INVESTIGATION OF THE DEGRADATION OF LIGNOCELLULOSIC MATERIALS IN ANAEROBIC DIGESTION By Juan Pablo Rojas - Sossa A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Biosystems Engineering Master of Science 20 20 ABSTRACT INVESTIGATION OF THE DEGRADATION OF LIGNOCELLULOSIC MATERIALS IN ANAEROBIC DIGESTION By Juan Pablo Rojas - Sossa Biogas produced through anaerobic digestion of lignocellulosic materials, is largely recognized as one of the only carbon negative fuel sources . The study consisted of two parts; 1 ) codigestion of AFEX TM - pretreated corn stover and cow manure and 2) degradation of plant cell wall components (different compounds) in agricultural biogas plants. It was concluded from the first part of the study that AFEX TM - pretreated corn stover promotes conversion of methane production in anaerobic digestion but has a smaller impact on consumption of the plant cell wall components. The corresponding biogas production (213 L/kg VS loading) of the AFEX treated co - digestion was 22% higher than that (175 L/kg VS loading) of the untreated co - digestion. The second part of the study led to the conclusion that biogas is produced mainly from nonstructural carbohydrates in the influent , and the plant cell wall makes a smaller contribution to biogas generation. Was observed g reater correlations between the biogas productivity and t he reduction of two organic components ( TOC= 17.8% & Protein= 18.1%) . On the other hand, lower correlations were detected between the consumptions of the plant cell was components ( Lignin = 12.5% , Cellulose= 3.7% & Xylan=0 %) and the biogas productivity. iii I want to dedicate this work to my friend Dr. Shenpan Lin, for being such an important part of iv ACKNOWLEDGEMENTS I owe a debt of gratitude to my supportive advisor Dr. Dana Kirk for believing in me and supporting my graduate studies during my entire enrollment at MSU. I am also grateful to Dr. Wei Liao for being an example and a great advisor during all the years wo rking together. I thank my parents, my brothers, and my nephew and nieces for being such an important part of my life and for supporting me at every step of my life. I acknowledge my dear friend Dr. Melissa Rojas - Downing for being such reliable support dur ing my enrollment at MSU and for being a great example. I thank all my classmates: graduate students Juan Sebastian Hernandez, Ian Kropp, Mahlet Garedew, Matthew Herman, among others, for being my partners and friends throughout this journey, and all the s I am grateful to the BAE Department for always helping me on various matters during my studies at MSU. v TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ........................ vii LIST OF FIGURES ................................ ................................ ................................ ..................... viii 1. INTRODUCTION ................................ ................................ ................................ ................... 1 1.1. Statement of the Problem ................................ ................................ ................................ . 1 1.2. Goals & Objectives ................................ ................................ ................................ .......... 2 2. LITERATURE REVIEW ................................ ................................ ................................ ........ 4 2.1. Transition to a renewable - based portfolio in the United States ................................ ....... 4 2.2. AD ................................ ................................ ................................ ................................ .... 6 2.3. AD principles ................................ ................................ ................................ ................... 7 2.4. AD: Microbial communities ................................ ................................ ........................... 11 2.5. Plant cell wall components ................................ ................................ ............................. 13 2.5.1. Cellulose ................................ ................................ ................................ ................. 14 2.5.2. Hemicelluloses ................................ ................................ ................................ ........ 15 2.5.3. Lignin ................................ ................................ ................................ ...................... 15 2.5.4. Pectin ................................ ................................ ................................ ....................... 16 2.6. Lignocellulosic - biomass degradation in AD ................................ ................................ .. 16 2.7. Pretreatment technologies for lignocellulosic materials ................................ ................ 17 2.8. AFEX TM ................................ ................................ ................................ ......................... 18 3. MATERIALS & METHODS ................................ ................................ ................................ 21 3.1. Dynamic microbiome assembly and the effect on the performance of AD of AFEX - pretreated corn stover and CS ................................ ................................ ................................ ... 22 3.1.1. Feedstock sample collection ................................ ................................ ................... 22 3.1.2. Feedstock mixture preparation ................................ ................................ ................ 22 3.1.3. Semicontinuous AD experiment ................................ ................................ ............. 23 3.1.4. Analytical methods ................................ ................................ ................................ . 25 3.1.5. Microbial Community Analysis ................................ ................................ .............. 27 3.1.6. Evaluation of Digestion Performance ................................ ................................ ..... 29 3.1.7. Statistical analysis of microbial - community data & of digestion performance ...... 30 3.2. Degradation of lignocellulosic feedstocks at ABPs ................................ ....................... 32 3.2.1. Influent and effluent sampling at ABPs ................................ ................................ .. 32 3.2.2. Analytical Methods ................................ ................................ ................................ . 34 3.2.3. Biochemical Methane Potential (BMP) in ABPs influents ................................ ..... 35 3.2.4. Statistical analysis of ABP data ................................ ................................ .............. 36 4. RESULTS ................................ ................................ ................................ .............................. 39 4.1. Dynamic microbiome assembly and the effect of the performance of AD of AFEX - pretreated corn stover and CS ................................ ................................ ................................ ... 39 4.1.1. Feedstock characterization and codigestion mixture mass ratios ........................... 39 4.1.2. Digestion Performance ................................ ................................ ............................ 40 4.1.3. Microbial Community Analysis ................................ ................................ .............. 45 vi 4.2. Degradation of Lignocellulosic Feedstocks at ABPs ................................ ..................... 52 4.2.1. Influent and effluent dry matter and characterization of raw samples .................... 52 4.2.2. BMP Experiments on ABP influents ................................ ................................ ...... 57 4.2.3. Variance - based sensitivity analysis of the downregulation of each compound and the effect on biogas productivity ................................ ................................ ........................... 60 5. CONCLUSIONS ................................ ................................ ................................ ................... 63 5.1. Dynamic microbiome assembly and the effect on the performance of AD of AFEX - pretreated corn stover and CS ................................ ................................ ................................ ... 63 5.2. Degradation of lignocellulosic feedstocks at the ABPs ................................ ................. 64 6. RECOMMENDATIONS ................................ ................................ ................................ ....... 65 APPENDIX ................................ ................................ ................................ ................................ ... 66 REFERENCES ................................ ................................ ................................ ............................. 89 vii LIST OF TABLES Table 1. Coefficients for stoichiometric equations for anaerobic treatment of various organic materials (Rittmann & McCarty, 2001). ................................ ................................ ....................... 11 Table 2. Description of the ABPs sampled ................................ ................................ ................... 33 Table 3. Characteristics of assay feedstocks ................................ ................................ ................. 40 Table 4. Decreases in the averages dry matter content of each parameter in the reactors. ........... 57 Table A1. Statistical results from the AD performance around the three HRT. . 77 Table A2. Statistical results from characterization of the influents and effluents of three different 88 viii LIST OF FIGURES Figure 1. Categorization of renewable fuels by the RFS Act. Adapted from ref. (EPA, 2017). .... 5 Figure 2. Functional roles in the biogas production. Each box shows a list of genes (and their abundance) identified for each metabolic pathway of all the processes (Campanaro et al., 2016). ................................ ................................ ................................ ................................ ....................... 12 Figure 3. Chemical structure of polysaccharides and polymers that constitute the plant cell wall; these materials are also called lignocellulosic biomass. a. Cellulose (V Balan et al., 2012). b. Hemicelluloses (V Balan et al., 2012). c. Lignin (Sarkanen & Ludwig, 1971). ........................... 14 Figure 4. AFM images of untreated a. and AFEX TM - pretreated b. corn stover. Confocal fluorescence imaging analysis of untreated c. and AFEX TM - pretreated d. corn stover. Adapted from ref. (Chundawat, Donohoe, et al., 2011) with permission. ................................ ........................... 19 Figure 5. Flow direction of biogas and water in the water displacement method. ....................... 24 Figure 6. Digestion performance of AFEX and CS during the digestion (three HRTs). a. Biogas productivity, b. methane content, c. VS content reduction, d. cellulose content reduction, e. xylan content reduction, f. total VFA concentration. ................................ ................................ ............. 40 Figure 7. Average acetate (Hac) and propioni c acid (Hpa) concentrations in the reactors. ......... 44 Figure 8. Diversity of microbial communities in both digestion reactions. a. The rank abundance (Whittaker) plots of relative abundance of OTUs in both digesters. The dots represent the logarithmic percentage of the relative abundance of each species, and then the lognormal curve was plotted on the data. b. Examples or the diversity curves seen in the digesters. ..................... 46 Figure 9. Ecological diversity indices (Shannon, Simpson, Inverse Simpson, and Fisher) for each AD reactor. ................................ ................................ ................................ ................................ .... 47 Figure 10. Relative abundance of different taxa found in the reactors. a. Relative abundance of the microbial domains. b. Relative bacterial phylum abundance. c. Relative Spirochaetes genera abundance. d. Relative Archaea genera abundance. ................................ ................................ ..... 49 Figure 11. Nonmetric multidimensional scaling of the relative abundance of microbial communities in the digesters. ................................ ................................ ................................ ........ 51 Figure 12. a. Influent and b. effluent composition analysis on a dry - matter basis. ...................... 52 Figure 13. Raw sample characterization. a. Chemical oxygen composition of a1. influents, a2. influents. b. Concentrations of nitrogen forms: b1. influents, b2. effluents. ................................ 55 Figure 14. TOC variation in the dry matter of the influents and effluents throughout the sampling period. ................................ ................................ ................................ ................................ ........... 56 ix Figure 15. BMPs: accumulated gas production on different dates of sampling of influents. a. Accumulated gas production of Farm A, b. accumulated gas production of Farm B, c. accumulated gas production of the MSU south campus digester. ................................ ................................ ..... 58 Figure 16. The ranking of contributions to the variance of biogas productivity by each of the parameters measured. ................................ ................................ ................................ .................... 61 1 1. INTRODUCTION 1.1. Statement of the Problem The world is currently transitioning from fossil fuel dominated energy sources to more renewable - energy based approach es , like bioenergy technologies. Bioenergy generation has been promoted in many countries (e.g., Germany, Sweden, Italy, and USA) by tax incentives and public policies. However, the extensive application of bioenergy technologies is not yet a reality around the wo rld. Many technologies such as pyrolysis, gasification, transesterification, and saccharification are still at the bench and pilot scale. Anaerobic digestion (AD) , however, the most widely applied bioenergy technology, has been deployed globally at residen tial and commercial scales. The advantages of AD over other bioenergy technologies are the production of renewable energy in addition to organic - waste disposal, environmental protection, and greenhouse emission reduction (Mao, Feng, Wang, & Ren, 2015) . Application of AD to convert recalcitrant carbon into usable carbon methane attracted increasing attention in recent years. Recalcitrant carbon is most commonly found in nonfood biomass, or lignocellulosic biomass (e.g., corn stover, wood biomass) , specifically in the plant cell wall (Keegstra, 2010) . This advancement may significantly increase the diversity of feedstocks for methane production and contribute to non food - based bioenergy generation. Because of the recalcitrant nature of lignocellulosic biomass, pretreatment needs to be utilized to trea t biomass before digestion. Numerous efforts have been made to develop pretreatment processes. One of the most developed pretreatments is Ammonia Fiber Expansion (AFEX TM ). The goal of the AFEX TM process is to break important linkages between plant cell wal l components, thus paving an accessible way to usable carbon sources. The AFEX TM - technology has reached the pilot scale operation and could be a valid pretreatment process for AD of lignocellulosic biomass. 2 Many countries worldwide have established publi c policies to promote the AD technology and its implementation. For example, countries like the United States (US) have the Renewable Fuel Standard (RFS). The RFS promotes the production of biogas from lignocellulosic sources D3 /D7 category known as cellulosic fuel (from a lignocellulosic source) (EPA, 2014) . This framework requires obligated parties involved in the vehicle fuel supply chain to source a portion of the fuel from renewable lignocellulosic feedstocks. For biogas producers utilizing animal manures, biosolids, or landfill gas this provides new opportunities to ma rket renewable natural gas (methane sourced from biogas). However, the degradation of cellulosic biomass and its contribution to biogas production via AD are not fully understood. To fulfill the needs mentioned above, this thesis covers two important topi cs: codigestion of pretreated lignocellulosic biomass and degradation of lignocellulosic biomass during the digestion process. 1.2. Goals & Objectives To achieve the project goals, the following specific objectives were pursued: Dynamic microbiome assembly and the effect on the performance of AD of AFEX - pretreated corn stover and conventional corn stover. a) To study the impact of codigestion of conventional corn stover and AFEX - pretreated corn stover on digestion performance. b) To elucidate dynamic changes of microbial communities during the codigestion. c) To identify correlations in order to explain the relation between digestion performance and microbial communities. Degradation of lignocellulosic feedstocks at agricultural biogas plants 3 d) To perform broad characterization of the influents and effluents of different commercial agricultural biogas plants ( ABPs ) by different methods from the literature. e) To carry out the biomethane potential testing of from influent samples. f) To determine the consumption of different components in the influent samples and their contribution to biogas production. g) Conduct a sensitivity analysis to rank the importance of the components in terms of biogas productivity. 4 2. LITERATURE REVIEW In 2017, renewable power accounted for 70% of net additions to global power - generating capacity (REN21, 2018) . The uses of this energy are diverse in such sectors as heating and cooling, power, and transportation. Contributing to this progress are advancements in enabling technologies and political efforts to develop the institutional conditions that encourage the use of more renewable sources in the worldwide energy matrix. This review focuses specifically on the status of this bioenergy transition in the US and the applic ation of AD of lignocellulosic biomass to produce biogas and eventually renewable natural gas (RNG) . 2.1. Transition to a renewable - based portfolio in the United States To produce more renewable energy, the US government has established the RFS , created under the Energy Policy Act of 2005, which amended the Clean Air Act (EPA, 2017). The Energy Policy Act establishes requirements for the minimum volume of renewable fuel production to replace fossil fuel production. In 2007, the program was renewed, and changes were made in the long - term goals to redefine various parameters such as greenhouse gas (GHG) emissions and the categorizing renewable fuel into different types dependin g on the sources. Fuel pathways include four critical parameters serving to categorize different fuels. The parameters are (1) the feedstock, (2) production process (type of technology), (3) fuel type (form of fuel), and (4) capacity for reduction of GHG e missions as compared to 2005 petroleum baseline. By statute, the RFS includes four categories of renewable fuel each with a specific fuel pathway and a renewable identification number (RIN D - Codes). The categories are advanced biofuel (D5), biomass - based d iesel (D4), cellulosic biofuel (D3 & D7), and renewable fuel (D6). Figure 1 shows how the Environmental Protection Agency (EPA) categorizes each fuel according 5 to the GHG emissions reduction capacity. Categories D4 and D5 (Advanced & Biodiesel fuel) must m eet 50% of lifecycle GHG reduction and can be produced from renewable biomass (excluding cornstarch). The renewable fuel (D6) typically refers to ethanol derived from corn starch and must meet 20% of the GHG reduction. Figure 1 . Categorization of renewable fuels by the RFS Act. Adapted from ref. (EPA, 2017) . The third category is D3 and D7, called Cellulosic Fuels. The sources of these types of fuel include cellulose, hemicellulose, and lignin (plant cell wall components also called lignocellulosic biomass). Cellulosic fuel can reach 60% of the GHG reduction a s compared to 2005 petroleum baseline. With this categorization, the EPA has approved several pathways such as ethanol made from sugarcane, cellulosic ethanol made from corn stover, biogas from landfills, municipal wastewater facilities, agricultural diges ters, and any other biogas from the cellulosic components. In the case of biogas, there are two kinds of possible pathways: Q and T. The Q pathway is designated for biogas produced from cellulosic biomass transformed by AD and is assigned the D6 D4 & D5 D3 & D7 6 D3 code (cel lulosic fuel). In pathway T, biogas 1 is produced from no lignocellulosic feedstocks (carbohydrates, proteins, and lipids among others) and will be assigned code D5. Nowadays, many US biogas plants are registered as energy producers at the EPA, splitting be tween D3 and D5, using both categorizations to produce and sell their energy on the market. Within this framework, biogas is required to be purified to pipeline quality natural gas standard, resulting in a fuel commonly called RNG. Since 2014, there has be en a linear positive increase in RNG production in the US, and a stable increase is predicted until 2022 (Hanson, 2017) . Hanson (2017) showed that in 2016, the RNG produced was used to reach 82% of the federal targets set for cellulosic fuel. Producers are utilizing feedstocks such as dairy manure (DM), biosolids, and landfill gas with codigestion of food waste or starch, to produce D5 RNG . Biogas producers have been able to increase their revenue by 78%, where as the ones being compared register only as D5 producers. Although AD and biogas production look promising, the capacity for energy production from plant cell wall co mponents and the rate at which this production is possible continue to be explored. At this moment, it is important to more deeply understand the ability of AD to process these materials to link them with the actual market of RNG. 2.2. AD AD is a microbiome - bas ed bioprocess that consumes organic matter and produces mainly a mixture of methane (CH 4 ) and carbon dioxide (CO 2 ) called biogas, under warm conditions (typically >35 C) and in reactors with minimal oxygen (<1%). This mixed culture forms a microbial community where ecological principles drive metabolic fluxes and affect the operation 1 7 and performance of the process. The biogas can be transformed into usable energy in sev eral ways, including engine combustion or purification to a natural gas standard to produce RNG. AD can turn organic residues to high - value resources such as renewable energy and fertilizer in comparison with conventional disposal as wastewater (activated sludge) or solid waste (landfill) (McCarty, Bae, & Kim, 2011) . Currently, over 2,000 AD plants produce bioga s in the United States (USDA, USEPA, & USDOE, 2015) , whereas Europe has over 17,000 biogas plants (Brijde, Dumont, & Blume, 2014) . China is the world leader in biogas production, with 17 million digesters (Changda, Xiang, Wu, & Yifeng, 1994) . Countries in Africa and South and Central Americ a have been developing (and investing in) biogas technologies in the last decade owing to their high potential in those regions (Flavin et al., 2014; Roopnarain & Adeleke, 2017) . These regions have high viability of residual organic biomass and high solar ra diation: both conditions are favorable for integration of these technologies with AD exploitation (Aguilar Alvarez et al., 2016) . 2.3. AD principles AD involves a complex syntrophic association of producers/consumers that interact to attain effective transport of electrons. The digestion process is based upon three stages: (1) hydrolysis of complex compounds, (2) volatile fatty acid (VFA) formation, and (3) methane production (Fang & Liu, 2002) . AD proceeds in the absence of oxygen; therefore, organic acids assume the role of an electron acceptor in the fermentation process and turn into methane at the methanogenic stage (Rittmann & McCarty, 2001) . There exists a direct correlation between carbo n mineralization and microbial community structure. The first metabolic pathway involved here is hydrolysis. The hydrolyzed compounds are broken into monomers and oligomers by different enzymes such as amylase, cellulose, and protease, among others y, 2009) . The hydrolysis flux 8 is proportional to the organic loading rate (OLR) applied to a reactor. Penaud et al. ( 1997) showed t hat the best conditions for high consumption of organic matter and high VFA concentrations occur at pH 8.5 and at high OLRs (~5 kg of chemical oxygen demand [COD] per m 3 per day) with solid x ) higher than 2 days; this condition is normall y not considered in the design of AD reactors, owing to the high variability in other factors (like feedstock composition) with time. With respect to modeling of AD, the hydrolysis step normally is not considered because of the complex waste pool. The comb ination of lysis, nonenzymatic decay, phase separation, and physical breakdown is defined as hydrolysis thus creating a lot of complexities when modeling is attempted (Batstone et al., 2002) . After this initial step, the products of hydrolysis are fermented. Acidogenesis (VFA formation) follows hydrolysis. During acidogenesis, fermentative bacteria consume monosugars and fatty acids formed during the hydrolysis to produce organic acids, hydrogen, and CO 2 (Batstone et al., 2002) . During acidogenesis, most of the energy and mass flow from the organic polymers thereby going to organic acids (76%), such as acetic, propionic, isobutyric, butyric, valeric, isovaleric, isocaproic, and caproic, amon g others. There is less molar production of H 2 , acetate, and CO 2 (Speece, 1983) . The acidogenesis rate grows with the OLR. Goux et al., (2015) increased the OLR of mesophilic reactors until reaching acidosis and studied the microbial community of the reactors before, during, and after, thus demonstrating that there is a shift of the fermenters unde r these unbalanced conditions. Acidogenesis is a crucial upstream process for methane production. It should to be controlled, monitored, and maintained at low concentrations. Nevertheless, it is most important to monitor the consumption of VFA by the downs tream processes. Moreover, maintaining low VFA concentrations keeps alkalinity and pH in the appropriate range for the downstream processes of acetogenesis and methanogenesis. 9 Acetogenesis proceeds simultaneously with acidogenesis. Acetogenesis uses produc ts from acidogenic fermentation (e.g., propionate, butyrate, and valerate) and generates more acetic acid. From a thermodynamic standpoint, this process does not happen spontaneously because the 0 > 0). This condition changes its window from pH ~4 to pH ~6. Elevated temperatures will also influence the process of switching the reaction into a spontaneous state. Digestion of VFAs will be more effective with supplemental heat. This whole process may ha ppen at 35 °C for mesophilic conditions or at 47 °C under thermophilic conditions (Oh & Martin, 2010) . Finally, m ethanogenesis is the last metabolic process necessary to create methane. This process derives from two main metabolic pathways: the hydrogenotrophic pathway and acetotrophic pathway (Demirel & Scherer, 2008) . They are classified as based on two different groups of methanogens: acetate fermenters and hydrogen oxidizers (Rittmann & McCarty, 2001) . The acetate fermenters employ acetate as an electron donor and as a carbon source and are slow growers. Hydrogen oxidizers grow faster but require a high concentration of H 2 to favor the process (Demirel & Scherer, 2008) . Because methanogens must be present to produce methane, the normal pH requirement is in the range of 6.5 to 8.2 (Safferman, Kirk, Faivor, & Wu - haan, 2012) . Furthermore, due to the energy available from the electron donor acceptor setup, AD has a slow growth rate (f s 0 x growth requirements (Rittmann & McCa rty, 2001) . During AD startup, the concentration of VFAs increases, and the conditions become less favorable for the growth of methanogens. In this stage, it is also important to maintain pH in the right range for methanogens. If pH is not maintained, t he methanogenic population could be hurt and gas production may decrease (Goux et al., 2015) . Normally, during the establishment process, it is possible to observe low gas production. During 10 this establishment, VFA concentrations are normally high (12,000 mg kg 1 ). Then, the concentration of VFAs decreases, and the abundance of methanogens increases, leading to increased alkalinity and CO 2 concentration (Rittmann & McCarty, 2001) . The increase in alkalinity directly affects pH ; the initial alkali nity depends directly on the feedstock but can increase during the process owing to the formation of cation salts. Regardless of this stable state, the systems always require normal evaluation of the parameters. AD systems are stable when the culture strik es a balance between VFA formation and methane production (Y. Chen, Cheng, & Creamer, 2008; Demirel & Scherer, 2008) . This condition is the most important factor to ensure good operation and effective ener gy production. By following this recommendation, it is possible to maintain a healthy process and a productive anaerobic digester. On the other hand, the process has also been studied from the energetic or stoichiometric point of view. After assuming that carbon dioxide is the electron acceptor, one can write the stoichiometric equation for generalized organic waste (Rittma nn & McCarty, 2001) : (1) where and represents the fraction of waste organic matter synthesized or converted to cells, and denotes the portion converted to energy, such that . The value of depends on the energy generation and synthesis reactions as well as the decay rat e and . For a reactor in operation steady state, f s can be estimated using equation 2: 11 (2) Table 1 summarizes the values of and for methane - producing fermentation of common organic compounds. Table 1 . Coefficients for stoichiometric equations for anaerobic treatment of various organic materials (Rittmann & McCarty, 2001) . Waste component Typical Chemical Formula Y (gVSS a /g BOD L removed) (d - 1 ) Carbohydrates C 6 H 10 O 5 0.28 0.20 0.05 Proteins C 16 H 24 O 5 N 4 0.08 0.056 0.02 Fatty acids C 16 H 32 O 2 0.06 0.042 0.03 Municipal sludge C 10 H 19 O 3 N 0.11 0.077 0.05 Ethanol CH 3 CH 2 OH 0.11 0.077 0.05 Methanol CH 3 OH 0.15 0.11 0 0.05 Benzoic acid C 6 H 5 COOH 0.11 0.077 0.05 2.4. AD: Microbial communities As explained before, biogas is produced by a biological process called AD. The process is performed by a specialized and sophisticated microbial community, where different actors have different roles in the structure of the joint orga nization (Campanaro et al., 2016) . The different members of the organization have different functions, and they are linked by i nteractions forming networks. If these interactions are established, then the consortium can turn diverse kinds of macromolecules into methane and carbon dioxide mainly. Figure 2 shows the gene diversity present in an AD microbial community, and how the or 12 Figure 2 . Functional roles in the biogas production. Each box shows a list of genes (and their abundance) identified for each metabolic pathway of all the processes (Campanaro et al., 2016) . This functional structure manifests an increase of specialization as it reaches the final metabolic pathway of the process. This means that there are more diverse kinds of genes able to act in the firs t main steps of AD, and later steps require more unique kinds of microbes to finally produce biogas. That is why it is possible to see high species diversity in the fermentation steps and more unique species of microbes that are involved in methane formati on. That is why it is mandatory to study the guild (metabolic diversity) and the clade (phylogenetic diversity) of an AD microbial community. The phylogeny in an AD community is based on two main domains: bacteria and archaea. The archaea population mainly belongs to the phylum Euryarchaeota and is mainly dominated by acetoclastic methanogenic genera such as Methanosaeta , followed by hydrogenoclastic genera 13 such as Methanospirillum and Methanobrevibacter (Demirel & Scherer, 2008) . Thermophilic reactors are dominated by the hydrogenoclastic genus Methanothermobacter followed by versatile Methanosarcina (Kirkegaard et al., 2017) . On the other hand, the phylogeny of the bacterial (Kirke gaard et al., 2017; Rojas - Sossa et al., 2017) . Phyla such as Firmicutes, Proteobacteria, Chloroflexi, Actinobacteria, Bacteroidetes, Synergistetes, and Acidobacteria, are the most abundant in AD reactors. What is still not clear is whether the microbes are enriched in the reactors or if the microbes immigrate to the reactors via the feedstocks. Kirkegaard et al., (2017) tested whether the immigrating microorganisms tend to die, survive, or grow in the reactors. It was found that there was a peak of microorganisms who were highly enriched in the reactor. Those authors found that the microorganisms that were enriched represented 60% of the total abundance in the reactors. Improving the understanding of these relationships will allow scientists and engineers to focus on control ling and changing those relations with the objective of improving the feedstock - to - methane conversion or attenuating the inhibitory effects in the process. Some of the substrates whose conversion to methan e needs to be improved are plant cell wall components , lignocellulosic biomass. These compounds are the most abundant carbon source on Earth (Tye, Le e, Wan Abdullah, & Leh, 2016) . 2.5. Plant cell wall components composed mainly of structural carbohydrates that form the plant cell wall (Chundawat, Donohoe, et al., 2011) . These structures have been mostly categorized into three groups two kinds of polysaccharides (cellulose and hemicellulose) and one type of polymer (lignin) that form the co mplex and strong structures providing support to a plant as well as protection against microbial 14 invasion (Keegstra, 2010 ) . These materials have been driven by evolution to become highly recalcitrant toward bacterial enzymes. The main components of the plant cell wall are cellulose, hemicelluloses, lignin, and pectin (V Balan, Sousa, Chundawat, Humpula, & Dale, 2012) . In Figure 3, readers can see the reported chemical structures of cellulose, hemicelluloses, and lignin. 2.5.1. Cellulose In Figure 3a, there is a diagram of the crystalline cellulose nanofibril structure, with the formula (C 6 H 10 O 5 ) n . These compounds form straight cylinders strongly bound and giving rigidity to the plant cell wall. Glucose is a homogeneous polysaccharide held together via covalently (1,4) - - D - glucans. These carbon chains interact with one another via hydrogen bonds to form a crystalline structure (Keegstra, 2010) . This material is the main component of the plant cell wall (V Balan et al., 2012) . a. b. Figure 3 . Chemical structure of polysaccharides and polymers that constitute the plant cell wall; these materials are also called lignocellulosic biomass. a. Cellulose (V Balan et al., 2012) . b. Hemicelluloses (V Balan et al., 2012) . c. Lignin (Sarkanen & Ludwig, 1971) . 15 Figure 3 c. 2.5.2. Hemicelluloses Hemicellulose has two common structures: xyloglucan and arabinoxylan (Figure 3b). Scheller & Ulvskov, (2010) described hemicelluloses as a group of polysaccharides that are neither cellulose nor pe ctin and contain linked backbones of glucose, mannose, or xylose. These polysaccharides are more heterogeneous in their structure and in their physicochemical properties than cellulose is (Scheller & Ulvskov, 2010) . The presence of the different kinds of hemic elluloses varies among plant families or species, but it is known that hemicelluloses bind tightly to cellulose microfibrils via hydrogen bonds (Keegstra, 2010) . Hemicellulose engages in complex binding to lignin, called the lignin carbohydrate complex (LCC). 2.5.3. Lignin Lignin is the second most abundant constituent after cellulose and is reported to be more abundant (by mass) in plant cell walls (Norgren & Edlund, 2014) . Figure 3c provides an overview of lignin structure. It is a heterogeneous and amorphous macromolecule with variable composition 16 dependent on the plant source (Sj str m, 1993) . However, lignin can be classified by three monomers: p - coumaryl alcohol, coniferyl alcohol, and sinapyl alcohol, which all differ in the positions of the aromatic rings (V Balan et al., 2012; Norgren & Edlund, 2014) . 2.5.4. Pectin Voragen, Coenen, Verhoef, & Schols, (2009) described pectin as one of the most complex macromolecules in nature, because it can be composed of as many as 17 different monosaccharides and with 20 different linkages. Due their anionic natur e, these polysaccharides are involved in the transport of ions and in the porosity of the plant cell wall. On the other hand, several studies indicate the importance of the characterization of these compounds in the plant cell wall and of their complex int eractions with the other components of the cell wall (V Balan et al., 2012; Keegstra, 2010; S cheller & Ulvskov, 2010) . 2.6. Lignocellulosic - biomass degradation in AD AD of lignocellulosic materials is inefficient because most of the carbon is recalcitrant. These materials show poor solubilization, thus resulting in low biogas production and poor dig estion performance (Y. Chen et al., 2008; Mao et al., 2015; Raposo, De La Rubia, Fernández - Cegrí, & Bor ja, 2012) . Important findings on this topic started with advances in the quantification of fiber content in forage samples (Goering & Van Soest, 1975) . The publication of this method allowed researchers to address the degradation of different components of the plant cell wall. Hills, (1979) showed how pure microcrystalline cellulose is consumed in AD and how cellulose combined with other organic compounds in complex waste may not be totally bioavailable for bacteria. Later, Hills & Roberts, (1981) p ublished one of the first optimization efforts to maximize gas production during the degradation of dairy manure and field crop residues. They found optimum performance was achieved when the nonlignin carbon - to - nitrogen ratio was ~25 and ~32. On the other hand, 17 Yue et al., (2013) demonstrated how the codigestion of dairy manure and corn stover promotes biogas productivity. However, the conversion of nonrecalcitrant carbon from corn stover directly correlates with the microbial community structure. Besides, some plants generate resin extracts, and these extracts could be inhibitory for AD (Speece, 1987) . Finally, there is evidence of inhibition of AD by lignin derivates obtained with aldehyde; they are highly toxic to methanogens (Y. C hen et al., 2008) . To optimize this process, it is necessary to pretreat these materials to improve the bacterial enzyme accessibility or detoxify the feedstock to make it bioavailable as an energy source to microbes. 2.7. Pretreatment technologies for lign ocellulosic materials The nature of lignocellulosic materials makes them very resistant to an enzymatic attack. The main objective of the pretreatment is to change or weaken these properties to prepare the material for the downstream process (AD and ethanol production, among ot hers). Different kinds of pretreatment methods exist, including physical, chemical, physicochemical, and biological (Taherzadeh & Karimi, 2008) . Different challenges make the pretr eatment of these materials effective , this due the high disparity presented in the lignocellulose structure between specific plants, specific plant tissues, and plant cells . Actually this materials shows a complexity paradigm across scale of systems who are tried to be stud ied (Chundawat, Donohoe, et al., 2011) . Factors such as crystallinity of cellulose, surface area protection by lignin and hemicellulose, the degree of cell ulose polymerization, and acetylation of hemicelluloses affect the pretreatment of lignocellulosic biomass (Karimi & Taherzadeh, 2016) . Analytical methods, imaging, and crystallinity analyses are normally used to evalua te pretreatment methods. Existing pretreatments include milling, irradiation, microwaving, steam explosion, supercritical CO 2 , alkaline and acid hydrolysis, AFEX TM , and others (Taherz adeh & Karimi, 2008) . 18 2.8. AFEX TM AFEX TM is an important pretreatment technology. It is an ammonia - based pretreatment that has proven to be a cost - effective way to reduce the recalcitrance of plant cell wall components (e.g., lignocellulosic materials) and improve microbial fermentation (Chundawat et al., 2013) . This technology is currently being scaled up for potential commercialization. One of the important features of AFEX TM is the reversible nature of the interaction of ammonia (NH 3 ) with biomass; this feature allows most of the ammonia to be recovered (Chundawat et al., 2013) . The conventional method is to use liquid ammonia (0.3 2.0 [g NH 3 ]/[g dry biomass]) in moist biomass (0.1 2.0 g in H 2 O per gram of dry biomass) with supplemental heat at high pressure (~ 2.25 MPa ). The process takes place in a batch reactor , but could occur in a plug flow reactor or in a packed bed reactor (Chundawat et al., 2013) . First, ammonia reacts with water to cause a rapi d increase in temperature. Later, more heating is supplied to maintain a constant reaction temperature (~100 °C) ; this process lasts for 30 min, then NH 3 is exhausted, thereby releasing the pressure from the reactor (Perez - Pimienta et al., 2016) . Three physicochemical changes exist in AFEX ; (1) t he LCC split and product formation, (2) lignin/hemicellulose redistribution, and (3) cellulose decrystallization (Chundawat, Beckham, Himmel, & Dale, 2011; da Costa Sousa, Chundawat, Balan, & Dale, 2009) . The LCC linkages are some of the most important impediment s to cellulase. By ammonolysis (an NH 3 reaction with esters) and hydrolysis (acid formation), AFEX TM breaks these bonds and produces amides and organic acids (Balan et al., 2012). The decrystallization effect could be observed scanning sample surfaces at t he microscale by atomic force microscopy (AFM). In Figure 4, AFM pictures of conventional corn stover are presented (Figure 4a), as is AFEX TM - pretreated corn stover (Figure 4b). In the AFM images, is possible to observe visually how AFEX TM changed the geometry on 19 the architecture of the cell wall . With the AFEX TM technology is not possible to observes the normal array of the typical microfibril. Here is possible to observe changes in the normal crevices and cracks that formed part of the na tural cell wall surface landscape , changing the AFM surface roughness factor from 16 to 30 . The action mechanism of AFEX TM consists in several chemical processes. Ammonia first reacts and evaporates; this event results in the formation of nanoporous ( 10 to 500 nm width) tunnel - like networks due to the rapid decompression and volatilization of ammonia; then, the extractives redeposit out of the cell wall surface. Chundawat et al., ( 2011) found on AFEX TM - reaction products a strong enrichment of lignin - derived phenolics. This, due a strong stain by Safranin die between samples of untreated corn stover ( Figure 4c) and AFEX TM corn stover ( Figure 4d). a. b. Figure 4 . AFM images of untreated a. and AFEX TM - pretreated b. corn stover. Confocal fluorescence imaging analysis of untreated c. and AFEX TM - pretreated d. corn stover. Adapted from ref. (Chundawat, Donohoe, et a l., 2011) with permission. 20 Figure 4 c. d. The nanoporous networks enhance a microbial attack (Lau & Dale, 2009) without removing any of lignin and hemicelluloses into separate liquid streams . Interest in the testing of AFEX TM materials as an AD substrate comes from the ability of the AFEX TM process to improve activity of the enzymes toward the cellulose /hemicellulose/lignin system at the same time releasing acetic and lactic acids (V Balan et al., 2012; Venkatesh Balan, Bals, Chundawat, Marshall, & Dale, 2009) . In conclusion, it is necessary to optimize AD. This optimization requires the use of different social and scientific disciplines. For example, it is crucial t o develop better policies regarding biogas production where the framework is aligned to the abilities of the technology. Additionally, it is important to understand the ecological principles of the microbial communities that control the process. Research s hould focus on understanding the nature of these cooperating species and the relations among them. Finally, it is necessary to improve analytical characterization of the AD feedstocks, specifically lignocellulosic compounds, in order to better determine wh at they are and improve their exploitation in AD. 21 3. MATERIALS & METHODS To obtain the results mentioned in the Objectives section, two experiments were performed and are described in detail in this section. The experimental design was different between the two projects. Nonetheless, the experiments are structured in the same way: sample collection and description, experimental setup, analytical methods, biological analysis (if necessary), performance evaluation, and statistical analysis. The first experiment is a novel evaluation of two kinds of semicontinuous codigestion of manure and corn stover , conventional corn stover ( CS ) and AFEX TM - pretreated corn stover ( AFEX ) . The continuous digestion was monitored for performance and microbial commun ity dynamics for 75 days. All the statistical analyses were conducted using nonparametric statistics and ecological statistics. The second experiment was evaluation of the digestion of plant cell wall components at ABPs in the state of Michigan. Here sampl es were collected from the influent and effluent and characterized ; then , the samples were evaluated in terms of the biogas potential, and finally, statistical analysis was performed to try to explain the correlation of these characteristics with the obser ved biogas productivity. The main objective was to examine the contribution of the plant cell wall components to the biogas production from the plants. Below is a description of both experiments. 22 3.1. Dynamic microbiome assembly and the effect on the performanc e of AD of AFEX - pretreated corn stover and CS 3.1.1. Feedstock sample collection C orn S tover , corn stover pretreated 2 with AFEX TM , and DM samples served as feedstocks for this study. CS samples were collected from the Michigan State University (MSU) Beef Cattle Teaching & Research Center in November 2015; AFEX TM samples were obtained from the Michigan Biotechnology Institute in October 2015, and DM samples were collected from the MSU Dairy Cattle Teaching & Research Center in January 2016. After collection the samples, CS and AFEX were stored at room temperature (~20 °C) in airtight bags, after which they were air dried and milled using a Willey Mill (Standard Model No. 3; Arthur H. Thomas, Philadelphia, PA). Finally, CS and AFEX samples were sieved through a 2 mm coarse mesh (No. 8, W.S. Tyler, Cleveland, Ohio) prior to use. Meanwhile, the DM samples were stored at 18 °C and thawed 2 days before the experiment. The AFEX samples were pretreated in a high - pressure Parr® stainless - steel reactor. For this purp ose, anhydrous liquid ammonia was added into the reactor at a 2:1 ratio (dry mater basis); the reaction was allowed to proceed for 30 min at a temperature of 102 °C and pressure 2.25 MPa . 3.1.2. Feedstock mixture preparation Two feedstock mixtures containing DM, one with CS and the other with AFEX, were prepared for the laboratory scale AD experiment. Both feedstock mixtures were prepared 2 to 5 days prior to use and stored at 4 °C. The feedstock mixtures were a 4:1 dry matte r ratio of DM to CS or AFEX . D istilled water was added to the mixtures to prepare a group with a TS content of 5%. The mass of manure needed was calculated using equation 3: 2 For the benefit of usage, AFEX TM is going to continue being called AFEX. 23 (3) where is the mass of raw manure needed to obtain a feedstock mixture containing 5% of TS (g), and is the TS content of raw manure (%). On the other hand, the necessary mass of CS or AFEX was calculated via equation 4: (4) where is the mass of CS or AFEX needed to obtain a feedstock mixture containing 5% of TS (g), and is the TS content of CS or AFEX (%). 3.1.3. S emicontinuous AD exp eriment Nine semi continuous feed completely stirred tank reactors (CSTRs) as anaerobic digesters with a liquid volume of 0.75 L were setup in triplicate for each feedstock mixture. Wheaton® bottles with rubber septa screw caps served as the CSTR vessels. The working volume of the digesters was 0. 7 5 L. Needles were used to puncture the septa to release the biogas. The biogas production was measured by the wate r displacement method. Figure 5 shows the configuration of the digestion unit. 24 Figure 5 . Flow direction of biogas and water in the water displacement method. Experiments were carried out on MaxQ 4000 incubator shakers (Thermo Scientific, Odessa, TX) at a temperature of 35 ± 0.5 °C and a shaking speed of 150 rpm. The hydraulic retention time (HRT) of the digesters was 20 days, and the duration of the digestion experiment was 75 days. Fifty milliliters of the AD effluent was discharged, and 50 mL of the feedstock mixture was fed every other day. pH was controlled in a range from 6.9 to 7.1 using a 30% (v/v) sodium hydroxide solution. All these operations were carried out in an ana erobic chamber (PLAS Lab, Lansing, MI), and the chamber was purged with a medical grade specialty gas (85% N 2 , 10% H 2 , 5% CO 2 ). A palladium catalyst heater was employed to ensure that the chamber was completely anaerobic. Two milliliters of the AD effluent were stored at 80 °C for microbial - community analysis. Ten milliliters of this effluent was used to quantitate TS and volatile solids (VS) ; the rest of the AD effluent was employed for quantitation of VFAs and structural carbohydrates. 25 3.1.4. Analytical methods 3.1.4.1. Gas Chromatography (GC) Biogas concentrations of CH 4 and CO 2 were measured on an SRI 8610C GC system equipped with a HayeSep® column and a thermal conductivity detector. Biogas sample s were collected at standard temperature from the bottle headspac e using a 5 mL Hamilton® syringe after gas production was recorded. Hydrogen (H 2 ) and helium served as carrier gases with pressure set to 145 kPa . The thermal conductivity detector was kept at a constant temperature of 150 °C. The injection volume was 3 mL with 100 µL transferred to the GC column. 3.1.4.2. TS/VS Ten - milliliter samples were used to measure the TS/VS ratio following the standard method (APHA, 1989) . The samples were dried for 24 h at 105 °C in a convection oven to quantify TS; the dried sample was then volatilized at 550 °C to obtain the VS. 3.1.4.3. National Renewable Energy Laboratory (NREL) Structural Carbohydrates Fiber content was measured by the NREL m (A. Sluiter et al., 2012) . Raw samples were dried at 45 °C in a food dehydrator (Tribest Sedona® SD - P9000). Then, low - concentration hydrolysis was performed in an autoclave, Getinge® 533LS. A Shimadzu® UV - 1800 spectrophotometer wa s employed to measure absorbance for lignin quantitation. The monosugars from the hydrolysis of cellulose and hemicellulose were quantified using a Shimadzu® HPLC system equipped with a Bio - Rad Aminex HPX - 87H analytical column and a refractive index detect or. The mobile phase was 0.005 mol/L sulfuric acid at a flow rate of 0.6 mL/min. The column temperature was set to 65 °C. 26 3.1.4.4. Concentrations of VFAs Ten milliliters of the AD effluent was centrifuged at 7,025 × g for 10 min in a Beckman® centrifuge (Allegra X - 12R, Beckman Coulter, Inc., Brea, CA) to collect the supernatant for measurement of VFA concentration. The supernatant was washed with 25% (w/w) metaphosphoric acid at a ratio of 1 to 5 (acid to sample) to remove remaining solids. The VFAs were quantified on a Shimadzu GC system (GC - 2010, Shimadzu Corp., Kyoto, Japan) equipped with a capillary column (122 - 3232 DB - FFAP, Agilent Technologies, Santa C lara, CA) and a flame ionization detector (Shimadzu Corp., Kyoto, Japan). Helium served as a carrier gas with the pressure set to 79 kPa . The injection volume was 10 µL with 1 µL transferred to the GC column. The column temperature was set to 150 °C for 2 min and raised to 220 °C at a rate of 15 °C/min, then maintained at 220 °C for 1 min. The temperatures of the injector and detector were set to 250 and 270 °C, respectively. The volatile free acid mixture (CRM46975, Sigma Aldrich, St. Louis, MO) served as the VFA standard. The acids quantified were acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, valeric acid, isocaproic acid, caproic acid, and heptanoic acid. A final VFA concentration was determined by summing all the concentra tions of the VFA profile described above. Due to the importance of the behavior of acetic (Hac) and propionic acid (Hpa), these were separately studied and plotted. 3.1.4.5. Carbon and N itrogen Raw materials dried at 45 °C were used to measure carbon and nitrogen c ontent. Total organic carbon (TOC) (TMECC 04.01 - A) and total nitrogen (TMECC 04.02 - D) were measured to calculate the carbon - to - nitrogen ratio (TMECC 05.02 - A) by the Test Methods for the Examination of Composting and Compost (USDA & CCREF, 2001). 27 3.1.5. Microbial C ommunity A nalysis 3.1.5.1. Amplicon preparation and sequencing procedures The PowerLyzer® PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Carlsbad) was used to extract DNA from the DM sample and DNA from the AD effluent of the CS and AFEX reactors at various time points. The DNA concentrations were measured on a NanoDrop Lite spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA). The DNA samples with the DNA concentration less than 25 ng/µL were concentrated using 5 M NaCl and cold ethanol (200 proof). The extracted DNA was stored at 80 °C before use as a template fo r amplicon preparation - CCTACGGGNBGCASCAG - - GACTACNVGGGTATCTAATCC - bacterial DNAs (Takahashi, Tomita, Nishioka, Hisada, & Nishijima, 2014) . Twenty - five microliters of th ( 1 ) ( ) , 0.5 (~ 40 ng/ ) - and RNase - free water for PCR. The PCR program started with a denaturing step at 95 °C for 5 min, followed by 30 cycles of the touchdown steps (denaturing at 95 °C for 2 min, annealing at 58 °C for 5 s, and elongation at 48 °C for 5 s), and ended with a final e xtension at 72 °C for 5 min. PCR products were loaded onto a Bio - Rad® 1% TAE Mini ReadyAgarose TM precast gel with 1% ethidium bromide and were visualized using an electrophoresis unit (Bio - Rad, Hercules, CA). After the PCR products showed correct bands on the electrophoresis gel (1% agarose, dyed with ethidium bromide) for all the samples, samples containing the original DNA template were analyzed at the Research Technology Support Facility at MSU. Then, at this facility, the V3 - V4 region (positions 341 806 ) of the 16S rRNA gene was amplified by nested PCR with a set of 28 primers designed to detect bacteria and archaea (Takahashi et al., 2014). First, primary PCR was carried out with chimeric primers containing target - specific portions (as described in Takahas carried out to add sequences necessary for Illumina sequencing and unique barcodes. The PCR products were normalized via Invitrogen SequalPrep DNA normaliz ation plates, and normalized eluates from the plates were pooled. After validation and quantification, a pool was sequenced in an Illumina MiSeq flow cell (v2) with a 500 - cycle reagent kit (2×250 bp paired - end reads). Custom sequencing primers matching the Fluidigm CS1 and CS2 oligos were used. Base calling was done by Illumina Real Time Analysis (RTA) v.1.18.54 software, and the output of RTA was demultiplexed and converted to FASTQ format in Illumina Bcl2fastq v.1.8.4. 3.1.5.2. Bioinformatics The FASTQ files from Illumina sequencing were analyzed with BION, a semi - commercial open - source package for microbial - community analysis from the Danish Genomic Institute, Aarhus, Denmark. Primer sequences were utilized to extract the paired sequences from the raw reads, and m inimum quality of 99% was set as a requirement for at least 14 of 15 bases for forward reads and 28 of 30 for reverse reads. A minimum length of 50 was imposed. Paired reads were joined where there was at least a 25 - base overlap and 85% similarity. Sequenc es were then filtered for length (250 minimum) and quality (99.6%), dereplicated, preclustered at 99%, and checked for chimeras by an algorithm unique to BION. Nonchimeric sequences were clustered at 99% stringency and a minimum length of 300. The sequence s were then matched to reference sequences using a K - mer length of 8 with a step size of 4 and were compared with the region 340 807 in RDP 11.04. The sequence similarities of each sample were converted to a taxonomic profile, using 29 the RDP taxonomy, and t he profiles were combined into abundance tables and counted as operational taxonomic units (OTU). 3.1.6. Evaluation of Digestion Performance To evaluate the performance of the digesters, it was assumed that the reactors are under steady - state conditions, where th e mass is being conserved and there is no change in stored mass with time. Under these assumptions, four performance parameters were determined to evaluate the digestion: biogas productivity, VS concentration reduction, and xylan and cellulose content redu ctions. Biogas productivity was determined using equation 5: (5) where is biogas productivity (m 3 of biogas/[kg VS]), is the accumulated produced biogas for each HRT (mL), is the total mass fed into the reactors around each HRT (g), and is the VS percentage in feed mixtures (%). Meanwhile, the VS content reduction was determined via equation 6: (6) where is the VS reduction in the reactor (%), is the VS percentage in feed mixtures (%), is the total mass of the AD effluent (g), and is the VS percentage in the AD effluent (%). The cellulose reduction was determined by means of equation 7: (7) 30 where is the cellulose reduction in the reactor (%), is the cellulose content (%) of feed mixtures (%), and is the cellulose content (%) in the effluent (%). The xylan reduction was determined using equation 8: (8) where is the xylan reduction in the reactor (%), is the xylan content (%) of feed mixtures, and is the xylan content (%) of the effluent. Because we assumed steady - state conditions (i.e., = ), the mass terms are cancelled out in equations 6 to 8, making the calculation of the performance parameters simpler. 3.1.7. Statistical analysis of microbial - community data & of digestion performance To understand the microbial ecological conditions of the reactors and their effect on the operation of the reactors, a series a statistical analysis was performed. First ANOVA was conducted, with the objective to find statistically significant differences among the operational - Diversity describes the structure of the community by itself, in the context of the study; this is the structure of each sample sequenced (Bolker, 2008) - diversity measures the turnover of species - diversity is the comparison between different reactor communities (Bolker, 2008) . The OTU table from the bioinformatics pipelines was employed to interpret the microbial data and correlate the microbial communities with digestion performance. The R statistical software (version 3.5.0) was used to carry out the analysis. Software code is presented in the Appendix. 31 Three data files the OTU table, taxonomy table, and metadata table were generated for R to run the analysis. All these tables were saved as tab limited TXT files. 3.1.7.1. ANOVA & Normality The Shapiro Wilk normality test was performed o n the digestion performance data transformed to a logarithmic scale (R function log ). The digestion performance data included biogas productivity, VS content reduction, VFA concentration, cellulose content in the AD effluent, xylan content in the AD efflue nt, and lignin content of the AD effluent. The R function shapiro.test was executed. A one - way ANOVA was then performed separately for each HRT (1, 2, & 3) using the R function aov . On the ANOVA results, the Tukey pairwise comparison was performed to find statistically significant differences between the various operational parameters via the R function TukeyHSD. 3.1.7.2. - Diversity By means of R libraries vegan (Oksanen et al., 2016) , phyloseq (McMurdie & Holmes, 2013) , MASS (Venables & Ripley, 2002) , and tidyverse (Wickham, 2009) , a detailed - diversity analysis was performed, to study the diversity in each sample. Several diversity indices (i.e. , ted using the diversity function. The sampling curve or richness curve was calculated using the rarecurve formula; this approach allows us to observe the change in diversity. Both formulas are a part of the vegan package. The sampling curve is a plot of th e species accumulation versus the sample size; the rate at which new species are added reflects useful richness diversity (Bolker, 2008) . 32 3.1.7.3. - Diversity A class object was constructed to unify all data at the experiment level using phyloseq() from library phyloseq . Before the samples were analyzed, the dataset was rarefied. Rarefication allows a researcher to normalize a dataset thereby enabling a comparis on of the relative species richness in a normalized base. Because the rarefication procedure requires random subsampling, a constant random number was set using set.seed(). This technique allowed to reproduce the results every time the same code was runnin g. The dataset was rarefied by means of the function rarefy_even_depth from phyloseq . Four hundred thirty OTUs were removed from the datasets thereby allowing me to normalize the abundance of all the datasets. The function merge_samples() was executed to m erge the duplicates of each run. The data were transformed to relative values using the function transform_sample_counts in phyloseq . The relative values served to compare relative abundance levels of individual taxa between different samples. Based on the normalized data, abundance bar plots were built. In addition, some ordination analysis was performed as well. Finally, to obtain a multivariate understanding of this effect in the microbial communities, nonmetric multidimensional scaling analysis was performed on the OTU data at the family level using the metaMDS() function from vegan . Bray distance was chosen as a dissimilarity index. Functions subset_taxa() and taxa_glom() were employed to cluster similar taxa at different levels from a domain to genu s. 3.2. Degradation of lignocellulosic feedstocks at ABPs 3.2.1. Influent and effluent sampling at ABPs Fresh influent and effluent samples were collected at three ABPs in Michigan. The samples were obtained on two farms (Farm A and Farm B) and from the MSU South Cam pus Anaerobic 33 Digester (SCAD) on a weekly basis. Farms A and B were sampled on May 22 and 30 and June 6, whereas the SCAD samples were collected on June 5, 11, and 19, 2018. The main characteristics of the sampled ABPs are listed in Table 2. The samples we re transported and stored at a temperature of 3 °C prior to the analysis at the Anaerobic Digestion Research and Education Center (ADREC) of MSU. Table 2 . Description of the ABPs sampled Name Reactor Units Effluent streams per reactor Feedstocks Type Farm A 3 1 Dairy manure/FOG CSTR Farm B 2 1 Dairy manure Plug - Flow SCAD 1 1 Dairy manure/FOG CSTR The ABPs have different reactor designs, effluents streams and configurations. For instance, Farm A and SCAD are CSTR reactors. For this type, the feedstock is introduced in tanks which are generally stirred by impellers or pumps (Doran, 1995) . The performance of these reactors is dependent of reaction kinetics and the HRT in which the reactor operates (Fogler, 1999) . The digester at Farm A consists of three reactor units each with a mass flow of feedstock of 604,74 m 3 per day . Therefore, one sample from each the influent and the effluent were sampled for each unit. Then, the samples were mixed using a volume ratio 1:1:1 in the laboratory. Regarding SCAD, this ABP consists of only one reactor tank and a single input output s tream; therefore, the samples obtained from the streamlines were used as is. On the other hand, Farm B is a plug - flow digester, in which the fluid is pumped through a pipe or tunnel where the feedstock reacts. Here, the chemical reactions proceed as the fe edstock travels through the reactor volume like a piston. Additionally, Farm B has two effluent lines (Table 2). Thus, in the case of Farm B, the composites were prepared at a volume ratio 1:1. In addition, jugs of 5 L with the composites were prepared and stored at 4 °C. 34 3.2.2. Analytical Methods Influent and effluent samples were characterized in terms of a variety of parameters. The samples were dried at 45 °C for 3 days. Then, the dry sample material was characterized f o llowing the method presented by Templeton, et al. , ( 2010) . Such parameters as cellulose, xylan, lignin, ash, protein content, water extractives (H 2 O.Ext), and ethanol extractives (C 2 H 6 O.Ext) w ere measured. Quantification of the extractives in the samples was based on the method - 510 - 42619 (A Sluiter, Ruiz, Scarlata, Sluiter, & Templeton, 2008) . The extraction procedures were performed in 6 00 mL Tall Form Kimble® Berzelius Beakers using a reflux system called Labconco® Crude Fiber Apparatus (Kansas City, MO); the process lasted for 2 h. For the aqueous extraction, 2 g of a raw dry sample was placed in 100 mL of deionized (DI) water mixed wit h 4 mL of 30% v/v thermostable Novozymes Thermamyl® 120 L (Franklinton, North Carolina). For the ethanol extraction, 1 g of a sample already water extracted was placed with 100 mL of 71.25% v/v ethanol and digested for 2 h again. After the samples were dig ested, they were transferred to 50 mL vials. Next, these samples were centrifuged at 4,427 × g on a HERMLE Z206A (Wehingen, Germany). After that, the supernatant was discarded, and the tubes were centrifuged again with 50 mL of DI water. This process was repeated twice. Finally, the precipitate and some of the supernatant were transferred to a 20 m L tube for drying and were dehydrated at 45 °C. In both extraction procedures, a small sample was dried at 105 °C to correct the final extractive content for moisture. Protein content was calculated using a nitrogen protein conversion multiplier of 6.25 (J. B. Sluiter, Ruiz, Scarlata, Sluiter, & Templeton, 2010) . Parameters such as COD, soluble chemical oxygen demand (sCOD), total Kjeldahl nitrogen (TKN), and ammonia (NH 3 ) were analyzed in the wet samples (Patel & Nakhla, 2006) . COD was 35 measured by an EPA - approved Hach method 8000. For measuring sCOD, the samples were centrifuged at 4427 x g and then passed through a 0.45 µm filter with the help of a vacuum pump. Next, the filtered samples served as input for COD Hach method 8000. TKN was quantified by the EPA - approved Hach method 10242. Ammonia analysis involved EPA - approved Hach method 10205 . The samples were analyzed at room temperature. All the described Hach methods involved a Hatch DRB200 reactor and Hatch DR5000 spectrophotometer (Loveland, Colorado). In addition, TS/VS, TOC, total organic nitrogen, structural carbohydrates, methane cont ent, and biogas were quantitated by the method described in subsection 1.1.4. 3.2.3. Biochemical Methane Potential (BMP) in ABP s influents BMP assays were performed on the samples from the ABP s influents. Considering the need for a relatively large amount of a d ry sample for fiber content analysis, a modified BMP method was adopted (Faivor & Kirk, 2011) . The samples were blended using a Nutri - Ninja Professional BL450 900 Watt without the addition of water (Hansen et al., 2004) . The blended samples were mixed with the digestion filtrate at a sample - filtrate VS ratio of 2:1. The mixtures contained 200 g of the filtrate and blended sample. DI water was u sed to bring the reactor volume to 1 L. Each sample was analyzed in triplicates . The control contained 200 mL of filtrate and 800 mL of DI water without the blended samples. The samples of the mixtures were mixed on a stir plate for 10 min and then poured into a 0.75 L graduated Wheaton® bottle. The bottles were sealed with Wheaton® screw caps that have a septum for poking needles to measure the gas production. The bottles were flushed with nitrogen at a constant flow rate of 750 mL min 1 for 15 min and incubated under mesophilic conditions (35 °C) on a Therm o ly ne Bigger Bill Oscillator shaker. After 2 h of incubation, the gas was released from the bottles. Then, the experiment was started, lasting for 50 36 days. Finally, once the experiment al data were obtained, biogas productivity was calculated using equation 9, usually called the raw method productivity: (9) where is the biogas productivity of the raw sample (L/[kg initial VS]), is the accumulated volume of the gas produced in the BMP (mL), denotes the average accumulated volume of gas produced by the control BMP (mL), is the VS of feedstock mixtures (mg/kg), and represents the sample mass added into the BMP (kg). 3.2.4. Statistical analysis of ABP data 3.2.4.1. ANOVA & Normality An R script was written in - house to analyze the dry matter, characterize the raw samples, and to model the possible effect of this characterization on biogas productivity. First, all the data were transformed to dry matter basis by dividing each obtained dat a point by the TS determined in the sample. The parameters studied were sCOD, COD, NH 3 , TKN, protein, TOC, H 2 O.Ext, C 2 H 6 O.Ext, cellulose, xylan, acid lignin, and biogas productivity. Then, the Shapiro Wilk test was performed with transformation to the loga rithmic scale for each parameter measured, with clustering of the samples for each biogas plant, and by taking a difference between the influent and effluent. Function shapiro.test was utilized to perform this analysis. Later, one - way ANOVA was performed separately for each parameter measured by means of the factors: function aov . Due to the high variation observed in the data, for significance, a p value of 0.1 was selected. On the ANOVA results, the Tukey pairwise comparison was performed to find 37 statistically significant differences in various operational parameters, using the TukeyHSD function . 3.2.4.2. Average percentage of a parameter decrease on dry matter In addition, an average percentage of reduction in each parameter was calculated via the average of the influent and effluent data, by means of equation 10: (10) where is the average content reduction percentage (%), is the average concentration of each substance (% dry matter), and is the average effluent concentration of the same substance (% dry matter basis). These values were used as inputs of the variance - ba sed sensitivity analysis. 3.2.4.3. Variance - based sensitivity analysis To gain a deeper insight into the impact of the characteristics of the influents and their decrease on the BMP biogas productivity, a sensitivity analysis was performed assuming a linear model a nd with calculation of the covariance matrix of the reduction samples. The model assumed noninteraction among the parameters. The description of the model is given in equation 11: (11) where is the biogas productivity obtained in the BMP experiments (L/[kg initial VS]), is the average decrease in sCOD (%), is the average decrease in COD (%), is the average decrease in NH 3 content (%), is the average decrea se in TKN content (%), is the average decrease in protein content (%), is the average reduction in TOC content (%), is the average reduction in (H 2 O) Ext (%), is the average reduction in 38 (C 2 H 6 O) Ext co ntent (%), is the average decrease in cellulose content (%), is the average decrease in xylan content (%), and is the average decrease in acid lignin content (%). The programming code for the analysis is pr esented in Appendix 7.4 in part 9 of the Rscript developed. The code was written via Oracle Crystal Ball specifically , and this analysis was performed using the newly developed function contribution_to_variance() 3 . Here, the variance contribution is calculated by squaring the rank correlation coefficients and is normalized to obtain 100%; these results are only an approximation and are not precisely variance decomposition. The rank correlation shows that positive coefficients indicate an increase in the assumption, whereas negative coefficients imply the opposite situation. The larger the absolute value of the correlation coefficient, the stronger is the relation. Finally, the constructed plot showed a contribution and rank correlation. 3 http://mattgrogan.info/stats/contribution - to - variance/ 39 4. RESULTS T he results section will be described in two subsections (4.1 and 4.2). The first subsection describes the evaluation of the codigestion of manure with the two kinds of corn stover , CS and AFEX . Here, readers can see a description of feedstock mixtures and their characterization, the digestion performance observed, and the details about the microbial community established during both kinds of digestion. On the other hand, in subsection 4.2, readers will see dry - matter and raw character ization of the ABPs studied, the results of the BMP experiments performed with the ABP influents, and the sensitivity analysis of biogas productivity on the basis of the reductions observed in the parameters being measured. 4.1. Dynamic microbiome assembly and the effect of the performance of AD of AFEX - pretreated corn stover and CS 4.1.1. Feedstock characterization and codigestion mixture mass ratios Table 3 summarizes the characteristics of DM, CS, and AFEX used in the study. It is app arent that the TS concentrations of the CS and AFEX samples are much higher in comparison with DM samples. This phenomenon is mainly caused by rumen digestion. DM is the digested residue of plant biomass and other nutrients in the animal feed. These three feedstocks all contain significant amounts of cellulose, xylan, and lignin and are considered lignocellulosic biomass. Among them, CS has the highest cellulose content ; at 37.30%, AFEX has the highest hemicellulose content (21.73%), and DM has the highest lignin concentration (21.86%). Meanwhile, DM has a much lower C/N ratio (18.2) than CS and AFEX do. This result makes it possible to mix DM with CS and AFEX to obtain feedstock mixtures with desired C/N ratios (between 15 and 30) for healthy and efficient AD. In addition, DM has a much higher moisture content of 85.55% than do CS 40 (7.40%) and AFEX (2.64%). Mixing DM, CS, and AFEX can significantly reduce the demand for water needed to carry out the digestion. Table 3 . C haracteristics of assay feedstocks . Characteristic DM CS AFEX Raw Total solids (%) 14.45 ± 0.59 92.60 ± 0.19 92.68 ± 0.14 Volatile solids (% raw TS) 88.40 ± 0.64 95.66 ± 0.23 97.36 ± 0.07 TOC (% raw TS) 37.6 46.61 ± 0.16 46. 40 ± 0. 04 Nitrogen (% raw TS) 2 . 07 0.51 ± 0. 08 1. 55 ± 0. 0 8 C/N 1 8 .2 91.39 29.94 Cellulose (% TS) 22.63 ± 0.27 37.30 ± 0.24 28.96 ± 1.60 Xylan (% raw TS) 9.29 ± 0.32 19.61 ± 0.55 21.73 ± 0.21 Lignin (% raw TS) 21.86 ± 1.14 17.62 ± 1.06 18.44 ± 0.68 4.1.2. Digestion Performance Figure 6 summarizes different patterns of biogas production, methane content, VS content reduction, cellulose/xylan content reduction, and VFA content as parameters of the anaerobic codigestion of two feedstock mixtures (CS and AFEX) . During th e 75 - day semi - continuous digestion, three stages: 1 (days 0 20), 2 (days 21 40), and 3 (days 41 75), were chosen based on the HRT of 20 days to investigate dynamic changes during the anaerobic codigestion. a. b. Figure 6 . Digestion performance of AFEX and CS during the digestion (three HRTs). a. Biogas productivity, b. methane content, c. VS content reduction, d. cellulose content reduction, e. xylan content reduction, f. total VFA concentration. 41 Figure 6 c. d. e. f. The experimental data revealed that the factors of feedstock and digestion time had a significant ( p < 0.05) influence on biogas production (Table 3), and both digestion procedures had a lag phase in biogas production (Figure 6a). The lag phases for the CS and AFEX - pretreated corn stover digestion were approximately one and two HRTs, respectively. Biogas production by the CS codigestion rapidly increased to 225 ± 5 L/[kg VS loading] per day in the 2 nd HRT relative to the 1 st HRT (91 ± 5 L/[kg VS loading] per day), and then leveled off in the 2nd HRT to reach stable biogas production of 175 ± 8 L/[kg VS loa ding] per day (Figure 6a). As for the AFEX - pretreated corn stover codigestion, biogas production kept increasing from 79 ± 4 L/[kg VS loading] per day in the 1 st HRT gradually to 130 ± 6 L/[kg VS loading] per day, and then stabilized 42 at 213 ± 0 L/[kg VS lo ading] per day in the 3 rd HRT. In the stabilized 3 rd HRT, the AFEX - pretreated corn stover codigestion indicated significantly ( p < 0.05) higher biogas production in comparison with the CS codigestion. In terms of CH 4 content, feedstock and HRT had a signi ficant ( p < 0.05) impact on methane content in the biogas during the digestion (Table A1). Both codigestion procedures started at slightly but significantly ( p < 0.05) higher methane contents (62% ± 1% and 65% ± 2% for the AFEX - pretreated corn stover and C S digestion reactions, respectively) in the 1 st HRT than the subsequent HRTs (Figure 6b). Methane contents stabilized at 61% ± 1% and 62% ± 0% for the corresponding digestion reactions in the 3 rd HRT without a significant ( p > 0.05) difference between the two digestion groups (Figure 6b). The variation observed in the content of the biogas samples is normal during the establishment of methanogenic communities. As for the VS content reduction, it was similar to VFA in that feedstock and digestion time genera lly had no significant ( p > 0.05) influences on the VS content reduction in both digestion reactions (Table A1). However, the VS content reductions largely fluctuated in the first two HRTs (Figure 6c). After the 1 st HRT, the AFEX - pretreated corn stover cod igestion yielded a greater VS content reduction than did the CS codigestion, particularly in the 2 nd HRT where the VS content reduction of the AFEX - pretreated corn stover codigestion (47% ± 5%) was significantly higher than that (31% ± 1%) in the CS codige stion. This observation along with the higher biogas productivity of the AFEX - pretreated corn stover codigestion could be explained by the chemical and structural changes of the corn stover during the AFEX treatment. Anhydrous ammonia in the AFEX pretreatm ent reacts with feruloyl or coumaryl ester bonds in biomass to form amides, especially acetamide (Chundawat et al., 2013; Chundawat, Donohoe, et al., 2011) . Guyot et al. demonstrated that amide compounds can be easily degraded by anaerobes and methanogens 43 ( Guyot, Ferrer, & Florina, 1995) . Accordingly, the AFEX - pretreated corn stover codigestion manifested more efficient biogas formation from VS than did the control CS codigestion. Both factors digestion time and feeds tock generally had no significant ( p > 0.05) impact on cellulose and xylan content reduction, except that the digestion time was a significant factor for cellulose content reduction. However, high variations of cellulose and xylan degradation were observed during AD (Figure 6d & 6e). The AD of the AFEX - pretreated corn stover yielded significantly ( p < 0.05) higher cellulose and xylan content reductions in the 1 st HRT than that of the CS digestion. After the 2 nd HRT, the differences in cellulose and xylan co ntent reductions between the two codigestion experiments were not significant ( p > 0.05). The original mixtures of both codigestion reactions had higher concentrations of fresh AFEX - pretreated corn stover and CS in the reactor at the beginning of the diges tion (the 1 st HRT). Accordingly, more cellulose and xylan from AFEX - pretreated corn stover and CS were released into the reactors. Because the cellulose and xylan in the AFEX - pretreated corn stover were relatively easy to digest by microbes owing to the lo ose carbohydrate lignin bonds (V Balan et al., 2012) , the larger amount of the AFEX - pretreated corn stover in the 1 st HRT led to significantly greater cellulose and xylan content reductions than did the CS digestion. After the 1 st HRT, 20% of AFEX - pretreated corn stover and CS in the feed for codigestion might be too little to reveal significant impacts on the overall ce llulose and xylan content reductions (Figure 6d & 6e). The VFA data further verified the performance patterns of these two digestion reactions (Figure 2f). A large variation (2.6 ± 2.7 g/L) between replicates of the AFEX corn stover digestion and a high VF A concentration (4.3 ± 0.8 g/L) of the CS digestion indicated unstable digestion during the 1st HRT. With progression of the digestion, the VFA concentrations stabilized at 2.8 ± 44 0.1 and 3.4 ± 1.4 g/L for the AFEX - pretreated corn stover and CS digestion re actions, respectively (Figure 2f). In Figure 7, the concentration of two VFAs being quantified is plotted: Hac and Hpa. On average, the sum of the concentrations of these two VFAs represents 80% of the total acid - producing fermentation measured in the eff luents. The ANOVA results (Table A1) suggested that in the 2 nd HRT, the concentration of Hac was significantly lower (p = 0.01) in group AFEX (0.41 g/L) in comparison with group CS (1.82 g/L). Figure 7 . Average acetate (Hac) and propionic acid (Hpa) concentrations in the reactors. On the other hand, Hpa concentration did not show significant differences between the different digestion reactions and at the same time did not undergo any reduction throughou t the whole experiment. The concentrations of these two acids could be having a strong effect on biogas 45 productivity and on microbial ecology. Wang, Zhang, Wang, & Meng, (2009) rep orted that 0.9 also found that this inhibition resulted in the accumulation of acetate and affected the total methane yield. By contrast, such accumulation was n ot seen in the AFEX reactors according to Figure 7, and the concentration of acetate was significantly lower in the AFEX reactors in comparison with CS reactors. Moreover, Chundawat, Beckham, et al., (2011) described how the AFEX pretreatment produces a series of ammonolytic and hydrolytic reactions that cleave various ester linkages, thereby resulting in the formation of amides and acids such as acetate. This evidence may confirm a possible influence of AFEX pretreatment on the acetate formation and consumption by codigestion in the AFEX reactors. It is important to remember that the group of methanogens are acetate reducers, and the concentratio n of acetate will have a major impact on the growth of this group and on methane formation (Rittmann & McCarty, 2001) . 4.1.3. Microbial Community Analysis Figure 8 illustrates the evenness and richness diversity of the microbial species in each reactor - diversity). The average relative abundance of microbial species indicates a good fit to the lognormal distribution of the microbial communities and reveals high rarity of the species (Figure 8a). Similar evenness was observed during codigestion of different feedstock mixtures elsewhere (R. Chen et al., 2016; Rojas - Sossa et al., 2017) . The sampling richness curves for each digester are presented in Figure 8b. The diversity of the microbial communities in the codigestion experiment was significantly dif ferent from that of the inoculum. The shifts could be caused by changes of nutrients in the feedstock mixture as well as introduction of new microbial species in groups CS and AFEX (Kirkegaard et al., 2017) . 46 a. b. Figure 8 . Diversity of microbial communities in both digestion reactions. a. The rank abundance (Whittaker) plots of relative abundance of OTUs in both digesters. The dots represent the logarithmic percentage of the relative abundance of each species, and then the lognormal curve was plotted on the data. b. Examples or the diversity curves seen in the digesters. Figure 8b also indicates that during the stable digestion performance (the 3 rd HRT), microbial communities of the AFEX codigestion reaction were slightly more diverse than those of 47 the CS codigestion reaction. This is because the sampling curves showed higher steepness in the 3 rd HRT. This finding indicates that AFEX - pretreated corn stover allowed for digestion to maintain a higher number of microbial species and a large number metabolic fluxes could be happening inside the reactors (Colwell & Rangel, 2009) . The diversity indices of the digestion reaction for three HRTs are presented in Figure 9. Figure 9 . Ecological diversity indices (Shannon, Simpson, Inverse Simpson, and Fisher) for each AD reactor. Much larger variation in Shannon, Simpson, and Inverse Simpson indices for the AFEX replicates was observed in the 1 st and 2 nd HRTs as compared to the CS replicates. The variation significantly diminished in the 3 rd HRT. The results once again show that AFEX certainly has a bigger impact on the microbial communities of AD. Meanwhile, the Fisher diversity index indicates a large d ifference between the seed and digestion samples, consistently with the 48 rarefaction curve (Figure 8b), where the seed diversity was much lower than that of the digestion samples. Figure 10 shows the relative abundance of AFEX and CS samples at four taxonom ic levels. At the domain level, as presented in Figure 10a, the total relative abundance of archaea was higher in the AFEX digesters than CS reactors in the 2 nd and 3 rd HRTs. The data suggest that AFEX codigestion has a potential to enrich the archaea popu lation. The distribution of dominant phyla (Bacteroidetes, Proteobacteria, Spirochaetes, and Verru co microbia) is depicted in Figure 10b. These phyla have been detected in other digestion studies. The phylum Bacteroidetes of AFEX and CS digesters showed sig nificant increases in its relative abundance as compared to the seed sample. Considering the high carbohydrate content in groups AFEX and CS, Bacteroidetes as carbohydrate - degrading microbes correspondingly increased in number at the beginning of the diges tion to satisfy the need for nutrient utilization to support healthy digestion. With progression of the digestion, stable digestion was achieved, and balanced communities formed correspondingly. The abundance of the phylum Bacteroidetes decreased. In contr ast to Bacteroidetes, phyla Verru co microbia and Proteobacteria did not undergo enrichment relative to the seed. Both decreased in abundance with the digestion duration. This phenomenon may be caused by the increased carbohydrate content and reduced protein amount of AFEX and CS feedstocks. Of note, the phylum Spirochaetes got enriched during the digestion of AFEX and CS (Figure 10b). In the 3 rd HRT, the AFEX digesters showed much higher relative abundance of these genera compared to other genera. The rich n onrecalcitrant carbohydrates of the AFEX reaction could have played a key role in this shift. Two main genera from the phylum Spirochaetes were found: Treponema and Sphaerochaeta (Figure 10c). Turroni et al., (2016) detected a significant 49 increase in the abundance of Tre ponema in human gut microbiomes of hunter - gatherers who lived on high - plant - fiber diets. The relative abundance of archaeal genera is illustrated in Figure 10d. Figure 10 . Relative abundance of different taxa found in the reacto rs. a. Relative abundance of the microbial domains. b. Relative bacterial phylum abundance. c. Relative Spirochaetes genera abundance. d. Relative Archaea genera abundance. Methanosarcina , Methanocorpusculum , and Methanobrevibacter were three dominant archaea in both AFEX and CS digestion reactions. Methanocorpusculum and Methanosarcina got enriched with the digestion duration. Greater enrichment of Methanosarcina was present in the AFEX digestion reaction than in the CS digestion, whereas more Methanocorpusculum was present in the CS digestion reaction than in the AFEX digestion reaction, however total archaea enrichment in group CS was lower than that in group AFEX (Figure 10a). On the other hand, the genus Methanobrevibacter showe d a decrease in the relative abundance with an increase in HRTs. 50 This genus is known as a hydrogenotroph that oxidizes hydrogen to produce methane. This could be happening because of propionate accumulation, which could be affecting the growth of this grou p. Propionate is an important intermediate during AD, and this degradation produces Hac, H 2 , and CO 2 (Li, Ban, Zhang, & Jha, 2012) . In addit ion, the genus Methanobacterium manifested good enrichment in the AFEX digestion reaction as the digestion progressed. The changes in relative microbial abundance of this Archaea genus are consistent with the observed performance and the structural charact eristics of lignocellulosic materials as the feed. The microbial community analysis led to the conclusion that the AFEX reaction significantly enriched Archaea communities during the digestion, and accordingly, biogas productivity significantly increased. Finally, nonmetric multidimensional scaling visualization of the microbial abundance was conducted to elucidate the relations between microbial communities and digestion performance; the results are presented in Figure 11. The visualization uncovered impo rtant correlations among microbial communities, biogas productivity, feedstocks, and xylan/VS content reduction. A similar trend has been observed in other similar studies. There was a possible inverse linear correlation between the AFEX biogas productivit y and Hac metabolic fluxes in the reactor. 51 Figure 11 . Nonmetric multidimensional scaling of the relative abundance of microbial communities in the digesters. 52 4.2. Degradation of Lignocellulosic Feedstocks at ABPs 4.2.1. Influent and effluent dry matter and characterization of raw samples Figure 12 is a plot of the average influent and effluent composition observed at different ABPs. a. b. Figure 12 . a . Influent and b . effluent composition analysis on a dry - matter basis. In Figure 12a, the relative composition of the influents of biogas plants. Table A2 shows a pairwise comparison for this analysis. According to the data, there were significant differenc es in the protein content and ethanol extractives. The SCAD protein content was significantly lower in 53 comparison with Farm A (p = 0.002) and Farm B (p = 0.014). In contrast, the amount of SCAD ethanol extractives was significantly higher than that on Farm A (p = 0.09); there were no significant differences in the pairwise comparison SCAD Farm B. Meanwhile, there was some variation of this composition, and we found very similar data on lignocellulosic content: 12% cellulose, 5% xylan, and 18% lignin. These values are lower than the ones EPA recommends as AD feedstocks: 22% cellulose, 36% hemicellulose, and 21% lignin (EPA, 2014; Sun & Cheng, 2003) . Most of the ABPs studied practice recycling of the effluent. The digestate is normally filtered in a liquid solid separator and is mixed with a fresh influent ; then , it is fed into the reactor. This procedure increases alkalinity of the influent and maintains healthy digestion. This procedure may dilute the influents and could be responsible for low content of plant cell wall components at the ABPs studied. On the oth er hand, in the case of effluents, more significant differences were found (Table A2). Significant differences were found in protein, water extractives, ethanol extractives, and lignin content of the effluent samples. As for protein, we found again a lower concentration in the SCAD reactor in comparison with Farm A (p = 0.02), suggesting possible major consumption of protein by Farm B in comparison with the other two biogas plants. Moreover, H 2 O.Ext content was significantly lower at the SCAD ABP in compari son with the Farm B effluents (p = 0.07). The C 2 H 6 O.Ext concentration of the effluents showed significant differences between SCAD and Farm A (p = 0.04); if this parameter is compared between influents and effluents, any reactor showed substantial consumpt ion of this substance . Significant differences were found in lignin content between SCAD and Farm B (p = 0.05); this pattern is suggestive of possible storage of lignin in the Farm A and B reactors, and this storage seems to be higher in the Farm A reactor . Yue et al. (2013) reported that AD codigestion of DM 54 normally can homogenize the AD effluents, producing very similar concentrations of plant cell components relative to one another in the AD effluent. Finally, ash content was evaluated here: all the rea ctors had similar contents of ash in the effluent, which contained ~2% of dry matter. Figure 13 presents plots of the proportions (%) of sCOD, COD, NH 3 (ammonia), and TKN in the influent samples and effluent samples from the three ABPs sampled. In Table A 2, pairwise comparisons of these raw parameters are detailed. There was higher COD relative content in the SCAD reactor in comparison with Farm A (p = 0.06) and Farm B (p = 0.06), and this concentration was similar between Farms A and B. As Figure 13A show s, effluent sCOD concentration was lower on Farm A than at SCAD (p = 0.09). On the other hand, Figure 13B shows the averages of nitrogen forms quantified in the streams: NH 3 and TKN. NH 3 content of influents was higher on Farm B than at SCAD (p = 0.08). NH 3 content of effluents was higher at SCAD than on Farm B (p = 0.02), indicating a possibly higher ammonification rate in the SCAD reactor in comparison with Farm B. As for Farm A, almost nonexistence of ammonification was found. The TKN content of effluent s was higher on Farm A than on Farm B (p = 0.02). TKN concentration (Figure 13B) increased in all the reactors, and on Farms A and B, this could be happening because these ABPs had a higher protein concentration in the diet. 55 a. a1. a2. b. b1. b2. Figure 13 . Raw sample characterization. a. Chemical oxygen composition of a 1. influents, a2. influents. b. Concentrations of nitrogen forms: b1. influents, b2. effluents . 56 Finally, the last parameter studied was TOC; ANOVA analysis uncovered significantly higher TOC contents in Farm B (p = 0.08) and SCAD (p = 0.005) influents in comparison with Farm A influents ; significant differences were not found in the effluent samples. Figure 14 depicts the variation of this parameter in the scatterplot on the different dates of sampling. These data indicate how the only reactor that showed clearly substantial consumption of TOC throughout the dates of sampling was the SCAD reactor. Farm A and Farm B influents fluctuated a lot throughout the experiment; actually, very similar TOC influent and effluent concentrations were observed. Figure 14 . TOC variation in the dry matter of the influents and effluents throughout the sampling period. Table 4 presents an average reduction in each parameter across the sampling time points for the different ABPs. All the data showed degradation of sCOD, COD, protein, and TOC. By contrast, almost all the ABPs showed a n increase in NH 3 , TKN, H 2 O.Ext, and C 2 H 6 O.Ext concentrations . Finally, in terms of decreases in cellulose, xylan, and lignin concentrations, the 57 reactors showed differences among the different ABPs. Farm A did not show a reduction in the amount of plant cell wall components; in contrast, Farm B manifested a stronger decrease in cellulose (11% ± 6%) and xylan ( 18% ± 3%) and lignin . Table 4 . Decreases in the averages dry matter content of each parameter in the reactors. Percentage of decrease Variables Farm A Farm B SCAD 53% ± 11% 44% ± 31% 42% ± 16% 14% ± 9% 7% ± 21% 11% ± 62% - 14% ± 1% 34% ± 2% - 208% ± 3% - 206% ± 9% - 33% ± 2% - 102% ± 7% 8% ± 2% 6% ± 1% - 10% ± 3% 0% ± 3% 1% ± 2% 13% ± 8% 17% ± 30% - 6% ± 28% - 6% ± 10% - 26% ± 7% - 10% ± 5% - 13% ± 16% - 9% ± 9% 11% ± 6% 4% ± 11% - 9% ± 5% 18% ± 3% - 3% ± 5% - 23% ± 9% - 1% ± 7% 3% ± 6% Finally, SCAD showed a reduction in cellulose ( 4% ± 11%) and lignin ( 3% ± 6%) contents but yielded an increase in xylan concentration (negative reduction: 3% ± 5%). It should be amounts seen in and . 4.2.2. BMP Experiments on ABP influents The BMP experiments revealed a significant difference in biogas productivity from the influents between SCAD and Farm A, with SCAD being significantly better than Farm A (p = 0.06): almost threefold higher biogas pr oductivity in comparison with Farm A. Nonsignificant 58 differences were noted between Farm B or SCAD and Farm A BMP. Figure 15 is a plot of the accumulation of gas production in the reactors; it is not possible to detect any inhibition at the beginning of th e experiment. The biogas productivity obtained was 278 ± 68, 390 ± 117, and 659 ± 71 L/[kg initial VS] for Farm A, Farm B, and SCAD, respectively. a. Figure 15 . BMPs: accumulated gas production on different dates of sampling of influents. a. Accumulated gas production of Farm A, b. accumulated gas production of Farm B, c. accumulated gas production of the MSU south campus digester. 59 Figure 15 b. c. 60 4.2.3. Variance - based sensitivity analysis of the downregulation of each compound and the effect on biogas productivity The objective of this analysis was to explain the biogas productivity observed in the BMP with the data on downregulation of substances obtained above. The input of the model (equation 11) is the data on concentration reductions (Table 4). Figure 16 illustrates the output of the analysis. The contribution to the variance was ranked for each parameter. This appr oach allows us to infer possible sensitivity of biogas productivity to each parameter. There were possible positive correlations with , , , , and . On the other hand, there was a possible ne gative correlation of biogas productivity with , , , and . The magnitude of a correlation is dependent on the absolute value of the contribution to the variance. It can be noted that there were greate r correlations of biogas productivity with (17.8%) and ( 18.1%), relative to the other parameters, with showing a positive correlation and a negative one. These two parameters are direct measurements of c arbon (TOC) and nitrogen (protein). The ABPs have different diets namely Farm A and Farm B diets contain a low concentration of carbon and high content of protein. On the other hand, SCAD nutrition contains more TOC in comparison with the other two ABPs st udied. Then, the other parameters (with lesser correlation) were ranked: (12.5%), (6.6%) , (3.7%), and (2.9%). The negative ranking was ( 9.9%), ( 13.0%), and ( 14 .7%). These results can be explained by the high productivity observed at SCAD in comparison with the other ABPs. SCAD had a higher content of lignin and C 2 H 6 O.Ext in the influents and effluents and manifested substantial production of NH 3 (Table 4). 61 Figure 16 . The ranking of contributions to the variance of biogas productivity by each of the parameters measured. Moreover, the negative correlation could be explained by the different reductions in concentrations of substances among the different ABPs. For example , there is a substantial correlation with and ; these phenomena could be due to the activities on Farm A, where production of NH 3 was important and consumption of H 2 O.Ext was important too. Nonetheless, this ABP s also has lower biogas productivity. Finally, regarding , most of the biogas plants studied had very similar values but different biogas productivity; therefore, this parameter is not important for the explanation of differences in bio gas productivity. 62 Finally, the importance of for biogas productivity was equal to 0%, indicating insignificance change in the amount of this plant cell wall component for biogas productivity observed at the ABPs under study. 63 5. CONCLUSIONS The results show a big picture of actual stages in the use of plant cell wall components as an energy source in AD. Presented below are specific conclusions from both research projects described above. 5.1. Dynamic microbiome assembly and the effect on the performance of AD of AFEX - pretreated corn stover and CS The digestion performance and the microbial community analysis indicate that AFEX - pretreated corn stover promotes a positive linear correlation betwee n a reduction in cellulose content and biogas productivity. AFEX does not promote degradation of lignocellulosic materials in the reactors studied ; rather , the impact of AFEX manifest ed itself in chemical reduction of acetic acid and its production rate in the reactors. AFEX codigestion promotes the enrichment with Methanosarcina and possibly increases the role of acetic acid as an electron acceptor because Methanosarcina is an acetic acid reducer. The tested CS mixtures promote the enrichment with acid fer menter Treponema ; this genus has been proven to get enriched with high lignocellulosic inputs. There is possible migration of various bacteria found in the reactors, which is important. However, there is no migration for Archaea genera. The enrichment could be a consequence of AFEX codigestion. 64 5.2. Degradation of lignocellulosic feedstocks at the ABPs There is big variability of the inputs of the digesters sampled. The results revealed positive and negative effects on biogas productivity. Positive influences were seen where carbon concentration diminished, and negative influences were detected where nitrogen was abundant in the diet. The ABPs studied have different diets, where Farms A and B show low content of carbon and high concentration of protein in the diet. On the other hand, SCAD has high TOC concentration in comparison with the other two ABPs and is the most productive plant among those studied. R anking the AD influents constituents that correlate with biogas producti vity. Furthermore, it was revealed that the recalcitrant carbon in ADP influents does not contribute to biogas productivity. Finally, at the biogas plants sampled, there is no evidence of a possible contribution to biogas productivity from degradation of lignocellulosic materials: most of the contribution results from destruction of carbon - based materials that are not structural carbohydrates. 65 6. RECOMMENDATIONS To improve the results of operation of the AD reactors, it is recommended to operate semicontin uous AD for five or six HRTs and with a DM control to measure against a true baseline . In this case, it will be possible to attain a bigger change in the dominate microbial taxa and at the same time to observe uniformity in the characteristics of inputs an d outputs. It will also be important to examine these results with a control setup that involves manure digestion. Feedback from the control will enable better monitoring and observation of the metabolic fluxes that take place inside the reactors. Furtherm ore, a manure control will allow investigators to examine the seed dynamics and the positive effect of corn stover on the diversity of possible enrichment with microbial immigrants fed in with the feedstock from outside. It is strongly recommended to expan d research into possible lignocellulosic bacterial fermenter spirilla like Treponema . These studies may allow researchers to use this genus as a possible indicator of healthy lignocellulosic AD. Besides, it is important to investigate more thoroughly the relation between Treponema enrichment and AFEX TM digestion and their acid formation e ffects. To decrease the high variation observed at the ABPs sampled, it will be advisable to expand the study to more biogas plants and more sampling dates. Moreover, it will be useful to study this degradation in a more controlled model, for example, in a BMP reactor, where it is possible to change the number of experimental conditions and create triplicates. These two recommendations will decrease the observed statistical insignificance of the results. 66 APPENDIX 67 APPENDIX RSCRIPT for microbial community analysis of digestion from a n AD reactor ## Dynamic microbiome assembly in performance ## of co - anaerobic digestion of AFEX corn stover ## MICHIGAN STATE UNIVERSITY ## ADREC ##Version 2.0 ## Using as inputs 16S OTU tables and Perfo rmance reactors data ## Made by Juan Pablo Rojas, 2018 cat ( " \ 014" ) #Erase console # 1. LOADING LIBRARY AND TABLES ------ library (vegan) library (phyloseq) library (MASS) library (ggplot2) library (grid) library (gridExtra) library (readr) library (VennDiagram) library (tidyverse) library (Rmisc) library (gtable) library (devtools) library (proto) library (reshape2) ##Choose the OTU.Table should be a .txt con < - file.choose ( new = FALSE ) ##Now choose the Taxanomy table should be .txt to con1 < - file.choose ( new = FALSE ) OTU_Table < - read.table (con, header = T, row.names = 1 ) OTU_Table_taxonomy < - read.delim (con1, header = T, row.names = 1 ) metadata < - read.delim ( "~/Thesis/ADonAFEXfiber(2016)/DNA - data/No_manure/Metadata_v2_no_seed.txt" , row.names= 1 ) Methane < - read_delim ( "Methane.txt" , " \ t" , escape_double = FALSE , trim_ws = TRUE ) # 2 NORMALITY TEST FOR EACH DATA SET ---- metadata $ Lignin < - NULL #CONVENTIONAL CORN STOVER for (i in 4 : 10 ) { print ( "#CONVENTIONAL CORN STOVER" ) fit < - metadata %>% filter (Mix == "Conventional Corn Stover" ) print ( colnames (fit[i])) print ( shapiro.test ( log ( unlist (fit[,i])))) } 68 #AFEX CORN STOVER for (i in 4 : 10 ) { print ( "#AFEX CORN STOVER" ) fit < - metadata %>% filter (Mix == "AFEX Corn Stover" ) print ( colnames (fit[i])) print ( shapiro.test ( log ( unlist (fit[,i])))) } # 3 ANOVA TEST OF THE PARAMETERS ---- metadata $ Mix < - factor (metadata $ Mix) ##Factor Statement #Biogas_Productivity for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (Biogas_Productivity)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #VS_Reduction for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (VS_Reduction)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #VFA for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (VFA)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #Hac for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (Hac)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #Hpa for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (Hpa)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #Cellulose for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) 69 print (i) ANOVA< - aov ( log ( unlist (Cellulose)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #Xylan for (i in 1 : 3 ) { fit < - metadata %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (Xylan)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } #Methane for (i in 1 : 3 ) { fit < - Methane %>% filter (Mix == "AFEX Corn Stover" ) print ( shapiro.test ( log ( unlist (fit $ CH4)))) fit < - Methane %>% filter (Mix == "Conventional Corn Stover" ) print ( shapiro.test ( log ( unlist (fit $ CH4)))) fit < - Methane %>% filter (HRT == i) print ( "HRT" ) print (i) ANOVA< - aov ( log ( unlist (CH4)) ~ Mix, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results } # 4. PLOTTING PERFORMANCE PARAMETERS VS HRT ----- metadata < - read.delim ( "~/Thesis/ADonAFEXfiber(2016)/DNA - data/No_manure/Metadata_v2_no_seed.txt" , row.names= 1 ) #Gas Production Gas_prod < - read_delim ( "Gas_prod.txt" , " \ t" , escape_double = FALSE , col_types = cols ( Date = col_date ( format = "%m/%d/%Y" )), trim_ws = TRUE ) gasprod< - ggplot (Gas_prod, aes (Date, Gas, color= Mix)) + geom_point ( aes ( shape= Mix, color= Mix)) + labs ( x = "Date" , y= "mL" , title= "a. Biogas Production" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "none" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) gasprod metadata< - metadata[ - 13 , ] ##Calculation of summary of the data tgc_Biogas < - summarySE (metadata, measurevar= "Biogas_Productivity" , groupvars= c ( "Mix" , "HRT" )) tgc_VS < - summarySE (metadata, measurevar= "VS_Reduction" , groupvars= c ( "Mix" , "HRT" )) tgc_VFA < - summarySE (metadata, measurevar= "VFA" , groupvars= c ( "Mix" , "HRT" )) tgc_Hac < - summarySE (metadata, measurevar= "Hac" , groupvars= c ( "Mix" , "HRT" )) tgc_Hpa < - summarySE (metadata, measurevar= "Hpa" , groupvars= c ( "Mix" , "HRT" )) tgc_Cellulose < - summarySE (metadata, measurevar= "Cellulose" , groupvars= c ( "Mix" , "HRT" )) tgc_Xylan < - summarySE (metadata, measurevar= "Xylan" , groupvars= c ( "Mix" , "HRT" )) tgc_Lignin < - summarySE (metadata, measurevar= "Lignin" , groupvars= c ( "Mix" , "HRT" )) tgc_CH4 < - summarySE (Methane, measurevar= "CH4" , groupvars= c ( "Mix" , "HRT" )) ##Make a table tgc_head < - tgc_Biogas[, 1 : 4 ] sum_table< - data.frame (tgc_head, tgc_VS $ VS_Reduction, tgc_VFA $ VFA,tgc_Hac $ Hac,tgc_Hpa $ Hpa, tgc_Cellulose $ Cellulose,tgc_Xylan $ Xylan,tgc_Lignin $ Lignin) sum_table $ N< - NULL 70 names (sum_table)[ 4 ]< - paste ( "VS_Reduction" ) names (sum_table)[ 5 ]< - paste ( "VFA" ) names (sum_table)[ 6 ]< - paste ( "Cellulose" ) names (sum_table)[ 7 ]< - paste ( "Xylan" ) names (sum_table)[ 8 ]< - paste ( "Lignin" ) names (sum_table)[ 9 ]< - paste ( "Hac" ) names (sum_table)[ 10 ]< - paste ( "Hpa" ) head (sum_table) #CH4 Productivity vs HRT Biogas< - ggplot (tgc_Biogas, aes ( x= HRT, y= Biogas_Productivity, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + geom_errorbar ( aes ( ymin= Biogas_Productivity - se, ymax= Biogas_Productivity + se), show.legend= F, width= . 1 , position= position_dodge ( 0.9 )) + labs ( x = " " , y= expression (m ^ 3 ~ kg ~ VS ^ - 1 ), title= "b. Biogas Productivity" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) # Methane tgc_CH4 $ CH4< - tgc_CH4 $ CH4 * 100 tgc_CH4 $ se< - tgc_CH4 $ se * 100 CH4< - ggplot (tgc_CH4, aes ( x= HRT, y= CH4, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = " " , y= "%" , title= "c. Methane content" ) + geom_errorbar ( aes ( ymin= CH4 - se, ymax= CH4 + se), show.legend= F, width= . 1 , position= position_dodge ( 0.9 )) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) CH4 #VS Reduction vs HRT tgc_VS $ VS_Reduction< - tgc_VS $ VS_Reduction * 100 tgc_VS $ se< - tgc_VS $ se * 100 VS< - ggplot (tgc_VS, aes (HRT, VS_Reduction, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + geom_errorbar ( aes ( ymin= VS_Reduction - se, ymax= VS_Reduction + se), show.legend= F, width= . 1 , position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = " " , y= "%" , title= "d. VS Reduction" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) #VFA Concentration vs HRT VFA< - ggplot (tgc_VFA, aes (HRT, VFA, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + geom_errorbar ( aes ( ymin= VFA - se, ymax= VFA + se), width= . 1 , show.legend= F, position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = " " , y= "g/L" , title= "e. Total VFA Concentration" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), 71 legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) #Cellulose vs HRT tgc_Cellulose $ Cellulose< - tgc_Cellulose $ Cellulose * 100 tgc_Cellulose $ se< - tgc_Cellulose $ se * 100 Cellulose< - ggplot (tgc_Cellulose, aes (HRT, Cellulose, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + geom_errorbar ( aes ( ymin= Cellulose - se, ymax= Cellulose + se), width= . 1 , show.legend= F, position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = " " , y= "%" , title= "f. Cellulose Reduction" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) #Xylan vs HRT tgc_Xylan $ Xylan< - tgc_Xylan $ Xylan * 100 tgc_Xylan $ se< - tgc_Xylan $ se * 100 Xylan< - ggplot (tgc_Xylan, aes (HRT, Xylan, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + geom_errorbar ( aes ( ymin= Xylan - se, ymax= Xylan + se), width= . 1 , show.legend= F, position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = "Reactors HRT" , y= "%" , title= "g. Xylan Reduction" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) #Lignin vs HRT tgc_Lignin $ Lignin< - tgc_Lignin $ Lignin * 100 tgc_Lignin $ se< - tgc_Lignin $ se * 100 Lignin< - ggplot (tgc_Lignin, aes (HRT, Lignin, fill= Mix)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + geom_errorbar ( aes ( ymin= Lignin - se, ymax= Lignin + se), width= . 1 , show.legend= F, position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = "Reactors HRT" , y= "%" , title= "g. Lignin Reduction" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) get_legend< - function (myggplot){ tmp < - ggplot_gtable ( ggplot_build (myggplot)) leg < - which ( sapply (tmp $ grobs, function (x) x $ name) == "guide - box" ) legend < - tmp $ grobs[[leg]] return (legend) } legend < - get_legend (Biogas) #Save the plot and addition to all plots in one page ga < - grid.arrange (gasprod, Biogas + guides ( fill= FALSE ) , CH4, VS, VFA, Cellulose, Xylan, legend, ncol= 2 ) 72 tgc_Hac $ HRT < - NULL tgc_Hac $ N < - NULL tgc_Hac $ sd< - NULL tgc_Hac $ se< - NULL tgc_Hac $ ci< - NULL tgc_Hac< - melt (tgc_Hac, na.rm = FALSE , value.name = "value" ) ## Using Mix as id variables tgc_Hpa $ HRT < - NULL tgc_Hpa $ N < - NULL tgc_Hpa $ sd< - NULL tgc_Hpa $ se< - NULL tgc_Hpa $ ci< - NULL tgc_Hpa< - melt (tgc_Hpa, na.rm = FALSE , value.name = "value" ) ## Using Mix as id variables tgc_Ac_Pa< - rbind (tgc_Hac,tgc_Hpa) HRT< - c ( 1 , 2 , 3 , 1 , 2 , 3 , 1 , 2 , 3 , 1 , 2 , 3 ) tgc_Ac_Pa< - data.frame (tgc_Ac_Pa,HRT) #Acetate & Propionate vs HRT Ac_Pa< - ggplot (tgc_Ac_Pa, aes ( fill= variable, y= value, x= HRT)) + geom_bar ( stat= "identity" ) + facet_grid (. ~ Mix) + ylim ( 0 , 4 ) + labs ( x = "HRT" , y= "Acid concentratio (g/L)" , title= "Acetate:Propionic Ratio" ) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) Ac_Pa # 5. ALPHA DIVERSITY ---- #UPLOAD OTU.Table with seeded ##Choose the OTU.Table should be a .txt OTU_Table_seeded < - read.delim ( "~/Thesis/ADonAFEXfiber(2016)/DNA - data/No_manure/OTU_v1.txt" , row.names= 1 ) set.seed ( 711 ) ## Now we create the data.frame used for OTU Table, let's watch it! ## Now we create a matrix object with the data frame t.OTU.table.seeded < - t (OTU_Table_seeded) # Conversion a matriz y transposición de tabla # Let's make some Alpha diversity analysis indexes #First Shannon H < - diversity (t.OTU.table.seeded, index = "shannon" , MARGIN = 1 , base = exp ( 1 )) #Then Simpson D < - diversity (t.OTU.table.seeded, "simpson" , MARGIN = 1 , base = exp ( 1 )) #Third inverse Simpson iD < - diversity (t.OTU.table.seeded, "inv" ) # The last is Pielou's evenness J< - H / log ( specnumber (t.OTU.table.seeded)) #Pielou's evenness print ( "ALPHA DIVERISTY WITH SEED SAMPLE" ) #Sampling Curve col < - c ( "#00BFC4" , "#00BFC4" , "#F8766D" , "#F8766D" , "#00BFC4" , "#00BFC4" , "#F8766D" , "#F8766D" , "#00BFC4" , "#00BFC4" , "#F8766D" , "#F8766D" , "forestgreen" ) lty < - c ( "solid" ) pars < - expand.grid ( col = col, lty = lty, stringsAsFactors = FALSE ) ra < - rarecurve (t.OTU.table.seeded, step = 20 , col = col, lty = lty, cex = 0.6 ) # curvas de rarefracción 73 rad < - rad.lognormal (t.OTU.table.seeded) # Rank of Abundance rad1 < - plot (rad, xlab = "Rank" , ylab = "Abundance" ) # Plotting the rank S < - specnumber (t.OTU.table.seeded) # observed number of species (raremax < - min ( rowSums (t.OTU.table.seeded))) Srare < - rarefy (t.OTU.table.seeded, raremax) rarecurve (t.OTU.table.seeded, step = 20 , col = col, lty = lty, sample = raremax, label = FALSE ) # 6. BETA DIVERSITY ---- ## 6.1 ABUNDANCE PLOTS AND RICHNESS metadata_seeded < - read.delim ( "~/Thesis/ADonAFEXfiber(2016)/DNA - data/No_manure/Metadata_v2_no_seed.txt" , row.names= 1 ) #Phyloseq OTU < - otu_table (OTU_Table_seeded, taxa_are_rows = TRUE ) # OTU Table production for phyloseq TAX < - tax_table ( as.matrix (OTU_Table_taxonomy)) ## Taxanomy production for phyloseq SAM < - sample_data (metadata_seeded) physeq < - phyloseq (OTU, TAX, SAM) ##physeq document production #Rarefication and normalization of abundance data physeq.r = rarefy_even_depth (physeq, rngseed = TRUE ) #Function for normalize physeq object #Richness r= plot_richness (physeq, x = "Duplicate" , measures = c ( "Shannon" , "Simpson" , "InvSimpson" , "Fisher" ), color = "Mix" ) + geom_boxplot () r + geom_point ( size = 5 , alpha = 0.7 ) + xlab ( "" ) + theme ( legend.position= "bottom" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) #Normalize Abundace Plotbar Bacteria physeq1 < - tax_glom (physeq.r, taxrank= rank_names (physeq.r)[ 4 ], NArm= TRUE , bad_empty= c ( NA , "" , " " , " \ t" )) mergedGP = merge_samples (physeq1, "Duplicate" ) physeq_ 1 = transform_sample_counts (mergedGP, function (x) x / sum (x)) p = plot_bar (physeq_ 1 , fill = "Phylum" ) p + geom_bar ( aes ( color= Phylum, fill= Phylum), stat = "identity" , position = "stack" ) + scale_y_continuous ( labels= scales :: percent) + ylab ( "Relative Abundace" ) + labs ( title = "Bacteria Abundance" ) + theme ( legend.position= "right" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) #Abundace Plotbar Spirochaetes physeq6 < - subset_taxa (physeq.r, Phylum == "Spirochaetes" ) physeq6_ 1 < - tax_glom (physeq6, taxrank= rank_names (physeq6)[ 6 ], NArm= TRUE , bad_empty= c ( NA , "" , " " , " \ t" )) l = plot_bar (physeq6_ 1 , x= "Duplicate" , fill = "Genus" ) + geom_bar ( aes ( color= Genus, fill= Genus), stat = "identity" , position = "stack" ) + ylab ( "Microbial Abundance" ) + xlab ( "Samples" ) + scale_fill_brewer ( palette= "Set1" ) + scale_colour_brewer ( palette= "Set1" ) + labs ( title = "c. Spirochaetes Abundance" ) + scale_y_continuous ( labels= scales :: percent) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.x = element_text ( 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), 74 legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 ), legend.direction= "vertical" ) l # Top 20 Abundace Plotbar Bacteria topN = 20 most_abundant_taxa = sort ( taxa_sums (physeq.r), TRUE )[ 1 : topN] physeq.r20 = prune_taxa ( names (most_abundant_taxa), physeq.r) ntaxa (physeq.r20) length ( get_taxa_unique (physeq.r20, "Phylum" )) topp ( 0.3 ) f1 = filterfun_sample ( topp ( 0.3 )) print (f1) ## function (x) ## { ## fun = flist[[1]] ## fval = fun(x) ## for (fun in flist[ - 1]) { ## fval = fval & fun(x) ## } ## return(fval) ## } ## ## ## attr(,"class") ## [1] "filterfun" wh1 = genefilter_sample (physeq.r20, f1, A = round ( 0.5 * nsamples (physeq.r20))) sum (wh1) ex2 = prune_taxa (wh1, physeq.r20) mergedGP = merge_samples (ex2, "Duplicate" ) ex2_r = transform_sample_counts (mergedGP, function (x) x / sum (x)) physeq8_ 1 < - tax_glom (ex2_r, taxrank= rank_names (ex2_r)[ 2 ], NArm= TRUE , bad_empty= c ( NA , "" , " " , " \ t" )) b = plot_bar (physeq8_ 1 , fill = "Phylum" ) + geom_bar ( aes ( color= Phylum, fill= Phylum), stat = "identity" , position = "stack" ) + ylab ( "Relative Abundace" ) + labs ( title = "b. Bacteria Abundance" ) + scale_y_continuous ( labels= scales :: percent) + theme ( legend.position= "right" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) b #Abundance Plotbar Domain physeq1 < - tax_glom (physeq.r, taxrank= rank_names (physeq.r)[ 1 ], NArm= TRUE , bad_empty= c ( NA , "" , " " , " \ t" )) mergedGP = merge_samples (physeq1, "Duplicate" ) physeq_dom = transform_sample_counts (mergedGP, function (x) x / sum (x)) a = plot_bar (physeq_dom, fill = "Domain" ) + geom_bar ( aes ( color= Domain, fill= Domain), stat = "identity" , position = "stack" ) + ylab ( "Microbial Abundance" ) + labs ( title= "a. Domain Abundance" ) + scale_y_continuous ( labels= scales :: percent) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.x = element_blank (), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 75 15 ), plot.title= element_text ( size = 15 )) a #Abundace Plotbar Archaea physeq5 = subset_taxa (physeq.r, Domain == "Archaea" ) physeq5_ 1 < - tax_glom (physeq5, taxrank= rank_names (physeq5)[ 6 ], NArm= TRUE , bad_empty= c ( NA , "" , " " , " \ t" )) mergedGP = merge_samples (physeq5_ 1 , "Duplicate" ) physeq5_ 2 = transform_sample_counts (mergedGP, function (x) x / sum (x)) d = plot_bar (physeq5_ 2 , fill = "Genus" ) + geom_bar ( aes ( color= Genus, fill= Genus), stat = "identity" , position = "stack" ) + ylab ( "Relative Abundance" ) + xlab ( "Samples" ) + labs ( title = "d. Archaea Abundance" ) + scale_y_continuous ( labels= scales :: percent) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.x = element_text ( 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 ), legend.direction= "vertical" ) d ec < - grid.arrange (a,b, l,d, ncol= 2 ) ## 6.2 PCOA WITH_SEED_FILE GP.ord < - ordinate (physeq.r, "NMDS" , "bray" ) p1 = plot_ordination (physeq.r, GP.ord, type= "taxa" , color= "Phylum" , title= "taxa" ) print (p1) p2 = plot_ordination (physeq.r, GP.ord, type= "samples" , color = "HRT" , shape= "Mix" ) p2 + stat_ellipse ( geom = "polygon" , alpha = 0.45 , aes ( fill = Mix)) + geom_point ( size= 5 ) ## Too few points to calculate an ellipse HRT < - list () HRT[[ 1 ]] < - c ( "CS_HRT_1_A" , "CS_HRT_1_B" , "AFEX_HRT_1_A" , "AFEX_HRT_1_B" , "SEED_DM" ) HRT[[ 2 ]] < - c ( "CS_HRT_2_A" , "CS_HRT_2_B" , "AFEX_HRT_2_A" , "AFEX_HRT_2_B" , "SEED_DM" ) HRT[[ 3 ]]< - c ( "CS_HRT_3_A" , "CS_HRT_3_B" , "AFEX_HRT_3_A" , "AFEX_HRT_3_B" , "SEED_DM" ) # Principal components analysis for each HRT beta < - vegdist (t.OTU.table.seeded, binary = TRUE ) pcoa.obj < - capscale (t.OTU.table.seeded ~ 1 , distance = "bray" ) plot (pcoa.obj) #plot the PcoA plot text ( scores (pcoa.obj) $ sites[, 1 ], scores (pcoa.obj) $ sites[, 2 ]) # change of the labes #labels=row.names(t.OTU.table.seeded) #SECOND trial with metaMDS vare.mds < - metaMDS (t.OTU.table.seeded, trace = FALSE ) vare.mds stressplot (vare.mds) metadata_fil < - read.delim ( "~/Thesis/ADonAFEXfiber(2016)/DNA - data/No_manure/Metadata_v2_no_seed.txt" , row.names= 1 ) ef < - envfit (vare.mds, env = metadata_seeded[ 4 : 11 ], permutations = 999 , p.max = 0.95 , na.rm = 76 TRUE ) plt < - plot (vare.mds, display = "sites" , type = "p" ) identify (plt, what = "sites" ) pl < - plot (ef) with (metadata_fil, ordiellipse (vare.mds, HRT, col= "forestgreen" , kind = "se" , conf = 0.95 , label = TRUE )) with (metadata_fil, ordiellipse (vare.mds, Mix, col= "red" , kind = "se" , conf = 0.95 , label = TRUE )) 77 Shapiro - Wilk & ANOVA results of performance data of AD reactor Table A1. Statistical results from the AD performance around the three HRT. Parameter Degree of freedom Feedstock HRT Feedstock & HRT Residuals Biogas productivity Sum square 1 , 564 .0 27 , 646 .0 9 , 037 .0 180 .0 F value 52.3 0 462.1 0 151.0 0 P (>F) * 0.0004 * 0.00003 * 0.000007 Methane content Sum square 25.5 22.2 2.3 11.3 F value 13.6 5.9 0.6 P (>F) * 0.01 * 0.04 0.57 VS reduction Sum square 2.7 470.8 528.1 440.2 F value 0.04 3.2 0 3.6 0 P (>F) 0.85 0.11 0.09 Cellulose reduction Sum square 18.8 1 , 070.7 464.2 399.2 F value 0.28 8.05 3.50 P (>F) 0.61 * 0.02 0.10 Xylan reduction Sum square 142.6 405.8 261.8 427.7 F value 2.00 2.85 1.84 P (>F) 0.21 0.14 0.24 VFA concentration Sum square 2. 5 1.6 1. 2 15. 4 F value 0.97 0.32 0.23 P (>F) 0.36 0.74 0.80 * Significance is selected with a p - value of 0.01 78 Calculation and assumptions for increasing BMP volume for the ABPs experiments The objective of this calculation is to determine a dry - matter mass balance of anaerobic degradation of structural carbohydrates. The requirements of dry mass necessitate changing the volume of the normal BMP. The calculation requirements are based on the material for one sample and on dry - matter material. With this sample size, it is possible to obtain two NREL content points per sample. In addition, it is assumed that predigestion TS is 7,000 mg/L and postdig estion TS is 4,000 mg/L for each BMP, and the TS concentration of our control/seed is ~4,000 mg/L. Similarly, it is assumed that the samples will be dried at 45 °C, and therefore , the sample will have ~10% moisture content. Finally, the BMP VS:VS Inoculum: Feed ratio will be 2. The experiments are based on the protocols developed by the NREL for quantitation of structural carbohydrates and determination of extractives in the biomass that could affect the results on the carbohydrates (A Sluiter et al., 2008; Amie Sluiter et al., 2004) . The sample size calculation is described below: 79 The above calculation means that it is not possible to obtain enough dry material required to run NREL with the actual volume size of a BMP test (0.150 L). The standard volume has to be increased for the preparation for the experiment (Faivor & Kirk, 2011), to obtain the dry mass required to run the BMP analysis and obtain the mass required for characterization of the lignocellulosic content in the BMP experiment . This observation implies that it is necessary to increase the sample volume of the BMP from 0.150 to 0.5 L. Accordingly, the final mass for the BMP pre - and postdigestion will be as follows: Blend the preparation for 1 L predigestion. The final volume of the BMP test for postdigestion characterization is 0.5 L. One should use 7 g of a dry sample for one sample as explained in the following calculation: 80 RSCRIPT for ABP variance - based sensitivity analysis ## Fiber Analysis in ABP ## MICHIGAN STATE UNIVERSITY ## ADREC ## Version 2.0 ## Made by Juan Pablo Rojas, 2018 cat ( " \ 014" ) #Erase console # 1. Loading Library and Tables ------ library (vegan) library (MASS) library (ggplot2) library (grid) library (gridExtra) library (readr) library (VennDiagram) library (tidyverse) library (Rmisc) library (reshape2) Biogas_Plants < - read.delim ( "~/Thesis/Fiber Study/Biogas_Plants_071618.txt" , row.names= 1 ) # 2. Evaluate the normal test of the independen variables ---------- # Influent_Farm_A Influent_Farm_A < - Biogas_Plants %>% filter (Flow == "Influent" & Plant == "Farm_A" ) for (i in 4 : 14 ) { print ( colnames (Influent_Farm_A[i])) print ( shapiro.test ( log ( unlist (Influent_Farm_A[,i])))) } # Influent_Farm_B Influent_Farm_B < - Biogas_Plants %>% filter (Flow == "Influent" & Plant == "Farm_B" ) for (i in 4 : 14 ) { print ( colnames (Influent_Farm_B[i])) print ( shapiro.test ( log ( unlist (Influent_Farm_B[,i])))) } # Influent_SCAD Influent_SCAD < - Biogas_Plants %>% filter (Flow == "Influent" & Plant == "SCAD" ) for (i in 4 : 14 ) { print ( colnames (Influent_SCAD[i])) print ( shapiro.test ( log ( unlist (Influent_SCAD[,i])))) } # Effluent_Farm_A Effluent_Farm_A < - Biogas_Plants %>% filter (Flow == "Effluent" & Plant == "Farm_A" ) for (i in 4 : 14 ) { print ( colnames (Effluent_Farm_A[i])) print ( shapiro.test ( log ( unlist (Effluent_Farm_A[,i])))) } # Effluent_Farm_B Effluent_Farm_B < - Biogas_Plants %>% filter (Flow == "Effluent" & Plant == "Farm_B" ) for (i in 4 : 14 ) { print ( colnames (Effluent_Farm_B[i])) 81 print ( shapiro.test ( log ( unlist (Effluent_Farm_B[,i])))) } # Effluent_SCAD Effluent_SCAD < - Biogas_Plants %>% filter (Flow == "Effluent" & Plant == "SCAD" ) for (i in 4 : 14 ) { print ( colnames (Effluent_SCAD[i])) print ( shapiro.test ( log ( unlist (Effluent_SCAD[,i])))) } # 3. Plotting BMP Results ---- Farm_A_BMP < - read.delim ( "~/Thesis/Fiber Study/Farm_A_BMP_080318.txt" ) Farm_B_BMP < - read.delim ( "~/Thesis/Fiber Study/Farm_B_BMP_080318.txt" ) SCAD_BMP < - read.delim ( "~/Thesis/Fiber Study/SCAD_BMP_080318.txt" ) #Farm_A Gas Production Farm_A_gasprod< - ggplot (Farm_A_BMP, aes (Time, Gas, color= Date)) + geom_point ( aes ( shape= Date, color= Date)) + ylab ( "mL of Biogas" ) + geom_line ( data= Farm_A_BMP[Farm_A_BMP $ Date != "Gas" , ]) + geom_smooth () + labs ( title = "a. Acumulated Gas Production Farm A" ) + xlab ( "Lapsed Time (h)" ) + ylim ( 0 , 1500 ) #Farm_B Gas Production Farm_B_gasprod< - ggplot (Farm_B_BMP, aes (Time, Gas, color= Date)) + geom_point ( aes ( shape= Date, color= Date)) + ylab ( "mL of Biogas" ) + geom_line ( data= Farm_B_BMP[Farm_B_BMP $ Date != "Gas" , ]) + geom_smooth () + labs ( title = "b. Acumulated Gas Production Farm B" ) + xlab ( "Lapsed Time (h)" ) + ylim ( 0 , 1500 ) #SCAD Gas Production SCAD_gasprod< - ggplot (SCAD_BMP, aes (Time, Gas, color= Date)) + geom_point ( aes ( shape= Date, color= Date)) + ylab ( "mL of Biogas" ) + geom_line ( data= SCAD_BMP[SCAD_BMP $ Date != "Gas" , ]) + geom_smooth () + labs ( title = "c. Acumulated Gas Production SCAD" ) + xlab ( "Lapsed Time (h)" ) + ylim ( 0 , 1500 ) BMPS < - grid.arrange (Farm_A_gasprod, Farm_B_gasprod,SCAD_gasprod , ncol= 3 ) # 4. ANOVA test of the independen variables ---------- Biogas_Plants $ Plant < - factor (Biogas_Plants $ Plant) ##Factor Statement Biogas_Plants $ Flow < - factor (Biogas_Plants $ Flow) ##Factor Statement #ANOVA Test for parameters #Characterization Data for (i in 4 : 14 ) { #Influent fit < - Biogas_Plants %>% filter (Flow == "Influent" ) print (fit[ 1 , 3 ]) print ( colnames (fit[i])) ANOVA< - aov ( log ( unlist (fit[,i])) ~ Plant, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results #Effluent fit < - Biogas_Plants %>% filter (Flow == "Effluent" ) 82 print (fit[ 1 , 3 ]) print ( colnames (fit[i])) ANOVA< - aov ( log ( unlist (fit[,i])) ~ Plant, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results print ( " --------------------------------------------------------------------- " ) } #Plant Operation for (i in 15 : 16 ) { #Influent fit < - Biogas_Plants %>% filter (Flow == "Influent" ) print (fit[ 1 , 3 ]) print ( colnames (fit[i])) ANOVA< - aov ( log ( unlist (fit[,i])) ~ Plant, fit) #ONE WAY ANOVA for Productivity print ( TukeyHSD (ANOVA)) #Plot results print ( " --------------------------------------------------------------------- " ) } # 5. Plotting the the dry matter constituents. ---- #Dry Matter Biogas_Plants $ Date< - NULL Biogas_Plants $ sCOD< - NULL Biogas_Plants $ COD< - NULL Biogas_Plants $ Ammonia< - NULL Biogas_Plants $ TKN< - NULL Biogas_Plants $ TOC< - NULL Biogas_Plants $ CH4_Prod< - NULL Biogas_Plants $ Mass_Flow< - NULL dat< - melt (Biogas_Plants, na.rm = FALSE , value.name = "value" ) ## Using Plant, Flow, Reactor as id variables Influent< - dat %>% filter (Flow == "Influent" ) Effluent< - dat %>% filter (Flow == "Effluent" ) Dry_matter_influent< - ggplot (Influent, aes ( fill= variable, y= value, x= Plant)) + geom_bar ( stat= "identity" , position= "fill" ) + labs ( x = "Biogas Plants" , y= "Dry matter (%)" , title= "a. Influents" ) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) Dry_matter_effluent< - ggplot (Effluent, aes ( fill= variable, y= value, x= Plant)) + geom_bar ( stat= "identity" , position= "fill" ) + guides ( fill= FALSE ) + labs ( x = "Biogas Plants" , y= "Dry matter (%)" , title= "b. Effluents" ) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) get_legend< - function (myggplot){ tmp < - ggplot_gtable ( ggplot_build (myggplot)) leg < - which ( sapply (tmp $ grobs, function (x) x $ name) == "guide - box" ) legend < - tmp $ grobs[[leg]] return (legend) 83 } legend < - get_legend (Dry_matter_influent) #Save the plot and addition to all plots in one page dry_ 1 < - grid.arrange (Dry_matter_influent + guides ( fill= FALSE ), Dry_matter_effluent, ncol= 2 , right= legend) # 6. Plotting the the raw constituents. ---- Biogas_Plants < - read.delim ( "~/Thesis/Fiber Study/Biogas_Plants_071618.txt" , row.names= 1 ) Biogas_Plants $ Ash< - NULL Biogas_Plants $ TKN< - NULL Biogas_Plants $ Ammonia< - NULL Biogas_Plants $ TOC< - NULL Biogas_Plants $ Protein< - NULL Biogas_Plants $ H2O.Ext< - NULL Biogas_Plants $ C2H6O.Ext< - NULL Biogas_Plants $ Cellulose< - NULL Biogas_Plants $ Xylan< - NULL Biogas_Plants $ Lignin< - NULL Biogas_Plants $ CH4_Prod< - NULL Biogas_Plants $ Mass_Flow< - NULL dat< - melt (Biogas_Plants, na.rm = FALSE , value.name = "value" ) ## Using Date, Plant, Flow, Reactor as id variables Influent< - dat %>% filter (Flow == "Influent" ) Effluent< - dat %>% filter (Flow == "Effluent" ) C_influent< - ggplot (Influent, aes ( fill= variable, y= value, x= Plant)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + labs ( x = "Biogas Plants" , y= "Dry matter (%)" , title= "a. Influent " ) + theme ( plot.title = element_text ( hjust = 0.5 )) + scale_fill_brewer ( palette= "Spectral" ) + theme ( legend.position= "right" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) C_effluent< - ggplot (Effluent, aes ( fill= variable, y= value, x= Plant)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = "Biogas Plants " , y= "Dry matter (%)" , title= "b. Effluent" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + scale_fill_brewer ( palette= "Spectral" ) + theme ( legend.position= "right" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) 84 get_legend< - function (myggplot){ tmp < - ggplot_gtable ( ggplot_build (myggplot)) leg < - which ( sapply (tmp $ grobs, function (x) x $ name) == "guide - box" ) legend < - tmp $ grobs[[leg]] return (legend) } legend < - get_legend (C_influent) #Save the plot and addition to all plots in one page liq_ 1 < - grid.arrange (C_influent + guides ( fill= FALSE ) , C_effluent, ncol= 2 , right= legend) #Nitrogen Balance Biogas_Plants < - read.delim ( "~/Thesis/Fiber Study/Biogas_Plants_071618.txt" , row.names= 1 ) Biogas_Plants $ Ash< - NULL Biogas_Plants $ sCOD< - NULL Biogas_Plants $ COD< - NULL Biogas_Plants $ TOC< - NULL Biogas_Plants $ Protein< - NULL Biogas_Plants $ H2O.Ext< - NULL Biogas_Plants $ C2H6O.Ext< - NULL Biogas_Plants $ Cellulose< - NULL Biogas_Plants $ Xylan< - NULL Biogas_Plants $ Lignin< - NULL Biogas_Plants $ CH4_Prod< - NULL Biogas_Plants $ Mass_Flow< - NULL dat< - melt (Biogas_Plants, na.rm = FALSE , value.name = "value" ) ## Using Date, Plant, Flow, Reactor as id variables Influent< - dat %>% filter (Flow == "Influent" ) Effluent< - dat %>% filter (Flow == "Effluent" ) N_influent< - ggplot (Influent, aes ( fill= variable, y= value, x= Plant)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + labs ( x = "Biogas Plants" , y= "Dry matter (%)" , title= "a. Influent " ) + theme ( plot.title = element_text ( hjust = 0.5 )) + ylim ( 0 , 1 ) + scale_fill_brewer ( palette= "Set1" ) + theme ( legend.position= "right" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 15 ), plot.title= element_text ( size = 15 )) N_effluent< - ggplot (Effluent, aes ( fill= variable, y= value, x= Plant)) + geom_bar ( stat= "identity" , position= position_dodge ( 0.9 )) + guides ( fill= FALSE ) + labs ( x = "Biogas Plants " , y= "Dry matter (%)" , title= "b. Effluent" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + scale_fill_brewer ( palette= "Set1" ) + theme ( legend.position= "right" , axis.title.x = element_blank (), axis.text.x = element_text ( size = 15 ), axis.title.y = element_text ( size = 15 ), axis.text.y = element_text ( size = 15 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 85 15 ), plot.title= element_text ( size = 15 )) legend < - get_legend (N_influent) #Save the plot and addition to all plots in one page liq_ 1 < - grid.arrange (N_influent + guides ( fill= FALSE ) , N_effluent, ncol= 2 , right= legend) # 7. Plotting TOC influent and Effluent ----- Biogas_Plants < - read_delim ( "Biogas_Plants_071618.txt" , " \ t" , escape_double = FALSE , col_types = cols ( Date = col_date ( format = "%m/%d/%Y" )), trim_ws = TRUE ) Biogas_Plants $ sCOD< - NULL Biogas_Plants $ COD< - NULL Biogas_Plants $ TKN< - NULL Biogas_Plants $ Ammonia< - NULL Biogas_Plants $ Reactor< - NULL Biogas_Plants $ Protein< - NULL Biogas_Plants $ H2O.Ext< - NULL Biogas_Plants $ C2H6O.Ext< - NULL Biogas_Plants $ Cellulose< - NULL Biogas_Plants $ Xylan< - NULL Biogas_Plants $ Lignin< - NULL Biogas_Plants $ CH4_Prod< - NULL Biogas_Plants $ Mass_Flow< - NULL TOC< - ggplot (Biogas_Plants, aes ( x= Date, y= TOC, shape= factor (Flow), colour = factor (Plant))) + geom_point ( size = 3 ) + geom_line () + theme ( plot.title = element_text ( hjust = 0.5 )) TOC # 8. Reduction of average influet and Effluent ---- Biogas_Plants < - read.delim ( "~/Thesis/Fiber Study/Biogas_Plants_071618.txt" , row.names= 1 ) Biogas_Plants $ CH4_Prod< - NULL Biogas_Plants $ Mass_Flow< - NULL Biogas_Plants $ Date< - NULL Biogas_Plants $ Reactor< - NULL #Calculating averages tgc_Reduction < - data.frame () for (i in 3 : 13 ) { fit< - summarySE (Biogas_Plants, measurevar= colnames (Biogas_Plants[i]), groupvars= c ( "Flow" , "Plant" )) #Nulling non requiring columns fit $ N< - NULL fit $ se< - NULL fit $ ci< - NULL dat< - melt (fit, na.rm = FALSE , value.name = "value" ) #Average reduction Influent< - dat %>% filter (Flow == "Influent" & variable != "sd" ) Effluent< - dat %>% filter (Flow == "Effluent" & variable != "sd" ) 86 Substract < - (Influent $ value - Effluent $ value) Total< - Influent $ value Reduction< - Substract / Total tgc_head < - Influent[, 2 : 3 ] tgc_Reduction< - data.frame (tgc_head, Reduction) print ( colnames (Biogas_Plants[i])) print ( head (tgc_Reduction)) #Standard deviation Influentsd< - dat %>% filter (Flow == "Influent" & variable == "sd" ) Effluentsd< - dat %>% filter (Flow == "Effluent" & variable == "sd" ) Substractsd < - (Influentsd $ value) + (Effluentsd $ value) rel_Substractsd < - Substractsd rel_Total < - Influentsd $ value SD < - rel_Substractsd + rel_Total tgc_Reduction< - data.frame (tgc_head, SD) print ( colnames (Biogas_Plants[i])) print ( head (tgc_Reduction)) print ( " --------------------------------------------------------------------- " ) } for (i in 3 : 13 ) { fit< - summarySE (Biogas_Plants, measurevar= colnames (Biogas_Plants[i]), groupvars= c ( "Flow" , "Plant" )) #Nulling non requiring columns fit $ N< - NULL fit $ value< - NULL fit $ se< - NULL fit $ ci< - NULL dat< - melt (fit, na.rm = FALSE , value.name = "sd" ) Influent< - dat %>% filter (Flow == "Influent" ) Effluent< - dat %>% filter (Flow == "Effluent" ) Reduction < - (Influent $ sd - Effluent $ sd) / Influent $ sd tgc_head < - Influent[, 2 : 3 ] tgc_Reduction< - data.frame (tgc_head, Reduction) print ( colnames (Biogas_Plants[i])) print ( head (tgc_Reduction)) # tgc_Reduction< - rbind(output, mtcars[i, ]) print ( " --------------------------------------------------------------------- " ) } #9. Contribution of the variance ---- rm ( list= ls ()) setwd ( '/Users/rojasju2/Documents/Thesis/Fiber Study/R_Scripts/Code/Code' ) source ( '/Users/rojasju2/Documents/Thesis/Fiber Study/R_Scripts/Code/Code/contribution_variance.R' ) # Load function filename < - 'Reduction_Biogas_Plants_082918.csv' input.data < - read.csv (filename) var_data< - contribution_to_variance (CH4 ~ sCOD + COD + NH3 + TKN + Protein + TOC + H2O.Ext + C2H6O.Ext + Cellulose + Xylan + Li 87 gnin, data= input.data) h < - ggplot (var_data, aes ( x = var, y = pct, fill = var)) + geom_bar ( alpha = 0.50 , stat = "identity" ) + geom_text ( aes ( y = pct / 2 , label = paste0 ( round (pct * 100 , 1 ), "%" ))) + coord_flip () + labs ( x = "Biogas Plants" , y= "Dry matter (%)" , title= "Contribution to Variance" ) + theme ( plot.title = element_text ( hjust = 0.5 )) + theme ( legend.position= "right" , axis.text.x = element_text ( size = 11 ), axis.title.y = element_text ( size = 11 ), axis.text.y = element_text ( size = 11 ), legend.text = element_text ( size = 11 ), legend.title= element_text ( size = 11 ), plot.title= element_text ( size = 11 )) print (h) 88 Shapiro Wilk & ANOVA results on ABP characterization Table A2. Statistical results from characterization of the influents and effluents of three different reactors Parameter Flow Farm A Farm B SCAD W T W T W T sCOD In 0.9 0.26±0.05 A 0.9 0.27±0.15 A 0.7 0.35±0.04 A [% dry matter] Eff 0.9 0.12±0.01 B 0.8 0.15±0.01 AB 0.9 0.21+0.07 A COD In 0.9 0.84±0.03 B 0.7 0.84±0.08 B 0.9 1.04±0.13 A [% dry matter] Eff 0.9 0.72±0.02 A 0.8 0.78±0.04 A 0.8 0.78±0.04 A Ammonia In 0.9 0.02±0.01 AB 0.9 0.03±0.01 A 0.9 0.01±0.01 B [% dry matter] Eff 0.9 0.02±0.00 AB 0.9 0.02±0.01 B 0.7 0.05±0.02 A TKN In 0.9 0.04±0.01 A 0.9 0.03±0.01 A 0.9 0.03±0.02 A [% dry matter] Eff 0.7 0.14±0.06 A 0.9 0.04±0.00 B 0.9 0.06±0.03 AB Protein In 0.9 0.18±0.00 A 0.8 0.17±0.01 A 0.9 0.14±.01 B [% dry matter] Eff 0.9 0.17±0.01 A 0.8 0.16±0.00 AB 0.9 0.15±0.01 B TOC In 0.7 0.36±0.01 B 0.9 0.39±0.01 A 0.9 0.43±0.02 A [% dry matter] Eff 0.9 0.36±0.00 A 0.7 0.39±0.01 A 0.9 0.37±0.03 A H 2 O.Ext In 0.9 0.40±0.14 A 0.8 0.32±0.10 A 0.9 0.23±0.04 A [% dry matter] Eff 0.8 0.33±0.02 AB 0.9 0.35±0.08 A 0.9 0.24±0.02 B C 2 H 6 O.Ext In 0.8 0.06±0.03 B 0.9 0.09±0.02 AB 0.8 0.14±0.05 A [% dry matter] Eff 0.9 0.08±0.01 B 0.9 0.10±0.01 AB 0.9 0.16±0.05 A Cellulose In 0.7 0.10±0.04 A 0.8 0.13±0.02 A 0.9 0.13±0.05 A [% dry matter] Eff 0.9 0.10±0.00 A 0.9 0.11±0.02 A 0.8 0.13 ±0.02 A Xylan In 0.7 0.05±0.02 A 0.8 0.06±0.01 A 0.9 0.06±0.02 A [% dry matter] Eff 0.9 0.06±0.00 A 0.9 0.05±0.01 A 0.8 0.06±0.01 A Lignin In 0.9 0.15±0.04 A 0.8 0.15±0.02 A 0.9 0.21±0.02 A [% dry matter] Eff 0.7 0.18±0.01 AB 0.9 0.16±0.03 B 0.9 0.21±0.02 A Productivity In 0.8 53.4±72 B 0.9 86.9±99 AB 0.9 154.13±170 A [m 3 /kg VS] p value of 0.01. 89 REFERENCES 90 REFERENCES Aguilar Alvarez, R. E., Bustamante Roman, M., Kirk, D., Miranda Chavarria, J. A., Baudrit, D., Liao, W. (2016). Technical and economic feasibility of a solar - bio - powered waste utilizati on and treatment system in Central America. Journal of Environmental Management , 184 , 371 379. https://doi.org/10.1016/j.jenvman.2016.10.007 APHA. (1989). Standard methods for the examination of water and wastewater. 17th edition. Retrieved from https://ta mug - ir.tdl.org/tamug - ir/handle/1969.3/24401 Balan, V, Sousa, L., Chundawat, S. P. S., Humpula, J., & Dale, B. E. (2012). Overview to Ammonia Pretreatments for Lignocellulosic Biorefineries. Dynamic Biochemistry, Process Biotechnology and Molecular Biology , 6 (Special Issue 2), 1 11. Balan, Venkatesh, Bals, B., Chundawat, S. P. S., Marshall, D., & Dale, B. E. (2009). Lignocellulosic Biomass Pretreatment Using AFEX. In J. R. Mielenz (Ed.), Biofuels: Methods and Protocols (pp. 61 77). https://doi.org/10.1007/97 8 - 1 - 60761 - 214 - 8_5 Vavilin, V. A. (2002). The IWA Anaerobic Digestion Model No 1 (ADM1). Water Science and Technology , 45 (10), 65 73. https://doi.org/10.2166/w st.2008.678 Bolker, B. M. (2008). Ecological Models and Data in R (Princeton). Retrieved from pup.princeton.edu Brijde, M., Dumont, M., & Blume, A. (2014). Contribution Greengasgrids Project To Development in Biomethane Markets . 48. Retrieved from http://w ww.greengasgrid.eu Campanaro, S., Treu, L., Kougias, P. G., De Francisci, D., Valle, G., & Angelidaki, I. (2016). Metagenomic analysis and functional characterization of the biogas microbiome using high throughput shotgun sequencing and a novel binning str ategy. Biotechnology for Biofuels , 9 (1), 1 17. https://doi.org/10.1186/s13068 - 016 - 0441 - 1 Changda, C., Xiang, Z., Wu, H., & Yifeng, S. (1994). Fushan Farm an example taking biogas as a link. Energy for Sustainable Development , 1 (2), 38 40. https://doi.org/10.1016/S0973 - 0826(08)60031 - 9 Chen, R., Murillo Roos, M., Zhong, Y., Marsh, T., Bustaman te Roman, M., Hernandez Liao, W. (2016). Responses of anaerobic microorganisms to different culture conditions and corresponding effects on biogas production and solid digestate quality. Biomass and Bioenergy , 85 , 84 93. https://doi.org/10. 1016/j.biombioe.2015.11.028 Chen, Y., Cheng, J. J., & Creamer, K. (2008). Inhibition of anaerobic digestion process: a review. Bioresource Technology , 99 (10), 4044 4064. https://doi.org/10.1016/j.biortech.2007.01.057 91 Chundawat, S., Bals, B., Campbell, T., on Ammonia Fiber Expansion Pretreatment. Aqueous Pretreatment of Plant Biomass for Biological and Chemical Conversion to Fuels and Chemicals , 169 200. https://doi.org/10.1002/9780470975831.ch9 Chund awat, S., Beckham, G., Himmel, M., & Dale, B. (2011). Deconstruction of Lignocellulosic Biomass to Fuels and Chemicals. Annual Review of Chemical and Biomolecular Engineering , 2 (1), 121 145. https://doi.org/10.1146/annurev - chembioeng - 061010 - 114205 Chundawa Dale, B. E. (2011). Multi - scale visualization and characterization of lignocellulosic plant cell wall deconstruction during thermochemical pretreatment. Energy and Environmenta l Science , 4 (3), 973 984. https://doi.org/10.1039/c0ee00574f Proceedings of the National Academy of Sciences , 106 (Supplement_2), 19651 19658. https://doi.org/10.1073/pnas.0901650106 da Costa Sousa, L., Chundawat, S. P., Balan, V., & Dale, B. E. (2009). - to - assessment of existing lignocellulose pretreatment technologies. Current Opinion in Biotechnology , 20 (3), 339 347. https://doi. org/10.1016/J.COPBIO.2009.05.003 Demirel, B., & Scherer, P. (2008). The roles of acetotrophic and hydrogenotrophic methanogens during anaerobic conversion of biomass to methane: a review. Reviews in Environmental Science and Bio/Technology , 7 (2), 173 190. https://doi.org/10.1007/s11157 - 008 - 9131 - 1 Doran, P. M. (1995). Bioprocess Engineering Principles. In Bioprocess Engineering Principles . https://doi.org/10.1016/B978 - 0 - 12 - 220855 - 3.50020 - 1 EPA. Regulation of fuels and fuel additives: RFS Pathways II, and tec hnical amendments to the RFS standards and E15 misfueling mitigation requirements. , 79 Federal Register § (2014). EPA. (2017). Overview for Renewable Fuel Standard. Retrieved September 6, 2018, from https://www.epa.gov/renewable - fuel - standard - program/overview - renewable - fuel - standard Faivor, L., & Kirk, D. (2011). Statistical Verification of a Biochemical Methane Potential Test. American Society of Agricultural and Biological Engineers , 6 (11), 5200 521 9. Louisville, KY: American Society of Agricultural and Biological Engineers. Fang, H. H. P., & Liu, H. (2002). Effect of pH on hydrogen production from glucose by a mixed culture. Bioresource Technology , 82 (1), 87 93. https://doi.org/10.1016/S0960 - 8524(01 )00110 - 9 Flavin, C., Gonzalez, M., Majano, A. M., Ochs, A., Rocha, M. da, & Tagwerker, P. (2014). Study on the Development of the Renewable Energy Market in Latin America and the 92 Caribbean. Inter - American Development Bank , (November), 1 67. Retrieved from https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact =8&ved=0ahUKEwjtyvjy - c3KAhUN3mMKHf - _AbYQFggcMAA&url=https://publications.iadb.org/bitstream/handle/11319/6711/Study - on - the - Development - of - the - Renewable - Energy - Market - in - Latin - Amer Fogler, H. S. (1999). Elements of chemical reaction engineering . U[[er Saddle River, H.J: Prentice Hall PTR. Goering, H. D., & Van Soest, P. J. (1975). Forage fiber analyses . Washington, D.C. Goux, X., Calusinska, M., Lemaigre, S., Marynowska, M., Klocke, Delfosse, P. (2015). Microbial community dynamics in replicate anaerobic digesters exposed sequentially to increasing organic loading rate, acidosis, and process recovery. Biotechnology for Biofuels , 8 , 122. https://doi.org/10.1186/s13 068 - 015 - 0309 - 9 Guyot, J. P., Ferrer, H., & Florina, R. (1995). Methane production from acetamide in an upflow anaerobic sludge - blanket reactor based on a synergistic association between an aerobic rod and methanogens. Applied Microbiology and Biotechnology , 43 (6), 1107 1111. https://doi.org/10.1007/BF00166933 Hansen, T. L., Schmidt, J. E., Angelidaki, I., Marca, E., Jansen, J. L. C., Mosbæk, H., & Christensen, T. H. (2004). Method for determination of methane potentials of solid organic waste. Waste Managem ent , 24 (4), 393 400. https://doi.org/10.1016/j.wasman.2003.09.009 fuel requirements. Retrieved June 9, 2018, from U.S. Energy Information Administration website: htt ps://www.eia.gov/todayinenergy/detail.php?id=33212 Hills, D. J. (1979). Effects of carbon: Nitrogen ratio on anaerobic digestion of dairy manure. Agricultural Wastes , 1 (4), 267 278. https://doi.org/10.1016/0141 - 4607(79)90011 - 8 Hills, D. J., & Roberts, D. W . (1981). Anaerobic digestion of dairy manure and field crop residues. Agricultural Wastes , 3 (3), 179 189. https://doi.org/10.1016/0141 - 4607(81)90026 - 3 Karimi, K., & Taherzadeh, M. J. (2016). A critical review of analytical methods in pretreatment of ligno celluloses: Composition, imaging, and crystallinity. Bioresource Technology , Vol. 200, pp. 1008 1018. https://doi.org/10.1016/j.biortech.2015.11.022 Keegstra, K. (2010). Plant Cell Walls. Plant Physiology , 154 (2), 483 486. https://doi.org/10.1104/pp.110.16 1240 Kirkegaard, R. H., McIlroy, S. J., Kristensen, J. M., Nierychlo, M., Karst, S. M., Dueholm, M. composition in full - scale anaerobic digesters. Scientific Reports , 7 (1), 1 11. https://doi.org/10.1038/s41598 - 017 - 09303 - 0 93 Lau, M. W., & Dale, B. E. (2009). Cellulosic ethanol production from AFEX - treated corn stover using Saccharomyces cerevisiae 424A(LNH - ST). Proceedings of the National Academy of Sciences of the United States of America , 106 (5), 1368 1373. https://doi.org/10.1073/pnas.0812364106 Li, J., Ban, Q., Zhang, L., & Jha, A. K. (2012). Syntrophic propionate degradation in anaerobic digestion: A review. International Journal of Agriculture and Biology , 14 (5), 843 850. M ao, C., Feng, Y., Wang, X., & Ren, G. (2015). Review on research achievements of biogas from anaerobic digestion. Renewable and Sustainable Energy Reviews , 45 , 540 555. https://doi.org/10.1016/j.rser.2015.02.032 McCarty, P. L., Bae, J., & Kim, J. (2011). D omestic wastewater treatment as a net energy producer - can this be achieved? Environmental Science & Technology , 45 (17), 7100 7106. https://doi.org/10.1021/es2014264 McMurdie, P. J., & Holmes, S. (2013). Phyloseq: An R Package for Reproducible Interactive Analysis and Graphics of Microbiome Census Data. PLoS ONE , 8 (4). https://doi.org/10.1371/journal.pone.0061217 Nayono, S. E. (2010). Anaerobic Digestion of Organic Solid Waste for Energy Production . Retrieved from https://books.google.com/books?hl=en&lr=&i d=cyZAAbGI6X0C&pgis=1 Norgren, M., & Edlund, H. (2014). Lignin: Recent advances and emerging applications. Current Opinion in Colloid and Interface Science , 19 (5), 409 416. https://doi.org/10.1016/j.cocis.2014.08.004 Oh, S. T., & Martin, A. D. (2010). Long chain fatty acids degradation in anaerobic digester: Thermodynamic equilibrium consideration. Process Biochemistry , 45 (3), 335 345. https://doi.org/10.1016/j.procbio.2009.1 0.006 H. (2016). vegan: Community Ecology Package . Retrieved from https://cran.r - project.org/package=vegan Patel, J., & Nakhla, G. (2006). Simultaneous Nitrific ation and Denitrification with Anoxic Phosphorus Uptake in a Membrane Bioreactor System. Water Environment Research , 78 (11), 2193 2202. https://doi.org/10.2175/106143005X82226 Penaud, V., Delgenes, J. P., Torrijos, M., Moletta, R., Vanhoutte, B., & Cans, P . (1997). Definition of optimal conditions for the hydrolysis and acidogenesis of a pharmaceutical microbial biomass. Process Biochemistry , 32 (6), 515 521. https://doi.org/10.1016/S0032 - 9592(97)00006 - X Perez - Pimienta, J. A., Flores - Gómez, C. A., Ruiz, H. A ., Sathitsuksanoh, N., Balan, V., da Costa Simmons, B. A. (2016). Evaluation of agave bagasse recalcitrance using AFEX TM , autohydrolysis, and ionic liquid pretreatments. Bioresource Technology , 211 , 94 216 223. https://doi.org/10.1016/j.biortech.2016.03.103 Raposo, F., De La Rubia, M. a., Fernández - Cegrí, V., & Borja, R. (2012). Anaerobic digestion of solid organic substrates in batch mode: An overview relating to methane yields and experimental procedures. Renew able and Sustainable Energy Reviews , 16 (1), 861 877. https://doi.org/10.1016/j.rser.2011.09.008 REN21. (2018). Renewables 2018 Global Status Report . https://doi.org/978 - 3 - 9818911 - 3 - 3 Rittmann, B. E., & McCarty, P. L. (2001). rinciples and applications . Rojas - Sossa, J. P., Murillo - Roos, M., Uribe, L., Uribe - Liao, W. (2017). Effects of coffee processing residues on anaerobic microorganisms and corresponding digestion performance. Bioresource T echnology , 245 (August), 714 723. https://doi.org/10.1016/j.biortech.2017.08.098 Roopnarain, A., & Adeleke, R. (2017). Current status, hurdles and future prospects of biogas digestion technology in Africa. Renewable and Sustainable Energy Reviews , 67 (Octobe r 2016), 1162 1179. https://doi.org/10.1016/j.rser.2016.09.087 Safferman, S., Kirk, D. M., Faivor, L. L., & Wu - haan, W. (2012). Anaerobic Digestion Processes. In Anaerobic Digestion Processes, in Bioremdiation and Sustainability: Research and Applications (R. Mohee a). https://doi.org/10.1002/9781118371220.ch3 Sarkanen, K. V., & Ludwig, C. H. (1971). Lignins: Occurrence, formation, structure and reactions. New York: Wiley - Interscience . Scheller, H. V., & Ulvskov, P. (2010). Hemicelluloses. Annual Review of Plant Biology , 61 (1), 263 289. https://doi.org/10.1146/annurev - arplant - 042809 - 112315 Biotechnology for Agro - Industrial Residues Utili sation . https://doi.org/10.1007/978 - 1 - 4020 - 9942 - 7 Sj str m, E. (1993). . Sluiter, A., Hames, B., Ruiz, R., Scarlata, C., Sluiter, J., Templeton, D., & Crocker, D. (2012). Determination of structural carbohydr ates and lignin in Biomass. In NREL/TP - 510 - 42618 . https://doi.org/doi.org/NREL/TP - 510 - 42621 Sluiter, A, Ruiz, R., Scarlata, C., Sluiter, J., & Templeton, D. (2008). Determination of Extractives in Biomass. In NREL/TP - 510 - 42619 . https://doi.org/NREL/TP - 510 - 42621 Sluiter, Amie, Hames, B., Ruiz, R. O., Scarlata, C., Sluiter, J., Templeton, D., & Crocker, D. (2004). Determination of Structural Carbohydrates and Lignin in Biomass. Biomass Analysis Technology Team Laboratory Analytical Procedure , 2011 (July), 1 14 . https://doi.org/NREL/TP - 510 - 42618 95 Sluiter, J. B., Ruiz, R. O., Scarlata, C. J., Sluiter, A. D., & Templeton, D. W. (2010). Compositional analysis of lignocellulosic feedstocks. 1. Review and description of methods. Journal of Agricultural and Food Chemis try , 58 (16), 9043 9053. https://doi.org/10.1021/jf1008023 Speece, R. E. (1983). Anaerobic biotechnology for industrial wastewater. Environmental Science and Technology , 17 (9), 416A - 427A. https://doi.org/10.1139/l96 - 121 Speece, R. E. (1987). Anaerobic diges tion of biomass. In D. P. Chynoweth & R. Isaacson (Eds.), Anaerobic digestion of biomass (pp. 129 140). Retrieved from https://www.springer.com/us/book/9781851660698 Sun, Y., & Cheng, J. (2003). Hydrolysis of Lignocellulosic Materials for Ethanol Productio n: A Review. ChemInform , 34 (1), no - no. https://doi.org/10.1002/chin.200301272 Taherzadeh, M. J., & Karimi, K. (2008, September 1). Pretreatment of lignocellulosic wastes to improve ethanol and biogas production: A review. International Journal of Molecular Sciences , Vol. 9, pp. 1621 1651. https://doi.org/10.3390/ijms9091621 Takahashi, S., Tomita, J., Nishioka, K., Hisada, T., & Nishijima, M. (2014). Development of a Prokaryotic Universal Primer for Simultaneous Analysis of Bacteria and Archaea Using Next - Ge neration Sequencing. PLoS ONE , 9 (8), e105592. https://doi.org/10.1371/journal.pone.0105592 Templeton, D. W., Scarlata, C. J., Sluiter, J. B., & Wolfrum, E. J. (2010). Compositional analysis of lignocellulosic feedstocks. 2. Method uncertainties. Journal of Agricultural and Food Chemistry , 58 (16), 9054 9062. https://doi.org/10.1021/jf100807b Candela, M. (2016). Enterocyte - associated microbiome of the Hadza hunter - gatherers. Frontiers in Microbiology , 7 (JUN), 1 12. https://doi.org/10.3389/fmicb.2016.00865 Tye, Y. Y., Lee, K. T., Wan Abdullah, W. N., & Leh, C. P. (2016). The world availability of non - wood lignocellulosic biomass for the production of cellulosic ethanol and potential pretreatments for the enhancement of enzymatic saccharification. Renewable and Sustainable Energy Reviews , 60 , 155 172. https://doi.org/10.1016/J.RSER.2016.01.072 USDA, USEPA, & USDOE. (2015). Biogas Opportunities Roadmap Progress Report . (December). Venables, W. N., & Ripley, B. D. (2002). Modern Applied Statistics with S . https ://doi.org/10.1007/978 - 0 - 387 - 21706 - 2 Voragen, A. G. J., Coenen, G. J., Verhoef, R. P., & Schols, H. A. (2009). Pectin, a versatile polysaccharide present in plant cell walls. Structural Chemistry , 20 (2), 263 275. ttps://doi.org/10.1007/s11224 - 009 - 9442 - z Wa ng, Y., Zhang, Y., Wang, J., & Meng, L. (2009). Effects of volatile fatty acid concentrations 96 on methane yield and methanogenic bacteria. Biomass and Bioenergy , 33 (5), 848 853. https://doi.org/10.1016/j.biombioe.2009.01.007 Wickham, H. (2009). ggplot2. In Applied Spatial Data Analysis with R . https://doi.org/10.1007/978 - 0 - 387 - 98141 - 3 Yue, Z., Chen, R., Yang, F., Maclellan, J., Marsh, T., Liu, Y., & Liao, W. (2013). Effects of dairy manure and corn stover co - digestion on anaerobic microbes and corresponding digestion performance. Bioresource Technology , 128 , 65 71. https://doi.org/dx.doi.org/10.1016/j.biortech.2012.10.115