avoided even for institutions like family and marriage. It affects the lives of children, adolescents, and adults throughout the family dynamic, resulting in a risk of emotional development, or less social control around neighborhoods. The obvious effects of technological advancement are labor organizations, specialization, high speed of life, and increase in production (Sharma, 2007; Weichold & Barber, 2009). Michigan State University Sustainability explains that the advancement of electronic devices such as the latest cell phones and televisions have altered the behavior of students at every corner of campus resulting in the consumption of more energy. During 1965, a student consumed 66 watts of electricity for lighting and playing vinyl records in a student’s dorm room. By 1978, the number rose to 255 watts to turn on small fridge, television, light, and radio. Moving forward to 2013, the number significantly increased to 3,671 watts to power computers, television, audio speakers, microwaves, cell phones, and mini fridges. The increasing appetite for power from student in Residence Halls is causing the University to improve energy systems to support campus growth (Michigan State University Sustainability, 2014). 4 Societal changes in technology have affected change in other related fields, such as agriculture, food, and the environment. Technology advancement has also improved crop production. Moreover, technology helps with food processing, prolonging expiration dates. Nevertheless, agriculture and food processing demand more resources. Agriculture has been correlated with land use changes from forest to farmland, as well as an increase in water demand. The food industry has increased the amount of food waste in landfills, creating new sources of diseases caused by rotten food. Eventually, these issues contribute to climate change which will be discussed in section 2.1.2. Various social movements were conducted to address climate change issues, such as reducing plastic bag use, saving electricity, and providing more mass transportation. Michigan State University enforced several breakthroughs to accomplish this mission, one of them is the establishment of SCAD in 2013 to address the needs of renewable energy sources and food waste management. SCAD will be discussed in section 2.3. 2.1.2 Climate change Traditional energy sources that mostly consist of fossil fuels such as petroleum, coal, and natural gas consequently contribute to environmental pollution (Zhou et al., 2016). A consequence of fossil fuel use is the significant increase in greenhouse gas (GHG) in the atmosphere which has been observed since the late nineteenth century. GHG effect is a natural phenomenon where sun heat radiation on earth surface gets absorbed by gases in the atmosphere and re-emitted to all directions, increasing earth surface temperature as a result (Quinto et al., 2016). GHG, which includes compounds such as carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), and ozone (O3), have contributed to raising global temperature and potentially became the cause of climate change and global warming (Hosseini & Wahid, 2014). 5 The United Nations, through The Sustainable Development Goals Reports in 2019 reported that temperature is increasing globally at an average of 0.85 °C from 1880 to 2012. This increase has affected major crop yield, contributed to the melting of snow and ice which has caused an increase in sea level. If this trend continues, global temperature might exceed 1.5 °C causing a sea level rise of 40-63 cm by 2100. Not just causing environmental pollution, but traditional energy sources are also non-renewable. The increase of human activities, particularly in industry and economic sector, requires a larger energy supply. Increasing energy is demanded in industrialized countries while significant populations in developing regions lack reliable energy. Currently, approximately 3 billion people have limited access to clean-cooking solutions and 840 million people have restricted access to electricity. Therefore, increasing clean and renewable energy use is essential to create more resilience communities to face climate change (United Nations, 2019). 2.1.3 Food waste diversion Food waste has become a global environmental issue over the past decade due to its environmental impact. Additionally, there are approximately 800 million people who suffer from hunger. A third of the food produced for global human consumption is lost or wasted every year, approximately 1.6 billion tons. This loss costs $2.6 trillion, of which $1 trillion is incurred from GHG emissions, water scarcity, biodiversity loss, increased conflicts, and loss of livelihood. Contributing factors are soil erosion, nutrient loss, reduced yields, wind erosion, and pesticides exposures. Moreover, food waste is responsible for 4.4 gigatons of CO2-e per year. That number represents 8% of the global anthropogenic GHG emissions (World Biogas Association, 2018). The United States Environmental Protection Agency (U.S. EPA) reported in 2018 that over 35 million tons of food were sent to landfills, which is the equivalent to half a pound per person per day. This waste costs the commercial food service industry about $100 billion annually. The 6 number significantly increased almost a decade later. The country spent $218 billion (about $670 per person in the U.S.) to grow, process, transport, and dispose uneaten food. Each year, the number of foods dumped to the landfill is approximately 52.4 million tons. Meanwhile, there is an additional 10.1 million tons of unharvested food remaining on farms, thus the total of wasted food is roughly 63 million tons annually. From this number, the U.S. restaurant sector itself contributes 11.4 million tons of food waste annually. This number consists of 7.3 million tons from full-service restaurants and 4.1 million tons from limited-services restaurants (ReFED, 2018; U.S. Environmental Protection Agency, 2018). According to the Michigan Recycling Coalition, of all municipal solid waste disposed in Michigan landfills, 13.6% is food waste. Food waste prevention is the most essential process to reduce the number of organic wastes going through the landfills, which is the top priority from environmental, social, and economic perspectives. The challenge is that the prevention efforts are difficult to quantify considering the diverse waste composition from each household. Therefore, the State of Michigan puts their effort into food surplus diversion by feeding hungry people and then animals. The Food Bank Council of Michigan cooperates with multiple partners such as farmers, individuals, non-profit organizations, corporations, and government to support 7 regional food banks and advance the access to healthy food for people in need in all 83 Michigan counties (O’Brien, 2017). The U.S EPA explained that food waste diversion should be considered seriously because of the environmental impact of food waste can be damaging. It starts with food waste rooting in landfills that create CH4, a greenhouse gas (GHG) 25 times stronger than carbon dioxide. Among GHG emissions in the United States, 13% results from the growth, manufacture, sale, transport, 7 and disposal of food. Not only that, wasted foods consumes more than a quarter of the total freshwater in the country (United States Environmental Protection Agency, 2014). Shifting the management system of food waste from linear to circular can be an option to optimize food waste as a valuable resource. For example, utilizing food waste for biogas production through anaerobic digestion can provide many benefits, such as generating renewable energy, reducing GHG emissions, protection of water bodies, additional revenue streams from sale of electricity and compost, creating more self-sufficient and resilient communities and sustainable industrialization (World Biogas Association, 2018). 2.1.4 Manure management The existing waste treatment solutions for manure and slurries systems still have several drawbacks regarding soil contamination and negative environmental impacts, such the contamination of water streams and food crops with pathological entities (Goldstein, 2018). Animal handling sectors such as farms contribute about 18% of GHG emissions when not responsibly managed (Esfandiari et al., 2011). Meanwhile in the United States, approximately 14% of the ammonia emissions come from livestock manure management which is one component of acid rain (Eckert et al., 2018). The perspective of manure as a waste product should be changed, as manure has valuable benefits to save fertilizer cost with proper storage, handling, and application (Courneya, 2010). A critical aspect of sustainable manure management is by constructing housing and manure storage systems thus helping to conserve and maintain high concentrations of nutrients (International Atomic Energy Agency, 2008). To do so, animal manure should not be mixed with human waste to control disease and parasites (Lorimor & Powers, 2018). Moreover, improper manure 8 management can lead to environmental issues such as water pollution and source of odor, flies, and parasites (Bradley, 2019). Lorimor and Powers (2018) classified manure into four different solid contents, as explained in Table 2.1 below. Table 2.1 Manure treatments based on solid content Solid content Treatment 4% or below Treated as a liquid with irrigation equipment, such as liquids which majority of solids have been removed or diluted manure 4 to 10% • Handled as a slurry • It may require special pumps, for instance swine pit manure and dairy manure with milking parlor wash water added 10 to 20% • This is typical of many dairy operations which are too thick to pump but too thin to scoop • Adding water to handle it as liquid and special pumps will be used to agitate and move the manure 20% or more Treated as a solid which can be stacked and picked up with a fork or bucket loader Bradley (2019) emphasizes that manure management and utilization plan must include these components: 1) Quantity of manure and bedding generated annually from all livestock on the farm; 2) Manure handling, collection methods, and equipment used. This includes the handling from barns, stalls, paddocks, and pastures; 3) Size and location of storage and/or composting facilities; 4) Methods used to prevent drainage into storage areas, paddocks, and pastures; 5) Nutrient analysis of manure prior to application (if applying to land); 6) Soil analysis for lands in which raw or composted manure will be applied; and 7) Utilization records: land application, compost monitoring, or off-sites uses. 9 2.1.5 Renewable energy Renewable energy is energy generated partially or entirely from natural resources which are available on a renewable basis and inexhaustible (naturally replenished), for instance hydropower, wind, solar, geothermal, and bioenergy (plant-based). Renewable energy can be utilized without transforming it into other forms, such as cooking gas. Otherwise, renewable energy would be processed for electricity generation (Gorjian, 2017; The U.S. Environmental Protection Agency (EPA) State and Local Energy and Environment Program, 2018). Renewable energy has received noticeable interest from many countries as a solution to limited energy resources and environmental issues related to fossil fuels consumption (Mohammadrezaei et al., 2018). According to Abolhosseini et al. (2014), the development of renewable energy technologies was driven by three factors: energy security, economic impacts, and carbon dioxide emissions reduction. The Arab oil embargo in 1973 played a vital role in raising the awareness of energy supply security. This was supported by high oil prices, increasing dependency on oil imports, depletion of fossil fuels, increasing competition from emerging economies, political instability in major oil producers, a high impact due to any disruption in energy supply on developed, and rapidly developing countries. Moreover, energy security issues must address climate change concerns, which demands the diversification of energy sources. That means if energy security is fulfilled only by fossil fuels, it will significantly increase GHG emissions which will worsen climate change. Therefore, energy demand must be fulfilled by renewable energy to embody energy security while lowering GHG emissions. Shifting to economic impacts, renewable energy opens job creation opportunities, industrial innovation, and a balance of payment. 10 Additionally, renewable energy technologies should reduce CO2 emissions by substituting fossil fuels in the power generation industry and transportation sector. Biogas is one of renewable energy sources that has become an alternative fuel for transportation, industrial engines, and residential electricity and heat. Biogas as an alternative fuel has the capacity to reduce the impact of pollution. Additionally, biogas can be utilized further to produce electricity and/or heat through combustion by combining heat and power (CHP) generation systems. The efficiency could reach 34-40% by using large turbines and 25% using small generators (Hosseini and Wahid, 2014). In Chile, electricity generated from biogas could fulfill 50 megawatts (MW) of demand. Moreover, the slurry and residues could be used as fertilizer (Passos et al., 2017; Zhou et al., 2016). Biogas composition primarily consists of CH4 and CO2 with a small amount of hydrogen sulfide (H2S) and water vapor (H2O). Biogas typically contains 55-70% of CH4 that can be used as fuel in a variety of purity levels and efficiencies (Chynoweth et al., 2001; Somers et al., 2018). Biogas which contains high concentration of CO2 will lower its quality as cooking gas since CO2 contributes to carbon monoxide (CO) formation which could weaken biogas combustion. Additionally, H2S reduction level (i.e., via desulphurization processes) must happen before biogas can be used as H2S is a toxic gas which poses safety concerns for people and may result in additional equipment maintenance (Hosseini & Wahid, 2014). Biogas is produced through anaerobic digestion (AD) which will be discussed further in section 2.2. 11 2.1.6 Policy drivers – focus in the U.S. 2.1.6.1 Food waste diversion According to O’Brien (2017), an amendment of the Solid Waste Management Act in 1990 attempted to minimize the hazardous landfill gases yield by banning the disposal of yard clippings. At the same time, this effort promoted the conversion of valuable organics into beneficial resources for municipalities, agriculture, and industry. The Michigan Legislature enacted Public Act 264 in 1990 which define “yard clippings” as “leaves, grass clippings, vegetable or other garden debris, shrubbery, or bush or tree trimmings less than four feet in length and two inches in diameter that possibly to convert into compost humus.” Due to this amendment, the Natural Resources and Environmental Protection Act (NREPA) prohibited owners or operators of landfills and municipal solid waste incinerators from accepting solid waste if it was known that the solid waste included yard clippings generated or collected on land owned by a county, municipality, or a state facility. Granholm and Chester (2007) use the three principles of sustainability – economic vitality, ecological integrity, and improved quality of life – to guide solid waste management decision- making. According to the Policy, Michigan’s preference is to first avoid waste generation, then to utilize generated waste for beneficial purposes, and properly dispose of what remains. According to EPA (2014), much of food waste that reaches landfills is still edible. Wasted food can be divided into three categories as explained in Table 2.2. 12 Table 2.2 Food waste classification Category Description Example Avoidable Food that can be easily prevented from an entire tray of lasagna is going to waste. Reasons for waste include left over every day at a overpreparation, improper storage, or buffet spoilage. Understanding the cause of this waste is key to preventing it. Possibly avoidable Food that may seem inedible but can be beet tops can be cooked used or repurposed similarly to collard greens or spinach instead of discarded. Also, slightly stale bread can be used for croutons or breadcrumbs Unavoidable Food that cannot be consumed by people banana peels or peach pits and should be used for animal feed, compost, or anaerobic digestion Moreover, EPA also created the Food Recovery Hierarchy to reduce wasted food. Based on this hierarchy, there are five stages to be considered when utilizing excess food: a. Source reduction: reduce the volume of food waste generated b. Feed people: donate extra food to food banks, soup kitchens and shelters c. Feed animals: provide food to farmers for animal feed d. Industrial uses: provide fat for rendering, biofuel, and food discards for animal feed production e. Composting/digesting: convert food scraps into a nutrient rich soil amendment 13 2.1.6.2 Renewable Portfolio Standard (RPS) A renewable portfolio standard (RPS) is a policy that demands that electricity providers include a minimum percentage of their electricity supplies from renewable energy sources, such as wind, solar, geothermal, and various forms of biomass and ocean energy. Electricity suppliers have two options to comply with this requirement. First, suppliers can own a renewable energy facility and produce their own electricity. Second, they can purchase renewable electricity from a renewable facility. Furthermore, RPS enables the market to choose any renewable energy resources to fulfill the mandate since renewable energy availability varies depending on the regional climate and geographies (Cory & Swezey, 2007; Rader & Hempling, 2001). From an environmental perspective, RPS promotes climate change mitigation by reducing air and carbon pollution, waste reduction, and conserving water and other valuable natural resources. Moreover, it can boost local economic development by creating more jobs, taxes, and revenues associated with renewable energy (Environmental Protection Agency, 2015). According to Lawrence Berkeley National Laboratory and U.S. Department of Energy, 29 States have implemented RPS, including Michigan, California, Nevada, and Minnesota (Figure 2.1). Additionally, Washington DC and 3 territories (Northern Mariana Island, Puerto Rico, and US Virgin Islands) have also had RPS. Meanwhile, 8 states (including Kansas, Oklahoma, and Indiana) and 1 territory (Guam) have established their renewable portfolio goals (Barbose, 2019; U.S. Department of Energy, 2016). 14 Figure 2.1 RPS Policies Established in the United States (Barbose, 2019) RPS implementation’s most prominent mechanism is renewable energy certificates (RECs). RECs are payments that electricity suppliers offer to renewable energy companies for the renewable electricity they provide to the grid. One REC typically represents 1 megawatt-hour (MWh) of renewable electricity. The implementation of REC provides an accurate, durable record of electricity produced, also a fungible commodity that can be traded among providers to fulfill RPS (Cory & Swezey, 2007). 2.1.6.3 Renewable fuel standard The United States is the biggest consumer of crude oil in the world. Consequently, the country is facing two major concerns which are low energy security and high greenhouse-gas emissions. Furthermore, the United States imported between 52 to 60% of oil consumed from 2005 to 2009 with roughly 30% of CO2 emissions resulting from transportation fuels (The National Academy of Sciences, 2011). Biofuels emerge as an alternative to petroleum-based fuels due to its production from renewable domestic sources which has the potential to improve U.S. energy security. It also 15 provides life-cycle greenhouse-gas benefits when compared to fossil fuel. As a result, the U.S. Congress enacted the Energy Independence and Security Act (EISA) in 2007 to lead the country toward energy independence and security while improving clean renewable fuels production. EISA then resulted in the Renewable Fuel Standard (RFS) program which was created by Congress to address greenhouse gas emissions, broaden renewable fuel sector in the country, and dwindle the demand for imported oil (America’s Oil and Natural Gas Industry, 2017). RFS demands that U.S. transportation fuel contains a minimum volume of renewable fuel. It sets the target for consuming 35 billion gallons of ethanol-equivalent biofuel and 1 billion gallons of biomass-based diesel by 2022. The Environmental Protection Agency (EPA) holds statutory authority on volume amounts after 2022. There are four categories of renewable fuels according to RFS: total renewable, advanced, biomass-based diesel (BBD), and cellulosic. These categories are determined based on the reductions in life-cycle emissions of GHG, relative to petroleum, feedstock, and fuel characteristics (Congressional Research Service, 2020; Stock, 2015). Nevertheless, many reports and studies are doubtful that the implementation of RFS will be successful and meet the target. First, there is a limit placed on car use of gasoline at 10% ethanol (E10) which is the maximum acceptance under the manufacturer’s warranty. This caused an E10 plateau in the US fuel supply in 2013. Second, cellulosic biofuels experience policy uncertainty and high production costs that might hinder investors from supporting this initiative. Eventually, it depends on how biofuels are produced. There are many factors that can affect it that might be unpredictable. These factors include technical expertise, weather conditions, market stability, tax incentives, and trade disputes. These factors could impact the entire industry (Congressional Research Service, 2020; Stock, 2015). 16 To facilitate RFS compliance, Renewable Identification Number (RIN) systems were created by the U.S. EPA. A RIN is a 38-character numeric code that represents a volume of renewable fuel produced in or imported into the United States. One RIN is equivalent to one gallon of biofuel produced and reported to EPA. A RIN contains information about biofuel sources like, production year, biofuel producer, and the type of fuel (Table 2.3). At the end of the calendar year, fuel suppliers must fulfil minimum requirements of RINs to be compliant with RFS. Similar to REC, RIN also can be traded like other commodities. RINs are valid for use during its production year and the following year (Christensen et al., 2014; Cooper, 2018; McPhail et al., 2011; Yacobucci, 2013). Table 2.3 RIN code definitions (McPhail et al., 2011) 38-character code: KYYYYCCCCFFFFFBBBBBRRDSSSSSSSSEEEEEEEE K RIN assignment code YYYY Year batch is produced/imported CCCC Company registration ID FFFFF Facility registration ID BBBBB Producer-assigned bath number RR Equivalence value for renewable fuel D Renewable type code1 SSSSSSSS RIN block starting number EEEEEEEE RIN block ending number 1 Five separate RIN categories: D=3 for cellulosic biofuel; D=4 for biomass-based diesel; D=5 for advanced biofuel; D=6 for other renewable fuel; D=7 for cellulosic diesel 2.1.6.4 California low carbon fuel standard The California Low Carbon Fuel Standard (LCFS) is a program created to reduce greenhouse gas emissions (GHG) and other air pollutants from the state’s transportation sector, diversify the State’s fuel mix, and decrease the dependence on petroleum. It is one of the policies in California that comes from the implementation of the Global Warming Act of 2006. LCFS evaluates the full life cycle emissions of transportation fuels and includes all GHG emissions that 17 are the result of production, distribution, and consumption expressed as grams of CO2e per megajoule (Renewable Fuels Association, 2021; Townsend & Havercroft, 2019). Recently, the transportation sector has contributed 50% of GHG emissions, 80% of nitrogen oxide emissions, and 95% of particulate matter emissions in California (Center for Law, 2019). LCFS reduces GHG emissions by establishing annual standards to be followed by fuel producers and distributors. It emphasizes the reduction of the average life-cycle carbon intensity (CI) of the fuels supplied to the market. Fuels that gain lower CI values regulated by California Air Resources Board (CARB) will get compliance credits. On the other hand, fuels with CI values higher than the standard will get compliance deficits (Renewable Fuels Association, 2021). Additionally, certain requirements might be needed to fulfill LCFS such as life cycle assessment (LCA) which accounts for the environmental impact of fuel production from feedstock production to end use stage (Congressional Research Service, 2021). The implementation of LCFS began in 2011. It was amended in 2015 to address fuel CI reduction standard adjustment and extension. Currently, LCFS requires a 10% reduction by 2022 and a 20% reduction by 2030. Since its first implementation, LCFS has achieved more than 77 million credits with each credit referencing a metric ton of GHG emissions reduction contra the annual standard (Renewable Fuels Association, 2021). It has helped to incentivize production of low-carbon fuels and generate additional revenue to encourage investment in statewide low-carbon transportation fuel infrastructure (Center for Law, 2019). Despite its merits LCFS has some implementation challenges. These challenges include determining the appropriate energy related to GHG target, developing a robust LCA, as well as constructing a transparent compliance system. Moreover, part of Congress opposed LCFS due to concerns about economic effects such as job loss, limited affordable lower carbon fuel options, 18 and increasing fuel price as an effect of fulfilling the certain standards required. Therefore, further improvement of the LCFS is essential (Center for Law, 2019; Congressional Research Service, 2021). 2.1.6.5 Carbon Intensity Carbon intensity (CI) is the number of GHG emission generated throughout production and use of life cycle energy sources, such as transportation fuel. CI is represented in units of grams of CO2 per megajoule of energy (gCO2e/MJ). CI calculation considers extraction, refinement, distribution, storage, and combustion of energy. Therefore, the calculation can be included in life cycle assessments (Ingram, 2015; United Nations Environment Programme, 2019). CI is calculated using the Greenhouse Gases, Regulated Emissions and Energy Use in Transportation (GREET) model which is developed by U.S. Department of Energy and Argonne National Laboratory. The current model is the CA-GREET 3.0 Model and Tier 1 Simplified Carbon Intensity calculators. The GREET model aims to evaluate and compare energy, environmental impacts of transportation fuels, and vehicle technologies on a life-cycle basis. Over 100 alternative fuel pathways and over 80 vehicle technologies have been evaluated using this model: including aviation fuel, aircraft operation, marine fuels, and vessel operation. Furthermore, GREET evaluates total energy consumption, GHG emissions (CO2, N2O, CH4), air pollutants (SOx, NOx, VOC, CO, PM10, PM2.5), and water consumption (Argonne National Laboratory, 2014; M. Wang, 2007). 19 2.2 Anaerobic digestion 2.2.1 Introduction The anaerobic digestion (AD) process is defined as a biological process in which a lack of O2 causes a degrade in carbon source materials through the help of various microorganism consortiums (Chynoweth et al., 2001). It transforms organic matters into CH4 gas, thus reducing odor and pathogen risks (R. Chen et al., 2016). AD does not only provide a renewable alternative energy source, but also an alternate pathway to process organic waste, reduce GHG emissions from landfills, and mitigate the demand of fossil fuel and chemical fertilizers (Li et al., 2018; Zou et al., 2018). In terms of efficiency and costs, AD energy could compete with other biomass energy sources, such as heat, synthesis gases, and ethanol (Chynoweth et al., 2001). Moreover, AD could provide electricity storage via upgrading biogas to high purity of methane which might be stored and used for other purposes (Jürgensen et al., 2018). One factor that makes AD attractive is the technology used in AD is scalable (Hosseini & Wahid, 2014). This means AD technology can be applied in various digester capacities; thus, the digesters can be less costly to build, operate, and maintain. This scalability also opens the chance to implement AD technology for small scale farms or communities. For example, the small-scale AD project might be carried out starting from 100 dairy cows or 200 cows or between 200 to 5,000 tons of organic waste per year which is expected to produce 80 kW electricity (Marjolaine, 2019). AD consists of four main stages: hydrolysis, acidogenesis, acetogenesis, and methanogenesis. The first three stages are maintained by bacterial communities to yield acetate, hydrogen, and carbon dioxide, while the last stage is maintained by methanogenic archaea communities to produce methane from acetate, or from hydrogen and carbon dioxide as alternative sources (Z. Yu et al., 2018; Zhou et al., 2016). 20 In the hydrolysis process, organic materials such as cellulose and hemicellulose are hydrolyzed, becoming soluble sugars, alcohol, and other organic substances. The hydrolysis products process will be transformed by acid producing bacteria to yield volatile fatty acids that consist of formic, acetic, propionic, and butyric acid. Hydrolysis and acidification are continuous biochemical reactions. When these processes are performed ideally, high yields of organic acids are achieved, degradation or loss of the organic acids is avoided, fermentation inhibitors are minimized, and cellulose and hemicellulose degradation are improved (J. Yu et al., 2017). Methanogens utilize formic and acetic acid directly, whereas propionic and butyric acid need to be converted into acetic acid by acetic acid producing bacteria (Shen et al., 2018). Temperature and pH play key roles in developing appropriate AD systems. In most studies, AD can be developed in mesophilic and thermophilic temperatures. Mesophilic temperature is preferable if hydrolysis and acidification are done within high solid content condition, while thermophilic temperature is recommended for improving AD efficiency under low solid content in 3 days (J. Yu et al., 2017). An ideal digestion should have a pH range of 6.5-8.0 to provide a convenient environment for anaerobic microbes, especially for archaea in degrading organic compounds and producing methane (Zhong et al., 2015). Appropriate pH for hydrolysis and acidogenesis is 5.5 and 6.5, respectively (Zhou et al., 2016). 2.2.2 Feedstock 2.2.2.1 Manures Historically, manure has been utilized for many purposes, such as fertilizer, soil amendment, energy source, and even construction material. There are many recyclable components in manure, including solids, organic matter, nutrients, and fiber. The contents in 21 manure can be affected by several key factors such as species, digestibility, protein and fiber content, diet, age, housing, environment, and production stage. Moreover, manure contains nitrogen, phosphorus, and potassium which are critical nutrient sources for crops. In general, manure with higher solids concentrations have higher nutrition content (Kostic et al., 2020; Lorimor & Powers, 2018; U.S. Environmental Protection Agency, 2015). Nutrients in manure are available in soluble and insoluble forms. Soluble nutrients can be consumed by the crops right away, while insoluble nutrients take up to a year or more to be available. Each nutrient has a different characteristic. For example, phosphorus is typically 80% available in the settled solids of manure storage and insoluble. Meanwhile, potassium is typically 80% found in the liquid and is highly soluble. Nitrogen is split almost evenly between liquid and solids (Lorimor & Powers, 2018). Among the various methods to utilize manure, anaerobic digestion is a technology that processes manure into biogas as source of heat or electricity. The energy produced can be used on the farm or sold to the local power grid (U.S. Environmental Protection Agency, 2015). Diverse types of manure contribute slightly different percentages of CH4 in biogas, as shown in Table 2.4. Table 2.4 Percentage composition of CH4 from anaerobic digestion of various manures (Anukam et al., 2019). Manure CH4 Composition (%) Cattle 50-60 Pig 60 Poultry 68 Sheep 65 Horse 66 Manure’s characteristics are critical factors for biogas production and process stability during anaerobic digestion. These characteristics are moisture content, total solids (TS), volatile 22 solids (VS; organic compounds from plant or animal that lost when the dry solids are burnt at 550 °C), biodegradability, particle size, pH, biological oxygen demand (BOD), chemical oxygen demand (COD), also carbon and nitrogen contents (L. Chen & Neibling, 2014). Additionally, manure contains variative nutrition contents as shown in Table 2.5. These contents are important for the digestate once it is land-applied. The anaerobic digestion process slightly changes the nutritional contents of manure. Most nitrogen is found in the form of organic N and NH4 in liquid phase of digester sludge. A negligible amount of nitrogen might be emitted as NH3. Therefore, digester sludge will have higher ammonium content than raw manure. Meanwhile, phosphorus content does not change significantly due to anaerobic digestion process. All the P present in the manure will still be present in the digester sludge. The difference is that the dissolved portion of P will be moved into bacteria bodies that perform anaerobic digestion. Additionally, C in the form of simple sugars, volatile fatty acids, and alcohol is converted to CO2 and CH4. As a result, effluent has less C or organic matter compared to raw manure (Manitoba, 2015; Natural Resource Conservation Services, 2007) . Table 2.5 Nutrition content of Manures (Chandra, 2005) Percentage content Manure Nitrogen (N) Phosphoric acid (P2O5) Potash (K2O) Cattle, fresh 0.4-0.5 0.3-0.4 0.3-0.4 Horse, fresh 0.5 0.4-0.6 0.3-1.0 Poultry, fresh 1.0-1.8 1.4-1.8 0.8-0.9 2.2.2.2 Food waste In this case, material contents are essential as they provide nutrition for the microbial community in the digester. The carbon/nitrogen (C/N) ratio in slurry has a critical contribution in the AD process. If slurry contains high C/N ratio, microbe will consume nitrogen sources rapidly 23 thus decrease biogas productivity. On the other hand, methanogen bacteria reproduction and metabolism will be inhibited as carbon deficit, ammonia accumulation, and pH increase occur in low C/N ratio slurry. Typically, C/N ratio range for AD is 10-90, but 30 is the most common (H. Chen et al., 2016; X. Wang et al., 2012). To achieve an ideal C/N ratio, previous research combined several resources as AD materials what is known as co-digestion. The co-digestion of various substrates enhances methane yield, improves buffering capacity, and prevents acidification (Li et al., 2018; X. Wang et al., 2018). Compared to animal manure only, applying the AD process to the co-digestion between animal manure and feedstock provides more stable operational performance and produces more methane. This happens because feedstock with a higher C/N ratio balance out with manure which has low C/N ratio achieving an ideal C/N ratio range (X. Wang et al., 2012). The mixture of food waste and dairy manure also attracts attention as this combination potentially has an ideal C/N achieving productive and efficient AD processes (H. Chen et al., 2016; X. Wang et al., 2012). Li et al. (2018) revealed that methane production could be improved by co- digesting dairy manure and corn stover with tomato residues at 20-40% volatile solid based. However, this process will happen once tomato residue is added beyond 60% since it may trigger pH drop and over produce VFA (Volatile Fatty Acids). 2.2.3 Testing and analysis – Operation data 2.2.3.1 Operational parameters 2.2.3.1.1 pH pH is a measure of acidity or basicity of a solution. Specifically, pH measures the hydrogen concentration, [H+] which the value ranges from 0 to 14. Additionally, pH 7 is called “neutral” 24 because it is the center of the measurement scale, where the ratio of [H+] and [OH-] (hydroxide ion concentration) is equal. A solution is determined to be acidic if the pH is below 7 because [H+] is greater than [OH-], while a solution with pH above 7 is grouped as basic or alkaline which means [OH-] is greater than [H+] (Hach, 2018). An anaerobic digester typically has a pH range between 6.4 and 8.2 to maintain methanogens population. This can be achieved by maintaining the balance of acetogens and methanogens. Acetogens are needed to produce acids, while methanogens consume acids to yield methane gas and increase alkalinity. If acetogens surpass methanogens population, it will drop the pH which then inhibit methanogens performance, resulting in a “sour” digester (L. Chen & Neibling, 2014; MSU Anaerobic Digestion and Research Center, 2019). 2.2.3.1.2 VFA Volatile Fatty Acids (VFAs) are intermediate products of anaerobic digestion process which emerge after polymer hydrolysis and acidogenesis prior to being degraded into acetic acid for methanogenesis stage. VFA belongs to carboxylate which has a low molecular weight consisting of 2 to 6 carbon atoms. VFAs are produced from various feedstocks, for example, agricultural waste, food waste, milk sewage, dairy whey effluent, municipal waste, and cellulose sewage (Mayer et al., 2010; Szacherska et al., 2021; Wainaina et al., 2019). VFA acts in the biopolymers of biofuels production such as methane and hydrogen. Other applications of VFA include carbon sources in biological denitrification, production of biodiesel, and electricity production through microbial fuel cells. Moreover, there are diverse types of microorganisms that contribute to VFA production as shown in Table 2.6 (Lukitawesa et al., 2020; Magdalena et al., 2019). Several factors are crucial for VFA productivity, which are hydraulic retention time (HRT), organic loading rate (OLR), temperature, pH, and pretreatment (Wainaina et al., 2019). 25 Table 2.6 Production of volatile fatty acids by microorganisms (Szacherska et al., 2021; Wainaina et al., 2019) Volatile Fatty Acid Bacteria Substrate Acetobacter aceti Cheese whey Mixed gas (4% H2:18% Clostridium acetium Argon:78% CO) Acetic Acid Clostridium lentocellum SG6 Paddy straw Moorela thermoaceatica Sugarcane straw hydrolysate Saccharomyces cerevisiae + Glucose Acetobacter pasteurianus Lactate Propionibacterium acidipropionici Glycerol (ATCC 4965) Sugarcane molasses Propionibacterium acidipropionici Glycerol (CGMCC 1.223) Propionic Acid Propionibacterium acidipropionici Hemicellulose hydrolysate (ATCC 4875) Cheese whey Propionibacterium freudenreichii Glucose CCTCC M207015 Pripionibacterium freudenreichii spp. Glycerol shermanii Clostridium butyricum S21 Sucrose Clostridium butyricum ZJUCB Glucose Clostridium thermobutyricum Glucose Butyric Acid JW171K Corn husk hydrolysate Clostridium tyrobutyricum Sugarcane bagasse hydrolysate VFA is expressed in equivalent milligrams of acetic acid per liter. VFA is accounted together with total alkalinity (TA), which is expressed in milligrams equivalent of calcium 26 carbonate per liter. VFA/TA ratio will determine the follow up action regarding feedstock supply into the digester, as explained in Table 2.7 (Lossie & Pütz, 2011; MSU Anaerobic Digestion and Research Center, 2019). Table 2.7 VFA/TA ratio and its correlation to feedstock supply into the digester VFA/TA Ratio Background Corrective Action >0.6 Highly excessive biomass input Stop adding biomass 0.5-0.6 Excessive biomass input Add less biomass 0.4-0.5 Plant is heavily loaded Monitor plant more closely 0.3-0.4 Biogas production at a maximum Keep biomass input constant 0.2-0.3 Biomass input is too low Slowly increase biomass input <0.2 Biomass input is far too low Rapidly increase biomass input There is a correlation between VFA and pH. If VFA level increases, that will decline pH, alkalinity, and biogas production. In a normal operating system, hydrogen and acetic acid formed by acidogenic and acetogenic bacteria is immediately converted into methane by methanogens. However, if VFA is increased, an unbalanced condition between acidogenic and methanogenic activities will reduce methanogens performance. The acceptable range for VFA is between 50 to 300 mg/L as acetic, meanwhile the acceptable range for alkalinity is between 1,500 to 5,000 mg/L as CaCO3 (Krakat et al., 2017; MSU Anaerobic Digestion and Research Center, 2019; Schnaars, 2012). 2.2.3.1.3 Ammonia Anaerobic degradation of proteins or amino acids produces ammonia through the degradation of nitrogenous matter. Ammonia is present at elevated levels in certain feedstock, such as meat processing by-products, food waste, also swine and poultry manure. Free ammonia (NH3) form is more toxic to methanogens than the ionized form (NH4+). Moreover, anaerobes are more 27 sensitive to ammonia toxicity at higher temperatures. Lowering digester pH will relieve NH3 toxicity. Additionally, dilution may be needed to measure samples with extreme NH3 levels. The acceptable range for ammonia content is between 1,500 to 3,000 mg/L. Table 2.8 explains the effect of ammonia concentration on digester. In low concentration (between 50-200 mg), ammonia is beneficial for amino acids, proteins, and nucleic acids synthesis. Furthermore, ammonia maintains neutral pH conditions by neutralizing the organic acids yielded by fermentative bacteria. These conditions are essential for bacterial growth (Y. Jiang et al., 2019; MSU Anaerobic Digestion and Research Center, 2019; Walker et al., 2011). Table 2.8 Effect of Ammonia-N concentrations on digester Effects Ammonia-N (mg/L) Beneficial 50-100 No adverse effect 200-1,000 Inhibitory effect at higher pH values 1,500-3,000 Toxic > 3,000 Anaerobes, especially methanogens, are sensitive to ammonia toxicity due to ammonia can freely pass-through methanogens’ cell membranes hence cause a proton imbalance. Consequently, the intracellular pH of methanogenic bacteria changes then inhibits specific enzymatic reactions. Moreover, acetate degradation is hindered by high ammonia concentration, leading to acetate accumulation, buffer capacity depletion, methane yield decrease, VFA concentration increase, and pH drop. Digesters with high concentrations of ammonia will experience methanogenesis inhibition and lead to complete failure (H. Chen et al., 2016; Morozova et al., 2020). 28 2.2.3.1.4 Organic Loading Rate Organic loading rate (OLR) is defined as the quantity of organics fed to a continuous digester per day (Meegoda et al., 2018). OLR data talks about digester health and space. It is important because OLR indicates the quantity of volatile solids to be fed into the digester each day (Babayee and Shayegan, 2011). OLR can be counted with formula (1) as follows (MSU Anaerobic Digestion and Research Center, 2019): OLR = mass of Volatile Solids (VS) / volume of reactor (1) Material characteristics and operation conditions are key points to in determining OLR. In general, the biogas yield increase is in line with the OLR increase. If OLR is optimized, the processing efficiency of anaerobic digestion can be improved. Moreover, optimum OLR can reduce the plant capital cost, improve biogas yield, and ensure operation stability (J. Jiang et al., 2020). However, excess OLR can contribute to biogas production inhibition (Musa et al., 2018). OLR is essential for AD operation since it is related to system stability, waste treatment ability, and biogas production. Ideal OLR is favorable for cell activity, thus contributing to increasing methane production and improving substrate degradation. Nevertheless, increasing OLR has a risk of creating excessive VFA production and lowering the pH that inhibits the entire process (Moguel-Castañeda et al., 2020). 2.2.3.2 Performance indicators 2.2.3.2.1 Mixing There are many aspects to consider in choosing a reactor type and mixing condition. For example, a conventional AD reactor is working for 20-30 days (about four and a half weeks) of hydraulic retention time, getting minimum once feeding per day with proper mixing at 35 °C (Chynoweth et al., 2001). Mixing in biogas digester aims to achieve balanced nutrient and heat 29 circulation, to minimize precipitation of materials by shaping an adequate mixture of solid particle and liquid suspension and facilitate gas lifting from fermentation substrates. The quality of mixing in AD plant will affect in choosing the mixer type (Eshkaftaki & Ebrahemi, 2019; Kress et al., 2018; Naegele et al., 2014). Mixing also helps create homogenous temperature and bacteria- nutrient composition in the digester. Currently, this process is considered to become an important stage in biomass-methane conversion (Mohammadrezaei et al., 2018). Nevertheless, mixing has also become a major challenge in AD field application since it consumes approximately 51% of total electricity needs in the complete process. It makes several research focused on analyzing whether mixing time and rate have a significant impact on methane production. A previous study demonstrated that reducing mixing time in a full-scale AD reactor fed with crops did not strongly affect the nutrient distribution that endangers biogas production (Kress et al., 2018). A moderate stirrer rate at 80 rpm is considered as the best condition of stirring process in AD since it provides appropriate mixing pattern. The low-level mixing rate causes high death species near reactor wall. On the other hand, a high mixing rate would affect microorganism structures in the reactor, thus declining biogas production (Mohammadrezaei et al., 2018). Moreover, appropriate mixing for digester with high solid content is beneficial but not to low solid content condition (J. Yu et al., 2017). Therefore, it is essential to determine the type of material that would be digested to choose an ideal reactor and mixing type. 2.2.3.2.2 Economic Analysis Renewable energy production often faces challenges regarding its economic feasibility. There are several processes in AD that need attention in cost efficiency. First is pretreatment. The type of pretreatment method used, and materials condition are critical for economic feasibility 30 analysis (Fu et al., 2018). According to Passos et al. (2017), the costs that should be anticipated come from extra energy and chemical agent used for pretreatment. These costs might fluctuate depending on the market conditions and negotiation terms between the companies. In addressing this issue, reuse or energy recovery technologies during pretreatment process might be considered to improve the economic performance (Fu et al., 2018). Second, transportation of starting materials to the digester location can be accounted as energy consumption. Therefore, if the digester is placed near the livestock farms, it can eliminate the initial energy used and the cost of biogas production (Mohammadrezaei et al., 2018). Besides pretreatment and transportation, another method to reduce substantial costs is to improve the efficiency of process steps before feeding biogas, such as harvesting and collection. Simultaneous harvesting and mechanical pretreatment of biomass might become an efficient method to increase energy yield per hectare by selecting an appropriate harvesting machine (Tsapekos et al., 2017). 2.3 Michigan State University South Campus Anaerobic Digester (MSU SCAD) Michigan State University’s South Campus Anaerobic Digester (MSU SCAD) is part of the “Keeping it Green, Recycling Waste to Resource” campus-based projects that concentrate on reducing and reusing waste. It is a single-tank complete mixed anaerobic digester which is projected to utilize approximately 17,000 tons of organic waste per year from MSU and the greater Lansing area to yield biogas than can be converted into more than 2.8 million kWh of electricity per year (Stuever, 2013). Approximately 10% of the energy powers the facility, then the rest offsets energy production in 10 south campus buildings, which is enough energy to power about 250-300 31 homes (MSU Sustainability). The total cost of the project is about $5 million, and it is expected to pay off itself in less than 15 years (Oswald, 2013). Figure 2.2 MSU South Campus Anaerobic Digester (personal documentation) Feedstock materials include dairy manure from MSU Dairy Teaching and Research Center, food waste from campus dining halls, food manufacturing waste from southern Michigan, also fats, oil, and grease (FOG) from local restaurants. The energy produced is used to power several buildings on South Campus (Michigan State University Sustainability, 2014). These feedstocks are received in two reception tanks, one is for manure, and the other one is for other materials. Relying on the delivery schedule and the target blend, feedstock is pumped from each reception tank into a central mix tank to be homogenized. Then, this blended material is pumped through a heat exchanger to raise up the temperature to 37.78 °C before entering the anaerobic digester. The digester is an aboveground steel tank with a liquid capacity of more than 1.7 million liters. Two hydraulically powered submersible mixers are used to keep the digester contents well blended for 25-day hydraulic retention time (Oswald, 2013). The electricity is generated through powering a 450-kW combined heat and power (CHP) system using biogas yielded from the digester. Hot water produced by the CHP is used to maintain 32 the digester temperature to stay at 37.78 °C and to provide heat to the other buildings at the site. The digestate—remaining mixture of solids and liquid after digestion—is pumped to a solid-liquid separator. Separated solids will be composted while liquid will be transferred to digestate holding tank (DHT) which is an aboveground steel tank with a 7.6-million-liter capacity. An airtight membrane will allow the headspace—the space above the digestate—to be used as biogas storage and minimize odors from the systems. The digestate will be land-applied seasonally as carbon-rich fertilizer. This project provides many benefits, such as renewable energy, emissions reduction, landfill and wastewater diversion, and enhanced fertilizer with few weed seeds and first year- available plant nutrients (Stuever, 2013). 2.4 Life Cycle Assessment (LCA) 2.4.1 What is it? Life Cycle Assessment (LCA) is a methodology to evaluate environmental loads of processes and products based on their whole life cycles. The assessment includes the extraction and processing of raw materials, manufacturing, transportation, distribution, use, reuse, maintenance, recycling, and final disposal of a product, process, or system. Therefore, LCA has been widely used due to its integrated way of managing the framework, impact assessment, and data quality. LCA is often called as a “cradle-to-grave" method which explains this process by starting with gathering of raw materials from the earth to create product and ends at a stage where all materials are returned to the earth (Khasreen et al., 2009; Odey et al., 2021; Ram & Sharma, 2017). LCA has gained support from various institutions, including United Nations Environment Programme (UNEP) / Society of Environmental Toxicology and Chemistry (SETAC) Life Cycle 33 Initiative, the Forum for Sustainability through Life Cycle Assessment (FLSCI), International reference Life Cycle Database System (ILCD), the European reference Life Cycle Database System (ELCD) (Odey et al., 2021). According to DEAT (Department of Environmental Affairs and Tourism) (2004), LCA is currently implemented by large groups of users, such as: • Industry and other commercial enterprises • National governments and local, national, and intergovernmental regulatory bodies • NGOs (consumer organizations and environmental groups) • Consumers (which includes governments as consumers) Nevertheless, LCA has several drawbacks to address, which are the absence of a perceived need for LCA, scarcity of LCA expertise, access to high-quality data, and incorrect perception of the application of LCA in relation to other tools (Department of Environmental Affairs and Tourism of South Africa, 2004). 2.4.2 Process LCA implicates a thorough assessment of environmental aspects of a product system, including all stages of its life cycle, by three major processes: 1) collecting an inventory of relevant inputs and outputs of a system; 2) formulating a thorough evaluation of the potential environmental impacts correlated with those inputs and outputs; and 3) interpreting the results in correlation to objectives of the study (Jensen et al., 1997). LCA is conducted based on ISO 14040 which consists of four analytical stages (Khasreen et al., 2009): • Defining goal and scope • Creating the life-cycle inventory • Assessing the impact 34 • Interpreting the results 2.4.3 Parameters 2.4.3.1 Water The amount of freshwater on Earth is only about 2% of all water on the planet. In terms of access to water, approximately one out of six people on Earth lacks access to drinking water. A product’s life cycle, either directly or indirectly, has a strong correlation with water consumption. Based on LCA methodologies mentioned in ISO 14040:2006, water is a parameter assessed in water consumption potential (WCP) section, which is described as water that has been removed from the watershed and cannot be returned (Arosemena, 2021). 2.4.3.2 Nutrients The primary concern about nutrients in the LCA study is the impact of excess nutrients that pollute the environment, such as land, water body, and air. This correlates nutrients with several impacts studied in LCA, which are Water Eutrophication Potential (WEP) and Air Acidification Potential (AAP). Eutrophication is a situation where a water body contains excessive nutrients that affect the dense growth of plant life and the death of water animals due to a lack of oxygen. It is due to nutrients runoff from the land, such as nitrogen and phosphorus, which then accumulate in the water. As the consequence, it creates a “dead zone” which is an area with low oxygen content that suffocates marine life (Mueller and Helsel, 1996). Water eutrophication potential (WEP) is the impacts resulting from excessive nutrient supplies on terrestrial and aquatic environments, particularly the most important substances such as nitrogen (N) and phosphorus (P). WEP can be presented as either mass of nitrogen equivalents (kg N-eq.) or phosphate equivalents (PO4-eq.) (Guinee, 2002). 35 According to the EPA, atmospheric acidification can be defined as: “the result of the oxidation of sulfur, nitrogen, and organic compounds to form their corresponding acids” (Durham & Demerjian, 1985). When absorbed by the atmosphere, these acids can lead to conditions such as acid rain. Air acidification potential (AAP) is an impact category used to convert processes or materials that form acid rain into common units of sulfur dioxide equivalents (SO2-eq.). 2.4.3.3 Greenhouse Gases (GHG) Greenhouse gases is a parameter in LCA that is assessed into Global Warming Potential (GWP) which is the amount of GHG released during the life cycle of a process. Carbon dioxide is commonly used as a reference gas to compare the impact of various greenhouse (Shine, 2009). Fossil fuel consumption contributes approximately 65% of GHG emissions (Environmental Protection Agency (EPA), 2019). Therefore, renewable energy is expected to address this concern. 2.5 Revenue Value As an emerging renewable energy source in the United States, anaerobic digestion provides various forms of revenues from renewable electricity, digestate, energy or fuel credits, and feeding fees. Nevertheless, the prices vary depending on location, operating procedures, and state regulations. One example is tipping fees, which are rates that a company should pay for disposing of waste materials as digester input. Tipping fees depend on the water content of materials. Fats, Oils, and Grease (FOG) have a monetary value of $0.10 per gallon due to high water content. However, the price can be about $0.05 per gallon if it has a lower water concentration or as dry material. On the other hand, digestate can be traded to farms or composting facilities for approximately $7.00 per ton (Dr. Dana Kirk, 2020). 36 Other potential revenues are RINs and RECs as mentioned previously. RIN price ranges between $0.01 and $3.50 depending on the biofuel source. RECs prices are various following changes in policy and the availability of RECs within the state. Currently, most RECs are fulfilled through sources such as wind and solar which creates a deviation in the market for prices. In Michigan itself, RPS increased from 10% in 2015 to 15% in 2021, while it has compliance requirements of 12.5% in 2019 and 2020 (Michigan Public Service Commission, 2021). REC is regulated by The Michigan Renewable Energy Certification System (MIRECS) as a tracking and certification system. According to Consumers Energy, REC price is $0.014 per kWh or $14/MWh. Additionally, Michigan’s electric providers in total retired 12,812,152 RECs in 2019 (Scripps et al., 2016). 37 CHAPTER 3. MATERIAL AND METHODS 3.1 Manure Feedstock The South Campus Anaerobic Digester (SCAD) digester uses various kinds of feedstock, including dairy, beef, waste feed, poultry, and swine manure. Most feedstocks are from the Michigan State University dairy farm which is located adjacent to the SCAD. 3.2 Food Waste Feedstock Food waste used in the digester included pineapple, pulp, FOG (fat, oil, and grease), and waste feed. A small portion of the food waste comes from Michigan State University dining halls, while the rest comes from off-campus food processors and manufacturers. Most of the food waste comes from southern Michigan, but a small portion of food waste was also gained from the neighboring states such as Indiana and Ohio. 3.3 Real-time (commercial system) sampling methods The process flow diagram shown in Figure 3.1 provides a better view of the sampling locations. 3.3.1 Gas quantity and quality, and electricity production Gas quantity was measured by using Endress Hausser Proline t-mass 65 flow meter in real time 24 hours as CHP (Combined Heat and Power) daily total in standard cubic foot (SCF) unit. Gas quality was measured by using AwiFLEX Cool+ gas analyzer every hour. Gas was taken from 2 distinct locations which were before and after the carbon activator scrubber. Three gases were measured which were CH4 (%), O2 (%), and H2S (ppm). Electricity was measured as daily electrical power generated in kilowatts hour (kWh) unit by using SATEC PM172E. The 38 measurement was recorded by the software in the morning until May 2015, then it changed to be in the midnight afterwards. Figure 3.1. Sampling and measurement maps (Source: MSU Anaerobic Research and Education Center) (Notes: black line: feedstock; yellow line: separated solids; blue line: filtrate; green dashed line: biogas phase; orange line: sample collection and measurement) 3.3.2 Digestate, filtrate, and solids sampling Digestate or effluent from the digester consists of a mix of cow manure from MSU Dairy Farms and food waste. The effluent was collected from a piping line that connects the digester to the solid separator. The effluent was taken before the slurry went through a solid separator. The ratio between cow manure and food waste changes daily based upon material received. The filtrate was collected after solid liquid separation from a line that connects the solid separator to the digestate holding tank (DHT). Solids were taken at the solid separator station. These samples were regularly sent to ADREC for laboratory analysis. The frequency of sampling varied depending on 39 the workforce availability in the laboratory. However, in general, the sampling was at least once a month. 3.3.3 Temperature sampling The temperature of digester was measured continuously by using Endress Hausser T13 RTD probe that is located about 3 feet off the floor on the southwest area of the digester. The measurement results were recorded by the Allen Bradley/Rockwell compact logix PLC processor every 10 minutes and put into a CSV file. The measurement results were also recorded by the digester manager twice a day in the morning and at the end of working hours. 3.3.4 pH sampling The pH of slurry in the digester was measured continuously by using Endress Hauser Liquiline C CM42, which is in the pipe that goes through a heat exchanger. The measurement results were recorded by the Allen Bradley/Rockwell compact logix PLC processor every 10 minutes and put into a CSV file. The measurement results were also recorded by the digester manager twice a day in the morning and at the end of working hours. 3.4 Laboratory Analysis The laboratory analysis for SCAD samples was done at the MSU Anaerobic Digestion Research and Education Center (ADREC). 3.4.1 Total Solids and Volatile Solids Total and volatile solids (TS and VS) are the fundamental feedstock and digestate measure for SCAD operational management. In addition to TS and VS, this test also yields information regarding the sample moisture content (MC) and fixed solids (FS). TS and VS analysis were performed following the EPA accepted Hach methods 8271 and 8276, respectively. The procedure 40 for TS was modified from a 6-hour oven holding time to 24 hours to ensure complete drying. The procedure for VS was also adjusted by increasing the time from 1 hour to 6 hours to ensure complete sample combustion. Materials and equipment needed for the test included the digester’s samples (filtrate, effluent, and solids), 50 mL glass beakers (3 per sample), laboratory analytical balances (Scientech SA 120), oven (Precision Scientific, Catalog No. 31578-10), furnace (Lindberg, Model No. CBFM516C), stir plate (Cole-Parmer Instrument Company, Catalog No. 03406-10), magnetic stir bar, desiccator (Boekel 1342), desiccant (Drierite 22001), syringe, spoon, marker, and white board. 3.4.2 Chemical Oxygen Demand Chemical Oxygen Demand (COD) is used as a measure of pollutant in wastewater or effluent. The results indicate the concentration of pollutants in the sample. The higher COD content, the more polluted the sample. The test was performed following the EPA accepted Hach method 8000. Materials and equipment needed were diluted samples using DI water, HACH heated reactor DRB 200, HACH spectrophotometer DR 5000, HACH COD test vials (Catalog No. 2125915), blank test vial, stir bar and stir plate (Cole-Parmer Instrument Company, Catalog No. 03406-10), micropipette, microtips, and delicate wipes. 3.4.3 Total Suspended Solids / Volatile Suspended Solids Total Suspended Solids / Volatile Suspended Solids (TSS/VSS) test is a method to determine the amount of total suspended solids and total volatile suspended solids found within a sample. TSS/VSS tests were performed following the EPA accepted Hach methods 8158 and 8164, respectively. Materials and equipment needed were laboratory analytical balances (Scientech SA 120), desiccator, oven (Precision Scientific, Catalog No. 31578-10), furnace (Lindberg, Model No. CBFM516C), stir plate (Cole-Parmer Instrument Company, Catalog No. 03406-10), magnetic stir 41 bar, vacuum filtration system, glass-microfibre discs filter 47 mm (HACH, Catalog No. 253000), tweezers, watch glass, aluminum crucible, DI water, 1 ml syringe, and tongs. 3.4.4 pH and Electrical Conductivity pH and electrical conductivity (EC) are key measurements to monitor the biological health of a digester hence liquid samples of the digester’s filtrate and effluent were evaluated weekly. The recommended range for pH is between 6.4 and 8.2 for a healthy digester (MSU ADREC Operator Training, 2019). Prior to use, the pH probe was calibrated using three calibration solutions at pH 4.01, 7.00, and 10.00. After calibration, pH, and electronic conductivity (EC) probes were rinsed with DI water and wiped using delicate wipes. Sample was stirred on the stir plate using magnetic stirrer; then, the probes were dipped into sample. Both probes were rinsed and wiped after use and the pH probe was stored in a storage solution. Materials and equipment needed were pH/Conductivity meter (Orion Star A215), pH probe (Orion 8157BNUMD), conductivity probe (Orion 013005MD), double ionized (DI) water, calibration solutions for pH 4.01 (Millipore Sigma BX1634), 7.00 (Millipore Sigma 7BX1635), and 10.00 (Millipore Sigma BX1642), delicate wipes, stir bar, stir plate (Cole-Parmer Instrument Company, Catalog No. 03406-10), magnetic collection stick, and empty beakers. 3.4.5 Alkalinity and Volatile Fatty Acids Alkalinity and Volatile Fatty Acids (VFA) tests were conducted using the titration method to understand the susceptibility of the digester towards the change within its internal environment. Alkalinity and VFA tests were conducted by referencing O’Brien and Donlan (1977) methods, as follows: • Samples were centrifuged for 30-40 minutes 42 • After centrifugation, liquids were filtered using 23 µm then 11 µm filters to gain 50 mL of liquids only • Filtered samples were then poured into 150 ml and stirred using stir bar on the stir plate • pH was measured using pH probe prior to titration to know the initial pH • Samples were gently stirred during titration using 1.0 N H2SO4 to a pH of 3.3, then the volume reading is noted. All processes up to this point were part of the alkalinity test • After the first titration, sample beaker was covered with 65 mm watch glass and then heated on the heated stir plate to a gentle boiling point for 3 minutes • The sample was then cooled down to room temperature. After that, the watch glass was rinsed into the beaker with DI water • Sample was then titrated again using 0.05 N NaOH to pH 4.0, then volume reading was noted • Lastly, sample was titrated again using 0.05 N NaOH to pH 5.1 without refilling the solution from the previous titration, then the volume reading is totaled with the second titration. This entire process was part of the VFA test Materials and equipment needed were centrifuge (Hermle Labnet Z 206 A), centrifuge vials 50 mL, 150 ml sample beaker, empty beaker, 23 µm (Whatman, Catalog No. 1441-047) and 11 µm filters (Whatman, Catalog No. 1001-047), filter flask (brand), tweezers, pH/Conductivity meter (Orion Star A215), pH probe (Orion 8157BNUMD), delicate wipes, stir bar, heated stir plate (Cole-Parmer Catalog No. EW-03407-36), cone hood, stir plate (Cole-Parmer Catalog No. 03406- 10), 65 mm watch glass, double ionized (DI) water, 1.0 N H2SO4, and 0.05 N NaOH. 43 3.4.6 Total Nitrogen Nitrogen is an important nutrient to both digester performance and fertilizer use of digestate. Total Nitrogen (TN) is a test to determine the total nitrogen by the per sulfate digestion method. The test was performed following the instructions available for the HACH TNT 827 test kit. Materials and equipment used were dilution of samples for range used, HACH heated reactor DRB 200, HACH nitrogen test kit high range (TNT 827), HACH spectrophotometer DR 5000, 20 mm reaction tube, stir bar and stir plate (Cole-Parmer Instrument Company, Catalog No. 03406- 10), glass beakers, double ionized (DI water), delicate wipes, micropipette, and microtips. 3.4.7 Ammonia Ammonia can cause toxicity in digesters if levels exceed 3,000 mg/L (MSU ADREC Operator Training, 2019). An ammonia test was conducted to analyze the ammonia content in the samples. The test was performed following the instructions available for the HACH TNT 832 test kit. Materials and equipment needed were diluted samples using DI water, HACH ammonia test kit high range (TNT 832), HACH spectrophotometer (DR 5000), stir bar and stir plate (Cole- Parmer Instrument Company, Catalog No. 03406-10), micropipette, microtips, and delicate wipes. 3.4.8 Total Phosphorus Phosphorus is the key limiting nutrient for land application of digestate as a fertilizer. In organic wastes, phosphates are present in organic and condensed inorganic forms. Phosphorus can be obtained from treating the samples with acid and heat providing conditions for hydrolysis of the condensed inorganic forms. Total Phosphorus (TP) test was conducted to measure the phosphorus content in filtrate and effluent samples. The test was performed following the instructions available for the HACH TNT 844 test kit. Materials and equipment needed were diluted samples using DI water, HACH heated reactor DRB 200, HACH phosphorus test kit high 44 range (TNT 844), HACH spectrophotometer DR 5000, stir bar and stir plate (Cole-Parmer Instrument Company, Catalog No. 03406-10), micropipette, microtips, and delicate wipes. 3.4.9 Other laboratory tests There were several tests conducted outside of ADREC laboratory due to facilities availability. Those tests were conducted by A&L Great Lakes Laboratory (algreatlakes.com) in Fort Wayne, Indiana. The tests included manure nutrition analysis which were moisture, solids, Total Kjeldahl Nitrogen (TKN), phosphorus, potassium, sulfur, calcium, magnesium, sodium, iron, aluminum, manganese, copper, and zinc. All test methods were referred to Recommended Methods of Manure Analysis, UW A3769, summarized in Table 3.1 below. Table 3.1 List of test methods done by A&L Great Lakes Laboratory Parameter Method Moisture UW A3769 III.2 Solids were calculated from Solids moisture Total Kjeldahl Nitrogen (TKN) UW A3769 III.3.2 Phosphorus, Potassium, Sulfur, Magnesium, Calcium, UW A3769 III.6.3. All minerals ran Sodium, Aluminum, Copper, Iron, Manganese, Zinc on Thermo iCAP 6500. 3.5 Statistical Analysis Three analyses were conducted for this research: mass and energy balance, life cycle impact assessment, and economic analysis. The original data from SCAD manager were reorganized to ease data calculation and further analysis. Data was organized in an Excel spreadsheet and used to calculate the descriptive analysis such as min, max, mean, average, standard deviation, number of samples, and coefficient of variation. Other spreadsheets were created from the reorganized spreadsheet to provide data sources for statistical analysis by using the R Studio programming software. 45 Multi linear regression (MLR) in R Studio programming software was used to determine the model for gas production from available feedstock combinations. The codes utilized for MLR analysis are provided in Appendix A. MLR formula is defined as equation below: (2) Furthermore, R Studio programming software was also used to create radar and violin charts to check data distribution of each feedstock year to year. A radar chart, also known as a spider plot, is used to visualize the amount of feedstock received by the digester each month and year. Radar chart was created by using fmsb and ggradar packages in R software. The codes utilized for Radar chart are provided in Appendix B. A violin chart is used to visualize the distribution of individual feedstock amount year to year, also output parameters of the digester. Violin chart is created by using ggplot2 and geom_violin packages in R software. ANOVA (analysis of variance) Tukey multiple comparison was used to determine statistically significant differences between the various operational parameters via the R function TukeyHSD. The codes utilized for Violin chart and ANOVA Tukey multiple comparison are provided in Appendix C. 46 CHAPTER 4. ANALYSIS OF SCAD OPERATIONAL DATA 4.1 Feedstock Amount During the operation period January 1, 2014, and ending December 31, 2020, the SCAD utilized 18 different feedstocks. Feedstocks were managed in two reception tanks, one for manure (low energy materials) and another for food waste (high energy materials). Table 4.1 summarizes the feedstocks and reception tank used to manage the inflow. The composition of feedstock received by the digester is variative each year. Table 4.1 Feedstocks in SCAD Manure Pit and Food Pit No Manure Pit Food Pit 1 Digestate (recycle) Filtrate (recycle) 2 Filtrate (recycle) Cart Food 3 ANS Other Fat, Oil, and Grease (FOG) 4 Beef Manure Other 5 Dairy Gutter Manure (Dairy G.) Pineapple (P.A.) 6 Dairy Freestall Manure (Parlor) Pulp 7 Poultry Manure SLS Solids 8 SLS Solids Waste Feed 9 Swine Manure 10 Waste Feed 4.1.1 Yearly total The yearly total of feedstock processed in SCAD is shown in Table 4.2 for manure pit and Table 4.3 for food pit. Digestate is recycled directly from the digester effluent without solid-liquid separation. Filtrate is the liquid generated after coarse solids are separated from digestate. Digestate and filtrate are recycled to thin out pits and to dilute high TS feedstocks. SLS Solids are 47 Table 4.2 Total Feedstock in Manure Pit Year 2014-2020 Feedstock (metric ton) Year SLS Dairy Waste ANS Digestate Filtrate Parlor Beef Poultry Swine Solids G. Feed Other 2014 5,643 4,104 4,273 697 17 25 4 2015 311 3,706 192 4,542 4,754 121 3 144 30 2016 1,001 412 56 5,091 4,200 243 34 5 17 2017 656 171 229 4,878 4,787 222 45 13 78 29 2018 296 179 42 4,855 5,110 220 151 5 2019 94 50 4,885 5,877 391 33 20 3 2020 23 45 59 4,664 5,388 7 56 15 75 Max 1,001 5,643 229 5,091 5,877 697 56 20 151 75 Min 23 45 42 4,104 4,200 7 3 5 25 3 Mean 268 380 82 4,707 4,882 160 23 12 81 13 Average 457 1,465 105 4,717 4,913 272 31 13 100 23 St. Dev 378 2,266 83 322 599 221 19 6 59 26 Coefficient 141% 597% 101% 7% 12% 139% 83% 54% 73% 191% of Variation 48 the coarse fiber separated from digestate via solid-liquid separation. SLS Solids are added solids back into the digester to thicken thin, low TS, feedstocks. ANS stands for animal science, meaning that there were research materials added into the digester in low quantities, for example eggs from research chickens. There were several years where the digester did not receive certain types of feedstocks. SLS Solids were not used in 2014. Poultry manure did not come to the digester in 2014, 2015, and 2018. Furthermore, there was no swine manure in 2016, 2019, and 2020. Pineapple was only available until 2016. Waste feed was absent in 2018. Several feedstocks, Dairy Gutter, and Parlor as well as FOG, produced the largest annual mass; meanwhile, the others had significant changes in a certain period. Moreover, each feedstock has a different peak of receiving in the digester. According to manure pit data (Table 4.2), the digester received the highest amount of SLS Solids in 2017; Dairy G in 2016; Parlor in 2019; Beef in 2014; Waste Feed in 2020; Poultry in 2019; and Swine in 2018. The average of each feedstock received by SCAD during 2014-2020 is 105 metric tons for SLS Solids, 4,717 metric tons for Dairy Gutter, 4,913 metric tons for Parlor, 272 metric tons for Beef, 31 metric tons for Waste Feed, 13 metric tons for Poultry, 100 metric tons for Swine, and 23 metric tons for ANS Other. Regarding the coefficient variation of feedstock, it showed that Dairy Gutter and Parlor had the most consistent amount that the digester received year by year. The coefficient variation for Dairy G and Parlor was 7% and 12% respectively, showing that the number did not significantly change year over year (Figure 4.1). These two feedstocks were also the major feedstocks in manure pit. Based on calculation of Dairy Gutter and Parlor in the total of feedstocks received in manure pit, the percentage of these feedstocks increased every year. Dairy G and Parlor contributed 57% of total feedstock in manure pit in 2014, 67% in 2015, 84% in 2016, 87% in 2017, 49 92% in 2018, 95% in 2019, and 97% in 2020. It shows that the digester improves the feedstock composition which affects gas production. On the other hand, Filtrate significantly decreased since 2016 due to the conclusion of pineapples waste reception. The primary reason for the addition of filtrate and SLS Solids to the reception pits is to create suitable feedstock for pumping and mixing, TS less than 8%. In 2016 and prior, the digester received a vast amount of pineapples waste that required excessive amount of filtrate to dilute and breakdown the waste before pumped to the mix tank. Meanwhile, on food pit data (Table 4.3), the digester received the highest amount of SLS Solids in 2019, Pineapple in 2014, Pulp in 2014, FOG in 2020, Waste Feed in 2015, Other Feedstock in 2015, and Cart Food in 2015. The average amount of feedstocks received by the digester in 2014-2020 was 152 metric tons for SLS Solids, 2,098 metric tons for Pineapple, 78 metric tons for Pulp, 8,380 metric tons for FOG, 18 metric tons for Waste Feed, 388 metric tons for Other Feedstocks, and 145 metric tons for Cart Food. Regarding coefficients of variation, there was no feedstock in the food pit that was as consistent as Dairy G or Parlor in manure pit. FOG was the most consistent followed by Pulp and Pineapple, with coefficients of variation of 31%, 38% and 44%, respectively (Figure 4.2). Pulp feedstock which comes from Brody dining hall experienced a great reduction in 2020 due to COVID-19 pandemic that limited students' activities in the residence halls after March 15, 2020. Moreover, the Pineapple contract with the digester was only until 2016. The supplier proceeded to compost the feedstock afterwards, hence there were no more supplies starting in 2017. Food carts also experienced a significant reduction starting in 2017 due to contamination with “debris.” Overall, FOG dominated feedstock supplies in the food pit. 50 Table 4.3 Total Feedstock in Food Pit Year 2014-2020 Feedstock (metric ton) Year SLS Waste Cart Filtrate Pineapple Pulp FOG Other Solids Feed Food 2014 1,321 2,681 102 4,005 24 117 283 2015 1,804 53 2,493 94 7,792 28 171 366 2016 1,312 271 1,118 80 8,470 154 320 2017 577 85 77 8,330 45 15 2018 201 196 80 8,068 2 44 14 2019 194 275 92 9,875 92 12 2020 256 35 20 12,122 2,094 3 Max 1,804 275 2,681 102 12,122 28 2,094 366 Min 194 35 1,118 20 4,005 2 44 3 Mean 564 115 1,955 71 8,016 10 142 43 Average 809 152 2,098 78 8,380 18 388 145 St. Dev 660 109 853 27 2,444 14 754 169 Coefficient 117% 94% 44% 38% 30% 138% 530% 396% of Variation 51 6,000 5,000 Quantity (metric ton/year) 4,000 3,000 2,000 1,000 0 2014 2015 2016 2017 2018 2019 2020 Year Digestate Filtrate SLS Solids Dairy G. Parlor Beef W. Feed Poultry Swine ANS Other Figure 4.1 Comparison of feedstock received by SCAD in manure pit 14,000 12,000 Quantity (metric ton/year) 10,000 8,000 6,000 4,000 2,000 0 2014 2015 2016 2017 2018 2019 2020 Year Filtrate SLS Solids P.A. Pulp FOG W. Feed Other Cart Food Figure 4.2 Comparison of feedstock received by SCAD in food pit A ratio of total feedstock between manure pit and food pit was calculated. Based on Table 4.4, the ratio of feedstock from food pit ranged from 37% to 58% with the highest percentage being in 2020. On average, SCAD has received 11,897 metric tons of feedstock in manure pit each 52 year and 10,838 metric ton of feedstock in food pit, for a total of 22,375 metric ton/year with the average ratio of feedstock in food pit was 47%. Table 4.4 Ratio of Manure Pit and Food Pit Manure Pit Food Pit Total Food Year (Metric (Metric (Metric Pit ton/year) ton/year) ton/year) (%) 2014 14,763 8,533 23,297 37% 2015 13,805 12,800 26,605 48% 2016 11,059 11,726 22,785 51% 2017 11,109 9,129 20,238 45% 2018 10,859 8,605 19,464 44% 2019 11,353 10,539 21,893 48% 2020 10,332 14,531 24,863 58% Max 14,763 14,531 26,605 58% Min 10,332 8,533 19,464 37% Mean 11,802 10,637 22,618 47% Average 11,897 10,838 22,735 47% St. Dev 1,683 2,294 2,499 7% 4.1.2 Monthly Average The monthly average of feedstock processed in SCAD is shown in Table 4.5 and 4.6. FOG, Parlor, and Dairy Gutter are three major feedstocks that the digester received each month during the years 2014-2020. Radar charts in Figure 4.3 to 4.9 also give a better picture of feedstock peaks for each month. In 2014 to 2016, Parlor, FOG, and Filtrate consistently had higher peaks compared to the rest of feedstocks. However, from 2017 to 2020, Parlor and FOG remained as the only feedstocks that have consistent peaks of supply. There were several outliers that happened during 2014-2020, which were Digestate (July 2016), Swine (December 2017), Filtrate in manure pit (August 2017 and May 2018), Filtrate in food pit (July 2018), and FOG (October 2019). 53 Table 4.5 Average Mass Delivered to Digester Each Month (Manure Pit), 2014-2020 Feedstock (kg) Year SLS Dairy Waste ANS Digestate Filtrate Parlor Beef Poultry Swine Solids G. Feed Other 2014 25,840 11,463 25,466 6,354 4,069 12,645 179 2015 20,592 21,259 2,916 12,647 27,034 4,784 1,069 4,537 741 2016 18,979 12,123 3,263 13,844 21,659 3,914 3,426 568 441 2017 15,338 14,884 3,187 13,307 20,612 2,394 1,451 1,071 25,682 919 2018 16,210 15,258 2,749 13,205 23,907 2,878 17,275 688 2019 10,184 3,159 13,287 21,993 3,432 1,847 1,935 1,562 2020 11,259 12,507 3,637 12,678 21,269 3,053 3,159 1,950 3,022 Max 20,592 25,840 3,637 13,844 27,034 6,354 4,069 1,950 25,682 3,022 Min 11,259 10,184 2,749 11,463 20,612 2,394 1,069 568 4,537 179 Mean 16,136 15,259 3,139 12,899 23,031 3,648 2,239 1,231 12,631 779 Average 16,475 16,008 3,152 12,919 23,134 3,830 2,503 1,381 15,035 1,079 St. Dev 3,218 5,173 279 705 2,222 1,255 1,105 589 7,653 888 Coefficient 20% 34% 9% 5% 10% 34% 49% 48% 61% 114% of Variation 54 Table 4.6 Average Mass Delivered to Digester Each Month (Food Pit), 2014-2020 Feedstock (kg) Year SLS Waste Cart Filtrate Pineapple Pulp FOG Other Solids Feed Food 2014 19,922 11,547 1,112 18,952 2,017 10,870 2,151 2015 18,478 3,032 11,093 1,272 35,270 2,146 2,849 1,845 2016 14,955 3,386 11,345 1,317 35,052 1,341 2,417 2017 18,812 2,844 2,013 32,703 1,067 121 2018 22,917 2,802 1,656 31,780 1,620 1,498 117 2019 14,565 3,228 2,088 34,277 2,857 92 2020 14,102 3,160 1,630 40,428 13,222 67 Max 22,917 3,386 11,547 2,088 40,428 2,146 13,222 2,417 Min 14,102 2,802 11,093 1,112 18,952 1,620 1,067 67 Mean 17,424 3,069 11,326 1,546 31,917 1,914 3,059 364 Average 17,679 3,076 11,328 1,584 32,637 1,928 4,815 973 St. Dev 3,031 207 186 346 6,143 224 4,663 1,020 Coefficient 17% 7% 2% 22% 19% 12% 152% 280% of Variation 55 Figure 4.3 Radar chart for feedstock supplies receiveed by SCAD in 2020 Figure 4.4 Radar chart for feedstock supplies receiveed by SCAD in 2019 56 Figure 4.5 Radar chart for feedstock supplies receiveed by SCAD in 2018 Figure 4.6 Radar chart for feedstock supplies receiveed by SCAD in 2017 57 Figure 4.7 Radar chart for feedstock supplies receiveed by SCAD in 2016 Figure 4.8 Radar chart for feedstock supplies receiveed by SCAD in 2015 58 Figure 4.9 Radar chart for feedstock supplies receiveed by SCAD in 2014 4.1.3 Feedstock Distribution This study also observes the distribution of feedstock received by SCAD for each feedstock from 2014-to 2020. Violin charts were used to check the data distribution of each feedstock year to year. The violin chart consisted of a colored area, a white box, and a straight line with top and bottom points. The colored area represents data points that the parameter has. The more colored the area, the larger the data points. A large part of the colored area shows where most data is located. The white box is in the middle of the colored area. The top part of the box represents the third quartile of data points, a line inside the box represents the median, while the bottom part represents the first quartile. The straight line with top and bottom points represents the range of data with the highest and lowest value, respectively. ANOVA (analysis of variance) Tukey multiple comparisons were used to determine the significance of data compared year to year. 4.1.3.1 Manure Pit From manure pit, Filtrate manure pit, Dairy Gutter and Parlor are three feedstocks with the largest data points. Figure 4.10 shows the violin chart for Filtrate manure pit. In general, the charts 59 in 2014, 2015, 2019, and 2020 have distinctive shapes which means the data in these years were distributed differently. Meanwhile, the chart shapes in 2016-2018 show were quite similar. There was a significant range of filtrate circulating into manure pit in 2015 with median and highest data point were about 23,000 kg/day and 40,000 kg/day, respectively. The range got smaller and more consistent in 2016-2018 which the medians were roughly 13,000 kg/day. In 2019 and 2020, data distribution for filtrate supplied to manure pit became more consistent with the range of data points was approximately 5,000 kg/day to 15,000. Lower supply of filtrate shows that digester received fewer solid feedstocks; thus, it did not need much dilution using the filtrate. Figure 4.10 Data distribution of Filtrate manure pit Table 4.7 shows the ANOVA Tukey multiple comparison performed for Filtrate manure pit data distribution. It shows that there was a significant difference in the quantity of Filtrate manure pit received by the digester at the p<0.05 level from year to year [F (6, 49) = 3.729, p = 0.00392). Table 4.7 ANOVA Tukey Multiple Comparison results for Filtrate manure pit data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 1.533e+09 258,810,814 3.729 0.00392 Residuals 49 3.401e+09 69,412,500 60 The violin chart of Dairy Gutter (Figure 4.11) shows different shapes for each year. In 2014, the chart shows an extensive range of data points with most data points located slightly below 12,500 kg/day, while the highest data point was about 16,000 kg/day. There was an outlier in 2017, with data points located around 8,000 kg/day. That means there was a moment where the digester received lower supplies of this feedstock. Meanwhile, feedstock supplies were typically stable in 2018-2020 which the median between 12,500 and 15,000 kg/day. Figure 4.11 Data distribution of Dairy Gutter Table 4.8 shows the ANOVA Tukey multiple comparison performed for Daily Gutter data distribution. It shows that there was a significant difference in the quantity of Dairy Gutter received by the digester at the p<0.05 level from year to year [F (6, 77) = 3.554, p = 0.00368). Table 4.8 ANOVA Tukey Multiple Comparison results for Dairy Gutter data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 42,372,311 7,062,052 3.554 0.00368 Residuals 77 152,983,389 1,986,797 Figure 4.12 shows the data distribution for Parlor. The chart shape was significantly different from year to year. The chart in 2015 shows the most extensive range with the highest peak reaching 40,000 kg/day, becoming the highest data point of all years. Data median also shows 61 two different trends. Median in 2014, 2015, and 2018 were around 25,000 kg/day, while the rest have median around 20,000 kg/day. Figure 4.12 Data distribution of Parlor Table 4.9 shows the ANOVA Tukey multiple comparison performed for Parlor data distribution. It shows that there was a significant difference in the quantity of Parlor received by the digester at the p<0.05 level from year to year [F (6, 77) = 5.017, p = 0.000). Table 4.9 ANOVA Tukey Multiple Comparison results for Parlor data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 4.214e+08 70,235,648 5.016 0.000 Residuals 77 1.078e+09 14,000,917 Figure 4.13 shows the data distribution for Total Manure Pit. The chart shows three different trends. Data distribution in 2014 and 2015 share a similar pattern where the curves have clear violin shapes. Data distribution in 2016, 2017, and 2018 share similar shapes where the curves look like a coke bottle. Data distribution in 2019 and 2020 look similar where the curves have shorter shapes, meaning that the data have more consistent ranges. Data median shows a constant slight decrease from 2014 to 2018. There was quite a significant decrease in data median 62 from 2018 to 2019, then a slight decrease in 2020. Data distribution in 2015 and 2016 shows the most extensive range, from 40,000 to 100,000 kg/day. Data distribution from 2019 to 2020 shows the most consistent data distribution, ranging from roughly 30,000 to 60,000 kg/day. Figure 4.13 Data distribution for total manure pit Table 4.10 shows the ANOVA Tukey multiple comparison performed for total manure pit data distribution. It shows that there was a significant difference in the quantity of total manure pit received by the digester at the p<0.05 level from year to year [F (6, 77) = 7.098, p = 0.000). Table 4.10 ANOVA Tukey Multiple Comparison results for Total Manure Pit data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 8.058e+09 1.343e+09 7.098 0.000 Residuals 77 1.457e+10 1.892e+08 Data for other feedstocks in manure pit are provided in Appendix C.1. In general, some of the feedstocks in the manure pit did not have large enough data points, therefore they show little or no colored area around the white box (SLS Solids, Beef and ANS Other). Meanwhile, other feedstocks did not show a complete chart for each year due to the digester did not receiving them in certain years (Digestate, Waste Feed Manure, Poultry, and Swine). 63 4.1.3.2 Food Pit From the manure pit, Pulp and FOG were feedstocks with the largest data points. Figure 4.14 shows data distribution for Pulp. The charts show different shapes each year with 2015 being the year with the most consistent data distribution, ranging from roughly 1,000 to 1,600 kg/day. The peak of pulp received by the digester happened in 2019 exceeding 3,000 kg/day. Data distribution in 2020 was typically consistent although the supply was disrupted by COVID-19 pandemic. Figure 4.14 Data distribution for Pulp Table 4.11 shows the ANOVA Tukey multiple comparison performed for Pulp data distribution. It shows that there was a significant difference in the quantity of Pulp received by the digester at the p<0.05 level from year to year [F (6, 67) = 8.78, p = 0.000). Table 4.11 ANOVA Tukey Multiple Comparison results for Pulp data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 10,479,022 1,746,504 8.78 0.000 Residuals 67 13,326,784 198,907 64 Figure 4.15 shows the data distribution for FOG. Data distribution was typically consistent from 2015 to 2019. The data median in 2014 was the lowest at around 20,000 kg/day, while data median for 2020 was the highest which was slightly above 40,000 kg/day. As previously shown in the radar chart, the year 2019 had an outlier for the data peak which was around 80,000 kg/day, and outlier compared to the rest of data points which ranged from about 25,000 kg/day to 40,000 kg/day. Figure 4.15 Data distribution for FOG Table 4.12 shows the ANOVA Tukey multiple comparison performed for FOG data distribution. It shows that there was a significant difference in the quantity of FOG received by the digester at the p<0.05 level from year to year [F (6, 77) = 8.095, p = 0.000). Table 4.12 ANOVA Tukey Multiple Comparison results for FOG data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 3.215e+09 535,833,184 8.095 0.000 Residuals 77 5.097e+09 66,195,896 Figure 4.16 shows the data distribution for Total Food Pit. The chart in 2020 shows a perfect upside-down violin, while 2014 looks like a violin shape. Data distribution in 2015, 2016, and 2018 share similar chart shapes. The highest data point happened in 2019 which was around 65 90,000 kg/day, while the lowest data point happened in 2017 which was around 20,000 kg/day, associated with the engine outage. The data range from 2014 to 2016 was typically consistent from slightly above 40,000 kg/day to slightly above 80,000 kg/day, meanwhile the data range from 2018 to 2020 was typically consistent with most data points were approximately between 30,000 kg/day and 80,000 kg/day. Figure 4.16 Data distribution for Total Food Pit Table 4.13 shows the ANOVA Tukey multiple comparison performed for Total Food Pit data distribution. It shows that there was a significant difference in the quantity of FOG received by the digester at the p<0.05 level from year to year [F (6, 77) = 3.641, p = 0.003). Table 4.13 ANOVA Tukey Multiple Comparison results for Total Food Pit data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 4.559e+09 759,802,790 3.641 0.003 Residuals 77 1.607e+10 208,668,601 Data for other feedstocks in food pit are provided in Appendix C.2. In general, some of the feedstocks in the manure pit did not have large enough data points, therefore they show little or no colored area around the white box (Filtrate Food Pit, SLS Solids Food, Other, and Cart Food). 66 Meanwhile, other feedstocks did not show a complete curve for each year due to the digester did not receiving them in certain years (Pineapples and Waste Feed). 4.1.3.3 Total Feedstock (Manure Pit and Food Pit) Figure 4.17 shows the data distribution for Total Feedstock. Data distribution charts in 2015 and 2018 look similar but in a different data range, while the rest of the charts look quite distinctive in shape. Data distribution in 2014 and 2017 were typically similar, ranged approximately from 95,000 kg/day to 155,000 kg/day. The highest data point happened in 2016, while the lowest data point happened in 2019. Data curves in 2016 and 2019 show the most extensive distribution but in a different range. Data distribution in 2018-2020 shows quite a similar range, which means total feedstocks received by the digester were considered consistent in the last three years. Figure 4.17 Data distribution for Total Feedstock Table 4.14 shows the ANOVA Tukey multiple comparison performed for Total Feedstock data distribution. It shows that there was a significant difference in the quantity of Total Feedstock received by the digester at the p<0.05 level from year to year [F (6, 77) = 5.736, p = 0.000). 67 Table 4.14 ANOVA Tukey Multiple Comparison results for Total Feedstock data distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 1.47e+10 2.451e+09 5.736 0.000 Residuals 77 3.29e+10 4.273e+08 4.1.4 Feedstock Characterization Table 4.15 shows the characterization of feedstocks received by SCAD during 2014-2020. Based on the data available at the digester, there are only seven feedstocks that have laboratory results for characterization: Parlor Manure, Beef, Dairy Gutter, FOG, Food Other, Pineapple, and Pulp. Characteristics measured were total solids (TS), volatile solids (VS), pH, electron conductivity (EC), soluble COD (sCOD), total nitrogen (TN), total phosphorus (TP), and ammonia. The data on table represents average, standard deviation, and number of samples in the brackets. According to several samples, TS, VS, pH, and EC were the parameters which were measured the most for Parlor Manure, FOG, and Food Other feedstocks. For Parlor Manure, data collected was 11 for TS, VS, and pH, while it had 10 data for EC. They were taken from 2015 to 2018. Characterization data of Parlor Manure were 63,844±20,998 mg/L, 52,742±18,966 mg/L, 7.01±0.30, and 13.72±1.37 mS/cm for TS, VS, pH, and EC, respectively. Data collected for FOG were 28 for TS and VS, 27 for pH, and 23 for EC. They were taken from 2014 to 2019. Characterization data of FOG were 120,191±172,277 mg/L, 105,384±142,739 mg/L, 5.50±1.51, and 12.17±38.04 mS/cm for TS, VS, pH, and EC, respectively. Standard deviation for FOG samples was large due to the wide range of laboratory results. For example, TS ranged from 1,520 to 689,323 mg/L, VS ranged from 532 to 542,778 mg/L, and EC ranged from 0.95 to 189.90 68 mS/cm. The sample variation is driven largely by the grease interceptor management and the material collection practices. Meanwhile, data collected for Food Other were 24 for TS and VS, 23 for pH, and 22 for EC. They were taken from 2014 to 2019. Characterization data of Food Other were 219,447±286,460 mg/L, 193,795±265,043 mg/L, 5.41±1.45, and 8.52±6.03 mS/cm for TS, VS, pH, and EC, respectively. Similar to FOG, standard deviation of Food Other feedstock characterizations was large due to significant difference between each data point. For example, TS ranged from 24,072 to 950,355 mg/L, VS ranged from 18,045 to 933,805 mg/L, pH ranged from 3.58 to 7.88, and EC ranged from 0.31 to 18.53 mS/cm. Other feedstocks and parameters were collected at a minimum data point. For example, the digester only has 1 data point for Dairy Gutter and Pulp, which was taken in 2016 and 2014, respectively. Beef has 3 data points for TS, VS, pH, and EC yet does not have any data for SCOD, TN, TP, and ammonia. These data points were collected in 2014, 2016, and 2018. Characterization data of Beef were 462,152±109,098 mg/L, 393,620±100,761 mg/L, 8.64±0.06, and 1.89±0.76 mS/cm for TS, VS, pH, and EC, respectively. Pineapple has 2 data points for TS, VS, pH, and EC; one data point for TN, TP, and ammonia, while it does not have any data for SCOD. These data points were taken in 2014 and 2016. Characterization data of Pineapple were 127,389±14,312 mg/L, 114,749±5,820 mg/L, 3.91±0.04, and 2.26±0.62 mS/cm for TS, VS, pH, and EC, respectively. The lack of workforce becomes the main reason for minimum characterization analysis for SCAD feedstocks. In general, feedstock received by SCAD has significant differences in terms of characterization results. 69 Table 4.15 Characterization of SCAD Feedstock a TS VS EC Feedstock pH (mg/L) (mg/L) (mS/cm) Parlor Manure 63,844±20,998 (11) 52,742±18,966 (11) 7.01±0.30 (11) 13.72±1.37 (10) Beef 462,152±109,098 (3) 393,620±100,761 (3) 8.64±0.06 (3) 1.89±0.76 (3) Dairy Gutter 162,268 (1) 142,940 (1) 8.23 (1) 7.74 (1) FOG 120,191±172,277 (28) 105,384±142,739 (28) 5.50±1.51 (27) 12.17±38.04 (23) Food Other 219,447±286,460 (24) 193,795±265,043 (24) 5.41±1.45 (23) 8.52±6.03 (22) Pineapple 127,389±14,312 (2) 114,749±5,820 (2) 3.91±0.04 (2) 2.26±0.62 (2) Pulp 275,459 (1) 262,105 (1) 4.36 (1) 1.49 (1) a data including average ± standard deviation, and (number of sample) Table 4.15 TN TP NH3 Feedstock SCOD (mg/L) (mg/L) (mg/L) Parlor Manure 25,900 (1) 2,190±214 (2) 1,210 (1) 932 (1) Beef - - - - Dairy Gutter - 5,050 (1) 1,386 (1) 1,085 (1) FOG 109,380±167,383 (4) 15,250 (1) 300 (1) 253.75 (1) Food Other - 9,139±10,502 (4) 1,032±1,360 (7) 195±363 (7) Pineapple - 595 (1) 109 (1) 62.70 (1) Pulp - - - - 70 4.2 Biogas and Electricity Production Table 4.16 below shows the total amount of biogas produced by the digester in the year 2014-2020, Methane (CH4) and Hydrogen Sulfide (H2S) average. The highest biogas production was in 2016, reaching 1,418,746 Standard Cubic Meter (SCM). In terms of CH4 percentage, the years 2018 and 2020 had the highest average which was 66%. The lowest H2S average was in 2014 at 360 ppm. Table 4.16 Biogas production, methane and H2S average year 2014-2020 CH4 H2S Electricity Biogas Total Year Average Average production (SCM) (%) (ppm) (kWh) 2014 846,232 63% 360 1,727,073 2015 1,103,695 61% 433 2,118,966 2016 1,418,746 62% 667 1,470,356 2017 1,326,335 65% 421 2,169,693 2018 1,348,024 66% 387 2,680,954 2019 1,280,438 64% 520 2,333,449 2020 1,340,179 66% 652 2,664,665 The digester experienced an outage of the CHP engine twice. The first outage occurred from December 2015 to April 2016, when the engine had a rod bearing failure. Consequently, the engine was rebuilt and back to daily operation from May 2016 to November 2016. Furthermore, the engine experienced a second failure with a similar issue. Therefore, CHP was not operating from December 2016 to January 2017. To overcome this condition, the digester installed a new engine in February 2017; thus, the electricity production is running again on the fourth week of February 2017. Therefore, gas production in 2016, as shown in Figure 4.18, experienced a significant decrease compared to the other years. Additionally, biogas produced during the engine 71 failure solely went to flare to burn the methane and reduce GHG released to the atmosphere; and to the boiler to heat the slurry. Another disturbance happened in 2017 when the feed additive was added to the digester and affected biogas production. July had the lowest production compared to the other months. 1,600,000 3,000,000 1,400,000 2,500,000 1,200,000 Electricity (kWh) 2,000,000 Biogas (SCM) 1,000,000 800,000 1,500,000 600,000 1,000,000 400,000 500,000 200,000 0 0 2014 2015 2016 2017 2018 2019 2020 Year Biogas Electricity Figure 4.18 Correlation of SCAD biogas and electricity over the year 4.3 SCAD Output Data Distribution Data distribution analysis for SCAD output, including biogas production, CH4 content, H2S content, electricity production, and laboratory analysis of effluent (total solids, total nitrogen, total phosphorus, pH, and VFA). Like feedstock analysis, the violin chart was also used to check the data distribution of each parameter year to year. Data distribution for biogas production is shown in Figure 4.19. Among all years, biogas production in 2018 has the most consistent chart with a closed data range between 3,500 to 4,000 m3/day, followed by 2020. Meanwhile, biogas production data distribution in 2015 has the largest 72 range between 700 to 4,700 m3/day. Biogas production in 2017 has an outlier due to feed additive added to the digester during summer. In terms of the median of data, biogas production in 2017, 2018, and 2020 had a similar median of around 3,600 m3/day, while 2019 had a slightly lower median than those three years, which was around 3,400 m3/day. Figure 4.19 Data distribution for biogas production Table 4.17 shows the ANOVA Tukey multiple comparison performed for biogas production data distribution. It shows that there was a significant difference in the quantity of biogas produced by the digester at the p<0.05 level from year to year [F (6, 77) = 7.678, p = 0.000). Table 4.17 ANOVA Tukey Multiple Comparison results for biogas production distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 20,642,104 3,440,351 7.678 0.000 Residuals 77 34,500,609 448,060 Data distribution for CH4 content is shown in Figure 4.20. In general, methane content in biogas produced by SCAD ranged from 58% to 69%. Methane content in 2018 has the most consistent distribution, ranging from 64% to 67%. Meanwhile, methane content in 2014 has the 73 widest distribution, from 60% to 68%. Overall, the methane content is in a good range which biogas typically contains 55-70% of CH4 (Chynoweth et al., 2001; Somers et al., 2018). Figure 4.20 Data distribution for CH4 content Table 4.18 shows the ANOVA Tukey multiple comparison performed for CH4 content data distribution. It shows that there was a significant difference in the percentage of CH4 content at the p<0.05 level from year to year [F (6, 77) = 10.79, p = 0.000). Table 4.18 ANOVA Tukey Multiple Comparison results for CH4 content distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 268.5 44.75 10.79 0.000 Residuals 77 319.3 4.15 Figure 4.21 provides the data distribution for H2S content. H2S is a toxic gas which poses safety concerns for people and may result in additional equipment maintenance (Hosseini and Wahid, 2014). H2S can cause corrosion to stainless steel and copper and nickel alloys. It also can produce H2SO3 from reaction with water vapor and form sulfur dioxide (SO2) from its combustion, leading to metal pipe and engine corrosion and gas leaks. H2S concentration in biogas ranges between 50 and 10,000 parts per million (ppm), depending on the feedstocks (X. Wang et al., 74 2018). H2S content in SCAD biogas ranging from 3 to 2,251 ppm, which is still in an acceptable range. H2S content in 2016 has the most extensive range, followed by 2020 and 2019. Meanwhile, H2S content in 2017 shows the most stable data distribution, ranging from 98 to 799 ppm. Figure 4.21 Data distribution for H2S content Table 4.19 shows the ANOVA Tukey multiple comparison performed for H2S content data distribution. It shows that there was not any significant difference in H2S content (p value>0.05) from year to year [F (6, 77) = 0.979, p = 0.445). Table 4.19 ANOVA Tukey Multiple Comparison results for H2S content distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 1,130,830 188,472 0.979 0.445 Residuals 77 14,823,077 192,507 Figure 4.22 provides the data distribution for electricity production per day. In terms of chart shape, data in 2014 and 2018 have quite similar shape but in a different data range. Data distribution for electricity was typically consistent in 2016-2020 with data median ranging between 6,500 to 7,600 kWh/day. Electricity production in 2015 had the highest peak but also the widest 75 range of all years, which the data ranges from 2,200 to 9,700 kWh/day. Meanwhile, violin curves in 2016 and 2017 show the time when SCAD did not produce electricity due to engine issues, showed by outliers happened in 2017. Figure 4.22 Data distribution for electricity production Table 4.20 shows the ANOVA Tukey multiple comparison performed for electricity production data distribution. It shows that there was a significant difference in electricity production at the p<0.05 level from year to year [F (6, 71) = 7.397, p = 0.000). Table 4.20 ANOVA Tukey Multiple Comparison results for electricity production distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 58,173,817 9,695,636 7.397 0.000 Residuals 71 93,063,092 1,310,748 Data distribution for effluent TS is shown in Figure 4.23. In general, TS content ranges 40,000 to 85,000 mg/L. Data distribution in 2018 has the most consistent shape, ranging from 53,000 to 65,000 mg/L. Meanwhile, there was an outlier that happened in 2016 where TS content was 85,000 mg/L. TS content in effluent had an increase trend from 2014 to 2017, then became more consistent in 2018-2020 with range from 50,000 to 73,000 mg/L. Data median also increased 76 from 2014 to 2017 from 50,000 to 68,000 mg/L, then became more consistent in 2018 to 2020 which was around 60,000 mg/L. Figure 4.23 Data distribution for effluent TS Table 4.21 shows the ANOVA Tukey multiple comparison performed for effluent TS data distribution. It shows that there was a significant difference in effluent TS at the p<0.05 level from year to year [F (6, 64) = 10.62, p = 0.000). Table 4.21 ANOVA Tukey Multiple Comparison results for effluent TS distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 2.907e+09 484,541,468 10.62 0.000 Residuals 64 2.919e+09 45,609,925 Figure 4.24 shows the data distribution for effluent TN. TN data distribution ranges from 1,400 to 5,000 mg/L. Data distribution fluctuated year by year, but it shows a constant decrease from 2016 to 2018. There was an outlier that happened in 2020 where TN was close to 5,000 mg/L. Data median in 2018-2020 were consistent around 3,000 mg/L. TN content in manure was not changed during anaerobic digestion. The process transforms protein and urea nitrogen into inorganic nitrogen such as ammonia gas (NH3) and ammonium ion (NH4+) (Field et al., 1984). 77 Figure 4.24 Data distribution for effluent TN Table 4.22 shows the ANOVA Tukey multiple comparison performed for effluent TN data distribution. It shows that there was a significant difference in effluent TN at the p<0.05 level from year to year [F (5, 44) = 4.072, p = 0.004). Table 4.22 ANOVA Tukey Multiple Comparison results for effluent TN distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 5 9,543,738 1,908,748 4.072 0.004 Residuals 44 20,623,980 468,727 Figure 4.25 shows the data distribution for effluent TP. Data available for TP content was the least compared to the rest of the parameters. TP effluent content ranges from 200 to 2,100 mg/L. There was a small data point in 2015 thus it did not show any charts. TP content was typically consistent in 2017-2019 with data median around 360 mg/L. TP content in manure was not changed during anaerobic digestion process. This process converts organic phosphorus (Org- P) to phosphate (PO4+) phosphorus. Most total P losses in digesters were related to solids accumulation in the reactor (Field et al., 1984). 78 Figure 4.25 Data distribution for effluent TP Table 4.23 shows the ANOVA Tukey multiple comparison performed for effluent TP data distribution. It shows that there was a significant difference in effluent TP at the p<0.05 level from year to year [F (6, 44) = 9.616, p = 0.000). Table 4.23 ANOVA Tukey Multiple Comparison results for effluent TP distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 4,015,385 669,231 9.616 0.000 Residuals 44 3,062,221 69,596 Data distribution for effluent pH is shown in Figure 4.26. In general, SCAD effluent pH ranges from 7.4 to 8.1. Effluent pH in 2017 and 2018 are the most consistent in the range of 7.7 to 8, while pH in 2019 and 2020 are consistent in the range of 7.4 to 7.9. These values are in an acceptable range for pH which is between 6.4 and 8.2 for a healthy digester (MSU Anaerobic Digestion and Research Center, 2019). 79 Figure 4.26 Data distribution for effluent pH Table 4.24 shows the ANOVA Tukey multiple comparison performed for effluent pH data distribution. It shows that there was a significant difference in effluent pH at the p<0.05 level from year to year [F (6, 64) = 2.58, p = 0.0266). Table 4.24 ANOVA Tukey Multiple Comparison results for effluent pH distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 0.333 0.05550 2.58 0.0266 Residuals 64 1.377 0.02151 Data distribution for effluent VFA is shown in Figure 4.27. VFA acts in the biopolymers of biofuels production such as methane and hydrogen (Lukitawesa et al., 2020; Magdalena et al., 2019). Several factors are crucial for VFA productivity, which are hydraulic retention time (HRT), organic loading rate (OLR), temperature, pH, and pretreatment (Wainaina et al., 2019). Data distribution of VFA in 2018-2020 was relatively in a consistent range, while VFA content in 2016 has the widest range and highest peak, ranging from 2,300 to 7,400 mg/L. Some outliers happened in 2014 which the values were around 3,200 and 2,900 mg/L, while majority of the data were around 1,000 mg/L. 80 Figure 4.27 Data distribution for effluent VFA Table 4.25 shows the ANOVA Tukey multiple comparison performed for effluent VFA data distribution. It shows that there was a significant difference in effluent VFA at the p<0.05 level from year to year [F (6, 63) = 7.911, p = 0.000). Table 4.25 ANOVA Tukey Multiple Comparison results for effluent VFA distribution Df Sum Sq Mean Sq F Value Pr (>F) Year 6 66,702,800 11,117,133 7.911 0.000 Residuals 63 88,531,222 1,405,257 The acceptable range for VFA is 50-300 mg/L (Schnaars, 2012). Based on the average result of SCAD effluent measurement, VFA content in SCAD effluent ranges from 817 to 3,785 mg/L. These values are higher than the reference. However, VFA/TA ratio is also important to know whether the digester performed at its best ambience. Higher alkalinity values show that the digester has a better capacity to resist pH changes which is associated to the buffering capacity to maintain the digester stability. Alkalinity values in an anaerobic digester range between 1,500 and 5,000 mg/L (Krakat et al., 2017; Schnaars, 2012). Table 4.26 provides effluent results for VFA and Alkalinity of SCAD during its operation from 2014 to 2020. Alkalinity ranges from 7.583 to 12,779 mg/L, which was higher than the 81 reference value. After calculating the ratio of VFA and Alkalinity, SCAD had VFA/TA ratio between 0.09 and 0.30. The ideal ratio is between 0.3 and 0.4 in the digester and would be 0.2 to 0.3 in the post-digester (Lossie & Pütz, 2011). However, Schnaars (2012) puts a broader range for VFA/TA ratio between 0.1 and 0.35 to maintain a digester in a well-operated condition. Furthermore, it is always better to maintain the ratio to be below 0.35 for a proper digester operation. Therefore, SCAD performance, in general, is well-operated in terms of VFA/TA ratio. The ratio in 2016 was the only year where SCAD performed in its optimum condition based on both references, while the ratio in 2018 was a bit below standard. Table 4.26 SCAD effluent VFA and Alkalinity VFA Alkalinity VFA/TAC Year (mg/L) (mg/L) Ratio 2014 1,301 11,007 0.12 2015 1,580 10,477 0.15 2016 3,785 12,779 0.30 2017 1,834 10,413 0.18 2018 817 9,524 0.09 2019 1,187 8,880 0.13 2020 1,378 7,583 0.18 Data distribution for the digester temperature is available in Figure 4.28. Temperature is an important operational parameter as it provides a suitable ambience for microbial growth. Microbial colonies inside the digester are sensitive to temperature changes. Earlier studies found that digesters should not experience any temperature changes more than 2oC within 24 hours, otherwise the microbial colonies will be highly disturbed (Meegoda et al., 2018; Schnaars, 2012). Based on the temperature setting, an anaerobic digester can be divided into three temperature conditions: psychrophilic (15–23°C), mesophilic (35–41°C), and thermophilic (52–58°C). Based on data from the digester temperature, SCAD is operating in mesophilic temperature. Diverse studies presented in academics revealed that temperature in the mesophilic 82 range is considered as the ideal temperature to promote bacterial activity within the digester (Arosemena, 2021). As shown in the Figure 4.28, the digester temperature ranges from 32.5 to 42.5 oC. The highest temperature recorded was in 2015, while the lowest temperature was in 2020. Moreover, the temperature range in 2015 and 2020 was the most significant, while temperature range in 2019 was the most stable. Figure 4.28 Data distribution for digester temperature Table 4.27 shows the ANOVA Tukey multiple comparison performed for digester temperature data distribution. It shows that there was a significant difference in digester temperature at the p<0.05 level from year to year [F (6, 77) = 3.277, p = 0.006). Table 4.27 ANOVA Tukey Multiple Comparison results for digester temperature Df Sum Sq Mean Sq F Value Pr (>F) Year 6 55.9 9.317 3.277 0.006 Residuals 77 218.9 2.843 SCAD experienced a decrease in the digester temperature during winter months each year. Figure 4.29 shows the decrease in digester temperature occurring typically during the November to February timeframe. The digester should be maintained at temperature 35–41°C for mesophilic condition to optimize microbial performance during anaerobic digestion process. However, there 83 were some moments when SCAD temperature went down to be lower than 35 °C, which were in winter 2015, 2019, and 2020. It shows that the digester struggled to maintain the digester temperature in the last two years of this study. There are two lines added in the figure which represent the highest and lowest temperature in East Lansing to check the gap between air and digester temperatures. It shows that despite the digester experiencing an inconsistent operational temperature lately, it was still above the highest air temperature all the time. Moreover, it emphasizes that more power was needed to heat up the digester during winter since the temperature significantly dropped after September each year. Therefore, heating system is critical for SCAD operational to maintain the mesophilic condition for anaerobic digestion process. Figure 4.29 Digester temperature average in 2014-2020 4.4 SCAD Laboratory Analysis This section compares SCAD laboratory analysis for influent and effluent of the digester. Influent samples were obtained from the mix tank before the mixture was pumped inside the digester. Effluent laboratory analysis was done more consistently than influent. Influent analysis was done only until 2018. The main reasons were due to the time and workforce constraints. Therefore, the digester decided to conduct laboratory analysis for filtrate, effluent, and solids only which were done once a month. Another reason to conclude the mix tank sampling was because 84 the digester received feedstocks that had not been changed tremendously from the previous years. Furthermore, other SCAD laboratory results will be provided in Appendix E. 4.4.1 Solids Table 4.28 shows the comparison between total solids of influent and effluent. TS reduction ranged from 35% to 48%, with the highest reduction being in 2018. In average, SCAD TS for influent and effluent were 105,405 mg/L and 60,744 mg/L, respectively. Moreover, TS destroyed in average was 45,132 mg/L or 43%. Table 4.28 Comparison of total solids between influent and effluent a TS TS Reduction Destroyed Year (mg/L) (mg/L) % Influent n Effluent n 2014 93,359±12,280 12 49,085±5,709 11 44,274 47 2015 107,615±4,216 4 57,082±8,282 10 50,533 47 2016 104,673±20,115 7 65,928±7,110 12 38,745 37 2017 107,881±17,283 3 70,359±5,929 9 37,522 35 2018 113,498 1 58,913±3,944 12 54,585 48 2019 - 59,702±6,768 10 - - 2020 - 64,138±6,668 7 - - Max 113,498 70,359 54,585 48 Min 93,359 49,085 37,522 35 Mean 105,186 60,398 44,654 42 Average 105,405 60,744 45,132 43 St. Dev 6,666 6,377 6,603 6 a average ± standard deviation Table 4.29 shows the comparison between volatile solids of influent and effluent. VS reduction ranged from 40% to 54%, with the highest reduction being in 2018. In average, SCAD VS for influent and effluent were 92,803 mg/L and 48,209 mg/L, respectively. Moreover, VS destroyed on average was 44,754 mg/L or 48%. 85 Table 4.29 Comparison of volatile solids between influent and effluent a VS VS Reduction Destroyed Year (mg/L) (mg/L) % Influent n Effluent n 2014 80,623±10,397 12 38,018±4,767 11 42,605 53 2015 93,101±3,727 4 43,803±7,684 10 49,298 53 2016 90,718±17,274 7 52,485±6,158 12 37,873 42 2017 96,555±16,237 3 58,056±4,682 9 38,499 40 2018 103,020 1 47,524±3,339 12 55,496 54 2019 - 47,354±6,291 10 - - 2020 - 51,626±5,843 7 - - Max 103,020 58,056 55,496 54 Min 80,623 38,018 37,873 40 Mean 92,504 48,029 44,266 48 Average 92,803 48,209 44,754 48 St. Dev 7,366 5,989 6,740 6 a average ± standard deviation 4.4.2 pH and Electronic Conductivity Table 4.30 shows the comparison between influent and effluent. Influent had a more acidic pH with a range from 5.18 to 6.44. Anaerobic digestion process increased the mixture pH thus the effluent had pH between 7.70 and 7.90. Table 4.31 shows electron conductivity comparison between influent and effluent. Influent has a range of EC from 13.57 to 19.03, while effluent has a range of EC from 14.00 to 20.40. 86 Table 4.30 Comparison of pH between influent and effluent a pH pH Year Change Influent n Effluent n 2014 6.44±0.32 12 7.90±0.1 11 1.46 2015 5.82±0.84 4 7.86±0.14 10 2.04 2016 5.18±0.67 7 7.80±0.2 12 2.62 2017 5.74±0.51 3 7.80±0.1 9 2.06 2018 5.27 1 7.80±0.1 12 2.53 2019 - - 7.70±0.1 10 - 2020 - - 7.70±0.2 7 - Max 6.44 7.90 2.62 Min 5.18 7.70 1.46 Mean 5.67 7.79 2.10 Average 5.69 7.79 2.14 St. Dev 0.45 0.07 0.42 a average ± standard deviation Table 4.31 Comparison of electronic conductivity between influent and effluent a EC Year (mS/cm) Influent n Effluent n 2014 13.86±2.07 12 20.00±0.9 6 2015 13.57±3.63 3 17.00±1.74 10 2016 15.55±2.01 7 20.40±3.1 9 2017 19,03±5.86 3 18.00±3.3 9 2018 9.52 1 17.00±0.8 12 2019 - - 16.80±2.3 10 2020 - - 14.00±3.3 6 Max 19.03 20.40 Min 9.52 14.00 Mean 13.96 17.48 Average 14.31 17.60 St. Dev 3.45 2.00 a average ± standard deviation 4.4.3 Chemicals Table 4.32 shows the comparison of soluble COD (sCOD) results between influent and effluent. The digester did not do frequent tests for (sCOD) of influent since the results were relatively consistent in 2014. Based on data in 2014 and 2018 which the only years (sCOD) was conducted for influent, (sCOD) reduction in 2014 and 2018 were 29,556 mg/L 87 and 29,729 ml/L or 79.57% and 76.82%, respectively. It shows that the (sCOD) reduction was consistent. Table 4.32 Comparison of soluble COD between influent and effluent a sCOD Year (mg/L) Influent n Effluent n 2014 37,144±7,692 8 7,588±763 2 2015 - - - - 2016 - - 15,418±3,812 10 2017 - - 19,075±5,289 4 2018 38,700 1 8,971±5,855 11 2019 - - 6,816±1,525 7 2020 - - - - Max 38,700 19,075 Min 37,144 6,816 Mean 37,914 10,641 Average 37,922 11,574 St. Dev 778 4,824 a average ± standard deviation 4.4.4 Fiber Results Table 4.33 shows the laboratory results for SCAD fiber. In average, total solids and volatile solids are 264,388 and 235,204, respectively with VS content in TS ranging from 87% to 90%. Table 4.33 Fiber laboratory results TS VS TS VS VS n Year (mg/L) (mg/L) % % % of TS 2014 238,755 215,338 23.9 21.5 90% 4 2015 264,328 236,833 26.4 23.6 89% 9 2016 249,041 221,187 24.9 22.1 89% 12 2017 277,084 248,754 27.7 24.9 90% 9 2018 268,160 235,912 27.1 23.6 87% 12 2019 268,436 238,752 26.9 23.9 89% 10 2020 284,911 249,653 28.5 25.0 88% 7 Max 284,911 249,653 28.5 25.0 90% Min 238,755 215,338 23.9 21.5 87% Mean 263,973 234,896 26.4 23.5 89% Average 264,388 235,204 26.5 23.5 89% St. Dev 14,666 11,948 1.5 1.2 1% 88 4.4.5 Other laboratory results Several laboratory analyses for different parameters were done outside of ADREC. The tests were conducted by A&L Great Lakes Laboratory in Fort Wayne, Indiana. The tests included manure nutrition analysis which were moisture, solids, Total Kjeldahl Nitrogen (TKN), phosphorus, potassium, sulfur, calcium, magnesium, sodium, iron, aluminum, manganese, copper, and zinc. Analysis of filtrate was done in 2017 and 2018; effluent was done in 2018; and solids were analyzed in all year except in 2017. The complete results of A&L Great Lakes Laboratory are provided in Appendix F. 4.5 Discussions 4.5.1 Correlation between feedstock and biogas production As mentioned in the beginning, SCAD was fed with 18 different feedstocks from 2014 to 2020. Figure 4.30 shows a correlation between food waste amount and biogas production. There were different trends occurring years. Using 2014 as the starting point, 2015 and 2020 showed similar trends where the increase in food waste from the previous year is followed by an increase in biogas production. This trend also happened in 2016 but in the opposite direction. A different trend occurred in 2017 and 2018, where the decrease in food waste affected an increase in biogas production. Meanwhile, in 2019, the trend was reversed: an increase in food waste affected by a decrease in biogas production. The correlation between food waste and biogas production is not clear yet. Therefore, further analysis was done. 89 1,600,000 16,000 1,400,000 14,000 Biogas Production (SCM) Food Waste (metric ton/year) 1,200,000 12,000 1,000,000 10,000 800,000 8,000 600,000 6,000 400,000 4,000 200,000 2,000 0 0 2014 2015 2016 2017 2018 2019 2020 Year Biogas Food Figure 4.30 Correlation between food waste and biogas production Multi linear regression (MLR) was done to determine specific feedstock that has significant impact on biogas production. There are two steps in determining the most significant parameter. First is Stepwise Procedure which calculates the top list of feedstocks that significantly affects output among all feedstocks available. Second is the Final Model which calculates the most significant factor among the list of feedstocks from Stepwise Procedure. Among all feedstock, five feedstocks were included as the result of Stepwise Procedure, which are Filtrate in manure pit, Dairy Gutter, Parlor, ANS Other, and Pineapples (PA). The formula generated was Equation 2, while coefficients for each feedstock are provided in Table 4.34. Biogas production = Filtrate_manure_pit + Dairy_G + Parlor + ANS_Other + P_A (Equation 2) 90 Table 4.34 Coefficient of each parameter Parameter Coefficient Intercept 2330.91251 Filtrate_manure_pit -0.01688 Dairy_G 0.18163 Parlor -0.04628 ANS_Other 0.08850 P_A -0.02158 The final model was run to determine the key feedstocks in Equation 2 that influence gas production. As shown in Table 4.35, Filtrate in the manure pit, Dairy Gutter, and Parlor are feedstocks that have the most significant impact on biogas production, with a p-value<0.05. However, this statistical analysis result does not mean that the other feedstocks are not important, for example, FOG and other manures. Those are the baseline in building biogas production due to their vast amount in feedstock composition. However, Filtrate in the manure pit, Dairy Gutter, and Parlor provide a critical role in SCAD feedstock matrices in terms of biogas production variation during the operation years. Therefore, food waste and manure are important baselines for biogas production, and it will significantly change once those three feedstocks are added to the feedstock composition. This MLR result shows Multiple R-squared as 47.15%, and Adjusted R-squared as 43.75%. It means that the model was able to explain 43% to 47% of the data variance to predict the biogas production. 91 Table 4.35 MLR results for feedstock impact on the biogas production Estimate Std. Error t value Pr(>|t|) (Intercept) 2.331e+03 7.506e+02 3.105 0.002648** Filtrate_manure_pit -1.688e-02 7.464e-03 -2.261 0.026549* Dairy_G 1.816e-01 4.507e-02 4.030 0.000129*** Parlor -4.627e-02 1.748e-02 -2.648 0.009804** ANS_Other 8.850e-02 5.779e-02 1.531 0.129752 P_A -2.158e-02 1.501e-02 -1.438 0.154510 Residual standard error 611.3 on 78 degrees of freedom Multiple R-squared 0.4714 Adjusted R-squared 0.4375 F-statistic 13.91 on 5 and 78 DF p-value 1.03e-09 Signif. Codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Table 4.36 presents the result of ANOVA Type II test to support the result of MLR test. Filtrate in manure pit, Dairy Gutter, and Parlor gave a significant impact on biogas production at the level of p<0.05. Table 4.36 ANOVA Type II Results for feedstock correlation with the biogas production Sum Sq Df F value Pr(>F) Filtrate_manure_pit 1,910,236 1 5.1119 0.0265489* Dairy_G 6,069,859 1 16.2433 0.0001286*** Parlor 2,619,518 1 7.0100 0.0098044** ANS_Other 876,194 1 2.3488 0.1297516 P_A 772,430 1 2.0671 0.1545103 Residuals 29,147,290 78 After the first MLR analysis, the data set was treated by using modelDFFITS to take out the data with high leverage or outliers to check whether the R-squared values can be improved. The result shows that Filtrate in manure pit, Dairy Gutter, Parlor, and FOG become four feedstocks that give the most significant impact in biogas production with a p-value<0.05, as shown in Table 4.37. Furthermore, R-squared values improved to be 58.24% and 56.01% for Multiple R-squared and Adjusted R-squared values, respectively. It means that the model was able to explain 56% to 58% of the data variance to predict the biogas production. 92 This percentage is understandable since SCAD received a wide variety of feedstock with uncertain frequency. Several feedstocks did not come every month; some of them were even only available in certain years. This uncertainty affects the R-squared values. Moreover, feedstock supplies are beyond the digester’s control. It depends on whether the supplier still produces the same food waste, for example. However, the study can still draw a conclusion since there are independent variables that are statistically significant to correlate the relationships between the variables. Table 4.37 MLR results for feedstock impact on the biogas production (after data treatment) Estimate Std. Error t value Pr(>|t|) (Intercept) 1.804e+03 6.544e+02 2.756 0.007333 ** Filtrate_manure_pit -2.207e-02 5.581e-03 -3.953 0.000173 *** Dairy_G 1.915e-01 3.994e-02 4.795 8.05e-06 *** Parlor -4.410e-02 1.520e-02 -2.901 0.004880 ** FOG 1.327e-02 6.526e-03 2.034 0.045470 * Residual standard error 509.4 on 75 degrees of freedom Multiple R-squared 0.5824 Adjusted R-squared 0.5601 F-statistic 26.15 on 4 and 75 DF p-value 1.372e-13 Signif. Codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 4.5.2 The impact of OLR, HRT, and digester pH and temperature on biogas production and methane percentage OLR, HRT, and digester pH and temperatures are key parameters in determining optimum biogas production and methane percentage. Regarding OLR and HRT, SCAD data is organized based on daily numbers, an average of 7 days and an average of 30 days. The formula for determining OLR and HRT is provided in Appendix G. This section only includes data from 2018- to 2020, as the digester performance in those years was considered settled. Table 4.38 shows data related to OLR in 2018-2020. It shows that data on 30 days average have the smallest standard deviation, which means that the data is the most stable, though the 93 difference between daily, 7 days average, and 30 days average are not significantly different. In general, the average OLR increased from 2018 to 2020. OLR average values in 2018 are 3.67, 3.65, and 3.59 g VS/L-d for daily, 7 days average, and 30 days average, respectively. In 2019, the values were 3.75, 3.76, and 3.80 g VS/L-d for daily, 7 days average, and 30 days average, respectively. The values in 2020 are 4.16, 4.14, and 4.16 g VS/L-d for daily, 7 days average, and 30 days average, respectively. Moreover, the table also shows that the standard deviation of OLR in 30 days average in 2018-2020 is very stable, which are 0.50 for 2018 and 2019, and 0.49 for 2020. Therefore, OLR in 30 days average is chosen to represent SCAD OLR. Figure 4.31 shows the OLR in 30 days on average in 2018-2020. Table 4.38 Comparison of OLR in 2018-2020: daily, 7 days average, and 30 average OLR g VS/L-d Month 2018 2019 2020 30 7 30 7 30 7 Days Daily Days Daily Days Days Daily Days Days AVG AVG AVG AVG AVG AVG January 3.37 3.38 3.28 4.41 4.70 4.89 4.36 4.26 4.01 February 3.40 3.40 3.36 2.69 2.67 3.40 4.56 4.67 4.58 March 3.67 3.69 3.54 4.10 3.88 3.30 3.94 3.89 4.07 April 3.41 3.49 3.62 4.11 4.08 4.06 3.84 3.82 3.98 May 2.85 2.84 3.13 4.33 4.30 4.26 4.09 4.22 4.13 June 3.72 3.61 3.18 3.97 4.03 4.22 3.30 3.23 3.59 July 2.97 3.09 3.49 3.63 3.72 3.69 3.88 3.67 3.34 August 3.23 3.11 2.95 3.44 3.37 3.48 4.40 4.32 4.05 September 3.75 3.76 3.44 4.11 4.05 3.95 5.04 5.06 4.82 October 4.10 3.97 3.99 3.86 3.81 3.96 5.01 5.07 5.14 November 4.53 4.48 4.37 2.86 3.06 3.09 4.12 4.11 4.47 December 5.00 5.00 4.71 3.52 3.40 3.30 3.34 3.43 3.80 Average 3.67 3.65 3.59 3.75 3.76 3.80 4.16 4.14 4.16 St. Dev 0.60 0.58 0.50 0.52 0.53 0.50 0.53 0.56 0.49 94 Figure 4.31 OLR in 30 days average 2018-2020 Table 4.39 shows data related to HRT in 2018-2020. It shows that data in 7 days average and 30 days average are relative stable, while daily data has a notable change from 2018 to 2019. HRT in 30 days average has the smallest standard deviation in 2018 and 2019, while HRT in 7 days average has the smallest standard deviation in 2020. HRT average values in 2018 are 21.7, 26.5, and 26.0 days for daily, 7 days average, and 30 days average, respectively. In 2019, the values are 31.2, 25.6, and 23.8 days for daily, 7 days average, and 30 days average, respectively. The values in 2020 are 27.3, 21.8, and 20.3 days for daily, 7 days average, and 30 days average, respectively. Moreover, the table also shows that the standard deviation of OLR in 30 days average in 2018-2020 is slightly more stable than 7 days average, which is 2.7 for 2018, 3.5 for 2019, and 4.8 for 2020. Therefore, OLR in 30 days average is chosen to represent SCAD OLR. Figure 4.32 shows the HRT in 30 days average in 2018-2020. 95 Table 4.39 Comparison of HRT in 2018-2020: daily, 7 days average, and 30 average HRT Day Month 2018 2019 2020 7 30 7 30 7 30 Daily Days Days Daily Days Days Daily Days Days AVG AVG AVG AVG AVG AVG January 24 24 24 29 21 21 20 20 20 February 28 25 25 71 42 29 19 19 19 March 23 23 23 26 25 30 30 24 22 April 21 26 24 27 26 25 28 24 23 May 20 28 25 18 19 21 20 22 22 June 24 29 28 20 22 19 27 28 25 July 20 28 29 28 25 24 28 21 24 August 25 34 30 35 26 25 18 16 7 September 18 27 31 20 24 20 25 19 19 October 20 28 26 35 21 20 18 16 16 November 21 24 25 40 31 27 27 26 21 December 18 22 22 26 25 25 67 26 26 Average 21.7 26.5 25.9 31.2 25.69 23.8 27.3 21.8 20.3 St. Dev 2.8 3.1 2.7 13.6 5.9 3.5 12.6 3.8 4.8 Figure 4.32 HRT in 30 days average 2018-2020 MLR analysis was done to determine the operational parameter that has significant impact on the biogas production (Table 4.40) and the methane percentage (Table 4.41). The results show 96 that none of the parameters tested have a significant impact on both two dependent variables. R- squared values for the biogas production analysis was 6.05% and -6.08% for Multiple R-squared and Adjusted R-squared, respectively. Meanwhile, R-squared values for the methane percentage analysis was 9.93% and -1.69% for Multiple R-squared and Adjusted R-squared, respectively. It means that the model was barely able to explain any of the data variance to predict the biogas production and the methane percentage. It might show an indication that in the commercial scale, the operational parameters have some impacts on the dependent variables. However, any changes in one parameter will not be significantly impactful since there are many parameters that influence the biogas production and the methane percentage. For example, previous study revealed that any temperature changes more than 2oC within 24 hours in the digester will negatively affect the microbial community in the digester (Meegoda et al., 2018; Schnaars, 2012). It certainly has any impact on the biogas production. Nevertheless, the other operational parameters can still backup the digester performance; therefore, it avoids the digester from a significant disturbance. A recent study conducted by Rossi et al. (2022) can be a good example on predicting the digester performance based on the operational parameters. The study was performed by considering 55 explanatory variables to predict the specific methane production (SMP) through the MLR model. The study eventually narrowed down the parameters to include only 9 out of 55. These parameters were included as the feedstock characteristics (total solids (TS), total volatile solids (TVS), C/N ratio, and lignin content), the operating parameters (volumetric flow of inlet (Qin), HRT, and OLR), and the inhibitory compounds (N-NH+4 and total VFA concentration). Through this process, the study was able to gain R2 as 0.87 by including TVS, OLR, C/N ratio, and lignin as the predictors. Although this study might not be representative enough as the working 97 volume was only around 28 L capacity (SCAD working volume is roughly 74,000 L), this can be a reference for SCAD to better predict the impact of the operational parameters on the biogas production and the methane percentage for the future study. Table 4.40 MLR results for operational parameter impact on the biogas production Estimate Std. Error t value Pr(>|t|) (Intercept) 3085.543 1696.503 1.819 0.0786 . OLR -31.360 150.304 -0.209 0.8361 HRT -23.090 19.758 -1.169 0.2515 Temperature 3.676 41.488 0.089 0.9300 pH 135.762 195.569 0.694 0.4927 Residual standard error 365.4 on 31 degrees of freedom Multiple R-squared 0.06047 Adjusted R-squared -0.06076 F-statistic 0.4988 on 4 and 31 DF p-value 0.7367 Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Table 4.41 MLR results for operational parameter impact on the methane percentage Estimate Std. Error t value Pr(>|t|) (Intercept) 63.23486 8.17748 7.733 1.01e-08 *** OLR 1.05067 0.72449 1.450 0.157 HRT 0.10655 0.09524 1.119 0.272 Temperature 0.07043 0.19998 0.352 0.727 pH -0.91353 0.94268 -0.969 0.340 Residual standard error 1.761 on 31 degrees of freedom Multiple R-squared 0.09931 Adjusted R-squared -0.01691 F-statistic 0.8545 on 4 and 31 DF p-value 0.5019 Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 4.5.3 Digester performance from year to year Table 4.42 provides a summary of SCAD operation in terms of feedstock input and output production. In 2014, SCAD received 14,763 metric tons of manure feedstock and 8,533 metric tons of food feedstock, with a total of feedstock received 23,297 metric tons. This feedstock produced 846,232 SCM of biogas, 1,727,073 kWh of electricity, 11,273 metric tons of effluent, and 3,920 metric tons of wet fiber. In 2015, SCAD received 13,805 metric tons of manure 98 feedstock and 12,800 metric tons of food feedstock, with a total of 26,605 metric tons. This feedstock produced 1,103,695 SCM of biogas, 2,118,966 kWh of electricity, 14,100 metric tons of effluent, and 4,858 metric tons of wet fiber. In 2016, SCAD received 11,059 metric tons of manure feedstock and 11,726 metric tons of food feedstock, with a total of feedstock received was 22,785 metric tons. This feedstock produced 1,418,746 SCM of biogas, 1,470,356 kWh of electricity, 14,007 metric tons of effluent, and 3,583 metric tons of wet fiber. In 2017, SCAD received 11,109 metric tons of manure feedstock and 9,129 metric tons of food feedstock, with a total of feedstock received 20,238 metric tons. This feedstock produced 1,326,335 SCM of biogas, 2,169,693 kWh of electricity, 12,751 metric tons of effluent, and 2,779 metric tons of wet fiber. In 2018, SCAD received 10,859 metric tons of manure feedstock and 8,605 metric tons of food feedstock, with a total of feedstock received was 19,464 metric tons. This feedstock produced 1,348,024 SCM of biogas, 2,680,954 kWh of electricity, 14,857 metric tons of effluent, and 1,884 metric tons of wet fiber. In 2019, SCAD received 11,353 metric tons of manure feedstock and 10,539 metric tons of food feedstock, with the total feedstock receiving 21,893 metric tons. This feedstock produced 1,280,438 SCM of biogas, 2,333,449 kWh of electricity, 15,762 metric tons of effluent, and 2,758 metric tons of wet fiber. In 2020, SCAD received 10,332 metric tons of manure feedstock and 14,531 metric tons of food feedstock, with a total of feedstock received was 24,863 metric tons. This feedstock produced 1,340,180 SCM of biogas, 2,664,665 kWh of electricity, 17,745 metric tons of effluent, and 3,082 tons of wet fiber. The range of input acquired by SCAD from 2014 to 2020 was 10,332 to 14,763 metric tons for manure feedstock, 8,533 to 14,531 metric tons for food feedstock, and 19,464 to 26,605 metric tons for a total of both manure and food feedstock. The range of output produced by SCAD from 2014 to 2020 was 846,232 to 1,418,746 SCM of biogas, 1,470,356 to 2,680,954 kWh of electricity, 99 11,273 to 17,745 metric tons of effluent, and 1,884 to 4,858 metric tons of wet fiber. During seven years of operation, SCAD has processed 83,281 metric tons of manure feedstock and 75,864 metric tons of food feedstock with a total of 159,145 metric tons of organic materials to yield 8,663,649 SCM of biogas, 15,165,156 kWh of electricity, 100,495 metric ton of effluent, and 22,864 metric ton of wet fiber. The range of input acquired by SCAD during 2014 to 2020 was 10,332 to 14,763 metric tons for manure feedstock, 8,533 to 14,531 metric tons for food feedstock, and 19,464 to 26,605 metric tons for total of both manure and food feedstock. The range of output produced by SCAD during 2014 to 2020 was 846,232 to 1,418,746 SCM of biogas, 1,470,356 to 2,680,954 kWh of electricity, 11,273 to 17,745 metric ton of effluent, and 1,884 to 4,858 metric ton of wet fiber. During 7 years of operation, SCAD has processed 83,281 metric ton of manure feedstock and 75,864 metric ton of food feedstock with total of 159,145 metric ton of organic materials to yield 8,663,649 SCM of biogas, 15,165,156 kWh of electricity, 100,495 metric ton of effluent, and 22,864 metric ton of wet fiber. In general, the operation of SCAD went through several trends. During 2014-2015, SCAD built good trends since its establishment in 2013 due to the output production increased from 2014 to 2015. SCAD operation was disrupted during 2016 and 2017 in terms of electricity production due to the CHP engine outage that happened in those years. Moreover, the SCAD operation in 2018-2020 was more consistent and settled. In summary, the operation of SCAD has been one of the waste management solutions at MSU by processing organic waste from farming and human consumption to provide renewable energy and fertilizer. Moreover, SCAD has provided emissions reduction, such as greenhouse gases and excess nutrients. To better understand the environmental 100 impact resulting by SCAD, a life cycle assessment was conducted, and the result will be provided in the next chapter. 101 Table 4.42 Summary of SCAD performance Input Output Total Manure Total Food Total Total Total Effluent Wet Fiber Year Pit Pit Feedstock Biogas Electricity Total Total Metric Ton Metric Ton Metric Ton SCM kWh Metric Ton Metric Ton 2014 14,763 8,533 23,297 846,232 1,727,073 11,273 3,920 2015 13,805 12,800 26,605 1,103,695 2,118,966 14,100 4,858 2016 11,059 11,726 22,785 1,418,746 1,470,356 14,007 3,583 2017 11,109 9,129 20,238 1,326,335 2,169,693 12,751 2,779 2018 10,859 8,605 19,464 1,348,024 2,680,954 14,857 1,884 2019 11,353 10,539 21,893 1,280,438 2,333,449 15,762 2,758 2020 10,332 14,531 24,863 1,340,179 2,664,665 17,745 3,082 Total 83,281 75,864 159,145 8,663,649 15,165,156 100,495 22,864 Average 11,897 10,838 22,735 1,237,664 2,680,954 14,356 3,266 St. Deviation 1,683 2,294 2,499 198,468 1,470,356 2,081 958 102 CHAPTER 5. LIFE CYCLE ASSESSMENT 5.1. Introduction Life Cycle Assessment (LCA) is a methodology that is developed to evaluate the environmental burdens of processes and products during the overall life cycle, which starts from raw materials handling and processing, manufacturing, transportation and distribution, consumption stage, recycling where needed, and final disposal. ISO 14040 defines this approach as a technique to determine the specific components of a product or process that create high environmental burdens and replace it with more sustainable and environmentally practices. LCA has been used by many institutions due to its integrated way of treating the framework, impact assessment, and data quality. Moreover, it enables the users to identify the potential environmental tradeoffs between stages by systematic analysis of the diverse impacts along the entire life cycle (Azapagic et al., 2006; Khasreen et al., 2009; Odey et al., 2021). This assessment will compare the commercial MSU South Campus Anaerobic Digester with the conventional approach of organic waste management system: landfilling the food waste and storing the manure in a long-term storage. The comparison will be based on global warming potential (GWP) and water eutrophication potential (WEP). The inputs and outputs represent the major variables that contribute to environmental impacts for both scenarios. 5.2 Goal, Scope, and Functional Unit for Life Cycle Assessment The LCA study was conducted based on the standard methodology provided by ISO 14040 series, Environmental Management Life-Cycle Assessment. The goal of this assessment was to evaluate the environmental impact of MSU SCAD which is an anaerobic co-digestion system using diverse types of manure and organic waste to produce electricity and agricultural coproducts. The objective was to determine how much the environmental impact of the co-digestion system 103 (Scenario 1) and to compare these impacts with the conventional reference method (Scenario 0) performance, which stored dairy manure in a long-term storage and landfilled food waste. This LCA only focused on the treatment of raw materials (Figure 5.1). Transportation and logistics of the waste to the treatment facility were not included in the assessment. The geographical scope occurred in the lower peninsula of Michigan. The temporal scope covers waste management for four years of SCAD operations between 2017-2020. The functional unit (FU) of this assessment was 10,913 metric tons of manure wastes and 10,701 metric tons of food wastes per year. Figure 5.1 System boundaries of Scenario 1 and 0 104 5.3 Life Cycle Data Inventory The life cycle inventory (LCI) provides all important information for life cycle impact assessment. In this study, LCI was divided into raw material and handling, anaerobic digestion and energy production, lagoon storage for manure, and food waste landfill. The raw material and handling section holds information regarding the material input into the system and processes before entering the digester. Additionally, this section also includes the emissions associated with manure storage if a digester was not in place. The anaerobic digestion and energy production section provides data regarding the process of anaerobic digestion and energy production from livestock manure and food waste. The third section provides data related to manure long-term storage in lagoon storage. The last section provides data related to the food waste landfill process. Data quality was evaluated using the Weidema method. It includes six indicators of evaluation: acquisition method, independence of data supplier, representativeness, data age, geographical correlation, and technological correlation. The score ranges from one to five, where one is the best quality and five is the most uncertain. Table 5.1 presents how to apply the indicators based on the pedigree matrix. Table 5.2 provides information about the inventory for this LCA study. This inventory is divided into raw materials, anaerobic digestion, and energy production, animal wastes lagoon storage and land application, and food wastes landfill inventory with landfill gas (LFG) combustion. 105 Table 5.1 Data Quality Evaluation Using the Weidema Method (Weidema et al., 2004) The first part of the inventory holds information and values related to manure waste, food waste, and its chemical compositions. Manure waste is considered as low energy material, while food waste is high energy material. Therefore, co-digestion of manure with food waste potentially increases biogas production in the anaerobic digestion process (Chen et al., 2015). Data for organic waste quantity were the average of feedstock supplies in 2017-2020, which is considered the stable period of SCAD operation. Based on the data average, SCAD was fed with 10,913 metric tons of manure waste and 10,701 metric tons of food waste per year. One thing to put into consideration is that CO2 from manure wastes and food wastes is not counted in the calculation of greenhouse gas emissions because the CO2 is considered of biogenic origin and therefore is assumed to be offset by CO2 capture by regrowth of the plants. The second part of the inventory provides information and values related to anaerobic digestion and energy production from this process. Based on SCAD operational data, biogas 106 composition consists of 65% (v/v) of CH4 and 35% (v/v) of CO2. The remaining consist of a small percentage of H2S, NH3, and H2. Anaerobic digestion process in SCAD produced 1,323,757 m3/year of biogas, which further converted biogas into 2,462,190 kWh-e/year of electricity and 5,584,551 kWh-e/year of heat. The remaining material was effluent as much as 19,948 metric tons/year. This part also provides information associated with the chemical contents of the effluent, including TN, TP, and soluble COD. These values will contribute to water eutrophication potential calculation. Meanwhile, biogas combustion and land application of effluent will contribute to global warming potential calculation. The third part of Table 5.2 provides information and values related to animal waste lagoon storage and land applications. They would include CH4 and N2O emissions if animal wastes were only stored in lagoon storage. Furthermore, this section also includes values related to water eutrophication potential from TN, TP, and COD of animal waste land application. The fourth section provides information and values related to food waste landfills with landfill gas combustion (LFG). Like the third section, this section includes CH4 and N2O emissions if food wastes were only landfilled, also values related to water eutrophication potential from TN, TP, and COD of food wastes landfill. The moisture content of the typical food waste is about 70% (EPA, 2018). 107 Table 5.2 Inventory for the life cycle assessment a Value Unit Source DQI Raw materials Manure wastes 10,913 Metric ton/year Operational data 1, 1, 2, 2, 1, 1 Total solids of 11.4 % Operational data 1, 1, 2, 2, 1, 1 manure wastes Volatile solids of 10.0 % Operational data 1, 1, 2, 2, 1, 1 manure wastes TN of manure 4,143 mg/kg Operational data 1, 1, 2, 2, 1, 1 wastes TP of manure 413 mg/kg Operational data 1, 1, 2, 2, 1, 1 wastes sCOD of manure 59,446 mg/kg Operational data 1, 1, 2, 2, 1, 1 wastes Food wastes 10,701 Metric ton/year Operational data 1, 1, 2, 2, 1, 1 Total solids of 10.1 % Operational data 1, 1, 2, 2, 1, 1 food wastes Volatile solids of 9.3 % Operational data 1, 1, 2, 2, 1, 1 food wastes TN of food 5,318 mg/kg Operational data 1, 1, 2, 2, 1, 1 wastes TP of food 449 mg/kg Operational data 1, 1, 2, 2, 1, 1 wastes sCOD of food 17,525 mg/kg Operational data 1, 1, 2, 2, 1, 1 wastes Anaerobic digestion and energy production inventory Biogas 1,323,757 m3/year Operational data 1, 1, 2, 2, 1, 1 production CH4 content in 65 % v/v Operational data 1, 1, 2, 2, 1, 1 biogas CO2 content in 34 % v/v Operational data 1, 1, 2, 2, 1, 1 biogas Electricity production from 2,462,190 kWh-e/year Operational data 1, 1, 2, 2, 1, 1 biogas Heat production 5,584,551 kWh-e/year Operational data 1, 1, 2, 2, 1, 1 from biogas Effluent 19,948 Metric ton/year Operational data 1, 1, 2, 2, 1, 1 TS of effluent 6.3 % (w/w) Operational data 1, 1, 2, 2, 1, 1 TN of effluent 3,246 mg/kg Operational data 1, 1, 2, 2, 1, 1 TP of effluent 584 mg/kg Operational data 1, 1, 2, 2, 1, 1 sCOD of effluent 7,894 mg/kg Operational data 1, 1, 2, 2, 1, 1 108 Table 5.2 (cont’d) (RTI N2O emission g N2O/g TN in 0.005 International, 2, 1, 1, 4, 1, 2 from the effluent the effluent 2010) (RTI GWP of N2O 298 g CO2-e/g N2O International, 2, 1, 1, 4, 1, 2 2010) Metric ton CO2- (Turnbull & CH4 emission 3.08×10-4 e/metric ton TS Komthunzi, 2, 1, 1, 4, 2, 2 from effluent in the effluent 2004) Water (RTI eutrophication g N-e/kg TN in 0.9864 International, 2, 1, 1, 4, 1, 2 potential (WEP) the effluent 2010) of TN Water (RTI eutrophication g N-e/kg TP in 7.29 International, 2, 1, 1, 4, 1, 2 potential (WEP) the effluent 2010) of TP Water (RTI eutrophication G N-e/kg COD 0.05 International, 2, 1, 1, 4, 1, 2 potential (WEP) in the effluent 2010) of COD Animal wastes lagoon storage and land application inventory Metric ton (Owen & Silver, CH4 emission 0.127 CH4/metric ton 2, 1, 1, 3, 2, 2 2015) VS (RTI g N2O/g TN in N2O emission 0.005 International, 2, 1, 1, 4, 1, 2 the waste 2010) Water (RTI eutrophication g N-e/kg TN in 0.9864 International, 2, 1, 1, 4, 1, 2 potential (WEP) the waste 2010) of TN Water (RTI eutrophication g N-e/kg TP in 7.29 International, 2, 1, 1, 4, 1, 2 potential (WEP) the waste 2010) of TP Water (RTI eutrophication g N-e/kg COD 0.05 International, 2, 1, 1, 4, 1, 2 potential (WEP) in the waste 2010) of COD 109 Table 5.2 (cont’d) Food wastes landfill inventory with landfill gas (LFG) combustion CH4 emission, Metric ton CO2- (Environmental food wastes 2.3 e/ton TS food Protection 2, 1, 1, 1, 2, 2 landfill waste Agency, 2020)b (RTI g N2O/g TN in N2O emission 0.005 International, 2, 1, 1, 4, 1, 2 the waste 2010) Water (RTI eutrophication g N-e/kg TN in 0.9864 International, 2, 1, 1, 4, 1, 2 potential (WEP) the waste 2010) of TN Water (RTI eutrophication g N-e/kg TP in 7.29 International, 2, 1, 1, 4, 1, 2 potential (WEP) the waste 2010) of TP Water (RTI eutrophication G N-e/kg COD 0.05 International, 2, 1, 1, 4, 1, 2 potential (WEP) in the waste 2010) of COD a. CO2 from manure wastes and food wastes is not counted in the calculation of greenhouse gas emissions because the CO2 is considered of biogenic origin and therefore is assumed to be offset by CO2 capture by regrowth of the plants. b. The moisture content of the typical food wastes in the reference is set at 70%. 5.4 Data Quality Evaluation This LCA study was supplied by legitimate sources, such as daily operational data from SCAD, research publications, and government annual reports. Table 5.3 provides the data quality evaluation for the life cycle inventory. Acquisition method, independence of data supplier, geographical correlation, and technological correlation were scored 1. Additionally, representativeness and data age were scored 2 and 3, respectively. Majority of data were from SCAD operational data which are primary data for the study. Furthermore, the study acquired data from verified institutions including EPA and RTI International. For data age, there was one source from RTI International which is less than 20 years 110 old. This data was used because the recent study related to this topic still refers to the data set in this publication. All data included in the study is geographically in the US. Table 5.3 Data Quality Evaluation Summary for LCI DQI Indicator Discussion Score Acquisition Method 1 Measured data Verified data, information from public or other Independence of Data Supplier 1 independent source Representative data from smaller number of Representativeness 2 sites but for adequate periods Data Age 3 Less than 10 years Geographical Correlation 1 Data from area under study Data from enterprises, processes, and materials Technological Correlation 1 under study 5.5 Mass and Energy Balance of the process at different months and years Mass and energy balance analysis was carried out based on the operational data from 2017 to 2020. The energy inputs for the digestion operation include heat (Wheat, kWh-e/year) to maintain the digestion temperature as well as electricity (Welectricity, kWh-e/year) to power pumps, mixers, and other accessary equipment. The energy inputs were calculated using the following equations modified from a previous study (Bustamante and Liao, 2017): 𝑊ℎ𝑒𝑎𝑡 = 𝑚 × 𝐶𝑝 × (𝑇𝑅 − 𝑇𝑂 ) × (1 + 20%) × 0.0002778 (3) 𝑊𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦 = 𝑚 × 0.00788 (4) Where m is the amount of the wet weight of the feedstock per year (kg); Cp is the heat capacity of the wet FM (4.12 kJ/(kg K)); TR is the reactor temperature (313 K); TO is the temperature of the wet feedstock based on the average environmental temperature in East Lansing, MI (288 K); 20% is the percentage of the additional heat needed to maintain the digestion temperature; 0.0002778 is the conversion factor of KJ to kWh; and 0.00788 is the average electricity demand unit of the digester operation (kWh/kg wet FM). 111 Figure 5.2 provides mass balance scheme that happened in SCAD. In general, the input comes from the feedstock then be processed in the digester to produce biogas and effluent. Furthermore, the effluent is divided into wet solid digestate and filtrate. These numbers represent the average values of each parameter in 2017-2020. From feedstock input, the amount of feedstock processed in the digester was 21,614 metric tons per year, which combined both manure pit and food pit. The amount of feedstock from manure pit and food pit were 10,913 metric ton and 10,701 metric ton per year, respectively. Furthermore, these feedstocks contain 10.6% (w/w) of total solids and 10.0% (w/w) of volatile solids. In terms of the nutrient contents, the amount of total nitrogen, total phosphorus, and soluble COD in the feedstock were 4,690 mg/kg, 633 mg/kg, and 38,700 mg/kg, respectively. These numbers were gained from a single measurement that happened in 2017, as the only measurement for feedstock during this period. Anaerobic digestion process transformed the raw materials to produce biogas and effluent. Biogas production was 1,323,757 m3/year with CH4, CO2, and H2S contents were 65% (v/v), 34% (v/v), and 495 ppm, respectively. Effluent yielded from anaerobic digestion process was 19,948 metric ton/year. The effluent contained 6.3% of total solids (w/w) and 5.1% of volatile solids (w/w). In terms of the nutrient contents, the amount of total nitrogen, total phosphorus, and soluble COD in the effluent were 3,246 mg/kg, 584 mg/kg, and 7,894 mg/kg, respectively. Through the separation process, effluent became wet solid digestate and filtrate for further application. The amount of filtrate produced was 17,081 metric ton/year with total solids and volatile solids contents were 4.5% (w/w) and 3.0% (w/w), respectively. Filtrate also contained 3,293 mg/kg of total nitrogen and 509 mg/kg of total phosphorus. For wet solid digestate, the yield was 2,867 metric ton/year with total solids and volatile solids contents were 27.6% (w/w) and 112 24.4% (w/w), respectively. Digestate also contained 529 mg/kg of total nitrogen and 171 mg/kg total phosphorus. Figure 5.2 Mass balance of the anaerobic digestion process *: Data are from a single measurement in 2017 Table 5.4 provides information about energy balance of anaerobic digestion process that happened in SCAD. The energy balance includes energy input, energy output, and net energy output coming from heat and electricity. SCAD consumed 742,090 kWh-e/year of heat to maintain the digester operational temperature and 170,320 kWh-e/year of electricity to power pumps, mixers, and other equipment on the site. Anaerobic digestion process yielded heat and electricity as much as 5,584,551 kWh-e/year and 2,462,190 kWh-e/year, respectively. Net energy output was gained from subtracting energy produced with energy consumed. The net energy output yielded by SCAD in the form of heat and electricity were 4,842,461 kWh-e/year and 2,291,870 kWh- e/year, respectively. Furthermore, from energy output values, the efficiency of electricity generation was calculated which was 29%. 113 Table 5.4 Energy balance of the anaerobic digestion process a Anaerobic digester Energy input Heat input (Wheat, kWh-e/year) b -742,090 c -170,320 Electricity input (Welectricity, kWh-e/year) Energy output Energy output as heat (Eheat, kWh-e/year) d 5,584,551 e 2,462,190 Energy output as electricity (Eelectricity, kWh-e/year) Net energy output Net heat output (kWh-e/year) f 4,842,461 g 2,291,870 Net electricity output (kWh-e/year) a. Negative numbers mean energy inputs, and positive numbers mean energy outputs. b. Eq. 1 was used to calculate the heat input. c. Eq. 2 was used to calculate the electricity input. d. The annual biogas production of 1,323,757 m3 with 65% (v/v) of methane was used to calculate the energy content of the biogas. The low heating value of methane is 35.8 MJ/m3 methane. The thermal conversion efficiency of the CHP unit is 65%. e. The electricity output is the metered number of the digestion operation. f. The net heat output = Eheat - Wheat g. The net electricity output = Eelectricity - Welectricity 5.6 Impact Assessment Two impact categories were chosen for Life Cycle Impact Assessment: Global Warming Potential (GWP) and Water Eutrophication Potential (WEP). The classification of each category is defined by the US Environmental Protection Agency (EPA). The LCIA provides an analysis of environmental impacts on both scenarios as a comparison to the emissions of waste management methods for animal wastes and food wastes. This could be done by calculating the environmental impacts yielded from processes or products associated with the proposed systems. 5.6.1 Global Warming Potential (GWP) Global warming is defined as the raising of Earth’s temperature due to GHG emissions globally, which mainly coming from human activities. The main GHGs are CO2, CH4, and N2O. 114 Currently, the main energy supplies are still from fossil fuels such as oil and coal, which contribute roughly 65% of GHG emissions (EPA, 2019). Therefore, renewable energy is expected to address this concern. Anaerobic digestion has been promoted as a renewable energy system that potentially lower the global warming potential by reducing emissions from manure and food waste. Global Warming Potential (GWP) is the amount of GHG released during the life cycle of a process. Carbon dioxide is commonly used as a reference gas to compare the impact of various greenhouse (Shine, 2009). For this LCA study, data were collected for CH4, and N2O emissions. They were normalized to 1 ton of CO2 equivalent (CO2-e) based on the following conversions: 1 kg CH4 = 25 kg CO2-e; and 1 kg N2O = 298 kg CO2-e (RTI International, 2010). CO2 from manure wastes and food wastes is not counted in the calculation of greenhouse gas emissions because the CO2 is considered of biogenic origin and therefore is assumed to be offset by CO2 capture by regrowth of the plants. Figure 5.3 provides information related to GWP contribution analysis for each scenario. According to the impact assessment, the AD system has an overall GWP of 1,842-ton CO2-e/year, while the land applications system produces 6,190-ton CO2-e/year. This result shows that the AD system can reduce GWP up to 70% lower than the landfills system. Furthermore, biogas and AD effluent were the only parameters to be considered for calculating GWP. Electricity, heat utilization, digestate, and filtrate were not calculated as they are outside of the system boundaries. For the AD scenario, biogas combustion contributed 92% of the total GWP, while land application of effluent only contributed 8%. For the landfill scenario, biogas production was counted as CH4 and N2O emissions from manure land application with lagoon storage and food wastes land application with landfill gas combustion. Lagoon storage contributed 115 58% of the total GWP, followed by landfill gas combustion with 42% of contribution towards the total GWP. The result of this study was supported by previous result from Chen et al. (2015), where a co-digestion of dairy manure and bakery wastes in the AD system was compared to the AD system with dairy manure only and landfilling the bakery waste. The study revealed that co-digestion of 7,147-ton dairy manure and 2,382-ton food waste had an overall GWP of 1.6x104 ton CO2-e, while another scenario yielded an overall GWP of 2.7x104 ton CO2-e, which means that AD scenario resulted in a 42% reduction in GWP. This could be evidence that the anaerobic co-digestion system has a capability to mitigate the global climate change rather than the conservative methods such as landfilling the food waste or digestion of manure only. 8,000 Global warming potential (metric 7,000 6,000 Landfill of the 5,000 food wastes: 4,000 2,619 ton CO2-e/year) Land application of 3,000 the AD effluent: 152 Manure land 2,000 application with Biogas 1,000 lagoon: 3,571 combustion: 0 1,690 AD Land applications Figure 5.3 GWP contribution analysis of the anaerobic digestion process and land applications 5.6.2 Water Eutrophication Potential (WEP) Eutrophication is a situation where a water body contains excessive nutrients that affect to a dense growth of plant life and death of water animals due to lack of oxygen. It is due to nutrients 116 runoff from the land, such as nitrogen and phosphorus, which then accumulate in the water. Consequently, it creates a “dead zone,” which is an area with low oxygen content that suffocates marine life (Mueller & Helsel, 1996). Water eutrophication potential (WEP) is the impact resulting from excessive nutrient supplies on terrestrial and aquatic environments, particularly the most important substances such as nitrogen (N) and phosphorus (P). WEP can be presented as either nitrogen equivalents (N-eq.) or phosphate equivalents (PO4-eq.) (Guinee, 2002). A kilogram of nitrogen equivalents (kg N-eq.) units was used to assess WEP in this study. TN, TP, and COD were the three parameters used for WEP assessment. All values related to TN, TP, and COD contents were measured as SCAD operational data, except for COD concentration of manure which was back calculated. WEP conversion values for TN, TP, and COD are 0.9864 g N-e/kg TN in the waste, 7.29 g N-e/kg TP in the waste, and 0.05 g N-e/kg COD in the waste, respectively (RTI International, 2010). Figure 5.4 provides information related to WEP contribution analysis for each scenario. According to the impact assessment, the AD system has an overall WEP of 173 kg N-e/year, while the land applications system produces 232 kg N-e/year. This result shows that the AD system can reduce WEP up to 25% lower than the landfills system. WEP from the AD system came from three parameters, which were TN, TP, and COD of AD effluent. TP became the most significant WEP contributor in the AD system, which was 54% of the total WEP, followed by TN, which contributed 41% of the total WEP. Meanwhile, COD only contributed 9% of the total WEP. In the landfills system, the chemical content was divided into manure waste and food waste. Both wastes have TN as the most significant contributor in WEP. Food waste TN became the highest contributor of WEP in landfills system, which was 27%, followed by Manure TN at 21%. Food waste TP, manure TP, and manure COD were closed 117 together, which were 17%, 16%, and 15% of WEP contribution, respectively. Food waste COD was the least contributor of WEP with 4% of contribution. AD system does not contribute much to reducing the nutrient content in the organic materials. WEP in AD is more correlated to feedstock quality and the digester’s HRT. It barely had any correlation to biogas productivity. Therefore, nutrient content in the feedstock did not significantly alter. Nevertheless, AD transformed the elemental nitrogen and phosphorus to become ammonia and phosphate, respectively. Compared to elemental nitrogen and phosphorus, ammonia and phosphate are more beneficial for soil conversions and plant adsorption (Arosemena, 2021; R. Chen et al., 2015; Field et al., 1984). Figure 5.4 WEP contribution analysis of the anaerobic digestion process and land applications 5.7 Interpretation 5.7.1 Sensitivity Analysis Sensitivity analysis is a method to measure any changes in a certain impact by simulating any changes in key parameters that influence the model. This method can report which parameters 118 greatly affect any changes in each impact category. By doing a sensitivity analysis with LCA, a study can depict further actions to conduct to lower the environmental burdens of a product or a system. 5.7.1.1 Global Warming Potential (GWP) The sensitivity analysis for the anaerobic digestion and the land applications systems were performed by modifying ±25% of each variable of interest while keeping the other variables constant for the base case. The parameters analyzed for the anaerobic digestion were biogas combustion and AD effluent, while the parameters analyzed for the land applications were NO2 and CH4 emissions from manure and food waste. The sensitivity analysis result for the anaerobic digestion system, as presented in Figure 5.5, shows that biogas combustion becomes the most sensitive parameter towards the impact category value, while effluent becomes the least sensitive parameter. The change in biogas combustion (±25%) resulted the highest change in GWP values at 29.44%. It is due to biogas combustion is the main process in converting methane into electricity. Meanwhile, the change in AD effluent only resulted 2.31% change in GWP values. The sensitivity analysis result for the land application system, as presented in Figure 5.6, shows that CH4 emission from manure becomes the most sensitive parameter towards the GWP value, followed by CH4 emission from food waste. The change in CH4 emissions from manure resulted 14% change in GWP value, while the change in CH4 emission from food waste resulted 10% change in GWP value. The emissions of NO2 from food waste and manure become the least sensitive parameters, which only affect 0.54% and 0.43% of change in GWP value, respectively. 119 Figure 5.5 Anaerobic digestion sensitivity analysis for GWP Figure 5.6 Land applications sensitivity analysis for GWP 120 5.7.1.2 Water Eutrophication Potential (WEP) The sensitivity analysis for the anaerobic digestion and the land applications systems were performed by modifying ±25% of each variable of interest while keeping the other variables constant for the base case. The parameters analyzed for the anaerobic digestion were TN, TP, and COD of the effluent. Meanwhile, the parameters analyzed for the land applications were TN, TP, and COD of both manure and food waste. The sensitivity analysis result for the anaerobic digestion system, as presented in Figure 5.7, shows that effluent TP becomes the most sensitive parameter towards the impact category value, while effluent COD becomes the least sensitive parameter. The change in effluent TP (±25%) resulted the highest change in GWP values, at 13.5%, followed by effluent TN at 10.2%. Meanwhile, the change in effluent COD only resulted 1.25% change in GWP values. The sensitivity analysis result for the land application system, as presented in Figure 5.8, does not show significant difference between most of parameters. Food waste TN becomes the most sensitive parameter, followed by manure TN and food waste TP. These three parameters resulted 6.7%, 5.3%, and 4.1% change in WEP value, respectively. On the other hand, food waste COD becomes the least sensitive parameter with only 1.1% change in the WEP value. Manure TP and COD contribute similar impact to WEP value, which are 3.9%. 121 Figure 5.7 Anaerobic digestion sensitivity analysis for WEP Figure 5.8 Land applications sensitivity analysis for WEP 122 5.7.2 Consistency and Completeness Check The consistency check aims to demonstrate that assumptions, methods, and data used throughout the LCA process are aligned with the goal and scope of the study. It can show where the data consistency can be improved to compare systems to one another. The consistency check and explanations of inconsistency are explained within Table 5.5. The overall data adequately shows consistency to support the goal and scope of the study. Table 5.5 Checklist and Inconsistencies based on Data Quality Category Checklist and Inconsistencies Both scenarios have most data based on operational data and Data Source legitimate sources Both scenarios are supplied with accurate data from measurement or Data Accuracy calculation based on previous studies Technological Both scenarios have data available for conducting the study Representation Temporal Both technologies are utilized up to date Representation Geographical Both technologies include data from the United States Representation System Boundary, Both systems serve as a waste management system Assumption and Model A completeness check ensures that the study has complete available sources for data interpretation. In case there are some gaps in the completeness of the data, it should be verified whether the incompleteness will affect the goal and scope of the study. Table 5.6 provides a completeness check for the AD system, while Table 5.7 provides a completeness check for the landfill system. In general, all data required for the study was completed. SCAD operational data contains most parameters needed for the study, even though there were several parameters that rely on a single measurement and one parameter needed a back-calculation. Manure storage and food waste landfill are two common methods for waste management in the US, therefore providing 123 legitimate sources for the impact assessment. There might be potential data gaps if the system boundary is extended since this study only focused on the waste treatment. Table 5.6 Completeness check for AD system Life cycle AD Complete Required Actions stage Raw X Yes - materials AD and energy X Yes - production Output: X Yes - Effluent X: data available n.a.: not applicable Table 5.7 Completeness check for landfills system Life cycle Landfill Complete Required Actions stage Raw X Yes - materials Manure X Yes - storage Food waste X Yes - landfill X: data available n.a.: not applicable 5.8 Technoeconomic Analysis Technoeconomic analysis is conducted to evaluate whether renewable energy production can also attract investors based on its financial benefits of it. One of the significant issues in producing renewable energy on a large scale is the large investment for capital and/or operational costs, which makes renewable energy seems environmentally favorable but less economically competitive (Carneiro and Ferreira, 2012; Fersi et al., 2012). This section will analyze the cost and profit analysis of SCAD as a commercial digester, including capital expenditures (CapEX), operational expenditures (OpEX), and revenues. The payback period was calculated according to capex and total net profit. 124 Table 5.8 provides information for the whole performance of technoeconomic analysis. The first section holds information about CapEX. Digester construction cost became the highest expense in this section, contributing 40.21% to the total CapEX. Feedstock receiving and combined heating and power (CHP) costs were the other major expenses with relatively similar values, contributing 20.29% and 21.71% to total CapEX, respectively. The total system CapEX was $3,586,861. The second section of Table 5.8 provides information about the OpEX, which is calculated per year. In this study, OpEX was taken from operational data in 2019-2020 as the recent years of the study. Among 12 items described in OpEX, labor cost became the highest expense which contributed roughly 41.5% to total OpEX. CHP engine service became the second-highest expense, contributing about 25% to total OpEX. DHT transport means that every gallon of food waste brought to the digester has to be exported to maintain nutrient balance on campus. It became the third-largest expense, contributing about 17.75% to total OpEX. The total system OpEX was $298,156. The third section belongs to revenue. Tipping fees and electricity were the two revenue sources of SCAD. Both commodities had similar values during 2019-2020, which were $237,746 for electricity and $217,854 for tipping fees. The total system revenue was $455,600. SCAD has experienced volatile revenue from electricity due to some changes in electricity pricing. The model was proposed to the board of trustees at the price of $0.123/kWh. Prior to 2018, the price was between $0.04 and $0.06/kWh due to the standby charge implemented by Consumers Energy. Standby charge is a measurement of energy production in a 15-minute interval and during peak time from 09:00 AM to 06:00 PM. If the digester does not produce power during that 15 minute, the digester gets the lower rate, and it is for the entire month, not only for that 15 minute. 125 In 2018, SCAD was directly connected to the campus power supply, which resulted in a flat rate of $0.1017/kWh. Total net revenue was calculated by subtracting total revenue from total OPEX. The net revenue value of SCAD was $157,444. The payback period was then calculated by considering 5- year average local inflation of 3.2% in the U.S. as the inflation rate and 20 years of depreciation period on CapEx. The annual depreciation rates from Modified Accelerated Cost Recovery System (MARCRS) are: 0.100, 0.188, 0.144, 0.115, 0.092, 0.074, 0.066, 0.066, 0.065, 0.065, and 0.033 (after 10 years). Based on cash flow calculation, the payback period will be in 21.5 years; then, the digester will start to gain profit afterward. This can be considered a quite promising payback time and economically competitive. An economic sensitivity analysis would be strongly recommended to be conducted to know which parameter is the most sensitive; thus, it can shorten the length of the payback period. Table 5.8. Economic performance of the digestion process a Capital expenditure (CapEX) Cost Reference Feedstock Receiving $727,927 Data Digester $1,442,140 Data CHP $778,651 Data Bond $38,143 Data Interconnection $300,000 Data Site improvements & excavation $300,000 Data Total CapEX $3,586,861 OpEX (per year) AD Repairs $28,373 Data ADMIN Fee $2,948 Data Bio Analysis $2,827 Data CHPS $74,226 Data Labor $123,616 Data Laundry $378 Data Maintenance and Repair $6,482 Data MISC $4,064 Data Motor Pool / Vehicle $1,165 Data Supplies $396 Data Telephone $772 Data 126 Table 5.8 (cont’d) Transport (DHT) $52,910 Data Total OpEX (per year) $298,156 Revenue (per year) Electricity $237,746 Data Tipping $217,854 Data Total revenue (per year) $455,600 Total net revenue (per year) b $157,444 c Payback time (Years) 21.5 a. The OpEX and revenue are the operational data from 2019-2020. b. The net revenue = Total revenue – Total OpEx c. The 5-year average local inflation of 3.2% in the U.S. is used as the inflation rate. The depreciation period is set at 20 years. The depreciation is just on CapEx. The annual depreciation rates from MARCRS (Modified Accelerated Cost Recovery System) are: 0.100, 0.188, 0.144, 0.115, 0.092, 0.074, 0.066, 0.066, 0.065, 0.065, and 0.033 (after 10 years). Figure 5.9 Total net cash flow and payback period of SCAD There are three basic scenarios that can be considered to shorten the payback period. The first scenario would be fixing the electricity price; the second scenario would include digestate as part of revenue, while the third scenario would be the combination of the first and the second scenario. As mentioned earlier, the proposed electricity price for SCAD to the board of trustees was $0.123/kWh, while the current price for electricity is $0.1017/kWh. That means there is a price gap of as much as $0.0213/kWh. If the proposed price can be achieved, the revenue from electricity 127 would be $307,713. Because of it, the total net revenue will be increased to $227,412, and it will lower the payback period to be in 15 years. The second scenario would be including digestate as part of the revenue. As per the current study, digestate is still part of OpEX, which is DHT transport. Digestate is a source of nutrients for the soil. If the digester can manage the digestate to be part of the revenue, then the DHT transport cost will be excluded from OpEX. Furthermore, the digestate price is assumed to be $7.00 per metric ton. The average digestate production in 2019-2020 was about 18,471 metric tons. Therefore, the revenue from selling the digestate can be up to $129,297. This additional revenue will increase the total revenue to $584,897 and the total net revenue to $339,652. It will also reduce the total OpEX to $245,246. These changes can reduce the payback period to be in 10.5 years. The third scenario would be the best scenario that SCAD can afford. With electricity price fixed and digested becoming part of the revenue, the total revenue will become $654,865, total net revenue will be $409,619, and the payback period will be in 9 years. Regardless of which scenario seems to be the most feasible, it shows that SCAD has a competitive economic advantage as a commercial-scale anaerobic digester. 5.8.1 Sensitivity Analysis Sensitivity analysis was done to check which parameters significantly affect the payback period. Nine variables – feedstock receiving, digester, CHP, interconnections, site improvements, CHP maintenance, labor, electricity, and tipping fees – were taken into consideration for the economic sensitivity analysis of this study. The analysis was done by modifying each variable of interest by ± 25% while keeping the other variables constant for the baseline scenario. The number 128 changed represents how much each variable affects the increasing or decreasing of the payback period compared to the baseline scenario. The result (Figure 5.6) shows that electricity was the most sensitive among all variables, followed by tipping fees and labor costs as the second and third most sensitive variables. Increasing the revenue from electricity by 25% helps to decrease the payback period by 11.41 years, while the decrease in this parameter by 25% contributes to increasing the payback period by 26.06 years. The increasing of tipping fees by 25% reduces the payback period by 10.69 years while decreasing the tipping fees by 25% increases the payback period by 22.67 years. Meanwhile, the increase of labor cost by 25% increases the payback period by 17.61 years, while reducing the cost by 25% lowers the payback period by 9.44 years. Interconnections, site improvements, and feedstock receiving become the least sensitive among all parameters. Figure. 5.10 Sensitivity analysis chart for SCAD payback period 129 CHAPTER 6. OVERALL CONCLUSION AND RECOMMENDATIONS 6.1 SCAD operational performance The goal of this study is to evaluate the operational performance of Michigan State University’s South Campus Anaerobic Digester (SCAD) as a commercial digester, as well as the environmental impact during its operation from 2014-to 2020. This study concluded that SCAD received a total of 18 different feedstocks throughout its operation thus far. Among all feedstocks, Dairy Gutter, Parlor, and FOG become the major feedstocks for the digester. Multilinear regression was conducted to determine feedstocks that have the most significant impact on biogas production. The filtrate in the manure pit, Dairy Gutter, Parlor, and FOG are feedstocks that have the most significant impact on biogas production, with a p-value<0.05. The Multiple R-squared was 58.24%, and the Adjusted R-squared was 56.01%, showing that the model was able to explain 56% to 58% of the data variance to predict the biogas production. It is due to SCAD receiving a wide variety of feedstock with an uncertain frequency which affects the R-squared values. However, the study can still draw a conclusion since there are independent variables that are statistically significant to correlate the relationships between the variables. In general, the operation of SCAD went through several trends. From 2014 to 2015, SCAD built good trends since its establishment in 2013 due to the output production increased from 2014 to 2015. SCAD operation was disrupted from 2016 to 2017 in terms of electricity production due to the CHP engine outage that happened in those years. Moreover, the SCAD operation from 2018 to 2020 was more consistent and settled. In summary, the operation of SCAD has been one of the waste management solutions at MSU by processing organic waste from farming and human consumption to provide renewable energy and fertilizer. During its operation years from 2014 to 2020, SCAD has processed 159,145 130 metric tons of feedstock that consist of 83,281 metric tons of manure wastes and 75,864 metric tons of food waste to produce 8,663,649 SCM of biogas, 15,165,156 kWh of electricity, 100,495 metric tons of effluent, and 22,864 metric tons of wet fiber. 6.2 Life Cycle Assessment and Technoeconomic Analysis A life cycle assessment was conducted to compare the environmental impact of the AD system with a conventional system that combines lagoon storage for manure wastes and landfills for food wastes. This LCA study was supplied by legitimate sources, such as daily operational data from SCAD, research publications, and annual government reports. Acquisition method, independence of data supplier, geographical correlation, and technological correlation were scored 1. Additionally, representativeness and data age were scored 2 and 3, respectively. The result showed that the AD system possesses fewer environmental burdens in both GWP and WEP compared to the conventional system. AD system has an overall GWP of 1,842- ton CO2-e/year, while the land applications system produces 6,190-ton CO2-e/year. This result shows that the AD system can reduce GWP up to 70% lower than the landfills system. AD system has an overall WEP of 157 kg N-e/year, while the land applications system produces 210 kg N- e/year. This result shows that the AD system has a WEP 26% lower than the landfills system. Sensitivity analysis was conducted to determine parameters that give the most significant impact on the LCA study. From GWP, biogas combustion became the most sensitive parameter on the AD system. Meanwhile, CH4 emissions from manure became the most sensitive parameter on the land application system. From WEP, effluent TP became the most sensitive parameter on the AD system. Meanwhile, food waste TN became the most sensitive parameter on the land application system. 131 Technoeconomic analysis was conducted to understand the financial feasibility of SCAD as a commercial digester. The result showed that SCAD needs 21.5 years to accomplish its payback time, which is considered quite economically competitive. Three basic scenarios can be done to gain a better payback time for up to 9 years. Economic sensitivity analysis shows that the revenue from electricity is the most sensitive parameter to affect the payback period, followed by tipping fees and labor costs. Increasing the revenue from electricity can lower the payback period by 11.41 years. Meanwhile, interconnections, site improvements, and feedstock receiving are the least sensitive parameters. 6.3 Further recommendations Future work for LCA study can include more impact parameters for the study, such as air acidification potential and smog potential. Since the current study has a system boundary that only includes organic waste treatment at the site, it would be interesting to know the dynamics in the LCA study once the system boundary is extended, such as including digestate land application or transportation to the treatment site. As for technoeconomic analysis, digestate is not included in the revenue as per the current study. Including digestate as an organic fertilizer can be additional revenue that can shorten the payback time of the digester. Moreover, electricity prices are critical to maintaining SCAD revenue. Getting an ideal price as proposed will be another improvement for SCAD, although it depends on the policy implementation for renewable electricity pricing. Data availability can be another aspect to improve, especially laboratory chemical analysis. The current study revealed that SCAD has fewer laboratory results for influent chemical analysis and several chemical analyses for effluent due to workforce constraints. If this challenge can be 132 addressed and future work can provide more frequent chemical analysis for the digester influent and effluent, it will help to provide more accurate results for mass balance analysis and life cycle impact assessment. 133 APPENDIX 134 A. MLR Codes ## Statistical analysis - Multiple regression analysis ## SCAD operation ## Wei Liao, October 23, 2021 ## Fahmi Nov 23, 2021 ## Fahmi Dec 9, 2021 ## Fahmi Dec 14, 2021 # Load libraries ----- library(dplyr) library(FSA) library(psych) library(car) library(rcompanion) # Choose data file of "BiogasProduction_v3.txt" ----- con <-file.choose(new = FALSE) metadata <- read.table(con, header = T, row.names = 1, fill = TRUE) head(metadata) ############################# # select data data.num = select(metadata, Digestate, Filtrate_manure_pit, SLS_Solids, Dairy_G, Parlor, Beef, W_Feed_manure, Poultry, Swine, ANS_Other, Filtrate_food_pit, SLS_Solids_food, P_A, Pulp,FOG, W_Feed, Other, Cart_Food, Biogas_production, CH4_production, H2S_content, Electricity, Effluent_TS, Effluent_TN, Effluent_TP, Effluent_pH, Effluent_VFA) data.num # corr.test(data.num, use = "pairwise", method="pearson", adjust="none", alpha="0.05") ############################ # Stepwise procedure model.null = lm(Biogas_production~1, data=data.num) model.full = lm(Biogas_production ~ Digestate + Filtrate_manure_pit + SLS_Solids+Dairy_G + Parlor + Beef + W_Feed_manure + Poultry + Swine + ANS_Other+ SLS_Solids_food + P_A + Pulp + FOG + W_Feed + Other + Cart_Food, data=data.num) step(model.null, scope = list(upper=model.full), direction="both", data=data.num) # Define the final model from the results of the stepwise procedure model.final = lm(Biogas_production ~ Filtrate_manure_pit + Dairy_G + Parlor + ANS_Other + P_A, 135 data=data.num) summary(model.final) # Analysis of variance for individual feed components Anova(model.final, Type="II") # Simple plot of predicted values with 1-to-1 line data.num$predy = predict(model.final) plot(predy ~ Biogas_production, data=data.num, pch = 16, xlab="Actual response value", ylab="Predicted response value") abline(0,1, col="blue", lwd=2) # Checking assumptions of the model hist(residuals(model.final), col="darkgray") plot(fitted(model.final),residuals(model.final)) RESULTS Call: lm(formula = Biogas_production ~ Filtrate_manure_pit + Dairy_G + Parlor + ANS_Other + P_A, data = data.num) Coefficients: (Intercept) Filtrate_manure_pit Dairy_G 2330.91251 -0.01688 0.18163 Parlor ANS_Other P_A -0.04628 0.08850 -0.02158 > # Define the final model from the results of the stepwise procedure > model.final = lm(Biogas_production ~ Filtrate_manure_pit + Dairy_G + Parlor + ANS_Other + P_A, + data=data.num) > summary(model.final) Call: lm(formula = Biogas_production ~ Filtrate_manure_pit + Dairy_G + Parlor + ANS_Other + P_A, data = data.num) Residuals: Min 1Q Median 3Q Max -2064.80 -335.13 -26.08 292.74 1659.95 136 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 2.331e+03 7.506e+02 3.105 0.002648 ** Filtrate_manure_pit -1.688e-02 7.464e-03 -2.261 0.026549 * Dairy_G 1.816e-01 4.507e-02 4.030 0.000129 *** Parlor -4.627e-02 1.748e-02 -2.648 0.009804 ** ANS_Other 8.850e-02 5.779e-02 1.531 0.129752 P_A -2.158e-02 1.501e-02 -1.438 0.154510 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 611.3 on 78 degrees of freedom Multiple R-squared: 0.4714, Adjusted R-squared: 0.4375 F-statistic: 13.91 on 5 and 78 DF, p-value: 1.03e-09 > # Analysis of variance for individual feed components > Anova(model.final, Type="II") Anova Table (Type II tests) Response: Biogas_production Sum Sq Df F value Pr(>F) Filtrate_manure_pit 1910236 1 5.1119 0.0265489 * Dairy_G 6069859 1 16.2433 0.0001286 *** Parlor 2619518 1 7.0100 0.0098044 ** ANS_Other 876194 1 2.3448 0.1297516 P_A 772430 1 2.0671 0.1545103 Residuals 29147290 78 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 137 Figure A.1 Final model of MLR for biogas production Figure A.2 Residual model of MLR for biogas production B. Radar Chart Codes ## Feed amount - Radar analysis ## SCAD operation ## Wei Liao, October 23, 2021 ## Fahmi, Noveberm 5, 2021 updated ## Wei Liao, December 9, 2021 updated # Load libraries ----- library(fmsb) 138 # Choose data file of "FeedAmount_Radar.txt" ----- con <-file.choose(new = FALSE) metadata <- read.table(con, header = T, row.names = 1, fill = TRUE) head(metadata) ### Set up the bound for the radargraph data0<-metadata[,c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data0 maxmin <-data.frame(Digestate=c(max(data0), min(data0)), Filtrate_manure_pit=c(max(data0), min(data0)), SLS_Solids=c(max(data0), min(data0)), Dairy_G=c(max(data0), min(data0)), Parlor=c(max(data0), min(data0)), Beef=c(max(data0), min(data0)), W_Feed_manure=c(max(data0), min(data0)), Poultry=c(max(data0), min(data0)), Swine=c(max(data0), min(data0)), ANS_Other=c(max(data0), min(data0)), #Total_manure_pit=c(max(data0), min(data0)), Filtrate_food_pit=c(max(data0), min(data0)), SLS_Solids_food=c(max(data0), min(data0)), P_A=c(max(data0), min(data0)), Pulp=c(max(data0), min(data0)), FOG=c(max(data0), min(data0)), W_Feed=c(max(data0), min(data0)), Other=c(max(data0), min(data0)), Cart_Food=c(max(data0), min(data0)) #Total_food_pit=c(max(data0), min(data0)) ) maxmin ############################# ## 2020 data # select data data1<- metadata[which(metadata$Year=="2020"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data1 data1 <- rbind(maxmin, data1) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) 139 # Provide the names of columns colnames(data1) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data1, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2020", bty = "n", pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ############################# ## 2019 data # select data data2<- metadata[which(metadata$Year=="2019"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data2 data2 <- rbind(maxmin, data2) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) # Provide the names of columns colnames(data2) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data2, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2019", bty = "n", pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ############################# ## 2018 data # select data data3<- metadata[which(metadata$Year=="2018"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data3 140 data3 <- rbind(maxmin, data3) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) # Provide the names of columns colnames(data3) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data3, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2018", bty = "n", pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ############################# ## 2017 data # select data data4<- metadata[which(metadata$Year=="2017"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data4 data4 <- rbind(maxmin, data4) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) # Provide the names of columns colnames(data4) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data4, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2017", bty = "n", 141 pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ############################# ## 2016 data # select data data5<- metadata[which(metadata$Year=="2016"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data5 data5 <- rbind(maxmin, data5) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) # Provide the names of columns colnames(data5) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data5, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2016", bty = "n", pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ############################# ## 2015 data # select data data6<- metadata[which(metadata$Year=="2015"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data6 data6 <- rbind(maxmin, data6) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) # Provide the names of columns colnames(data6) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", 142 "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data6, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2015", bty = "n", pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ############################# ## 2014 data # select data data7<- metadata[which(metadata$Year=="2014"),c(3,4,5,6,7,8,9,10,11,12,14,15,16,17,18,19,20,21)] data7 data7 <- rbind(maxmin, data7) # Create the radar chart # Set up the font windowsFonts(A=windowsFont("Times New Roman")) #Import font op <- par(family = "A", font =1) # Provide the names of columns colnames(data7) <-c("Digestate", "Filtrate in manure pit", "SLS solids", "Dairy G", "Parlor", "Beef", "Feed wastes", "Poultry", "Swine", "Other animal wastes", "Filtrate in food pit", "SLS food solids", "P & A", "Food pulp", "FOG", "Food feed wastes", "Others", "Cart food wastes") radarchart(data7, axistype=2, pty=32, plty=1, axislabcol="grey", na.itp=FALSE, title="", vlcex=1) legend(x=1.5, y=1, legend = unique(metadata$Month), title="Months of 2014", bty = "n", pch=20, col=unique(metadata$Month), cex=1, pt.cex=2) ####### C. Violin Chart and ANOVA Tukey Multiple Comparison Codes for Feedstock FULL CODES ## Feed amount - Violin analysis ## SCAD operation ## Wei Liao, December 9, 2021 ## Fahmi Dwilaksono, December 12, 2021 Updated ## Fahmi, Jan 4,2022 143 # Load libraries ----- library (MASS) library(ggplot2) library(grid) library(gridExtra) library(ggpubr) library(plyr) library(inferr) # Plot bar chart with standard deviation ----- #data : a data frame #varname : the name of a column containing the variable to be summarized #groupnames : vector of column names to be used as #grouping variables data_summary <- function(data, varname, groupnames){ require(plyr) summary_func <- function(x, col){ c(mean = mean(x[[col]], na.rm=TRUE), sd = sd(x[[col]], na.rm=TRUE)) } data_sum<-ddply(data, groupnames, .fun=summary_func, varname) data_sum <- rename(data_sum, c("mean" = varname)) return(data_sum) } # Choose data file FeedAmount_Violin.txt ----- con <-file.choose(new = FALSE) metadata <- read.table(con, header = T, row.names = 1, fill = TRUE) head(metadata) # Define factors for metadata ----- metadata$Feed_type <- factor(metadata$Feed_type) metadata$Year<- factor(metadata$Year) metadata$Month <- factor(metadata$Month) #Anova #1. Digstate ##Data selection data1<-metadata[which(metadata$Feed_type=="Digestate"),] data1 ## Anova fit1 <- aov(Daily_Feed~Year, data1) 144 summary(fit1) Tukey1 <- TukeyHSD(fit1, conf.level=0.95) #Tukey multiple comparison Tukey1 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_1 <- ggplot(data1, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Digestate", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_1 box_1_1 <- box_1 + geom_boxplot(width=0.2) # Add median and quartile box_1_1 ## Mean and standard deviation box_1_data <- data_summary(data1, varname="Daily_Feed", groupnames=c("Year")) box_1_data #2. Filtrate manure pit ##Data selection data2<-metadata[which(metadata$Feed_type=="Filtrate_manure_pit"),] data2 ## Anova fit2 <- aov(Daily_Feed~Year, data2) summary(fit2) Tukey2 <- TukeyHSD(fit2, conf.level=0.95) #Tukey multiple comparison Tukey2 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_2 <- ggplot(data2, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Filtrate manure pit", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), 145 axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_2 box_2_1 <- box_2 + geom_boxplot(width=0.2) # Add median and quartile box_2_1 ## Mean and standard deviation box_2_data <- data_summary(data2, varname="Daily_Feed", groupnames=c("Year")) box_2_data #3. SLS Solids ##Data selection data3<-metadata[which(metadata$Feed_type=="SLS_Solids"),] data3 ## Anova fit3 <- aov(Daily_Feed~Year, data3) summary(fit3) Tukey3 <- TukeyHSD(fit3, conf.level=0.95) #Tukey multiple comparison Tukey3 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_3 <- ggplot(data3, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "SLS Solids", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_3 box_3_1 <- box_3 + geom_boxplot(width=0.2) # Add median and quartile box_3_1 ## Mean and standard deviation box_3_data <- data_summary(data3, varname="Daily_Feed", groupnames=c("Year")) box_3_data 146 #4. Dairy G ##Data selection data4<-metadata[which(metadata$Feed_type=="Dairy_G"),] data4 ## Anova fit4 <- aov(Daily_Feed~Year, data4) summary(fit4) Tukey4 <- TukeyHSD(fit4, conf.level=0.95) #Tukey multiple comparison Tukey4 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_4 <- ggplot(data4, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Dairy G", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_4 box_4_1 <- box_4 + geom_boxplot(width=0.2) # Add median and quartile box_4_1 ## Mean and standard deviation box_4_data <- data_summary(data4, varname="Daily_Feed", groupnames=c("Year")) box_4_data #5. Parlor ##Data selection data5<-metadata[which(metadata$Feed_type=="Parlor"),] data5 ## Anova fit5 <- aov(Daily_Feed~Year, data5) summary(fit5) Tukey5 <- TukeyHSD(fit5, conf.level=0.95) #Tukey multiple comparison Tukey5 #Output Tukey results ## Plot 147 windowsFonts(A=windowsFont("Times New Roman")) #Import font box_5 <- ggplot(data5, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Parlor", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_5 box_5_1 <- box_5 + geom_boxplot(width=0.2) # Add median and quartile box_5_1 ## Mean and standard deviation box_5_data <- data_summary(data5, varname="Daily_Feed", groupnames=c("Year")) box_5_data #6. Beef ##Data selection data6<-metadata[which(metadata$Feed_type=="Beef"),] data6 ## Anova fit6 <- aov(Daily_Feed~Year, data6) summary(fit6) Tukey6 <- TukeyHSD(fit6, conf.level=0.95) #Tukey multiple comparison Tukey6 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_6 <- ggplot(data6, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Beef", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_6 box_6_1 <- box_6 + geom_boxplot(width=0.2) # Add median and quartile 148 box_6_1 ## Mean and standard deviation box_6_data <- data_summary(data6, varname="Daily_Feed", groupnames=c("Year")) box_6_data #7. Waste Feed_Manure ##Data selection data7<-metadata[which(metadata$Feed_type=="W_Feed_manure"),] data7 ## Anova fit7 <- aov(Daily_Feed~Year, data7) summary(fit7) Tukey7 <- TukeyHSD(fit7, conf.level=0.95) #Tukey multiple comparison Tukey7 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_7 <- ggplot(data7, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "W Feed Manure", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_7 box_7_1 <- box_7 + geom_boxplot(width=0.2) # Add median and quartile box_7_1 ## Mean and standard deviation box_7_data <- data_summary(data7, varname="Daily_Feed", groupnames=c("Year")) box_7_data #8. Poultry ##Data selection data8<-metadata[which(metadata$Feed_type=="Poultry"),] data8 149 ## Anova fit8 <- aov(Daily_Feed~Year, data8) summary(fit8) Tukey8 <- TukeyHSD(fit8, conf.level=0.95) #Tukey multiple comparison Tukey8 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_8 <- ggplot(data8, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Poultry", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_8 box_8_1 <- box_8 + geom_boxplot(width=0.2) # Add median and quartile box_8_1 ## Mean and standard deviation box_8_data <- data_summary(data8, varname="Daily_Feed", groupnames=c("Year")) box_8_data #9. Swine ##Data selection data9<-metadata[which(metadata$Feed_type=="Swine"),] data9 ## Anova fit9 <- aov(Daily_Feed~Year, data9) summary(fit9) Tukey9 <- TukeyHSD(fit9, conf.level=0.95) #Tukey multiple comparison Tukey9 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_9 <- ggplot(data9, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Swine", subtitle=NULL) + 150 theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_9 box_9_1 <- box_9 + geom_boxplot(width=0.2) # Add median and quartile box_9_1 ## Mean and standard deviation box_9_data <- data_summary(data9, varname="Daily_Feed", groupnames=c("Year")) box_9_data #10. ANS Other ##Data selection data10<-metadata[which(metadata$Feed_type=="ANS_Other"),] data10 ## Anova fit10 <- aov(Daily_Feed~Year, data10) summary(fit10) Tukey10 <- TukeyHSD(fit10, conf.level=0.95) #Tukey multiple comparison Tukey10 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_10 <- ggplot(data10, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "ANS Other", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_10 box_10_1 <- box_10 + geom_boxplot(width=0.2) # Add median and quartile box_10_1 ## Mean and standard deviation box_10_data <- data_summary(data10, varname="Daily_Feed", groupnames=c("Year")) 151 box_10_data #11. Total Manure Pit ##Data selection data11<-metadata[which(metadata$Feed_type=="Total_manure_pit"),] data11 ## Anova fit11 <- aov(Daily_Feed~Year, data11) summary(fit11) Tukey11 <- TukeyHSD(fit11, conf.level=0.95) #Tukey multiple comparison Tukey11 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_11 <- ggplot(data11, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Total Manure Pit", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_11 box_11_1 <- box_11 + geom_boxplot(width=0.2) # Add median and quartile box_11_1 ## Mean and standard deviation box_11_data <- data_summary(data11, varname="Daily_Feed", groupnames=c("Year")) box_11_data #12. Filtrate Food Pit ##Data selection data12<-metadata[which(metadata$Feed_type=="Filtrate_food_pit"),] data12 ## Anova fit12 <- aov(Daily_Feed~Year, data12) summary(fit12) Tukey12 <- TukeyHSD(fit12, conf.level=0.95) #Tukey multiple comparison Tukey12 #Output Tukey results 152 ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_12 <- ggplot(data12, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Filtrate Food Pit", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_12 box_12_1 <- box_12 + geom_boxplot(width=0.2) # Add median and quartile box_12_1 ## Mean and standard deviation box_12_data <- data_summary(data12, varname="Daily_Feed", groupnames=c("Year")) box_12_data #13. SLS Solids Food ##Data selection data13<-metadata[which(metadata$Feed_type=="SLS_Solids_food"),] data13 ## Anova fit13 <- aov(Daily_Feed~Year, data13) summary(fit13) Tukey13 <- TukeyHSD(fit13, conf.level=0.95) #Tukey multiple comparison Tukey13 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_13 <- ggplot(data13, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "SLS Solids Food", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") 153 box_13 box_13_1 <- box_13 + geom_boxplot(width=0.2) # Add median and quartile box_13_1 ## Mean and standard deviation box_13_data <- data_summary(data13, varname="Daily_Feed", groupnames=c("Year")) box_13_data #14. P_A ##Data selection data14<-metadata[which(metadata$Feed_type=="P_A"),] data14 ## Anova fit14 <- aov(Daily_Feed~Year, data14) summary(fit14) Tukey14 <- TukeyHSD(fit14, conf.level=0.95) #Tukey multiple comparison Tukey14 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_14 <- ggplot(data14, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Pinnapples", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_14 box_14_1 <- box_14 + geom_boxplot(width=0.2) # Add median and quartile box_14_1 ## Mean and standard deviation box_14_data <- data_summary(data14, varname="Daily_Feed", groupnames=c("Year")) box_14_data 154 #15. Pulp ##Data selection data15<-metadata[which(metadata$Feed_type=="Pulp"),] data15 ## Anova fit15 <- aov(Daily_Feed~Year, data15) summary(fit15) Tukey15 <- TukeyHSD(fit15, conf.level=0.95) #Tukey multiple comparison Tukey15 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_15 <- ggplot(data15, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Pulp", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_15 box_15_1 <- box_15 + geom_boxplot(width=0.2) # Add median and quartile box_15_1 ## Mean and standard deviation box_15_data <- data_summary(data15, varname="Daily_Feed", groupnames=c("Year")) box_15_data #16. FOG ##Data selection data16<-metadata[which(metadata$Feed_type=="FOG"),] data16 ## Anova fit16 <- aov(Daily_Feed~Year, data16) summary(fit16) Tukey16 <- TukeyHSD(fit16, conf.level=0.95) #Tukey multiple comparison Tukey16 #Output Tukey results ## Plot 155 windowsFonts(A=windowsFont("Times New Roman")) #Import font box_16 <- ggplot(data16, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "FOG", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_16 box_16_1 <- box_16 + geom_boxplot(width=0.2) # Add median and quartile box_16_1 ## Mean and standard deviation box_16_data <- data_summary(data16, varname="Daily_Feed", groupnames=c("Year")) box_16_data #17. Waste Feed ##Data selection data17<-metadata[which(metadata$Feed_type=="W_Feed"),] data17 ## Anova fit17 <- aov(Daily_Feed~Year, data17) summary(fit17) Tukey17 <- TukeyHSD(fit17, conf.level=0.95) #Tukey multiple comparison Tukey17 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_17 <- ggplot(data17, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Waste Feed", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_17 box_17_1 <- box_17 + geom_boxplot(width=0.2) # Add median and quartile 156 box_17_1 ## Mean and standard deviation box_17_data <- data_summary(data17, varname="Daily_Feed", groupnames=c("Year")) box_17_data #18. Other ##Data selection data18<-metadata[which(metadata$Feed_type=="Other"),] data18 ## Anova fit18 <- aov(Daily_Feed~Year, data18) summary(fit18) Tukey18 <- TukeyHSD(fit18, conf.level=0.95) #Tukey multiple comparison Tukey18 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_18 <- ggplot(data18, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Other", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_18 box_18_1 <- box_18 + geom_boxplot(width=0.2) # Add median and quartile box_18_1 ## Mean and standard deviation box_18_data <- data_summary(data18, varname="Daily_Feed", groupnames=c("Year")) box_18_data #19. Cart Food ##Data selection data19<-metadata[which(metadata$Feed_type=="Cart_Food"),] data19 157 ## Anova fit19 <- aov(Daily_Feed~Year, data19) summary(fit19) Tukey19 <- TukeyHSD(fit19, conf.level=0.95) #Tukey multiple comparison Tukey19 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_19 <- ggplot(data19, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Cart Food", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_19 box_19_1 <- box_19 + geom_boxplot(width=0.2) # Add median and quartile box_19_1 ## Mean and standard deviation box_19_data <- data_summary(data19, varname="Daily_Feed", groupnames=c("Year")) box_19_data #20. Total Food Pit ##Data selection data20<-metadata[which(metadata$Feed_type=="Total_food_pit"),] data20 ## Anova fit20 <- aov(Daily_Feed~Year, data20) summary(fit20) Tukey20 <- TukeyHSD(fit20, conf.level=0.95) #Tukey multiple comparison Tukey20 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_20 <- ggplot(data20, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Total Food Pit", subtitle=NULL) + theme_classic() + 158 theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_20 box_20_1 <- box_20 + geom_boxplot(width=0.2) # Add median and quartile box_20_1 ## Mean and standard deviation box_20_data <- data_summary(data20, varname="Daily_Feed", groupnames=c("Year")) box_20_data #21. Total Feedstock ##Data selection data21<-metadata[which(metadata$Feed_type=="Total_feedstock"),] data21 ## Anova fit21 <- aov(Daily_Feed~Year, data21) summary(fit21) Tukey21 <- TukeyHSD(fit21, conf.level=0.95) #Tukey multiple comparison Tukey21 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_21 <- ggplot(data21, aes(x=Year, y=Daily_Feed)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Daily feed (kg/day)") + labs(title = "Total Feedstock", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_21 box_21_1 <- box_21 + geom_boxplot(width=0.2) # Add median and quartile box_21_1 ## Mean and standard deviation box_21_data <- data_summary(data21, varname="Daily_Feed", groupnames=c("Year")) 159 box_21_data ### RESULTS C.1 Violin Chart of Feedstock in Manure Pit C.1.1 Digestate > ## Anova > fit1 <- aov(Daily_Feed~Year, data1) > summary(fit1) Df Sum Sq Mean Sq F value Pr(>F) Year 4 145459560 36364890 0.906 0.476 Residuals 24 963661949 40152581 > Tukey1 <- TukeyHSD(fit1, conf.level=0.95) #Tukey multiple comparison > Tukey1 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data1) $Year diff lwr upr p adj 2016-2015 -1625.8753 -21303.490 18051.740 0.9991656 2017-2015 -5303.0678 -24800.975 14194.839 0.9276618 2018-2015 -4567.2934 -24730.849 15596.262 0.9615453 2020-2015 -9407.0445 -32270.367 13456.278 0.7445301 2017-2016 -3677.1925 -12067.756 4713.371 0.6989719 2018-2016 -2941.4181 -12780.226 6897.389 0.9011895 2020-2016 -7781.1692 -22374.479 6812.141 0.5291867 2018-2017 735.7743 -8738.506 10210.054 0.9993463 2020-2017 -4103.9767 -18454.048 10246.094 0.9145118 2020-2018 -4839.7510 -20081.966 10402.464 0.8802393 160 Figure C.1 Feedstock distribution for Digestate Year Daily_Feed sd 1 2015 20755.92 NA 2 2016 19130.04 7853.040 3 2017 15452.85 5927.523 4 2018 16188.62 4250.253 5 2020 11348.87 5349.909 C.1.2 Filtrate Manure Pit > ## Anova > fit2 <- aov(Daily_Feed~Year, data2) > summary(fit2) Df Sum Sq Mean Sq F value Pr(>F) Year 6 1.553e+09 258810814 3.729 0.00392 ** Residuals 49 3.401e+09 69412500 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey2 <- TukeyHSD(fit2, conf.level=0.95) #Tukey multiple comparison > Tukey2 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data2) $Year diff lwr upr p adj 2015-2014 -3603.9396 -14059.61 6851.7330 0.9367379 2016-2014 -11798.0489 -22764.05 -832.0469 0.0274157 2017-2014 -10023.4382 -21713.24 1666.3592 0.1376290 2018-2014 -9645.7095 -24432.26 5140.8446 0.4245674 2019-2014 -14760.8341 -26941.32 -2580.3440 0.0085485 161 2020-2014 -12419.0466 -28950.92 4112.8235 0.2604355 2016-2015 -8194.1092 -19160.11 2771.8928 0.2661668 2017-2015 -6419.4986 -18109.30 5270.2988 0.6269482 2018-2015 -6041.7698 -20828.32 8744.7843 0.8681169 2019-2015 -11156.8944 -23337.38 1023.5956 0.0926584 2020-2015 -8815.1069 -25346.98 7716.7631 0.6582698 2017-2016 1774.6107 -10373.78 13923.0045 0.9993180 2018-2016 2152.3394 -12999.37 17304.0486 0.9994188 2019-2016 -2962.7852 -15584.05 9658.4844 0.9905911 2020-2016 -620.9977 -17480.26 16238.2679 0.9999998 2018-2017 377.7287 -15305.78 16061.2377 1.0000000 2019-2017 -4737.3958 -17992.38 8517.5885 0.9255114 2020-2017 -2395.6084 -19734.38 14943.1632 0.9995043 2019-2018 -5115.1246 -21167.71 10937.4587 0.9561045 2020-2018 -2773.3371 -22334.11 16787.4354 0.9994251 2020-2019 2341.7875 -15331.52 20015.1000 0.9996104 Year Daily_Feed sd 1 2014 25025.88 7210.635 2 2015 21421.94 12445.617 3 2016 13227.83 5988.434 4 2017 15002.44 6870.595 5 2018 15380.17 10625.096 6 2019 10265.05 3988.524 7 2020 12606.83 4367.135 C.1.3 SLS Solids > ## Anova > fit3 <- aov(Daily_Feed~Year, data3) > summary(fit3) Df Sum Sq Mean Sq F value Pr(>F) Year 5 1589309 317862 0.187 0.965 Residuals 27 45912828 1700475 > Tukey3 <- TukeyHSD(fit3, conf.level=0.95) #Tukey multiple comparison > Tukey3 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data3) $Year diff lwr upr p adj 2016-2015 349.67407 -2280.370 2979.718 0.9983908 2017-2015 274.02501 -1561.700 2109.750 0.9972077 2018-2015 -168.21459 -2531.880 2195.451 0.9999236 2019-2015 244.98504 -1943.344 2433.314 0.9992972 2020-2015 727.06262 -2367.702 3821.827 0.9777973 162 2017-2016 -75.64907 -2739.198 2587.900 0.9999992 2018-2016 -517.88867 -3569.368 2533.590 0.9948985 2019-2016 -104.68903 -3022.461 2813.083 0.9999975 2020-2016 377.38854 -3269.827 4024.604 0.9995207 2018-2017 -442.23960 -2843.131 1958.651 0.9925485 2019-2017 -29.03997 -2257.525 2199.445 1.0000000 2020-2017 453.03761 -2670.251 3576.326 0.9975627 2019-2018 413.19963 -2266.945 3093.344 0.9967478 2020-2018 895.27721 -2564.775 4355.329 0.9663717 2020-2019 482.07758 -2860.650 3824.805 0.9976286 Figure C.2 Feedstock distribution for SLS Solids Year Daily_Feed sd 1 2015 2938.868 1252.5055 2 2016 3288.542 1312.8727 3 2017 3212.893 1673.5432 4 2018 2770.653 474.1477 5 2019 3183.853 949.3934 6 2020 3665.931 1288.7252 C.1.4 Dairy Gutter > ## Anova > fit4 <- aov(Daily_Feed~Year, data4) > summary(fit4) Df Sum Sq Mean Sq F value Pr(>F) Year 6 42372311 7062052 3.554 0.00368 ** Residuals 77 152983389 1986797 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey4 <- TukeyHSD(fit4, conf.level=0.95) #Tukey multiple comparison > Tukey4 #Output Tukey results 163 Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data4) $Year diff lwr upr p adj 2015-2014 1192.79576 -549.44637 2935.0379 0.3790328 2016-2014 2400.10647 657.86433 4142.3486 0.0014801 2017-2014 1858.36642 116.12429 3600.6086 0.0289095 2018-2014 1755.51444 13.27230 3497.7566 0.0470523 2019-2014 1838.33278 96.09064 3580.5749 0.0318566 2020-2014 1224.58500 -517.65714 2966.8271 0.3470729 2016-2015 1207.31071 -534.93143 2949.5528 0.3642678 2017-2015 665.57066 -1076.67147 2407.8128 0.9078985 2018-2015 562.71868 -1179.52346 2304.9608 0.9573103 2019-2015 645.53702 -1096.70512 2387.7792 0.9194903 2020-2015 31.78924 -1710.45290 1774.0314 1.0000000 2017-2016 -541.74005 -2283.98218 1200.5021 0.9644623 2018-2016 -644.59203 -2386.83417 1097.6501 0.9200129 2019-2016 -561.77369 -2304.01583 1180.4684 0.9576524 2020-2016 -1175.52147 -2917.76361 566.7207 0.3969598 2018-2017 -102.85199 -1845.09412 1639.3901 0.9999971 2019-2017 -20.03365 -1762.27578 1722.2085 1.0000000 2020-2017 -633.78142 -2376.02356 1108.4607 0.9258387 2019-2018 82.81834 -1659.42380 1825.0605 0.9999992 2020-2018 -530.92944 -2273.17157 1211.3127 0.9677964 2020-2019 -613.74778 -2355.98991 1128.4944 0.9358943 Year Daily_Feed sd 1 2014 11554.80 2239.1947 2 2015 12747.60 1283.0903 3 2016 13954.91 1216.0525 4 2017 13413.17 1876.9954 5 2018 13310.32 837.6477 6 2019 13393.14 809.3594 7 2020 12779.39 942.6853 C.1.5 Parlor > ## Anova > fit5 <- aov(Daily_Feed~Year, data5) > summary(fit5) Df Sum Sq Mean Sq F value Pr(>F) Year 6 4.214e+08 70235648 5.017 0.000219 *** Residuals 77 1.078e+09 14000917 164 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey5 <- TukeyHSD(fit5, conf.level=0.95) #Tukey multiple comparison > Tukey5 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data5) $Year diff lwr upr p adj 2015-2014 1580.2767 -3044.705 6205.2580 0.9443209 2016-2014 -3837.9931 -8462.974 786.9881 0.1693431 2017-2014 -4892.8213 -9517.803 -267.8401 0.0311264 2018-2014 -1571.4695 -6196.451 3053.5118 0.9457526 2019-2014 -3500.9365 -8125.918 1124.0448 0.2613178 2020-2014 -4230.8038 -8855.785 394.1775 0.0953095 2016-2015 -5418.2698 -10043.251 -793.2886 0.0113920 2017-2015 -6473.0980 -11098.079 -1848.1168 0.0011735 2018-2015 -3151.7462 -7776.727 1473.2350 0.3847257 2019-2015 -5081.2132 -9706.194 -456.2319 0.0219480 2020-2015 -5811.0805 -10436.062 -1186.0993 0.0050681 2017-2016 -1054.8282 -5679.809 3570.1530 0.9927745 2018-2016 2266.5236 -2358.458 6891.5049 0.7534924 2019-2016 337.0567 -4287.925 4962.0379 0.9999897 2020-2016 -392.8107 -5017.792 4232.1706 0.9999746 2018-2017 3321.3518 -1303.629 7946.3331 0.3214669 2019-2017 1391.8848 -3233.096 6016.8661 0.9697196 2020-2017 662.0175 -3962.964 5286.9988 0.9994637 2019-2018 -1929.4670 -6554.448 2695.5143 0.8663878 2020-2018 -2659.3343 -7284.316 1965.6469 0.5912253 2020-2019 -729.8673 -5354.849 3895.1139 0.9990631 Year Daily_Feed sd 1 2014 25669.45 3318.805 2 2015 27249.73 5947.718 3 2016 21831.46 3686.228 4 2017 20776.63 3026.019 5 2018 24097.98 4252.452 6 2019 22168.52 2156.050 7 2020 21438.65 2477.829 C.1.6 Beef > ## Anova > fit6 <- aov(Daily_Feed~Year, data6) > summary(fit6) 165 Df Sum Sq Mean Sq F value Pr(>F) Year 6 100115286 16685881 5.879 8.31e-05 *** Residuals 56 158938879 2838194 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey6 <- TukeyHSD(fit6, conf.level=0.95) #Tukey multiple comparison > Tukey6 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data6) $Year diff lwr upr p adj 2015-2014 -1583.1873 -5017.726 1851.35126 0.7943863 2016-2014 -2459.4665 -4826.559 -92.37421 0.0367349 2017-2014 -3991.3828 -6263.117 -1719.64909 0.0000309 2018-2014 -3503.7920 -5819.357 -1188.22661 0.0004279 2019-2014 -2945.2863 -5217.020 -673.55254 0.0037558 2020-2014 -3327.0973 -6042.338 -611.85619 0.0073262 2016-2015 -876.2793 -4267.614 2515.05581 0.9850225 2017-2015 -2408.1955 -5733.673 917.28212 0.3045062 2018-2015 -1920.6047 -5276.178 1434.96833 0.5861362 2019-2015 -1362.0990 -4687.577 1963.37867 0.8700467 2020-2015 -1743.9100 -5386.788 1898.96820 0.7644217 2017-2016 -1531.9163 -3737.789 673.95606 0.3534888 2018-2016 -1044.3255 -3295.312 1206.66137 0.7894796 2019-2016 -485.8197 -2691.692 1720.05261 0.9935552 2020-2016 -867.6308 -3528.013 1792.75134 0.9525220 2018-2017 487.5908 -1662.895 2638.07671 0.9924675 2019-2017 1046.0965 -1057.120 3149.31328 0.7311977 2020-2017 664.2855 -1911.618 3240.18939 0.9851726 2019-2018 558.5058 -1591.980 2708.99170 0.9846223 2020-2018 176.6947 -2437.947 2791.33618 0.9999928 2020-2019 -381.8111 -2957.715 2194.09284 0.9992895 166 Figure C.3 Feedstock distribution for Beef Year Daily_Feed sd 1 2014 6404.719 3618.6704 2 2015 4821.532 2269.4265 3 2016 3945.252 1348.7279 4 2017 2413.336 770.9569 5 2018 2900.927 1055.3299 6 2019 3459.433 929.5379 7 2020 3077.622 256.3979 C.1.7 Waste Feed Manure > ## Anova > fit7 <- aov(Daily_Feed~Year, data7) > summary(fit7) Df Sum Sq Mean Sq F value Pr(>F) Year 5 23261394 4652279 2.906 0.0472 * Residuals 16 25618186 1601137 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey7 <- TukeyHSD(fit7, conf.level=0.95) #Tukey multiple comparison > Tukey7 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data7) $Year diff lwr upr p adj 2015-2014 -3023.9467 -6745.8869 697.9935 0.1491613 2016-2014 -648.2586 -4370.1988 3073.6816 0.9922582 2017-2014 -2638.3935 -5615.9456 339.1587 0.0985172 167 2019-2014 -2239.7768 -5217.3290 737.7753 0.2056048 2020-2014 -916.8304 -3894.3825 2060.7218 0.9138622 2016-2015 2375.6881 -1701.4931 6452.8693 0.4489565 2017-2015 385.5532 -3025.6613 3796.7677 0.9989874 2019-2015 784.1698 -2627.0447 4195.3844 0.9735442 2020-2015 2107.1163 -1304.0983 5518.3308 0.3888309 2017-2016 -1990.1349 -5401.3494 1421.0796 0.4476599 2019-2016 -1591.5183 -5002.7328 1819.6963 0.6671881 2020-2016 -268.5718 -3679.7864 3142.6427 0.9998256 2019-2017 398.6166 -2180.0192 2977.2525 0.9955334 2020-2017 1721.5631 -857.0727 4300.1989 0.3114649 2020-2019 1322.9464 -1255.6894 3901.5822 0.5784162 Figure C.4 Feedstock distribution for Waste Feed Manure Year Daily_Feed sd 1 2014 4101.228 2328.1781 2 2015 1077.281 208.4797 3 2016 3452.969 152.3505 4 2017 1462.834 425.8712 5 2019 1861.451 597.4175 6 2020 3184.397 1771.8344 C.1.8 Poultry > ## Anova > fit8 <- aov(Daily_Feed~Year, data8) > summary(fit8) Df Sum Sq Mean Sq F value Pr(>F) Year 3 4476064 1492021 27.26 3.96e-05 *** Residuals 10 547330 54733 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 168 > Tukey8 <- TukeyHSD(fit8, conf.level=0.95) #Tukey multiple comparison > Tukey8 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data8) $Year diff lwr upr p adj 2017-2016 507.16122 27.02892 987.2935 0.0380139 2019-2016 1377.78570 831.13103 1924.4404 0.0000800 2020-2016 1392.98103 773.13290 2012.8292 0.0002110 2019-2017 870.62448 347.92265 1393.3263 0.0021976 2020-2017 885.81982 286.98963 1484.6500 0.0050653 2020-2019 15.19533 -638.18197 668.5726 0.9998615 Figure C.5 Feedstock distribution for Poultry Year Daily_Feed sd 1 2016 572.6599 315.0070 2 2017 1079.8211 225.7625 3 2019 1950.4456 0.0000 4 2020 1965.6409 213.9322 C.1.9 Swine ## Anova > fit9 <- aov(Daily_Feed~Year, data9) > summary(fit9) Df Sum Sq Mean Sq F value Pr(>F) Year 3 471835620 157278540 6.639 0.034 * Residuals 5 118459221 23691844 --- 169 Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey9 <- TukeyHSD(fit9, conf.level=0.95) #Tukey multiple comparison > Tukey9 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data9) $Year diff lwr upr p adj 2015-2014 -8172.480 -23726.627 7381.666 0.3207542 2017-2014 13141.467 -8855.417 35138.352 0.2408659 2018-2014 4667.462 -13292.920 22627.843 0.7773253 2017-2015 21313.948 1233.631 41394.264 0.0400869 2018-2015 12839.942 -2714.204 28394.089 0.0967451 2018-2017 -8474.006 -30470.890 13522.879 0.5388782 Figure C.6 Feedstock distribution for Swine Year Daily_Feed sd 1 2014 12745.482 7762.501 2 2015 4573.001 3664.656 3 2017 25886.949 NA 4 2018 17412.943 4232.458 C.1.10 ANS Other ## Anova > fit10 <- aov(Daily_Feed~Year, data10) > summary(fit10) Df Sum Sq Mean Sq F value Pr(>F) Year 6 39084480 6514080 4.285 0.00281 ** 170 Residuals 32 48650416 1520325 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey10 <- TukeyHSD(fit10, conf.level=0.95) #Tukey multiple comparison > Tukey10 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data10) $Year diff lwr upr p adj 2015-2014 589.27271 -1784.0243 2962.570 0.9852771 2016-2014 286.21655 -2087.0805 2659.514 0.9997286 2017-2014 768.33625 -1660.8107 3197.483 0.9516286 2018-2014 534.89837 -2821.4505 3891.247 0.9986656 2019-2014 1416.00083 -1940.3480 4772.350 0.8347201 2020-2014 2887.73677 514.4398 5261.034 0.0092548 2016-2015 -303.05615 -2240.8451 1634.733 0.9988001 2017-2015 179.06354 -1826.7385 2184.866 0.9999531 2018-2015 -54.37434 -3118.2876 3009.539 1.0000000 2019-2015 826.72812 -2237.1851 3890.641 0.9775607 2020-2015 2298.46406 360.6752 4236.253 0.0118835 2017-2016 482.11970 -1523.6824 2487.922 0.9875485 2018-2016 248.68181 -2815.2315 3312.595 0.9999732 2019-2016 1129.78427 -1934.1290 4193.698 0.9040714 2020-2016 2601.52022 663.7313 4539.309 0.0032201 2018-2017 -233.43788 -3340.8131 2873.937 0.9999831 2019-2017 647.66458 -2459.7106 3755.040 0.9941396 2020-2017 2119.40052 113.5985 4125.203 0.0328451 2019-2018 881.10246 -2994.4753 4756.680 0.9906926 2020-2018 2352.83840 -711.0749 5416.752 0.2259142 2020-2019 1471.73594 -1592.1773 4535.649 0.7370542 171 Figure C.7 Feedstock distribution for ANS Other Year Daily_Feed sd 1 2014 157.9634 99.61785 2 2015 747.2361 709.21161 3 2016 444.1800 562.77269 4 2017 926.2997 1015.13921 5 2018 692.8618 736.09366 6 2019 1573.9642 147.53947 7 2020 3045.7002 2272.07072 C.1.11 Total Manure Pit ## Anova > fit11 <- aov(Daily_Feed~Year, data11) > summary(fit11) Df Sum Sq Mean Sq F value Pr(>F) Year 6 8.058e+09 1.343e+09 7.098 4.84e-06 *** Residuals 77 1.457e+10 1.892e+08 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey11 <- TukeyHSD(fit11, conf.level=0.95) #Tukey multiple comparison > Tukey11 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data11) $Year diff lwr upr p adj 2015-2014 -2271.4375 -19272.71 14729.840 0.9996395 2016-2014 -4954.4342 -21955.71 12046.843 0.9741916 2017-2014 -4340.4218 -21341.70 12660.855 0.9868731 172 2018-2014 -13978.4959 -30979.77 3022.781 0.1776912 2019-2014 -23422.4703 -40423.75 -6421.193 0.0014787 2020-2014 -26187.7939 -43189.07 -9186.517 0.0002514 2016-2015 -2682.9967 -19684.27 14318.281 0.9990631 2017-2015 -2068.9843 -19070.26 14932.293 0.9997901 2018-2015 -11707.0583 -28708.34 5294.219 0.3719687 2019-2015 -21151.0328 -38152.31 -4149.756 0.0057191 2020-2015 -23916.3564 -40917.63 -6915.079 0.0010877 2017-2016 614.0124 -16387.26 17615.290 0.9999998 2018-2016 -9024.0616 -26025.34 7977.216 0.6782326 2019-2016 -18468.0361 -35469.31 -1466.759 0.0244380 2020-2016 -21233.3597 -38234.64 -4232.083 0.0054555 2018-2017 -9638.0740 -26639.35 7363.203 0.6073685 2019-2017 -19082.0485 -36083.33 -2080.771 0.0178006 2020-2017 -21847.3721 -38848.65 -4846.095 0.0038195 2019-2018 -9443.9744 -26445.25 7557.303 0.6300296 2020-2018 -12209.2981 -29210.58 4791.979 0.3214553 2020-2019 -2765.3236 -19766.60 14235.954 0.9988876 Year Daily_Feed sd 1 2014 71283.72 11368.291 2 2015 69012.28 17246.236 3 2016 66329.29 18413.578 4 2017 66943.30 15397.431 5 2018 57305.23 14024.063 6 2019 47861.25 8318.581 7 2020 45095.93 7459.882 C.2 Violin Chart of Feedstock in Food Pit C.2.1 Filtrate Food Pit ## Anova > fit12 <- aov(Daily_Feed~Year, data12) > summary(fit12) Df Sum Sq Mean Sq F value Pr(>F) Year 6 5.047e+08 84113595 2.001 0.0802 . Residuals 58 2.438e+09 42039631 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey12 <- TukeyHSD(fit12, conf.level=0.95) #Tukey multiple comparison > Tukey12 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data12) $Year 173 diff lwr upr p adj 2015-2014 -178.2101 -8444.244 8087.824 1.0000000 2016-2014 -3729.6602 -11814.001 4354.680 0.7947930 2017-2014 158.3603 -8880.207 9196.928 1.0000000 2018-2014 4296.0265 -4742.541 13334.594 0.7714169 2019-2014 -4123.0379 -14024.292 5778.217 0.8618493 2020-2014 -4589.5006 -13628.068 4449.067 0.7132479 2016-2015 -3551.4501 -11817.484 4714.584 0.8435061 2017-2015 336.5704 -8864.867 9538.008 0.9999998 2018-2015 4474.2366 -4727.201 13675.674 0.7523395 2019-2015 -3944.8278 -13994.982 6105.326 0.8918034 2020-2015 -4411.2904 -13612.728 4790.147 0.7643244 2017-2016 3888.0205 -5150.547 12926.588 0.8427671 2018-2016 8025.6868 -1012.881 17064.254 0.1135614 2019-2016 -393.3777 -10294.632 9507.877 0.9999997 2020-2016 -859.8403 -9898.408 8178.727 0.9999465 2018-2017 4137.6662 -5763.588 14038.921 0.8598617 2019-2017 -4281.3982 -14975.975 6413.179 0.8824651 2020-2017 -4747.8609 -14649.115 5153.394 0.7641371 2019-2018 -8419.0644 -19113.642 2275.513 0.2154937 2020-2018 -8885.5271 -18786.782 1015.727 0.1063931 2020-2019 -466.4627 -11161.040 10228.115 0.9999995 Figure C.8 Feedstock distribution for Filtrate Food Pit Year Daily_Feed sd 1 2014 18804.22 6936.435 2 2015 18626.01 6325.763 3 2016 15074.56 5643.998 4 2017 18962.58 6082.968 5 2018 23100.25 9488.308 6 2019 14681.19 4057.150 174 7 2020 14214.72 5168.211 C.2.2 SLS Solids Food > ## Anova > fit13 <- aov(Daily_Feed~Year, data13) > summary(fit13) Df Sum Sq Mean Sq F value Pr(>F) Year 5 2340050 468010 0.472 0.795 Residuals 38 37707433 992301 > Tukey13 <- TukeyHSD(fit13, conf.level=0.95) #Tukey multiple comparison > Tukey13 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data13) $Year diff lwr upr p adj 2016-2015 357.09030 -2005.4356 2719.6162 0.9974016 2017-2015 -189.04707 -2525.1752 2147.0811 0.9998748 2018-2015 -231.78551 -2546.5787 2083.0077 0.9996439 2019-2015 197.65271 -2084.7641 2480.0696 0.9998253 2020-2015 129.42492 -2598.5850 2857.4348 0.9999911 2017-2016 -546.13737 -1998.2315 905.9568 0.8665252 2018-2016 -588.87581 -2006.3913 828.6397 0.8112956 2019-2016 -159.43759 -1523.4425 1204.5674 0.9992436 2020-2016 -227.66538 -2250.8117 1795.4809 0.9993710 2018-2017 -42.73845 -1415.8067 1330.3298 0.9999989 2019-2017 386.69978 -931.0542 1704.4538 0.9489365 2020-2017 318.47198 -1673.7848 2310.7287 0.9966128 2019-2018 429.43823 -850.1118 1708.9883 0.9127921 2020-2018 361.21043 -1605.9855 2328.4064 0.9935241 2020-2019 -68.22780 -1997.2221 1860.7665 0.9999979 175 Figure C.9 Feedstock distribution for SLS Solids Food Year Daily_Feed sd 1 2015 3056.303 516.3881 2 2016 3413.393 1016.5215 3 2017 2867.256 930.6221 4 2018 2824.517 1268.9409 5 2019 3253.956 888.4573 6 2020 3185.728 228.3480 C.2.3 Pineapple (PA) > ## Anova > fit14 <- aov(Daily_Feed~Year, data14) > summary(fit14) Df Sum Sq Mean Sq F value Pr(>F) Year 2 1362345 681173 0.14 0.87 Residuals 26 126273793 4856684 > Tukey14 <- TukeyHSD(fit14, conf.level=0.95) #Tukey multiple comparison > Tukey14 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data14) $Year diff lwr upr p adj 2015-2014 -476.3094 -2711.951 1759.332 0.8576849 2016-2014 -221.6931 -3136.613 2693.227 0.9805173 2016-2015 254.6163 -2660.303 3169.536 0.9743880 176 Figure C.10 Feedstock distribution for Pineapple Year Daily_Feed sd 1 2014 11657.20 2206.999 2 2015 11180.89 2370.419 3 2016 11435.51 1649.753 C.2.4 Pulp ## Anova > fit15 <- aov(Daily_Feed~Year, data15) > summary(fit15) Df Sum Sq Mean Sq F value Pr(>F) Year 6 10479022 1746504 8.78 4.63e-07 *** Residuals 67 13326784 198907 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey15 <- TukeyHSD(fit15, conf.level=0.95) #Tukey multiple comparison > Tukey15 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data15) $Year diff lwr upr p adj 2015-2014 203.13362 -350.31903 756.5863 0.9211267 2016-2014 249.21100 -304.24164 802.6637 0.8161929 2017-2014 950.72196 384.83062 1516.6133 0.0000582 2018-2014 590.38779 36.93514 1143.8404 0.0290603 2019-2014 1026.10070 472.64805 1579.5534 0.0000077 2020-2014 564.94884 -310.13664 1440.0343 0.4477754 177 2016-2015 46.07739 -507.37526 599.5300 0.9999766 2017-2015 747.58834 181.69700 1313.4797 0.0027771 2018-2015 387.25417 -166.19848 940.7068 0.3493295 2019-2015 822.96709 269.51444 1376.4197 0.0004964 2020-2015 361.81522 -513.27025 1236.9007 0.8687476 2017-2016 701.51095 135.61961 1267.4023 0.0061482 2018-2016 341.17678 -212.27587 894.6294 0.5044607 2019-2016 776.88970 223.43705 1330.3423 0.0011967 2020-2016 315.73783 -559.34764 1190.8233 0.9269321 2018-2017 -360.33417 -926.22551 205.5572 0.4647631 2019-2017 75.37874 -490.51260 641.2701 0.9996339 2020-2017 -385.77312 -1268.77808 497.2318 0.8363818 2019-2018 435.71292 -117.73973 989.1656 0.2177536 2020-2018 -25.43895 -900.52442 849.6465 1.0000000 2020-2019 -461.15187 -1336.23734 413.9336 0.6815731 Year Daily_Feed sd 1 2014 1078.566 332.9565 2 2015 1281.700 167.5954 3 2016 1327.777 440.1656 4 2017 2029.288 466.0263 5 2018 1668.954 574.3428 6 2019 2104.667 558.1254 7 2020 1643.515 469.1668 C.2.5 FOG ## Anova > fit16 <- aov(Daily_Feed~Year, data16) > summary(fit16) Df Sum Sq Mean Sq F value Pr(>F) Year 6 3.215e+09 535831329 8.095 8.63e-07 *** Residuals 77 5.097e+09 66195940 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey16 <- TukeyHSD(fit16, conf.level=0.95) #Tukey multiple comparison > Tukey16 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data16) $Year diff lwr upr p adj 2015-2014 16422.8276 6366.315 26479.340 0.0000869 2016-2014 16203.2890 6146.776 26259.802 0.0001120 2017-2014 13835.6144 3779.102 23892.127 0.0015085 178 2018-2014 12791.7480 2735.235 22848.261 0.0043489 2019-2014 15421.9768 5365.464 25478.490 0.0002714 2020-2014 21622.0881 11565.575 31678.601 0.0000001 2016-2015 -219.5385 -10276.051 9836.974 1.0000000 2017-2015 -2587.2132 -12643.726 7469.300 0.9863374 2018-2015 -3631.0796 -13687.592 6425.433 0.9282912 2019-2015 -1000.8507 -11057.364 9055.662 0.9999353 2020-2015 5199.2605 -4857.252 15255.773 0.7044122 2017-2016 -2367.6746 -12424.188 7688.838 0.9914359 2018-2016 -3411.5410 -13468.054 6644.972 0.9461552 2019-2016 -781.3122 -10837.825 9275.201 0.9999850 2020-2016 5418.7990 -4637.714 15475.312 0.6626833 2018-2017 -1043.8664 -11100.379 9012.646 0.9999171 2019-2017 1586.3624 -8470.150 11642.875 0.9990653 2020-2017 7786.4737 -2270.039 17842.987 0.2369671 2019-2018 2630.2288 -7426.284 12686.742 0.9851162 2020-2018 8830.3401 -1226.173 18886.853 0.1232539 2020-2019 6200.1113 -3856.402 16256.624 0.5083279 Year Daily_Feed sd 1 2014 19128.24 2959.616 2 2015 35551.07 8151.082 3 2016 35331.53 8097.009 4 2017 32963.85 7527.845 5 2018 31919.99 4847.275 6 2019 34550.22 14652.833 7 2020 40750.33 5266.926 C.2.6 Waste Feed ## Anova > fit17 <- aov(Daily_Feed~Year, data17) > summary(fit17) Df Sum Sq Mean Sq F value Pr(>F) Year 2 162532 81266 0.123 0.891 Residuals 2 1326492 663246 > Tukey17 <- TukeyHSD(fit17, conf.level=0.95) #Tukey multiple comparison > Tukey17 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data17) $Year diff lwr upr p adj 2015-2014 129.8785 -5409.711 5669.468 0.9895938 2018-2014 -400.3705 -5939.960 5139.219 0.9092011 179 2018-2015 -530.2490 -7314.833 6254.335 0.8954420 Figure C.11 Feedstock distribution for Waste Feed Year Daily_Feed sd 1 2014 2033.302 814.3993 2 2015 2163.180 NA 3 2018 1632.931 NA C.2.7 Other ## Anova > fit18 <- aov(Daily_Feed~Year, data18) > summary(fit18) Df Sum Sq Mean Sq F value Pr(>F) Year 6 1.515e+09 252446057 17.69 1.34e-11 *** Residuals 59 8.421e+08 14272946 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey18 <- TukeyHSD(fit18, conf.level=0.95) #Tukey multiple comparison > Tukey18 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data18) $Year diff lwr upr p adj 2015-2014 -10337.815271 -17781.398 -2894.233 0.0014831 2016-2014 -11858.180056 -19301.762 -4414.598 0.0001748 2017-2014 -12134.719979 -20288.756 -3980.684 0.0005340 2018-2014 -11700.216642 -19143.799 -4256.634 0.0002197 2019-2014 -10330.305803 -18018.005 -2642.607 0.0023257 180 2020-2014 118.311914 -7325.270 7561.894 1.0000000 2016-2015 -1520.364785 -6228.100 3187.370 0.9551763 2017-2015 -1796.904708 -7562.679 3968.869 0.9622032 2018-2015 -1362.401371 -6070.136 3345.333 0.9737142 2019-2015 7.509468 -5077.425 5092.444 1.0000000 2020-2015 10456.127185 5748.392 15163.862 0.0000001 2017-2016 -276.539923 -6042.314 5489.234 0.9999991 2018-2016 157.963414 -4549.771 4865.698 0.9999999 2019-2016 1527.874253 -3557.060 6612.809 0.9683570 2020-2016 11976.491970 7268.757 16684.227 0.0000000 2018-2017 434.503337 -5331.271 6200.277 0.9999866 2019-2017 1804.414176 -4273.245 7882.074 0.9701660 2020-2017 12253.031893 6487.258 18018.806 0.0000004 2019-2018 1369.910839 -3715.024 6454.846 0.9816758 2020-2018 11818.528556 7110.794 16526.263 0.0000000 2020-2019 10448.617717 5363.683 15533.552 0.0000009 Figure C.12 Feedstock distribution for Other Year Daily_Feed sd 1 2014 13209.657 10452.4314 2 2015 2871.842 5042.0168 3 2016 1351.477 853.5117 4 2017 1074.937 216.1485 5 2018 1509.441 996.2938 6 2019 2879.352 1944.4467 7 2020 13327.969 5174.6180 181 C.2.8 Cart Food ## Anova > fit19 <- aov(Daily_Feed~Year, data19) > summary(fit19) Df Sum Sq Mean Sq F value Pr(>F) Year 6 91209454 15201576 29.92 <2e-16 *** Residuals 77 39124552 508111 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey19 <- TukeyHSD(fit19, conf.level=0.95) #Tukey multiple comparison > Tukey19 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data19) $Year diff lwr upr p adj 2015-2014 -388.577147 -1269.6485 492.4942 0.8331802 2016-2014 187.560292 -693.5110 1068.6316 0.9950230 2017-2014 -2126.439296 -3007.5106 -1245.3680 0.0000000 2018-2014 -2130.332627 -3011.4039 -1249.2613 0.0000000 2019-2014 -2155.620381 -3036.6917 -1274.5491 0.0000000 2020-2014 -2180.945935 -3062.0172 -1299.8746 0.0000000 2016-2015 576.137439 -304.9339 1457.2088 0.4357217 2017-2015 -1737.862149 -2618.9335 -856.7908 0.0000014 2018-2015 -1741.755481 -2622.8268 -860.6842 0.0000013 2019-2015 -1767.043235 -2648.1145 -885.9719 0.0000009 2020-2015 -1792.368788 -2673.4401 -911.2975 0.0000006 2017-2016 -2313.999588 -3195.0709 -1432.9283 0.0000000 2018-2016 -2317.892919 -3198.9642 -1436.8216 0.0000000 2019-2016 -2343.180673 -3224.2520 -1462.1094 0.0000000 2020-2016 -2368.506227 -3249.5775 -1487.4349 0.0000000 2018-2017 -3.893331 -884.9646 877.1780 1.0000000 2019-2017 -29.181085 -910.2524 851.8902 0.9999999 2020-2017 -54.506639 -935.5780 826.5647 0.9999961 2019-2018 -25.287754 -906.3591 855.7836 1.0000000 2020-2018 -50.613307 -931.6846 830.4580 0.9999975 2020-2019 -25.325553 -906.3969 855.7458 1.0000000 182 Figure C.13 Feedstock distribution for Cart Food Year Daily_Feed sd 1 2014 2248.64454 949.06204 2 2015 1860.06739 1201.71795 3 2016 2436.20483 1097.05343 4 2017 122.20524 57.59513 5 2018 118.31191 59.13605 6 2019 93.02416 32.70010 7 2020 67.69861 22.86615 C.2.9 Total Food Pit ## Anova > fit20 <- aov(Daily_Feed~Year, data20) > summary(fit20) Df Sum Sq Mean Sq F value Pr(>F) Year 6 4.559e+09 759802970 3.641 0.00311 ** Residuals 77 1.607e+10 208668601 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey20 <- TukeyHSD(fit20, conf.level=0.95) #Tukey multiple comparison > Tukey20 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data20) $Year diff lwr upr p adj 2015-2014 13781.6369 -4073.381 31636.6548 0.2402617 2016-2014 5834.7051 -12020.313 23689.7229 0.9548619 2017-2014 -6451.4390 -24306.457 11403.5788 0.9280614 183 2018-2014 -3507.1355 -21362.153 14347.8823 0.9967990 2019-2014 -7225.5316 -25080.549 10629.4863 0.8821913 2020-2014 8102.4005 -9752.617 25957.4184 0.8138326 2016-2015 -7946.9318 -25801.950 9908.0860 0.8271496 2017-2015 -20233.0759 -38088.094 -2378.0581 0.0161612 2018-2015 -17288.7725 -35143.790 566.2454 0.0640479 2019-2015 -21007.1685 -38862.186 -3152.1506 0.0108739 2020-2015 -5679.2364 -23534.254 12175.7814 0.9603379 2017-2016 -12286.1441 -30141.162 5568.8737 0.3728430 2018-2016 -9341.8406 -27196.858 8513.1772 0.6927108 2019-2016 -13060.2367 -30915.255 4794.7812 0.2999843 2020-2016 2267.6954 -15587.322 20122.7133 0.9997311 2018-2017 2944.3035 -14910.714 20799.3213 0.9987976 2019-2017 -774.0925 -18629.110 17080.9253 0.9999995 2020-2017 14553.8395 -3301.178 32408.8574 0.1857086 2019-2018 -3718.3960 -21573.414 14136.6218 0.9955849 2020-2018 11609.5360 -6245.482 29464.5539 0.4427454 2020-2019 15327.9321 -2527.086 33182.9499 0.1406406 Year Daily_Feed sd 1 2014 56727.27 13338.33 2 2015 70508.91 11747.97 3 2016 62561.98 15610.50 4 2017 50275.83 14544.59 5 2018 53220.14 14366.86 6 2019 49501.74 19066.94 7 2020 64829.67 10934.65 C.2.10 Total Feedstock ## Anova > fit21 <- aov(Daily_Feed~Year, data21) > summary(fit21) Df Sum Sq Mean Sq F value Pr(>F) Year 6 1.47e+10 2.451e+09 5.736 5.68e-05 *** Residuals 77 3.29e+10 4.273e+08 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey21 <- TukeyHSD(fit21, conf.level=0.95) #Tukey multiple comparison > Tukey21 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Daily_Feed ~ Year, data = data21) $Year diff lwr upr p adj 184 2015-2014 11511.2200 -14038.03 37060.475 0.8188421 2016-2014 881.2536 -24668.00 26430.509 0.9999999 2017-2014 -10790.8025 -36340.06 14758.452 0.8595263 2018-2014 -17484.5730 -43033.83 8064.682 0.3795427 2019-2014 -30647.0569 -56196.31 -5097.802 0.0087674 2020-2014 -18084.4862 -43633.74 7464.769 0.3386491 2016-2015 -10629.9663 -36179.22 14919.289 0.8678570 2017-2015 -22302.0224 -47851.28 3247.232 0.1276646 2018-2015 -28995.7930 -54545.05 -3446.538 0.0159145 2019-2015 -42158.2768 -67707.53 -16609.022 0.0000713 2020-2015 -29595.7062 -55144.96 -4046.451 0.0128581 2017-2016 -11672.0561 -37221.31 13877.199 0.8090433 2018-2016 -18365.8267 -43915.08 7183.428 0.3203061 2019-2016 -31528.3105 -57077.57 -5979.056 0.0063055 2020-2016 -18965.7399 -44514.99 6583.515 0.2831181 2018-2017 -6693.7706 -32243.03 18855.484 0.9849832 2019-2017 -19856.2544 -45405.51 5693.000 0.2330349 2020-2017 -7293.6838 -32842.94 18255.571 0.9767234 2019-2018 -13162.4838 -38711.74 12386.771 0.7078275 2020-2018 -599.9132 -26149.17 24949.342 1.0000000 2020-2019 12562.5706 -12986.68 38111.826 0.7505992 Year Daily_Feed sd 1 2014 128010.01 16625.39 2 2015 139521.23 22888.56 3 2016 128891.27 25993.66 4 2017 117219.21 16957.01 5 2018 110525.44 21605.39 6 2019 97362.96 23159.83 7 2020 109925.53 14971.79 D. Violin Chart and ANOVA Tukey Multiple Comparison Codes for Digester Output Parameters FULL CODES ## Biogas - Violin analysis ## SCAD operation ## Wei Liao, December 9, 2021 ## Fahmi Dwilaksono, December 12, 2021 # Load libraries ----- library (MASS) library(ggplot2) library(grid) library(gridExtra) library(ggpubr) library(plyr) library(inferr) 185 # Plot bar chart with standard deviation ----- #data : a data frame #varname : the name of a column containing the variable to be summarized #groupnames : vector of column names to be used as #grouping variables data_summary <- function(data, varname, groupnames){ require(plyr) summary_func <- function(x, col){ c(mean = mean(x[[col]], na.rm=TRUE), sd = sd(x[[col]], na.rm=TRUE)) } data_sum<-ddply(data, groupnames, .fun=summary_func, varname) data_sum <- rename(data_sum, c("mean" = varname)) return(data_sum) } # Choose data file Biogas_Violin.txt ----- con <-file.choose(new = FALSE) metadata <- read.table(con, header = T, row.names = 1, fill = TRUE) head(metadata) # Define factors for metadata ----- metadata$Year<- factor(metadata$Year) metadata$Month <- factor(metadata$Month) #Anova #1. Biogas produciton ## Anova fit1 <- aov(Biogas_production~Year, metadata) summary(fit1) Tukey1 <- TukeyHSD(fit1, conf.level=0.95) #Tukey multiple comparison Tukey1 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_1 <- ggplot(metadata, aes(x=Year, y=Biogas_production)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Biogas production (m3/day)") + labs(title = "Biogas production", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), 186 axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_1 box_1_1 <- box_1 + geom_boxplot(width=0.2) # Add median and quartile box_1_1 ## Mean and standard deviation box_1_data <- data_summary(metadata, varname="Biogas_production", groupnames=c("Year")) box_1_data #2. Methane content ## Anova fit2 <- aov(Methane_content~Year, metadata) summary(fit2) Tukey2 <- TukeyHSD(fit2, conf.level=0.95) #Tukey multiple comparison Tukey2 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_2 <- ggplot(metadata, aes(x=Year, y=Methane_content)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Methane content (%)") + labs(title = "Methane content", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_2 box_2_1 <- box_2 + geom_boxplot(width=0.2) # Add median and quartile box_2_1 ## Mean and standard deviation box_2_data <- data_summary(metadata, varname="Methane_content", groupnames=c("Year")) box_2_data #3. H2S ## Anova fit3 <- aov(H2S_content~Year, metadata) summary(fit3) Tukey3 <- TukeyHSD(fit3, conf.level=0.95) #Tukey multiple comparison 187 Tukey3 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_3 <- ggplot(metadata, aes(x=Year, y=H2S_content)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("H2S content (ppmv)") + labs(title = "H2S content", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_3 box_3_1 <- box_3 + geom_boxplot(width=0.2) # Add median and quartile box_3_1 ## Mean and standard deviation box_3_data <- data_summary(metadata, varname="H2S_content", groupnames=c("Year")) box_3_data #4. Electricity ## Anova fit4 <- aov(Electricity~Year, metadata) summary(fit4) Tukey4 <- TukeyHSD(fit4, conf.level=0.95) #Tukey multiple comparison Tukey4 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_4 <- ggplot(metadata, aes(x=Year, y=Electricity)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Electricity (kWh)") + labs(title = "Electricity", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_4 box_4_1 <- box_4 + geom_boxplot(width=0.2) # Add median and quartile box_4_1 188 ## Mean and standard deviation box_4_data <- data_summary(metadata, varname="Electricity", groupnames=c("Year")) box_4_data #5. Effluent_TS ## Anova fit5 <- aov(Effluent_TS~Year, metadata) summary(fit5) Tukey5 <- TukeyHSD(fit5, conf.level=0.95) #Tukey multiple comparison Tukey5 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_5 <- ggplot(metadata, aes(x=Year, y=Effluent_TS)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Effluent_TS (mg/L)") + labs(title = "Effluent_TS", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_5 box_5_1 <- box_5 + geom_boxplot(width=0.2) # Add median and quartile box_5_1 ## Mean and standard deviation box_5_data <- data_summary(metadata, varname="Effluent_TS", groupnames=c("Year")) box_5_data #6. Effluent_TN ## Anova fit6 <- aov(Effluent_TN~Year, metadata) summary(fit6) Tukey6 <- TukeyHSD(fit6, conf.level=0.95) #Tukey multiple comparison Tukey6 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_6 <- ggplot(metadata, aes(x=Year, y=Effluent_TN)) + 189 geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Effluent_TN (mg/L)") + labs(title = "Effluent_TN", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_6 box_6_1 <- box_6 + geom_boxplot(width=0.2) # Add median and quartile box_6_1 ## Mean and standard deviation box_6_data <- data_summary(metadata, varname="Effluent_TN", groupnames=c("Year")) box_6_data #7. Effluent_TP ## Anova fit7 <- aov(Effluent_TP~Year, metadata) summary(fit7) Tukey7 <- TukeyHSD(fit7, conf.level=0.95) #Tukey multiple comparison Tukey7 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_7 <- ggplot(metadata, aes(x=Year, y=Effluent_TP)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Effluent_TP (mg/L)") + labs(title = "Effluent_TP", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_7 box_7_1 <- box_7 + geom_boxplot(width=0.2) # Add median and quartile box_7_1 ## Mean and standard deviation box_7_data <- data_summary(metadata, varname="Effluent_TP", groupnames=c("Year")) 190 box_7_data #8. Effluent_pH ## Anova fit8 <- aov(Effluent_pH~Year, metadata) summary(fit8) Tukey8 <- TukeyHSD(fit8, conf.level=0.95) #Tukey multiple comparison Tukey8 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_8 <- ggplot(metadata, aes(x=Year, y=Effluent_pH)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Effluent_pH") + labs(title = "Effluent_pH", subtitle=NULL) + theme_classic() + theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_8 box_8_1 <- box_8 + geom_boxplot(width=0.2) # Add median and quartile box_8_1 ## Mean and standard deviation box_8_data <- data_summary(metadata, varname="Effluent_TP", groupnames=c("Year")) box_8_data #9. Effluent_VFA ## Anova fit9 <- aov(Effluent_VFA~Year, metadata) summary(fit9) Tukey9 <- TukeyHSD(fit9, conf.level=0.95) #Tukey multiple comparison Tukey9 #Output Tukey results ## Plot windowsFonts(A=windowsFont("Times New Roman")) #Import font box_9 <- ggplot(metadata, aes(x=Year, y=Effluent_VFA)) + geom_violin(trim=TRUE, fill="green") + xlab("Year")+ ylab("Effluent_VFA (mg/L)") + labs(title = "Effluent_VFA", subtitle=NULL) + theme_classic() + 191 theme(title=element_text(size=20, family="A"), axis.text.x = element_text(size=20, family="A"), axis.text.y=element_text(size=20, family="A"), axis.title.y = element_text(size = 20, family="A"), axis.title.x=element_text(size=20, family="A"), legend.position = "top") box_9 box_9_1 <- box_9 + geom_boxplot(width=0.2) # Add median and quartile box_9_1 ## Mean and standard deviation box_9_data <- data_summary(metadata, varname="Effluent_TP", groupnames=c("Year")) box_9_data RESULTS D.1 Biogas Production Year 6 20642104 3440351 7.678 1.76e-06 *** Residuals 77 34500609 448060 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey1 <- TukeyHSD(fit1, conf.level=0.95) #Tukey multiple comparison > Tukey1 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Biogas_production ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 694.00000 -133.36997 1521.3700 0.1599213 2016-2014 1541.66667 714.29669 2369.0366 0.0000055 2017-2014 1308.91667 481.54669 2136.2866 0.0001566 2018-2014 1358.50000 531.13003 2185.8700 0.0000784 2019-2014 1128.33333 300.96336 1955.7033 0.0017106 2020-2014 1334.41667 507.04669 2161.7866 0.0001099 2016-2015 847.66667 20.29669 1675.0366 0.0410426 2017-2015 614.91667 -212.45331 1442.2866 0.2817500 2018-2015 664.50000 -162.86997 1491.8700 0.1998249 2019-2015 434.33333 -393.03664 1261.7033 0.6893826 2020-2015 640.41667 -186.95331 1467.7866 0.2372841 2017-2016 -232.75000 -1060.11997 594.6200 0.9783861 2018-2016 -183.16667 -1010.53664 644.2033 0.9938444 2019-2016 -413.33333 -1240.70331 414.0366 0.7365206 2020-2016 -207.25000 -1034.61997 620.1200 0.9881153 2018-2017 49.58333 -777.78664 876.9533 0.9999968 2019-2017 -180.58333 -1007.95331 646.7866 0.9942997 192 2020-2017 25.50000 -801.86997 852.8700 0.9999999 2019-2018 -230.16667 -1057.53664 597.2033 0.9795748 2020-2018 -24.08333 -851.45331 803.2866 1.0000000 2020-2019 206.08333 -621.28664 1033.4533 0.9884622 Year Biogas_production sd 1 2014 2330.500 553.7258 2 2015 3024.500 1143.9877 3 2016 3872.167 794.9921 4 2017 3639.417 702.3386 5 2018 3689.000 199.2326 6 2019 3458.833 523.6508 7 2020 3664.917 286.1870 D.2 Methane Content Df Sum Sq Mean Sq F value Pr(>F) Year 6 268.5 44.75 10.79 1.11e-08 *** Residuals 77 319.3 4.15 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey2 <- TukeyHSD(fit2, conf.level=0.95) #Tukey multiple comparison > Tukey2 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Methane_content ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 -2.7500000 -5.26714655 -0.2328534 0.0231534 2016-2014 -1.1666667 -3.68381322 1.3504799 0.7984379 2017-2014 1.6666667 -0.85047989 4.1838132 0.4202415 2018-2014 2.2500000 -0.26714655 4.7671466 0.1105162 2019-2014 0.6666667 -1.85047989 3.1838132 0.9841179 2020-2014 2.5833333 0.06618678 5.1004799 0.0404567 2016-2015 1.5833333 -0.93381322 4.1004799 0.4837268 2017-2015 4.4166667 1.89952011 6.9338132 0.0000206 2018-2015 5.0000000 2.48285345 7.5171466 0.0000012 2019-2015 3.4166667 0.89952011 5.9338132 0.0018285 2020-2015 5.3333333 2.81618678 7.8504799 0.0000002 2017-2016 2.8333333 0.31618678 5.3504799 0.0172959 2018-2016 3.4166667 0.89952011 5.9338132 0.0018285 2019-2016 1.8333333 -0.68381322 4.3504799 0.3049407 2020-2016 3.7500000 1.23285345 6.2671466 0.0004418 2018-2017 0.5833333 -1.93381322 3.1004799 0.9921290 2019-2017 -1.0000000 -3.51714655 1.5171466 0.8910075 193 2020-2017 0.9166667 -1.60047989 3.4338132 0.9254758 2019-2018 -1.5833333 -4.10047989 0.9338132 0.4837268 2020-2018 0.3333333 -2.18381322 2.8504799 0.9996575 2020-2019 1.9166667 -0.60047989 4.4338132 0.2548637 Year Methane_content sd 1 2014 63.58333 2.9374799 2 2015 60.83333 2.3290003 3 2016 62.41667 2.2746961 4 2017 65.25000 1.5447860 5 2018 65.83333 0.9374369 6 2019 64.25000 0.9653073 7 2020 66.16667 2.3677121 D.3 H2S Content Df Sum Sq Mean Sq F value Pr(>F) Year 6 1130830 188472 0.979 0.445 Residuals 77 14823077 192507 > Tukey3 <- TukeyHSD(fit3, conf.level=0.95) #Tukey multiple comparison > Tukey3 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = H2S_content ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 71.83333 -470.4866 614.1532 0.9996570 2016-2014 307.66667 -234.6532 849.9866 0.6065461 2017-2014 61.16667 -481.1532 603.4866 0.9998653 2018-2014 27.66667 -514.6532 569.9866 0.9999988 2019-2014 160.83333 -381.4866 703.1532 0.9718399 2020-2014 292.00000 -250.3199 834.3199 0.6634742 2016-2015 235.83333 -306.4866 778.1532 0.8421465 2017-2015 -10.66667 -552.9866 531.6532 1.0000000 2018-2015 -44.16667 -586.4866 498.1532 0.9999802 2019-2015 89.00000 -453.3199 631.3199 0.9988299 2020-2015 220.16667 -322.1532 762.4866 0.8806089 2017-2016 -246.50000 -788.8199 295.8199 0.8126780 2018-2016 -280.00000 -822.3199 262.3199 0.7057319 2019-2016 -146.83333 -689.1532 395.4866 0.9822128 2020-2016 -15.66667 -557.9866 526.6532 1.0000000 2018-2017 -33.50000 -575.8199 508.8199 0.9999962 2019-2017 99.66667 -442.6532 641.9866 0.9977881 2020-2017 230.83333 -311.4866 773.1532 0.8550638 2019-2018 133.16667 -409.1532 675.4866 0.9892975 194 2020-2018 264.33333 -277.9866 806.6532 0.7581408 2020-2019 131.16667 -411.1532 673.4866 0.9901197 Year H2S_content sd 1 2014 359.8333 273.7334 2 2015 431.6667 291.7634 3 2016 667.5000 614.5482 4 2017 421.0000 214.4625 5 2018 387.5000 290.1907 6 2019 520.6667 512.9892 7 2020 651.8333 645.3405 D.4 Electricity > ## Anova > fit4 <- aov(Electricity~Year, metadata) > summary(fit4) Df Sum Sq Mean Sq F value Pr(>F) Year 6 58173817 9695636 7.397 3.61e-06 *** Residuals 71 93063092 1310748 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey4 <- TukeyHSD(fit4, conf.level=0.95) #Tukey multiple comparison > Tukey4 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Electricity ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 1550.33333 132.0368 2968.6299 0.0231496 2016-2014 2526.72619 874.4607 4178.9917 0.0003011 2017-2014 2114.03788 663.8655 3564.2102 0.0006559 2018-2014 2462.00000 1043.7035 3880.2965 0.0000284 2019-2014 1706.41667 288.1201 3124.7132 0.0085860 2020-2014 2637.08333 1218.7868 4055.3799 0.0000065 2016-2015 976.39286 -675.8726 2628.6583 0.5569849 2017-2015 563.70455 -886.4678 2013.8769 0.8995423 2018-2015 911.66667 -506.6299 2329.9632 0.4547983 2019-2015 156.08333 -1262.2132 1574.3799 0.9998810 2020-2015 1086.75000 -331.5465 2505.0465 0.2467925 2017-2016 -412.68831 -2092.3954 1267.0188 0.9890915 2018-2016 -64.72619 -1716.9917 1587.5393 0.9999997 2019-2016 -820.30952 -2472.5750 831.9559 0.7400160 2020-2016 110.35714 -1541.9083 1762.6226 0.9999938 2018-2017 347.96212 -1102.2102 1798.1345 0.9903727 195 2019-2017 -407.62121 -1857.7936 1042.5511 0.9781490 2020-2017 523.04545 -927.1269 1973.2178 0.9277426 2019-2018 -755.58333 -2173.8799 662.7132 0.6722154 2020-2018 175.08333 -1243.2132 1593.3799 0.9997677 2020-2019 930.66667 -487.6299 2348.9632 0.4293029 Year Electricity sd 1 2014 4880.417 728.5998 2 2015 6430.750 2041.8703 3 2016 7407.143 1080.8314 4 2017 6994.455 1037.2193 5 2018 7342.417 692.6983 6 2019 6586.833 952.5799 7 2020 7517.500 870.4768 D.5 Effluent TS > ## Anova > fit5 <- aov(Effluent_TS~Year, metadata) > summary(fit5) Df Sum Sq Mean Sq F value Pr(>F) Year 6 2.907e+09 484541468 10.62 3.69e-08 *** Residuals 64 2.919e+09 45609925 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey5 <- TukeyHSD(fit5, conf.level=0.95) #Tukey multiple comparison > Tukey5 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Effluent_TS ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 7997.682 -984.79256 16980.156 0.1122739 2016-2014 16843.182 8261.75241 25424.611 0.0000023 2017-2014 21274.293 12034.12035 30514.466 0.0000000 2018-2014 9827.932 1246.50241 18409.361 0.0147911 2019-2014 10598.882 1616.40744 19581.356 0.0108218 2020-2014 15052.896 5113.18934 24992.603 0.0003813 2016-2015 8845.500 43.05339 17647.947 0.0481143 2017-2015 13276.611 3830.82238 22722.400 0.0012007 2018-2015 1830.250 -6972.19661 10632.697 0.9954403 2019-2015 2601.200 -6592.65563 11795.056 0.9769386 2020-2015 7055.214 -3075.92119 17186.350 0.3538760 2017-2016 4431.111 -4634.15279 13496.375 0.7508227 2018-2016 -7015.250 -15408.05353 1377.554 0.1611800 196 2019-2016 -6244.300 -15046.74661 2558.147 0.3318569 2020-2016 -1790.286 -11567.60594 7987.035 0.9977393 2018-2017 -11446.361 -20511.62502 -2381.097 0.0049887 2019-2017 -10675.411 -20121.19984 -1229.622 0.0169013 2020-2017 -6221.397 -16581.69843 4138.905 0.5346432 2019-2018 770.950 -8031.49661 9573.397 0.9999681 2020-2018 5224.964 -4552.35594 15002.285 0.6658345 2020-2019 4454.014 -5677.12119 14585.150 0.8312859 Year Effluent_TS sd 1 2014 49084.82 5988.019 2 2015 57082.50 8729.659 3 2016 65928.00 7426.269 4 2017 70359.11 6288.440 5 2018 58912.75 4119.301 6 2019 59683.70 7100.303 7 2020 64137.71 7202.142 D.6 Effluent TN > ## Anova > fit6 <- aov(Effluent_TN~Year, metadata) > summary(fit6) Df Sum Sq Mean Sq F value Pr(>F) Year 5 9543738 1908748 4.072 0.004 ** Residuals 44 20623980 468727 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey6 <- TukeyHSD(fit6, conf.level=0.95) #Tukey multiple comparison > Tukey6 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Effluent_TN ~ Year, data = metadata) $Year diff lwr upr p adj 2016-2014 497.666667 -844.8832 1840.2166 0.8769966 2017-2014 9.777778 -1349.8754 1369.4309 1.0000000 2018-2014 -560.000000 -1876.4785 756.4785 0.8008652 2019-2014 -698.033333 -2040.5832 644.5166 0.6354059 2020-2014 -350.333333 -1792.4633 1091.7966 0.9779748 2017-2016 -487.888889 -1424.9651 449.1874 0.6340520 2018-2016 -1057.666667 -1930.9197 -184.4136 0.0095733 2019-2016 -1195.700000 -2107.7830 -283.6170 0.0040739 2020-2016 -848.000000 -1901.1828 205.1828 0.1790566 2018-2017 -569.777778 -1469.1038 329.5482 0.4232430 197 2019-2017 -707.811111 -1644.8874 229.2651 0.2364522 2020-2017 -360.111111 -1435.0113 714.7891 0.9161162 2019-2018 -138.033333 -1011.2864 735.2197 0.9969255 2020-2018 209.666667 -810.0732 1229.4065 0.9895575 2020-2019 347.700000 -705.4828 1400.8828 0.9207828 Year Effluent_TN sd 1 2014 3645.333 258.6220 2 2016 4143.000 653.1854 3 2017 3655.111 577.1333 4 2018 3085.333 581.5000 5 2019 2947.300 876.2909 6 2020 3295.000 819.1642 D.7 Effluent TP > ## Anova > fit7 <- aov(Effluent_TP~Year, metadata) > summary(fit7) Df Sum Sq Mean Sq F value Pr(>F) Year 6 4015385 669231 9.616 9.39e-07 *** Residuals 44 3062221 69596 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey7 <- TukeyHSD(fit7, conf.level=0.95) #Tukey multiple comparison > Tukey7 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Effluent_TP ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 -352.333333 -1293.2144 588.54775 0.9062776 2016-2014 -628.033333 -1164.4181 -91.64859 0.0125302 2017-2014 -792.000000 -1335.2179 -248.78205 0.0009006 2018-2014 -724.250000 -1250.2185 -198.28149 0.0019504 2019-2014 -794.733333 -1331.1181 -258.34859 0.0007161 2020-2014 -27.000000 -603.1696 549.16964 0.9999991 2016-2015 -275.700000 -1130.2977 578.89768 0.9521294 2017-2015 -439.666667 -1298.5697 419.23632 0.6946584 2018-2015 -371.916667 -1220.0154 476.18208 0.8221838 2019-2015 -442.400000 -1296.9977 412.19768 0.6836316 2020-2015 325.333333 -554.7803 1205.44699 0.9113842 2017-2016 -163.966667 -538.3538 210.42047 0.8230582 2018-2016 -96.216667 -445.1047 252.67138 0.9776756 2019-2016 -166.700000 -531.1017 197.70168 0.7919737 198 2020-2016 601.033333 180.2585 1021.80814 0.0011970 2018-2017 67.750000 -291.5549 427.05490 0.9970225 2019-2017 -2.733333 -377.1205 371.65380 1.0000000 2020-2017 765.000000 335.5485 1194.45149 0.0000356 2019-2018 -70.483333 -419.3714 278.40471 0.9956522 2020-2018 697.250000 289.8365 1104.66346 0.0000726 2020-2019 767.733333 346.9585 1188.50814 0.0000228 Year Effluent_TP sd 1 2014 1168.3333 51.98397 2 2015 816.0000 NA 3 2016 540.3000 129.16918 4 2017 376.3333 78.57640 5 2018 444.0833 325.76915 6 2019 373.6000 77.72916 7 2020 1141.3333 571.92715 *group with fewer than 2 data points will be deleted. 2015 only has 1 data point so it was removed D.8 Effluent pH > ## Anova > fit8 <- aov(Effluent_pH~Year, metadata) > summary(fit8) Df Sum Sq Mean Sq F value Pr(>F) Year 6 0.333 0.05550 2.58 0.0266 * Residuals 64 1.377 0.02151 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey8 <- TukeyHSD(fit8, conf.level=0.95) #Tukey multiple comparison > Tukey8 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Effluent_pH ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 -0.031818182 -0.2269049 0.163268578 0.9988202 2016-2014 -0.056818182 -0.2431948 0.129558443 0.9666104 2017-2014 -0.048484848 -0.2491685 0.152198755 0.9897882 2018-2014 -0.065151515 -0.2515281 0.121225110 0.9361855 2019-2014 -0.191818182 -0.3869049 0.003268578 0.0569543 2020-2014 -0.197532468 -0.4134090 0.018344039 0.0944586 2016-2015 -0.025000000 -0.2161768 0.166176809 0.9996669 2017-2015 -0.016666667 -0.2218160 0.188482632 0.9999795 199 2018-2015 -0.033333333 -0.2245101 0.157843476 0.9982809 2019-2015 -0.160000000 -0.3596777 0.039677665 0.1996821 2020-2015 -0.165714286 -0.3857484 0.054319784 0.2635888 2017-2016 0.008333333 -0.1885515 0.205218166 0.9999996 2018-2016 -0.008333333 -0.1906133 0.173946602 0.9999993 2019-2016 -0.135000000 -0.3261768 0.056176809 0.3372597 2020-2016 -0.140714286 -0.3530640 0.071635413 0.4143088 2018-2017 -0.016666667 -0.2135515 0.180218166 0.9999739 2019-2017 -0.143333333 -0.3484826 0.061815965 0.3499429 2020-2017 -0.149047619 -0.3740589 0.075963618 0.4147777 2019-2018 -0.126666667 -0.3178435 0.064510142 0.4144769 2020-2018 -0.132380952 -0.3447307 0.079968746 0.4893366 2020-2019 -0.005714286 -0.2257484 0.214319784 1.0000000 Year Effluent_pH sd 1 2014 7.881818 0.1250454 2 2015 7.850000 0.1269296 3 2016 7.825000 0.1959824 4 2017 7.833333 0.1000000 5 2018 7.816667 0.1267304 6 2019 7.690000 0.1523884 7 2020 7.684286 0.1767161 D.9 Effluent VFA > ## Anova > fit9 <- aov(Effluent_VFA~Year, metadata) > summary(fit9) Df Sum Sq Mean Sq F value Pr(>F) Year 6 66702800 11117133 7.911 2.26e-06 *** Residuals 63 88531222 1405257 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 > Tukey9 <- TukeyHSD(fit9, conf.level=0.95) #Tukey multiple comparison > Tukey9 #Output Tukey results Tukey multiple comparisons of means 95% family-wise confidence level Fit: aov(formula = Effluent_VFA ~ Year, data = metadata) $Year diff lwr upr p adj 2015-2014 278.85455 -1298.6290 1856.3381 0.9981357 2016-2014 2483.37121 976.3183 3990.4241 0.0000896 2017-2014 532.89899 -1089.8410 2155.6390 0.9521232 2018-2014 -484.87879 -1991.9317 1022.1741 0.9565537 2019-2014 -114.54545 -1692.0290 1462.9381 0.9999894 200 2020-2014 76.78788 -1755.5436 1909.1193 0.9999996 2016-2015 2204.51667 658.6492 3750.3841 0.0009793 2017-2015 254.04444 -1404.8055 1912.8943 0.9991716 2018-2015 -763.73333 -2309.6008 782.1341 0.7409955 2019-2015 -393.40000 -2008.0059 1221.2059 0.9892838 2020-2015 -202.06667 -2066.4530 1662.3197 0.9998878 2017-2016 -1950.47222 -3542.4951 -358.4493 0.0071468 2018-2016 -2968.25000 -4442.1768 -1494.3232 0.0000013 2019-2016 -2597.91667 -4143.7841 -1052.0492 0.0000621 2020-2016 -2406.58333 -4211.7676 -601.3990 0.0025172 2018-2017 -1017.77778 -2609.8007 574.2451 0.4579787 2019-2017 -647.44444 -2306.2943 1011.4055 0.8958544 2020-2017 -456.11111 -2358.9424 1446.7202 0.9901690 2019-2018 370.33333 -1175.5341 1916.2008 0.9901989 2020-2018 561.66667 -1243.5176 2366.8510 0.9630106 2020-2019 191.33333 -1673.0530 2055.7197 0.9999185 Year Effluent_VFA sd 1 2014 1301.5455 918.1749 2 2015 1580.4000 1312.9674 3 2016 3784.9167 1859.0623 4 2017 1834.4444 1103.4844 5 2018 816.6667 597.9409 6 2019 1187.0000 1039.4021 7 2020 1378.3333 796.3772 E. Other SCAD laboratory results E.1 Total Suspended Solids (TSS) Table E.1 Summary of TSS results Filtrate Effluent Year (mg/L) (mg/L) 2014 22,972 33,604 2015 39,000 2016 23,862 35,556 2017 26,633 33,378 2018 33,172 32,819 2019 26,995 33,328 2020 49,236 48,833 Max 49,236 48,833 Min 22,972 32,819 Mean 29,384 36,302 Average 30,478 36,645 St. Dev 9,001 5,356 201 E.2 Volatile Suspended Solids (VSS) Table E.2 Summary of VSS results Filtrate Effluent Year (mg/L) (mg/L) 2014 18,556 27,247 2015 33,333 2016 19,795 30,828 2017 20,567 28,311 2018 22,901 27,587 2019 22,858 29,650 2020 34,542 42,389 Max 34,542 42,389 Min 18,556 27,247 Mean 22,694 30,995 Average 23,203 31,335 St. Dev 5,306 4,916 E.3 Ammonia Table E.3 Summary of ammonia results Filtrate Effluent Year (mg/L) (mg/L) 2014 1,942 1,850 2015 1,825 1,566 2016 2,368 2,493 2017 1,917 2,029 2018 1,829 1,845 2019 1,875 1,755 2020 1,609 1,547 Max 2,368 2,493 Min 1,609 1,547 Mean 1,898 1,847 Average 1,909 1,869 St. Dev 213 299 202 E.4 Alkalinity Table E.4 Summary of alkalinity results Filtrate Effluent Year (mg/L) (mg/L) 2014 1,942 1,850 2015 1,825 1,566 2016 2,368 2,493 2017 1,917 2,029 2018 1,829 1,845 2019 1,875 1,755 2020 1,609 1,547 Max 2,368 2,493 Min 1,609 1,547 Mean 1,898 1,847 Average 1,909 1,869 St. Dev 213 299 E.5 Total Dissolved Solids (TDS) Table E.5 Summary of TDS results Filtrate Effluent Year (mg/L) (mg/L) 2014 17,667 17,653 2015 20,317 2016 17,535 30,988 2017 14,070 34,948 2018 11,299 26,094 2019 13,515 26,374 2020 6,379 16,871 Max 17,667 34,948 Min 6,379 16,871 Mean 12,725 23,942 Average 13,411 24,749 St. Dev 3,862 6,318 203 F. A&L Great Lakes Laboratory Results F.1 Solids Table F.1 Summary of Solids results Filtrate Effluent Solids Year (%) (%) (%) 2014 24.74 2015 39.35 2016 26.23 2017 3.87 2018 5.24 3.87 27.48 2019 28.25 2020 31.25 Max 5.24 3.87 39.35 Min 3.87 3.87 24.74 Mean 4.50 3.87 29.20 Average 4.56 3.87 29.55 St. Dev 0.69 - 4.81 F.2 Total Kjeldahl Nitrogen (TKN) Table F.2 Summary of TKN results Filtrate Effluent Solids Year (%) (%) (%) 2014 0.43 2015 0.72 2016 0.63 2017 0.29 2018 0.311 0.311 0.45 2019 0.55 2020 0.59 Max 0.31 0.31 0.72 Min 0.29 0.31 0.43 Mean 0.30 0.31 0.55 Average 0.30 0.31 0.56 St. Dev 0.01 - 0.10 204 F.3 Phosphorus (P) Table F.3 Summary of Phosphorus results Filtrate Effluent Solids Year (%) (%) (%) 2014 0.14 2015 0.35 2016 0.23 2017 0.03 2018 0.037 0.033 0.147 2019 0.19 2020 0.18 Max 0.04 0.03 0.35 Min 0.03 0.03 0.14 Mean 0.03 0.03 0.20 Average 0.03 0.03 0.21 St. Dev 0.00 - 0.07 F.4 Potassium (K) Table F.4 Summary of Potassium results Filtrate Effluent Solids Year (%) (%) (%) 2014 0.26 2015 0.65 2016 0.31 2017 0.14 2018 0.144 0.159 0.143 2019 0.22 2020 0.19 Max 0.14 0.16 0.65 Min 0.14 0.16 0.14 Mean 0.14 0.16 0.26 Average 0.14 0.16 0.30 St. Dev 0.00 - 0.17 205 F.5 Moisture Table F.5 Summary of Moisture results Filtrate Effluent Year (%) (%) 2017 96.13 2018 94.76 0.159 Max 96.13 0.16 Min 94.76 0.16 Mean 95.44 0.16 Average 95.45 0.16 St. Dev 0.68 - F.6 Sulphur (S) Table F.6 Summary of Sulphur results Filtrate Effluent Year (%) (%) 2017 0.03 2018 0.03 0.02 Max 0.03 0.02 Min 0.03 0.02 Mean 0.03 0.02 Average 0.03 0.02 St. Dev - - F.7 Magnesium (Mg) Table F.7 Summary of Magnesium results Filtrate Effluent Year (%) (%) 2017 0.05 2018 0.05 0.07 Max 0.05 0.07 Min 0.05 0.07 Mean 0.05 0.07 Average 0.05 0.07 St. Dev - - 206 F.8 Calcium (Ca) Table F.8 Summary of Calcium results Filtrate Effluent Year (%) (%) 2017 0.11 2018 0.12 0.13 Max 0.12 0.13 Min 0.11 0.13 Mean 0.11 0.13 Average 0.12 0.13 St. Dev 0.01 - F.9 Natrium (Na) Table F.9 Summary of Natrium results Filtrate Effluent Year (%) (%) 2017 0.07 2018 0.08 0.09 Max 0.08 0.09 Min 0.07 0.09 Mean 0.07 0.09 Average 0.08 0.09 St. Dev 0.01 - F.10 Aluminum (Al) Table F.10 Summary of Aluminum results Filtrate Effluent Year (ppm) (ppm) 2017 48.5 2018 44 67 Max 48.50 67.00 Min 44.00 67.00 Mean 46.20 67.00 Average 46.25 67.00 St. Dev 2.25 - 207 F.11 Copper (Cu) Table F.11 Summary of Copper results Filtrate Effluent Year (ppm) (ppm) 2017 4.35 2018 4.3 4.9 Max 4.35 4.90 Min 4.30 4.90 Mean 4.32 4.90 Average 4.33 4.90 St. Dev 0.02 - F.12 Iron (Fe) Table F.12 Summary of Iron results Filtrate Effluent Year (ppm) (ppm) 2017 90.5 2018 106 162 Max 106.00 162.00 Min 90.50 162.00 Mean 97.94 162.00 Average 98.25 162.00 St. Dev 7.75 - F.13 Manganese (Mn) Table F.13 Summary of Manganese results Filtrate Effluent Year (ppm) (ppm) 2017 9.65 2018 11 12 Max 11.00 12.00 Min 9.65 12.00 Mean 10.30 12.00 Average 10.33 12.00 St. Dev 0.68 - 208 F.14 Zinc (Zn) Table F.14 Summary of Zinc results Filtrate Effluent Year (ppm) (ppm) 2017 61 2018 53 57 Max 61.00 57.00 Min 53.00 57.00 Mean 56.86 57.00 Average 57.00 57.00 St. Dev 4.00 - F.15 Nitrogen (N) Table F.15 Summary of Nitrogen results Solids Year (%) 2017 0.28 2018 0.45 Max 0.45 Min 0.28 Mean 0.35 Average 0.37 St. Dev 0.09 F.16 Ammonia (NH3) Table F.16 Summary of Ammonia results Solids Year (%) 2017 0.18 2018 0.14 Max 0.18 Min 0.14 Mean 0.16 Average 0.16 St. Dev 0.02 209 G. Formula for OLR and HRT Calculation G.1 OLR Formula OLR = IF (Total VS g=0, “”, (Total VS g / (Working Volume Liters*Days))) G.2 HRT Formula HRT = IF (Volume of Digestate Wasted=0, “”, ((Working Volume Gallons/Volume of Digestate Wasted)) * Days) H. R-Code of MLR for feedstock impact on biogas production (after data treated) RegressionData1 <- read.table("C:/BIOGAS-AHAY/MLR_ORIGINAL DATA - REGRESSION.csv", header = TRUE, sep=',') validationData <- read.table("C:/BIOGAS-AHAY/MLR_ORIGINAL DATA - VALIDATION.csv", header = TRUE, sep=',') #BL2020a <- read.table("C:/MSU/Biogas data/BiogasProduction_REMOVED.csv", header = TRUE, sep=',') names(RegressionData1) = c('row.name','Year','Mon','Digestate','FMP','SLSS','DG','Parlor','Beef','WFM','Poultry','Swine','A NS','TMP','FFP','SLSSF','PA','Pulp','FOG','WFEED','Other','CartF','TFP','TFS','BiogasP','CH4P',' MethaneC','H2SC','Elec','TS','TN','TP','pH','VFA') summary(RegressionData1) RegressionData <- RegressionData1 #BL2020[138,] RegressionData$Year = as.factor(RegressionData$Year) RegressionData$Mon = as.factor(RegressionData$Mon) row.name <- RegressionData[1] Year <- RegressionData[2] MoN <- RegressionData[3] Digestate <- RegressionData[4] FMP <- RegressionData[5] SLSS <- RegressionData[6] DG <- RegressionData[7] Parlor <- RegressionData[8] Beef <- RegressionData[9] WFM <- RegressionData[10] Poultry <- RegressionData[11] Swine <- RegressionData[12] ANS <- RegressionData[13] 210 TMP <- RegressionData[14] FFP <- RegressionData[15] SLSSF <- RegressionData[16] PA <- RegressionData[17] Pulp <- RegressionData[18] FOG <- RegressionData[19] WFEED <- RegressionData[20] Other <- RegressionData[21] CartF <- RegressionData[22] TFP <- RegressionData[23] TFS <- RegressionData[24] BiogasP <- RegressionData[25] CH4 <- RegressionData[26] MethaneC <- RegressionData[27] H2SC <- RegressionData[28] Elec <- RegressionData[29] TS <- RegressionData[30] TN <- RegressionData[31] TP <- RegressionData[32] pH <- RegressionData[33] VFA <- RegressionData[34] library(tidyverse) library(GGally) # prints first set of rows showing works as a visual check if data were read correctly head(RegressionData) #shows the data type and sample of the data str(RegressionData) ############################################################################## ########################## # Standardize variables in a dataset RegressionData_std <- data.frame(scale(RegressionData[,c('Digestate','FMP','SLSS','DG','Parlor','Beef','WFM','Poultry',' Swine','ANS','TMP','FFP','SLSSF','PA','Pulp','FOG','WFEED','Other','CartF','TFP','TFS','BiogasP ','CH4P','MethaneC','H2SC','Elec','TS','TN','TP','pH','VFA')])) RegressionData_std["Year"] <- RegressionData[,"Year"] RegressionData_std["Month"] <- RegressionData[,"Mon"] head(RegressionData_std) ############################################################################## ########################## #outliers testing install.packages("e1071") install.packages("outliers") 211 library(e1071) library(outliers) #histograms of weights hist(RegressionData$FMP, breaks= 12, freq = F, xlab = 'FMP', main="") hist(sqrt(RegressionData$FMP), breaks= 12, freq = F, xlab = 'SQRT(FMP)', main="") ggpairs(columns = c('FMP','DG','Parlor','Beef','TMP','Pulp','FOG','Other','CartF'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8))) ggpairs(columns = c('FFP','Pulp','FOG','Other','CartF','TFP','TFS'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8))) ggpairs(columns = c('BiogasP','CH4P','MethaneC','H2SC','Elec','TS','TN','TP','pH','VFA'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8))) ggpairs(columns = c('BiogasP','TMP','TFP','FMP','MethaneC','H2SC','TS','TN','TP','pH','VFA'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8))) ggpairs(columns = c('BiogasP', 'sqrtFMP'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8))) ggpairs(columns = c('BiogasP', 'TFP'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8)),aes(color = Mon, alpha = 0.5)) ggpairs(columns = c('BiogasP', 'TMP'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8)),aes(color = Mon, alpha = 0.5)) ggpairs(columns = c('BiogasP', 'TFS'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8)),aes(color = Mon, alpha = 0.5)) ggpairs(columns = c('BiogasP', 'TFS'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8)),aes(color = Year, alpha = 0.5)) ggpairs(columns = c('BiogasP', 'TMP'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8)),aes(color = Year, alpha = 0.5)) ggpairs(columns = c('BiogasP', 'TFP'), data = RegressionData, upper = list(continuous = wrap('cor', size = 8)),aes(color = Year, alpha = 0.5)) #Comparisons between the different Months boxplot(RegressionData$BiogasP ~RegressionData$Mon, ylab ="Biogas Production", xlab ="Mon") boxplot(RegressionData$BiogasP ~RegressionData$Year, ylab ="Biogas Production", xlab ="Year") #Analysis of variance 212 #Conduct H0 : Mean Biogas production is the same for Months # Ha: at least one Month is different aov_Months <- aov(RegressionData$BiogasP~RegressionData$Mon) summary(aov_Months) #attributes(aov_species) aov_Months$coefficients TukeyHSD(aov_Months) plot(TukeyHSD(aov_Months), las=1) ############################################################################## ########################## #Specifying the preliminary full model / parameters we wish to investigate #Full model for original data #'Year','Mon','Digestate','FMP','SLSS','DG','Parlor','Beef','WFM','Poultry','Swine','ANS','TMP','F FP','SLSSF','PA','Pulp','FOG','WFEED','Other','CartF','TFP','TFS','BiogasP','CH4P','MethaneC',' H2SC','Elec','TS','TN','TP','pH','VFA' full_model <- lm(BiogasP ~ FMP + DG + Digestate + SLSS + SLSSF + WFM + Poultry + Parlor + Beef + FFP + Swine + ANS + PA + WFEED + Pulp + FOG + Other + CartF, data=RegressionData) #Full model for standardized data full_model_std <- lm(BiogasP ~ FMP + DG + PA + Beef + FFP + Pulp + FOG + Other + CartF , data=RegressionData_std) modelDFFITS <- dffits(multiple.regression) DFITTSThreshold <- sqrt(4*nparameters/ndata) # not a statistical test but to provide a general judgment which(abs(modelDFFITS)>DFITTSThreshold) RegressionData[index(RegressionData)==62,] RegressionData_noinfluence <- RegressionData[-c(49,50,71,72),] full_model <- lm(BiogasP ~ FMP + DG + Digestate + SLSS + SLSSF + WFM + Poultry + Parlor + Beef + FFP + Swine + ANS + PA + WFEED + Pulp + FOG + Other + CartF, data=RegressionData_noinfluence) Bothfit.p <- ols_step_both_p(full_model, pent = .01, prem = .05, details = TRUE) Bothfit.p 213 multiple.regression <- lm(BiogasP ~ FMP + DG + Parlor + FOG, data=RegressionData_noinfluence) summary(multiple.regression) RESULT Figure H.1 Correlation plot between the parameters part 1 (Note: FMP = filtrate manure pit, DG = Dairy Gutter, TMP = Total Manure Pit, CartF = Cart Food) 214 Figure H.2 Correlation plot between the parameters part 2 (Note: FFP = Filtrate Food Pit, TFP = Total Food Pit, TFS = Total Feedstock) Figure H.3 Correlation plot between the operational parameters (Note: BiogasP = Biogas Production, CH4P = Methane production, MethaneC = Methane Concentration, H2SC = hydrogen sulfide concentration, Elec = Electricity) 215 Figure H.4 Correlation plot between the total feedstock and the operational parameters Figure H.5 Correlation plot between the total manure pit and the biogas production 216 Figure H.6 Correlation plot between the total food pit and the biogas production Figure H.7 Correlation plot between the total feedstock and the biogas production 217 Figure H.8 The data distribution of the biogas production from 2014 to 2020 Figure H.9 Linear model plot of biogas production (residuals vs fitted) 218 I. R-Code of MLR for operational parameters impact on biogas production # Load libraries ----- library(dplyr) library(FSA) library(psych) library(car) library(rcompanion) # Choose data file of "MLR_operational_parameters.txt" ----- con <-file.choose(new = FALSE) metadata <- read.table(con, header = T, row.names = 1, fill = TRUE) head(metadata) ############################# # select data data.num = select(metadata, OLR, HRT, Temperature, pH, Biogas, Methane) head(data.num) # corr.test(data.num, use = "pairwise", method="pearson", adjust="none", alpha="0.05") ########################################### # Correlations # ########################################### library(psych) pairs.panels(data.num, method = "pearson", # correlation method hist.col = "#00AFBB", density = TRUE, # show density plots ellipses = TRUE # show correlation ellipses ) summary(data.num$Methane) plot(data.num$Methane) hist(data.num$Methane) model1<-lm(Methane~OLR+HRT+Temperature+pH,data=data.num) summary(model1) model2<-lm(Biogas~OLR+HRT+Temperature+pH,data=data.num) summary(model2) 219 # Scatterplots of variables # regression line : Green # non-parametric mean : Red # non-parametric variance: Blue scatterplotMatrix(data.num, diagonal=FALSE, regLine = list(col="green",lwd=3), smooth=list(col.smooth="red", col.spread="blue") ) RESULTS Figure I.1 Correlation plot between the operational parameters and the dependent variables (biogas production and methane concentration) 220 Figure I.2 Scatter plots of variables (Note: green = regression line, red = non-parametric mean, blue = non-parametric variance) 221 REFERENCES 222 REFERENCES Abolhosseini, S., Heshmati, A., & Altmann, J. (2014). A Review of Renewable Energy Supply and Energy Efficiency Technologies. America’s Oil and Natural Gas Industry. (2017). Renewable Fuel Standard Fill Up on Facts. Anukam, A., Mohammadi, A., Naqvi, M., & Granström, K. (2019). A Review of the Chemistry of Anaerobic Digestion: Methods of Accelerating and Optimizing Process Efficiency. Processes, 7(8), 504. https://doi.org/10.3390/pr7080504 Argonne National Laboratory. (2014). Greenhouse Gases, Regulated Emissions, and Energy use in Transportation (GREET) Model. Arosemena, M. I. B. (2021). Exploring The Impact of Biogas Quantity and Quality in Different Digester Types with Variations in Temperature. Azapagic, A., Millington, A., & Collett, A. (2006). A Methodology for Integrating Sustainability Considerations into Process Design. Chemical Engineering Research and Design, 84(6), 439–452. https://doi.org/10.1205/cherd05007 Barbose, G. (2019). U.S. Renewable Portfolio Standards: 2019 Annual Status Updates. Bradley, A. L. (2019). Manure Management for Small and Hobby Farms. Center for Law, E. and the E. (2019). California Climate Policy Fact Sheet: Low Carbon Fuel Standard. Chandra, K. (2005). Organic Manures. Chen, H., Wang, W., Xue, L., Chen, C., Liu, G., & Zhang, R. (2016). Effects of Ammonia on Anaerobic Digestion of Food Waste: Process Performance and Microbial Community. Energy & Fuels, 30(7), 5749–5757. https://doi.org/10.1021/acs.energyfuels.6b00715 Chen, L., & Neibling, H. (2014). Anaerobic Digestion Basics. Chen, R., Rojas-Downing, M. M., Zhong, Y., Saffron, C. M., & Liao, W. (2015). Life Cycle and Economic Assessment of Anaerobic Co-digestion of Dairy Manure and Food Waste. Industrial Biotechnology, 11(2), 127–139. https://doi.org/10.1089/ind.2014.0029 Chen, R., Roos, M. M., Zhong, Y., Marsh, T., Roman, M. B., Hernandez Ascencio, W., Uribe, L., Uribe Lorio, L., Kirk, D., Reinhold, D. M., Miranda Chavarria, J. A., Baudrit Ruiz, D., Pereira, J. F. A., Montero, W. R., Srivastava, A., & Liao, W. (2016). Responses of anaerobic microorganisms to different culture conditions and corresponding effects on 223 biogas production and solid digestate quality. Biomass and Bioenergy, 85, 84–93. https://doi.org/10.1016/j.biombioe.2015.11.028 Christensen, A., Searle, S., & Malins, C. (2014). A Conversational Guide to Renewable Identification Numbers (RINs) in the U.S. Renewable Fuel Standard. Chynoweth, D. P., Owens, J. M., & Legrand, R. (2001). Renewable methane from anaerobic digestion of biomass. Renewable Energy, 22(1–3), 1–8. https://doi.org/10.1016/S0960- 1481(00)00019-7 Congressional Research Service. (2020). The Renewable Fuel Standard (RFS): An Overview. Congressional Research Service. (2021). A Low Carbon Fuel Standard: In Brief. Cooper, G. (2018). Understanding the RFS and RINs. Cory, K. S., & Swezey, B. G. (2007). Renewable Portfolio Standards in the States: Balancing Goals and Implementation Strategies. Courneya, J. (2010). Manure Management Plan: A Step-by-Step Guide for Minnesota Feedlot Operators. Department of Environmental Affairs and Tourism of South Africa. (2004). Life Cycle Assessment, Integrated Environmental Management, Information Series 9. Durham, J. L., & Demerjian, K. L. (1985). Atmospheric Acidification Chemistry: A Review. In Acid Deposition: Environmental, Economic, and Policy Issues (pp. 17–30). Springer US. https://doi.org/10.1007/978-1-4615-8350-9_2 Eckert, M., Bell, M., Potterton, S., Craigon, J., Saunders, N., Wilcox, R., Hunter, M., Goodman, J., & Garnsworthy, P. (2018). Effect of Feeding System on Enteric Methane Emissions from Individual Dairy Cows on Commercial Farms. Land, 7(1), 26. https://doi.org/10.3390/land7010026 Environmental Protection Agency. (2015). Energy and Environment Guide to Action. Environmental Protection Agency. (2020). Documentation for greenhouse gas emission and energy factors used in the waste reduction model (WARM): Organic materials chapters. Environmental Protection Agency (EPA). (2019, September 13). Global Greenhouse Gas Emissions Data. Https://Www.Epa.Gov/Ghgemissions/Global-Greenhouse-Gas-Emissions- Data. Esfandiari, S., Khosrokhavar, R., & Sekhavat, M. (2011). Greenhouse gas emissions reduction through a biogas plant: A case study of waste management systems at FEKA dairy farm. 2nd International Conference on Environmental Science and Technology, 6, 445–448. 224 Eshkaftaki, M. M., & Ebrahemi, R. (2019). Assess a new strategy and develop a new mixer to improve anaerobic microbial activities and clean biogas production. Journal of Cleaner Production, 797–807. Field, J. A., Caldwell, J. S., Jeyanayagam, S., Reneau, R. B., Jr., Kroontje, W., Collins, E. R., & Jr. (1984). Fertilizer Recovery from Anaerobic Digesters. Transactions of the ASAE, 27(6), 1871–1876. https://doi.org/10.13031/2013.33060 Fu, S.-F., Chen, K.-Q., Zhu, R., Sun, W.-X., Zou, H., & Guo, R.-B. (2018). Improved anaerobic digestion performance of Miscanthus floridulus by different pretreatment methods and preliminary economic analysis. Energy Conversion and Management, 159, 121–128. https://doi.org/10.1016/j.enconman.2018.01.014 Goldstein, N. (2018, July 6). Co-digestion of Food Waste In The U.S. BioCycle. https://www.biocycle.net/codigestion-food-waste-u-s/ Gorjian, S. (2017). An Introduction to Renewable Energy Sources. Tarbiat Modares University. Granholm, J. M., & Chester, S. E. (2007). Michigan Solid Waste Policy. Greenwood, J., & Guner, N. (2008). Social Change. The Institute for the Study of Labor. http://ideas.repec.org/p/eag/rereps/9.html Guinee, J. B. (2002). Handbook on life cycle assessment operational guide to the ISO standards. The International Journal of Life Cycle Assessment, 7(5), 311. https://doi.org/10.1007/BF02978897 Hach. (2018). What is pH and how is it measured? (1). Hosseini, S. E., & Wahid, M. A. (2014). Development of biogas combustion in combined heat and power generation. Renewable and Sustainable Energy Reviews, 40, 868–875. https://doi.org/10.1016/j.rser.2014.07.204 Ingram, W. (2015). Staff Report: Calculating Life Cycle Carbon Intensity of Transportation Fuels in California. International Atomic Energy Agency. (2008). Guidelines for Sustainable Manure Management in Asian Livestock Production Systems. Jensen, A. A., Hoffman, L., Møller, B. T., & Schmidt, A. (1997). Life Cycle Assessment. Jiang, J., He, S., Kang, X., Sun, Y., Yuan, Z., Xing, T., Guo, Y., & Li, L. (2020). Effect of Organic Loading Rate and Temperature on the Anaerobic Digestion of Municipal Solid Waste: Process Performance and Energy Recovery. Frontiers in Energy Research, 8. https://doi.org/10.3389/fenrg.2020.00089 225 Jiang, Y., McAdam, E., Zhang, Y., Heaven, S., Banks, C., & Longhurst, P. (2019). Ammonia inhibition and toxicity in anaerobic digestion: A critical review. Journal of Water Process Engineering, 32, 100899. https://doi.org/10.1016/j.jwpe.2019.100899 Jürgensen, L., Ehimen, E. A., Born, J., & Holm-Nielsen, J. B. (2018). A combination anaerobic digestion scheme for biogas production from dairy effluent—CSTR and ABR, and biogas upgrading. Biomass and Bioenergy, 111, 241–247. https://doi.org/10.1016/j.biombioe.2017.04.007 Khasreen, M., Banfill, P. F., & Menzies, G. (2009). Life-Cycle Assessment and the Environmental Impact of Buildings: A Review. Sustainability, 1(3), 674–701. https://doi.org/10.3390/su1030674 Kostic, B., Stevanovic, G., Lutovac, M., Lutovac, B., Ketin, S., & Biocanin, R. (2020). Animal Manure and Environment. Fresenius Environmental Bulletin, 29(3), 1289–1296. Krakat, N., Demirel, B., Anjum, R., & Dietz, D. (2017). Methods of ammonia removal in anaerobic digestion: a review. Water Science and Technology, 76(8), 1925–1938. https://doi.org/10.2166/wst.2017.406 Kress, P., Nägele, H.-J., Oechsner, H., & Ruile, S. (2018). Effect of agitation time on nutrient distribution in full-scale CSTR biogas digesters. Bioresource Technology, 247, 1–6. https://doi.org/10.1016/j.biortech.2017.09.054 Li, Y., Xu, F., Li, Y., Lu, J., Li, S., Shah, A., Zhang, X., Zhang, H., Gong, X., & Li, G. (2018). Reactor performance and energy analysis of solid state anaerobic co-digestion of dairy manure with corn stover and tomato residues. Waste Management, 73, 130–139. https://doi.org/10.1016/j.wasman.2017.11.041 Lorimor, J., & Powers, W. (2018). Manure Characteristics. In Manure Management Systems (2nd ed.). MidWest Plan Service. Lossie, U., & Pütz, P. (2011). Targeted Control of Biogas Plants with the Help of FOS/TAC. Lukitawesa, Patinvoh, R. J., Millati, R., Sárvári-Horváth, I., & Taherzadeh, M. J. (2020). Factors influencing volatile fatty acids production from food wastes via anaerobic digestion. Bioengineered, 11(1), 39–52. https://doi.org/10.1080/21655979.2019.1703544 Magdalena, J. A., Greses, S., & González-Fernández, C. (2019). Impact of Organic Loading Rate in Volatile Fatty Acids Production and Population Dynamics Using Microalgae Biomass as Substrate. Scientific Reports, 9. Manitoba. (2015). Properties of Manure. 226 Marjolaine. (2019). What is the future of small-scale anaerobic digestion? . Biogas World. https://www.biogasworld.com/news/future-small-scale-anaerobic-digestion/ Mayer, F., Guignard, C., Noo, A., & Hoffman, L. (2010). Monitoring volatile fatty acid production during mesophilic anaerobic digestion exposed to increasing feeding rates. Third International Symposium on Energy from Biomass and Waste. McPhail, L., Westcott, P., & Lutman, H. (2011). The Renewable Identification Number System and U.S. Biofuel Mandates. Meegoda, J., Li, B., Patel, K., & Wang, L. (2018). A Review of the Processes, Parameters, and Optimization of Anaerobic Digestion. International Journal of Environmental Research and Public Health, 15(10), 2224. https://doi.org/10.3390/ijerph15102224 Michigan Public Service Commission. (2021). Renewable Energy. Https://Www.Michigan.Gov/Mpsc/Consumer/Electricity/Renewable-Energy. https://www.michigan.gov/mpsc/consumer/electricity/renewable-energy Michigan State University Sustainability. (2014). It all adds up: Conserving energy at MSU. Https://Sustainability.Msu.Edu/Discover/Discover-E-Conserving-Energy.Html. Moguel-Castañeda, J. G., González-Salomón, M., Hernández-García, H., Morales-Zarate, E., Puebla, H., & Hernandez-Martinez, E. (2020). Effect of organic loading rate on anaerobic digestion of raw cheese whey: experimental evaluation and mathematical modeling. International Journal of Chemical Reactor Engineering, 18(7). https://doi.org/10.1515/ijcre-2020-0022 Mohammadrezaei, R., Zareei, S., & Behroozi- Khazaei, N. (2018). Optimum mixing rate in biogas reactors: Energy balance calculations and computational fluid dynamics simulation. Energy, 159, 54–60. https://doi.org/10.1016/j.energy.2018.06.132 Morozova, I., Nikulina, N., Oechsner, H., Krümpel, J., & Lemmer, A. (2020). Effects of Increasing Nitrogen Content on Process Stability and Reactor Performance in Anaerobic Digestion. Energies, 13(5), 1139. https://doi.org/10.3390/en13051139 MSU Anaerobic Digestion and Research Center. (2019). MSU Anaerobic Digestion Operator Training. Michigan State University. Mueller, D. K., & Helsel, D. R. (1996). Nutrients in the Nation’s Waters--Too Much of A Good Thing? Musa, M., Idrus, S., Hasfalina, C., & Daud, N. (2018). Effect of Organic Loading Rate on Anaerobic Digestion Performance of Mesophilic (UASB) Reactor Using Cattle Slaughterhouse Wastewater as Substrate. International Journal of Environmental Research and Public Health, 15(10), 2220. https://doi.org/10.3390/ijerph15102220 227 Naegele, H.-J., Mönch-Tegeder, M., Haag, N. L., & Oechsner, H. (2014). Effect of substrate pretreatment on particle size distribution in a full-scale research biogas plant. Bioresource Technology, 172, 396–402. https://doi.org/10.1016/j.biortech.2014.09.037 Natural Resource Conservation Services. (2007). Manure Chemistry – Nitrogen, Phosphorus, and Carbon. O’Brien, K. (2017). The State of Organics Recovery in Michigan: Priorities and Recommendations. Odey, G., Adelodun, B., Kim, S.-H., & Choi, K.-S. (2021). Status of Environmental Life Cycle Assessment (LCA): A Case Study of South Korea. Sustainability, 13(11), 6234. https://doi.org/10.3390/su13116234 Oswald, T. (2013, August 13). New MSU anaerobic digester to supply power for south campus buildings. Https://Msutoday.Msu.Edu/News/2013/New-Msu-Anaerobic-Digester-to- Supply-Power-for-South-Campus-Buildings. https://msutoday.msu.edu/news/2013/new- msu-anaerobic-digester-to-supply-power-for-south-campus-buildings Owen, J. J., & Silver, W. L. (2015). Greenhouse gas emissions from dairy manure management: a review of field-based studies. Global Change Biology, 21(2), 550–565. https://doi.org/10.1111/gcb.12687 Passos, F., Ortega, V., & Donoso-Bravo, A. (2017). Thermochemical pretreatment and anaerobic digestion of dairy cow manure: Experimental and economic evaluation. Bioresource Technology, 227, 239–246. https://doi.org/10.1016/j.biortech.2016.12.034 Quinto, P., Aguilar, H., López, R., & Reyes, A. (2016). Potential of biogas production in an anaerobic digester solar assisted. Renewable Energy and Power Quality Journal, 580–582. https://doi.org/10.24084/repqj14.399 Rader, N., & Hempling, S. (2001). The Renewable Portfolio Standard: A Practical Guide. Ram, A., & Sharma, P. (2017). A study on life cycle assessment. International Journal of Engineering and Advanced Technology, 6(17), 197–201. ReFED. (2018). Restaurant Food Waste Action Guide. Renewable Fuels Association. (2021). The California Low Carbon Fuel Standard and Ethanol: A Decade of Reducing Greenhouse Gas Emissions. Rossi, E., Pecorini, I., & Iannelli, R. (2022). Multilinear Regression Model for Biogas Production Prediction from Dry Anaerobic Digestion of OFMSW. Sustainability, 14(8), 4393. https://doi.org/10.3390/su14084393 228 RTI International. (2010). Greenhouse Gas Emissions Estimation Methodologies for Biogenic Emissions from Selected Source Categories: Solid Waste Disposal, Wastewater Treatment, Ethanol Fermentation. Schnaars, K. (2012, December). What every operator should know about anaerobic digestion. Operator Essentials. Operator Essentials. Scripps, D., Phillips, T., & Peretick, K. (2016). Report on the Implementation and Cost- Effectiveness of the P.A. 295 Renewable Energy Standard. Sharma, R. K. (2007). Social Change and Social Control. Atlantic Publishers & Distributors (P) Ltd. Shen, F., Li, H., Wu, X., Wang, Y., & Zhang, Q. (2018). Effect of organic loading rate on anaerobic co-digestion of rice straw and pig manure with or without biological pretreatment. Bioresource Technology, 250, 155–162. https://doi.org/10.1016/j.biortech.2017.11.037 Shine, K. P. (2009). The global warming potential—the need for an interdisciplinary retrial. Climatic Change, 96(4), 467–472. https://doi.org/10.1007/s10584-009-9647-6 Somers, M. H., Azman, S., Sigurnjak, I., Ghyselbrecht, K., Meers, E., Meesschaert, B., & Appels, L. (2018). Effect of digestate disintegration on anaerobic digestion of organic waste. Bioresource Technology, 268, 568–576. https://doi.org/10.1016/j.biortech.2018.08.036 Stock, J. H. (2015). The Renewable Fuel Standard: A Path Forward. Stuever, B. (2013, July 11). MSU Builds Combined Heat and Power System using Anaerobic Digestion. Https://Www.Egr.Msu.Edu/Bae/Adrec/Feature/South-Campus-Anaerobic- Digester/. https://www.egr.msu.edu/bae/adrec/feature/south-campus-anaerobic-digester/ Szacherska, K., Oleskowicz-Popiel, P., Ciesielski, S., & Mozejko-Ciesielska, J. (2021). Volatile Fatty Acids as Carbon Sources for Polyhydroxyalkanoates Production. Polymers, 13(3), 321. https://doi.org/10.3390/polym13030321 The National Academy of Sciences. (2011). Renewable Fuel Standard: Potential Economic and Environmental Effects of U.S. Biofuel Policy. The U.S. Environmental Protection Agency (EPA) State and Local Energy and Environment Program. (2018). Quantifying the Multiple Benefits of Energy Efficiency and Renewable Energy: A Guide for State and Local Governments. Thoma, G., Popp, J., Nutter, D., Shonnard, D., Ulrich, R., Matlock, M., Kim, D. S., Neiderman, Z., Kemper, N., East, C., & Adom, F. (2013). Greenhouse gas emissions from milk production and consumption in the United States: A cradle-to-grave life cycle assessment 229 circa 2008. International Dairy Journal, 31, S3–S14. https://doi.org/10.1016/j.idairyj.2012.08.013 Townsend, A., & Havercroft, I. (2019). The LCFS and CCS Protocol: An Overview for Policymakers and Project Developers. Tsapekos, P., Kougias, P. G., Egelund, H., Larsen, U., Pedersen, J., Trénel, P., & Angelidaki, I. (2017). Improving the energy balance of grass-based anaerobic digestion through combined harvesting and pretreatment. Anaerobe, 46, 131–137. https://doi.org/10.1016/j.anaerobe.2016.12.005 Turnbull, J. H., & Komthunzi, W. (2004). Greenhouse Gas Emission Reduction Associated with Livestock Waste Management Systems: A Case Study for the Langerwerf Dairy Waste Management System. IEA Bioenergy. United Nations. (2019). The Sustainable Development Goals Report. United Nations Environment Programme. (2019). Emissions Gap Report 2019. United States Environmental Protection Agency. (2014). Reducing Wasted Food and Packaging: A Guide for Food Services and Restaurants. https://nepis.epa.gov/Exe/ZyNET.exe/P100I7UW.TXT?ZyActionD=ZyDocument&Client= EPA&Index=2011+Thru+2015&Docs=&Query=&Time=&EndTime=&SearchMethod=1& TocRestrict=n&Toc=&TocEntry=&QField=&QFieldYear=&QFieldMonth=&QFieldDay= &IntQFieldOp=0&ExtQFieldOp=0&XmlQuery=&File=D%3A%5Czyfiles%5CIndex%20D ata%5C11thru15%5CTxt%5C00000009%5CP100I7UW.txt&User=ANONYMOUS&Pass word=anonymous&SortMethod=h%7C- &MaximumDocuments=1&FuzzyDegree=0&ImageQuality=r75g8/r75g8/x150y150g16/i42 5&Display=hpfr&DefSeekPage=x&SearchBack=ZyActionL&Back=ZyActionS&BackDes c=Results%20page&MaximumPages=1&ZyEntry=1&SeekPage=x&ZyPURL U.S. Department of Energy. (2016). Renewable Portfolio Standard Policies. U.S. Environmental Protection Agency. (2015). Beneficial Uses of Manure and Environmental Protection. U.S. Environmental Protection Agency. (2018). Food: Material-Specific Data. Https://Www.Epa.Gov/Facts-and-Figures-about-Materials-Waste-and-Recycling/Food- Material-Specific-Data#wfr. https://www.epa.gov/facts-and-figures-about-materials-waste- and-recycling/food-material-specific-data#wfr Wainaina, S., Lukitawesa, Kumar Awasthi, M., & Taherzadeh, M. J. (2019). Bioengineering of anaerobic digestion for volatile fatty acids, hydrogen or methane production: A critical review. Bioengineered, 10(1), 437–458. https://doi.org/10.1080/21655979.2019.1673937 230 Walker, M., Iyer, K., Heaven, S., & Banks, C. J. (2011). Ammonia removal in anaerobic digestion by biogas stripping: An evaluation of process alternatives using a first order rate model based on experimental findings. Chemical Engineering Journal, 178, 138–145. https://doi.org/10.1016/j.cej.2011.10.027 Wang, M. (2007). Overview of GREET model development in Argonne. Wang, X., Li, Z., Bai, X., Zhou, X., Cheng, S., Gao, R., & Sun, J. (2018). Study on improving anaerobic co-digestion of cow manure and corn straw by fruit and vegetable waste: Methane production and microbial community in CSTR process. Bioresource Technology, 249, 290–297. https://doi.org/10.1016/j.biortech.2017.10.038 Wang, X., Yang, G., Feng, Y., Ren, G., & Han, X. (2012). Optimizing feeding composition and carbon–nitrogen ratios for improved methane yield during anaerobic co-digestion of dairy, chicken manure and wheat straw. Bioresource Technology, 120, 78–83. https://doi.org/10.1016/j.biortech.2012.06.058 Weichold, K., & Barber, B. L. (2009). Introduction to Social Change and Human Development. The International Society for the Study of Behavioural Development. World Biogas Association. (2018). GLOBAL FOOD WASTE MANAGEMENT: Full Report AN IMPLEMENTATION GUIDE FOR CITIES. http://www.waste.ccacoalition.org/document/white-paper-waste-and-climate-change-iswa- key- Yacobucci, B. C. (2013). Analysis of Renewable Identification Numbers (RINs) in the Renewable Fuel Standard (RFS). Yu, J., Zhao, Y., Zhang, H., Hua, B., Yuan, X., Zhu, W., Wang, X., & Cui, Z. (2017). Hydrolysis and acidification of agricultural waste in a non-airtight system: Effect of solid content, temperature, and mixing mode. Waste Management, 59, 487–497. https://doi.org/10.1016/j.wasman.2016.10.019 Yu, Z., Leng, X., Zhao, S., Ji, J., Zhou, T., Khan, A., Kakde, A., Liu, P., & Li, X. (2018). A review on the applications of microbial electrolysis cells in anaerobic digestion. Bioresource Technology, 255, 340–348. https://doi.org/10.1016/j.biortech.2018.02.003 Zhong, Y., Bustamante Roman, M., Zhong, Y., Archer, S., Chen, R., Deitz, L., Hochhalter, D., Balaze, K., Sperry, M., Werner, E., Kirk, D., & Liao, W. (2015). Using anaerobic digestion of organic wastes to biochemically store solar thermal energy. Energy, 83, 638–646. https://doi.org/10.1016/j.energy.2015.02.070 Zhou, J., Zhang, R., Liu, F., Yong, X., Wu, X., Zheng, T., Jiang, M., & Jia, H. (2016). Biogas production and microbial community shift through neutral pH control during the anaerobic digestion of pig manure. Bioresource Technology, 217, 44–49. https://doi.org/10.1016/j.biortech.2016.02.077 231 Zou, H., Chen, Y., Shi, J., Zhao, T., Yu, Q., Yu, S., Shi, D., Chai, H., Gu, L., He, Q., & Ai, H. (2018). Mesophilic anaerobic co-digestion of residual sludge with different lignocellulosic wastes in the batch digester. Bioresource Technology, 268, 371–381. https://doi.org/10.1016/j.biortech.2018.07.129 232