THE ROLE OF INTERNATIONAL SOYBEAN TRADE IN TELECOUPLED HUMAN AND NATURAL SYSTEMS By Anna Jean Herzberger A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirement s for the degree of Fisheries & Wildlife Doctor of Philosophy 20 20 ABSTRACT THE ROLE OF INTERNATIONAL SOYBEAN TRADE IN TELECOUPLED HUMAN AND NATURAL SYSTEMS By Anna Jean Herzberger International food trade and globalized agriculture production connects humans and the environment around the world. As consumption demands are in creasingly met abroad by importing food products, environmental and socioeconomic effects of production are left in the producing to the complexity and low vi sibility of the impacts. As population growth increases the caloric demand and rising affluence drives changes in consumption patterns, connections via food trade will continue to increase. Therefore, to identify local impacts of global phenomena, this dis sertation analyses the environmental and socioeconomic impacts of international soybean trade within Brazil (e.g., largest producer), China (e.g., largest consumer) and the U.S. (e.g., the former largest producer). Drawing from both natural and social scie nce disciplines, global trade data, satellite - imagery, farmer interviews and soil samples were combined for an interdisciplinary assessment of how international soybean trade couples distant human and the environment systems, the true extent of land - use ch ange driven by soybean trade and the resulting impacts within each respective country. Chapter 1 provides a review of the published literature and background on international soybean trade and production. In chapter 2 nd - oriented soybean expansion in Brazil displaced the production of other crops and increased imports from nearby countries. For chapter 3 , the impact of imported soybeans on pr agricultural region was explored. Competition from imported soybeans has resulted in many farmers switching cultivation to corn or to abandon farming in search of more lucrative options. This cultivation shift requires changes in m anagement that involve increased nitrogen inputs and residual crop biomass both of which have resulted in environmental spillovers. Chapter 4 furthered the analysis by considering the impacts of farmer cultivation and management decisions on soil propert ies. Soil texture, pH, total organic carbon and 16 S rRNA gene sequenc e were used in combination with detailed farmer management surveys to understand how changes in residue management effect efficiency, productivity, profitability and sustainability of the system. The results indicated that the accumulation of residual corn biomass has increased the use of residue fires and decreased the amount of crop residue being returned to the soil. The culminating chapter used an agent - based modeling (ABM) to integrate the above chapters into a TeleABM. The teleABM models land use chang e in Brazil and China based on global soybean demand. Land - use change decisions are made by farmer agents which have parametrized using the farmer interviews. Next, the farmer agent cultivation and management decisions have environmental impacts that were determined by analyzing the soil samples under the context of management decisions. Finally, future cultivation and management decisions. Because of the eco nomic, environmental and political importance of international soybean trade, the results of this dissertation are of great interest for future soybean production and trade between the specified countries as well as food security and environmental sustaina bility across the world. iv To my parents and my husband . v ACKNOWLEDGEMENTS To my major advisor, Dr. Jack Liu, thank you for believing in me even before I believed in myself . To my husband, Dr. Sheng Chen, thank you for being my morning coffee and evening whiskey, my daily motivation and my nightly resolve and available anytime for comic relief. To Mom and Dad, thanks for buying me books and letting me dig holes in the yard. To my brothers, thank s for being proud of me. I will never stop trying to impress you. To Sue Nichols , thanks for always being around to remind me who I am, what my story is and why that is valuable. To Dr. Scott & Missy Meiners , thank you for showing me a way off the farm a nd the way back. To my high school science teacher, Dr. Jason Vic ich , thank you for being so annoying and pushing me to finish this work . score a lot of points (or publications) , there is real value in being on the team. vi TABLE OF CONTENT S LIST OF TABLES ................................ ................................ ................................ ..................... viii LIST OF FIGURES ................................ ................................ ................................ ...................... x INTRODUCTION AND LITERATURE REVIEW ................................ .......................... 1 Overview ................................ ................................ ................................ ................ 1 Conceptual Framework ................................ ................................ ........................ 3 Study Areas ................................ ................................ ................................ ............ 5 Objectives of this dissertation ................................ ................................ .............. 6 Significance ................................ ................................ ................................ ............ 7 TELECOUPLED FOOD TRADE AFFECTS PERICOUPLED TRADE AND INTRACOUPLED PRODUCTION ................................ ................................ ..................... 8 Abstract ................................ ................................ ................................ .................. 8 Introduction ................................ ................................ ................................ ........... 9 Methods ................................ ................................ ................................ ................ 11 2.3.1. Conceptual Framework ................................ ................................ .................. 11 2.3.2. Relevant Theories ................................ ................................ ............................ 12 2.3.3. Data Collection ................................ ................................ ................................ 13 2.3.4. Data Analysis ................................ ................................ ................................ ... 14 Results ................................ ................................ ................................ .................. 16 2.4.1. Trend Analysis ................................ ................................ ................................ 16 2.4.2. Differences in Trade Between the Pre - and Post - Period Trends ................ 21 Discussion ................................ ................................ ................................ ............. 26 Conclusions ................................ ................................ ................................ .......... 29 APPENDIX ................................ ................................ ................................ ....................... 30 REFERENCES ................................ ................................ ................................ ................ 37 INTERNATIONAL FOOD TRADE LED TO AIR POLLUTION IN NON - TRADING COUNTRY ................................ ................................ ................................ ........................... 45 Overview ................................ ................................ ................................ .............. 45 Results ................................ ................................ ................................ .................. 46 Discussion ................................ ................................ ................................ ............. 51 Materials and Methods ................................ ................................ ....................... 53 3.4.1. Study Site ................................ ................................ ................................ ......... 53 3.4.2. Land use change analysis ................................ ................................ ............... 54 3.4.3. Household surveys ................................ ................................ .......................... 54 3.4.4. Fire ................................ ................................ ................................ ................... 55 3.4.5. Aerosol Optical Thickness ................................ ................................ .............. 55 3.4.6. Statistical Analysis ................................ ................................ .......................... 56 APPENDIX ................................ ................................ ................................ ....................... 57 REFERENCES ................................ ................................ ................................ ................ 61 vii TELECOUPLED MANAGEMENT DECSIONS DRIVE MICROBIAL COMMUNITY COMPOSITION ................................ ................................ ................................ .................. 67 Overview ................................ ................................ ................................ .............. 67 Results ................................ ................................ ................................ .................. 69 4.2.1. Changes in crop rotation practices shift regional cultivation patterns ...... 69 4.2.2. Changes in crop rotation practices have cascading impacts on management ................................ ................................ ................................ ......... 71 4.2.3. Changes in crop rotation practices influence microbial community structure, diversity, species abundance ................................ ............................ 71 Discussion ................................ ................................ ................................ ............. 78 Conclusion ................................ ................................ ................................ ........... 81 Materials & Methods ................................ ................................ .......................... 82 4.5.1. Site Description ................................ ................................ ............................... 82 4.5.2. Sampling Design ................................ ................................ .............................. 82 4.5.3. DNA extraction, PCR, and DNA sequencing ................................ ............... 83 4.5.4. Microbial Data Analysis ................................ ................................ ................. 85 REFERENCES ................................ ................................ ................................ ................ 86 USING AGENT - BASED MODELLING TO MEASURE THE IMPACT OF LAND MANAGEMENT ON EVIRONMNETAL HEALTH IN TELECOUPLED HUMAN AND NATURAL SYSTEMS ................................ ................................ .............................. 93 Overview ................................ ................................ ................................ .............. 93 Model Description ................................ ................................ ............................... 93 5.2.1. Purpose ................................ ................................ ................................ ............. 93 5.2.2. Study System ................................ ................................ ................................ ... 94 5.2.3. Agents ................................ ................................ ................................ ............... 94 5.2.4. Environment ................................ ................................ ................................ .... 94 5.2.5. Process Overview ................................ ................................ ............................ 98 5.2.6. Design Concepts ................................ ................................ .............................. 99 5.2.7. Validation ................................ ................................ ................................ ....... 103 Results ................................ ................................ ................................ ................ 103 Discussion ................................ ................................ ................................ ........... 108 Future Work ................................ ................................ ................................ ...... 109 REFERENCES ................................ ................................ ................................ .............. 111 SYNTHESIS ................................ ................................ ................................ ....................... 114 viii LIST OF TABLES tonnes (MT) from Argentina, Paraguay, Uruguay and the U.S. The table includes the trade value for the pre period (the average trade value between 1991 and 1995), 2016 and the percent change between the pre period and 2016. ................................ ................................ ................................ . 21 Table 2. Description of final ARIMA models including p (# of lags), d (# of differences), q (# of error terms). Followed by parameter descriptions and p - values, (***p < 0.001 , **p< 0.05 and *p < 0.01). AR and MA parameters refer to the autoregressive and moving average model terms. The pericoupled parameters are relative to the country in the dependent variable and are the aggregated imports from the adjacent countries. For ex ample, if Y = Soybean exports from Argentina to China, and Paraguay. The fourth column includes the sum of the differences between the observed and predicted val ues for 1991 - 1995 in MT. Positive numbers can be interpreted as a predicted increase in trade during the pre period of X MT compared to the observed values; negative numbers indicate a predicted decrease of X MT during the pre period relative to the observ ed values. Finally, normalized root mean squared error (NRMSE) and mean absolute scaled error (MASE) were included as measures of model validity. NRMSE was normalized by dividing the RMSE by the standard deviations of the dependent variable. In both cases, values close to zero indicate a high performing model that captures the variation in the data. Values around 1 indicate the model is explaining some of the variation in the data, while values beyond 1 indicated noise and a low - performing model. ................................ ................................ ................................ ......................... 35 Table 3. Farmer Survey Responses ................................ ................................ ............................... 49 Table 4. Crop residue management in Heilongjiang, China. ................................ ........................ 51 Table 5. Logistic regression analysis with clustered SE at the village level. Includes factor names, coefficients, p values and VIF ................................ ................................ ................................ ...... 59 Table 6. Factor description. ................................ ................................ ................................ .......... 60 Table 7. Mean and (standard deviation) of influential variables grouped by crop rotation practice. The influential factors determined by PCA include: pH, Total organic carbon, silt, Nitrogen use, crop residue weight returned and yield. The lower case letters designated significant groups differences (p>0.05). ................................ ................................ ................................ ..................... 71 Table 8. The extracted PCoA components (e.g., Dim1 and 2) represent the major sources of variation in the data and determine the most influential variables and factors. ............................ 74 Table 9. Mean and (standard deviation) for the alpha diversity analysis. ................................ .... 75 Table 10. Bacterial groups whose abundance differed significantly (p<0.001) between crop rotation practices. ................................ ................................ ................................ .......................... 77 ix Table 11. Household and farm characteristics of the TeleABM receiving system. Including variable names, variable description, variable distribution, parameters and independence. *denotes variables selected for Uncertainty and Sensitivity Analysis. ................................ ......... 96 x LIST OF FIGURES Figure 1. Adapted from Liu et al.[1]) Summary of the 5 interrelated components of the telecoupling framework. ................................ ................................ ................................ ................................ ...... 4 Figure 2. World map highlighting the study locations for this dissertation. Brazil, the sending system is shown in yellow while China, the receiving system is shown in green. Heilongjiang China, was the site of field work and the core study area is highlig hted in lime green. Additionally, the spillover systems are shown in grey and include Argentina, Paraguay and Uruguay as well as the Russian Provinces that border Heilongjiang. ................................ ................................ ............ 6 Figure 3. Conceptual overview of the metacoupling framework adapted from Liu (2017). The green circles represent natural components while the orange circles represent human components of the system. Grey arrows are interactions within systems (i.e., intracoupling) and black arrows are interactions between systems (i.e., peri - and telecouplings). ................................ .................. 12 Figure 4. Timeline that defines the pre period (1991 - 1995) and post period (1996 - 2016) and highlights important events. ................................ ................................ ................................ .......... 18 Fi gure 5. Flow maps depicting telecoupled soybean trade and pericoupled trade of soybeans (green arrow), corn (orange arrow) and wheat (blue arrow) from 1991 - 2016. Panels are split into 5 - year time periods: 1991 - 1995 (A), 1996 - 2000 (B), 2001 - 2005 (C), 2006 - 2 010 (D) and 2011 - 2016 (E). The width of the line represents average commodity flow from a single trade partner during the respective time period and the arrows are weighted according to total import from all trade partners. All values are presented in mill ions of metric tonnes. Trade amounts averaging below 25,000 MT per year were excluded. ................................ ................................ ................................ ................ 19 Figure 6. ARIMA for Argentina, Brazil and country and commodity results. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded gr ey area is the 95% CI. P - values, NRMSE (normalized by standard deviation) and MASE can be found in Table S1. ................................ ................................ ................................ ............. 22 Figure 7. ARIMA for Uruguay (2) and Paraguay (3). The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA mo dels. The shaded grey area is the 95% CI. ................................ ................................ ................................ .... 24 Figure 8. Summary of ARIMA results for all bilateral trade relationships. Black lines are cases where trade flows were enhanced (hindcast values were higher than the observed trade) and red lines indicate situations where trade was offset (hindcast values were lower than the observed trade) during the pre period. Solid lines are soybe an flows, dashed are corn flows and dotted are wheat flows. ................................ ................................ ................................ ................................ ............. 26 xi ts from Argentina, Brazil, Uruguay and Paraguay. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. ................................ ................................ ....................... 31 Figure 10. Full ARIMA results fo Uruguay and Paraguay. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. ................................ ................................ ....................... 32 Brazil and Paraguay. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from th e ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. ................................ ................................ ....................... 33 Brazil and Uruguay. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. ................................ ................................ ....................... 34 Figure 13. Temporal dynamics of AOT aggregated across Heilongjiang province, China and across neighboring provinces in Russia (Amur, Khabarovsk, Primorsky and Yevrey). Panel A shows multi - annual (2006 - 2016) monthly average AOT values and Panel B shows the average AOT values for the month of November of each year from 2006 to 2016. ................................ ........... 47 Figure 14. Average seasonal dynamics of the areas where fires were detected in Heilongjiang province, Northeast China during the 2006 - 2016 period. Error bars correspond to 1 SEM. ....... 47 Figure 15. Location of the study site, Heilongjiang, China (in yellow) and neighboring Russia Provinces (in grey) with inset snapshots of the China - Russia border from NASA worldview[14 ]. The inset image snapshot is from 10/14/16 and captures smoke from agricultural fires blowing across the international border. Panel A shows 1:5000, Panel B shows 1:2500 and Panel C shows 1:1500. The red dots on the inset image map are satellite detectio ns of fires, derived from MODIS and VIIRS instruments, in the area. https://go.nasa.gov/30UQllB. ................................ .............. 48 Figure 16. Spatio - temporal dynam ics of agricultural land use (maize, soybean and mixed maize/soybean in Heilongjiang province, Northeast China. (A) Spatial distribution of agricultural land use classes during 2006. (B) Spatial distribution of agricultural land use classes during 2016. (C ) Temporal dynamics of agricultural land use classes between 2006 and 2016. ...................... 49 Figure 17. Spatio - temporal dynamics of fires in Heilon gjiang province, Northeast China. (A) Spatial distribution of fires during 2006. (B) Spatial distribution of fires during 2016. (C) Temporal dynamics of the area exposed to fire (combining low, nominal and high confidence fires) between 2006 and 2016. ................................ ................................ ................................ .............................. 51 xii Figure 18. Underlying cropscape map from 2010 showing farm survey and soil sample locations in Heilongjiang for 2016. Green is soybean and yellows is corn. Point shape corresponds to crop rotation practice. ................................ ................................ ................................ ........................... 70 Figure 19. Principle coordinates analysis based on Unifrac distances for soil microbial communities sampled from farm fields in Heilongjiang China. The samples are colored by rotation method - which include C, S and M. The distance of 3% was used to define operational taxonomic units. Significant variable s are in red and significant factors are in black. The vector points toward the direction of increase for a given variable and the arrow length indicates the correlation between the variable and the ordination scores. ................................ ................................ .......................... 73 Figure 20. Box plots for alpha diversity analysis: (A) observed species, (B) Chao 1, (C) Shannon, (D) Simpson and (E) Pielou indexes of samples grouped by rotation type. ................................ . 75 Figure 21. Relative abundance for the top 8 most abundant phyla, representing 92% of the taxa across 3 rotation pr actices: continuous corn, mixed rotations and continuous soybean. .............. 76 Figure 22. Adapted from Dou et al. Figure 63, the decision - makin g process making process in the receiving system of teleABM. The dotted lines represent linkages to the sending system which can be turned on or off to simulate different trade scenarios. The present version of the model extends beyond the cultivation deci sions and adds management decision which decisions affect soil properties that then feedback to the cultivation decisions via a profit function (i.e., yield * price). ................................ ................................ ................................ ................................ ..................... 102 Figure 23. The area planted to corn, soybean and rice overtime. ................................ ............... 103 Figure 24. Simulated land use maps showing the initial and final model tick. Soybeans are green, corn is yellow and rice is white. ................................ ................................ ................................ .. 104 Figure 25. (A) total emissions from all farms over time (b) change in total emissions from all farms over time. PM2.5 is red, CO2 is black and NOx is grey. ................................ ............................ 105 Figure 26. Average emissions of CO2 (left column), NOx (center column), and PM2.5 (right column) from farms shown by crop planting p roportion. Corn is yellow, soybeans green and rice blue. ................................ ................................ ................................ ................................ ............. 106 Figure 27. Average total organic carbon by crop type overtime. Corn is y ellow, soybeans green and rice blue. ................................ ................................ ................................ ............................... 107 Figure 28. Simulated land use maps showing the initial and final model tick. Green indicates high levels of total organic carbon, followed by yellow, orange and finally red. Red indicates degraded soil. ................................ ................................ ................................ ................................ .............. 107 Figure 29. Yield by crop type over time. Corn is shown in the left panel followed by soybean and rice. The grey shaded area represents standard deviation. ................................ ......................... 108 Figure 30. Average nitrogen overuse on farms shown by planting proportion. Corn is yellow, soybeans green and rice blue. ................................ ................................ ................................ ..... 108 1 I NTRODUC TION AND LITERATURE REVIEW Overview International food trade and globalized agriculture production connects humans and the environment around the world [ 1, 2 ] . The rate and scale of food trade has significantly increased in the past several decades, even though similar food types e xist in both importing and exporting countries [ 1, 3 - 5 ] . As consumption demands are increasingly met abroad by importing food products, environmental and socioeconomic effects of production are left in the producing region [ 6 - 8 ] . However, the effects of food trade beyond the immediate trade partners, known as spillover effects, are often understudied due to the complexity and low visibility of the socioeconomic and environmental impacts. Connections via food trade and the associated spillover effects are expected to continue to increase as p opulation growth increases the caloric demand increased affluence drives changes in consumption patterns. The highly interconnected and complex trade networks that society has come to rely pose threats that are difficult to foresee, understand and control. food and sustainability [ 9 - 11 ] . International soybean trade is representative of how distant people and places are connected. Soybean production and trade are dom inated by a small number of countries; 97% of the area planted to soybeans, globally resides in Brazil, the United States, China, Argentina and India [ 12 ] . and exporter during much of the 19 th and early 20 th centuries [ 13 ] . However, due to domestic population growth and rising affluence, China enacted a series of polices which encouraged importing 80% of its total soybean supply, importing 88 million tonnes in 201 8 [ 14 - 16 ] . Much of 2 producers and exporters [ 1, 14 ] . Much research has been conducted on either the socioeconomic or environmental impacts of international soybean trade at international and national scales [ 6, 17 - 19 ] . That is, we know broadly soybean production is increasing in Brazil a s well as several other countries (Argentina, Canada, Uruguay, Paraguay, etc. and several African countries such as Mozambique, South Africa and Zimbabwe [ 12, 15, 20, 21 ] Several studies have focused on the environmental and socioeconomic consequences of soybean expansion in the sending system, Brazil. Briefly, land use change associated with soybean expansion has led to unfavorable environmental impacts including deforestation, displacement of cattle production, biodiversity loss and increased greenhouse gas emissions [ 3, 12, 22, 23 ] . While the socioeconomic effects include economic growth of the agricultural sector including the agribusiness sector (production, proc essing, transportation, trade, etc.), increased farmer incomes, creation of on - farm employment as well as jobs within the supply chain (inputs, processing, transportation, etc.). Less research has been within the receiving system, China, but the literature does highlight low domestic soybean price s , land use change [ 24 - 26 ] , increased nitrogen pollution [ 27 ] , improved l evels of soil carbon [ 28 ] and access to livestock feed for the pork sector as results of international imported soybeans. However, little is known of the impacts of international soybean trade beyond the trading partners of Brazil and China. Specifically, how has land use change in the main production regions of Brazil and China has affected neighboring countries. Further, t he internal mechanisms behind production and farmer cultivation decisions within Heilonjiang are not clear, the environmental and socioeconomic impacts of soybean decline are not well quantified and farmer response and 3 ability to adapt is largely unknown. production region are of great importance as they may influence future soybean production and trade as well as food security and environmental sustainability across the world. The information gained from this research will be relevant to Brazil, China and the US as they may inform agricultural and environmental policy across local to international scales. Conce ptual Framework The telecoupling framework is an ideal tool for understanding the complex interactions within global soybean trade [ 1 ] , offering a novel integration of natural and social systems across distances that historically would have been studied in their respective disciplines. The telecoupling framework consists of five interrelated components: systems, flows, agents, causes, and effects [ 1 ] ( Figure 1). Systems are coupled human and natural systems that exchange information, material, and energy with other sys tems, which may be located relatively close or far away [ 1 ] . Systems can be further classified as sending (origins, exporting country), receiving (destinations, importing countries), and spillov er systems. Global soybean trade is a perfect example of this, where the importing country is the receiving system and the exporting country is the sending system. Movement of soybeans represents the flows , which could also be forms of information, materia ls, or energy responsible for connecting the two systems. Spillover systems are a byproduct of the coupled relationship between the sending and receiving systems. They may be connected to the sending and receiving systems in multiple ways: as a stopover be tween the two systems (e.g., port, truck stop), as the pathway between the two systems (e.g., transportation route), as an outside entity that is connected to the telecoupling (e.g., third party in trade negotiations), or the recipient of externalities pro duced by the telecoupling [ 1 ] (e.g., GHG emissions, food price fluctuations). Agents are decision - making entities involved in the telecoupling, which facilitate or hinder flows 4 between the systems; agents may take the form of domestic and international farmers, governments and agribusiness companies. Causes are drivers or factors originating from any system (sending, receiving, and spillover) and generate a telecoupling between a minimum of two coup led human and natural systems. Telecouplings often have multiple causes (e.g., economic, political, technological, cultural, and ecological), which are responsible for altering the dynamics (e.g., emergence, patterns, changes) of the telecoupling. Effects are the resulting socioeconomic and environmental impacts of a telecoupling, which can appear in the sending, receiving, and/or spillover system. Effects may result in feedbacks loops (to the cause), time lags (showing up years to decades later), and legacy effects (persisting for years or decades) [ 1 ] . Figure 1 . Adapte d from Liu et al.[1]) Summary of the 5 interrelated components of the telecoupling framework. 5 Study Areas dissertation analyzes spillover effects of international soybean trade between Brazil and China. In chapter 2, the trade relationships between Brazil (shown in yellow, Figure 1) and other South American soybean sending systems (e.g., Argentina, Paraguay and Uruguay shown in grey, Figure 1) were analyzed for cha move from the sending system to the receiving system to analyzes how land use change patterns in Heilongjiang, China (shown in lime green, Figure 1) resulted in air pollution spilling a cross the border to Russian Provinces (shown in grey, Figure 1). Next, chapter 4 considers the effects of land use change in Heilongjiang on the soil conditions (shown in lime green, Figure 1). Finally, chapter 5 integrates the research done in previous ch apters by simulating international trade and land use dynamics using an agent - based model (tele - ABM). 6 Figure 2 . World map highlighting the study locations for this dissertation. Brazil, the sending system is shown in yellow while China, the receiving system is shown in green. Heilongjiang China, was the site of field work and the core study area is highlighted in lime green. Additionally, the spillover systems are shown in grey and include Argentina, Paraguay and Urugu ay as well as the Russian Provinces that border Heilongjiang. Objectives of this dissertation Chapter 2 Analyzes the tele - and peri - coupled flows between South American soybean sending systems before and after China joined the W orld T rade O rganization . Ch apter 3 Explore s air pollution spillover s in Russia that are a result of land use conversion in Heilongjiang , China. Chapter 4 Assess es the impact of farmer cultivation and management decisions in Heilongjiang on the soil microbial community. Chapter 5 Eva luate s to what degree land - use changes contributes to greenhouse gas emissions and soil carbon in Heilon g jiang using the TeleABM. 7 Significance This project will fill research gaps by providing insight on regional and local cultivation patterns, their relationship to microbial community composition and function and the environmental effects of international soybean trade in the receiving country. The telecoupling framework will serve as the conceptual framework for this work and will be used to better u nderstand and identify the complex and interrelated causes and effects of international soybean trade [ 1 ] . Starting Chapter 2, the international soybean trade network will be organized conceptually using the telecoupling framework. The systems are countries that send flows of soybeans between them, the causes are the factors influencing soybean export or import behavior and the effects are soybean flows under future scenarios. This chapter will fill research gaps on how to apply quantitative methods to components of the telecoupling framework and how to integrate cross - scale data. Chapter 3 sees the continuation of the telecoupling concept by exploring how regional crop cultivation patterns, influenc ed by soybean imports, vary across Heilongjiang. This chapter will address the land use effects production at the regional scale. In addition to a regional comparison, Chapter 4 will consider the effects of ma nagement decisions (influenced by soybean imports) on microbial community composition and function at the local level. Lastly, Chapter 5 culminates this research project by integrating the previous chapters in an agent - based model. The model, created from information gathered during the household survey and soil sampling, builds upon empirical data to forecast future crop cultivation patterns and soil dynamics under different scenarios. By considering micro - scale effects of a macro - scale process we will int egrate cross - scale data to address a major research challenge. 8 TELECOUPLED FOOD TRADE AFFECTS PERICOUPLED TRADE AND INTRACOUPLED PRODUCTION Herzberger, A.; Chung, M.G.; Kapsar, K.; Frank, K.A.; Liu, J. Telecoupled Food Trade Affects Pericou pled Trade and Intracoupled Production. Sustainability 2019 , 11 , 2908. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited Abstract Technology, tran sportation and global appetites have transformed trade relationships between nearby and distant countries. The impact of distant food demand has had on local agricultural production and trade has attracted considerable scientific scrutiny, yet still little is known about how distant trade affects trade relationships between and production in adjacent countries. In this paper, we explore this important issue by examining international food trade and agriculture production, which represent how distant places are connected through trade networks. By analyzing patterns of soybean, corn and wheat trade from 1991 - 2016 under the framework of metacoupling (human - nature interactions within as well as between adjacent and distant systems), this study provides new insi ghts on the spatio - temporal dynamics of trade flows. Results reveal that telecoupled (between distant countries) trade interacts with the geo - political landscape to enhance or offset intracoupled production (within country) and pericoupled (between neighbo ring countries) trade. Evidence from the literature and the results of autoregressive integrated moving average models indicate that when restrictions are placed on direct export routes, pericoupled trade increased. The extent to which telecoupled food tra de affected pericoupled trade and intracoupled processes holds implications for the true extent of production driven by distant demands. Keywords: soybeans; international trade; metacoupling; telecoupling 9 Introduction International food trade and globali zed agricultural production connect humans and the environment around the world [ 37 - 41 ] . The rate and scale of food trade have significantly in creased in the past several decades, even though both importing and exporting countries produce and consume similar food types [ 42 ] . These interconnections may continue increasing as population growth increases caloric demand and greater affluence drives changes in consumption patterns [ 43 - 46 ] . Because these phenomena have been implicated as the main drivers of increased crop production and trade as well as their associated environmental and socioeconomic impacts, the conceptual fr amework of telecoupling has emerged to help disentangle distant human - environment interactions [ 47 ] . Soybean trade is representative of the ways distant people and places are connected through telecouplings. In the context of global food trade, much research has been conducted on the large flow o f soybeans from the Americas to China [ 12, 27, 48 - 50 ] . For example, many studies have documented production increases in Brazil and the United States (U.S.) as well as several other countries (e.g., Argentina, Canada, Uruguay, Paraguay [ 51, 52 ] ) to meet the rising demand from middle class [ 53 ] , however, the catalyst that initiat ed the flow of soybeans from west to east was China lowering the soybean import tariff from 130% to 3% in 1995 [ 48, 51 ] . This reduction quickly ed demand for, and therefore production of, soybeans around the world. At the time, the U.S. was the in 1995, but has since declined to 40% in 2016 [ 54 ] . Brazil first surpassed the U.S. in terms of Chinese market share in 2006 and then again in terms of total soybean production in 2013 [ 54 ] . Between 1995 (when China lowered their soybean import tariff) and 2016, total soybean exports 10 from Brazil to China increased by over 750,000% [ 54 ] . Numerous previous studies have highlighted a highly concentrated trade network where Brazil and the U.S. account for 80% of global soybean exports and China makes up 64% of global soybean imports [ 42, 46, 50, 54 ] . Given the economic [ 24, 53 ] , environmental [ 12, 55 ] and political [ 56, 57 ] importance of soybean trade, the dynamics among Brazil, China and the U.S. have been widely studied by academics [ 49, 58, 59 ] , governments [ 60 - 62 ] , industry [ 63 ] and NGOs [ 64 ] . While the literature documents production increases in South American countries in response [ 52, 56 ] altered trade relationships among South American co untries. Few studies have looked beyond structure of trade between medium - size producers or the effects that high volume soybean trade has had on them. To address this research gap, this study uses the metacoupling framework [ 65 ] , which is an extension of the telecoupling framework [ 47 ] , to explore how telecoupled (e.g., distant) soybean trade between Brazil and China has influenced pericoupled (e.g., trade between neighboring countries) and intracoupled (e.g., production within a country) processes within South America. To study the interaction between telecouplings and pericouplings, we identified soybean exports from Brazil to China as our focal telecoupling, because it is the largest bilateral exchange of soybeans [ 54 ] . To explore the influence of telecouplings on pericouplings, Argentina, Paraguay and Uruguay were identified for several reasons. First, they share a border and joint membership in the Mercosur trade agreement with Brazil allowing feasible pericoupled trade both, geographically and politically [ 66, 67 ] . Second, Argentina, Paraguay and Uruguay are net soybean exporters to Brazil, while Bolivia, Chile, Colombia, Ecuador, Peru and Venezuela are net 11 importers from Brazil and do not send soybeans to China. Because of their lack of participation in the Chinese soybean trade, these countries were excluded from further analysis. Third, Argentina, Paraguay and Uruguay have ideal climatic and environmental conditions for soybean pro duction and were the respective 3 rd , 4 th , and 7 th largest soybean exporters in 2016 [ 54 ] . Last, the literature - scale soybean production on regional cropping patterns [ 55, 68, 69 ] . Corn and wheat trade were identified as possible relationships to be affected by soybean expansion based on production - substitute suitability, status as globally important food crops, literature review, preliminary analyses as well as data availability. After analyzing production and trade da ta to identify patterns in bilateral trade relationships of soybeans, corn and wheat, literature review and autoregressive integrated moving average models (ARIMA) were used to identify changes in trade patterns before and after China lowered their soybean tariff. Methods 2.3.1. Conceptual Framework This study treats the South American food trade network as a metacoupled system. The metacoupling framework is an umbrella framework that examines three types of human - nature interactions. Human - nature interactions oc curring within a coupled human and natural system are called intracouplings. Pericouplings occur when human - nature interactions cross boundaries between adjacent systems, while telecouplings examine human - nature interactions across distance. By providing a typology that categorizes processes as intracouplings, pericouplings, and telecouplings (Figure 3 ), the metacoupling framework provides a structure for developing a more complete understanding of the complexity within trade networks. In relation to teleco upled trade between South American soybean - sending systems (e.g., Brazil, Argentina, Paraguay and Uruguay) and China, trade relationships among the neighboring countries in South America are 12 classified as pericouplings and production within a country is co nsidered as intracoupling. Because 30% of all global soybean trade flows in 2016 occurred between Brazil and China [ 54 ] , this relationship is considered the main soybean tel ecoupling and a driver of structural change in the metacoupled soybean trade network. Figure 3 . Conceptual overview of the metacoupling framework adapted from Liu (2017). The green circles represent natural components while the orange circles represent human components of the system. Grey arrows are interactions within systems (i.e., intracoupling) and black arrows are interactions between systems (i.e., peri - and telecouplings). 2.3.2. Relevant Theories Originating in the interdis ciplinary field of land system science, the metacoupling framework aims to capture various socioeconomic and environmental interactions as well as their impacts, at multiple distances. The metacoupling concept and framework [ 65 ] is an extension of the telecoupling concept and framework [ 47 ] which was derived from the integration of the concepts such as telec onnections (e.g., distant climatic connections [ 70 ] ) and globalization (e.g., distant connections among human systems [ 71 ] ). The metacoupling concept and framework are also supported through the integration of a variety of interdisciplinary concepts and theories [ 72 ] . For 13 example, metacoupled systems can be treated as complex adaptive systems [ 73, 74 ] with interacting feedbacks and spillovers at multiple scales. The metacoupling framework can also be integrated with traditional theories of international trade. International trade theory suggests that every country has a comparative advantage in terms of a good or service that could be produced at a relatively lowe r cost than other countries. Countries producing similar goods and services still trade with one another, because comparative advantages stem from not only differences in climate as well as natural and human resources but also from differences in technolog y, economies of scale, preference for variety, and other factors [ 75 ] . By framing this study under the metacoupling fram ework and incorporating interdisciplinary concepts, such as theories of international trade, we were able to identify multiple drivers that influence tele - and pericoupled trade relationships. 2.3.3. Data Collection Soybean, corn and wheat production data were collected from FAOstat for the years 1991 to 2016 [ 15 ] . Total imports, total exports, and bilateral crop trade data were collected from UNComTrade [ 54 ] under HS code 1201 (soybeans), 1005 (corn), and 1001 (wheat) for the same time period. The study period 1991 - 2016 was chosen to capture change in the soybea n trade network since the Chinese soybean tariff was lowered from 130% to 3% in 1995 [ 27, 51 ] . While China did not join to the World Trade Organization (WTO) until 2001, the study period was chosen thus split into two periods, be fore (1991 - 1995) and after (1996 - entered the global market. Due to limited trade during the pre - period values are reported as average to avoid comparing trade values in 2016 to a trade value of 0. Trade values below 25,000 MT, which is 0.0007% of the soybeans exported from Brazil to China in 2016, were excluded from the 14 analysis. Furthermore, many countries did not report bilateral trade data prior to 1991 which limited further historical analysis. While this study discusses fou r countries (i.e., Brazil, Argentina, Paraguay and Uruguay) in detail, data were collected and analyzed for all countries that were available during the study period (~75 countries). 2.3.4. Data Analysis To identify trends in the production and trade of soybeans, corn, and wheat both within and between country pairs from 1991 to 2016, descriptive statistics, data visualization and trend analyses were performed. To understand the drivers behind these trade trends, literature was collected from academic, government and NGO sources. Finally, to capture the interaction between tele - between South America sending systems, time series autoregressive integrated moving average (ARIMA) models [ 76 - 78 ] were specified for each commodity (i.e., soybeans, corn and wheat) and bilateral country pair (e.g., China, Brazil, Argentina, Uruguay and Paraguay). When the dependent variable was telecoupling (e.g., soybean export to China), pericoup ling (e.g., cross - border trade) was added as an independent variable and vice versa, to test for significance. Several country pairs had very little or no trade for certain commodities during the study period and are not discussed in the manuscript, full A RIMA results can be found in the appendix (Figures 9 - 12 , Table 2 ). Following the protocol specified in [ 77 ] , outliers were smoothed and missing values were linearly interpolated before diagnostic tests were used to determine parameters and the best model fit using R packages tseries [ 79 ] , forecast [ 77 ] and ggplot2 [ 80 ] . The ARIMA models capture time series trends and predict future val ues of trade ( at time by including autoregressive (AR( p )), integrated (I( d )) and moving average (MA( q )) model terms. The AR model term assumes the 15 current value of trade between two countries is a linear function of the previous trade values and therefore includes p time lags to predict future values of (1). (1) AR w here is a constant, are parameters and is AR polynomial order. The integrated model term I refers to the differencing degree which is calculated by subtracting the current of trade between two countries and previous values of trade times to stabilize the time ser ies , w here is differenced times (2a - b). (2a) I (2b) AR+I The MA model term calculates future error terms, at time , by combining previous errors terms, , where indicates the number of error terms included (3). (3) MA w here is a constant, are parameters and is the MA polynomial order. Through the combination of model terms, ARIMA models (4) are commonly used to recreate a time series trend and then project into the future for predictive purposes. (4) AR+I+MA Last, the ARIMA models were modified to test the influence of pericoupling processes ( ) on telecoupling processes ( ) (5). (5) AR+I+MA+X Where is a covariate at time and is its coefficient. While ARIMA models are traditionally used for the purpose of forecasting, in the present case it is more informative to hindcast the predicted values in order to examine tariff reduction on soybean trade. Therefore, the ARIMA models were trained on trade data in the 16 post period (e.g., 1996 - 2016) and then hindcast for the pre period (e.g., 1991 - 1995). This method allowed for comparison between the o bserved and hindcast trade data among South American sending systems before China lowered the soybean import tariff. Because the hindcast values were calculated from the trend in the post period, differences between the observed and hindcasted data approxi Brazil and nearby countries (e.g., pericoupling). Results 2.4.1. Trend Analysis Overview While several countries in South America, as well as around the world, increased intracoupled the global market [ 59 ] sending system may be due to strategies used by the Brazilian government to liberalize trade and increase im ports of and farmer access to fertilizers, pesticides, and seeds [ 81 ] . Since the pre - period average (1991 - 1995), Brazil steadily increased soybean imports in 2016. The remaining 2016 market shares belonged to the U.S. (40%), Argentina (9%) followed by Uruguay and then Canada, who each accounted for less than 2% of [ 54 ] . While, Argentina, Paraguay and Uruguay share a border with Brazil as well as suitable land for soybean cultivation, their varying shares of the Chinese soybean market are largel y reflective of the unique geo - government . 17 Argentina Exports of soybeans from Argentina to China started in 1995 and experienced rapid growth during the first decade of the study period. However, in 2 006, the Argentine government placed a domestic tariff on exports of soybean of 23.5% which resulted in slower growth in soybean exports to China [82] (Figure 4 ). Soybean exports to other South American countries were exempt from the tariff through joint membership in the Mercosur trade agreement [66] . Between the pre - period corresponding with a cessation of soybean exports from Argentina to Brazil in 1995 that did not contrast, since 2006 soybean exports from Argentina to Brazil, Paraguay and Uruguay have increased by 128,766%, 128% and 88%, respectively, while exports to China increased by 23% ( Figure 5, Table 1 ) [54, 83] . Furthermore, the rate of soybean expansion by area planted in Argentina was much higher before the domestic tariff started in 2006 (e.g., 215% increase between 1995 - 2005 vs. 45% increase between 2006 - 2016). In contrast, the rate of corn expansion by area planted was much higher after 2006 (e.g., 80% increase between 1995 - 2005 vs. 175% increase between 2006 - 2016), while the rate of wheat expansion by area planted occurred at approximately increase between 1995 - 2005 vs. 46% increase between 2006 - 2016) (Figure 5, Table 1 ). Along with soybean expansion and drove corn expansion while wheat intra coupling has remained relatively stable during the study period [ 24, 84 - 88 ] . 18 Figure 4 . Timeline that defines the pre period (1991 - 1995) and post period (1996 - 2016) and highlights important events. Uruguay Differing from Argentina, Uruguay has tariff - free access to the Chinese soybean market as well as land prices that are about 50% cheaper than soybean land in Argentina. These factors, in combination with the financial crisis during the early 2000s, promoted investment in Uruguayan agriculture from nearby producers. In addition to investments from multinational companies, soybean expansion i n Uruguay was driven primarily by investments from Argentina [63] . The flows of investment capi increasing exports to China by 1,250% between 2005 and 2016 [54] . In 2007, two years after Uruguay began expor ting soybeans to China, its soybean imports from Argentina, Brazil and Paraguay were equal to 60% of the amount Uruguay exported to China. As Uruguay increased soybean intracoupling, the contribution from nearby countries had become consistent but small. H owever, the rapid establishment of soybean production and foreign land ownership limited the development of soybean processing industry. Therefore, Uruguay is a net importer of both soybean oil and meal to support its a booming livestock industry, 100% of which comes from Argentina, Brazil and Paraguay [54] . 19 Figure 5 . Flow maps depicting telecoupled soybean trade and pericoupled trade of soy beans (green arrow), corn (orange arrow) and wheat (blue arrow) from 1991 - 2016. Panels are split into 5 - year time periods: 1991 - 1995 (A), 1996 - 2000 (B), 2001 - 2005 (C), 2006 - 2010 (D) and 2011 - 2016 (E). The width of the line represents average commodity flow from a single trade partner during the respective time period and the arrows are weighted according to total import from all trade partners. All values are presented in millions of metric tonnes. Trade amounts averaging below 25,000 MT per year were exclu ded. 20 Paraguay Unlike Argentina and Uruguay, Paraguay chose to forgo diplomatic ties with China and market access due to ties with Taiwan [ 89 ] . Despite having restrict ed access to the largest soybean more than half of which were exported [ 15, 64, 90 ] . Due to the geographic and political landscape, Paraguayan soybeans were sent via barge to river ports in Argentina and Uruguay [ 89, 91 ] . Prior to 2009, exports to Argentina , Brazil and Uruguay accounted for between 40% and 70% of and Turkey. Russia and Turkey are net exporters of processed soybean products and export soybean oil to China [ 54 ] and Uruguay, while 15% went to Russia and 13% to Turkey [ 54 ] . While re - export analysis was not performed in the present study, the literature confirms that even though Paraguay does not directly engage in soybean trade with China, intracoupled soybean production, and pericoupled [ 89 , 91, 92 ] . 21 Table 1 . om Argentina, Paraguay, Uruguay and the U.S. The table includes the trade value for the pre period (the average trade value between 1991 and 1995), 2016 and the percent change between the pre period and 2016. Brazil's Soybean Imports Argentina Paraguay Uruguay USA Pre Period 129,079 177,765 33,016 187,397 2016 670 381,448 0 0 Percent Change - 99.5% 114.6% - 100% - 100% Brazil's Corn Imports Argentina Paraguay Uruguay USA Pre Period 811,167 80,831 5,188 123,894 2016 1,436,245 1,465,053 0 532 Percent Change 77% 1712% - 100% - 99.5% Brazil's Wheat Imports Argentina Paraguay Uruguay USA Pre Period 1,074,906 14,544 47,287 173,355 2016 3,950,036 956,125 577,415 1,226,210 Percent Change 268 6474% 1121% 607% 2.4.2. Differences in Trade Between the Pre - and Post - Period Trends Overview To explore differences in trade between the pre - period trend (e.g., 1991 - soybean tariff reduction) and the post - period trend (e.g., 1996 - an tariff reduction), autoregressive integrated moving average (e.g., ARIMA) models were constructed. For illustrative purposes, Figure 6 shows the ARIMA results for soybean exports from Argentina, Brazil and Uruguay to China. In all three cases, the hindc ast values (e.g., dashed line) for the pre period were higher than the observed values (e.g., solid black line), indicating that based on the trend during the post period, more trade was expected during the pre period. Differences in the hindcast values an d observed trade values during the pre period indicate that the trade trends before pericoupled imports are shown in Table 2 22 relationship with their pericoupled soybean imports. Many other factors could drive bilateral trade relationships, such as population a nd economic growth, however because the models were constructed from the observed data the impact of major factors is endogenized in the ARIMA trend. The varying degrees of access to the Chinese soybean market by Argentina (somewhat restricted access), Uru guay (unrestricted access) and Paraguay (completely restricted) allow for a unique analysis of how telecoupled, pericoupled and intracoupled processes interact. Figure 6 to China. See SF1 for full country and commodity results. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. P - v alues, NRMSE (normalized by standard deviation) and MASE can be found in Table S1. Argentina The observed trade data show that soybean exports from Argentina to Brazil have declined by 99% between the pre period (average of 1991 - 1995) and 2016 while export s of corn and wheat have increased 77% and 267%, respectively (Table 1) [ 54 ] . Using the trend from post period to hindcast the values for the pre period, the ARIMA results indicate that the observed soybean and corn exports from Argentina to Brazil during the pre period were higher than the hindcast values while the observed wheat exports were lower than the ARIMA values (Figure 7 A1). With respect to soybean trade, the lower hindcast values reflect when the observed soybean exports from Argentina to Brazil dropped to zero in 1995 and did not start again consecutively until 2005, w hich 23 respectively correspond with the reduction of the Chinese import tariff and the start of the Argentine export tariff. The ARIMA results suggest that fewer soybean exports from Argentina to Brazil reflect the observed trend during the first half of th e study period. However, between 2006 and 2016 soybean exports from Argentina to Brazil increased by over 400% [ 54 ] . The opposite trend observed during the second half of th e study period may have masked the and contributed to the relatively small difference between the observed and hindcast values (Table 2 ). These results suggest that pericoupled soybean trade between Argentina and Brazil was reduced between 1995 and 2005 an d then enhanced between 2006 and 2016. Similarly, the hindcast values for corn exports from Argentina to Brazil are slightly lower than the observed data (Figure 7 B1) which supports the literature finding that corn expansion in Argentina was driven by bio fuel mandates in Europe and the U.S. during the mid - [ 88 ] . Competition from biofuel mandates, as well as other drivers, may have offset pericoupled corn trade between Argentina and Brazi l during the post period. Differing from soybeans and corn, the observed wheat imports from Argentina to Brazil were much lower than the hindcast values (Figure 7 C1), indicating that while pericoupled corn and soybean trade were offset during the post per iod, pericoupled wheat trade was enhanced. 24 Figure 7 Paraguay (3). The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Uruguay Based on hindcasting the post - period trend, the ARIMA values for soybean and corn exports from Uruguay to Br azil were lower than the observed values during the pre period (Figure 7 & Table 2 ). This is because observed soybean and corn exports from Uruguay to Brazil during the wered, soybean exports from Uruguay to Brazil decline to zero and do not start again consistently . Further, corn exports from Uruguay to Brazil both increase and decrease but trade only occurred intermittently during the post period. The highly variable pa ttern and lower hindcast values indicate that exports of corn and soybean from Uruguay to Brazil were reduced in the post period. In contrast, the 25 observed values for wheat exports from Uruguay to Brazil were substantially lower than the hindcast values (F igure 7 C2). The observed exports from Uruguay to Brazil increased by 1121% since the pre period (Table 1), indicating that pericoupled wheat trade was enhanced during the post period. Additionally , there was a significant relationship ybean exports to China and their soybean imports from Argentina, Brazil and Uruguay. Further, the hindcast values were higher than the observed trade for exports of corn and wheat from Argentina to Uruguay and exports of corn and wheat from Paraguay to Uru guay (Figure 11 ), suggesting pericoupled imports were enhanced during the post period. Paraguay In contrast to Argentina and Uruguay, Paraguay showed a consistent pattern across soybean, corn and wheat exports to Brazil. Paraguay is the only cou ntry that is completely restricted from accessing the Chinese market and experienced increases in exports to Brazil of all three crops. Specifically, since the pre - period (average of 1991 - 1995) exports of soybeans, corn and wheat to Brazil have increased b y 114%, 1,712% and 6,474% , respectively (Table 1). The observed values for soybean, corn and wheat exports from Paraguay to Brazil, during the pre period were all lower than the ARIMA hindcast values (Figure 7 , C1, C2 & C3). The consistent pattern across crops indicates that pericoupled trade between Paraguay and Brazil was enhanced during the post period. significant exports to China (Table 2 ). Further, in supplement al Figures 10 and 11 , the hindcast values are higher than the observed soybean exports from Paraguay to Argentina and soybean and corn exports from Paraguay to Uruguay. 26 Figure 8 . Summary of ARIMA results for all bilateral trade relationships. Black lines are cases where trade flows were enhanced (hindcast values were higher than the observed trade) and red lines indicate situations where trade was offset (hindcast values were lower than the observed trade) during the pre period. Solid lines are soybean flows, dashed are corn flows and dotted are wheat flows. Discussion demand, domestic policies and international geo - politic al relationships (Figure 8 imports of wheat substantially increased from Argentina, Uruguay and Paraguay during the study period. The ARIMA models confirm that wheat trade during the post period was enhanced compared to the pre period. In ay and Paraguay varied and may be related to the respective government strategies used to facilitate or inhibit trade. Soybean exports from Argentina to Brazil decreased in the first half of the study period (e.g., 1995 - 2005) and increased in the second ha lf (e.g., 2006 - 2016), corresponding with 27 exports from Argentina to Brazil increased during the study period, differences between the observed and hindcas t values suggest that pericoupled corn trade was offset in the post period. demand may have interacted to offset and enhance pericoupled soybean and corn trade betw een Argentina and Brazil during the post period. Uruguay did not start exporting soybeans to China until 2005, but since then has increased exports. Both soybean and corn exports from Uruguay to Brazil declined after the pre period and only occur intermitt ently. Additionally, Uruguay increased imports of corn from Argentina and Paraguay. The ARIMA models indicate differences in the trend between the pre and post period which supports the observation that pericoupled trade was offset in the post period. Thir d, soybean and corn exports from Paraguay to Brazil increased and post period trends and suggest pericoupled trade was enhanced during the post period. The inability to access the Chinese market directly may have contributed to increased pericoupled trade with Brazil as well as Argentina and Uruguay, which offer conveni ent export routes [ 89, 93 ] . and soybean products, may compete with pericoupled exports and provide further evidence that Paraguay is indirectly influenced by the Chinese market. In the meta - and telecoupling literature there are no studies that specifically address the synergy between telecoupling and pe ricoupling processes. However, a number of studies document spillover effects of Brazilian soybean expansion. For example, Bicudo da Silva et al. found that as a result of increased soybean production in Brazil, farmers increased operations from growing 1 to 2 crops each year which increases risk of precipitation anomalies. Soybeans are 28 planted first, followed by corn, most of which is for domestic consumption rather than export [ 24 ] . Further, Dou et al. found that conservation efforts (i.e., Soy Moratorium and zero - deforestation beef agreement) meant to contain deforestation in the Amazon biome displaced deforestation to the Cerrado region [ 55 ] . In the respective cases, soybean expansion temporally displ aced corn the present study indicate that similar interactions may occur across system boundaries where telecouplings interact with pericouplings to enhance or offset one another. On the other hand, a network analysis of global soybean trade [ 50 ] identified Brazil, China and the USA as the key players but did not find regional or continental geography to be predictor of trade patterns. The present findings suggest that pericoupled relationships are important in global trade patterns and may be dwarfed, contextually and computationally, by the massive flows among Brazil, China and the USA. Therefore, analyses that only consider direct trade routes or bilateral exchanges could mask the true value of the flow and therefore underrepresent the land - use impact of production. Because soybean expansion has already threatened ecosystems and displaced land - use in Brazi l [ 24 ] , Argentina [ 56 ] and Uruguay [ 69 ] , stakeholders should consider pericoupled trade when designing policy interventions. ses in soybean demand on pericoupled trade relationships between Brazil, the largest soybean exporter, and the adjacent soybean - sending systems of Paraguay, Uruguay and Argentina. We note that this analysis lacks systematic consideration of other factors t hat could have driven differences between the pre and post period. For example, population growth and income level have been shown to be significant factors for food trade [ 43 ] and were not explicitly considered in this analysis. While the lack of other consideration is valid, ARIMA models accurately capture time - series trends by 29 endogenizing the impact of major drivers through accounting for non - stationarity, seasonality and au to - correlation. ARIMA models use lags of the dependent variable or lags of errors as independent variables to capture the underlying systematic patterns in the data [ 76, 94 ] . Additional independent variables can be added to ARIMA - X models, however those factors should not have been affected by the intervention in question, i.e. China acceding to the WTO, and can result in over - specification. Further, ARIMA models have been shown to capture the same trend with only slight differences from ARIMA - X models [ 76 ] . Therefore, the ARIMA models can determine if the trends between the pre and post periods differ, however, those differences cannot be casually literature support this h ypothesis [ 95 ] . One possible future research direction is to consider the impact of telecouplings on pericoupled trade among receiving systems. Conclusions The present research highlights the importance of pericoupled relationships in global trade and provides insights on how commodities flow in a metacoupled world. Despite government (e.g., demand affected intracoupling and trade both directly, via telecouplings, and indirectly through pericouplings. Th is suggests that in a metacoupled world commodities flow from areas of supply to areas of demand using the most cost - effective route [ 96 ] . If that route is restricted, the commodity will likely still flow to the area of demand, but via a different route. These results should be considered by and may stimulate future studies on metacoupled systems (e.g., beyond soybean trade, such as migration, tourism, species invasion) and provide insights for governing metacoupling systems. 30 APPENDIX 31 Figure 9 Paraguay. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. 32 Figure 10 soybean, corn and wheat imports from Brazil, Uruguay and Paraguay. The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95 % CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. 33 Figure 11 . The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank gr aphs indicate no or very little trade occurred. 34 Figure 12 The solid line represents the natural log of imports in metric tons and the dashed line represents the natural log of the trend derived from the ARIMA models. The shaded grey area is the 95% CI. Trade values below 25,000 MT per year were excluded. Blank graphs indicate no or very little trade occurred. 35 Table 2 . Description of final ARIMA models including p (# of lags), d (# of differences), q (# of error terms). Followed by parameter descriptions and p - values, (***p < 0.001, **p< 0.05 and *p < 0.01). AR and MA parameters refer to the autoregressiv e and moving average model terms. The pericoupled parameters are relative to the country in the dependent variable and are the aggregated imports from the adjacent countries. For example, if Y = Soybean exports from Argentina to China, then pericoupled soy Uruguay and Paraguay. The fourth column includes the sum of the differences between the observed and predicted values for 1991 - 1995 in MT. Positive numbers can be interpreted as a predicted increase in trade during the pre period of X MT compared to the observed values; negative numbers indicate a predicted decrease of X MT during the pre period relative to the observed values. Finally, normalized root mean squared error (NRMSE) and mean abs olute scaled error (MASE) were included as measures of model validity. NRMSE was normalized by dividing the RMSE by the standard deviations of the dependent variable. In both cases, values close to zero indicate a high performing model that captures the va riation in the data. Values around 1 indicate the model is explaining some of the variation in the data, while values beyond 1 indicated noise and a low - performing model. ARIMA Model (p, d, q) Parameter p - value - Predicted values) NRMSE MASE Soybean exports from Argentina to China (0, 1, 1) MA1 0.00*** 21,065,499 0.55 0.82 Pericoupled soybean imports 0.27 Pericoupled corn imports 0.63 Pericoupled wheat imports 0.69 Soybean exports from Brazil to China (0, 2, 1) MA1 0.00*** 12,115,029 0.11 0.71 Pericoupled soybean imports 0.05* Pericoupled corn imports 0.7 Pericoupled wheat imports 0.1 Soybean exports from Uruguay to China (1, 1, 0) AR1 0.00*** 174,735 0.06 0.4 Pericoupled soybean imports 0.05* Pericoupled corn imports 0.9 Pericoupled wheat imports 0.69 Soybean exports from Argentina to Brazil (0, 1, 1) MA1 0.00*** - 515,895 0.78 0.7 soybean exports to China) 0.8 Soybean exports from Uruguay to Brazil (0, 2, 1) MA1 0.00*** - 180,406 0.37 1.1 soybean exports to China) 0.8 Soybean exports from Paraguay to Brazil (0, 2, 1) MA1 0.00*** 2,217,646 0.71 0.92 soybean exports to China) 0.3 Corn exports from Argentina to Brazil (0, 1, 0) soybean exports to China) 0.38 - 2,054,664 0.63 0.9 Corn exports from Uruguay to Brazil (0, 1, 0) soybean exports to China) 0.9 - 18,305 0.4 0.97 Corn exports from Paraguay to Brazil (0, 1, 1) MA1 0.05* 423,823 0.51 1 soybean exports to China) 0.34 Wheat exports from Argentina to Brazil (0, 1, 0) soybean exports to China) 0.58 12,012,323 0.6 0.94 36 Table 2 Wheat exports from Uruguay to Brazil (1, 2, 1) AR1 0.00*** 911,268 0.3 0.7 MA1 0.00*** soybean exports to China) 0.46 Wheat exports from Paraguay to Brazil (3, 1, 1) AR1 0.00*** 505,229 0.45 0.63 AR2 0.00*** AR3 0.00*** MA1 0.00*** soybean exports to China) 0.00*** Soybean exports from Paraguay to Argentina (1, 1, 0) AR1 0.00*** 248,227 0.12 0.8 Soybean exports from Paraguay to Uruguay (2, 0, 1) AR1 0.00*** 417,083 0.23 0.79 AR2 0.06 MA1 0.00*** Corn exports from Argentina to Uruguay (1, 0, 0) AR1 0.00*** 37,552 0.48 0.89 Corn exports from Paraguay to Uruguay (0, 1, 0) 1,565 0.47 0.9 Wheat exports from Argentina to Uruguay (1, 0, 0) AR1 0.00*** 65,416 0.37 0.9 Wheat exports from Argentina to Paraguay (0, 1, 0) 271,462 0.45 0.96 Author Contributions: All authors were involved in the conceptualization, writing, reviewing and editing of the manuscript; A.H. and M.C. curated the data, developed the methodology and performed the analysis.; A.H. and K.K. created the visualizations in the manuscript; K.F. and J.L. provided supervision; J.L. is responsible for project administration and funding acquisition. Funding: This resea rch was funded by the National Science Foundation and Michigan State University AgBio Research 37 REFERENCES 38 REFERENCES 1. Liu, J.; Hull, V.; Batistella, M.; DeFries, R.; Dietz, T.; Fu, F.; Hertel, T.W.; Izaurralde, R.C.; Lambin , E.F.; Li, S.; Martinelli, L.A.; Mcconnell, W.J.; Moran, E.; Naylor, R.L.; Ouyang, Z.; Polenske, K.R.; Reenberg, A.; de Miranda Rocha, G.; Simmons, C.S.; Verburg, P.H.; Vitousel, P.M.; Zhang, F.; Zhu, C., Framing Sustainability in a Telecoupled World Ecol ogy and Soceity 2013. 18 (2). 2. Young, O.R.; Berkhout, F.; Gallopin, G.C.; Janssen, M.A.; Ostrom, E.; van der Leeuw, S., The globalization of socio - ecological systems: An agenda for scienific research Global Environmental Change, 2006. 16 : p. 304 - 316. 3. Garnett, T.; Appleby, M.C.; Balmford, A.; Bateman, I.J.; Benton, T.G.; Bloomer, P.; Burlingame, B.; Dawkins, M.; Dolan, L.; Fraser, D.; Herrero, M.; Hoffmann, I.; Smith, P.; Thornton, P.K.; Toulmin, C.; Vermeulen, S.J.; Godfray, H.C.J., Sustain able Intensifi cation in Agriculture : Premises and Policies. Science, 2013. 341 . 4. Richards, P.D.; Myers, R.J.; Swinton, S.M.; Walker, R.T., Exchange rates, soybean supply response, and deforestation in South America. Global Environmental Change 2012. 22 : p. 454 - 462. 5. Reenberg, A.; Fener, N.A., Globalizing land use transistions: the soybean acceleration. Geografisk Tidsskrift - Danish Journal of Geography, 2011. 11 (1): p. 85 - 92. 6. Garrett, R.D.; Lambin, E.F.; Naylor, R.L., The new economic geography of l and use change: Supply chain configurations and land use in the Brazilian Amazon. Land Use Policy, 2013. 34 : p. 265 - 275. 7. Lathuilliere, M.J.; Johnson, M.S.; Galford, G.L.; Couto, E.G., Environmental footprints propriation for soybean production in Mato Grosso, Brazil. Environmental Research Letters 2014. 9 . 8. MacDonald, G.K.; Brauman, K.A.; Sun, S.; Carlson, K.M.; Cassidy, E.S.; Gerber, J.S.; West, P.C., Rethinking agricultural trade relationships in an era of globalization. BioScience 2015. 9. Cleveland, D., Balancing on a planet: The future of food and agriculuture . 2014, Berkeley and Los Angeles, California University of California Press. 10. DeFries, R.; Rudel, T.K.; Uriarte, M.; Hansen, M., Deforestation dr iven by urban populatio nand agriculutural trade in the twenty - first century nature geoscience 2010. 3 : p. 178 - 181. 11. Brown, L.R. Plan B Updates Can the World Feed China? 2014 [cited 2015; Available from: http://www.earth - policy.org/plan_b_updates/2014/ update121 . 39 12. Reenberg, A.; Fener, N.A., Globalizing land use transistions: the soybean acceleraition Geografisk Tidsskrift - Danish Journal of Geography 2011. 111 (1): p. 85 - 92. 13. Pratap, A.; Gupta, S.K.; Kumar, J.; Solanki, R.K., Soybean in Technological Innovations in Major World Oil Crops 2012, Spriner Science+Business Media p. 293 - 321. 14. Salin, D.L., Soybean Transportation Guide: Brazil 2014 2015, U.S. Dept. of Agriculture, Agricultural Marketing Service . 15. Food and Agriculture Organization of the United Nations, FAOSTAT . 2014, FAO. 16. Sun, J.; Wu, W.; Tang, H.; Liu, J., Spaitiotemporal patterns of non - genetically modified crops in the era of expansion of genticaly modifed food Scientific Reports 2015. 17. Brown - Lima, C., An overview of the Brazil - China soybean trade and trategic implications for conservation . The Nature Conservancy Latin America n Region. 18. Garrett, R.D.; Rueda, X.; Lambin, E.F., production in South America: linkages between preferences for non - genetically modified crops, eco - certifications, and land use. Environmental Research Letter s, 2013. 8 . 19. Karstensen, J.; Peters, G.P.; Andrew, R.M., Attribution of CO2 emissions from Brazilian deforestation to consumers between 1990 and 2010. Environmental Research Letters, 2013(8). 20. Gasparri, N.I.; Kuemmerle, T.; Meyfroidt, P.; le Polain d eWaroux, Y.; Kreft, H., The emerging soybean production frontier in Sothern Africa: Conservation Challeneges and the role of south - south Telecouplings. . Conservation Letters, 2015. 21. Rudel, T.K.; Schneider, L.; Uriarte, M.; Turner, B.L.; DeFries, R.; La werence, D.; Geoghegan, J.; Hecht, S.; Ickowitz, A.; Lambin, E.F.; Birkenholtz, T.; Baptista, S.; Grau, R., Agriculutral Intensifitication and changes in cultivated areas, 1970 - 2005 PNAS, 2009. 106 (49 ): p. 20675 - 20680 . 22. Nepstad, D.; McGrath, D.; Stickl er, C., Slowing Amazon deforestation through public policy and interventions in beef and soy supply chains. Science, 2014. 344 . 23. Martinelli, L.A.; Naylor, R.L.; Vitousel, P.M.; Moutinho, P., Agriculture in Brazil: impacts, costs, and opportunities for a sustainable future Current opinion in Environmental Sustainaibility, 2010. 2 : p. 431 - 438. 24. Bicudo da Silva, R.F.; Batistella, M.; Dou, Y.; Moran, E.F.; Torres, S.M.; Liu, J., The Sino - Brazil ian Telecoupled Soybean System and Cascading Effects for the Exporting Country. Land 2017. 6 (53): p. 1 - 19. 25. Dou, Y.; Millington, J.D.A.; Bicudo Da Silva, R.F.; McCord, P.; Viña, A.; Song, Q.; Yu, Q.; Wu, W.; Batistella, M.; Moran, E.; Liu, J., Land - use changes across distant places: design of a telecoupled agent - based model. Journal of Land Use Science, 2019. 14 (3): p. 191 - 209. 40 26. Dou, Y.; Yao, G.; Herzberger, A.; da Silva, R.; Song, Q.; Hovis, C.; Batistella, M.; Moran, E.; Wenbin, W.; Liu, J., Land - Us e Changes in Distant Places: Implementation of a Telecoupled Agent - Based Model. Journal of Artificial Societies and Social Simulation, 2020. 23 . 27. Sun, J.; Tong, Y.; Liu, J., Telecoupled land - use changes in distant countries. Journal of Integrative Agric ulture, 2017. 16 (2): p. 368 - 376. 28. Tong, Y.; Liu, J.; Li, X.; Sun, J.; Herzberger, A.; Wei, D.; Zhang, W.; Dou, Z.; Zhang, F., Scientific Reports, 2017. 7 (1): p. 18064. 29. Cellini, F.; Chesson, a.; Colquhoun, I.; Constable, a.; Davies, H.V.; Engel, K.H.; Gatehouse, a.M.R.; Kärenlampi, S.; Kok, E.J.; Leguay, J. - J.; Lehesranta, S.; Noteborn, H.P.J.M.; Pedersen, J.; Smith, M., Unintended effects and their detection in gene tically modified crops. Food and chemical toxicology : an international journal published for the British Industrial Biological Research Association, 2004. 42 : p. 1089 - 125. 30. Lambin, E.F.; Meyfroidt, P., Land use transitions: Socio - ecological feedback ve rsus socio - economic change. Land Use Policy, 2010. 27 : p. 108 - 118. 31. Meyfroidt, P.; Rudel, T.K.; Lambin, E.F., Forest transitions, trade, and the global displacment of land use. PNAS, 2010. 7 (49): p. 20917 - 20922. 32. Börner, J.; Mendoza, A.; Vosti, S.a., Ecosystem services, agriculture, and rural poverty in the Eastern Brazilian Amazon: Interrelationships and policy prescriptions. Ecological Economics, 2007. 64 : p. 356 - 373. 33. Li, J., Statistical Yearbook of Heilongjiang State Farms 2011, Statistical Bureau of Heilongjiang Agricutlural Reclaimation p. 365 . 34. Simms, E.L.; Taylor, D.L., Partner Choice in Nirtogen - Fixation Mutualisms of Legumes and Rhizobia Integrative and Comparative Bi ology 2002. 42 : p. 369 - 380 . 35. Herzberger, A.; Duncan, D.; Jackson, R., Bouncing Back: Plant - Associated Soil Microbes respond rapidly to Prairie Eastablishment PLOS one, 2014 9 (12). 36. Richter - Heitmann, T.; Eickhorst, T.; Knauth, S.; Friedrich, M.W.; Sc hmidt, H., Evaluation of Strategies tp Seperate Root - associated Microbial Communities: A cruical choice in rhizobiome research. Frontier in Microbiology, 2016. 7 . 37. Wood, S.A.; Smith, M.R.; Franzo, J.; Remans, R.; DeFries, R., Trade and the equitability of global food nutrient distribution. Nature Sustainability, 2018. 1 (1): p. 34 - 37. 38. Lensen, M.; Moran, D.; Kanemoto, K.; Foran, B.; Lobefaro, L.; Geschke, A., International trade drives biodiversity threats in developing nations. Nature 2012. 486 (7401): p. 109 - 112. 39. Dalin, C.; Wada, Y.; Kastner, T.; Puma, M.J., Groundwater depletion embedded in international food trade. Nature, 2017. 543 (7647): p. 700 - 704. 41 40. Chaudhary, A.; Katsner, T., Land use biodiversity impacts embodied in international food trade. Global Environmental Change, 2016. 38 p. 195 - 204. 41. Moran, D.; Kanemoto, K., Identifying species threat hotspots from global supply chains. Nature Ecology & Evoloution, 2017. 1 (23). 42. MacDonald, G.K.; Brauman, K.A.; Sun, S.; Carlson, K.M.; Cassidy, E.S.; Gerber, J.S.; West, P.C., Rethinking agricultural trade relationships in an era of globalization. BiosSciences, 2015. 65 (3): p. 275 - 289. 43. Crist, E.; Moran, C.; Engelman, R., The inte raction of human population, food production, and biodiversity protection. Science, 2017. 356 (6335): p. 260 - 264. 44. DeFries, R.; Rudel, T.K.; Uriarte, M.; Hansen, M., Deforestation driven by urban population growth and agricultural trade in the twenty - fir st century. Nature Geoscience, 2010. 3 (3): p. 178 - 181. 45. Lambin, E.F.; Meyfroidt, P., Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences 2011. 108 (9): p. 3465 - 3472. 46. Foley, J. A.; Ramankutty, K.A.; Brauman, K.A.; Cassidy, E.S.; Gerber, J.S.; Johnston, M.; S.R.; Hill, J.; Monfreda, C.; Polasky, S.; Rockstrom, J.; Sheehan, J.; Siebert, S. ; Tilman, D.; Zaks, D.P.M., Solutions for a cultivated planet. Nature, 2011. 478 (7369): p. 337 - 342. 47. Liu, J.; Hull, V.; Batistella, M.; DeFries, R.; Dietz, T.; Fu, F.; Hertel, T.W.; Izaurralde, E.F.; Lambin, S.L.; Li, S.; Martinelli, L.A.; McConnel, W.J .; Moran, E.F.; Naylor, R.; Ouyang, Z.; Polenske, K.R.; Reenberg, A.; de Miranda Rocha, G.; Simmons, C.S.; Verburg, P.H.; Vitousek, P.M.; Zhang, F.; Zhu, C., Framing sustainability in a telecoupled world. Ecology and Society 2013. 18 (2). 48. Sun, J.; Moone y, H.; Wu, W.; Tang, H.; Tong, Y.; Xu, Z.; Huang, B.; Cheng, Y.; Yang, X.; Wei, D.; Zhang, F.; Liu, J., Importing food damages domestic environment: Evidence from global soybean trade. PNAS, 2018. 115 (21): p. 5415 - 5419. 49. Martinelli, A.L.; Batistella, M. ; Silva, F.R.; Moran, E., Soy Expansion and Socioeconomic Development in Municipalities of Brazil. Land, 2017. 6 (3). 50. Schaffer - Smith, D.; Tomscha, S.A.; Jarvis, K.J.; Maguire, D.Y.; Treglia, M.L.; Liu, J., Network analysis as a tool for quantifying the dynamics of metacoupled systems: an example using global soybean trade. Ecology and Society, 2018. 23 (4). 51. Sun, J.; Wu, W.; Tang, H.; Liu, J., Spatiotemporal patterns of non - genetically modified crops in the era of expansion of genetically modified food. Scientific Reports, 2015. 5 : p. 14180. 42 52. Gasparri, N.I.; Kuemmerle, T.; Meyfroidt, P.; de Waroux, Y.l.P.; Kreft, H., The Emerging Soybean Production Frontier in Southern Africa: Conservation Challenges and the Role of South - South Telecoupl ings. Conservation Letters, 2016. 9 (1): p. 21 - 31. 53. Bo, X.; Yi, F.; Li, Y., livestock sector mean for its vegetable oil market? China Agricultural Economic Review, 2017. 9 (2): p. 303 - 316. 54 . UN International Statistics Division; International Trade Database, UNcomtrade , United Nations, Editor. 2018: New York. 55. Dou, Y.; Bicudo da Silva, R.F.; Yang, H.; Liu, J., Spillover effect offsets the conservation effort in the Amazon. Journal of Geog raphical Sciences 2018. 28 (11): p. 1715 - 1732. 56. Li, A.S., Argentina and China: the Soybean Trade and its Consequences. Chicago Unbound, 2016. Law School: International Immersion Program Papers (11). 57. Peine, E.K., Trading on Pork and Beans: Agribusiness and the Construction of the Brazil - China - Soy - Pork Commodity Complex , in The Ethics and Economics of Agrifood Competition , J.H.S. James, Editor. 2013, Springer Netherlands: Dordrecht. p. 193 - 210. 58. Jordan, G.; Navin, R.; Oliver, T.C., Increasing expansio n of large - scale crop production onto deforested land in sub - Andean South America. Environmental Research Letters, 2018. 13 (8): p. 084021. 59. Reenberg, A.; Fenger, N.A., Globalizing land use transitions: the soybean acceleration. Geografisk Tidsskrift - Dan ish Journal of Geography, 2011. 111 (1): p. 85 - 92. 60. FAS, , in Oilseeds: World Markets and Trade . 2017, USDA Forgein Agriculutral Service. 61. FAS, EU - 27 Soybean Imports from the United States Still Impeded , in Oilseeds and Products . 2009, USDA Forgein Agricultural Service. 62. Westcott, P.; Trostle, R., USDA Agriculutral Projections to 2023 . 2014, USDA Economic Research Service,. 63. Cargill. Growing a new soy industry in Uruguay . 2018 12/15/18]; Available from: http://150.cargill.com/150/en/URUGUAYAN - SOY.jsp . 64. Dutch Soy Coalition, Soy in Paraguay . 2009. 65. Liu, J., Integration across a metacoupled world. Ecology and Society, 2017. 22 (4). 66. Felter, C.; Renwick, D. . 2017 01/02/2019]; Available from: https://www.cfr.org/backgrounder/mercosur - south - america s - fractious - trade - bloc . 43 67. Reeder, J.; Torene, J.; Jabara, C.; Babula, R., REGIONAL TRADE AGREEMENTS: EFFECTS OF THE ANDEAN AND MERCOSUR PACTS ON THE VENEZUELAN SOYBEAN TRADE AND U.S. EXPORTS . 2005, U.S. International Trade Commission. 68. silva, F.R.; Ba tistella, M.; Dou, Y.; Moran, E.; Torres, S.M.; Liu, J., The Sino - Brazilian Telecoupled Soybean System and Cascading Effects for the Exporting Country. Land, 2017. 6 (53). 69. Williman, K., Soybean trade telecouplings and land use change in Uruguay: connecting the provincial to the global , in Stockholm Resilience Centre . 2015, Stockholm University. p. 50. 70. Wallace, J.M.; Gutzler, D.S., Teleconnections in the Geopotential Height Field during the Northern Hemisphere Winter. Americam Meteorological So ciety, 1981. 109 . 71. Sassen, S., Globalization and its Discontents: Essays on the New Mobility of People and Money . 1999, New York, NY: The New Press. 72. Liu, J.; Herzberger, A.; Kapsar, K.; Carlson, A.; Connor, T., What is Telecoupling? , in Telecoupling : Exploring Land - Use Change in a Globalised World (Palgrave Studies in Natural Resource Management) , C. Friis; J.Ø. Nielsen, Editors. 2019, Palgrave Macmillan. p. 19 - 48. 73. Levin, S.A., Ecosystems and the biosphere as complex adaptive systems. Ecosystems, 1998. 1 (5): p. 431 - 436. 74. Holland, J.H., Complex Adaptive Systems. Daedalus, 1992. 121 (1): p. 17 - 30. 75. International Relations Online. New Trade Theory . 2014. 76. Peter, D.; Silva, P. ARIMA vs. ARIMAX which approach is better to analyze and forecast macroeconomic time series? in Proceedings of 30th International Conference Mathematical Methods in Economics . 2012. Karviná, Czech Republic Silesian University in Opava, School of Business Administration . 77. Hyndman, R.J.; Athanasopoulos, G., Forecasting: principles and practice . 2014: OTexts. 78. Williams, B.M.; Hoel, L.A., Modeling and Forecasting Vehicular Traffic Flow as a Seasonal ARIMA Process: Theoretical Basis and Empirical Results. Journal of Transpor tation Engineering, 2003. 79. Trapletti, A.; Hornik, K., tseries: Time Series Analysis and Computational Finance. 2018(R package version 0.10 - 46). 80. Wickham, H., ggplot2: Elegant Graphics for Data Analysis . 2016, Springer - Verlag New York. 81. Lin, Y., Po st - crisis China impact on trade integration and manufacturing competitiveness between Argentina and Brazil. Journal of Chinese Economic and Business Studies, 2018. 16 (2): p. 147 - 170. 44 82. Deese, W.; Reeder, J., Export taxes on agricultural products: recent history and economic modeling of soybean export taxes in Argentina , U.S.I.T. Commission, Editor. 2007, Journal of International Commerce and Economics. 83. de Ridder, M.; ten Haaf, P.; Olah, N.; Bolscher, H., Soy supply security for the Netherlands , in Geo politics , C.o.E.o. Resources, Editor. 2015, The Hague Centre for Strategic Studies; the Netherlands Organization for Applied Scientifc Research; Trinomics. 84. Wright, B., Global Biofuels: Key to the Puzzle of Grain Market Behavior. Journal of Economic Per spectives, 2014. 28 (1): p. 73 - 98. 85. Villoria, N.B.H., Thomas W., Geography Matters: International Trade Patterns and the Indirect Land Use Effects of Biofuels. American Journal of Agricultural Economics, 2011. 93 (4): p. 919 - 935. 86. Babcock, B.A.C., M. A ., Prospects for Corn Ethanol in Argentina. CARD Staff Report Series, 2012(SR 107). 87. Vieira, L., in Successful Farming . 2018. 88. Liu, J.; Mooney, H.; Hull, V.; Davis, S.J.; Gaskell, J.; Hertel, T.; Lubchenco, J.; Seto, K.C.; Gleick, P.; Kremen, C.; Li, S., Systems integration for global sustainability. Science, 2015. 347 (6255). 89. Bronstein, H.; Desantis, D., INTERVIEW - Paraguay soy beans flow to China regardless of politics , in Reuters . 2018. 90. Solidaridad, Promoting Sustainability amoung soy family farmers in Paraguay . 2016, the sustainable trade initiative. 91. Bronstein, H.; Desantis, D., Paraguay soybean exports to top Argentin a's for first time , in Reuters . 2018. 92. Brazier, J., , in South America Monthly Insight . 2017, A2 Global. 93. Nickson, R.A. Brazil and Paraguay: A Protectorate in the making? in Brazil and Spanish America: Relationships and Comparisons . 2019. Insititue of Latin America Studies, University of London: bepress. 94. Lagarde, M., routine longitudinal data. Health Policy and Planning, 2011. 27 (1): p. 76 - 83. 95. Carlson, A.; Zaehringer, J.; Garrett, R.; Felipe Bicudo Silva, R.; Furumo, P.; Raya Rey, A.; Torres, A.; Gon Chung, M.; Li, Y.; Liu, J., Toward Rigorous Telecoupling Causal Attribution: A Systematic Review and Typology. Sustainability, 2 018. 10 (12): p. 4426. 96. Boulding, K., The Economics of Peace . Vol. 278. 1945, New York: Prentice - Hall Inc. 45 INTERNATIONAL FOOD TRADE LED TO AIR POLLUTION IN NON - TRADING COUNTRY Overview Almost all places around the world are interconnected via food trade [ 1 ] . Over the past three decades, global food production has do ubled [ 2 ] , with producing countries increasingly cultivating for exports rather than domestic consumption, resulting in a rise in food export s by an order of magnitude [ 3 ] . The increased g lobal trade of agriculture products have numerous, well - documented socioeconomic and environmental impacts affecting both exporting and importing regions [ 4 ] . However, little is known about the spillover effects that influence other regions outside the trading partners [ 5 - 7 ] . To explore spillover effects of international trade, we choose international soybean trade as an example because it has been rapidly growing in recent decades. Several studies have analyzed the impact of soybean trade with China, the top soybean consumer, on Brazil, the top soybean producer. In addition to pros perous economic growth, soybean expansion in Brazil competes with production of other crops and land for conservation efforts [ 8, 9 ] . Further, changes in crop production and trade relationships with countries bordering Brazil indicate soybean expansion has displaced other crops production to neigh boring countries [ 10 ] . On the other side of the world, increased soybean imports have prompted many farmers in China, particularly in Heilongjiang province (the most important soybean producing region in the coun try), to switch soybeans to alternative crops such as maize. In this study, we explore the underlying regional cultivation and management shifts that resulted in air pollution locally and across the China - Russia border. We quantified the spatial and tempor al patterns of PM 2.5 in Heilongjiang and in the nearby Russia Provinces. To confirm air pollutant transport, we used NASA Worldview images to capture smoke from the fires. Finally, 46 we analyzed the spatiotemporal patterns of agricultural fires and crop con version patterns and evaluated the household mechanisms behind the patterns. Results In this study, through a combination of remotely sensed data and household surveys, we found that air quality in Russia is unexpectedly affected by soybean trade between China and major soybean exporters such as Brazil and the USA. Specifically, the concentration of atmospheric fine particulate matter smaller than 2.5 microns in diameter (PM 2.5 ) [ 11 ] in several provinces of Russia - May), and with a consp icuous peak in late autumn (i.e., November; Fig. 1 3 ). During spring the P M 2.5 concentrations are mostly associated with household heating (with coal) over the winter months [ 12, 13 ] . The effects of pre - planting crop residue burning (Fig. 14 ) on PM 2.5 concentration in neighboring Russian provinces during spring seem to be lower than in November because they increased PM 2.5 concentration in Heilongjiang province but not in neighborin g Russian provinces (Fig. 1 3 ). In contrast, the conspicuous increase in PM 2.5 in November across the entire region (including areas in Russia) is mostly associated with post - harvest residue burning in Heilongjiang (Figs. 1 3 and 14 ). This seasonal differen ce in the effects of crop residue burning may be explained by thermal inversions characteristic of autumn months, impeding the vertical movement of air (convection) thus reducing the dilution of PM 2.5 pollutants [ 11 ] . Using images from NASA Worldview [ 14 ] (Fig. 15 ), we observed that sm oke from agricultural fires in Heilongjiang was blowing into neighboring Russian provinces. 47 Figure 13 . Temporal dynamics of AOT aggregated across Heilongjiang province, China and across neighboring provinces in Russia (Amur, Khabarovsk, Primorsky and Yevrey). Panel A shows multi - annual (2006 - 2016) monthly average AOT values and Panel B shows the average AOT values for the month of November of each year from 2006 to 2016. Figure 14 . Average sea sonal dynamics of the areas where fires were detected in Heilongjiang province, Northeast China during the 2006 - 2016 period. Error bars correspond to 1 SEM. 48 Figure 15 . Location of the study site, Heilongjiang, China (in yellow) and neighboring Russia Provinces (in grey) with inset snapshots of the China - Russia border from NASA worldview [14] . The inset image snapshot is from 10/14/16 and captures smoke from agricultural fires blowing across the international border. Panel A shows 1:5000, Panel B shows 1:2500 and P anel C shows 1:1500. The red dots on the inset image map are satellite detections of fires, derived from MODIS and VIIRS instruments, in the area. https://go.nasa.gov/30UQllB . The liberalization of soybean importation in China has pushed farmers, particularly in Heilongjiang province (Fig. 15 ; the most important soybean producing area in the country), to switch soybeans to alternative crops such as maize [ 15 ] . During the household surveys, nearly 50% of the farmers reported that international soybean imports affect their decisions compared to the 24% which reported no impact (Table 3 ). The majority of farmers were against international soybean trade often citing price competition as support while 31% supported soybean trade due to 49 the availability of cheaper soy - based animal feed. In follow - up questions regarding the decision to abandon soybean production, 50% of the farmers mentioned decreasing soybean price as a reason. Figure 16 . Spatio - temporal dynamics of agricultural land use (maize, soybean and mixed maize/soybean in Heilongjiang province, Northeast China. (A) Spatial distribution of agricultural land use classes during 2006. (B) Spatial distribution of agric ultural land use classes during 2016. (C) Temporal dynamics of agricultural land use classes between 2006 and 2016. Table 3 . Farmer Survey Responses 1. Are you aware that most of the soybeans in China are imported? 57% Yes 43% N o 2. If yes, do you know which countries soybeans are imported from? 43% U.S. >1% Brazil 3. Do imported soybeans affect your decisions? 24% No 49% Yes 4. If it affects your decision, does it make you anxious? 15% No 22%A little 31% A lot 32% Don't Know 5. What is your attitude towards soybean trade? 60% Against 31% Support Concurrent with the reduction in areas under soybean and increases in areas under maize, there is a concomitant increase in the areal extent of fires in Heilongjiang (Fig. 16 ). While fires are not restricted to agricultural areas, the proportion of fires o ccurring in agricultural areas increased from 60% in 2006 to 77% in 2016 (Fig. 1 7 ). Further, the area exposed to fire and the area under maize cultivation between 2006 and 2016 are significantly positively correlated ( r = 50 0.68, p = 0.02) and the area expos ed to fire and the area under soybean cultivation are significantly negatively correlated ( r = - 0.59, p = 0.05). The increase in agricultural fires is explained by not only the increase in residual crop biomass, which is more than 4 times higher for maize than for soybeans [ 16 ] but also, the greater tendenc y to burn maize residue compared to soybean residue. For example, based on the average crop yields obtained during the farm interviews, soybeans produce 2,401 kg/ha of residual straw biomass, while maize produces 11,164 kg/ha of residual straw biomass [ 17 ] . Further, during the househo ld surveys farmers revealed that they most often use soybean residue for household fuel, animal fodder or returning it to the soil (i.e., 92%) than burning it (i.e., 8%) (Table 4 ). Whereas over one - third of the farmers burned maize residue in the field whi ch explains why the fires in Heilongjiang are concentrated seasonally in two periods corresponding to pre - planting (around day of the year 100) and post - harvesting (around day of the year 300) (Fig. 2). Due to a residue burn ban and monitoring agricultural fires by the local government [ 18 ] , the percent of farmers burning their maize residue declined from 44% to 38% between 2005 and 2015 (Table 4 ). Despite this reduction, the area planted with maize increased by 83% (more than 30,000 km 2 ) between 2006 and 2016 resulting in an increase in residual crop biomass. Cultivating maize (instead of soybeans) was found to be the most significant and influential factor affe 5 & 6 ). 51 Figure 17 . Spatio - temporal dynamics of fires in Heilongjiang province, Northeast China. (A) Spatial distribution of fires during 2006. (B) Spatial distribution of fi res during 2016. (C) Temporal dynamics of the area exposed to fire (combining low, nominal and high confidence fires) between 2006 and 2016. Table 4 . Crop residue management in Heilongjiang, China. Return Burn Household Fuel Livest ock Fodder Maize 2015 31% 38% 27% 4% 2005 12% 44% 37% 7% Soybeans 2015 28% 8% 62% 2% 2005 26% 8% 67% 0% Discussion Managing crop residue with fire is an emerging issue of concern, posing environmental and human health hazards not only locally, but also at regionally and international scales [ 19 ] . The particles released during the inef ficient combustion of crop biomass can travel from the area of production to regions unassociated with the offending activity [ 20, 21 ] , which makes identifying the underlying land - use drivers of air pollution is difficult [ 22 ] . In the present study, we combine remotely sensed data an d farmer interviews to understand how international soybean trade increased crop residue fires in Heilongjiang and negatively affected air quality across the Russian border. 52 First, levels of PM 2.5 were elevated over Heilongjiang, China and the neighborin g Russian provinces, seasonally, corresponding to pre - plant and post - harvest activities. Seasonal spikes were also observed in the area exposed to active fires over Heilongjiang and both the levels of PM 2.5 and area exposed to active fires have increased over the past decade. Similar findings have been reported in the literature, where Northeastern China in general and Heilongjiang in particular, have been identified as one of the top contributing - regions to air pollution due to residue fires [ 23, 24 ] . Many studies have found patterns in emissions associated with crop residue fires in this region [ 23 - 26 ] and have documented an exponential increase in residue burning compared to other provinces [ 27 - 29 ] . Additionally, studies have documented air pollution spi llovers from crop residue fires that reduce air quality and visibility in nearby urban areas. These fires and subsequent air pollution are seasonal, in the fall, after harvest, and the spring, before planting [ 21, 30, 31 ] . Second, the pattern of active fires in Heilongjiang is sign ificantly and positively correlated with the spatial and temporal increase in maize cultivation and significantly and negatively correlated with the decline of soybean production. Farmers in Heilongjiang more often use fire to management maize residue than soybean residue and are increasing the area planted to maize and reducing the area planted to soybeans. From the household survey, we also know that many farmers are choosing to covert rice production, the residue of which is most often burned. The increa se in rice cultivation in the region, while not analyzed in this study, is likely also contributing to the levels of PM 2.5 . Additionally, beyond the pollutants linked residue fires, emissions of ammonia from nitrogen fertilizer contribute to PM 2.5 levels [ 32 ] . Both maize and rice require nitrogen application whereas soybeans do not and previous studies have documented the increase used of nitrogen alongside the land - use change in Heilongjiang [ 33, 34 ] . In fact, the nutritional value of legume residue is a driver for leaving soybean residue on the field compared to the more 53 common practice of removing maize and rice residu e [ 35, 36 ] . To manage the accumulating maize residue sustainable, farmer need access to crushing or mulching technology to ensure the straw breaks down on the field overwinter. Other uses for the residue include biochar or biomass gasification, however, this maybe limited by transportation costs to the nearest plant. Last, interviews with farmers confirmed that the reduction in soybean production is in response to low prices driven by international trade competition. The decline of soybean production in China is the result of several domestic policies that promoted the import of soybeans to support the domestic pork industry [ 37 ] . The dynamics observed in this study are illustrative of unexpected spillover effects that can arise from the international trade of an agricultur al commodity. Because international trade will continue connecting producers and consumers around the globe, it is crucial that future studies consider the impact beyond trade partners [ 7, 38 ] . Similarly, in order to develop effective policy for a more sustainable future, the impact of international trade on the domestic environment as well as the impacts that spillover to other countries must be considered [ 39, 40 ] . Materials and Methods 3.4.1. Study Si te To explore the patterns and mechanisms behind crop residue fires we chose Heilongjiang Province in Northeast China as our study area (Fig. 1). Heilongjiang is the top grain producing - area in China, producing one - quarter to one - n over the past decade [ 41, 42 ] . The farms in Heilongjiang average around 3.5 ha and typically lack mechanization. The growing season lasts from early May to mid - August [ 42 ] . While winters are harsh, the area has a humid continental climate with high temperatures and rainfall in the summer months. Previous analyses conduc ted by members of our research team suggested the total amount of cropland 54 planted to soybeans decreased by 24% from 43,722 km 2 in 2005 to 35,384 km 2 in 2010 [ 42 ] and indicated that soy - planting proportion is decreasing while maize planting proportion in increasing as a response to competition from international soybean imports [ 17, 33, 34, 42 ] . 3.4.2. Land use change analysis To classify agricultural land use dynamics in Heilongjiang Provin ce we used a bi - weekly composite image time series (250 m/pixel) of the Normalized Difference Vegetation Index derived from surface reflectance data acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) on board the National Aeronautics and Space Administration (NASA) Terra satellite, between January 2006 and December 2016 (MOD13Q1 product). The time series imagery for the year 2016, together with the geographic location of 76 soybean and 99 maize fields obtained during the summer of 2016, were input into MaxENT, a general - purpose algorithm that generates fuzzy inferences from an incomplete set of information. The field data were split into calibration (2/3) and validation (1/3) datasets, but to reduce the dependence on a single partition in to calibration and validation datasets, we performed five different partitions. Through this algorithm we obtained continuous maps of the probability distribution of agricultural fields under soybean, maize and mixed soybean/maize (with an average area und er the receiver operating characteristics (ROC) curve of 0.91). To change from continuous maps to a single choropleth output map , we used threshold values of 0.48 for maize and 0.43 for soybean. Values above these thresholds indicate that these crops are p soybean are present. Model coefficients were then applied to the MODIS - NDVI image time series of 2006 through 2015 to obtain annual maps of agricultural land use. 3.4.3. Household surveys 55 To g ain insight on the mechanism behind crop residue fires, household surveys were conducted during the summer of 2016 in Heilongjiang Province, China. The questionnaire was designed to acquire information on farmer demographics, land use decisions and the inf luence of international soybean trade on cultivation and management practices (See Tables 3 and 4 ). The surveys were conducted face - to - face with 1,190 household heads in 42 villages. Farming in Heilongjiang is dominated by households, so the household surv ey is the main method to collect relevant management information, such as residue management practices across the province. 3.4.4. Fire To assess spatio - temporal fire dynamics over the 2006 - 2016 period, we used the Terra - MODIS eight - day summary fire product (MOD 14A2). This product is a 1 km/pixel composite of fire pixels detected in each grid cell over each 8 - day compositing period, including measures of confidence (i.e., low, nominal, high). Fire occurrence was integrated annually, to obtain estimates of the ar eas showing evidence of fire along each year. Images of smoke from fires in Heilongjiang on 10/29/16 were captured using NASA worldview snapshot [ 14 ] . These photos provide evidence that smoke from the fires reject injection heights that allowed particulate transfer across the international border. 3.4.5. Aerosol Optical Thickness To assess spatial - temporal dynamics in aerosol optical thickness (AOT, a surrogate of atmospheric fine particulate matter, PM 2.5 [ 43 ] ) in Heilongjiang Province and in neighboring provinces in Russia, we used the Terra - MODIS aerosol product (MOD04L2). This product provides information on the atmospheric aerosol loading in 1 - km pixels, composited over a mont h. A value of 0.01 corresponds to a clean atmosphere, while values larger than 0.4 correspond to hazy atmospheric conditions. 56 3.4.6. Statistical Analysis D ata processing and analysis was done in the R environment using packages lme4 [44] , ggplot2 [45] , reshape2 [46] . To determine the factors that significantly influenced residue management, an ordinary least squares regression model was constructed and standards errors were clustered (1) to account for autocorrelation and heteroskedasticity that exists at the indivi dual level [47] . Data processing and analysis were done in the R environment using packages plm [48] , lmtest [49] , car [50] , reshape [46] , ggplot2 [45] and MASS [51] . We cons idered the decision to burn farmers reported what percent of their residue was returned to the soil, used on the farm for fuel or fodder and burned. In trying to determine the factors that led to increased residue burning, the data were simplified by combining the farmers that returned their residue to the soil with the farmers who used their residue on the farm for fuel or fodder. The regression analysis was conducted with 251 farmers (i) in 38 villages (v). + (1) Where, P is the probability that Y = 1 for farmer i in village v and Y = residue burning (1=yes, 0=no); are explanatory variables; and U is the error term. 57 APPENDIX 58 In addition to cultivating soybeans, larger planted areas and family size had a significant, suppressing impact on burning crop residue while increases in nitrogen use and time spent off the farm had a significant positive relationship with crop residue bu rning. Farmers with larger planted areas are likely more financially established which may have allowed them allocate more time, access to additional rented land and invest in labor saving - machinery. Further, having larger families would provide additional labor capital that is needed to use or return their crop residue. This is in contrast with farmers who spend much of their time in urban areas and supplement their income with off - farm employment. Needing to migrate for work after planting crops in the sp ring management strategy. Further, farmers in Heilongjiang are constrained by a lack of mechanization. Some villages communally owned a single tractor to be s hared amongst the farmers, in other cases farmers would hire someone else for planting, spraying and harvest activities, but often the farmer would perform these tasks by hand. The lack of mechanization consumes time and results in crop residue that is int act (i.e., whole maize stalks). The large pieces of crop residue do not break down during the harsh winters and it is more feasible for the farmer to collect and burn the maize stalks rather than chop them fine enough to be returned to the soil. 59 Table 5 . Logistic regression analysis with clustered SE at the village level. Includes factor names, coefficients, p values and VIF Variable Estimate SE p - value VIF (Intercept) 5.39 20.33 0.79 Crop (Soybean) - 0.31 0.06 0.00*** 1.4 Planted Area - 0.01 0.00 0.00** 1 Nitrogen Use 0.00 0.00 0.00*** 1.3 Family Size - 0.08 0.02 0.00*** 1 Time off - farm 0.02 0.01 0.05* 1 Farmer Edu. - 0.02 0.02 0.34 1 Farmer Age 0.00 0.00 0.08 1 Electricity Use 0.00 0.00 0.62 1 Coal Use 0.01 0.02 0.82 1 Maize Cob Use 0.00 0.00 0.63 1 Gas Use 0.00 0.01 0.79 1 Year 0.00 0.01 0.82 1 R2 = 0.35 and adjusted R2 = 0.31. Coefficient and p value (***p < 0.001 and *p < 0.01) indicated that crop type, planted area, nitrogen use, family size and time spent off the farm are significant and exhibited strong influences on residue management, further farmer age had a p - value of 0.08 and was still considered to be an influential factor. To determine the factors that significantly influenced residue managem ent, an ordinary least squares regression model was constructed and standards errors were clustered (1) to account for autocorrelation and heteroskedasticity that exists at the individual level [ 47 ] . Data processing and analysis were done in th e R environment using packages plm [ 48 ] , lmtest [ 49 ] , car [ 50 ] , reshape [ 46 ] , ggplot2 [ 45 ] and MASS [ 51 ] . We considered the decision to burn crop residue as a dichotomous dependent varia residue was not burned in the field. During the household interview farmers reported what percent of their residue was returned to the soil, used on the farm for fuel or fodd er and burned. In trying to determine the factors that led to increased residue burning, the data were simplified by combining the farmers that returned their residue to the soil with the farmers who used their residue 60 on the farm for fuel or fodder. The r egression analysis was conducted with 251 farmers (i) in 38 villages (v). + (1) Where, P is the probability that Y = 1 for farmer i in village v and Y = residue burning (1=yes, 0=no); are explanatory variables; and U is the error term . Table 6 . Factor description. Variable Description Mean SD Y Residue (Burn = 1, Use = 0) 0.26 0.44 X1 Crop Type (Soybean = 112, Maize = 138) X2 Planted Area (ha) 6.4 8.4 X3 Nitrogen Use (kg/ha) 160 120 X4 Family Size 3.4 1 X5 Off - farm employment (months) 0.96 2.6 X6 Farmer Education (1=illiterate, 2=elementary, 3=junior high, 4=high school) 2.3 1.1 X7 Farmer Age 49.5 10.7 X8 Electricity (degrees) 1692 1525 X9 Coal Use (MT) 2 1 X10 Maize Cob Use (truckloads) 14 28.4 X11 Gas Use (tanks) 1 3.6 X12 Year (2010=80, 2015=170) 61 REFERENCE S 62 REFERENCES 1. Lambin, E.F.; Meyfroidt, P., Global land use change, economic globalization, and the looming land scarcity. PNAS, 2011. 108 (9): p. 3465 - 3472. 2. FAO, FAOSTAT data . 2016, Rome: FAO ( http://www.fao.org/faostat/en/#home ). 3. UNcomtrade, United Nations Commodity Trade Statistics Database . 2010: United Nations Statistics Division ( https://comtrade.un.org/db/ ). 4. Schmitz, C.; Biewald, A.; Lotze - Campen, H.; Popp, A.; Dietrich, J.P.; Bodirs ky, B.; Krause, M.; Weindl, I., Trading more food: Implications for land use, greenhouse gas emissions, and the food system. Global Environmental Change, 2012. 22 (1): p. 189 - 209. 5. Lin, J.; Pan, D.; Davis, S.J.; Zhang, Q.; He, K.; Wang, C.; Streets, D.G.; Wuebbles, D.J.; Guan, D., Proceedings of the National Academy of Sciences, 2014. 111 (5): p. 1736 - 1741. 6. Zhao, H.; Zhang, Q.; Guan, D.; Davis, S.; Liu, Z.; Huo, H.; Lin, J.; Liu, W.; He, K., Assessment of China's virtual air pollution transport embodied in trade by using a consumption - based emission inventory. Atmospheric Chemistry and Physics, 2015. 15 (10): p. 5443 - 5456. 7. Liu, J.; Dou, Y.; Batistella, M.; Challies, E.; Connor, T.; Frii s, C.; Millington, J.D.A.; Parish, E.; Romulo, C.L.; Silva, R.F.B.; Triezenberg, H.; Yang, H.; Zhao, Z.; Zimmerer, K.S.; Huettmann, F.; Treglia, M.L.; Basher, Z.; Chung, M.G.; Herzberger, A.; Lenschow, A.; Mechiche - Alami, A.; Newig, J.; Roche, J.; Sun, J., Spillover systems in a telecoupled Anthropocene: typology, methods, and governance for global sustainability. Current Opinion in Environmental Sustainability, 2018. 33 : p. 58 - 69. 8. Bicudo da Silva, R.F.; Batistella, M.; Dou, Y.; Moran, E.F.; Torres, S.M. ; Liu, J., The Sino - Brazilian Telecoupled Soybean System and Cascading Effects for the Exporting Country. Land 2017. 6 (53): p. 1 - 19. 9. Dou, Y.; Bicudo da Silva, R.F.; Yang, H.; Liu, J., Spillover effect offsets the conservation effort in the Amazon. Journ al of Geographical Sciences 2018. 28 (11): p. 1715 - 1732. 10. Herzberger, A.; Chung, M.G.; Kapsar, K.; Frank, K.A.; Liu, J., Telecoupled Food Trade Affects Pericoupled Trade and Intracoupled Production. Sustainability, 2019. 11 (10): p. 2908. 11. Yin, S.; W ang, X.; Xiao, Y.; Tani, H.; Zhong, G.; Sun, Z., Study on spatial distribution of crop residue burning and PM 2.5 change in China. Environmental Pollution, 2017. 220 : p. 204 - 221. 63 12. Zhang, J.; Smith, K.R., Household Air Pollution from Coal and Biomass Fuels in China: Measurements, Health Impacts, and Interventions. Environmental Health Perspectives, 2007. 115 (6): p. 848 - 855. 13. Xing, X.; Zhou, Y.; Lang, J.; Chen, D.; Cheng, S.; Han, L.; Huang, D.; Zhang, Y., Spatiotemporal variation of domesti c biomass burning emissions in rural China based on a new estimation of fuel consumption. Science of The Total Environment, 2018. 626 : p. 274 - 286. 14. NASA, Earth Observing System Data and Information System (EOSDIS) . 2020. 15. Oliveira, G.d.L.; Schneide r, M., The politics of flexing soybeans: China, Brazil and global agroindustrial restructuring. The Journal of Peasant Studies, 2016. 43 (1): p. 167 - 194. 16. Jiang, D.; Zhuang, D.; Fu, J.; Huang, Y.; Wen, K., Bioenergy potential from crop residues in China : Availability and distribution. Renewable and sustainable energy reviews, 2012. 16 (3): p. 1377 - 1382. 17. Tong, Y.; Liu, J.; Li, X.; Sun, J.; Herzberger, A.; Wei, D.; Zhang, W.; Dou, Z.; Zhang, F., Cropping System Conversion led to Organic Carbon Change i Scientific Reports, 2017. 7 (1): p. 18064. 18. Fanchao, D., The Department of Environment: the air pollution in 11 northeastern cities are off the charts, among which Haerbin, Daqing, and Suihua have the worst pollution. 2016. 19. Shi, T.; Liu, Y.; Zhang, L.; Hao, L.; Gao, Z., Burning in agricultural landscapes: an emerging natural and human issue in China. Landscape ecology, 2014. 29 (10): p. 1785 - 1798. 20. Campbell - Lendrum, D.; Prüss - Ustün, A., Climate change, air pollution a nd noncommunicable diseases. Bulletin of the World Health Organization, 2019. 97 (2): p. 160 - 161. 21. Bikkina, S.; Andersson, A.; Kirillova, E.N.; Holmstrand, H.; Tiwari, S.; Srivastava, A.K.; Bisht, D.S.; Gustafsson, Ö., Air quality in megacity Delhi affe cted by countryside biomass burning. Nature Sustainability, 2019. 2 (3): p. 200 - 205. 22. Liu, J.; Hull, V.; Batistella, M.; DeFries, R.; Dietz, T.; Fu, F.; Hertel, T.W.; Izaurralde, E.F.; Lambin, S.L.; Li, S.; Martinelli, L.A.; McConnel, W.J.; Moran, E.F.; Naylor, R.; Ouyang, Z.; Polenske, K.R.; Reenberg, A.; de Miranda Rocha, G.; Simmons, C.S.; Verburg, P.H.; Vitousek, P.M.; Zhang, F.; Zhu, C., Framing sustainability in a telecoupled world. Ecology and Society 2013. 18 (2). 23. Sun, J.; Peng, H.; Chen, J.; Wang, X.; Wei, M.; Li, W.; Yang, L.; Zhang, Q.; Wang, W.; Mellouki, A., An estimation of CO2 emission via agricultural crop residue open field burning in China from 1996 to 2013. Journal of Cleaner Production, 2016. 112 : p. 262 5 - 2631. 24. Mehmood, K.; Chang, S.; Yu, S.; Wang, L.; Li, P.; Li, Z.; Liu, W.; Rosenfeld, D.; Seinfeld, J.H., Spatial and temporal distributions of air pollutant emissions from open crop straw and biomass burnings in China from 2002 to 2016. Environmental Chemistry Letters, 2018. 16 (1): p. 301 - 309. 64 25. Li, J.; Bo, Y.; Xie, S., Estimating emissions from crop residue open burning in China based on statistics and MODIS fire products. Journal of Environmental Sciences, 2016. 44 : p. 158 - 170. 26. Li, C.; Hu, Y .; Zhang, F.; Chen, J.; Ma, Z.; Ye, X.; Yang, X.; Wang, L.; Tang, X.; Zhang, R.; Mu, M.; Wang, G.; Kan, H.; Wang, X.; Mellouki, A., Multi - pollutant emissions from the burning of major agricultural residues in China and the related health - economic effects. Atmosphereic Chemistry and Physics, 2017. 17 : p. 4957 - 4988. 27. Xu, Y.; Huang, Z.; Jia, G.; Fan, M.; Cheng, L.; Chen, L.; Shao, M.; Zheng, J., Regional discrepancies in spatiotemporal variations and driving forces of open crop residue burning emissions in China. Science of The Total Environment, 2019. 671 : p. 536 - 547. 28. Li, J.; Li, Y.; Bo, Y.; Xie, S., High - resolution historical emission inventories of crop residue burning in fields in China for the period 1990 2013. Atmospheric Environment, 2016. 138 : p. 152 - 161. 29. Zhou, Y.; Xing, X.; Lang, J.; Chen, D.; Cheng, S.; Wei, L.; Wei, X.; Liu, C., A comprehensive biomass burning emission inventory with high spatial and temporal resolution in China. Atmos. Chem. Phys., 2 017. 17 (4): p. 2839 - 2864. 30. Kharol, S.K.; Badarinath, K.V.S.; Sharma, A.R.; Mahalakshmi, D.V.; Singh, D.; Prasad, V.K., Black carbon aerosol variations over Patiala city, Punjab, India A study during agriculture crop residue burning period using ground measurements and satellite data. Journal of Atmospheric and Solar - Terrestrial Physics, 2012. 84 - 85 : p. 45 - 51. 31. Duan, F.; Liu, X.; Yu, T.; Cachier, H., Identification and estimate of biomass burning contribution to the urban aerosol organic carbon conce ntrations in Beijing. Atmospheric Environment, 2004. 38 (9): p. 1275 - 1282. 32. Hill, J.; Goodkind, A.; Tessum, C.; Thakrar, S.; Tilman, D.; Polasky, S.; Smith, T.; Hunt, N.; Mullins, K.; Clark, M.; Marshall, J., Air - quality - related health damages of maize. Nature Sustainability, 2019. 2 (5): p. 397 - 403. 33. Sun, J.; Mooney, H.; Wu, W.; Tang, H.; Tong, Y.; Xu, Z.; Huang, B.; Cheng, Y.; Yang, X.; Wei, D.; Zhang, F.; Liu, J., Importing food damages domestic environment: Evidence from global soybean trade. PNAS, 2018. 115 (21): p. 5415 - 5419. 34. Sun, J.; Tong, Y. ; Liu, J., Telecoupled land - use changes in distant countries. Journal of Integrative Agriculture, 2017. 16 (2): p. 368 - 376. 35. Shyamsundar, P.; Springer, N.P.; Tallis, H.; Polasky, S.; Jat, M.L.; Sidhu, H.S.; Krishnapriya, P.P.; Skiba, N.; Ginn, W.; Ahuja , V.; Cummins, J.; Datta, I.; Dholakia, H.H.; Dixon, J.; Gerard, B.; Gupta, R.; Hellmann, J.; Jadhav, A.; Jat, H.S.; Keil, A.; Ladha, J.K.; Lopez - Ridaura, S.; Nandrajog, S.P.; Paul, S.; Ritter, A.; Sharma, P.C.; Singh, R.; Singh, D.; Somanathan, R., Fields on fire: Alternatives to crop residue burning in India. Science, 2019. 365 (6453): p. 536. 65 36. Ouyang, W.; Wei, X.; Hao, F., Long - term soil nutrient dynamics comparison under smallholding land and farmland policy in northeast of China. Science of The Tota l Environment, 2013. 450 - 451 : p. 129 - 139. 37. Schneider, M., Feeding China's Pigs: Implications for the Environment, China's Smallholder Farmers and Food Security . 2011. 38. Zhang, Q.; Jiang, X.; Tong, D.; Davis, S.J.; Zhao, H.; Geng, G.; Feng, T.; Zheng , B.; Lu, Z.; Streets, D.G., Transboundary health impacts of transported global air pollution and international trade. Nature, 2017. 543 (7647): p. 705 - 709. 39. Liu, J.; Hull, V.; Batistella, M.; DeFries, R.; Dietz, T.; Fu, F.; Hertel, T.; Izaurralde, R.C. ; Lambin, E.; Li, S., Framing sustainability in a telecoupled world. Ecology and Society, 2013. 18 (2). 40. Liu, J.; Mooney, H.; Hull, V.; Davis, S.J.; Gaskell, J.; Hertel, T.; Lubchenco, J.; Seto, K.C.; Gleick, P.; Kremen, C., Systems integration for glob al sustainability. Science, 2015. 347 (6225): p. 1258832. 41. Li, J., Statistical Yearbook of Heilongjiang State Farms 2011, Statistical Bureau of Heilongjiang Agricutlural Reclaimation p. 365 42. Sun, J.; Wu, W.; Tang, H.; Liu, J., Spatiotemporal pattern s of non - genetically modified crops in the era of expansion of genetically modified food. Scientific Reports, 2015. 5 : p. 14180. 43. Wang, J.; Christopher, S.A., thickness and PM2. 5 mass: implications for air quality studies. Geophysical research letters, 2003. 30 (21). 44. Bates, D.; Maechler, M.; Bolker, B.; Walker, S., Fitting Linear Mixed - Effects Models Using lme4. Journal of Statistical Software, 2015. 67 (1): p. 1 - 48. 45. Wickham, H., ggplot2: Elegant Graphics for Data Analysis . 2016, Springer - Verlag New York. 46. Wickham, H., Reshaping data with the reshape package. Journal of Statistical Software, 2007. 21 (2). 47 . Stock, J.H.; Watson, M.W., Heteroskedasticity - Robust Standard Errors for Fixed Effects Panel Data Regression. Econometrica, 2008. 76 (1): p. 155 - 174. 48. Croissant, Y.; Millo, G., Panel Data Econometrics with R: the plm package. Journal of Statistical Software, 2018. 27 (2): p. 1 - 43. 49. Zeileis, A.; Hothorn, T., Diagnosti c Checking in Regression Relationships. R news, 2002. 2 (3): p. 7 - 10. 66 50. Fox, J.; Weisberg, S., An R Companion to Applied Regression . second ed. 2011, Thousand Oaks, CA: Sage. 51. Venables, W.N.; Ripley, B.D., Modern Appled Statistics with S. Vol. Forth Edition. 2002, New York: Springer. 67 TELECOUPLED MANAGEMENT DECSIONS DRIVE MICROBIAL COMMUNITY COMPOSITION Overview Soil microbes are both economically and environmentally important in agriculture systems and many feedbacks exist between agriculture land management and microbial community composition 1 - 4 . Starting with crop choice, essentially every management decision has a cascading effect on the soil c ommunity that feedbacks to crop production via nutrient cycling processes and sometimes soil borne diseases. Understanding the factors impacting these decisions, as well as the potential synergies or trade - offs between different decision - making scenarios a nd their cascading impacts on microbial communities, can help researchers to better understand and promote environmentally sustainable and economically profitable practices. Farm cultivation and management decisions are shaped by endogenous socioeconomic and environmental factors as well as the exogenous market conditions (Figure S1). The literature identifies a variety of farmer demographics used to help inform cultivation decisions, such as age, education, phase in the family farm lifecycle, off - farm emp loyment, crop diversification, and farm size 5 - 10 . Beyond household attributes, environmental conditions and variability heavily influence cultivation decisions. Climate, rainfall, soil type and quality all must be taken into consideration prior to cultivation. Decisions are also influenced by far which is another component of the strategic decision - making process 11 . Farmers also use the crop cultivation and management pr actices of their peers to inform their own decision - making, adding elements of cooperation and competition to the decision process. Further, farmers must decide on their cultivation strategy for the year months before actual crop prices are available, nece ssitating the use of current market information (e.g., crop prices, subsidy amounts) while considering both domestic and international markets to estimate future prices at harvest. This complex decision - 68 making process plays out on the individual farm - scale , aggregates to regional cultivation patterns 12,13 and is affected by national and international policies and markets, resulting in both bottom - up and top - down processes. The soil microbial community is a crucial component of agricultural systems that is affected by and feedbacks to farmer decision - making. Soil microbes play a substantial role in both carbon and nitrogen cycling, which are major determinants in crop productivity as well as nutrient losses 1,4,14,15 . The a boveground plant community supplies the belowground community with carbon and nitrogen inputs via root exud at es and residual root and crop biomass which influences the microbial community by increasing the relative abundance of microbes that prefer nutrien t rich environments (e.g., copiotrophs) and decreasing the relative abundance of microbes that prefer nutrient depleted environments (e.g., oligotrophs) 16 . The microbial community decomp oses residual organic matter and mineralizes organic nitrogen and phosphorous into forms usable for plant and microbe uptake 4 , which improves nutrient cycling and reduces farmers dependences on nitrogen fertilizer 16 . Decomposition rates depend on the quality of the residual crop straw (e.g., C:N ratio, nitrogen concentration, cellulose and lignin content) 16 as well as the existing edaphic and nutrient conditions 17 . Crop choice 17,18 , crop rotation 3,19,20 , residue management 2,21 , fertilizer inputs 22 and tillage disturbance 3,21 all affect the concentration of nutrients in the soil and therefore the microbial community. Furthermore, excess nutrients associated with agricultural management select for microbes that accelerate the production of greenhou se gases such as N 2 O 2,23 - 25 . Many experimental studies 2,3,15,18,19,21,26,27 have reduced the complexity in the relationships between agricultural land management and the soil microbial community by using long - term test plots to limit variation and isolate the relationships in question. Often located on research farms, under University management, these test plots offer a semi - controlled environment for measuring 69 the impact of different management strategies on the microbial community. These studies the impact that alternative practices would have on the microbial community. While these studies provide a wealth of knowledge, in reality, only a fraction of farm ers report using best management practices 4,28 , which limits the real world applicability of their findings. Given that over 80% of anthropogenic N 2 O comes from agriculture fields 29 and the overwhelming majority of agriculture land is managed by farmers, there is a crucial need to understand the realized impact cultivation and management decisions have on the microbial community. To address this research gap and investigate the relationships between on - farm cultivation and management decisions and the soil microbial community, we assessed the results of 135 paired in - H eilongjiang Province. Historically, the province was dominated by continuous soybean production, however starting in the early 2000s low soybean prices, driven by competition from international imports, lead to regional declines in soybean production and c orn cultivation increased 30 - 32 . We ask: 1 ) How did changes in crop rotation practices shift regional cultivation patterns? 2) What were there cascading changes in management associated with the crop rotation practices ? and 3) Did changes in crop rotation and management practices influence the soil Results 4.2.1. Changes in crop rotation practices shift regional cultivation patterns The most pronounced response to competition from imported soybeans was a change in crop rotation practices among farmers that resulted in a regional cultivation shift away from a soybean - dominated and toward a corn - dominate agricultural system in Heilongjiang (Figure 1 8 ). Using the 70 past five years of cultivation history reporte d by the farmer, we grouped farmers and soil samples by rotation strategy (Table 7 ). Among respondents, continuous cultivation (5 or more years of cultivating the same crop) was 4 times more common than rotating annually between crops. Continuous corn (C) cultivation was the most frequent strategy, followed by continuous soybean (S) cultivation. Continuous corn farmers were found across Heilongjiang while continuous soybean farmers were found mainly in the northern part of province. Existing between the two continuous crop rotations, a fifth of the farmers had diversified by including both corn and soybeans in their crop rotation. The most common mixed rotations (M) were simple annual corn - soybean rotation (e.g., C - S - C - S - C) and bi - annual rotations (e.g., C - C - S - C - C; S - S - C - S - S) with a total of 10 unique rotation patterns recorded. Figure 18 . Underlying cropscape map from 2010 showing farm survey and soil sample locations in Heilongjiang for 2016. Green is soybean and yellows is corn. Point shape corresponds to crop rotation practice . Corn Soybean Soil sample + 71 Table 7 . Mean and (standard deviation) of influential variables grouped by crop rotation practice. The inf luential factors determined by PCA include: pH, Total organic carbon, silt, Nitrogen use, crop residue weight returned and yield. The lower case letters designated significant groups differences (p>0.05). Farm Interviews Soil Samples Group Label N Definition N - Use kg/ha Yield kg/ha Res idue kg/ha pH TOC Silt Continuous Corn C 60 5+ years of corn cultivation 234 (119)a 9930 (2091)a 42 (27)a 5.66 (0.86)a 2.15 (0.79) 0.28 (0.5) Mixed Rotation M 27 Any corn - soybean rotation 134 (129)b 5958 (5415)b 38 (26)a 5.43 (0.53)a,b 2.36 (0.68) 0.21 (0.4) Continuous Soybean S 48 5+ years of soybean cultivation 64 (27)c 1971 (363)c 56 (26)b 5.29 (0.50)b 2.46 (0.70) 0.09 (0.08) 4.2.2. Changes in crop rotation practices have cascading impacts on management Crop type and rotation practice significantly affected nitrogen inputs, yield, the amount of residue returned and soil pH (Table 7 ). Total organic carbon and the silt content did no t differ significantly between rotation practices. Farmers cultivating C reported higher nitrogen use and higher yields, followed M and S cultivation. Despite soybeans having only 1/10 th the residual biomass of corn, the amount of residue returned to the f ield was highest with S cultivation (> C > M). This is because farmers reported removing corn residue twice as often as soybean residue. The average soil pH significantly differed between C and S samples, but neither differed from M samples. Regardless of rotation practice, most samples were slightly acidic compared to the optimal pH range of 5.8 - 6.2. In addition, to the absence of lime application in the region, low soil pH can result from removing crop residue (and the nutrients that control acidity) and heavy application of nitrogen fertilizer 18,33 . 4.2.3. Changes in crop rotation practices influence microbial community structure, diversity, species abundance 72 Microbial community structure differed significantly between the three crop rotation practices (Figure 19 , Table 8 ). The influential factors from the farmer interviews and the abiotic soil analysis explained 25.1% (dim 1) and 12.2% (dim 2) of the variation in community structure for a total of 37.3%. Nitrogen use loads the highest on dimension 1 and is the most significant variable, likely capturin g C, M, S management gradient. Additionally, the residue management factor loads on dimension 1 with removing residue on the negative side and returning residue on the positive side, which supports the C, M, S management gradient. The remaining variables load on dimension 2, pH is the most significant and loads opposite of silt, as well as residue weight returned and TOC, which likely represent soil conditions along the C, M, S gradient. Temperature zone and sampling month were significant factors that exp lained regional variation within the soil community that may exist due to samples being collected across a large geographic extent. 73 Figure 19 . Principle coordinates analysis based on Unifrac distances for soil microbial communities sampled from farm fields in Heilongjiang China. The samples are colored by rotation method - which include C, S and M. The distance of 3% was used to define operational taxonomic units. Significant variables are in red and significant factors are in black. The vector points toward the direction of increase for a given variable and the arrow length indicates the correlation between the variable and the ordination scores. 74 Table 8 . The extracted PCoA components (e.g., Dim1 and 2) represent the major sources of variation in the data and determine the most influential variables and factors. Vector Dim 1 Dim 2 p - value Factors Dim 1 Dim 2 p - value pH - 0.08 0.28 0.003** Temp Zone 1 (warmest) - 0.007 - 0.007 0.001*** T OC 0.08 - 0.13 0.2 Temp Zone 2 (warmer) - 0.01 - 0.001 Silt - 0.13 - 0.20 0.03* Temp Zone 3 (colder) 0.002 0.001 Nitrogen use - 0.36 - 0.2 0.001*** Temp Zone 4 (coldest) 0.017 0.007 Residue Weight 0.13 - 0.15 0.07 May sample collection - 0.009 0.003 0.001*** June sample collection - 0.008 - 0.007 July sample collection 0.009 0.005 Removing Residue - 0.003 0.000 0.005** Returning Residue 0.012 - 0.000 Across all indices of alpha diversity, soil communities from continuous soybean rotations were the most diverse (Figure 20 , Table 9 ), followed by samples from mixed rotations and continuous corn rotations. 75 Figure 20 . Box plots f or alpha diversity analysis: (A) observed species, (B) Chao 1, (C) Shannon, (D) Simpson and (E) Pielou indexes of samples grouped by rotation type. Table 9 . Mean and (standard deviation) for the alpha diversity analysis. Group Observed Chao 1 Shannon Simpson Pielou C 1996 (229) 3314 (405) 6.3 (0.32) 0.988 (0) 0.83 (0.03) S 2167 (405) 3433 (725) 6.7 (0.33) 0.995 (0) 0.87 (0.02) M 2078 (194) 3382 (368) 6.4 (0.24) 0.991 (0) 0.84 (0.02) We recovered 23,315 operational taxonomic units (OTUs) from soil samples taken along a C, M and S management gradient. Across all rotation types 92% of the taxa were assigned to eight Phyla: Proteobacteria (36.3%), Acidobacteria (16.9 %), Bacteroidetes (12.4%), Actinobacteria (8.1%), Verrucomicrobia (8%), Chloroflexi (4.2%), Firmicutes (2.8%) and Gemmatimonadetes (2.6%). With the exception of Chloroflexi and Verrucomicrobia, which were the most abundant in S and M rotations respectively, the remaining phyla decreased in relati ve abundance from C > M > S rotations (Figure 21 ). The relative abundance of 9% of the sequences, representing 40 taxa or 0.5% of the total taxa identified, differed between rotations practices. Notably, more than 20 76 families were significantly associated with C and M rotations, while 2 families were significantly associated with S and M rotations and 3 families were found to significantly associated with only S rotations (p < 0.01) (Table 10 ). Figure 21 . Relative abundance for t he top 8 most abundant phyla, representing 92% of the taxa across 3 rotation practices: continuous corn, mixed rotations and continuous soybean. 77 Table 10 . Bacterial groups whose abundance differed significantly (p<0.001) between c rop rotation practices. Bacterial groups that significantly (p<0.001) associated with: Phyla/ (Family) C & M Rotations S & M Rotations S Rotations Type indicators/ associations Acidobacteria Oligotrophs Nutrient limited environments, negative correlation with C amendments 18,34 Gp1, Gp3, Gp4, Gp6 Gp 16 Actinobacteria Oligotrophs Nutrient limited environments, negative correlation with C amendments 35,36 Micromonosporaceae High pH & N inputs 14,37 Nocardioidaceae Pesticide degrader, Polluted soils 38 Streptomycetaceae Residue degrader 39 Micrococcaceae Residue degrader Armatimonadetes Chthonomonadaceae Bacteroidetes Degradation of OM, pesticides, soil fatigue 33,40 , 41,42 Chitinophagaceae Gemmatimonadetes Gemmatimonadaceae Low moisture conditions 43 , high pH 44 Proteobacteria Coxiellaceae Oxalobacteraceae Pesticide degrader 40 Sphingomonadaceae Degradation of OM 41 , pesticides 40,42 ,Corn cultivation 20 , Polluted soils 44 Rhizobiaceae Corn cultivation 20 Sinobacteraceae Polluted soils 45 Xanthomonadaceae Pesticide degrader 40 , Xanthobacteraceae Polluted soils 44 78 Caulobacteraceae Nitrosomonadaceae Nitrifying bacteria 46 High N inputs, Nitrification, polluted soil 44 Chloroflexi Anaerolineaceae Intensified agriculture, High N inputs, root worm infestation, burned residues, soil fatigue 30,47 Caldilineaceae Intensified agriculture Thaumarchaeota Ammonia oxidizing bacteria High N inputs, 42 Nitrification Nitrososphaeraceae Polluted soils 44 Verrucomicrobia Oligotrophs Nutrient limited environments, negative correlation with C amendments 18,34 c_Spartobacteria c_Subdivision3 Verrucomicrobiaceae Discussion Conventional wisdom dictates that soil under continuous corn management should receive higher inputs of carbon and nitrogen than the soil under a continuous soybean managment 31 . This is because corn produces 10 times more residual crop biomass than soybeans, which is theoretically returned to the soil post - harvest 31 . Crop residues provides labile sources of carbon and nitrogen to the microbial community and a re drivers of microbial community structure. Additionally, corn relies on external nitrogen inputs, whereas soybeans form associations with nitrogen fixing bacteria and do not require additional nitrogen fertilizer. However, the management 79 practices in the present study differ from previous research 2,3,15,17 - 19,21,26,27,48 and result in unexpected impacts on the microbial community. In response to competition from international soybean imports 31,32,49 , farmers in the region decreased soybean prod uction and shifted towards continuous corn cultivation. The shift towards continuous corn was accompanied by changes in residue management, where farmers commonly remove and burn corn residue after harvest. Additionally, farmers reported using the highest amount of nitrogen fertilizer with C cultivation, on average 30 kg/ha more than recommended by best management practices 50 . Farm ers also reported using nitrogen fertilizer in S production, on average 64 kg/ha. Because soybeans do not require nitrogen inputs, nitrogen overuse was more than double in S cultivation than C cultivation. The crop rotation and management strategies observ ed in the field resulted in different abundances of bacteria that are frequently associated with sources of carbon and nitrogen in soil. Due to a lack of mechanization in the region, crop residue is often left whole which prevents the biomass from breaking down over the winter. Therefore, the farmer often collected and remove the crop stalks/straw from the field. This practice was twice as likely to occur when farmers cultivated corn compared to soybeans and may explain why, contrary to previous research in the study region 31 , TOC did not significantly differ b etween C and S samples. Further, the levels of TOC measured in the sampled soils were relatively high and may make changes, especially increases, in soil carbon hard to detect. The tendency for farmers to remove crop residue and overuse nitrogen may explai n the acidic soil pH values and abundance of oligotrophic bacteria, which favor nutrient depleted environments 16 , in the soil community. Several of the most abundant phyla are found in i ntensified agricultural systems 33,40 . Within the most abundant phyla (Proteobacteria), several families were significantly associated with C & M (Table 10 ). Notably, the presence of Sphingomonadaceae, Oxalobacteraceae, Xanthomonadaceae (Proteobacteria), as 80 well as Chitinophagaceae ( Bacteroidetes ) has been associated with the degradation of organic matter 15,41 as well as pesticides 40 , and have been used as indicators of soil fatigue (caused by continuou s monocultures and inputs 33 ia, Actinobacteria and Verrucomicrobia are commonly regarded as oligotrophs and have frequently been shown to prefer nutrient limited environments and to be negatively correlated with carbon amendments 34,36,37 . In the present study, acidobacterial subgroups 1, 3, 4 and 6 significantly associated with C and M while acidobacte rial subgroup 16 was significantly associated with S and M. Additionally, Verrucomicrobia and Actinobacteria ( Micromonosporaceae) were also significantly associated with C and M rotations (Table 10 ). Further, previous research suggests that additions of le gume residue would increase the relative abundance of groups belonging to Actinobacteria while suppressing groups belonging to Proteobacteria, which may explain the significant associations between Micrococcaceae (Actinobacteria) , a known residue degrader 15,16 , and S cultivation. Nitrogen inputs and pollution were found to increase in Heilongjiang when farmers across the province shifted from soybean to corn cultivation 31,49 . This pattern held true in a meta - analysis that measured nitrogen balance in areas that had undergone soybean - corn conversions 49 . The obvious explanation for increase nitrogen use is that corn requires additional inputs of nitrogen while soybeans fix nitrogen through microbial associations. Among the sampled farms, n itrogen application rates were highest in continuous corn rotations, however, many of the farmers unnecessarily apply nitrogen to soybeans as well. The presence of several groups of bacteria found in soils of all rotations may indicate that there is more n itrogen present than can be used by the crops. For example, groups within the phylum Actinobacteria are known to associate with high soil pH and nitrogen inputs 26,37 . Nocardioidaceae and Streptomycetaceae significantly associated with soil from C and M while Micrococcaceae significantly associated with samples from S. 81 Additionally, two subphy la of Chloroflexi , ( Anaerolineaceae and Caldilineaceae) significantly associated with soil from S . This phylum is generally associated with intensified agriculture and can be associated with corn or soybean cultivation 20,47 . High abundance of Anaerolineaceae has been found by the literature in association with excess N inputs 26 , root worm infestation 20 , burned crop residues 26 and as an indicators for arable land fatigue 26 . Additionally ammonia oxidizing bacteria, which are key actors in the nitrogen cycle and indicate high levels of ammonia 46 , were found in soils from all rotations. In agri cultural systems these bacteria are essential for nutrient uptake by crops but can also lead to losses of ammonia - based fertilizer as N O and NO3 - nitrate pollution 23,26,46,51 . Nitrososphaeraceae (Thaumarchaeota) significantly associated with C and M rotations while Nitrosomonadaceae (Proteobacteria) significantly associated with S and M rotations (Table 10 ). The nitrification process is generally associated with continuous corn cultivation which requires high nitrogen inputs to maintain yields. Considering soybeans do not need external nitrogen inputs , the association of Nitrosomonadaceae with S rotations may indicate that there is more nitrogen present than can be used by the plant community 52 . Conclusion Overtime, agriculture management could impact the microbial community to such an extent that the changes feedback to the management decision. Depriving the soil community of the valuable nu trients in crop residues, regardless of crop type, can lead to a nutrient depleted environment. It is possible that removing crop residue from the field could lower yields over time and that would feedback to management decisions. Some farmers may invest t he time, labor and capital required to return the crop residue to the field. However, it is more likely that farmers may compensate for yield loss by increasing nitrogen fertilizer use 16,53 . Increased input cost could affect profitability, but given that nitrogen fertilizer is relatively cheap, widely available and easy 82 to apply, farmers may find it worth the cost 14,17,19 . Excess fertilizer that cannot be utilized by plants is lost as N O and NO 3 - pollution and can even result in a microbial community that is more likely to release nitrogen rather than convert it into a form useable by plants 25,54 . Overtime, agriculture management could impact the microbial community to such an extent that the changes feedback to the management decision. Materials & Methods 4.5.1. Site Description The study area is Heilongjiang Province located in northeastern China, between 44° and 53° north and 121° to 135° east, and is bordered by Russia o n the north and east sides. Heilongjiang has an area of 454,800 km 2 and is comprised of thirteen prefectures. While winters are harsh, the region has a humid continental climate with high temperatures and rainfall in the summer months. July temperatures r each 20° C [9], falling just short of the ideal temperature, 25° C [10], for from early May to mid - August which, in addition to a relatively low population and l arge contiguous land parcels, contributes to Heilongjiang producing one - third to one - quarter of the continuously declined. Previous analyses indicate that on - farm soy - planting proportion is decreasing while corn planting proportion is increasing in response to competition from international soybean imports and fertilizer subsidies 30 - 32,55 . 4.5.2. Sampling Design Farm surveys and soil samples were collected in Heilongjiang Province, China between May and July of 2016. Interview questions covered farmer demographics such as ed ucation, age, 83 income, off - farm employment, land - ownership status and questions on the influence of policy and price fluctuations that may be used to help explain management decisions. Management decisions included in the survey were land level, crop type, inputs (fertilizers, herbicides and pesticides types and application rates), growth period, planting and harvest date, production costs, farm gate price, yield, mechanization, subsides, residue management, crop - rotation and planting history. At the conclus ion of the survey, the interviewer would ask the farmer for permission to take a soil sample and the farmer would then identify their field boundaries. The boundary was marked with field flags before 6 (10 cm deep cores, 1 cm wide) samples were collected from the root - influenced soil (layer where the majority of crop roots are located, 0 - 10 cm from soil surface) and mixed to generate one composite sample per field. One hundred and thirty - five unique soil samples were collected and the field composite was u sed to measure abiotic attributes of the soil including soil pH, total organic carbon and soil texture. Samples were processed with a 3 mm soil sieve and soil texture was determined each night, so that the soil was field moist. Soil texture classifications were checked against published values for the region 56,57 . The samples were then dried at ambient temperatures and stored in marked plastic bags 58 . Soil pH was measured using pH meter at the Chinese Academy of Sciences, Institute of Soil Sciences in Nanjing. Soil organic carbon was determined by Yingtan Red Soil Ecological Experim ental Station. 4.5.3. DNA extraction, PCR, and DNA sequencing From the processed samples, 5 grams of soil was separated into small plastic tubes, labeled and stored at - 40 C upon returning to the research institute and prior to DNA extraction. DNA was extracted from 0.25 grams portions of the frozen soil samples at the Chinese Academy of Sciences, Institute of Soil Sciences in Nanjing, using an adapted protocol from PowerSoil TM DNA Isolation kit (MO BIO Laboratories, Inc.) [13]. A NanoDrop was used to make 84 sure all samples had A260/280 readings between 1.8 and 2.0. Aliquots of the extracted DNA were freeze - Ecology for further analysis. The samples were rehydrated with PCR - grade water to reach the original concentration, vortexed and allowed time to re - suspend. Rehydrating dry samples improves the quantity and quality of extracted DNA [12]. The aliquots were used for amplification of a portion of the 16S rRNA genes (~250 nucleotides) using the Schloss primers F515 and R806 (respectively, - GTGCCAGCMGCCGCGGTAA - - GGACTACVSGGGTATCTAAT - Sequencing Support Facility. Raw sequences were initially processed by pair - end assembled, chimeras removed and filtered for quality before being clustered into operational taxonomic units (OTUs) at 97% similarity . Raw paired - end sequences were merged and chimeras removed using de novo mode. The quality of the merged reads for all samples was then checked using FastQC 59 . A value above 30 indicates good quality and a value below 20 indicates poor quality. The average quality per read was 37, the minimum and maximum observed were 32 and 38, respec tively for all samples. Furthermore, with almost 100% expected overlap, the merged reads have a high rate of accuracy identifying a nucleotide. Furthermore, with almost 100% expected overlap, the reads have a high rate of accuracy identifying a nucleotide. Processing was done via the UPARSE pipeline (i.e., usearch, cluster_otus), chimeras were automatically removed. Merged reads were clustered using USEARCH version 8.1 with the usearch_glob al command and an id of 0.985 so that reads within a cluster differed by no more than 3%. Representative 60 with training set No.16. A 85 phylogenetic tree was constructed with FASTTREE 61 from the representative sequences aligned with INFERNAL 62 . 4.5.4. Microbial Data Analysis A principle coordinates analysis (PCoA) based on Unifrac distances was used to depict distances ( - diversity) among samples from farm fields in Heilongjiang China. Environmental 63 . Additionally, Chao, Shannon, Simpson and Pielou diversity indices were calculated. Last, species relative abundance was calculated and compared between crop rotation types. Bacterial groups that differed significantly between rotation groups were identified and interpreted. All statistical analysis w ere performed in the R environment 64 , using packages phyloseq 65 , RDPutils 66 , ade4 67 , ggplot2 68 , ggrepel 69 , ggordiplots 70 , QsRutils 71 , RVAideMemoire 72 , tidyverse 73 , indicspeci es 74 , vegan 63 , data.table 75 and nlme 76 . 86 REFERENCES 87 REFEREN CES 1. Barrios, E., Soil biota, ecosystem services and land productivity. Ecological Economics, 2007. 64 (2): p. 269 - 285. 2. Baruah, A., et al., Effect of Organic Residues with Varied Carbon Nitrogen Ratios on Grain Yield, Soil Health, and Nitrous Oxide Emission from a Rice Agroecosystem. Communications in Soil Science and Plant Analysis, 2016. 47 (11): p. 1417 - 1429. 3. Dorr de Quadros, P., et al., The Effect of Tillage System and Cr op Rotation on Soil Microbial Diversity and Composition in a Subtropical Acrisol. Diversity, 2012. 4 (4). 4. Hamilton, S.K., J. Doll, E., and G.P. Robertson, The Ecology of Agricultural Landscapes: Long - Term Research on the Path to Sustainability . 2015: Oxford University Press. 5. Anosike, N. and C.M. Coughhenour, The socioeconmic basis of farm enterprise divserification decsions. Rural Sociology 1990. 55 (1): p. 1 - 24. 6. Botterill, L. and N. Mazur, Risk and risk percpetion 2004, Australian Governemnt Rural Industires Research and Development Corporation. 7. OUTSTANDING EXPORT SALES AND EXPORTS BY COUNTRY, REGION AND MARKETING YEAR , U.S.D.o. Agriculture, Editor. 2014. 8. Ondersteijin, C.J.M., G.W.J. Giesen, and R.B.M. Hurine, Perceived environmental uncertainty in Dutch dair y farming context on stragetic choice Agricultural Systems 2006. 88 : p. 205 - 226 9. Ondersteijin, C.J.M., G.W.J. Giesen, and R.B.M. Hurine, Identification of farmer characterisitics and farm strategies explaining changes in enviornmental and economic performa nce of dairy farms Agricultural Systems, 2003. 78 : p. 31 - 55. 10. Parker, D.C., A. Hessl, and S.C. David, Complexity, land - use modeling, and the human dimension: Fundamental challenges for mapping unknown outcome spaces. Geoforum 2007. 29 : p. 789 - 804. 11. Ohlmer, B., K. Olson, and B. Brehmer, Understanding farmers' decsion making processes and improving managerial assistance Agricutlural Economics 1998. 18 : p. 273 - 290. 12. Willock, J., et al., Farmers' attitudes, objectives, behaviors, and personailty traits: T he Edinburgh study of decision making on farms Journal of Vocational Behavior 1999. 54 : p. 5 - 36. 13. Herman, A., Enchanting resilience: Relations of care and people - place connections in agriculture Journal of Rural Studies 2015 42 : p. 102 - 111. 88 14. Generalized bala nce and the emergence of network structures through influence and selection in Influence Amplifies the Effects of Selection: Generalized balance and the emergence of network structures through influence and selection I. progress., Editor. 2017. 15. De la Cruz - Barrón, M., et al., The Bacterial Community Structure and Dynamics of Carbon and Nitrogen when Maize (Zea mays L.) and Its Neutral Detergent Fibre Were Added to Soil from Zimbabwe with Contrasting Management Practices. Microbial Ecology, 2017. 73 (1): p. 13 5 - 152. 16. Ortiz - Cornejo, N.L., et al., Incorporation of bean plant residue in soil with different agricultural practices and its effect on the soil bacteria. Applied Soil Ecology, 2017. 119 : p. 417 - 427. 17. Lian, T., et al., The fate of soybean residue - carbon lin ks to changes of bacterial community composition in Mollisols differing in soil organic carbon. Soil Biology and Biochemistry, 2017. 109 : p. 50 - 58. 18. Behnke, G.D., et al., Acidification in corn monocultures favor fungi, ammonia oxidizing bacteria, and nirK - d enitrifier groups. Science of The Total Environment, 2020. 720 : p. 137514. 19. Li, C., et al., Effect of monoculture soybean on soil microbial community in the Northeast China. Plant and Soil, 2010. 330 (1): p. 423 - 433. 20. Benitez, M. - S., S.L. Osborne, and R.M. Le hman, Previous crop and rotation history effects on maize seedling health and associated rhizosphere microbiome. Scientific Reports, 2017. 7 (1): p. 15709. 21. Ceja - Navarro, J.A., et al., Phylogenetic and Multivariate Analyses To Determine the Effects of Different Tillage and Residue Management Practices on Soil Bacterial Communities. Applied and Environmental Microbiology, 2010. 76 (11): p. 3685. 22. Ramirez, K.S., J.M. Craine, and N. Fierer, Consistent effects of nitrogen amendments on soil microbial communit ies and processes across biomes. Global Change Biology, 2012. 18 (6): p. 1918 - 1927. 23. Ouyang, Y., et al., Ammonia - oxidizing bacteria are more responsive than archaea to nitrogen source in an agricultural soil. Soil Biology and Biochemistry, 2016. 96 : p. 4 - 15. 24. Al - Amin, A., et al., Price and Trade Mechanism for Palm and Soybean oils: A comparative analysis of economic developments between Asai and the contemporary world. . Asian Journal of Agricultural Research, 2015. 9 (1): p. 1 - 11. 25. Weese, D.J., et al., Long - term nitrogen addition causes the evolution of less - cooperative mutualists. Evolution, 2015. 69 (3): p. 631 - 642. 26. Jiménez - Bueno, N.G., et al., Bacterial indicator taxa in soils under different long - term agricultural management. Journal of Applied Microbiology, 2016. 120 (4): p. 921 - 933. 89 27. Li, T., et al., Changes in soil bacterial community structure and microbial function caused by straw retention in the North China Plain. Ar chives of Agronomy and Soil Science, 2019: p. 1 - 12. 28. Baumgart - Getz, A., L.S. Prokopy, and K. Floress, Why farmers adopt best management practice in the United States: A meta - analysis of the adoption literature. Journal of Environmental Management, 2012. 96 ( 1): p. 17 - 25. 29. Ollivier, Q.R., et al., Winter emissions of CO2, CH4, and N2O from temperate agricultural dams: fluxes, sources, and processes. Ecosphere, 2019. 10 (11): p. e02914. 30. Dou, Y., et al., Land - Use Changes in Distant Places: Implementation of a Telec oupled Agent - Based Model. Journal of Artificial Societies and Social Simulation, 2020. 23 . 31. Tong, Y., et al., Mollisols Regions. Scientific Reports, 2017. 7 (1): p. 18064. 32. Sun, J., Y. Tong, and J. Liu, Telecoupled land - use changes in distant countries. Journal of Integrative Agriculture, 2017. 16 (2): p. 368 - 376. 33. Indicators of arable soils fatigue Bacterial families and genera: A metagenomic approach. Ecological Indicators, 2018. 93 : p. 490 - 500. 34. Ramirez - Villanueva, D.A., et al., Bacterial community structure in maize residue amended soil with contrasting management practices. Applied Soil Ecology, 2015. 90 : p. 49 - 59. 35. Bicudo da Silva, R.F., et al., The S ino - Brazilian Telecoupled Soybean System and Cascading Effects for the Exporting Country. Land 2017. 6 (53): p. 1 - 19. 36. Liu, J., et al., Diversity and distribution patterns of acidobacterial communities in the black soil zone of northeast China. Soil Biology and Biochemistry, 2016. 95 : p. 212 - 222. 37. Anderson, C.R., et al., Biochar induced soil microbial community change: Implications for biogeochemical cycling of carbon, nitrogen and phosphorus. Pedobiologia, 2011. 54 (5): p. 309 - 320. 38. Satsuma, K., Mineralization of s - triazine herbicides by a newly isolated Nocardioides species strain DN36. Applied Microbiology and Biotechnology, 2010. 86 (5): p. 1585 - 1592. 39. Chouvenc, T., et al., The Termite Fecal Nest: A Framework for the Opportunistic Acquisition of Beneficial Soil Streptomyces (Actinomycetales: Streptomycetaceae). Environmental Entomology, 2018. 47 (6): p. 1431 - 1439. 40. Zhang, P., et al., Sorption, desorption and degradation of neonicotinoids in four agricultural soils and their effects on soil microorganisms. Science of The Total Environment, 2018. 615 : p. 59 - 69. 90 41. Fierer, N., M.A. Bradford, and R.B. Jackson, Toward an ecological classification of soil bacteria. Ecology, 2007(88): p. 1354 - 1364. 42. Sun, G., et al., Biodegradation of Dichlorodiphenyltrichloroethanes (DDTs) an d Hexachlorocyclohexanes (HCHs) with Plant and Nutrients and Their Effects on the Microbial Ecological Kinetics. Microbial Ecology, 2015. 69 (2): p. 281 - 292. 43. DeBruyn, J.M., et al., Global Biogeography and Quantitative Seasonal Dynamics of <span class=&qu ot;named - content genus - species" id="named - content - 1">Gemmatimonadetes</span> in Soil. Applied and Environmental Microbiology, 2011. 77 (17): p. 6295. 44. Hemmat - Jou, M.H., et al., Analysis of microbial communities in heavy metals - contaminated soils using the metagenomic approach. Ecotoxicology, 2018. 27 (9): p. 1281 - 1291. 45. Zhou, Y., et al., Description of Sinobacter flavus gen. nov., sp. nov., and proposal of Sinobacteraceae fam. nov. International Journal of Systematic and Ev olutionary Microbiology, 2008. 58 (1): p. 184 - 189. 46. Prosser, J.I., I.M. Head, and L.Y. Stein, The Family Nitrosomonadaceae , in The Prokaryotes: Alphaproteobacteria and Betaproteobacteria , E. Rosenberg, et al., Editors. 2014, Springer Berlin Heidelberg: Berli n, Heidelberg. p. 901 - 918. 47. Song, X., et al., Changes in the Microbial Community Structure and Soil Chemical Properties of Vertisols Under Different Cropping Systems in Northern China. Frontiers in Environmental Science, 2018. 6 : p. 132. 48. Navarro - Noya Y. E., et al., Relative impacts of tillage, residue management and crop - rotation on soil bacterial communities in a semi arid agroecosystem. Soil Biology & Biochemistry, 2013. 65 : p. 86 - 95. 49. Sun, J., et al., Importing food damages domestic environment: Evidence f rom global soybean trade. PNAS, 2018. 115 (21): p. 5415 - 5419. 50. Ju, X. and P. Christie, Calculation of theoretical nitrogen rate for simple nitrogen recommendations in intensive cropping systems: A case study on the North China Plain. Field Crops Research, 20 11. 124 (3): p. 450 - 458. 51. He, J. - Z., et al., A review of ammonia - oxidizing bacteria and archaea in Chinese soils. Frontiers in Microbiology, 2012. 3 : p. 296. 52. Staley, C., et al., Urea Amendment Decreases Microbial Diversity and Selects for Specific Nitrifying Strains in Eight Contrasting Agricultural Soils. Frontiers in Microbiology, 2018. 9 : p. 634. 53. Bechini, L., et al., Barriers and drivers towards the incorporation of crop residue in the soil. planned behaviour. Italian Journal of Agronomy, 2015. 10 (663). 91 54. Schmidt, J.E., D.J. Weese, and J.A. Lau, Long - term agricultural management does not alter the evolution of a soybean rhizobium mutualism. Ecological Applications, 2017. 27 (8): p. 2487 - 2496. 55. Sun, J., et al., Spatiotemporal patterns of non - genetically modified crops in the era of expansion of genetically modified food. Scientific Reports, 2015. 5 : p. 14180. 56. Shirazi, M.A. and L. Boersma, A Unifying Quantitative Analysis of Soil Texture. Soil Sci ence Society of America Journal, 1984. 48 (1): p. 142 - 147. 57. Xing, B., et al., Physical and Chemical Characteristics of a Typical Mollisol in China. Communications in Soil Science and Plant Analysis, 2005. 35 (13 - 14): p. 1829 - 1838. 58. Clark, I.M. and P.R. Hirsch, Survival of bacterial DNA and cultural bacteria in archived sils from the Rothamsted Broadbalk experiment Soil Soil Biology & Biochemistry 2008. 40 : p. 1090 - 1102 59. Team., R.C., R: A language and environment for statistical computing. 2016, R Foundation for Statistical Computing: Vienna, Austria. 60. McMurdie, P.J. and S. Holmes, phyloseq: An R package for reproducible interactive analysis and graphics of microbiome census data. PLoS ONE, 2013. 8 (4). 61. Quensen, J., RDPutils: R Utilities for Processing RDPTool Outp ut. R package version 1.4.1, 2018. 62. Chessel, D., A. Dufour, and J. Thioulouse, The ade4 Package - II: 63. Two - Table and K - Table Methods. R News, 2004. 7 (2): p. 47 - 52. 64. Wickham, H., ggplot2: Elegant Graphics for Data Analysis . 2016, Springer - Verlag New York. 65. Slow ikowski, K., ggrepel: Automatically Position Non - Overlapping Text Labels with 'ggplot2'. 2019. 66. Quensen, J.F., ggordiplots: Make ggplot Versions of Vegan's Ordiplots . 2018. 67. Quensen, J.F., QsRutils: R Functions Useful for Community Ecology . 2018. 68. Herve, M., RVAideMemoire: Testing and Plotting Procedures for Biostatistics . 2019. 69. Wickham, H., et al., Welcome to the {tidyverse}. Journal of Open Source Software, 2019. 4 (43). 70. De Caceres, M. and P. Legendre, Associations between species and groups of sites: indices and statistical inference. Ecology, 2009. 71. Oksanen, J., et al., vegan: Community Ecology Package . 2019. 92 72. Dowle, M. and A. Srinivasam, data.table: Extension of `data.frame` . 2019. 73. Pinheiro, J., et al., nlme}: Linear and Nonlinear Mixed Effects Models . 2019. 93 USING AGENT - BASED MODELLING TO M EASURE THE IMPACT OF LAND MANAGEMENT ON EVIRONMNETAL HEALTH IN TELECOUPLED HUMAN AND NATURAL SYSTEMS Overview One challenge of conducting interdisciplinary research is how to integrate the natural and social components of a system. Much of this complexity lies in how to characterize the environment and its relationship to human actions, which vary at different temporal and spatial scales 1 . One approach is to employ a systems dynam ics model using the impact of human action as an aggregated effect. However, agricultural land - use patterns are formed from cultivation decisions made by individual actors. Aggregating the impact of human action on a landscape may overlook the importance o f variations between actors and cause the researcher to miss key insights. This omission may result in a loss of predictive and explanatory power as well as a generalization of the underlying processes 1 . Agent - based models provide a key advantage for modeling an 2 . This allows for the integration and analysi s of human action into the causes, effects, and processes of landscape use and cover change while incorporating individual actor variation 2 . M odel Description 5.2.1. Purpose The present ABM is an adaption of TeleABM 3,4 , which simulates soybean trade and land - use change in the sending (e.g., Brazil) and receiving (e.g., China) systems. The TeleABM was written in Java using the Repast platform and designed to be easily adapted to other land systems. The model was cali brated with empirical land use data and farmer interviews for decision rules. Land 94 use and production patterns in each system are aggregated to the regional level to influence crop price in the trade model and, ultimately, land use is the telecoupled syste m. 5.2.2. Study System The sending system, Mato Grosso, Brazil, was not modified from the original TeleABM 3 . The receiving system, Heilongjiang a , is the top grain producing region in China. Driven by competition from international imports, farmers in Heilongjiang are converting to rice and corn at the expense of soybean production 3,5 - 7 . The data that was used to calibrate the receiving system was collected during the summers of 2016 and 20 17, which included ground - truth points, farmer surveys and soil samples. 5.2.3. Agents There are 3 types of agents within the receiving system, farmer agents, government agent and trader agents. Each farmer agent is given a list of properties (e.g., labor, capi tal, property) and actions (e.g., land use decision; calculate profit). Information gathered from the farmer interviews w as used for agent attributes. Government agents set domestic production quotas and subsidies in the receiving system and tariffs on imp orted soybeans from the sending system in the trade model. Local trade agents aggregate soybean production to the international trade agent where soybean price is determined from production in both systems. Soybean price is then passed from the internation al trade agent to the local trade agents and then to farmer agents for the next time step. Government and trader agents were not modified from the original 3 in this model adaption and are used for scenario analysis. 5.2.4. Environment 95 The environment is based on a grid of cells , representing 900 m 2 or 0.0625 km 2 based on 250m X 250m map resolution. Each cell is assigned biophysical properties (e.g., soil texture, pH, SOM, temperature, precipitation, elevation) based empirical data. Cell s are assigned to a farmer agent, who determines cultivation and management decisions. Crop yield is a function of crop type, crop rotation, fertilizer inputs and residue management. Table 11 provides a details on the household and farm characteristics in the TeleABM. 96 Table 11 . Household and farm characteristics of the TeleABM receiving system. Including variable names, variable description, variable distribution, parameters and independence. *denotes variables selected for Uncer tainty and Sensitivity Analysis. Household Characteristics Var_Name Var_Descripitio n (units) Var_Distributio n Parameters Independe nt (yes/no) SA/UA familyPopulation Number of family members Normal M (sd) = 3.46 (1.15) Yes * hhdHeadMale Gender Dummy for head of household (0 = M; 1 = F) Discrete 0=0.073; 1=0.027 Yes age Head of Household Age Normal M (sd) = 28.5 (10.8) Yes dependentRatio ((kids + elders)/ total family num) Normal M (sd) = 0.21 (0.1) No * genderRatio ratio of male to female (0 = M; 1 = F) Normal M (sd) = 0.1597 (0.18) No * hhdHeadunHealth Health Dummy for head of household (0=healthy; 1=not healthy) Discrete 0=0.716; 1=0.284 Yes occupation 1 full time farmer; 2 part - time farmers; 3 non farmer Discrete 1=0.82;2=0.1 4; 3=0.04 Yes knowInternationalTrade 1=know of soybean trade Discrete 0=0.46; 1=0.54 Yes whetherknow_soybean_ixYe s 1=know of imported soybean competition Discrete 0=0.28; 1=0.72 Yes * whetherknow_pericoupledper i 1=know Discrete 0=0.978; 1=0.022 Yes whetherknow_transgeneYes 1=know of GM Discrete 0=0.3; 1=0.7 Yes whether_know_import_gmo Yes 1=know that imported soybeans are GM Discrete 0=0.69; 1=0.31 Yes 97 noOffFarmIncome Scale 1 - 5 (1 = no off farm income) Disrete 0=0.664; 1=0.23; 2=0.08; 3=0.007; 4=0.007; 5=0.002 Yes * noBigMachine Number of machines (1 - 8) Discrete 0=0.165; 1=0.153; 2=0.158; 3=0.150; 4=0.148; 5=0.092; 6=0.056; 7=0.032; 8=0.032 Yes * Farm Characteristics Var_Name Var_Descripitio n (units) Var_Distributio n Parameters Independe nt (yes/no) cellSize Farm size Normal totalFertilizerInput Fertilizer input for all crops Normal No totalWaterInput Water input for all crops Normal No totalExtraN Fertilizer input that exceeds crop requirements for all crops Normal No totalExtraNsoy Fertilizer input that exceeds soybean requirements Normal No totalExtraNcorn Fertilizer input that exceeds corn requirements Normal No totalExtraNrice Fertilizer input that exceeds rice requirements Normal No farmPM25 PM 2.5 emissions Normal No farmNOX NOx emissions Normal No farmCO2 CO2 emissions Normal No cornProportion Corn cells/ total cells Normal No riceProportion rice cells/ total cells Normal No soyProportion soybean cells/ total cells Normal No meanTemp Soil temperature Uniform 2.6 Yes meanSoila Soil texture Uniform 22.02 Yes 98 meanSoilb Soil texture Uniform 57.14 Yes meanSoilc Soil texture Uniform 20.83 Yes farmpH Average farm soil pH Uniform No farmSOC Average farm soil organic carbon Uniform No elevation cell elevation from a map Yes soc farm soil organic carbon from a map Yes toc farm soill organic carbon/ cellSize Normal No soilpH cell soil pH Normal No soyYield Farm soybean yield Uniform 2000 + (0.141* Soc - 0.703) + planting area No cornYield Farm corn yield Uniform 9597 + (0.141* Soc - 0.703) + (1.195* Fertilizer) + planting area No riceYield Farm corn yield Uniform 8112 + (0.141* Soc - 0.703) + (1.1* Fertilizer) + planting area No burnRatecorn 0=nothing; 1=burn Discrete 0=0.45; 1=0.55 Yes returnRatecorn 0=nothing; 1 = return Discrete 0=0.55; 1=0.45 Yes burnRatesoy 0=nothing; 1=burn Discrete 0=0.94; 1=0.06 Yes returnRatesoy 0=nothing; 1 = return Discrete 0=0.06; 1=0.94 Yes burnRaterice 0=nothing; 1=burn Discrete 0=0.26; 1=0.74 Yes returnRaterice 0=nothing; 1 = return Discrete 0=0.74; 1=0.26 Yes 5.2.5. Process Overview To initialization the TeleABM the user must first set the global parameters by determining which systems to simulate (e.g., sending, receiving or both) and setting crop price as either static, 99 empirical or trad e scenario. When only one system is initialized, crop price is either set as a static price or read in as a dynamic price from file. Next, land use and suitability maps must be read in before initializing the agents, allocating land cells and setting up agent properties (Table 1 1 ). Farmer agent s determine their land use decision by deciding how much land to allocate to rice production and then by deciding the proportion of corn and soybeans from what land is left. This is because rice price is the highest but also requires suitable land (e.g., next to existing water) and investment capital to convert from dry to paddy land production. Farmer agents then pass their land use decision (e.g., corn, soybeans, rice) and inputs (e.g., fertilizer, crop residue) to the land cell and update land use in the current step. The previous land uses are recorded to the land cell property and undergo ecological processes to generate new crop yield, soil pH and SOM in response to management decisions. When all ag ents and land cells have been updated the model moves to the annual accounting step, where profit and environmental impacts are stored (e.g., crop production, GHG emissions and nutrient runoff) (Figure 22 ). Information on crop production is passed to the t rade agent then to soil degradation, soil quality feedback to cultivation and management decisions through the yield and profit functions. After the annual accoun ting step, the model moves to the next time step. 5.2.6. Design Concepts To extend this model, management decisions were added after cultivation decisions. Residue Management & GHG Emissions . Residue management starts by setting each farmer agents decision to burn or return the residual crop biomass post - harvest. The decision to burn/return is set randomly, per crop type, in proportion to the values observed by farmers in the field (Table 11 ). If the residue management decision is to burn, then GHG emissions i,t are calculated, which include 100 g/kg of CO2, NOx and PM 2.5 released from burning straw biomass for crop type c by farmer agent i at time t (equation 1) 8 - 12 . Alternatively, if the residue management decision is to return, the resid ual biomass is added to the soil as carbon(equation 4) 6 . Nitrogen Fertilizer Use & Overuse. For each crop nitrogen fertilizer input for farmer agent i at time t is set as the average input observed in the field for crop type c random deviation for crop type c (equation 2). Excess nitrogen fertilizer use for farmer agent i at time t is then calculated by subtracting the recommended fertilizer rate for crop type c from the fertilizer input for farmer agent i at time t (equation 3). SOC . SOC is the model component that connects agent management decision to the quality of the cell which in turn feeds back to cultivations decisions. SOC for land cell l at time t is equal to SOC for land cell l at time t - 1 plus the residual biomass, root biom ass and seed carbon of crop type c for farmer agent I at time t, plus fertilizer use at land cell l , plus rhizodeposition, temperature and perception in region r (equation 4). 1 4 2 3 101 Yield. Yield is the cumulating model step and is used to feedback to cultivation decisions between model ticks. Yield for farmer agent i at time t , is based off the observed yield for crop type c plus the nitrogen fertilizer input for farmer agent i at time t resp onse for crop type c plus SOC for land cell l at time t (equation 5). 5 102 Figure 22 . Adapted from Dou et al. Figure 6 3 , the decision - making process making process in the receiving system of teleABM. The dotted lines represent linkages to the sending system which can be turned on or off to simulate different trade scenarios. The present version of the model extends beyond the cultivatio n decisions and adds management decision which decisions affect soil properties that then feedback to the cultivation decisions via a profit function ( i.e ., yield * price). 103 5.2.7. Validation The land use decision has been validated by Dou et al. 3,4 Validation of the Management Decisions sub model was done by ensuring that simulated soil carbon values fell between the maximum and minimum value measured from the field ( Figure 27) . Results Based on the empirical data, model initialization starts with 50% of the simulated area in Heilongjiang planted to corn, 35% to soybeans and 15% planted to rice ( Figure 23 & 24 ). The area planted to corn increases at the expense of soybeans rice until model tick 7. Then Rice beg ins to increase the expense of corn and soybeans. From Figure 24 we can see the spatial hot spots of soybean decline on the east and west sides of Heilongjiang. Rice and corn expansion are somewhat clustered. Rice can only expand near water and when adjacent to other rice cells which creates the blocks of white (Fig ure 24) . Figure 23 . The area planted to corn, soybean and rice overtime. 104 Figure 24 . Simulated land use maps showing the initial and final model tick. Soybeans are green, corn is yellow and r ice is white . The pattern of greenhouse gas emissions as a result of crop residue burning is closely tied to the area planted to each crop ( Figure 2 5 ). Emissions of CO 2, NO x and PM 2.5 all increase over the model simulation. In the first few ticks the levels of CO 2 increase rapidly but slow around tick 7, following the increase in the area planted to corn ( Figure 23). Emissions of PM 2.5 also increase quickly in the first 7 ticks but do not slow as much as the emissions of CO 2 . At tick 13 the emission of PM 2.5 surpasse the emissions of CO 2 . This pattern is because burning rice residue produces more PM 2.5 while burning corn residue produces relatively mor e CO 2 . soybeans corn rice year 105 Figure 25 . (A) total emissions from all farms over time (b) change in total emissions from all farms over time. PM2.5 is red, CO2 is black and NOx is grey. Additionally, the emissions can be viewed based on the area planted to each crop ( Figure 2 6 ). The emissions of PM 2.5 are highest from farms planting high proportions of rice. The PM 2.5 emissions factor for rice is 14.73 while the PM 2.5 emissions factor for corn is 11.7 and soybeans is 5.6. CO 2 and NO x emi ssions were highest from farms planting high proportion s of corn. Farms planting high proportion of soybeans had the lowest emissions because soybean residue is seldomly burn t. A B 106 Figure 26 . Average emissions of CO2 (l eft column ), NOx ( center column ), and PM2.5 ( right column ) from farms shown by crop planting proportion. Corn is yellow, soybeans green and rice blue. The level of toc was empirically measured using soil samples and input maps , toc was higher in samples from corn and rice than samples taken from soybean fields ( Figure 2 6 & 27 ). When simulated using the residue return rates , the level of toc in corn and soybeans cells w as higher than rice, however, the level of toc in rice cells increased over the simulation wh ile the level of toc in corn and soybeans cells decreased. The spatial pattern of TOC is similar to the pattern of land use change ( Figure 24). Areas planted to rice experienced the most rapid soil TOC decline followed by areas planted to corn ( Figure 27). 107 Figure 27 . Average total organic carbon by crop type overtime. Corn is yellow, soybeans green and rice blue. Figure 28 . Simulated land use maps showing the initial and final model tick. Green indicates high levels of total organic carbon, followed by yellow, orange and finally red. Red indicates degraded soil. 108 On average, yield slightly decreased over the model simulation for all crops ( Figure 2 9 ). This is likely due to the small decrease in to c overtime which was driven by the removal of crop residue from fields. Figure 29 . Yield by crop type over time. Corn is shown in the left panel followed by soybean and rice. The grey shaded area represents standard deviation. N itrogen overuse was highest from farms with high planting proportions of corn and lowest from farms with high planting proportions of soybeans ( Figure 30 ). Excess nitrogen use contributes to the emissions of NO x ( Figure 2 5 & 2 6 ). Figure 30 . Average nitrogen overuse on farms shown by planting proportion. Corn is yellow, soybeans green and rice blue. Discussion The presented extension advances the TeleABM, which represents two different land systems that our connecte d through trade flows, by going beyond land use changes to assess the biophysical impact of international trade on soil properties and greenhouse gas emissions. T he sub - model 109 builds relationships between nitrogen use, residue management, soil TOC and green house gas emissions . Notably, a surprising land use pattern emerged from the parametrized cultivation and management decisions. Initially, across all farms the area planted to corn increased before decreasing while the area planted to rice increased. This land use pattern indicates that a threshold of rice conversion must of first been reached before rice expansion became notable. This is likely because rice expansion can only occur near existing rice production based on access to water. Therefore, early ri ce expansion feedbacks through the TeleABM and facilitates further rice expansion clearly indicating a network effect. The shift from corn expansion early in the model to rice expansion in the second half of the model can be seen in the simulated greenhous e gas emissions. Where corn cultivation contributes more heavily to CO 2 emissions and rice production to PM 2.5 emission. Lastly, the area planted to soybeans and the average soybean yield experienced a consistent decline throughout the model run. Soybean p roduction does not benefit from nitrogen fertilizer as do rice and corn cultivation but would be subject to the same level of soil carbon loss as the cells that grew corn and rice. Therefore, farmers could compensate for low soil carbon by adding nitrogen fertilizer to corn and rice but not soybeans. This mechanism lead to a negative feedback in farmer - decision making and may partially explain the decline in soybean production. Future Work Sensitivity analysis (SA) and uncertainty (UA) analysis are important steps to complete an ABM. SA is used to evaluate the variability of model outputs due to uncertainty in input variables. UA is done before SA to quantify the outcome variability of in model inputs 13 . These next steps will be used to determine which model inputs are most responsible for outcome results. To do so, seven potenti ally influential variables have been selected (Table 11). These variables will change 110 through multiple model runs while the rest of the model parameters are held constant . This allows us to assess the influence of the potentially influential variables on m odel outcomes. 111 REFERENCES 112 REFERENCES 1. An, L., et al., Explroing Complexity in a Huamn - Environment System: An agent - based spaitial model for multidiciplinary and mulitscale integration. Annals of the Association of American Geogrpahers 2005. 95 (1): p. 54 - 79. 2. Matthews, R.B., et al., Agent - based land - use models: a review of applications Landscape Ecology 2007. 22 : p. 1447 - 1459. 3. Dou, Y., et al., Land - Use Changes in Distant Places: Imp lementation of a Telecoupled Agent - Based Model. Journal of Artificial Societies and Social Simulation, 2020. 23 . 4. Dou, Y., et al., Land - use changes across distant places: design of a telecoupled agent - based model. Journal of Land Use Science, 2019. 14 (3) : p. 191 - 209. 5. Sun, J., Y. Tong, and J. Liu, Telecoupled land - use changes in distant countries. Journal of Integrative Agriculture, 2017. 16 (2): p. 368 - 376. 6. Tong, Y., et al., Mollisols Regions. Scientific Reports, 2017. 7 (1): p. 18064. 7. Sun, J., et al., Spatiotemporal patterns of non - genetically modified crops in the era of expansion of genetically modified food. Scientific Reports, 2015. 5 : p. 14180. 8. Gadde, B., et al., Air polluta nt emissions from rice straw open field burning in India, Thailand and the Philippines. Environmental Pollution, 2009. 157 (5): p. 1554 - 1558. 9. Jin, Q., et al., Dynamics of major air pollutants from crop residue burning in mainland China, 2000 2014. Journa l of Environmental Sciences, 2018. 70 : p. 190 - 205. 10. Li, X., et al., Particulate and Trace Gas Emissions from Open Burning of Wheat Straw and Corn Stover in China. Environmental Science & Technology, 2007. 41 (17): p. 6052 - 6058. 11. Ni, H., et al., Emission characteristics of carbonaceous particles and trace gases from open burning of crop residues in China. Atmospheric Environment, 2015. 123 : p. 399 - 406. 12. Ni, H., et al., PM2.5 emissions and source profiles from open burning of cro p residues. Atmospheric Environment, 2017. 169 : p. 229 - 237. 13. Ligmann - Zielinska, A., et al., Using uncertainty and sensitvity analyses in socioecological agent - based models to improve their analytical performance and policy relevance. Plos one 2014. 9 (10 ). 14. Parker, D.C., A. Hessl, and S.C. David, Complexity, land - use modeling, and the human dimension: Fundamental challenges for mapping unknown outcome spaces. Geoforum 2007. 29 : p. 789 - 804. 113 15. Willock, J., et al., Farmers' attitudes, objectives, behavi ors, and personailty traits: The Edinburgh study of decision making on farms Journal of Vocational Behavior 1999. 54 : p. 5 - 36. 16. Ohlmer, B., K. Olson, and B. Brehmer, Understanding farmers' decsion making processes and improving managerial assistance Agr icutlural Economics 1998. 18 : p. 273 - 290. 17. Richter - Heitmann, T., et al., Evaluation of Strategies tp Seperate Root - associated Microbial Communities: A cruical choice in rhizobiome research. Frontier in Microbiology, 2016. 7 . 114 SYNTHES IS This dissertation is part of a Belmont Forum - and NSF - funded Telecoupling Consortium, which brings scholars from four continents (Brazil, China, UK and US) together to study the dynamics of international soybean production and trade using the telecoupling framework. This d issertation contributes to the larger NSF project by integrat ing network analysis, genomic sequencing and agent - based modeling to explore the environmental and socioeconomic impact of international trade in the receiving country. This dissertation was grou nded in the telecoupling framework and quantitatively addresses all the components (e.g., systems, flows, causes, effects and agents). Furthermore, this is a cross - scale study addressing the impacts of international soybean trade at the macro, meso and mic ro - scale . For chapter 2, the influence of on trade networks among soybean - producing countries was analyzed . Using 15 years of global trade data, autoregressive integrated moving average models (ARIMA) were developed to measure demand on production and bilateral trade of corn, soybeans and wheat. The results reveal that - driven soybean production in Brazil and displaced production of corn and wheat to nearby count ries (e.g., Argentina, Paraguay and Uruguay). Additionally, when restrictions were placed on the direct export of soybeans to China, Argentina and Paraguay increased soybean exports to Brazil as an indirect export route to China. The results were published in Sustainability and presented at the Global Land Programme meeting in Taiwan where it also won the Best Paper award. The research results hold implications for the true extent of production driven by distant demands as well as the impact of the geo - po litics on trade networks. Chapter 3 analyzed how soybean imports to China compete with domestic production and have altered domestic cultivation and management practices. Satellite - imagery and farmer interviews were used to quantify the regional cultivati on shift from soybean to corn production. 115 To determine if the change in cultivation was influenced by international soybean trade an interrupted comparative time - series (ICTS) model was used. Interestingly, farmers in the region are causing air pollution t o spill across the China - Russia border by using fire to manage the residual biomass that is left after harvesting corn. This was confirmed by measuring the amount of carbon in the soils of farmers and level of particulate matter (PM 2.5) in the air. Last, a multivariate regression determined which factors significantly influenced residue management decisions. The results have been presented at the AAAS Annual Meeting and submitted to Science Advances . This research highlights the impact of distant trade on production in the often overlooked importing country. Chapter 4 of my dissertation furthered the above analysis by considering the impacts of farmer cultivation and management decisions on soil prope rties. Soil texture, pH, total organic carbon and 16s rRNA sequencing were used in combination with detailed farmer management surveys to understand how changes in residue management effect efficiency, productivity, profitability and sustainability of the system. The results indicated that the accumulation of residual corn biomass has increased the use of residue fires and decreased the amount of crop residue being returned to the soil. However, crop choice, planting history and nitrogen use has more notice able effects on the microbial community than the amount of residue returned. Samples from continuous soybeans fields were the most diverse followed by mixed rotations and then continuous corn samples. The results of chapter are under internal review at MSU and plan to be published due to the uniqueness of the dataset. The culminating chapter of this dissertation uses agent - based modeling (ABM) to integrate the above chapters into a TeleABM. The teleABM models land use change in Brazil and China based on glo bal soybean demand. Land - use change decisions are made by farmer agents which 116 have parametrized using the farmer interviews. Next, the farmer agent cultivation and management decisions have environmental impacts that were determined by analyzing the soil s amples under the context of management decisions. Finally, production and the impact of farmer agent decisions The results of chapter 5 highlight both a positive a nd negative feedback. First, rice expansion started off slow but quickly increased once a threshold of rice production had been met and allowed for conversion on adjacent land. Second, soybean production decreased in response to soybean yield declines tha Future work is needed to determine which model inputs most effect the model outcome .