IMPACTS OF DISTANT DRIVERS ON LANDSCAPES AND BIODIVERSITY By Ciara Layne Hovis A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife – Doctor of Philosophy 2022 ABSTRACT IMPACTS OF DISTANT DRIVERS ON LANDSCAPES AND BIODIVERSITY By Ciara Layne Hovis Global biodiversity is increasingly impacted by distant drivers. With societies more connected than ever before, natural resource consumption has expanded beyond administrative and political boundaries. International food trade in particular has profound impacts on land-use and socioeconomic and environmental outcomes. At the same time, global biodiversity is threatened at an unprecedented scale, with many of the causes obfuscated by complexities of distant, interacting socioecological systems. Understanding the ultimate drivers of biodiversity change and translating them to local biodiversity outcomes is integral to addressing conservation challenges in the age of globalization. This dissertation analyzes the impacts of international trade on biodiversity in an agroecosystem undergoing land-use change driven by global markets. Chapter 1 provides background on the study region, Heilongjiang Province, and describes disruption of soybean production in the area due to changes in global trade. Chapter 2 is a systematic review of studies on distant drivers of biodiversity change. Across all taxa, harmful impacts on biodiversity were the most frequent outcome reported, with distant impacts of trade and tourism most frequently studied. In Chapter 3, satellite imagery was classified into landcover classes to create high-fidelity maps of the agriculture-dominated study landscape. By utilizing phenological, synthetic aperture radar, and vegetation/soil index data, accuracies of 91%- 80% were achieved. In Chapter 4 these landcover maps were used to calculate landscape metrics. These metrics were then used to analyze relationships between landscape structure (i.e., composition and configuration) and bird communities. Functional biodiversity indices derived from life history and morphological traits were examined in addition to taxonomic measures. Though no discernable differences between taxonomic and functional community metrics were observed, several significant relationships between landscape structure and biodiversity metrics were found. Crop diversity, natural landcover, and edge metrics, were positively correlated with bird richness. Aggregation of patches, corn area, and soybean area were negatively correlated. We also compared landscape structure and biodiversity between two regions impacted by global soybean trade. Despite the more impacted region having lower crop diversity and natural area, there was no difference in biodiversity between the two regions. The more impacted region also had more rice area, demonstrating that negative biodiversity impacts may be mitigated by rice cultivation. Chapter 5 built on the previous chapter by modeling bird occupancy to assess species-specific relationships with landscape structure. Results indicated that increased crop diversity significantly increased occupancy of birds at both the taxonomic and functional level, particularly for birds belonging to less common functional groups. Percentage of natural area was not as important as expected, while metrics related to landscape configuration had very few significant impacts on occupancy. Increases in rice area were not as detrimental to bird occupancy as increases in corn and soybean. In fact, soybean area exhibited more significant negative relationships with bird occurrence than corn, suggesting that decreases in soybean area due to global trade may have benefitted bird biodiversity in the case of a monocultural landscape. However, due to the prevalence of small-scale farming practices, the more likely outcome would be a decrease in crop diversity due to soybean fields being converted to more profitable crops (e.g., corn, rice). By linking global trade, changes in landcover/use, landscape structure, and local bird communities in the same context, the results of this dissertation highlight the need for integrated biodiversity studies that place ecosystems in the broader context of globalization. This dissertation is dedicated to the common and unexciting species of birds that have adapted to living in human-modified landscapes and contribute to the health of the ecosystems we all rely on. I’ll still refer to you as ‘trash birds’, but it is from a place of love and respect (and a force of habit). I hope you continue to fill everyday life with color and song for generations to come. iv ACKNOWLEDGEMENTS This dissertation would not have been completed without the support, wisdom, and kindness I have received from those around me. I would first like to thank my advisor, Dr. Jianguo Liu, for teaching me how to think about ecological problems through the lens of coupled human and natural systems as well as giving me the freedom to design and conduct a study on a topic I am truly passionate about. I would also like to thank my other committee members, Dr. Gary Roloff, Dr. Kyla Dahlin, and Dr. Doug Landis, for their guidance, patience, and invaluable insight throughout these past few years. The contributions of my entire committee have helped me grow as a scientist, and their lessons will continue to impact me throughout the rest of my career. I have also been on the receiving end of financial support from a number of funding sources, which made this work possible. These include the National Science Foundation (Complex Dynamics of Telecoupled Human and Natural Systems, grant number 1518518), the MSU Enrichment Fellowship, the Taylor International fellowship, the Vera M. Wallach Memorial Scholarship, the MSU College of Agriculture and Natural Resources summer fellowship, the MSU Wallace Scholarship, and the MSU dissertation completion fellowship. I also owe so much to my colleagues and collaborators, both at Michigan State and in China. I’d like to thank my fellow members of the Center for Systems Integration and Sustainability (CSIS) for providing incredible academic discourse as well as a pleasant and friendly working environment. I’d especially like to thank James Vatter for always solving my computer crises and Sue Nichols for helping me communicate the value of my work. This work would not have been possible without the collaboration of scholars from the Chinese Academy of Agricultural Sciences. In particular, I’d like to thank Dr. Wenbin Wu for v his support of this work. I’d also like to thank the other students and researchers I worked with in Heilongjiang and Beijing. They endured my terrible Mandarin and welcomed me into their circles. I am especially grateful to my two field assistants, Ge Shen and Yuqiao Long, and our drivers for tolerating the extremely early mornings necessary for bird data collection. I also need to thank my lab mates for their comradery and support over the past six years. I especially want to thank Kelly Kapsar, for her willingness to listen and commiserate as we worked towards finishing our degrees during a global pandemic. Last, but certainly not least, I need to thank Dr. Anna Herzberger for her friendship since my first day as a grad student, when I hopped on a plane to China with zero understanding of what I was getting myself into. I don’t think I would have finished this degree without her words of encouragement and unwavering belief in my abilities. I want to thank the other amazing friends I have in my life. To Emily, Kiersten, and Carly, thank you for all the laughs, cries, and reunions of the past six years. I also want to thank my best friend, Brandon Fuhs, for the past 15 years of continued friendship. I’ll forever be impressed that we managed to grow from small town, dorky kids into somewhat well-adjusted and well-traveled adults. I am also incredibly grateful for unwavering support of my family. I would first like to express my gratitude to my soon-to-be in-laws for their kindness and generosity, as well as their restraint exhibited during our visits by not asking me when I’d finally be graduating. I also want to thank my sister, Savanna Hovis, for her ability to listen as well as make me laugh. I owe an unending gratitude to my parents. To my mother, Candee Morris, thank you for always being there, for instilling the value of education in me, and for your example of always living with an open mind. To my father, Joseph Hovis, thank you for the informal ecological vi education you gave me throughout my childhood, for picking up the phone when I needed someone to talk birds with, and for enabling my book addiction as both a child and an adult. To you both, I would never be where I am today without the sacrifices you made and your desire to provide Savanna and I with a better life. I hope you are proud of the person I have become. Finally, I need to thank my partner in life, Kelsen Case. Listing all the ways in which I received support from you would be longer than this entire dissertation, so I will restrain myself. Thank you for the past eight years and for persevering with me during the long-distance of the first five. Your endless patience, love, and sense of humor have been a constant source of joy, especially during the more difficult times. A bonus thank you to my furry companions that I’ve had throughout this process. In particular, to the late and great Catticus, who I adopted in my second month of grad school to ease the loneliness and uncertainty during that time in my life. Thanks buddy, wish you were here to see me graduate and to immediately lay on my diploma the second I turn around. vii TABLE OF CONTENTS LIST OF TABLES .......................................................................................................................... x LIST OF FIGURES ..................................................................................................................... xiii CHAPTER 1: INTRODUCTION ................................................................................................... 1 1.1 Overview .......................................................................................................................... 2 1.2 Global Soybean Trade ...................................................................................................... 3 1.3 Agroecosystems ............................................................................................................... 5 1.4 Conceptual Framework .................................................................................................... 6 1.5 Study System .................................................................................................................... 8 1.6 Dissertation Objectives .................................................................................................. 10 CHAPTER 2: COMPLEX IMPACTS OF TELECOUPLING ON GLOBAL BIODIVERSITY 11 Abstract ..................................................................................................................................... 12 2.1 Introduction ......................................................................................................................... 13 2.2 Methods............................................................................................................................... 16 2.2.1 Literature Search .......................................................................................................... 16 2.2.2 Inclusion Criteria ......................................................................................................... 17 2.2.3 Data extraction and analysis ........................................................................................ 18 2.3 Results ................................................................................................................................. 18 2.3.1 Impacts across taxa ...................................................................................................... 19 2.3.2 Impacts across telecoupling types ................................................................................ 24 2.3.3 Impacts across regions ................................................................................................. 25 2.4 Research Challenges & Recommendations ........................................................................ 27 2.4.1 Mechanisms behind impacts of telecoupling on biodiversity ...................................... 28 2.4.2 Explicitly identify and define potential telecouplings ................................................. 29 2.4.3 Connect environmental change to biodiversity indicators ........................................... 30 2.4.4 Apply the telecoupling framework to understudied taxa and geographic regions....... 30 2.5 Conclusions ......................................................................................................................... 31 APPENDIX ............................................................................................................................... 33 CHAPTER 3: LANDCOVER CLASSIFICATION OF A HETEROGENEOUS AGROECOSYSTEM ................................................................................................................... 42 Abstract ..................................................................................................................................... 43 3.1 Introduction ......................................................................................................................... 44 3.1.1 Background .................................................................................................................. 44 3.1.2 Objectives .................................................................................................................... 47 3.2 Methods............................................................................................................................... 49 3.2.1 Study System ............................................................................................................... 50 3.2.2 Ground Truth Data Collection ..................................................................................... 51 3.2.3 Imagery Acquisition & Processing .............................................................................. 52 3.2.4 Landcover Classification ............................................................................................. 56 3.3 Results & Discussion .......................................................................................................... 57 viii 3.3.1 Accuracy Assessment .................................................................................................. 57 3.3.2 Variable Importance..................................................................................................... 63 3.4 Conclusions ......................................................................................................................... 67 CHAPTER 4: LANDSCAPE STRUCTURE AND BIRD DIVERSITY IN AGROECOSYSTEMS OF NORTHEAST CHINA..................................................................... 68 Abstract ..................................................................................................................................... 69 4.1 Introduction ......................................................................................................................... 70 4.1.1 Background .................................................................................................................. 70 4.1.2 Research Questions ...................................................................................................... 71 4.2 Methods............................................................................................................................... 72 4.2.1 Bird Data Collection .................................................................................................... 72 4.2.2 Biodiversity Indices ..................................................................................................... 72 4.2.3 Landscape Metrics ....................................................................................................... 76 4.2.4 Prefecture Case Study .................................................................................................. 78 4.3 Results ................................................................................................................................. 81 4.3.1 Biodiversity results ...................................................................................................... 81 4.3.2 Landscape metric results .............................................................................................. 87 4.3.3 Bird Diversity & Landscape Relationship ................................................................... 90 4.3.4 Community Weighted Means and Landscape ........................................................... 100 4.4 Discussion ......................................................................................................................... 107 4.5 Conclusions ....................................................................................................................... 111 APPENDIX ............................................................................................................................. 113 CHAPTER 5: MODELING BIRD OCCUPANCY IN A CHANGING AGROECOSYSTEM 122 Abstract ................................................................................................................................... 123 5.1 Introduction ....................................................................................................................... 124 5.1.1 Background ................................................................................................................ 124 5.1.2 Research Goal ............................................................................................................ 126 5.2 Methods............................................................................................................................. 126 5.2.1 Bird Data .................................................................................................................... 126 5.2.2 Landscape Metrics ..................................................................................................... 127 5.2.3 Occurrence Model ...................................................................................................... 128 5.3 Results ............................................................................................................................... 130 5.3.1 Model data ................................................................................................................. 130 5.3.2 Taxonomic model results ........................................................................................... 135 5.3.3 Functional model results ............................................................................................ 146 5.4 Discussion & Conclusion.................................................................................................. 156 APPENDIX ............................................................................................................................. 160 CHAPTER 6: SYNTHESIS ........................................................................................................ 163 REFERENCES ........................................................................................................................... 167 ix LIST OF TABLES Table 1: Examples of biodiversity impacts present in our sample and their potential telecoupled mechanisms ................................................................................................................................... 28 Table 2: Codebook ........................................................................................................................ 36 Table 3: Taxa proportion at individual indicator level and article level (n = number of studies) 40 Table 4: Percentage of telecoupling impacts across taxa (n = number of studies) ....................... 41 Table 5: Percentage of telecoupling impacts across telecoupling types (n = number of studies) 41 Table 6: Percentage of telecoupling impacts across habitats (n = number of studies) ................. 41 Table 7: Percentage of telecoupling impacts across regions (n = number of studies) .................. 41 Table 8: Class definitions and ground truth totals ........................................................................ 52 Table 9: Total number of images in for each collection (2017 samples, 2018 samples, and ground truth samples) that were used for the image mosaicking process. ................................................ 53 Table 10: Data layers used for classification ................................................................................ 54 Table 11: Parameters for simple non-iterative clustering algorithm in GEE................................ 56 Table 12: Parameters for Random Forest Classifier in Google Earth Engine .............................. 57 Table 13: Accuracy assessment for training data: producer accuracy (PA), user accuracy (UA) 59 Table 14: Training confusion matrix ............................................................................................ 59 Table 15: Testing confusion matrix .............................................................................................. 59 Table 16: Merged classification schemes ..................................................................................... 60 Table 17: Accuracy assessment for training and testing data of rain-fed crop reclassification .... 61 Table 18: Accuracy assessment for training and testing data of reclass reclassification.............. 61 Table 19: Rain-fed crop training confusion matrix....................................................................... 61 Table 20: Rain-fed crop testing confusion matrix ........................................................................ 61 Table 21: Reclass training confusion matrix ................................................................................ 62 Table 22: Reclass testing confusion matrix .................................................................................. 62 x Table 23: F-scores for each class .................................................................................................. 62 Table 24: Data layer ranked by variable importance .................................................................... 65 Table 25: Summary of trait frequency for families (n=30) .......................................................... 75 Table 26: Landscape metric definitions, input, equations, and descriptions (Hesselbarth et al., 2019; Laliberté & Legendre, 2010). The ‘reclassified’ landcover input merged similar landcover classes so that there were only 3 classes contained in the raster (built-up, natural, and cropland). ....................................................................................................................................................... 77 Table 27: Family membership to functional groups and occurrence for all sampled prefectures in Heilongjiang, Jiamusi, and Qiqihar (OW=old world). ................................................................. 82 Table 28: Most frequent communities. ........................................................................................ 84 Table 29: Summary table of the richness metrics calculated at the point level for all Heilongjiang prefectures (n=188), Jiamusi (n=53), and Qiqihar (n=55) ...................................... 84 Table 30: Summary table of the abundance-weighted biodiversity metrics averaged across all intervals (n=564) and at the point-level for all samples (n=188). ................................................ 86 Table 31: Summary table of the abundance-weighted biodiversity metrics at the point-level for Jiamusi (n=53) and Qiqihar (n=55). ............................................................................................. 86 Table 32: Summary table of all landscape metrics at the small (250m), medium (500m), and large (1000m) extents. Bold values indicate significant differences between Jiamusi and Qiqihar (p-value < 0.1)............................................................................................................................... 88 Table 33: Table of foraging behavior definitions (M.-B. Lee & Martin, 2017; Remsen & Robinson, 1990; Ruxton & Hansell, 2011) ................................................................................. 114 Table 34: Trait matrix for families observed in study area (n=30). ............................................ 115 Table 35: Pearson’s correlation values between CWM for mass and habitat preference and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. ........................................... 118 Table 36: Pearson’s correlation values between CWM for diet preference and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. ................................................................... 119 Table 37: Pearson’s correlation values between CWM for foraging behavior and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. ................................................................... 120 xi Table 38: Pearson’s correlation values between CWM for the ground nesting trait/migration status and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. .................... 121 Table 39: Landscape metrics selected for model. The ‘Reclass’ input refers the landcover raster that merged crop classes (soybean, corn, rice, and other) into one ‘cropland’ class and the natural classes (forest, grassland, wetland, and water) into one ‘natural’ class. ..................................... 128 Table 40: Frequency of bird order and family in sample............................................................ 131 Table 41: Functional group composition and species frequency ................................................ 132 Table 42: Functional group frequency and group aliases. .......................................................... 133 Table 43: Data variation for each metric across all extents (250m, 500m, and 1000m). .......... 133 Table 44: Estimates of hyperparameters for the taxonomic occurrence model .......................... 135 Table 45: Estimates of hyperparameters for the functional occurrence model .......................... 146 xii LIST OF FIGURES Figure 1: Chinese soybean imports from 1960-2020 (FAOSTAT) ................................................ 4 Figure 2: Conceptual Framework Diagram of Telecoupling (Taken from Liu et al. 2013) ........... 7 Figure 3: Map of Heilongjiang Province (left), Google Earth Image showing village and cropland juxtaposition (top right), and example of typical farmland (bottom right) ..................................... 9 Figure 4: Overview of telecoupling types (bold) with examples. ................................................. 15 Figure 5: Reported impacts of telecouplings on biodiversity for each taxon in the examined articles. Each pie chart shows the relative proportion of impacts found across articles for each taxon and telecoupling type (Figure 2.1). Beneficial, harmful, and changed impacts indicate a significant relationship between biodiversity and telecoupling was found. Unclear impacts indicate non-significant, neutral, or disagreeing results within a study and mixed impacts mean an article had equal numbers of indicators in one or more categories (e.g., increased abundance of one species, decreased for another). The background of the pie chart indicates the number of articles in a given category (white to black). Articles that considered multiple telecoupling types were counted for each type reported. ............................................................................................ 21 Figure 6: Reported impacts of telecouplings on biodiversity by continent. Each pie chart shows the relative proportion of impacts (i.e., beneficial, harmful, changed, mixed, unclear) found across articles for each continent and telecoupling type (Figure 4). The background of the pie chart and the color of each continent indicates the number of articles in each category (white to black). Articles that consider multiple telecoupling types are replicated for each type’s pie chart. ....................................................................................................................................................... 26 Figure 7: The impact of different telecoupling types on biodiversity across taxa (n=788). For articles that considered multiple telecoupling types, their relationships appear in >1 column. ... 39 Figure 8: Typical farmland structure in the United States (left) versus Heilongjiang (right). For comparison, the red point in the center of the image is surrounded by a buffer of 1000m. ......... 45 Figure 9: Workflow diagram for the landcover classification in Google Earth Engine. .............. 49 Figure 10: Map of Heilongjiang Province, China. ........................................................................ 50 Figure 11: Climate of Heilongjiang Province from 1970-2000 (mean temperature in dark blue, min/max in light blue, and average precipitation in orange. ......................................................... 51 Figure 12: Map of Heilongjiang and location of field ground truth points .................................. 52 Figure 13: Example result from the simple non-iterative clustering algorithm ............................ 55 Figure 14: Bar graph of data layer importance for random forest classification .......................... 64 xiii Figure 15: Phenological difference between two of the most important data layers, VH (a-c) and NDTI (d-f) for all three time periods compared to the final classification (g). ............................ 66 Figure 16: Map of Heilongjiang and location of sites where bird biodiversity was sampled. Counties in the Qiqihar (west) and Jiamusi (east) prefectures are outlined in black. ................... 72 Figure 17: Comparison of the three extents chosen for landscape metric calculation at the same site ................................................................................................................................................. 76 Figure 18: Example landscape input for crop diversity. All non-crop classes (i.e., built-up, forest, grassland, wetland, and water were masked (NA). ....................................................................... 77 Figure 19: Soybean area decline for Jiamusi and Qiqihar prefectures. The orange lines highlight the difference in soybean area between 2000 and 2018. .............................................................. 79 Figure 20: Soybean area in 2000 (a,d), 2009 (b,e), and 2018 (c,f) for Qiqihar (a-c) and Jiamusi (d-f). Data from X. Li et al. 2021. Bird/landscape samples are depicted by purple points. ......... 80 Figure 21: Cluster dendrogram of families based on functional traits.......................................... 81 Figure 22: Histograms of the three richness metrics with the mean depicted by the red vertical line: number of families (n.sp) and functional group richness (FGR) .......................................... 85 Figure 23: Map of richness metrics: number of families (n.sp) and unctional group richness (FGR). Counties in the Qiqihar (west) and Jiamusi (east) prefectures are outlined in black. ...... 85 Figure 24: Histograms of the abundance-weighted biodiversity metrics at the point-level with the mean depicted by the red vertical line: Shannon-Weiner index (shannon) and functional dispersion (FDis). .......................................................................................................................... 87 Figure 25: Map of abundance-weighted biodiversity indices at the point-level: Shannon-Weiner index (shannon) and functional dispersion (FDis). Note that the gray points represent missing values (where 0 species are present for FDis. Counties in the Qiqihar (west) and Jiamusi (east) prefectures are outlined in black. .................................................................................................. 87 Figure 26: Histograms of percent soybean (p_soy) for Jiamusi (top) and Qiqihar (bottom). ...... 89 Figure 27: Correlations between taxonomic richness (n.sp) and functional richness (function group richness). The size of each sign corresponds to the number of extents (250m, 500m 1000m) that were significantly positive (+) or negative (-). Metrics with no significant relationships are denoted by the null sign (∅)............................................................................... 91 Figure 28: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with crop diversity (left) and percent natural landcover (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. ............................................................................................................................................... 92 xiv Figure 29: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with total edge (top left), edge density (top right) and aggregation index (bottom). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. ....................................................................................................................... 93 Figure 30: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with percent soybean (left) and percent corn (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot.................... 94 Figure 31: Correlations between Shannon’s diversity index (taxonomic) and functional dispersal (functional). The size of each sign corresponds to the number of extents (250m, 500m 1000m) that were significantly positive (+) or negative (-). Metrics with no significant relationships are denoted by the null sign (∅). ......................................................................................................... 96 Figure 32: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with crop diversity (left) and percent natural landcover (right). Pearson’s correlation (R) and p- value (p) for each relationship is also included in the top left corner of each plot. ...................... 97 Figure 33: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with total edge (top left), edge density (top right) and aggregation index (bottom). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. ............................................................................................................................................... 98 Figure 34: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with percent soybean (left) and percent corn (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. ........................................... 99 Figure 35: Linear relationships of crop diversity at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means.................................................. 101 Figure 36: Linear relationships of rice paddy percentage at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means. ......................................... 104 Figure 37: Linear relationships of soybean percentage at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means.................................................. 105 Figure 38: Linear relationships of corn percentage at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means.................................................. 106 Figure 39: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with mean patch size (left) and percent rice (right). Pearson’s correlation (R) and p- value (p) for each relationship is also included in the top left corner of each plot. .................... 116 Figure 40: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with mean patch size (left) and percent rice (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. ......................................... 117 Figure 41: Correlogram for metrics selected as model predictors. ............................................. 134 xv Figure 42: Number of bird families with significant and nearly significant coefficient estimate means (green, blue = positive, red, orange = negative) for each landscape metric: crop diversity (crop div), natural area (nat), rice area (rice), mean patch size (mn patch), aggregation of patches (ai), corn area (corn), and soybean area (soy). ............................................................................ 136 Figure 43: Dot plot showing the family-specific mean estimates with 90% Bayesian credible intervals for crop diversity. Five families had positive estimates with BCIs that did not cross 0 (Corvidae, Ardeidae, Anatidae, Rallidae, Paridae) ..................................................................... 137 Figure 44: Dot plots showing the family-specific mean estimates with 90% credible intervals for mean patch area and aggregation index. ..................................................................................... 138 Figure 45: Dot plots showing the family-specific mean estimates with 90% Bayesian credible intervals for percent area natural landcover (forest, wetland, grassland), soybean, corn, and rice paddy. .......................................................................................................................................... 140 Figure 46: Dot plots showing the family-specific mean estimates with 90% credible intervals for longitude and minutes after sunrise, which were used as covariates in the occurrence and detection models respectively. .................................................................................................... 142 Figure 47: Dot plots showing the family-specific mean estimates with 90% credible intervals for occurrence and detection............................................................................................................. 143 Figure 48: All landscape metric relationships with occurrence probability for the most abundant families (>30 observations)......................................................................................................... 145 Figure 49: Number of bird groups with significant and nearly significant coefficient estimate means (green, blue = positive, red, orange = negative) for each landscape metric: crop diversity (crop div), natural area (nat), rice area (rice), mean patch size (mn patch), aggregation of patches (ai), corn area (corn), and soybean area (soy). ............................................................................ 147 Figure 50: Dot plot showing the group-specific mean estimates with 90% Bayesian credible intervals for crop diversity. Six groups had positive estimates with BCIs that did not cross 0 (crow/magpie, shorebird/lapwing, rail, pipit/wagtail, hoopoe/starling, kingfisher) ................... 148 Figure 51: Dot plots showing the group-specific mean estimates with 90% credible intervals for mean patch area and aggregation index. ..................................................................................... 149 Figure 52: Dot plots showing the group-specific mean estimates with 90% Bayesian credible intervals for percent area natural landcover (forest, wetland, grassland), soybean, corn, and rice paddy. .......................................................................................................................................... 151 Figure 53: Dot plots showing the group-specific mean estimates with 90% credible intervals for longitude and minutes after sunrise, which were used as covariates in the occurrence and detection models respectively. .................................................................................................... 152 Figure 54: Dot plots showing the group-specific mean estimates with 90% credible intervals for occurrence and detection............................................................................................................. 153 xvi Figure 55: All landscape metric relationships with occurrence probability for the most abundant groups (>30 observations)........................................................................................................... 155 xvii CHAPTER 1: INTRODUCTION 1 1.1 Overview As the world continues to develop, boundaries between natural and human systems become less clear and more complicated (J. Liu et al., 2013). Today’s global trade market blurs these boundaries even further by connecting distant, coupled human and natural systems (CHANS) all around the globe. This presents new challenges for sustainability, as many countries now consume resources far beyond their borders, with consequences felt not by the consumers but the unseen ecosystems and individuals that provide these resources (Chaudhary & Kastner, 2016; Prell et al., 2017; Weinzettel et al., 2013; Xu et al., 2020). Concurrently, the human population is growing exponentially, thereby exacerbating resource demand. As of 2019, the world population was approximately 7.56 billion (U.S. Census Bureau, 2019). As many countries find their domestic resources inadequate to meet basic human needs, such as food, they turn to global trade (Odorico et al., 2014). Additionally, types of commodities demanded by countries is shifting to more animal-based products as development results in more affluent populations (Robinson & Pozzi, 2011). The combination of these factors has led to sky- rocketing demands in the food trade which subsequently impacts the CHANS that produce these highly traded commodities (Tilman et al., 2011). Importing countries often obtain environmental benefits (e.g., increased land allocation for biodiversity conservation and restoration rather than food production) (De Fraiture et al., 2004; Lenzen et al., 2012; Nimubona, 2012; Walz & Wellisch, 1997), whereas exporting countries experience environmental degradation (e.g., from converting land for food production) (Lambin & Meyfroidt, 2011; Lenzen et al., 2012). However, little empirical evidence on the environmental impacts of global trade in importing countries is available (but see Sun et al. 2018) as most research has been conducted on exporting countries (J. Sun et al., 2018). The 2 assumption that importing countries always receive environmental benefits from global trade must be further tested (J. Liu, 2020), for it is crucial to identify environmental costs and benefits borne by both importing and exporting countries to secure global environmental sustainability and food security (Tanentzap et al., 2015). 1.2 Global Soybean Trade One of the most globally traded crops is soybean (Glycine max). The soybean trade network is among the largest in the world (Leff et al., 2004). The three largest nodes within this network are the United States (US), Brazil, and China (Reenberg & Fenger, 2011; Schaffer-Smith et al., 2018). Soybeans are mainly used for human consumption, livestock feed, and industrial purposes. Soybeans are also valued for their ability to fix atmospheric nitrogen and are commonly grown in rotation with other crops to maintain soil fertility (Hymowitz, T., Newell, 1981). Soybeans are one of the only commodity crops to significantly increase in acreage, increasing 45% from 1980-2009 (Reenberg & Fenger, 2011). For comparison, corn acreage increased by 21% during the same time period. Acceleration of the soybean trade is unprecedented in terms of scale and consequences of this rapid growth are still unclear. Despite domesticating the soybean over 3000 years ago (Hymowitz, 1970), China is now the world’s largest importer of soybeans; importing more than 84 million tons of soybeans in 2015 or 64% of total global soybean imports (Figure 1) (FAOSTAT Database, 2017). The rise in soybean imports to China can be attributed to two major factors: reduction of the soybean tariff in 2001, when China joined the World Trade Organization (L. Li et al., 2017); and for livestock feed to meet the rising demand for animal products as China’s population becomes more affluent (Gale, 2015). The majority of China’s soybean imports are from Brazil and the US where climate, vast field sizes, and genetically modified varieties lower cost of production significantly 3 (Yao et al., 2018). The majority of domestically grown soybeans, however, are grown in northeast China where growing seasons are short and farms are small, with non-contiguous fields of various crops (L. Li et al., 2017). Additionally, cultivation of genetically modified soybeans (and other crops) is explicitly outlawed by the Chinese government (Reenberg & Fenger, 2011). Figure 1: Chinese soybean imports from 1960-2020 (FAOSTAT) In addition to socioeconomic effects, soybean production for global trade has numerous environmental impacts. Research identified threats in Brazil to both the Cerrado (savannah) and Amazonian forests as well as spillover feedbacks driving cattle expansion (Fearnside, 2002; le Polain de Waroux et al., 2017; Vanwey et al., 2013). However, less research has been conducted on environmental effects in soybean importing countries. The assumption of increased environmental benefits may not hold true in China due to the disruption of its domestic soybean production and the subsequent impact on the agroecosystem. 4 1.3 Agroecosystems Agricultural landscapes are oftentimes overlooked with regards to biodiversity conservation. This is largely due to the patch-matrix paradigm that has dominated the field of landscape ecology for the greater part of the field’s history (Lausch et al., 2015). Briefly, the patch-matrix model is that habitat is fragmented by disturbance, usually human driven, which breaks up the habitable areas into patches, which are then separated by unsuitable areas (i.e. the matrix, often human-dominated lands) (Rutledge, 2003). The patch-matrix model was conceptualized with the inhospitable (to terrestrial animals) ocean as the matrix separating hospitable islands (MacArthur & Wilson, 1967). However, terrestrial landscapes oftentimes offer a more subtle gradient between hospitable and inhospitable habitats. There is a growing body of work demonstrating that matrix habitats can play important roles in how species utilize and move about landscapes, and this literature calls for more comprehensive approaches to landscape characterization (Fahrig, 2001; Prugh et al., 2008). Recently, landscape ecologists have espoused treating the “matrix”, or non-habitat areas of the landscape, as part of the functioning ecosystem landscape mosaic, rather than disregarding it as barren a wasteland (Fahrig et al., 2011; Vasseur et al., 2013). Agricultural land is a prime example of a heavily human-modified landscape. Many natural elements persist, but humans manage a majority of the inputs and outputs as well as configuration and composition. Using the PMM framework, farmland would be treated as matrix for wildlife, while considering uncultivated forest or grassland areas as habitat patches. However, agricultural lands can house a multitude of species assemblages. The question remains, what kind of agricultural landscapes offer the greatest habitat potential? There is consensus that landscape heterogeneity positively contributes to biodiversity, however, widespread 5 intensification of farming operations has reduced heterogeneity of farmlands (i.e. larger fields, more agrichemicals, bigger machinery, mono-cropping etc.) (Benton et al., 2003). The fields of agroecology and landscape ecology are simultaneously moving towards functional views of ecosystem and biodiversity characteristics. Functional views consider the role an organism or habitat plays within the broader ecosystem. By utilizing mainstay landscape pattern and composition metrics, relationships between landscape, species diversity, and ecosystem services can be quantified to aid in comprehending mechanistic patterns and processes of human-modified landscapes (Fahrig et al., 2011). These concepts have been applied to both theoretical and empirical situations (Gil-Tena et al., 2015; Vasseur et al., 2013). Vasseur et al.’s (2013) approach demonstrated how the oftentimes hidden landscape mosaic (i.e., temporal changes or chemical inputs) affects arthropod habitat suitability in farmland. Gil-Tena et al. (2015) evaluated bird occurrence and landscape heterogeneity within and among landscape patches, and focused on functional biodiversity rather than taxonomic biodiversity (e.g. insectivorous, seed eating, raptors etc.) (Gil-Tena et al., 2015). Both studies found that higher heterogeneity across the landscape resulted in higher biodiversity opportunity (Gil-Tena et al., 2015; Vasseur et al., 2013). 1.4 Conceptual Framework In this dissertation, I utilized the telecoupling framework to guide my research questions, as the concept is well suited to understanding and quantifying complex interactions between coupled human and natural systems (J. Liu et al., 2013). The framework distinguishes sending, receiving, and spillover systems with defined boundaries as well as relevant environmental and socioeconomic causes, effects and agents (Figure 2). Systems are connected by flows which can be tangible (e.g., people, materials, organisms) or intangible (e.g., knowledge, technology, 6 capital). The role a system plays (i.e., sending, receiving, or spillover) depends on the flow of interest (e.g., soybeans) and direction(s) of said flow. Sending systems have outward direction of flow (e.g., origin, source, exporter) while receiving systems flow inward (e.g., destination, recipient, importer). Spillover systems affect or are being affected by flows between sending and receiving systems (e.g., stopover, third party). Agents are defined as decision-makers in a system and can have either direct or indirect impacts on flows. Causes and effects of a system are often linked via feedbacks, and they are defined as components that influence emergence/dynamics of telecouplings and as consequences of telecouplings, respectively. Figure 2: Conceptual Framework Diagram of Telecoupling (Taken from Liu et al. 2013) In the context of this dissertation, we focus on flow of soybeans from Brazil to China. On that basis, Brazil as the sending system and China is the receiving system. and the United States is the spillover system. For our purposes, the US is defined as a spillover system as it is a major competitor to Brazil in the soybean market and exports a large share of its soybeans to China, however it is not the focus of this specific study. Cause and effect linked to soybean price in China (i.e., low priced Brazil soybeans outcompete domestically grown soybeans in China) and 7 land use change (i.e., farmers in China shift crop types because they cannot compete with soybeans), respectively. A multitude of other cause and effects exist for this system, but this dissertation centers on soybean price and land use change. In this system, farmers are a primary agent as they make decisions on growing soybeans based on market prices. Land use change causes a secondary effect on biodiversity. The magnitude and significance of relationships between crop price, farmers, land use, and biodiversity are the focus of this dissertation. 1.5 Study System The study site is Heilongjiang province in northeast China (Figure 3). Heilongjiang’s biome is temperate broadleaf/mixed forests with a climate characterized by long, harsh winters and short, mild summers which allow for a single-crop growing season (Olson et al., 2001; Weather and Climate, 2016). Heilongjiang is the top agrarian producer in China. Until recently, the main crop was soybean, accounting for up to one third of total national production (Survey Office of the National Bureau of Statistics in Heilongjiang, 2013). Soybeans (Glycine max) are a nitrogen- fixing legume belonging to Fabaceae that typically grow to be ~50cm on average. The other two main crops are corn and paddy rice. Corn (Zea mays; Poaceae) is a tall cereal grass (2-3m) that produces 1-3 ears of corn per plant. Rice (Oryza sativa; Poaceae) is also a cereal grass traditionally grown in paddies that are flooded with 10cm or more of water throughout their growing cycle to limit competition from weeds. In recent years, this region has undergone significant changes in land use/cover change (LULCC) as influence of global trade grows in the Chinese agriculture market. Due to cheap soybean imports from abroad, Heilongjiang’s small-scale farmers struggle to sell their soybean harvests for profit. From 2009-2013, soybean cultivation in Heilongjiang dropped from 4 million ha to 2.4 million ha as farmers switched to growing more profitable crops such as corn and 8 paddy rice (J. Sun et al., 2015; Survey Office of the National Bureau of Statistics in Heilongjiang, 2013). Figure 3: Map of Heilongjiang Province (left), Google Earth Image showing village and cropland juxtaposition (top right), and example of typical farmland (bottom right) As previously stated, Heilongjiang’s primary industry is agriculture therefore the majority of land is dedicated to cultivation. The only remnants of primary forest, once home to Siberian tigers, exist in the northwest of the province where the main industry is forestry (Muldavin, 1997). Despite being dominated by farmland, Heilongjiang is a unique agroecosystem, dominated by small-scale farmers. These farmers tend to manage 3-10 ha and are quick to respond to changing crop markets. The landscape is typified by villages scattered across the landscape, with fields of varying crop type and size in the surrounding vicinity (Figure 3). This type of agricultural landscape creates a diverse crop mosaic landscape, which can host a number of species that support overall biodiversity of the region (Y. Liu et al., 2013). Under current trends of crop conversion, it is not clear what impacts on biodiversity will be. 9 1.6 Dissertation Objectives Chapter 2 synthesizes empirical evidence from scientific publications over the past decade that measured impacts of different telecoupling flows on biodiversity. Chapter 3 describes the process of producing high fidelity landcover maps of the study region using satellite imagery and machine learning. Chapter 4 evaluates relationships between structure (i.e., composition and configuration) of an agricultural landscape and taxonomic and functional biodiversity for birds at the community- level. Chapter 5 evaluates species-specific relationships between agricultural landscape structure and bird occupancy at taxonomic and functional levels. 10 CHAPTER 2: COMPLEX IMPACTS OF TELECOUPLING ON GLOBAL BIODIVERSITY In collaboration with Veronica Frans, Kelly Kapsar, Aurora Torres, Yuqian Zhang, Min-Gon Chung, Yingjie Li, Ruishan Chen, Emily Dean, Anna Herzberger, Ming Lei, Xutong Wu, Ying Xing, Jianguo Liu 11 Abstract As the world becomes more interconnected, underlying causes of the global biodiversity crisis often extend beyond local systems. Human-nature interactions over distances (i.e., telecouplings, which expand and integrate disciplinary concepts such as trade and tourism) are increasing in magnitude and complexity, yet knowledge about their impacts on biodiversity is scattered and fragmented. Here, we synthesized empirical evidence from scientific publications in the past decade that specifically measured impacts of telecouplings on biodiversity. Impacts reported as significant were categorized as “beneficial”, “harmful”, or “changed” based on direction of the biodiversity-telecoupling relationship (e.g., increased tourist disturbance linked to decreased species richness). Results indicated that harmful impacts were the most frequent outcome across all taxa (51%), though there were also beneficial impacts (11%). Among the 10 types of telecouplings, the most common telecouplings connected to observed impacts were tourism (46%) and trade (26%). We also found few studies occurring in Africa and Oceania, despite being the locations of many important biodiversity hotspots. Our study provides the first integrated assessment of evidence for impacts of telecouplings on biodiversity, and sheds light on how telecoupling-mediated mechanisms have complex cascading consequences for biodiversity. Further efforts to study telecoupling impacts on biodiversity will be critical for addressing the biodiversity crisis post-2020. 12 2.1 Introduction Global biodiversity is threatened at an unprecedented scale and addressing the causes of these threats is essential to slow or reverse biodiversity declines. Modern species extinctions are occurring 1,000 times more frequently than historical background extinction rates (Díaz et al., 2019; S. L. Pimm et al., 2014) with species declines extending across taxa (Alroy, 2015; Arthington et al., 2016; Bek et al., 2017; Böhm et al., 2013; Ceballos et al., 2015; Darwall & Freyhof, 2015; S. Pimm et al., 2006; Sodhi et al., 2008; van Klink et al., 2020; Wake & Vredenburg, 2008). Approximately one million species of plants and animals are currently facing extinction due to human activity (Díaz et al., 2019). Understanding the underlying mechanisms that cause biodiversity loss is a key priority for assuring and enhancing the wellbeing of both humans and nature (Hooper et al., 2012; Mace et al., 2012; Oliver et al., 2015). Ultimate causes of biodiversity loss may originate locally and/or emerge from distant regions. Traditionally, research has focused on local causes of biodiversity loss, such as deforestation and habitat fragmentation. However, globalization has drastically altered the use of natural resources as well as patterns of resource consumption (Chaudhary & Kastner, 2016; Dorninger et al., 2021). As the world becomes more interconnected, resources and goods are increasingly consumed outside production systems. For example, local farming in many places is predominantly driven by international demand for commodity crops rather than local food requirements (Headey, 2011). Identifying only the local causes of biodiversity loss will oftentimes not provide enough information to discern or address the mechanisms ultimately driving reductions in biodiversity (Carrasco et al., 2017). Understanding the impacts of globalization on biodiversity requires an integrated approach that includes both local and distant interactions between humans and nature. The telecoupling framework is such an integrated tool (J. Liu et al., 2013). The term ‘telecoupling’ refers to human-nature interactions over distances (J. Liu et al., 2013). It expands 13 and integrates disciplinary concepts such as trade, tourism, migration, and investment (Figure 4). It encompasses local and distant systems as well as underlying mechanisms driving biodiversity impacts. In the context of this paper, we consider the location experiencing biodiversity impacts as the ‘local’ system. Telecoupling types are categorized by the content of the flows transferred between systems (Figure 4). Many different types of telecouplings have been associated with threats to biodiversity. In rural areas, for example, waste transfer in the form of agricultural run- off results in distance effects on distant and biodiverse regions, such as the “dead zone” in the Gulf of Mexico (Del Giudice et al., 2020; Osterman et al., 2009). Similarly, wild species in southeast Asia are increasingly threatened by the palm oil trade as their habitats are destroyed and converted to plantations to meet global demand (Fitzherbert et al., 2008). The role of telecouplings in global biodiversity changes is increasingly recognized (Chan et al., 2020) and has been highlighted in important international venues and publications (e.g., 2021 Nobel Prize Summit and 2019 Global Assessment Report on Biodiversity and Ecosystem Services sponsored by the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services). However, knowledge about the impacts of telecoupling on biodiversity is fragmented and scattered, and research synthesizing impacts is lacking. 14 Figure 4: Overview of telecoupling types (bold) with examples. A global overview assessing the relationships between telecoupling and biodiversity is urgently needed. Previous studies have examined the influence of single telecouplings (e.g., trade (Green et al., 2019); tourism (Brandt & Buckley, 2018) on biodiversity or on specific taxa (e.g., birds (Marques et al., 2019)), yet no studies have synthesized the influence of telecouplings across taxa, telecoupling types, and geographic regions. Given the potentially unique responses of specific taxa to different telecouplings, understanding these relationships is crucial to more effectively conserve biodiversity (J. Liu, 2021; J. Liu et al., 2016). This issue is particularly salient and timely as scientists and stakeholders around the world are developing the Post-2020 Biodiversity Framework (Convention on Biological Diversity, 2020), which will serve as a global biodiversity conservation blueprint for the next decade. Establishing the degree to which telecouplings impact biodiversity across taxa, telecoupling types, and geographic regions is essential to inform the next era of conservation in a globalized world. To address this need, we conducted a global synthesis of telecoupling impacts on biodiversity. Our review assessed empirical studies from 2010 to 2019 that quantified the 15 impacts of telecouplings on biodiversity, defined here as the variety of organisms across levels of biological organization. Of the 7,306 unique articles resulting from our initial search, 131 articles (1.8%) reported biodiversity indicators and their relationship with telecouplings. From this synthesis, we assessed the impacts of telecouplings on biodiversity and highlighted knowledge gaps to inform future research and conservation efforts. 2.2 Methods We synthesized studies that quantified biodiversity change in the context of telecoupling published between 2010 and 2019. We curated studies for our synthesis by first conducting a literature search, then implementing multiple rounds of abstract and full text screenings to check that each study met our inclusion criteria. Once the initial screenings were complete, each article underwent a full-text assessment to extract data on the type of biodiversity and telecoupling studied, the study location, the metrics researchers reported, direction of the biodiversity impact, and significance of the impact. Overall impact was then determined for each article by taking the mode of all reported impact directions (e.g., beneficial, harmful) reported for each taxon. Each step of our research process – literature search, inclusion criteria, data extraction and analysis – is detailed below. 2.2.1 Literature Search We implemented a systematic review by adapting the workflow of PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) (Moher et al., 2010; Sato et al., 2013). To synthesize impacts of telecouplings on biodiversity using information from peer- reviewed literature, we focused on research articles published within the past decade (2010- 2019) and written in English. The literature search was conducted with Web of Science on December 20, 2019, using terms pertaining to four primary categories: biodiversity, impact, scale, and telecoupling. We used several keywords for each primary category. For example, 16 terms such as “impact”, “effect”, “consequence”, and “influence” were included for the impact category, and “trade”, “migration”, “technology transfer” were for the telecoupling category (Hull & Liu, 2018). We finalized the search string based on different combinations of keywords, while excluding potentially ambiguous ones (see Box 1 in appendix for search string). 2.2.2 Inclusion Criteria To only include articles relevant to the research objectives, we conducted two rounds of screening (title/abstract screening and full-text assessment) based on specific inclusion criteria we developed (Box 2 in appendix). We only included articles that met all five of the following criteria: (1) measured impacts on biodiversity of animal and plant species (excluding other taxa such as fungi, bacteria); (2) investigated the study species in relation to a particular telecoupling process; (3) reported direct outcomes in the form of a quantitative relationship, (4) were observational or experimental studies; and (5) were written in English. Subsequently, synthesis efforts, such as meta-analyses and literature reviews, fell outside the scope of this review. For example, a review by Scheele et al. provides strong evidence of the negative impacts of disease spread (chytridiomycosis) facilitated by humans but was not included in our synthesis due to the difficulty of identifying specific telecoupling processes related to species decline and the fact that many of the articles used to inform their meta-analysis were published prior to 2010 (Scheele et al., 2019). For the first round of screening, we checked only the title and abstract for these criteria using the metagear package in R (Lajeunesse, 2016). Articles that were immediately identified as missing one or more of our criteria were removed prior to the full-text assessment. If it was unclear whether an article met the five criteria from the abstract, a full-text assessment was conducted to provide more context prior to deciding on inclusion or exclusion. 17 2.2.3 Data extraction and analysis Once all articles were assessed for inclusion/exclusion criteria, each eligible paper was coded to extract information on study characteristics (e.g., telecoupling types, region; see Table 2 in appendix for codebook) and each biodiversity value reported (e.g., metric type, impact, significance; see Table 2 in appendix for codebook). To enhance inter-coder reliability (i.e., the concordance rate of individuals coding the same articles), all coders were initially trained by reading common abstracts and then having group discussions to determine an article’s eligibility based on inclusion and exclusion criteria. Group decisions were made on articles with unclear eligibility; three additional coders independently assessed them, and a majority rule was used for the article’s eligibility decision. After all eligible articles were coded, we documented number of biodiversity values reported for a given taxa (e.g., occurrence values for multiple species, composition changes for multiple community structures) and their relationship with the telecoupling studied (i.e., impact). Impacts reported as significant were categorized as “beneficial”, “harmful”, or “changed” based on direction of the biodiversity-telecoupling relationship. Statistically insignificant relationships, neutral relationships, or cases of model disagreement were categorized as “unclear” impacts. The category with the greatest number of relationships determined overall impact for a given article. Articles with equal numbers in one or more categories were considered ‘mixed’. All data were synthesized and visualized using R, version 4.1.0716 . 2.3 Results Overall, telecouplings were more frequently associated with statistically significant harmful (51%) rather than beneficial (11%) impacts on biodiversity, and several impacts showed statistically significant changes in species communities (6%). There were, however, articles where telecoupling impacts on biodiversity were not straightforward, with 25% of articles 18 resulting in ‘unclear’ (non-significant, neutral, disagreement) and 7% ‘mixed’ impacts (articles with equal numbers in one or more categories of impacts). The scale of the studies in our review ranged from local landscapes (Amira et al., 2018) to global assessments (Nishijima et al., 2016). We found biodiversity impacts driven by a wide array of complex mechanisms that ultimately affected many indicators of biodiversity (see Table 2 in appendix for list of indicator types). Some examples of impacted biodiversity indicators included changes in distribution (e.g., avoidance of recreation infrastructure in habitat (Coppes et al., 2017), density (e.g., trampling disturbance due to tourists (Sarmento & Santos, 2012), abundance (e.g., declining populations due to trapping for pet trade (Harris et al., 2017), and productivity (e.g., altered species interactions via alien species introduction (Nobre et al., 2019). Across the 131 articles, we recorded 788 cases of biodiversity indicators impacted by telecoupling (Figure 7, Table 3 in appendix). 2.3.1 Impacts across taxa Harmful impacts on biodiversity were more frequent than beneficial impacts across all taxa (Table 4 in appendix). Mammals had the largest proportion of harmful impacts reported (80%), followed by invertebrates (48%), birds and plants (both 46%), and fish (40%). We found 4 studies recorded for reptiles and amphibians with harmful impacts (3 and 1 respectively). Similarly, most (75%) studies using indicators combining multiple taxa found harmful impacts. Birds were the most studied taxon (28%), followed by invertebrates (24%), vascular plants (21%), mammals (11%), and fish (11%; Table 3 in appendix). A much smaller set of articles studied reptiles and amphibians (4% and 1% of articles, respectively). Only four articles (3%) analyzed telecoupling impacts on more than one taxon simultaneously. Our assessment also revealed that there were relatively more beneficial impacts on birds and plants, potentially because conservation initiatives usually contribute to creation or bolstering of habitats at smaller 19 scales, rather than those with large or restrictive habitat requirements (Balduino et al., 2017; Dale et al., 2017; Sanderson et al., 2016). Birds experienced more harmful (46%) and unclear (22%) impacts than beneficial (16%) (Figure 5, Table 4 in appendix). Harmful impacts to birds were associated with trade (9 articles; e.g., impact of trade agreements and farming intensification (Reif & Vermouzek, 2019)) in addition to knowledge transfer and human migration. Tourism was also associated with harmful impacts to birds (7 articles; e.g., effects of road or trail use (Wolf et al., 2013)). Beneficial impacts for birds were mainly associated with knowledge transfer (50% of beneficial bird impacts [3 articles]; e.g., biodiversity-friendly farming practices (Martín et al., 2012)), tourism (e.g., disturbed park areas had higher abundances for some species (Huhta & Sulkava, 2014)), and species dispersal (e.g., non-native shrub offered more food resources for woodland bird species (Arizaga et al., 2013)). Instances of trade interacting with other telecouplings also resulted in beneficial impacts: knowledge transfer (conservation policy informed by prices of crops for trade (Hyun et al., 2013)) and energy transfer (forest fragmentation for transferring natural gas increased abundances of some bird species (Farwell et al., 2016)). 20 Figure 5: Reported impacts of telecouplings on biodiversity for each taxon in the examined articles. Each pie chart shows the relative proportion of impacts found across articles for each taxon and telecoupling type (Figure 2.1). Beneficial, harmful, and changed impacts indicate a significant relationship between biodiversity and telecoupling was found. Unclear impacts indicate non-significant, neutral, or disagreeing results within a study and mixed impacts mean an article had equal numbers of indicators in one or more categories (e.g., increased abundance of one species, decreased for another). The background of the pie chart indicates the number of articles in a given category (white to black). Articles that considered multiple telecoupling types were counted for each type reported. Invertebrates were found to be most associated with harmful (48%) and unclear (39%) impacts, with only one study demonstrating beneficial impacts of telecouplings (Figure 5, Table 4 in appendix). Studies reporting harmful impacts were most often related to tourism (11 articles; e.g., impact of trampling on reef communities (Santos et al., 2015)), as well as species dispersal (e.g., release of popular aquarium corals that outcompete native species (Mantelatto et al., 2018)) and trade (e.g., decreased soil fauna diversity in tropical forest converted to rubber plantations 21 for trade (Singh et al., 2019)). The single instance of beneficial impacts was related to species dispersal (e.g., benthic community benefitting from the introduction of a non-native fish that altered the functioning of the lakes’ trophic and increased productivity (Nobre et al., 2019)). Plant biodiversity was associated more with harmful (46%) and unclear (29%) impacts than beneficial (14%) (Figure 5, Table 4 in appendix). Tourism was frequently associated with harmful impacts to plant biodiversity (7 plant articles, e.g., plant community disturbance due to tourist activity (Šilc et al., 2016)). In addition to tourism, plants were negatively impacted by telecouplings related to species dispersal (2 plant articles, non-native plants inhibiting native species diversity (Fukami et al., 2013)) as well as knowledge transfer (dissemination of information regarding mismatch between habitat needed and habitat protected (Assédé et al., 2018)), trade (e.g., community market access and local plant richness (Oldekop et al., 2018)), water transfer and waste transfer (run-off contaminants inhibit growth of saltmarsh plant species and allow for the encroachment of exotic and mangrove species (Geedicke et al., 2018)). For plants, most beneficial associations occurred with tourism (e.g., intermediate disturbance caused by tourists (Czortek et al., 2018)), investment, and knowledge transfer (policy and incentives to start game farms decreased vegetation loss (Mokotjomela & Nombewu, 2020)) with one instance related to trade (e.g., wood pellet trade resulted in improved overall forest condition (Dale et al., 2017)). For mammals, the majority of telecoupling impacts were harmful (80%), with remaining impacts being 7% beneficial and 13% unclear (Figure 5, Table 4 in appendix). Harmful impacts on mammal biodiversity were associated with tourism (e.g., tourist presence deterring species with low tolerance for human activity (Zhou et al., 2013)), trade (e.g., impact of commercial livestock on local species (Moraga et al., 2015)), human migration (e.g., increased hunting resulting from migrant populations increasing demand for bushmeat (Remis & Jost Robinson, 22 2012)), and energy transfer (e.g., effect of oil-spills on marine mammals (Balmer et al., 2018)). The only instance of beneficial impacts was related to tourism (more visitors increased wildlife protections resulting in higher elephant numbers (Selier et al., 2016)). Fish experienced mostly harmful impacts (40%) with some instances of significant community changes (20%) and even fewer beneficial outcomes (13%) (Figure 5, Table 4 in appendix). Harmful impacts on fish biodiversity were the most frequent (40%) and associated with trade (3 articles; e.g., unmanaged harvest for the aquarium trade(Santos et al. 2015)) as well as species dispersal (e.g., introduced aquaculture species led to a reduction of native species (Cuvin-Aralar, 2016)), energy transfer, and water transfer (e.g., cascaded dams and reservoirs impeded movement of eggs, larvae and young fish more than a single dam (Yang et al., 2017)). Most studies reporting significant changes to fish communities were associated with hydroelectric dams (i.e., water and energy transfer) and physical barriers to movement they created (Lima et al., 2018; Sá-Oliveira et al., 2015). Additionally, one study demonstrated significant changes to fish behavior due to tourist activity (Lima et al., 2014). Beneficial impacts on fish biodiversity were recorded for tourism (e.g., tourist feeding fish resulted in high fish abundance (Balduino et al., 2017)), water transfer, and investment (e.g., investment in restoration efforts for dam removal increased number of lamprey nests (Lasne et al., 2015)). Although reptiles and amphibians were studied in relatively fewer articles, some significant impacts of telecoupling were found (Figure 5, Table 4 in appendix). Impacts on reptile biodiversity were either harmful (60%; 3 articles) or unclear (40%; 2 articles). The three cases of harmful impacts were associated with species dispersal (e.g., spread of parasite from invasive turtle species to native species (Meyer et al., 2015)), tourism (e.g., tourism development resulted in reptile populations (Krawczyk et al., 2019)), and trade (e.g., population declines due to the pet trade (Flecks et al., 2012)). For amphibians, while we had a small sample size that 23 prevented us from assessing impacts adequately, one study related trade to harmful impacts on amphibian communities (e.g., community market access and frog richness (Oldekop et al., 2018)). 2.3.2 Impacts across telecoupling types Most articles investigated changes in biodiversity associated with impacts of tourism (46%) and trade (26%). Species dispersal and knowledge transfer were the next most common telecouplings studied, representing 13% and 10% of articles, respectively. The remaining telecouplings (energy transfer, investment, human migration, water transfer, and waste transfer) were each studied in <10 articles (Table 5 in appendix). No studies looked at the impact of technology transfers. Some articles studied impacts of multiple, interacting telecouplings on biodiversity (18%). In these cases, we attributed the impacts to all telecoupling types mentioned in the article. In general, biodiversity was negatively impacted by telecouplings. Over half of the articles for trade and tourism showed harmful impacts. Trade was reported as harmful in 64% of the articles (e.g., biodiversity loss due to human consumption of traded goods (Wilting et al., 2017)). For tourism, 53% were categorized as harmful (e.g., human trampling at tourist destinations (Farris et al., 2013)). Species dispersal, the third most frequent telecoupling studied, was associated with harmful impacts in 44% of the articles (e.g., impacts of invasive species on local species communities (Van Bocxlaer & Albrecht, 2015)). Some types of telecouplings had relatively more beneficial impacts on biodiversity. Compared to other telecoupling types, investment and knowledge transfer had higher proportions of beneficial impacts on biodiversity (43% and 21% respectively; Table 5 in appendix). Beneficial impacts on biodiversity were mostly linked with policy initiatives committed to halting biodiversity loss by driving conservation investments (Waldron et al., 2017), knowledge 24 transfer to promote ecological restoration efforts (Sieges et al., 2014), transitions from intensive production to ecological production methods (e.g., agro-environmental schemes (Gamero et al., 2017)), and the creation of protected areas (Kleijn et al., 2014). In some articles, we also observed mixed impacts (i.e., equal contributions of beneficial, harmful, changed, or unclear impacts). For example, tourism was sometimes beneficial (e.g., income from tourism funding conservation efforts (Selier et al., 2016)) or harmful (e.g., tourism disturbance (Coppes et al., 2017)). Differential impacts may also reflect different tolerance levels among species (Shochat et al., 2010), with sensitive species being lost (e.g., Bronze‐tailed Peacock‐pheasant [Polyplectron chalcurum] or capercaillie [Tetrao urogallus] avoiding roads and other infrastructure (Coppes et al., 2017; Harris et al., 2017)), less sensitive species resisting, and some non-native, introduced species thriving (e.g., the highly invasive snail, Melanoides tuberculata, in Lake Malawi (Van Bocxlaer & Albrecht, 2015)). Thus, depending on the relative proportion of declining, resisting, and thriving species within a focal system, telecouplings may change local species assemblages (Reyes-Martínez et al., 2015), leading to biotic homogenization or a loss of community uniqueness (Thomas et al., 2014). Likewise, conservation policies involving knowledge transfer and investment can also result in mixed impacts. For example, a review of the European Union’s agriculture policy showed that despite the intended beneficial biodiversity impacts of farmland conservation incentives, harmful impacts also occurred due to agricultural intensification encouraged by policymakers (Gamero et al., 2017). 2.3.3 Impacts across regions There was a bias in distribution of locations for studies regarding telecoupling impacts on biodiversity. Studies most commonly occurred in Brazil (15%), the United States (10%), and China (8%). However, when aggregated to the continental scale, Europe (23%) was the most 25 frequent study continent. The remaining articles were distributed across South America (20%), Asia (19%), North America (17%), Africa (8%), Oceania (8%), and Antarctica (2%), with global studies representing 4% of all the articles. The relative scarcity of studies in Africa and Oceania is concerning, as they are home to many of the world’s most valuable (and threatened) biodiversity hotspots (Habel et al., 2019). The distribution of studies across habitats was also biased. Most articles investigated impacts on biodiversity in terrestrial habitats (57%), whereas 19% and 17% of articles examined impacts on biodiversity in marine and freshwater habitats, respectively (Table 6 in appendix). The direction of telecoupling impacts on biodiversity also varied geographically. Harmful impacts were most frequent across all continents, apart from North America, where 50% of the impacts were unclear. Europe had the largest number of beneficial impacts compared to other Figure 6: Reported impacts of telecouplings on biodiversity by continent. Each pie chart shows the relative proportion of impacts (i.e., beneficial, harmful, changed, mixed, unclear) found across articles for each continent and telecoupling type (Figure 4). The background of the pie chart and the color of each continent indicates the number of articles in each category (white to black). Articles that consider multiple telecoupling types are replicated for each type’s pie chart. 26 continents. Of the five global studies, four reported harmful impacts of telecoupling while the remaining one demonstrated beneficial impacts (Figure 6, Table 7 in appendix). The types of telecouplings studied in the context of biodiversity impacts differed by continent as well. Tourism studies occurred mostly in Europe, Asia, and South America and trade studies were mainly concentrated in Asia and South America. Trade and tourism were also common telecouplings in North America, in addition to species dispersal, knowledge transfer, and energy transfers. For Africa, tourism, trade, species dispersal, and knowledge transfer were equally studied, whereas studies in Oceania focused on tourism and trade. The two studies that looked at telecoupling impacts on biodiversity in Antarctica were related to tourism and human migration. Finally, studies conducted on a global scale mostly focused on trade with one study each related to investment and human migration. Across all continents, investment, human migration, and waste transfer were the least studied (Figure 6, Table 7 in appendix). 2.4 Research Challenges & Recommendations Telecouplings are increasing in magnitude and scope as the world becomes more interconnected and populations grow (Borrelli et al., 2020; Powers & Jetz, 2019; Stehfest et al., 2019). They will likely continue to have significant impacts on global biodiversity. In our analysis, we found abundant evidence of telecouplings driving biodiversity change. However, the percentage of studies that explicitly acknowledged, identified, and quantified relevant telecouplings and their impacts on biodiversity was small in the literature (1.8%). Thus, more extensive and empirical studies on telecoupling impacts and mechanisms are urgently needed. Additionally, there is a need for studies that link environmental changes driven by telecoupling to specific indicators of biodiversity. Finally, of the studies that considered telecouplings as a driver of biodiversity changes, the majority were biased towards certain taxa, locations, and metrics. 27 2.4.1 Mechanisms behind impacts of telecoupling on biodiversity This study provides evidence demonstrating the profound impacts of telecoupling on biodiversity. However, the specific mechanisms that led to these impacts were less studied and thus are less clear. Here, we define mechanisms as the processes driving biodiversity impacts involved in a particular flow of materials, energy, people, products, capital, information, and organisms. Other important components of the telecoupling framework are factors influencing emergence and dynamics of flows (causes), socioeconomic and environmental consequences of the telecoupling (effects, which includes cascading impacts on biodiversity), and the relevant systems (local and distant) (Table 1). Table 1: Examples of biodiversity impacts present in our sample and their potential telecoupled mechanisms Telecoupling Species Knowledge Energy Trade Tourism type Dispersal Transfer Transfer Distant Food importing Country with Native range International Energy System countries eco-tourists of introduced regulators company species Local Food exporting Country with Region species Nation under Location of System country eco-tourism introduced to regulation energy attractions extraction Causes High global Eco-tourism Economic Alternative Increased food demand industry need in local production energy demand established community encouraged by regulators Effects Deforestation Tourist Non-native Widespread Contamination for more spending species use of practice event leading to cropland and reinforcing introduced for improves habitat habitat loss wildlife production, habitat quality degradation protections altered species community Biodiversity Increased Increased Decreased Increased Displacement of Impact extinction risk wildlife native species wildlife wildlife abundance dominance populations populations (Green et al., (Selier et al., (Cuvin-Aralar, (Martín et al., (Balmer et al., 2019) 2016) 2016) 2012) 2018) More studies should connect observed changes in biodiversity to both local and distant drivers so that the mechanisms can be better understood. To illustrate the full application of the 28 telecoupling framework, we outlined potential mechanisms of telecoupling impacts on biodiversity in five articles with examples for each component (Table 1). In general, mechanisms vary greatly across telecoupling types. For example, agricultural trade caused by international demand can result in harmful biodiversity impacts by increasing the need for land conversion and thus habitat loss in the local system. In the case of tourism, biodiversity often benefits from the economic incentives to preserve wildlife. It should be noted that, due to the inherent complexity of telecouplings, accurately identifying and quantifying each component of the telecoupling framework can be challenging. Thus, after an initial mechanism is established, further investigations should take place to validate the proposed model and uncover new research avenues. This holistic approach is necessary to fully understand biodiversity dynamics. 2.4.2 Explicitly identify and define potential telecouplings Many of the studies we found did not directly examine telecoupling impacts, with most telecouplings being briefly mentioned or inferred by us. For example, we classified studies involving production of globally traded commodities in the local system as influenced by trade, even if authors did not directly identify the regions that ultimately received the commodities. Explicit links between observed biodiversity changes due to telecouplings are essential to understand the mechanisms and achieve more effective management and sustainability interventions with less trade-offs. For instance, policies to mitigate deforestation in the Amazon for soybean production and trade have resulted in increased deforestation and threats to biodiversity in another biodiversity hotspot, the Brazilian Cerrado region (Dou et al., 2018). Findings from research like this can help avoid similar spillover effects on biodiversity in other systems. Whenever possible and appropriate, researchers should identify all relevant human- nature interactions in their local system and those relevant distant systems (Table 1). Doing so 29 ensures that all important drivers of change are accounted for and may also lead to unexpected avenues for further investigation. 2.4.3 Connect environmental change to biodiversity indicators There is a pressing need to bridge the gap between studies connecting telecouplings to environmental change and those quantifying biodiversity’s relationship to environmental variables. Many studies on telecouplings quantified direct and local environmental impacts, but rarely included biodiversity indicators in their assessment. Research that simultaneously measures telecouplings’ impact on the environment and biodiversity should be a priority for future research. To overcome logistical challenges of such an undertaking, researchers can also use existing studies, datasets, and/or models to link telecouplings with biodiversity. For example, researchers have successfully combined socioeconomic models for material flow and economic trade with biodiversity models to quantify extinction risk for species of concern in the Brazilian Cerrado (Green et al., 2019). Such innovative approaches to telecoupling and biodiversity research can play a large role in future biodiversity conservation. 2.4.4 Apply the telecoupling framework to understudied taxa and geographic regions More studies need to be conducted on those taxa and geographic regions underrepresented in the current literature. The scientific literature assessing telecoupling impacts on biodiversity focused on certain locations, taxa, and biodiversity metrics. Most articles focused on birds, plants, and invertebrates in terrestrial habitats. Amphibians, reptiles, fishes, and mammals were comparatively under-studied, which may have implications for understanding threatened species and conservation prioritization. Moreover, only four studies used biodiversity indicators including more than one taxon (Marquardt et al., 2019; Seshadri & Ganesh, 2011; Waldron et al., 2017; Wilting et al., 2017). Bridging this research gap with multi-taxa studies may facilitate more comprehensive conservation policies to limit biodiversity loss. Additionally, 30 many studies were concentrated in Brazil, the United States, and China. In contrast, studies rarely occurred in Africa, the Middle East, and South and Southeast Asia. The disproportionately low number of scientific studies on these regions is concerning, as they are forecasted to experience increased rates of development in the coming decades (Bren d’Amour et al., 2017). This limited number of studies could lead policy makers to underappreciate the importance of telecouplings in those regions. Finally, our study revealed differential use of various biodiversity metrics. Most articles reported compositional and structural components of biodiversity, using metrics such as species presence or absence, richness, and abundance. In contrast, impacts on movement, population dynamics, or functional metrics were rarely assessed. Assessments of biodiversity should be conducted using a wide variety of indicators to better understand telecoupling impacts on ecological processes that in turn shape biodiversity. Studies addressing one or more of these gaps will be essential for informing biodiversity conservation. 2.5 Conclusions This paper synthesizes impacts of 10 different types of telecoupling on 7 major taxa across all continents. It is also timely as the world is developing the Post-2020 Global Biodiversity Framework. Impacts of telecoupling are often beyond the control of local people, highlighting greater challenges in biodiversity conservation and the increasing need for global collective actions. Slowing the trajectory of biodiversity loss requires an integrative approach to conservation. Amid the current biodiversity crisis, it is imperative that policy and intervention strategies be tailored to the ultimate causes of species loss. The causes of biodiversity decline are often intertwined with both local drivers and telecouplings. Our synthesis demonstrates the need for more research that empirically assesses biodiversity change in the context of distant system interactions. Of the studies that identified telecouplings linked to biodiversity change, impacts 31 varied widely across taxa, regions, and telecoupling types. Despite some instances where telecouplings benefitted biodiversity, it was clear that most impacts of telecouplings were overwhelmingly harmful. This further highlights the importance of understanding the dynamics and interactions of the human and natural components of telecoupled systems if leverage points for intervention are to be found. Policy aiming to reduce biodiversity declines needs to consider telecouplings in addition to local drivers. Collaboration among governments, industries, and communities in local and distant systems will be essential. Halting biodiversity loss is a global challenge that will require system-based solutions. Utilizing integrated and interdisciplinary approaches, such as the telecoupling framework, will bring us closer to these solutions and a more sustainable future in a telecoupled world. 32 APPENDIX 33 2A.1 Supplemental Methods Box 1: Search string (TS = ((Biodiversity OR "Biological diversity") AND (abundan* OR densit* OR "population size" OR "population-level" OR "species distribution" OR *diversity OR "species richness" OR "species composition" OR "species evenness" OR "number of species" OR biomass) AND (impact* OR effect$ OR affect* OR influenc*) AND (transboundary OR intercontinental OR distal OR teleconnection$ OR telecoupl* OR "far away" OR external OR trade OR touris* OR migrat* OR "information dissemination" OR "technology transfer" OR "knowledge transfer" OR "foreign investment" OR "international aid" OR "international cash transfer" OR "international remittance" OR "payment for ecosystem services" OR "virtual land" OR "virtual water" OR "virtual energy" OR "human migration" OR refugee$ OR immigra* OR “disease spread” OR “foreign direct investment” OR “labor migration”)) NOT TS=(trade-off$ OR simulation))(TS = ((Biodiversity OR "Biological diversity") AND (abundan* OR densit* OR "population size" OR "population-level" OR "species distribution" OR *diversity OR "species richness" OR "species composition" OR "species evenness" OR "number of species" OR biomass) AND (impact* OR effect$ OR affect* OR influenc*) AND (transboundary OR intercontinental OR distal OR teleconnection$ OR telecoupl* OR "far away" OR external OR trade OR touris* OR migrat* OR "information dissemination" OR "technology transfer" OR "knowledge transfer" OR "foreign investment" OR "international aid" OR "international cash transfer" OR "international remittance" OR "payment for ecosystem services" OR "virtual land" OR "virtual water" OR "virtual energy" OR "human migration" OR refugee$ OR immigra* OR “disease spread” OR “foreign direct investment” OR “labor migration”)) NOT TS=(trade-off$ OR simulation)) 34 Box 2: Inclusion and Exclusion Criteria 1. BIODIVERSITY: Does it measure impacts on biodiversity? Here, our definition of biodiversity refers to the variety of organisms across levels of biological organization. This includes any animal and plant species, at any life stage, at terrestrial, freshwater, or marine realms. EXCLUDE (irrelevant) if there is no measurement of impacts on biodiversity or impacts are assessed for other taxa (e.g., fungi, bacteria). If unclear from the information provided in the title/abstract, answer YES (full paper to be screened). INCLUDE papers that look at proxy measures (e.g., land-use change or habitat) if it is in the context of a specific species or group of species. 2. INTERVENTION: Does the study investigate the impacts of telecouplings on the study species? (i.e., Effect/Impact of ______ on biodiversity). This includes “trade, tourism, species dispersal, human migration, the spread of diseases, knowledge transfer, information dissemination, technology transfer, international investments, international cash transfer, international remittance, payments for ecosystem services, virtual land or water or energy”… or direct mention of telecoupling INCLUDE papers that have inferred telecoupings (e.g. a paper discussed local biodiversity variation due to tourism without specifying tourist source; a paper mentioned local landscape change impact on biodiversity because of agricultural food trade in general). EXCLUDE (teleconnections) papers analyzing the impacts of teleconnections (e.g., natural long-distance seed dispersal) as this revision focuses only on human- nature interactions (couplings). EXCLUDE papers focused on “ecosystem services” generally (if no biodiversity metric is mentioned). 3. OUTCOME: Does this study report direct outcomes in the form of a quantitative effect of the intervention? Studies should quantify the relationship between couplings and biodiversity responses at metacommunity, community, population or species levels. This can include population abundance, species presence/absence, species richness/abundance among other response variables. We will include all outcomes (beneficial, harmful, direct and indirect) associated with any aspect of couplings. EXCLUDE studies that report qualitative outcomes or do not report outcomes at all. 4. STUDY CLASS: Is this an observational study? This can include manipulated experiments (i.e., researcher has control over the variable of interest and the response to/ manipulating it is measured), or non-manipulative studies (sometimes called observational or natural experiments) occur when the researcher takes advantage of changes that have happened (by using existing data) or are about to happen (by taking measurements, i.e. making observations) to understand its effect. If the study is a simulation study (i.e., individual-based models or population viability analysis) = EXCLUDE (modeling study); unless the study contains potentially useful field-collected data that was used to validate the model. Also, any recent reviews that you deem to have potentially relevant references (i.e., the review is relevant but simply not primary literature = EXCLUDE (relevant review) but write down the title, to screen later at full-text; time permitting, we will go through those references to ensure we’re capturing all relevant papers. Similarly, if the item is an entire conference proceeding including multiple abstracts of potential relevance = EXCLUDE. All other situations = EXCLUDE (i.e., policy discussions, conceptual frameworks, strictly modeling, or irrelevant reviews, or the coder determined the study is not relevant at all). 5. LANGUAGE: Is the full text written in English? 35 Table 2: Codebook Variable Variable Type of Category Response Options Description Name Type Question Biodiversity Text Short - What is the specific "biodiversity" being studied in Impact answer, text this article? (e.g., Polar bear population dynamics) Telecoupling Text Short - In your own words, what is the external process that answer, text is impacting biodiversity in the focal system(s)? Quantitative Text Short - What are the specific quantitative variables that were Screening answer, text collected/analyzed in relation to the impact of the Confirmation telecoupling on the studied biodiversity metric? Study Type Text Multiple True This study is based on observations and/or choice False experiments. It is not a simulation study (e.g., climate (single forecasting) or review paper. response) Paper ID Text Short - NA Code answer, numerical Coder Name Text Short - NA answer, text Identifiers Year Continuous Short 2010-2020 NA Published (Integer) answer, numerical First Text Short - NA Author's answer, text Last Name Taxonomic Text Checkboxes Bird Check the boxes for the taxonomic groups for which Focus Mammal data were collected, analyzed, and/or synthesized Reptile during the study. Amphibian Fish Invertebrate Plant Other Biodiversity Text Multiple Abundance/Density Abundance/Density – number of individuals, Indicator choice Occurrence individuals/unit area, biomass (single (single Within-Species species) response) diversity Occurrence – presence, range, persistence, NOT Population detections (if separate from occurrence) dynamics NA Within species diversity – genetic diversity, age structure, etc. Population dynamics – survival, fitness, reproduction, mortality, etc. If indicator is at multiple or habitat level, select NA Biodiversity Biodiversity Text Multiple Diversity index Diversity index – Shannon-wiener, Simpson’s, Indicator choice Richness Inverse Simpson’s, etc. (multiple (single Evenness species) response) Composition Richness – number of species Abundance/Density Occurrence Evenness – relative abundance NA Composition – what species make up the community? Abundance/Density - number of individuals, individuals/unit area, biomass Occurrence - presence, range, persistence, NOT detection (if separate from occurrence) If indicator is at single or habitat level, select NA Biodiversity Text Multiple Amount Amount – e.g. land use change from non-habitat to Indicator choice Quality habitat (must be in the context of specific species) (habitat) (single response) Quality – pollution, connectance, disturbance, etc. If indicator is at single or multiple level, select NA 36 Table 2 (cont’d) Specific Text Short - State the specific external flow(s) that were analyzed Telecoupled answer, text in relation to their impact on focal system Flow biodiversity. Type(s) For example, if international trade data were analyzed, which commodities were used to evaluate trade? Telecoupled Text Checkboxes Trade Check the boxes for the variables that were analyzed Flow Migration (human) as they flowed between countries, multi-country Category Migration (non- regions, or continents. human) Species Dispersal Tourism Knowledge Transfer Technology Transfer Investment Telecoupling Water Transfer Variables Waste Transfer Other Impact of Text Multiple Beneficial Beneficial = Biodiversity improved as a result of the Telecoupling choice Neutral telecoupling process (single Harmful Neutral = Biodiversity did not change a result of the response) Changed telecoupling process Unclear Harmful = Biodiversity declined as a result of the telecoupling process Changed = Biodiversity changed but no determination of direction could be made (e.g., change in community composition) Impact of Text Short - If you selected "Unclear" above, please briefly state Telecoupling answer, text the situation. (Unclear) Otherwise, leave blank. Significance Text Multiple True The impact of the telecoupling process on of choice False biodiversity was statistically significant (p<= 0/5). telecoupling (single Not evaluated response) Habitat Text Checkboxes Terrestrial For descriptions, see Freshwater https://askabiologist.asu.edu/explore/biomes Marine Not specified/Global study Number of Continuous Short 1-200 State the total number of countries included in the Countries (Integer) answer, analysis (regardless of the number of focal countries Analyzed numerical listed in the "Scale of Analysis" section). Study Countries Text Short - If <10 countries were analyzed, please write the Characteristics Analyzed answer, text country names here. Otherwise, leave blank. Continent(s) Text Checkboxes North America Check boxes for all continents from which data were represented South America collected/analyzed. Asia Europe Africa Antarctica Australia Unclear/unknown 37 2A.2 Supplemental Results 2A.2.1 Total vs Overall Impacts We documented a total of 788 telecoupling impacts on biodiversity within the 131 eligible articles from our review (Figure 7). Telecouplings were more frequently associated with statistically significant harmful (30%) rather than beneficial (10%) impacts with biodiversity, and a considerable number of impacts showed statistically significant changes in species communities (13%). However, most articles (70%) reported multiple cases of biodiversity impacts (e.g., occurrence values for multiple species, composition changes for multiple community structures). Furthermore, the total number of impacts per article was highly variable (range of 1-162). Therefore, we also evaluated the overall impact (see methods for details). When aggregated to the article level, the relative frequency of harmful impacts increased from 30% to 51%, while the proportion of beneficial impacts remained almost the same (from 10% to 11%). The percentage of unclear impacts decreased from 47% to 25%, indicating that there were some articles reporting many relationships associated with unclear biodiversity impacts. When these results were aggregated into an overall impact value for each study, the relative contribution of unclear relationships decreased (Table 3). 38 Figure 7: The impact of different telecoupling types on biodiversity across taxa (n=788). For articles that considered multiple telecoupling types, their relationships appear in >1 column. 39 Table 3: Taxa proportion at individual indicator level and article level (n = number of studies) Indicators Articles Taxa (n = 788) (n = 131*) n % n % Birds 361 47.07% 37 28.24% Invertebrates 165 21.51% 31 23.66% Fish 92 11.99% 15 11.45% Plants 88 11.47 28 21.37% Mammals 55 07.17% 15 11.45% Reptiles 10 1.30% 5 3.82% Amphibians 1 0.13% 1 0.76% Multiple 16 2.09% 4 3.05% *some articles reported on more than one taxon 40 Table 4: Percentage of telecoupling impacts across taxa (n = number of studies) Taxa (n) Beneficial Harmful Changed Mixed Unclear Birds (37) 16% (6) 46% (17) 5% (2) 11% (4) 22% (8) Plants (28) 14% (4) 46% (13) 4% (1) 7% (2) 29% (8) Invertebrates (31) 3% (1) 48% (15) 6% (2) 3% (1) 39% (12) Fish (15) 13% (2) 40% (6) 20% (3) 13% (2) 13% (2) Mammals (15) 7% (1) 80% (12) - - 13% (2) Reptiles (5) - 60% (3) - - 40% (2) Amphibians (1) - - 100% (1) - - Multiple (4) 25% (1) 75% (3) - - - Table 5: Percentage of telecoupling impacts across telecoupling types (n = number of studies) Telecoupling (n) Beneficial Harmful Changed Mixed Unclear Tourism (62) 8% (5) 53% (33) 5% (3) 6% (4) 27% (17) Trade (36) 8% (3) 64% (23) 3% (1) 3% (1) 22% (8) Spp. Dispersal (18) 11% (2) 44% (8) 6% (1) 6% (1) 33% (6) Investment (7) 43% (3) - - 29% (2) 29% (2) Knowledge Transfer (14) 29% (4) 29% (4) 7% (1) 14% (2) 21% (3) Water Transfer (9) 11% (1) 33% (3) 22% (2) 22% (2) 11% (1) Energy Transfer (9) 11% (1) 33% (3) 22% (2) 11% (1) 22% (2) Table 6: Percentage of telecoupling impacts across habitats (n = number of studies) Habitat (n) Beneficial Harmful Changed Mixed Unclear Terrestrial (75) 13% (10) 55% (41) 3% (2) 7% (5) 23% (17) Marine (25) 4% (1) 56% (14) 8% (2) 4% (1) 28% (7) Freshwater (22) 14% (3) 36% (8) 14% (3) 9% (2) 27% (6) Terrestrial/Freshwater (4) - - 25% (1) - 75% (3) Terrestrial/Marine (2) - 100% (2) - - - Freshwater/Marine (1) 100% (1) - - - - Not specified /Global study (2) 100% (1) 100% (1) - - - Table 7: Percentage of telecoupling impacts across regions (n = number of studies) Region (n) Beneficial Harmful Changed Mixed Unclear Europe (31) 19% (6) 42% (13) 3% (1) 10% (3) 26% (8) Asia (27) 4% (1) 67% (18) 4% (1) 7% (2) 19% (5) S. America (28) 7% (2) 61% (17) 11% (3) 7% (2) 14% (4) N. America (22) 9% (2) 32% (7) - 9% (2) 50% (11) Oceania (11) 9% (1) 55% (6) 9% (1) - 27% (3) Africa (11) 18% (2) 46% (5) 18% (2) - 18% (2) Antarctica (2) - 50% (1) - - 50% (1) Global (5) 20% (1) 80% (4) - - - 41 CHAPTER 3: LANDCOVER CLASSIFICATION OF A HETEROGENEOUS AGROECOSYSTEM 42 Abstract The global food trade drives changes, both environmental and economical, all over the world. China is the world’s largest importer of soybeans, importing over half of the world’s traded supply annually. However, their domestic soybean farmers have struggled to compete in the global market. This has resulted in a rapid conversion of land to other crops, and the complex landscape created by small-scale farming has made characterizing and understanding this change challenging. The ability to accurately map this region is the first step to understanding the environmental and social consequences of this globally driven change. By using high-resolution imagery across different periods of the growing season and utilizing a random forest classifier with an object-based approach, we were able to generate highly accurate landcover maps (91% training, 80% testing). Furthermore, merging similar landcover classes (e.g., rain-fed crop, natural area) increased the accuracy substantially. We found that most important parameters in our analysis consisted mainly of both synthetic aperture radar (SAR) and short-wave infrared (SWIR) indices from earlier in the growing season. The resulting map from this customizable and transferable workflow will be vital for understanding the impact of landscape change on biodiversity. 43 3.1 Introduction 3.1.1 Background Agricultural landscapes can oftentimes be overlooked in terms of biodiversity conservation efforts (Norris, 2008). This is largely due to the patch-matrix paradigm dominating the field of landscape ecology for the greater part of its history (Forman & Godron, 1981; Wiens, 1976). More recently, the “matrix”, or non-habitat areas of the landscape, is increasingly being treated as part of the functioning ecosystem, rather than disregarded as a barren wasteland. This has inspired numerous theories and frameworks in the fields of agro- and landscape ecology that aim to examine the biodiversity potential of agricultural landscapes that were traditionally labeled as part of the ‘matrix’(Benton et al., 2003; Fahrig et al., 2011; Perfecto & Vandermeer, 2010). In order to quantify the effects of landscape structure (i.e., composition and configuration) on ecological processes, a variety of landscape metrics have been developed, however they require reliable landcover/use maps to produce meaningful measures. Advances in remote-sensing technology and processing software have enabled researchers to produce more detailed analyses of heterogeneous landscapes. However, those advances are now at a point where the scale of images is so fine that traditional remote sensing methods are ineffective. Now, many scientists are looking to object-based remote sensing, which is a classification method based on objects rather than pixels. With the advent of widely available satellite imagery at fine spatial resolutions, researchers have the ability to characterize heterogeneous landscapes at resolutions that could not be achieved previously (Lausch et al., 2015). For instance, a commonly used resolution is NASA’s 30m/pixel Landsat mission (USGS US Dept of the Interior, 2017). Up until recently, 44 this was considered a fine spatial resolution. However, today there are many satellites, such as Sentinel-2, offering finer resolutions (10m/pixel) (European Space Agency, 2017). This is particularly relevant for regions with high landscape heterogeneity, like small-scale agriculture (Figure 8). In this example, farmers may grow multiple crops within a 30x30m area and wildlife species often utilize even smaller areas. However, increased resolution does present a challenge during classification due to increased pixel variation (Fahrig et al., 2011). Figure 8: Typical farmland structure in the United States (left) versus Heilongjiang (right). For comparison, the red point in the center of the image is surrounded by a buffer of 1000m. The area of GEOgraphic-Object-Based Image Analysis (GEOBIA) emerged as an alternative to pixel-based image analysis methods. Pixel-based methods of image classification treat each pixel as separate from its neighbors and leaves out the context of the pixel entirely. Pixel-based methods are still widely used and are extremely applicable when the target of an image is less than or equal to pixel size. However, there are many applications where this is not the case, and the recent onset of accessible and extremely high-resolution imagery has made GEOBIA more popular among geographers. When using an object-based approach, the image first goes through a process where similar pixels are grouped into objects based on their attributes. The same classification methods used for pixel-based methods can then be applied to these objects. The main benefit of segmenting the image first is that it retains its context and 45 there is less misclassification within objects due to pixel-pixel variation (i.e., salt and pepper effect). Machine-learning algorithms are commonly applied to GEOBIA. Duro et al. compared three different algorithms, decision tree (DT), random forest (RF), and support vector machine (SVM), using both pixel and object-based classification. They found that pixel-based showed no significant difference across the three algorithms, though object based showed a significantly different result with DT. Also, classifications using either RF or SVM gave the most valuable depiction of cropland, regardless of pixel or object based, though, the pixel-based method took less computational time. However, they concluded that there was no obvious advantage over the other when it comes to pixel or object based (Duro et al., 2012). Random forest is a machine learning method of classification that has shown to be one of the more reliable methods of classification due to it using both decision trees and sub-setting (similar to bootstrapping) to improve prediction while controlling for over-fitting (Belgiu & Drǎguţ, 2014; Gislason et al., 2006). The algorithm is trained on a subset of validated data (i.e., ground truth) and produces multiple decision trees for each observation, which then assign a class based on the majority outcome. Many studies and reviews have tested RF against other classification methods, and it reliably performs the best. Its superior performance is attributed to the fact that it can deal with both high data dimensionality as well as multicollinearity, things that are extremely common in large data sets. However, RF can be sensitive to sample design and tends to favor classes with a large number of training samples. It is also preferred because there are only two parameters that are user defined (number of trees and the number of features used to split a node), whereas other methods have much more ambiguity for the user to define. 46 Researchers have also explored and examined current methods strictly within the object- based sphere. After examining a case study in China, Li et al determined that among seven GEOBIA methods, they found that RF was the most accurate classifier when varying the different segmentation variables (M. Li et al., 2016). In fact, RF is one of the most commonly used methods in GEOBIA classification, and it is a promising avenue as more high quality and high-volume data are created. For example, satellite programs like Sentinel offer unprecedented ability to further discriminate landcover types as spectral differences can be parsed more easily. This is expected to be extremely useful for mapping agriculture (Lebourgeois et al., 2017). An example of a study that merged these two methods, did so by mapping irrigation in Africa (Vogels et al., 2019). By using a higher spatial resolution than would be allowed for pixel-based methods, they showed that there was more irrigation in the region than one would infer from maps made using coarser resolutions. This result brings into question our understanding of the world through the lens of these potentially inaccurate data products, and we should expect many of them to be challenged or corrected in the future. 3.1.2 Objectives Recent research, including empirical remote sensing studies in our study region of Heilongjiang, China, have indicated cropland conversions in importing countries in response to international trade (J. Sun et al., 2015; van Vliet et al., 2015). Farmer land-use decisions are affected by multiple factors such as climate, resources, and crop prices, among others (Lutz, 1998; S. A. Wood et al., 2014). As global trade increases, small farmers struggle to compete when their markets are flooded with cheap imports (Tilt, 2008). Crop price plays a large part in determining what small-scale farmers choose to grow from one year to another and whether they plant a single crop type or diversify (Di Falco & Perrings, 2005; Michler & Josephson, 2017; D. 47 Wood & Lenné, 1999). Additionally, as small farmers in developing countries become integrated in a global commodity market, they have to consider risk and market uncertainty more than ever before (Hao, 2010). With more farmers reacting to global markets by altering their cultivation strategies, the structure of agroecosystems changes as well. While some global landcover maps exist, they are not at the resolution needed to accurately capture the heterogeneity of the study region nor do they differentiate between crop types (Figure 8). The inherent inaccuracies and limitations of these products are also difficult to ascertain. Furthermore, accurately classifying crop types can be challenging as they often have very similar spectral characteristics, and many crop classification methods focus on using single- source optical data. However, recent advances in sensor capabilities and data processing techniques have enabled the use of both optical and radar data, which enhances the performance of crop classification, particularly in the context of crop identification (C. Liu et al., 2019; Orynbaikyzy et al., 2019). Additionally, many crop classification methods leave out non-crop classes which limits their applicability to landscape-level analyses (Kang et al., 2018; You et al., 2021; Zhang et al., 2020; Zhong et al., 2016). Therefore, the main objective of this analysis is to create high fidelity landcover maps of our study region that distinguish between the region’s major crops (soybean, corn, and rice) as well as other non-crop classes (wetland, forest, grassland, built-up) at a fine spatial resolution to facilitate calculation of landscape metrics relevant to agricultural structure and biodiversity. 48 3.2 Methods Excluding field data collection, the entire process of creating landcover maps occurred in Google Earth Engine (GEE). To see each step of the workflow, see the diagram in Figure 9. The accessibility of imagery in GEE as well as their ever-growing library of processing and analysis tools makes it an ideal platform to create custom landcover maps for many applications. Figure 9: Workflow diagram for the landcover classification in Google Earth Engine. 49 3.2.1 Study System Heilongjiang province in northeast China produces much of the country’s grain (Figure 10). With a climate characterized by long, harsh winters and short, mild summers, only a single crop growing season is possible (Figure 11) (Olson et al., 2001; Weather and Climate, 2016). The three main crops grown in the region are soybean, corn, and paddy rice. However, with the growing influence of global trade and cheap imports of soybeans, farmers have switched to growing more corn and rice in recent years. From 2009-2013, soybean cultivation decreased by nearly 40% (4 million to 2.4 million ha) (J. Sun et al., 2015; Survey Office of the National Bureau of Statistics in Heilongjiang, 2013). As previously stated, Heilongjiang’s main industry is agriculture therefore the majority of land is dedicated to agriculture. Heilongjiang is primarily cultivated by smallholder farmers (farm sizes around 3-10 ha). The landscape is typified by villages scattered across the landscape, with fields of varying crop type, size, and shape in the surrounding vicinity (Figure 8). Figure 10: Map of Heilongjiang Province, China. 50 Figure 11: Climate of Heilongjiang Province from 1970-2000 (mean temperature in dark blue, min/max in light blue, and average precipitation in orange. 3.2.2 Ground Truth Data Collection In the summer of 2018, we collected ground truth data across Heilongjiang (Figure 12). Points were selected in areas with at least 100m2 of a single landcover class for nine classes: built up, corn, forest, grassland, ‘other crop’, rice, soybean, water, & wetland (Table 8). Due to accessibility and logistical constraints, most points were collected along roadsides to increase the overall sample size for each class. In addition to ground truth points collected in the field, additional points were collected using satellite imagery to supplement the non-crop classes that were under-represented in the field data. Of the points collected, 30% were withheld as testing data for classification accuracy assessment. The number of samples per land class can be found in the table below (Table 8). 51 Figure 12: Map of Heilongjiang and location of field ground truth points Table 8: Class definitions and ground truth totals Class Training Testing Total Description Corn 1365 535 1900 Rain-fed cropland planted with corn (Zea mays) Soybean 855 390 1245 Rain-fed cropland planted with soybeans (Glycine max) Rice 1095 492 1587 Flooded paddy rice (Oryza sativa) Cropland planted with an alternative crop (e.g., potato, tobacco, ‘Other crop’ 117 71 188 rapeseed, cucumber, watermelon, etc.) Forestland, typically early successional. Either mixed species Forest 351 149 500 stands for single planted stands Both scrub and grassland (vegetation no taller than 3 m) in dry Grassland 362 143 505 soil. Wetland 183 86 269 Saturated land with some vegetation cover Water 188 63 251 Body of water >1m deep Built-up 332 167 499 Villages, roads, greenhouses, etc. Total 4848 2096 6944 3.2.3 Imagery Acquisition & Processing Shapefiles of each bird point location (collected in 2017 & 2018, see Chapter 4) as well as ground truth points (collected in 2018) with 2000m buffers were used to filter Sentinel-2 Level-1C (i.e., top of atmosphere, orthorectified) imagery in Google Earth Engine. Unfortunately, Level-2A (i.e., surface reflectance) images were not available in our study area for either year. Three time periods throughout the growing season were used to create the filtered 52 collections: early (5/1-6/30), mid (7/1-8/31), and late (9/1-10/31). The ground truth collection (the collection with the largest extent) contained around 35 Sentinel-2 tiles. Only images with <20% cloudy pixel percentages were included1. The resulting size of the image collections for 2017 and 2018 are presented in the table below (Table 9). Since Sentinel-2B was not launched until March 2017, there were significantly fewer images available in 2017 as the satellite was still in the ramp-up phase during most of the growing season (Sudmanns et al., 2020). The Sentinel-1 collections were filtered using the same workflow as the optical imagery (Table 9). Table 9: Total number of images in for each collection (2017 samples, 2018 samples, and ground truth samples) that were used for the image mosaicking process. Sentinel - 2 Sentinel - 1 Ground Ground 2017 2018 truth 2017 2018 truth (2018) (2018) Early 36 176 284 101 90 97 Mid 42 103 171 67 63 69 Late 76 187 293 70 48 53 Five vegetation indices were calculated for all images in each collection before calculating the median value for all pixels (see Table 10 for specific band designations and index definitions). In addition to the commonly used vegetation (NDVI)(Rouse & Haas, 1974) and water indices (NDWI)(Gao, 1996), we also derived three additional indices known to aid in crop type separation, particularly between corn and soybean (Kang et al., 2018; Zhang et al., 2020; Zhong et al., 2016). The first two combined the visible and SWIR bands to provide information on vegetation senescence (NDSVI)(Qi et al., 2002) and residue (NDRI)(Gelder et al., 2009). The final index was calculated using bands just in the SWIR region to provide information on the tillage state of a pixel/object (NDTI)(Van Deventer et al., 1997). 1 One site was filtered at 60% cloudy pixels to ensure image availability in the early season for 2017. Before adding them to the entire collection, we checked that no cloudy pixels occurred within the 2km sample area. 53 Table 10: Data layers used for classification Approx. Band Sensor (res.) Abbrev. Name Center/Calculation B2 Blue 490 nm B3 Green 560 nm B4 Red 660 nm Sentinel-2 B8 Near infrared (NIR) 830 nm (10 m) NDVI Normalized difference (B8-B4)/B8+B4 vegetation index NDWI Normalized difference (B3-B8)/B3+B8 water index B5 Red edge 1 700 nm B6 Red edge 2 740 nm B7 Red edge 3 780 nm B8A Near infrared 860 nm narrowband Sentinel-2 B11 Short wave infrared 1 1610 nm (20 m) (SWIR) B12 Short wave infrared 2 2200 nm (SWIR) NDTI Normalized differential (B11-B12)/(B11+B12) tillage index NDSVI Normalized difference (B11-B4)/(B11+B4) senescent vegetation Sentinel-2 index (10m & 20m)* NDRI Normalized difference (B4-B12)/(B4+B12) residue index VV Single co-polarization, 5.4 GHz vertical transmit/vertical (~5.5 cm wavelength Sentinel-1 receive C-band) C-band SAR VH Dual-band cross- 5.4 GHz (10 m) polarization, vertical (~5.5 cm wavelength transmit/horizontal C-band) receive *Final resolution of indices is 20m Since Sentinel data are collected at a relatively fine spatial resolution (10 and 20m), we segmented the images into objects prior to classification rather than just classifying individual pixels. Doing so reduces error due to spectral variability within a landcover class and preserves features such as field borders/margins. Before initializing segmentation, we stacked the early, mid, and late mosaics for each sample (2017, 2018, GT). To initiate the clustering algorithm, we 54 set a seed grid over the mosaics with 20 pixels between each seed and used B2, B3, B4, B8, NDVI, NDWI, NDSVI, NDRI, and NDTI as clustering variables. Finally, the target bands for classification were extracted. The 60m resolution bands (B1, B9, B10) and the SAR layers (HH, HV) were excluded. We then ran the mosaics through an image segmentation algorithm based on simple non-iterative clustering (Figure 13). The parameters used can be found in the table below (Table 11). After the segmentation, the band and index means were calculated for each segment. In GEE, composite (i.e., mosaicked) images do not have a specific resolution until it is specified by the user. For this application, we exported the final rasters at 10m for all data layers. Figure 13: Example result from the simple non-iterative clustering algorithm 55 Table 11: Parameters for simple non-iterative clustering algorithm in GEE Argument Value Details image image The input image for clustering. size 5 The superpixel seed location spacing, in pixels. If 'seeds' image is provided, no grid is produced. Default = 5 compactness 1 Compactness factor. Larger values cause clusters to be more compact (square). Setting this to 0 disables spatial distance weighting. Default = 1 connectivity 8 Connectivity. Either 4 or 8. Default = 8 neighborhoodSize 64 Tile neighborhood size (to avoid tile boundary artifacts). Defaults to 2 * size. Default = null seeds defined by If provided, any non-zero valued pixels are used as seed seed grid locations. Pixels that touch (as specified by 'connectivity') are considered to belong to the same cluster. Default = null 3.2.4 Landcover Classification For the classification of image segments, we used a RF classifier trained on ground truth (GT) points collected during the 2018 field season. The data layers used for the image segment classification were the mean value for bands 2-8, 8A, 11, 12, NDVI, NDWI, NDSVI, NDRI, NDTI, VV, & VH (17 for each time period, total of 51) in each cluster (Table 10). Seventy percent of the points were randomly chosen to train the RF classifier with the remaining 30% set aside for training. The number of trees was set to 450, with all other parameters set to default (Table 12). The trained classifier was then run on the 2018 and 2017 samples. To assess the performance of the classifier, we calculated overall, user, and producer accuracies for both the training and testing ground truth points as well as F- score metrics for each class. We also derived variable importance for the RF algorithm in GEE to compare across the bands used in the classifier. 56 Table 12: Parameters for Random Forest Classifier in Google Earth Engine Argument Type Details numberOfTrees Integer, default: The number of Rifle decision trees to create per class. 1 variablesPerSplit Integer, default: The number of variables per split. If set to 0 (default), 0 defaults to the square root of the number of variables. minLeafPopulation Integer, default: The minimum size of a terminal node. 1 bagFraction Float, default: The fraction of input to bag per tree. 0.5 outOfBagMode Boolean, Whether the classifier should run in out-of-bag mode. default: false seed Integer, default: Random seed. 0 3.3 Results & Discussion 3.3.1 Accuracy Assessment The re-substitution and validation accuracies (i.e., training and testing accuracies) were assessed for the final classified image. Overall training accuracy was 91% whereas the testing accuracy was 80%. The kappa value for the training data was 0.89 for the training data, reflecting a high level of agreement, and the testing data’s kappa value of 0.76, while slightly lower, still indicates a substantial level of agreement (Table 13). The accuracy of each class varied somewhat. For the training data, both the producer and user accuracies of built-up, forest, grassland, rice, water, and wetland were above 90%. Corn had a producer’s relatively high producer accuracy at 92%, but a lower user accuracy (83%) due to a higher number of commission errors (false positives). The soybean and ‘other crop’ class had low producer accuracies (76.7% and 76.9%) with slightly better user accuracies (86.8% and 90.6%) due to a higher number of omission errors (false negatives). For the testing data, wetland was the only class with both user and producer accuracies above 90%. Rice was relatively accurate with a producer and user accuracy of 92.9% and 89.9%, 57 respectively. Similarly, water had a producer and user accuracy of 85.7% and 92.1%. The forest, grassland, and built-up classes had moderate levels of accuracy with user accuracies all around 78% and producer accuracies ranging from 83-88%. In contrast, the three rain-fed -crop classes all performed relatively poorly. Corn, soybean, and ‘other crop’ had a producer’s accuracy of 78.3%, 62.1%, and 52.1% and a user accuracy of 68.6%, 77.9%, and 81.7%. Similar to the testing data, the difference in producer and user accuracies can be attributed to higher numbers of commission errors in corn and omission errors in soybean and ‘other crop’. Looking at the confusion matrices (Table 14,Table 15) for the training and testing data, we can see the majority of commission errors for corn occurred in the soybean class, with most omission errors in soybean being incorrectly classified as corn. The ‘other crop’ class was most often confused with both corn and soybean. The ‘other crop’ class’s low performance can also be attributed to the fact that it had a low number of ground truth points compared to the rest of the classes. Ground truth supplementation was conducted by visually inspecting satellite imagery so identifying pixels containing crops other than rice, corn, and soybean was not feasible. That being said the performance of this classification method to correctly identify the crops in Heilongjiang performs similarly to (if not better than) recent mapping efforts (You et al., 2021). 58 Table 13: Accuracy assessment for training data: producer accuracy (PA), user accuracy (UA) Training Testing PA UA PA UA built-up 95.783 93.373 82.635 78.443 corn 92.015 83.223 78.318 68.598 forest 97.443 98.580 87.919 85.906 grassland 93.094 96.685 86.014 78.322 othercrop 76.923 90.560 52.113 81.690 rice 96.986 97.260 92.886 89.837 soybean 76.725 86.784 62.051 77.949 water 96.277 100 85.714 92.063 wetland 100 99.454 93.023 94.186 OVERALL 91.277 - 80.2 - KAPPA 0.893 - 0.759 - Table 14: Training confusion matrix Predicted built-up corn forest grassland othercrop rice soy water wetland TOTAL built-up 318 3 1 1 0 5 4 0 0 332 corn 6 1256 1 3 1 4 94 0 0 1365 forest 1 7 343 0 1 0 0 0 0 352 Reference grassland 5 14 0 337 0 4 1 0 1 362 othercrop 1 13 1 0 90 4 8 0 0 117 rice 6 15 0 5 2 1062 5 0 0 1095 soybean 2 175 2 3 7 10 656 0 0 855 water 1 2 0 0 0 3 1 181 0 188 wetland 0 0 0 0 0 0 0 0 183 183 ALL 4849 Table 15: Testing confusion matrix Predicted built-up corn forest grassland othercrop rice soy water wetland TOTAL built-up 138 7 3 5 1 9 2 1 1 167 corn 12 419 9 6 4 21 64 0 0 535 forest 7 3 131 1 0 4 3 0 0 149 Reference grassland 2 10 2 123 0 2 1 0 3 143 othercrop 3 18 1 0 37 0 12 0 0 71 rice 6 9 3 8 4 457 4 1 0 492 soybean 5 118 3 8 4 10 242 0 0 390 water 0 2 0 2 0 4 0 54 1 63 wetland 1 1 0 1 0 0 0 3 80 86 ALL 2096 59 To improve the overall accuracy and quality of data derived from the classification map, the rain-fed crop classes (i.e., corn, soybean, and ‘other crop’) were merged into one class (Table 16). The original landcover maps that contained 9 landcover classes were also re-classified to just three classes (built-up, cropland, and natural landcover) to facilitate calculation of various landscape metrics (e.g., percent of natural landcover). Table 16: Merged classification schemes Original Classes Rain-fed crop Reclassified n=9 n=7 n=3 built-up built-up built-up rice rice cropland corn rain-fed crop cropland soybean rain-fed crop cropland ‘other crop’ rain-fed crop cropland forest forest natural grassland grassland natural water water natural wetland wetland natural For the rain-fed crop classification, overall training and testing accuracies increased substantially to 97.38% (+6.08%) and 90.70% (+10.5%) respectively (Table 17). The reclassification of just 3 classes (built-up, cropland, and natural) resulted in even more accurate classifications with a training accuracy of 98.23% (+6.93%) and a testing accuracy of 93.61% (+13.41%) (Table 18). For both reclassified maps, the producer and user’s accuracy for the training data was above 90% across all landcover classes. For the testing data, the merged rain-fed crop map had the highest accuracies in the wetland, rain-fed crop, and rice classes (accuracies ≥90%). The remaining classes all had accuracies at 78% or higher (Table 17). The accuracy assessment of the testing data for the simple reclassification found cropland class to be the most accurate, followed by natural and built-up classes (Table 18). The confusion matrices for each reclassification can be found in Table 19-Table 22. 60 Table 17: Accuracy assessment for training and testing data of rain-fed crop reclassification Training Testing PA UA PA UA built-up 97.853 92.638 82.635 78.443 forest 97.443 98.580 87.919 85.906 grassland 93.094 96.685 86.014 78.322 rain-fed crop 98.373 97.945 92.169 95.281 rice 96.986 97.260 92.886 89.837 water 95.767 100.000 85.714 92.063 wetland 100.000 99.454 93.023 94.186 OVERALL 97.381 - 90.697 - Table 18: Accuracy assessment for training and testing data of reclass reclassification Training Testing PA UA PA UA built-up 95.796 92.793 82.635 78.443 natural 96.225 98.435 90.930 88.889 cropland 99.096 98.688 95.632 96.707 OVERALL 98.227 - 93.607 - Table 19: Rain-fed crop training confusion matrix Predicted built-up forest grassland rain-fed crop rice water wetland TOTAL built-up 319 1 1 7 5 0 0 333 forest 1 343 0 8 0 0 0 352 grassland 5 0 337 15 4 0 1 362 Reference rain-fed crop 10 4 6 2298 18 0 0 2336 rice 6 0 5 22 1062 0 0 1095 water 2 0 0 3 3 181 0 189 wetland 0 0 0 0 0 0 183 183 ALL 4850 Table 20: Rain-fed crop testing confusion matrix Predicted built-up forest grassland rain-fed crop rice water wetland TOTAL built-up 138 3 5 10 9 1 1 167 forest 7 131 1 6 4 0 0 149 grassland 2 2 123 11 2 0 3 143 Reference rain-fed crop 20 13 14 918 31 0 0 996 rice 6 3 8 17 457 1 0 492 water 0 0 2 2 4 54 1 63 wetland 1 0 1 1 0 3 80 86 ALL 2096 61 Table 21: Reclass training confusion matrix Predicted built-up natural cropland TOTAL built-up 319 2 12 333 Reference natural 8 1045 33 1086 cropland 16 15 3400 3431 ALL 4850 Table 22: Reclass testing confusion matrix Predicted built-up natural cropland TOTAL built-up 138 10 19 167 Reference natural 10 401 30 441 cropland 26 39 1423 1488 ALL 2096 We also calculated the F-score for each of the original and merged classes. F-score is a metric that combines both precision and recall performance of a classifier and ranges from ‘worst’ to ‘best’ using values between 0 and 1 (Table 23). The majority of classes exhibited high F-score values, for both the training and testing data (>0.8). The classes with the lowest F-scores were corn, soybean, and ‘other crop’ classes in the original classified map. However, the F- scores of the merged rain-fed crop class for both training and testing data were above 0.9. Table 23: F-scores for each class All classes built-up corn soy other rice forest grass- wetland water Train 0.944 0.881 0.807 0.826 0.971 0.980 0.948 0.997 0.978 Test 0.809 0.747 0.674 0.612 0.915 0.870 0.828 0.936 0.885 Rain-fed crop built-up rain-fed crop rice forest grassland wetland water Train 0.944 0.980 0.971 0.980 0.948 0.997 0.978 Test 0.809 0.936 0.915 0.870 0.828 0.936 0.885 Reclassified built-up cropland natural Train 0.944 0.989 0.973 Test 0.809 0.961 0.900 62 3.3.2 Variable Importance Two variables were distinctly important compared to the rest of the variables used in the RF classifier (Figure 14, Table 24). The top-ranking variable was the Sentinel-1 C-band cross polarized VH data for the early time period. The next most important variable was NDTI (normalized difference tillage index) from the early time period. The next two variables in the ranking were also indices from the early period: NDWI (normalized difference water index and NDSVI (normalized difference senescent vegetation index). The NDTI for both the late and mid time periods were the next highest ranked, resulting in all three phenological periods for the same metric ranking in the top ten ranked variables. It is also notable that the seven of the top ten variables were derived from the early time period, highlighting the importance of indices from early in the growing season. Comparing two of the top data layers, VH and NDTI, across all three time periods one can see there is significant variability (Figure 15). For the VH data, the peak occurs early in the season, decreases significantly during the height of the growing season and the increases slightly around harvest time (Figure 15, a-c). NDTI showed a similar pattern, though not as pronounced (Figure 15, d-f). You can also observe the relatively constant values for the water (dark blue) and built-up (gray) regions in each image band image when comparing to the classified image (Figure 15, g). 63 Figure 14: Bar graph of data layer importance for random forest classification 64 Table 24: Data layer ranked by variable importance rank band importance rank band importance 1 VH_early 724.016 27 B4_mid 575.268 2 NDTI_early 691.432 28 B2_early 574.638 3 NDWI_early 640.533 29 B3_mid 573.473 4 NDSVI_early 639.248 30 B11_mid 568.819 5 NDTI_late 638.017 31 NDWI_late 568.045 6 NDTI_mid 636.149 32 B8A_early 567.087 7 VH_mid 628.957 33 NDVI_mid 563.617 8 NDVI_early 628.829 34 NDVI_late 558.595 9 VV_early 628.559 35 B12_late 551.397 10 NDRI_early 623.969 36 B6_mid 549.361 11 VV_mid 623.755 37 B8A_late 547.822 12 B5_mid 622.738 38 B7_mid 546.712 13 VH_late 619.534 39 B7_late 546.118 14 B11_early 602.069 40 B4_late 545.718 15 VV_late 600.015 41 B7_early 542.923 16 B12_mid 598.964 42 B6_early 540.137 17 NDSVI_late 593.742 43 B8_late 539.084 18 B2_late 592.968 44 B8_mid 531.535 19 NDRI_mid 590.811 45 B4_early 530.072 20 B12_early 585.410 46 B8_early 526.940 21 NDSVI_mid 583.504 47 B6_late 524.495 22 B2_mid 582.286 48 B3_early 522.843 23 NDWI_mid 581.532 49 B5_early 519.542 24 B3_late 581.379 50 B8A_mid 508.663 25 NDRI_late 580.274 51 B11_late 507.015 26 B5_late 576.418 65 Figure 15: Phenological difference between two of the most important data layers, VH (a-c) and NDTI (d-f) for all three time periods compared to the final classification (g). 66 3.4 Conclusions This chapter demonstrates a workflow that produces high fidelity landcover classification for a heterogeneous agricultural landscape. By utilizing fine spatial resolution data from both optical and SAR sensors at multiple points during the growing season and using an object-based approach, we were able to achieve an accuracy of >80% on the validation data. Merging the rain- fed crop classes further increased the accuracy to >90%. Evaluating the performance of each data layer in our RF classifier solidified the importance of phenological data, as many of the top bands were from earlier in the growing season. We also found that indices involving the SWIR region were among the top-ranking variables. This entire process was performed in Google Earth Engine and could easily be transferred to other regions and time periods, so long as the user supplies the appropriate amount of ground truth data for each target class. The final classified maps can be used for several applications, including landcover dynamics and inventories as well as the calculation of landscape metrics for landscape-level analyses. 67 CHAPTER 4: LANDSCAPE STRUCTURE AND BIRD DIVERSITY IN AGROECOSYSTEMS OF NORTHEAST CHINA 68 Abstract International food trade drives land use/cover change, having consequences for wildlife that utilize agroecosystems. To explore this, we assessed bird communities and landscape structure in the agricultural region of Heilongjiang Province in northeastern China. We examined the relationship between functional and taxonomic measures of bird diversity and landscape metrics (both compositional and configurational). We found that several landscape metrics significantly related to bird metrics, though there was little difference in the relationships of between bird diversity and landscape structure when comparing taxonomic versus functional diversity. Crop diversity as well as natural landcover and edge metrics positively correlated with bird richness, while patch aggregation and rain-fed crop (e.g., corn, soybean) percentage were negatively related. We also found that paddy rice in our study region supported functionally different bird species compared to soybean and corn. Finally, we compared regions in Heilongjiang impacted by global trade and found that those with a larger decrease in soybean cultivation had lower levels of crop diversity as well as higher amounts of rice area. There were no significant regional differences in bird diversity, however, which indicates that negative impacts to bird diversity due to soybean conversion (and a decrease in crop diversity) may be mitigated by increases in rice cultivation rather than corn. 69 4.1 Introduction 4.1.1 Background Agroecosystems are a multi-functional landscapes that not only can be utilized for food production, but also serve as habitat for a multitude of species that provide ecosystem services and support robust ecosystems (Godfray & Garnett, 2014). To understand species’ contributions to ecosystem function in agroecosystems, research is undergoing a shift that focuses on functional, rather than purely taxonomic diversity. This provides a more holistic perspective as functional diversity often serves as an indicator of the wide range of ecosystem services on which humans and food production rely (S. A. Wood et al., 2015). Scientists often study ecosystems using taxonomic definitions to group or separate species. However, taxonomic metrics alone may not provide insight about the functioning of an ecosystem (Cadotte et al., 2011). Many common functional biodiversity metrics are trait-based, though they can also be based on environmental gradients, biotic interactions, and energetic/reproductive performance (Mcgill et al., 2006). Furthermore, studies that examine both functional and taxonomic biodiversity often find contrasting results (Edie Stewart M. et al., 2018; Jacoboski & Hartz, 2020; W. Li et al., 2021). To understand landscape characteristics that promote biodiversity, we often use metrics to quantify various aspects of landscape heterogeneity. Here, we focus on structural landscape heterogeneity, which distinguishes landcover types by their geographic characteristics. Alternatively, some researchers classify landcover based on species-specific resource needs (Fahrig et al., 2011). There are two components of structural landscape heterogeneity that are typically considered: composition and configuration. Composition refers to the identity of landcover types found in a landscape while configuration is defined by the spatial layout of various landcover types (Fahrig & Nuttle, 2005). There are numerous metrics that can be used to 70 characterize a landscape, but care must be taken to avoid using correlated and redundant metrics simultaneously (Frazier & Kedron, 2017). Additionally, in the context of biodiversity, it remains unclear which spatial resolution is the most appropriate to apply, and what extent to calculate landscape metrics for the most informative results (Fahrig et al., 2011). There is a need to understand and quantify characteristics of agricultural landscapes that bolster biodiversity. There have been some attempts to create and assess habitat descriptors, mainly in Europe (Herzog et al., 2017). They recommend seven descriptors (richness, diversity, patch size, linearity, crop richness, shrub habitat amount, and tree habitat amount) as well as a semi-natural habitat descriptor which is interpreted differently across farm type. However, these descriptors only assess potential habitat quality and do not address realized biodiversity in terms of number of species (or functional groups) of a landscape. Other studies have found that general heterogeneity mediates farmland bird diversity rather than crop diversity (Redlich et al., 2018). However this does not seem to be true in regions that also farm paddy rice (M. B. Lee & Goodale, 2018). 4.1.2 Research Questions Using geo-located samples of bird diversity and landcover maps, we aim to determine the relationship between agricultural landscape characteristics and bird diversity in taxonomic and functional contexts in our study region of Heilongjiang Province in northeastern China. Our research questions are listed below. Q1: What is the impact of global trade on landscape structure? Q2: What is the relationship between landscape structure and bird diversity? Q3: What is the relationship between crop diversity and functional/taxonomic bird biodiversity? Q4: What is the relationship between landscape structure and functional bird community traits? 71 4.2 Methods 4.2.1 Bird Data Collection In the summer of 2017 and 2018, we collected bird data at 37 sites across Heilongjiang. At each site, we conducted a series of point counts (2-9 per site depending on conditions) for a total of 207 bird samples (Figure 16). These geolocated data were collected using a point count method (Bibby, 2000; Hutto, 1986). All birds seen or heard were recorded (both species and abundance) within 50m (2017) or 100m (2018) for three, three-minute intervals. Only those birds seen within a 50m range were used for this analysis. Additionally, birds flying over the sample point were not counted unless they were actively using the air space (e.g., swallows feeding over a field, kestrel hovering while hunting). Birds recorded as present in previous time intervals were re-recorded since each interval was treated as a separate sample. Figure 16: Map of Heilongjiang and location of sites where bird biodiversity was sampled. Counties in the Qiqihar (west) and Jiamusi (east) prefectures are outlined in black. 4.2.2 Biodiversity Indices We calculated taxonomic richness (TR) at the family-level as many species were difficult to distinguish in the field (e.g., swallow species often occurred in mixed flocks). The taxonomic diversity metrics we examined were richness (i.e., number of families) and the Shannon index. 72 As seen in ( 1.1 ), the Shannon index (H) is based on the summation across species abundances, where 𝑝𝑖 is the proportional abundance of species 𝑖 and 𝑏 is the log base (natural log), which balances the importance of both abundant and rare species (Morris et al., 2014). The vegan package in R was used for this calculation (Oksanen et al., 2020; R Core Team, 2021). 𝐻 = − ∑ 𝑝𝑖 𝑙𝑜𝑔𝑏 𝑝𝑖 ( 1.1 ) 𝑖 Before calculating functional diversity indices, we created a trait matrix for all families present in our sample (Table 34). In addition to average body mass, we also collected information on different suites of binary traits: habitat preference (wetland, water, grassland, open woodland, forest, and forest edge), diet (insects, fish, small mammals, small birds, other invertebrates, grains, nestlings/eggs, herps, plant matter, fruits, and carrion), foraging behavior (stalking, bark gleaning, hovering, hawking, scanning, foliage gleaning, continuous aerial feeding, ground foraging, sallying diving, dabbling, and probing [see Table 33 for definitions]), nesting behavior (ground nester), and migratory status (summer, winter, resident, migration route, and scarce). The majority of the information was gathered from the Birds of the World database as well as field guides (Billerman et al., 2020; Brazil, 2009). The final trait frequency for each family is listed in Table 25. Next, we created functional groups based on the trait matrix by inspecting the dendrogram created by clustering the families using the Ward method (Petchey & Gaston, 2006; Ward, 1963). Each suite of traits was weighted as follows: mass (1), habitat preference (1.25), diet (1.5), foraging behavior (1.5), ground nester (1), and migratory status (1). The diet and foraging traits were weighted the most to prioritize traits that may contribute to ecosystem services/disservices related to agriculture (e.g., pest-control, crop destruction, seed predation). 73 In addition to functional group richness, we calculated functional dispersal as well as the community weighted means (CWM) of each trait for each sample. Functional dispersion (FDis) is the mean distance in a functional space from the centroid, weighted towards the most abundant species (Laliberté & Legendre, 2010), and the CWM of each trait allowed us to assess community functional composition (Lavorel et al., 2008). All binary trait variables were treated as continuous, so higher values of CWM indicate higher frequencies of a given trait. The functional diversity metrics were calculated using the FD package in R (Laliberté et al., 2014; Laliberté & Legendre, 2010; R Core Team, 2021). 74 Table 25: Summary of trait frequency for families (n=30) freq perc (%) Wetland 20 66.7 Habitat Preference Open woodland 18 60.0 Grassland 17 56.7 Forest Edge 13 43.3 Forest 11 36.7 Water 8 26.7 Insects 27 90.0 Other invertebrates 27 90.0 Grains & seeds 14 46.7 Plant matter 12 40.0 Small herps 11 36.7 Diet Fruits & berries 10 33.3 Nestlings & eggs 8 26.7 Fish 7 23.3 Small mammals 5 16.7 Small birds 5 16.7 Carrion 2 6.7 Ground foraging 17 56.7 Foliage gleaning 15 50.0 Sallying 6 20.0 Hovering 5 16.7 Foraging Behavior Diving 5 16.7 Probing 5 16.7 Hawking 4 13.3 Scanning 3 10.0 Continuous aerial 3 10.0 Dabbling 2 6.7 Stalking 1 3.3 Bark gleaning 1 3.3 Ground nester 12 40.0 Nesting & Migration Summer 23 76.7 Migratory 23 76.7 Resident 10 33.3 Migration route 4 13.3 Scarce 1 3.3 Winter 0 0.0 75 4.2.3 Landscape Metrics The classified landcover rasters (see chapter 3) for 2017 and 2018 were exported at a GEE scale of 10 and used to calculate landscape metrics using the landscapemetrics R package (Hesselbarth et al., 2019; R Core Team, 2021). Nineteen samples were removed from this analysis due to uncertainty in the land cover classification, resulting in a final sample size of 188. All metrics were calculated at three different extents: 250m, 500m, & 1000m (Figure 17). Different extents were used to identify the single extent, by landscape metric, that incorporated the greatest amount of variability. Figure 17: Comparison of the three extents chosen for landscape metric calculation at the same site Six different metrics were calculated at each extent. Depending on the metric type, calculations were conducted at the landscape or class level (Table 26). For the crop diversity metric (MSIDI_crop), the input landscape contained only the four crop classes (corn, soybean, rice, and ‘other crop’) with the other classes masked (Figure 18). 76 Table 26: Landscape metric definitions, input, equations, and descriptions (Hesselbarth et al., 2019; Laliberté & Legendre, 2010). The ‘reclassified’ landcover input merged similar landcover classes so that there were only 3 classes contained in the raster (built-up, natural, and cropland). Landcover Metric Name input Formula Description (level) 𝑚 Modified Simpson’s Value of 0 indicates only one patch in landscape. diversity index −𝑙𝑛 ∑ 𝑃𝑖2 Metrics increases (no limit) as the number of patches Crop classes (with equal proportions) increases. (MSIDI) 𝑖=1 (landscape) 𝑃𝑖 = landscape area proportion of class 𝑖 𝑚 = number of classes Mean patch area All classes A landscape with all small patches has a value of 0 and (AREA_MN) 𝑚𝑒𝑎𝑛(𝐴𝑅𝐸𝐴[𝑝𝑎𝑡𝑐ℎ𝑖𝑗 ]) increases (no limit) as patch areas increase. Unit in (landscape) hectares. 𝑚 Ranges from 0-100. Values of 0 indicate completely Aggregation index 𝑔𝑖𝑖 [∑ ( ) 𝑃 ] ∗ 100 disaggregated classes. Values of 100 for completely (AI) max 𝑔𝑖𝑖 𝑖 aggregated classes in a landscape. Unit in percent. All classes 𝑖=1 (landscape) 𝑔𝑖𝑖 = the number of like adjacencies max 𝑔𝑖𝑖 = the max number of like adjacencies of class i 𝑚 Value of 0 if all cells in a landscape raster are edge Total edge length ∑ 𝑒𝑖𝑘 cells or if the landscape only contains one class. (TE) All classes Increases (no limit) as landscape fragmentation 𝑘=1 increases. Unit in meters. (landscape) 𝑒𝑖𝑘 = total edge length i = target class k = all other classes Edge density (ED) 𝐸 Value of 0 indicates only one patch is present All classes ∗ 10000 (landscape boundary not included). Value increases (no 𝐴 limit) as the number of patches increase. Unit in (landscape) 𝐸 = total edge length meters. 𝐴 = the total area Percent of landscape ∑𝑛𝑗=1 𝑎𝑖𝑗 Value of 100 indicates only one patch of target class is (PLAND) Reclassified ∗ 100 present in landscape. Approaches 0 as class area All classes 𝐴 decreases. Unit is percent. 𝑎𝑖𝑗 = the area of patch 𝑗 (class) 𝑛 = number of patches 𝐴 = the total landscape area Figure 18: Example landscape input for crop diversity. All non-crop classes (i.e., built-up, forest, grassland, wetland, and water were masked (NA). 77 4.2.4 Prefecture Case Study To evaluate impacts of global trade on landscape and bird biodiversity in Heilongjiang, we compared two prefectures (Jiamusi and Qiqihar) with differing degrees of landcover change as portrayed by soybean planting area. Due to limited availability of Sentinel imagery during our study period, we used annual soybean maps derived from Landsat imagery created by another group of researchers to inform our selection of the regions (X. Li et al., 2021). The entire province experienced growth in soybean area beginning in 2002. However, around 2010, there was a sharp decrease in soybean area across all prefectures, though the degree of decline was not equal. Qiqihar is located in the western half of the province and covers over 42,000 km2. Prior to 2000, Qiqihar had the most soybean planted area in the province (X. Li et al., 2021). After the increase in the early 2000s, the planted area dropped slightly in Qiqihar resulting in another prefecture surpassing the planted soybean area of Qiqihar (X. Li et al., 2021). The Jiamusi prefecture is located on the eastern side of Heilongjiang and is slightly smaller than Qiqihar (~31,000 km2). Compared to Qiqihar, Jiamusi grew fewer soybeans between 2000-2018, though it did go from being the 5th largest grower in 2000 to the 3rd in 2009 (X. Li et al., 2021). However, Jiamusi experienced a more severe drop in soybean area compared to Qiqihar, with soybean planting area dropping below the pre-increase level in 2018 (Figure 19). Using soybean area decline as a proxy for the impact of soybean imports on landscape and bird diversity, we selected Jiamusi as an example of greater impact and Qiqihar as an example of the impact. Each prefecture has a similar number of samples, with 53 and 55 in Jiamusi and Qiqihar, respectively (Figure 20). 78 Area (% of prefecture) Figure 19: Soybean area decline for Jiamusi and Qiqihar prefectures. The orange lines highlight the difference in soybean area between 2000 and 2018. 79 a) b) c) d) e) f) Figure 20: Soybean area in 2000 (a,d), 2009 (b,e), and 2018 (c,f) for Qiqihar (a-c) and Jiamusi (d-f). Data from X. Li et al. 2021. Bird/landscape samples are depicted by purple points. 80 4.3 Results 4.3.1 Biodiversity results After inspecting the dendrogram created by clustering the families based on the trait matrix, we chose to assess functional group diversity using 15 groups (Figure 21). This value was determined based on inspection of a range of group number values. Though other grouping schemes could be argued, the 15 groups we created made ecological sense without masking functionally unique families and were therefore well-suited to our research objectives. The group membership for each family can be found in Table 27. Figure 21: Cluster dendrogram of families based on functional traits 81 Table 27: Family membership to functional groups and occurrence for all sampled prefectures in Heilongjiang, Jiamusi, and Qiqihar (OW=old world). All Jiamusi Qiqihar (n=188) (n=53) (n=55) Fam Grp Fam Group Family Description Grp Occ Fam Occ Grp Occ Occ Occ Occ 1 Acrocephalidae Reed warbs 63 65 26 26 14 14 1 Muscicapidae OW Flycatchers 3 0 0 2 Alcedinidae Kingfisher 1 1 0 0 0 0 3 Anatidae Ducks 2 4 1 2 1 1 3 Ardeidae Herons 3 1 1 4 Campephagidae Cuckooshrike 1 14 1 3 0 7 4 Cuculidae Cuckoos 7 1 5 4 Oriolidae Orioles 3 1 2 4 Turdidae Thrushes 3 0 0 5 Charadriidae Shorebirds 5 5 3 3 0 0 6 Columbidae Doves 17 45 7 13 3 5 6 Emberizidae OW Buntings 29 7 1 6 Phasianidae Pheasants 1 0 1 7 Corvidae Crows & Jays 34 34 12 12 13 13 8 Falconidae Falcons 10 20 4 8 2 2 8 Laniidae Shrikes 10 4 0 9 Hirundinidae Swallows 65 70 10 10 26 30 9 Panuridae Reed Tits 1 0 1 9 Remizidae Penduline Tits 7 0 6 10 Laridae Gulls 1 4 0 2 1 2 10 Phalacrocoracidae Cormorants 2 2 0 10 Podicipedidae Grebes 2 0 2 11 Motacillidae Wagtails & Pipits 3 3 0 0 0 0 12 Paridae Tits 2 33 0 4 0 13 12 Passeridae OW Sparrows 30 3 13 12 Troglodytidae Wrens 1 1 0 13 Picidae Woodpeckers 1 1 0 0 1 1 14 Rallidae Rails 3 3 0 0 2 2 15 Sturnidae Starlings 1 3 1 1 0 0 15 Upupidae Hoopoe 2 0 0 82 Across all samples, the most frequently occurring families were Hirundinidae (swallows), Acrocephalidae (reed warblers), and Corvidae (corvids) (Table 27). These three families were the most common in Jiamusi and Qiqihar as well, with reedwarblers occurring more in Jiamusi and swallows occurring more in Qiqihar. The most common functional groups differed between the two prefecture. Across the entire sample area and in Qiqihar, swallows/reedtits (group 9) occurred more frequently while reedwarb/flycatchers (group 1) were more common in Jiamusi. We also investigated relative frequencies of bird communities based on functional groups. In general, many of the communities we found were unique, with only a couple of occurrences across all 188 points. The most common community only contained species from group 1 (reedwarb/flycatcher) followed by communities compromised solely of families from group 9 (swallow/reedtit) (Table 28). These two groups were also found in several other of the most common communities, including one where they co-occur. Unlike with the taxonomic frequency, Jiamusi had the same most common functional group as the rest of the region (reedwarb/flycatcher) while Qiqihar’s most common group was the swallow/reedtit group. 83 Table 28: Most frequent communities. ALL (n=188) Jiamusi (n=53) Qiqihar (n=55) comm. Freq description comm. Freq description comm. Freq description group1 21 reedwarb/flycatcher group1 10 reedwarb/flycatcher group9 6 swallow/reedtit group9 13 swallow/reedtit group6 3 dove/bunting/pheasant group12 5 sparrow/tit group1 reedwarb/flycatcher group12 12 sparrow/tit group7 3 crow/magpie 4 group9 swallow/reedtit group1 reedwarb/flycatcher group9 swallow/reedtit group6 10 dove/bunting/pheasant group5 2 shorebird/lapwing 3 group12 sparrow/tit group6 dove/bunting/pheasant group4 thrush/cuckoo/oriole group1 reedwarb/flycatcher group1 reedwarb/flycatcher 8 2 group7 2 crow/magpie group9 swallow/reedtit group8 falcon/shrike group9 swallow/reedtit group6 dove/bunting/pheasant group1 reedwarb/flycatcher 8 2 group7 2 crow/magpie group9 swallow/reedtit group9 swallow/reedtit group7 crow/magpie group9 swallow/reedtit group9 swallow/reedtit 8 2 group9 2 swallow/reedtit group12 sparrow/tit group12 sparrow/tit group12 sparrow/tit group7 6 crow/magpie group1 reedwarb/flycatcher 5 group6 dove/bunting/pheasant group8 4 falcon/shrike The richness metrics were calculated by aggregating observations for all 3 intervals recorded for each point. TR ranged from 0-9 with a mean of 1.66 (Table 29). The number of functional groups (FGR) had a similar mean to TR (1.62) and ranged from 0-8 (Table 29). Finally, we mapped the richness metrics across the study region (Figure 23) and found no significant differences between Jiamusi and Qiqihar. Table 29: Summary table of the richness metrics calculated at the point level for all Heilongjiang prefectures (n=188), Jiamusi (n=53), and Qiqihar (n=55) All Prefectures Jiamusi (more impacted) Qiqihar (less impacted) mean sd mean sd mean sd n.sp 1.66 1.37 1.60 1.55 1.73 1.58 FGR* 1.62 1.28 1.58 1.52 1.64 1.32 * Number of groups (15) determined based on visual inspection of dendrograms produced using the ward method of clustering. 84 Figure 22: Histograms of the three richness metrics with the mean depicted by the red vertical line: number of families (n.sp) and functional group richness (FGR) Figure 23: Map of richness metrics: number of families (n.sp) and unctional group richness (FGR). Counties in the Qiqihar (west) and Jiamusi (east) prefectures are outlined in black. 85 The abundance-weighted metrics were calculated at the interval level first, and then averaged at the point level. Comparing these two levels of analysis, the mean values of these metrics did not differ greatly (Table 30). Therefore, we will focus on the point-level results. Shannon (H) diversity ranged from 0-1.53 with a mean of 0.217 (Table 30) with a distribution clearly skewed to the right (Figure 24). For the 156 points where functional dispersal (FDis) could be calculated, many had a FDis value of zero, indicating the presence of a single species (Figure 24). The largest value of FDis in our sample was 0.221, with a mean of 0.043 (Table 31). Finally, we mapped the abundance-weighted metrics across the study region (Figure 25) and found no difference between the Qiqihar and Jiamusi (p-value > 0.05) (Table 31). Table 30: Summary table of the abundance-weighted biodiversity metrics averaged across all intervals (n=564) and at the point-level for all samples (n=188). Intervals Points mean sd mean sd H 0.198 0.354 0.217 0.309 FDis* 0.047 0.066 0.043 0.052 * Could not be calculated for communities with 0 species present (32 points) Table 31: Summary table of the abundance-weighted biodiversity metrics at the point-level for Jiamusi (n=53) and Qiqihar (n=55). Jiamusi (more impacted) Qiqihar (less impacted) mean sd mean sd H 0.22 0.34 0.22 0.31 * FDis 0.04 0.06 0.04 0.05 * Could not be calculated for communities with 0 species present (13 and 12 points for Jiamusi and Qiqihar respectively) 86 Figure 25: Histograms of the abundance-weighted biodiversity metrics at the point-level with the mean depicted by the red vertical line: Shannon-Weiner index (shannon) and functional dispersion (FDis). Figure 24: Map of abundance-weighted biodiversity indices at the point-level: Shannon-Weiner index (shannon) and functional dispersion (FDis). Note that the gray points represent missing values (where 0 species are present for FDis. Counties in the Qiqihar (west) and Jiamusi (east) prefectures are outlined in black. 4.3.2 Landscape metric results A summary of the landscape metrics calculated for each extent can be found in Table 32. Crop diversity at the largest extent was significantly higher in Qiqihar. Percent natural area was also significantly higher at the 500m and 1000m extents in Qiqihar, while rice at the highest extent was significantly higher in Jiamusi. Though soybean area mean was not significantly different between the Jiamusi and Qiqihar samples, Qiqihar had more samples with higher soybean percentage (Figure 26). 87 Table 32: Summary table of all landscape metrics at the small (250m), medium (500m), and large (1000m) extents. Bold values indicate significant differences between Jiamusi and Qiqihar (p-value < 0.1). All HLJ Jiamusi (more impacted) Qiqihar (less impacted) mean sd mean sd mean sd msidi_crop_250m 0.30 0.33 0.35 0.37 0.35 0.32 msidi_crop_500m 0.36 0.34 0.36 0.37 0.41 0.34 msidi_crop_1000m 0.42 0.33 0.38 0.34 0.49 0.31 mn_patch_250m 6.03 5.38 4.79 3.83 5.59 5.37 mn_patch_500m 10.32 9.68 8.71 4.81 9.02 7.29 mn_patch_1000m 14.11 6.45 13.61 6.90 12.98 5.41 ai_250m 96.70 2.02 96.26 1.75 96.36 1.98 ai_500m 96.82 1.51 96.50 1.45 96.55 1.39 ai_1000m 96.82 1.21 96.70 1.24 96.58 1.19 3 3 3 3 3 te_250m 2.51x10 1.51x10 2.86x10 1.37x10 2.72x10 1.51x103 te_500m 9.30x103 4.23x103 1.01x104 4.19x103 1.01x104 3.92x103 4 4 4 4 4 te_1000m 3.56x10 1.28x10 3.67x10 1.36x10 3.83x10 1.23x104 ed_250m 99.54 60.11 113.52 54.43 107.81 60.22 ed_500m 92.34 42.05 100.39 41.67 100.06 39.02 ed_1000m 88.41 31.75 90.96 33.76 94.74 30.83 p_nat_250m 15.35 22.22 10.78 12.72 14.29 20.00 p_nat_500m 15.63 19.64 9.96 8.94 14.44 15.49 p_nat_1000m 16.68 18.65 9.14 6.88 15.62 13.53 p_soy_250m 15.86 26.25 11.31 14.24 11.83 24.19 p_soy_500m 14.40 22.30 9.62 10.41 12.27 22.12 p_soy_1000m 13.41 18.81 8.57 8.39 11.53 18.45 p_corn_250m 29.35 32.02 32.55 34.91 37.19 32.51 p_corn_500m 29.93 29.50 35.57 34.68 35.43 28.35 p_corn_1000m 29.65 27.00 36.25 33.02 33.02 24.10 p_rice_250m 33.35 36.71 37.23 35.53 31.43 34.69 p_rice_500m 31.81 34.05 36.32 34.81 29.29 30.85 p_rice_1000m 29.62 30.15 37.29 33.64 26.70 25.54 88 Jiamusi Qiqihar Figure 26: Histograms of percent soybean (p_soy) for Jiamusi (top) and Qiqihar (bottom). 89 4.3.3 Bird Diversity & Landscape Relationship While examining relationships between landscape structure and bird diversity across all samples, we found little differences between taxonomic or functional measures of bird diversity. However, we observed some significant relationships (p-value < 0.1) between bird diversity and landscape structure. Taxonomic (n.sp) and functional group richness (FGR) exhibited similar relationships across all landscape metrics (Figure 27). Both were significantly and positively correlated with crop diversity and percent natural landcover for all extents, with the degree of correlation increasing slightly as extent decreased (Figure 28). Total edge and edge density were also positively correlated with these richness metrics (Figure 27). For TR, the edge metric relationships were stronger at the largest extent while the FGR relationships were strongest at the smallest extent (Figure 29). On the other hand, aggregation was significantly and negatively correlated with taxonomic and functional group richness (Figure 27). Percent soybean and corn were also negatively correlated with TR and FGR (Figure 27). Soybean area was only moderately significant (p-values < 0.1) for the 500m and 1000m extents, while corn was significantly correlated for all extents (Figure 30). Neither mean patch size nor percent rice were significantly correlated with these biodiversity metrics (Figure 27). 90 Figure 27: Correlations between taxonomic richness (n.sp) and functional richness (function group richness). The size of each sign corresponds to the number of extents (250m, 500m 1000m) that were significantly positive (+) or negative (-). Metrics with no significant relationships are denoted by the null sign (∅). 91 Figure 28: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with crop diversity (left) and percent natural landcover (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 92 Figure 29: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with total edge (top left), edge density (top right) and aggregation index (bottom). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 93 Figure 30: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with percent soybean (left) and percent corn (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 94 We found some significant correlations between the abundance-weighted bird diversity metrics, Shannon (H) and functional dispersion (FDis), and landscape structure (Figure 31). Unlike with richness, crop diversity was substantially less related to H and FDis, with just the 500m extent being positively related to FDis (Figure 32). On the other hand, percent natural landcover, total edge, and edge density were all positively correlated across all extents (Figure 31-Figure 33). Aggregation was negatively related across all extents for both H and FDis (Figure 31, Figure 33). H and FDis were negatively correlated across all extents of soybean percentage while corn area was only negatively correlated with H at the 1000m extent (Figure 31, Figure 34). Neither mean patch size nor percent rice were related to the H and FDis biodiversity metrics (Figure 31, Figure 40). 95 Figure 31: Correlations between Shannon’s diversity index (taxonomic) and functional dispersal (functional). The size of each sign corresponds to the number of extents (250m, 500m 1000m) that were significantly positive (+) or negative (-). Metrics with no significant relationships are denoted by the null sign (∅). 96 Figure 32: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with crop diversity (left) and percent natural landcover (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 97 Figure 33: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with total edge (top left), edge density (top right) and aggregation index (bottom). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 98 Figure 34: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with percent soybean (left) and percent corn (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 99 4.3.4 Community Weighted Means and Landscape We detected 29 significant correlations between crop diversity and various CWM values. Note that all these significant relationships were found to be positive except one. Average mass was positively correlated with crop diversity calculated at 250m and 500m (Table 35). Within the habitat preference traits, the forest trait was positively correlated with crop diversity at all extents (Table 35). The majority of the diet traits (6 out of 11) were positively related to crop diversity as well (Table 36). The impact of crop diversity was not as strong when considering the suite of foraging traits. Only bark gleaning, hawking, and scanning exhibited significant relationships in our sample (Table 37). The ground nesting trait was also correlated with crop diversity at the largest extent (Table 38). Finally, the presence of resident species was positively related with crop diversity at all extents, while those that passed through the region on migration were positively correlated at the medium and largest extents of crop diversity (Table 38). The relative relationships between the CWM for habitat, diet, foraging behavior, and migration and crop diversity at 1000m can be found in Figure 35. 100 Figure 35: Linear relationships of crop diversity at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means. 101 Percentage of soybean, corn, and paddy rice was found to be significantly related to many of the CWM traits (26, 54, and 55 for soybean, corn, and rice area respectively). We found that there were many instances where percentage of rain-fed crops (soybean and corn) and rice paddy had opposite relationships with various CWMs. Both corn and rice percent were correlated with average body mass, positively and negatively respectively (Table 35). Every habitat preference trait, except for forest was negatively correlated with corn and positively correlated with rice. Additionally, forest preference was negatively correlated with paddy rice area and positively related to soybean (Table 35). As with habitat preference, the direction of the significant relationships of the rain-fed crops and rice paddy within the diet trait CWMs were often reversed. Percent corn crop was correlated with 6 of the diet traits we recorded, 7 significant relationships were found with rice percentage, and 2 diet traits were correlated with soybean area (Table 36). Fruit and carrion diets CWMs were positively correlated with corn and negatively correlated with rice. Additionally, diets that include nestlings/eggs were negatively correlated with rice. Insect, fish, and invertebrate diet CWMs were positively related to rice and negatively with corn. Seed diets were positively correlated with soybean area and negatively with rice area. Diets with plant matter were positively correlated with soybean area and negatively with corn. Herptiles (reptiles and amphibians), small birds, and mammal diet CWMs were not significantly related to either crop percentage (Table 36). The significant foraging behavior CWMs that were positively correlated with rice (and negatively correlated with both corn and soybean) were hawking and ground foraging (Table 37). Foliage gleaning was also positively correlated with rice but only negatively correlated with corn. Scanning behavior was negatively correlated with rice and positively corelated with corn. Finally, sallying was negatively correlated with corn while dabbling was positively related with 102 rice area. Birds that typically nest on the ground were positively associated with soybean area but negatively correlated with corn (Table 38). Finally, the migration statuses of bird communities were significantly correlated with crop area (Table 38). Corn was negatively correlated with both summer migrants and those that pass through the regions during their migration. Soybean was negatively related to summer migrants and positively correlated with migration route and resident species. Rice area negatively correlated with resident species and positively correlated to summer migrants. The relative relationships between the CWM for habitat, diet, foraging behavior, and migration and the different crop areas at 1000m can be found in (Figure 36-Figure 38). The edge metrics, total edge and edge density were also significantly related to many CWM traits. The majority of the relationships were positive with a few negative relationships (Table 35-Table 38). Aggregation also had several significant relationships with a number of functional traits, with most of them being negative. Mean patch size and natural landcover area had relatively fewer significant correlations with CWM values, with mostly negative relationships for mean patch size (9 out of 10) and mostly positive for natural area (15 out of 16) (Table 35-Table 38). 103 Figure 36: Linear relationships of rice paddy percentage at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means. 104 Figure 37: Linear relationships of soybean percentage at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means. 105 Figure 38: Linear relationships of corn percentage at the 1000m extent and the habitat, diet, foraging behavior, and migration community weighted means. 106 4.4 Discussion This chapter examined the relationship between landscape structure (compositional and configurational) and bird diversity using both taxonomic and functional measures of biodiversity across multiple scales. For percent crop area, both soybean and corn had significant negative correlations with taxonomic and functional diversity while paddy rice had no significant relationships. This implies that landscapes with more rain-fed crop area (i.e., fields that are not flooded like with paddy rice) support fewer bird species/functional groups. This is further supported by studies reporting high levels of species diversity associated with paddy rice farming practices (Amira et al., 2018) and highlights the importance of considering crops separately as the resources they provide can differ by species/taxa (Gil-Tena et al., 2015). Soybean area positively correlated with some functional traits of birds. Soybean area was significantly positively correlated with ground nesting, while corn was negatively correlated (rice was not significant). Of the 30 families observed in our study, 40% have the ground nesting trait meaning soybean cultivation may be conducive to the breeding ecology for many species in this agroecosystem. Furthermore, there may be population-level consequences for these species if soybean area loss limits the amount of nesting habitat past a certain point. Additionally, both migratory (i.e., species that pass through during migration) and non-migratory traits positively correlated with soybean area, while summer migrants were positively correlated with rice area. In the context of these two important life history strategies, both soybean and rice area positively impact different bird groups, while corn decreases the frequency of these traits in the bird community. We also found that crop diversity and natural landcover were positively correlated with bird biodiversity. The fact that natural area has a positive relationship with biodiversity is expected, however the significant correlation of crop diversity with both taxonomic and 107 functional biodiversity is notable, as not all studies agree with this finding (Martin et al., 2020; Redlich et al., 2018). However there are a number of studies that produced similar results (M. B. Lee & Goodale, 2018; D. Li et al., 2020; Lindsay et al., 2013). A possible explanation for this discrepancy is the presence of paddy rice in our study region as it is able to support species that fields of rain-fed crops cannot (e.g., herons, ducks). Crop diversity also had several positive correlations with various functional traits which could have implications for various ecosystem services (Figure 35, Table 35-Table 38). For example, CWM for small mammal diet preference increased with crop diversity, which could help control local rodent populations (Fischer et al., 2018; Pejchar et al., 2018; O. M. Smith et al., 2021). Furthermore, five of the six significant traits were those associated with a carnivorous diet (e.g., small birds, carrion). Similar to soybean percentage, crop diversity was also significantly correlated with the ground nesting, migratory-route, and resident functional traits. Higher values of edge metrics (total edge and edge density) were correlated with biodiversity (both taxonomic and functional) while higher levels of aggregation were associated with lower biodiversity. This indicates that landscapes dominated by a small number of landcover classes do not support high levels of bird diversity, similar to our finding on crop diversity. It is interesting that both edge metrics were found to be important for biodiversity but not mean patch size despite the two being closely related (i.e., smaller patch size increases edge metrics) as well as other studies finding field size significantly related to species richness (Martin et al., 2020). However, the study by Martin et al. (2020) did not include natural landcover in their analysis. The semi-natural strips that occur in field margins are known to be beneficial to both bird and general biodiversity in a landscape (Blaix & Moonen, 2020; Kujawa et al., 2020; Meichtry-Stier et al., 2018). This disagreement may be evidence that semi-natural field edges 108 could mediate the effect of patch size on farmland biodiversity. However, it should be noted that we limited the potential effect of patch size by using a maximum extent of 1000m. In order to understand the impacts of global trade on Heilongjiang’s agroecosystem, this chapter links landcover change (i.e., soybean area loss) to landscape structure and bird biodiversity. From 2000-2018, cheap soybean imports drove many Chinese soybean farmers to switch to more profitable crops in our study region (Gale et al., 2019; X. Li et al., 2021; J. Sun et al., 2018). Using soybean area decrease as a proxy for global trade impact, we compared landscape and biodiversity metrics in two regions of Heilongjiang that experienced a decline in soybean area, with one exhibiting a greater degree of loss than the other. The samples in the more impacted region (Jiamusi) had significantly lower levels of crop diversity (i.e., number of crops and evenness of crop area). Additionally, the Jiamusi samples had less natural landcover percentage and higher rice percentage. However, the two regions had the same levels of both functional and taxonomic biodiversity. While the difference in crop diversity can be linked to the differing amounts of soybean area loss, the relationship between the amount of natural landcover and global trade is not as clear. From 2000-2018, cropland in Heilongjiang expanded substantially with most of the land being converted from wetland, grassland, and forests. However, the intensity of the expansion was not the same between our two case study prefectures. The more highly impacted region in our study, Jiamusi, is located in the Sanjiang plain, which experienced higher rates of cropland expansion (increase of 6.8%). In the Songnen plain where Qiqihar (the less impacted area) is located, cropland increased by 3.56% (Chen et al., 2022). The factors driving this expansion are linked to a number of factors including food security, economic pursuit, and national policies (Chen et al., 2022). The degree to which global soybean imports effects these factors is not yet clear. For example, the demand for the high-quality rice grown in Heilongjiang has recently 109 boomed within China while the profitability of soybeans has decreased due to global trade, pushing farmers towards rice cultivation rather than soybean. However, uncovering the causal link between these socioeconomic factors and landcover change is challenging and warrants further research. In the context of global trade, we observed that highly impacted regions (i.e., more soybean loss) had lower levels of crop diversity as well as lower levels of natural landcover. While we cannot say whether the difference of natural landcover amount is related to global trade, we would expect a region with less natural landcover and lower levels of crop diversity to have lower biodiversity based on the relationships we found. However, we did not find a significant difference for either taxonomic or functional biodiversity between the degrees of global trade impact. Therefore, it is possible that the higher amount of rice in the highly impacted region mitigated the expected decline in biodiversity due to global trade. In the typical heterogeneous landscape of Heilongjiang farmland, the conversion of soybean fields to another crop due to low prices, decreased the crop diversity of the landscape (i.e., reduced crop richness and/or evenness). However, if they were converted to rice instead of corn, the impact to bird communities was less detrimental. In general, we saw little to no discernable differences in the significance, direction, or strength of the relationships when using taxonomic versus functional indicators (e.g., species richness versus functional group richness). This finding does not agree with similar studies that do report differing relationships of taxonomic/functional diversity and landscape structure (Birkhofer et al., 2018; Flynn et al., 2009; M.-B. Lee & Martin, 2017; Morelli et al., 2018). This disparity could be related to the relatively small functional space due to environmental filtering inherent in our study region (Barragán et al., 2011; Duflot et al., 2014). Extending the amount of 110 time spent sampling (particularly for the abundance-weighted indices) may help reduce this issue. A quick note on the impact of scale. We found that some relationships became significant at different scales in some cases. However, we also had many instances where scale did not matter (i.e., all scales were either significant or insignificant). Therefore, if researchers wish to use landscape metrics to model biodiversity, we suggest exploring a number of extents to select the appropriate extent to use (while also accounting for collinearity among landscape variables). There were some additional limitations to this study. First, there is a need for data from across the seasons. For example, this region serves as an important migratory route for many species, including some cranes (Austin et al., 2018). Exploring these landscape-biodiversity relationships during the harvest/fall season may uncover important temporal dynamics associated this agroecosystem (M.-B. Lee & Martin, 2017; Vasseur et al., 2013). Additionally, collecting more data across a wider range of landscape metric values (e.g., natural landcover) may help to confirm these relationships as well as performing additional analyses that include multiple variables and account for imperfect species detections. 4.5 Conclusions This chapter helps contribute to our current understanding of how landscape structure mediates farmland biodiversity, which is necessary to promote the multi-functionality of agricultural systems. We found that many aspects of landscape structure can influence both taxonomic and functional biodiversity. Additionally, the fact that our findings are not in complete agreement with some existing studies suggest that these relationships are dependent on the characteristics of a given agroecosystem, and that broad generalizations are not possible using these methods alone. Finally, we show that global trade impacted regions growing 111 soybeans by decreasing crop diversity. This could translate to negative consequences for biodiversity but may be mitigated by rice cultivation. 112 APPENDIX 113 Table 33: Table of foraging behavior definitions (M.-B. Lee & Martin, 2017; Remsen & Robinson, 1990; Ruxton & Hansell, 2011) Behavior Definition stalking walking slowly or standing, waiting for prey/food to be within reach bark gleaning plucking or picking food from bark surface and underneath hovering aerial maneuver where bird maintains location while searching for prey, dives onto prey from aerial position hawking catching and feeding on prey in the air scanning watching an area before attacking prey foliage gleaning plucking or picking food from surface of leaves and other foliage continuous aerial feeding similar to hawking, but flies for much longer periods of time ground foraging plucking or picking food from the ground surface sallying catching prey in the air, then returning to perch to feed diving specific to aquatic locations, bird dives below surface for food (may dive from water surface or from the air) dabbling specific to aquatic locations, bird submerges head/neck below the surface to feed probing inserting bill into soft substrate to capture food underneath 114 Table 34: Trait matrix for families observed in study area (n=30). Mass Habitat Preference Diet Forage Behavior Nesting & Migration Avg. Mass (g) Wetland Water Grass Open Woodland Forest Forest Edge Insects Fish Small mammals Mass Low (g) Mass High (g) Small birds Other invertebrates Grain & seeds Nestlings & eggs Small herps Plant matter Fruits and berries Carrion Stalking Bark gleaning Hovering Hawking Scanning Foliage Gleaning Continuous Aerial Ground Sallying Diving Dabbling Probing Ground Nester Summer Winter Resident Migration route Scarce Migratory Acrocephalidae 18.25 29.5 23.88 1 0 1 1 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 1 0 0 0 0 1 Alcedinidae 19 40 29.5 1 1 0 0 0 0 1 1 0 0 1 0 0 0 1 1 0 0 0 1 0 1 0 0 0 1 1 0 0 1 1 0 1 0 0 1 Anatidae 750 1500 1125 1 1 0 0 0 0 1 1 0 0 1 1 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0 0 0 1 Ardeidae 649.4 1421.6 1035.5 1 1 0 0 0 0 1 1 1 1 1 0 1 1 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 Campephagidae 20 28 24 0 0 0 1 1 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 1 Charadriidae 80.5 665 372.75 1 1 1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 1 0 0 1 0 1 Columbidae 144.75 267 205.88 0 0 1 1 1 1 0 0 0 0 1 1 0 0 1 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1 0 0 1 Corvidae 245.33 354 299.67 0 0 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1 0 0 0 0 1 1 0 1 0 0 0 0 0 0 0 1 0 0 0 Cuculidae 106 133 119.5 1 0 1 1 1 1 1 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 1 Emberizidae 14 25.4 19.7 1 0 1 1 1 1 1 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 1 1 0 1 1 0 1 Falconidae 118.75 263.25 191 1 0 1 1 0 1 1 0 1 1 1 0 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 1 0 1 0 0 1 Hirundinidae 17 27.33 22.167 1 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 1 0 0 0 0 1 Laniidae 44.67 79 61.833 1 0 1 1 1 1 1 0 1 1 1 0 1 1 0 0 0 0 0 1 1 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 Laridae 69.5 112.5 91 1 1 0 0 0 0 1 1 0 0 1 0 0 1 0 0 0 0 0 1 1 0 0 1 0 0 1 0 0 1 1 0 0 1 0 1 Motacillidae 17.2 27.13 22.167 1 0 1 1 0 1 1 0 0 0 1 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1 0 0 1 1 1 0 0 0 0 1 Muscicapidae 13 17 15 1 0 1 1 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 0 0 0 0 1 Oriolidae 65 100 82.5 0 0 0 1 1 1 1 0 0 0 1 0 1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 1 Panuridae 12 18 15 1 0 0 0 0 0 1 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 1 0 1 1 0 0 0 0 1 0 0 0 0 1 Paridae 2 15 8.5 0 0 0 1 1 0 1 0 0 0 1 1 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 1 0 0 1 0 0 0 Passeridae 17 30 23.5 0 0 1 1 0 0 1 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 Phalacrocoracidae 1810 2810 2310 1 1 0 0 0 0 0 1 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 1 Phasianidae 545 1990 1267.5 1 0 1 1 0 1 1 0 0 0 0 1 0 0 1 1 0 0 0 0 0 0 1 0 1 0 0 0 0 1 0 0 1 0 0 0 Picidae 17 69 43 1 0 0 1 1 1 1 0 0 0 1 1 1 0 1 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 Podicipedidae 130 236 183 1 1 0 0 0 0 1 1 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 0 0 0 0 1 Rallidae 273 516 394.5 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 1 0 1 0 1 1 1 1 1 0 0 0 0 1 Remizidae 7.5 12 9.75 1 0 1 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 1 Sturnidae 55 100 77.5 0 0 1 1 0 0 1 0 0 0 1 1 0 1 1 1 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 Troglodytidae 6 12 9 0 0 0 1 1 0 1 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0 0 0 0 0 1 0 0 0 Turdidae 60 72 66 0 0 0 0 1 1 1 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 0 1 Upupidae 46 89 67.5 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0 0 1 115 Figure 39: Taxonomic richness (n.sp, red) and functional group richness (FGR, blue) relationships with mean patch size (left) and percent rice (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 116 Figure 40: Shannon-Weiner (shannon, red) and functional dispersal (FDis, blue) relationships with mean patch size (left) and percent rice (right). Pearson’s correlation (R) and p-value (p) for each relationship is also included in the top left corner of each plot. 117 Table 35: Pearson’s correlation values between CWM for mass and habitat preference and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. R.AvgMassGrams p.AvgMassGrams R.Open_Woodland p.Open_Woodland R.Forest_Edge p.Forest_Edge R.Wetland p.Wetland R.Water p.Water R.Grassland p.Grassland R.Forest p.Forest msidi_crop_250m 0.21 0.01 -0.08 0.29 0.11 0.15 -0.01 0.95 0.01 0.86 0.14 0.05 0.05 0.49 msidi_crop_500m 0.15 0.06 -0.07 0.36 0.07 0.33 -0.02 0.82 0.04 0.58 0.15 0.03 0.06 0.45 msidi_crop_1000m 0.10 0.20 -0.05 0.51 -0.01 0.90 -0.04 0.57 0.01 0.88 0.13 0.07 0.02 0.74 mn_patch_250m -0.08 0.30 0.04 0.58 -0.11 0.13 0.04 0.61 0.05 0.48 -0.01 0.87 0.05 0.50 mn_patch_500m -0.08 0.32 0.05 0.51 -0.08 0.25 0.03 0.70 0.00 0.98 0.00 0.95 0.04 0.58 mn_patch_1000m -0.04 0.65 0.05 0.54 0.00 0.97 -0.01 0.88 0.05 0.50 -0.05 0.52 0.06 0.41 ai_250m -0.14 0.09 0.07 0.35 -0.10 0.17 -0.01 0.88 -0.03 0.63 -0.01 0.91 0.00 0.99 ai_500m -0.17 0.04 0.11 0.14 -0.12 0.09 0.01 0.91 -0.04 0.63 0.01 0.87 0.02 0.75 ai_1000m -0.12 0.13 0.05 0.48 -0.01 0.88 -0.02 0.73 0.00 0.95 0.02 0.81 0.03 0.71 te_250m 0.16 0.05 -0.05 0.46 0.13 0.07 0.02 0.83 0.04 0.55 0.03 0.65 0.02 0.81 te_500m 0.17 0.03 -0.10 0.18 0.15 0.04 0.00 0.99 0.04 0.57 0.01 0.91 -0.02 0.80 te_1000m 0.14 0.09 -0.05 0.48 0.02 0.75 0.03 0.72 0.00 0.97 0.00 0.98 -0.03 0.72 ed_250m 0.16 0.05 -0.05 0.46 0.13 0.08 0.01 0.85 0.04 0.55 0.04 0.63 0.02 0.80 ed_500m 0.17 0.03 -0.10 0.19 0.15 0.05 0.00 0.99 0.04 0.60 0.01 0.91 -0.02 0.79 ed_1000m 0.13 0.10 -0.05 0.48 0.02 0.76 0.02 0.74 0.00 0.95 0.00 0.98 -0.03 0.71 p_nat_250m 0.05 0.55 0.07 0.37 0.03 0.70 0.06 0.39 0.05 0.53 0.19 0.01 0.08 0.28 p_nat_500m 0.02 0.79 0.05 0.49 0.02 0.79 0.02 0.74 0.00 0.99 0.17 0.02 0.03 0.65 p_nat_1000m -0.07 0.39 0.09 0.20 -0.03 0.70 0.05 0.53 -0.05 0.45 0.14 0.06 -0.02 0.83 p_soy_250m -0.03 0.67 -0.11 0.14 -0.05 0.51 -0.04 0.63 0.02 0.83 0.19 0.01 -0.04 0.57 p_soy_500m 0.00 0.98 -0.16 0.03 -0.05 0.52 -0.07 0.34 0.01 0.93 0.14 0.06 -0.07 0.34 p_soy_1000m 0.04 0.65 -0.21 0.00 -0.07 0.36 -0.11 0.12 0.01 0.93 0.14 0.06 -0.08 0.29 p_corn_250m 0.22 0.01 -0.35 0.00 -0.16 0.03 -0.25 0.00 -0.25 0.00 -0.09 0.20 -0.29 0.00 p_corn_500m 0.24 0.00 -0.35 0.00 -0.13 0.07 -0.23 0.00 -0.23 0.00 -0.02 0.82 -0.25 0.00 p_corn_1000m 0.21 0.01 -0.37 0.00 -0.13 0.08 -0.24 0.00 -0.21 0.00 -0.02 0.75 -0.25 0.00 p_rice_250m -0.20 0.01 0.32 0.00 0.11 0.13 0.17 0.02 0.16 0.03 -0.18 0.02 0.23 0.00 p_rice_500m -0.19 0.02 0.34 0.00 0.10 0.18 0.18 0.01 0.19 0.01 -0.16 0.03 0.26 0.00 p_rice_1000m -0.16 0.05 0.35 0.00 0.14 0.06 0.17 0.02 0.19 0.01 -0.15 0.04 0.26 0.00 118 Table 36: Pearson’s correlation values between CWM for diet preference and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. R.Small_Mammals p.Small_Mammals R.Small_Birds p.Small_Birds R.Invertebrates p.Invertebrates R.Seeds & Grains p.Seeds & Grains R.Nestlings/Eggs p.Nestlings/Eggs R.Plant Matter p.Plant Matter R.Fruits & Berries p.Fruits & Berries R.Insects p.Insects R.Fish p.Fish R.Herps p.Herps R.Carrion p.Carrion msidi_crop_250m -0.03 0.73 0.04 0.57 0.20 0.01 0.20 0.01 0.03 0.67 0.07 0.37 0.22 0.00 0.21 0.00 0.05 0.46 0.19 0.01 0.20 0.01 msidi_crop_500m -0.01 0.86 0.02 0.79 0.18 0.02 0.18 0.02 0.05 0.48 0.11 0.12 0.19 0.01 0.18 0.01 0.10 0.15 0.13 0.09 0.15 0.03 msidi_crop_1000m -0.01 0.91 -0.04 0.55 0.12 0.11 0.12 0.11 0.00 0.96 0.10 0.16 0.12 0.10 0.12 0.09 0.10 0.16 0.06 0.41 0.09 0.20 mn_patch_250m -0.02 0.80 -0.10 0.19 -0.19 0.01 -0.19 0.01 0.02 0.80 0.03 0.67 -0.21 0.00 -0.20 0.01 0.11 0.15 -0.06 0.42 -0.13 0.07 mn_patch_500m 0.01 0.94 -0.07 0.36 -0.09 0.21 -0.09 0.21 -0.02 0.83 -0.02 0.75 -0.10 0.18 -0.10 0.16 0.02 0.84 -0.03 0.71 -0.02 0.84 mn_patch_1000m 0.03 0.72 0.04 0.59 -0.09 0.20 -0.09 0.20 0.04 0.56 0.00 0.99 -0.08 0.27 -0.10 0.16 0.01 0.93 -0.04 0.61 -0.01 0.92 ai_250m -0.04 0.55 -0.08 0.26 -0.19 0.01 -0.19 0.01 -0.07 0.37 0.00 0.98 -0.23 0.00 -0.20 0.00 0.12 0.09 -0.12 0.09 -0.14 0.05 ai_500m -0.02 0.78 -0.09 0.24 -0.23 0.00 -0.23 0.00 -0.07 0.36 -0.01 0.87 -0.23 0.00 -0.24 0.00 0.12 0.09 -0.10 0.19 -0.13 0.07 ai_1000m -0.03 0.70 0.02 0.80 -0.19 0.01 -0.19 0.01 0.00 0.99 0.03 0.67 -0.18 0.02 -0.20 0.01 0.13 0.07 -0.07 0.37 -0.10 0.18 te_250m 0.05 0.46 0.10 0.16 0.21 0.00 0.21 0.00 0.08 0.26 0.01 0.86 0.25 0.00 0.22 0.00 -0.09 0.20 0.13 0.07 0.15 0.04 te_500m 0.04 0.63 0.11 0.15 0.23 0.00 0.23 0.00 0.08 0.28 0.03 0.67 0.24 0.00 0.25 0.00 -0.10 0.18 0.10 0.16 0.14 0.05 te_1000m 0.03 0.65 -0.01 0.92 0.20 0.01 0.20 0.01 0.00 0.97 -0.01 0.87 0.18 0.01 0.21 0.00 -0.11 0.13 0.08 0.30 0.11 0.14 ed_250m 0.05 0.47 0.10 0.16 0.21 0.00 0.21 0.00 0.08 0.26 0.01 0.86 0.25 0.00 0.23 0.00 -0.09 0.21 0.13 0.07 0.15 0.04 ed_500m 0.03 0.64 0.10 0.15 0.23 0.00 0.23 0.00 0.08 0.30 0.03 0.70 0.24 0.00 0.25 0.00 -0.10 0.18 0.10 0.17 0.14 0.06 ed_1000m 0.03 0.67 -0.01 0.91 0.20 0.01 0.20 0.01 0.00 0.99 -0.01 0.85 0.18 0.01 0.21 0.00 -0.11 0.13 0.07 0.31 0.10 0.15 p_nat_250m 0.08 0.25 0.04 0.55 0.10 0.18 0.10 0.18 0.08 0.27 0.03 0.67 0.11 0.12 0.11 0.12 0.12 0.11 0.10 0.16 0.00 0.95 p_nat_500m 0.04 0.55 0.03 0.69 0.03 0.65 0.03 0.65 0.03 0.65 0.03 0.69 0.04 0.55 0.05 0.53 0.14 0.05 0.09 0.22 -0.05 0.50 p_nat_1000m 0.06 0.40 -0.01 0.93 -0.03 0.65 -0.03 0.65 -0.03 0.66 0.03 0.68 -0.02 0.82 -0.03 0.71 0.15 0.04 0.03 0.64 -0.09 0.21 p_soy_250m -0.02 0.77 -0.06 0.42 -0.01 0.88 -0.01 0.88 0.01 0.93 0.25 0.00 -0.02 0.77 -0.02 0.84 0.20 0.01 0.01 0.89 0.06 0.39 p_soy_500m -0.05 0.50 -0.06 0.43 0.02 0.78 0.02 0.78 0.00 0.97 0.23 0.00 0.01 0.93 0.02 0.81 0.14 0.06 0.00 0.97 0.07 0.31 p_soy_1000m -0.10 0.18 -0.07 0.34 0.06 0.38 0.06 0.38 -0.01 0.93 0.21 0.00 0.04 0.56 0.06 0.39 0.11 0.14 0.03 0.71 0.08 0.27 p_corn_250m -0.27 0.00 -0.13 0.08 0.06 0.38 0.06 0.38 -0.30 0.00 0.04 0.58 0.08 0.26 0.05 0.47 -0.18 0.01 0.12 0.11 0.11 0.15 p_corn_500m -0.28 0.00 -0.12 0.09 0.10 0.17 0.10 0.17 -0.28 0.00 0.06 0.40 0.12 0.10 0.09 0.22 -0.12 0.09 0.15 0.04 0.12 0.09 p_corn_1000m -0.31 0.00 -0.14 0.06 0.06 0.38 0.06 0.38 -0.25 0.00 0.06 0.41 0.08 0.29 0.05 0.48 -0.10 0.16 0.13 0.09 0.08 0.27 p_rice_250m 0.17 0.02 0.10 0.18 -0.10 0.18 -0.10 0.18 0.19 0.01 -0.26 0.00 -0.12 0.10 -0.10 0.17 -0.08 0.27 -0.19 0.01 -0.14 0.06 p_rice_500m 0.19 0.01 0.09 0.21 -0.11 0.15 -0.11 0.15 0.21 0.00 -0.24 0.00 -0.12 0.10 -0.11 0.14 -0.07 0.37 -0.18 0.01 -0.13 0.08 p_rice_1000m 0.21 0.00 0.13 0.08 -0.08 0.28 -0.08 0.28 0.21 0.00 -0.24 0.00 -0.09 0.22 -0.08 0.27 -0.06 0.40 -0.17 0.02 -0.10 0.19 119 Table 37: Pearson’s correlation values between CWM for foraging behavior and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. R.Bark Gleaning p.Bark Gleaning R.Foliage_Gleaning p.Foliage_Gleaning R.Continuous_Aerial p.Continuous_Aerial R.Ground_Foraging p.Ground_Foraging R.Stalking p.Stalking R.Hovering p.Hovering R.Hawking p.Hawking R.Scanning p.Scanning R.Sallying p.Sallying R.Diving p.Diving R.Dabbling p.Dabbling R.Probing p.Probing msidi_crop_250m -0.02 0.76 0.09 0.20 0.04 0.58 -0.09 0.24 0.24 0.00 0.01 0.95 -0.02 0.79 -0.07 0.35 -0.04 0.57 0.04 0.61 0.05 0.50 0.04 0.56 msidi_crop_500m -0.02 0.81 0.14 0.06 0.06 0.39 -0.10 0.18 0.21 0.00 0.00 0.95 -0.06 0.40 -0.07 0.33 -0.04 0.60 0.01 0.86 0.03 0.70 0.05 0.52 msidi_crop_1000m 0.02 0.78 0.11 0.14 0.05 0.49 -0.12 0.09 0.16 0.03 -0.04 0.63 -0.03 0.69 -0.08 0.26 -0.02 0.76 -0.06 0.40 -0.05 0.51 -0.01 0.93 mn_patch_250m -0.04 0.59 -0.03 0.73 -0.14 0.06 0.04 0.58 -0.16 0.03 0.11 0.14 -0.03 0.63 0.07 0.32 -0.10 0.18 -0.08 0.27 -0.07 0.34 -0.09 0.22 mn_patch_500m -0.04 0.61 -0.02 0.79 -0.13 0.07 0.04 0.61 -0.05 0.46 0.10 0.16 0.02 0.77 0.05 0.47 -0.10 0.18 -0.05 0.48 -0.04 0.55 -0.08 0.27 mn_patch_1000m 0.05 0.51 0.09 0.20 -0.16 0.03 0.09 0.23 -0.09 0.23 0.16 0.03 -0.10 0.17 0.01 0.87 -0.07 0.35 0.04 0.59 0.04 0.56 -0.06 0.41 ai_250m -0.11 0.15 -0.06 0.42 -0.11 0.14 -0.02 0.76 -0.17 0.02 0.04 0.57 0.01 0.87 0.03 0.67 -0.07 0.33 -0.04 0.57 -0.02 0.76 -0.02 0.74 ai_500m -0.08 0.29 0.01 0.92 -0.17 0.02 -0.01 0.91 -0.20 0.01 0.12 0.10 0.05 0.53 0.05 0.46 -0.08 0.25 -0.05 0.52 -0.05 0.52 -0.07 0.35 ai_1000m -0.01 0.88 0.04 0.63 -0.18 0.01 -0.01 0.92 -0.20 0.01 0.13 0.07 -0.06 0.41 0.01 0.84 -0.03 0.70 0.04 0.60 0.03 0.64 -0.04 0.61 te_250m 0.11 0.15 0.08 0.30 0.12 0.09 0.02 0.78 0.19 0.01 -0.03 0.68 -0.02 0.81 -0.03 0.65 0.07 0.31 0.06 0.41 0.04 0.60 0.04 0.60 te_500m 0.08 0.26 0.02 0.84 0.18 0.01 0.00 0.96 0.21 0.00 -0.11 0.12 -0.04 0.57 -0.05 0.52 0.09 0.23 0.07 0.37 0.06 0.40 0.09 0.24 te_1000m 0.02 0.81 -0.02 0.77 0.18 0.01 -0.01 0.91 0.21 0.00 -0.13 0.09 0.06 0.41 -0.01 0.86 0.02 0.77 -0.03 0.69 -0.03 0.71 0.04 0.54 ed_250m 0.11 0.14 0.08 0.29 0.13 0.08 0.02 0.79 0.19 0.01 -0.03 0.67 -0.02 0.79 -0.04 0.63 0.08 0.30 0.06 0.42 0.04 0.60 0.04 0.60 ed_500m 0.08 0.27 0.01 0.84 0.18 0.01 0.00 0.95 0.20 0.00 -0.12 0.12 -0.04 0.59 -0.05 0.51 0.09 0.23 0.06 0.38 0.06 0.41 0.08 0.25 ed_1000m 0.02 0.80 -0.02 0.78 0.18 0.01 -0.01 0.91 0.21 0.00 -0.13 0.08 0.06 0.41 -0.02 0.83 0.02 0.75 -0.03 0.68 -0.03 0.70 0.04 0.55 p_nat_250m 0.06 0.42 -0.01 0.93 0.15 0.05 0.00 0.98 0.04 0.63 0.04 0.59 0.08 0.28 0.09 0.23 0.16 0.03 0.01 0.87 -0.02 0.84 -0.04 0.60 p_nat_500m 0.02 0.74 -0.01 0.94 0.13 0.08 -0.07 0.34 -0.03 0.70 -0.02 0.80 0.10 0.16 0.06 0.41 0.23 0.00 0.01 0.91 -0.02 0.79 0.00 0.98 p_nat_1000m -0.03 0.68 -0.01 0.86 0.09 0.20 -0.16 0.03 -0.06 0.43 -0.08 0.26 0.19 0.01 0.04 0.58 0.21 0.00 -0.01 0.92 -0.03 0.69 0.00 0.95 p_soy_250m -0.05 0.50 0.00 0.99 -0.07 0.35 -0.29 0.00 0.04 0.56 -0.06 0.38 -0.05 0.47 -0.04 0.55 0.02 0.81 -0.05 0.51 -0.04 0.61 0.01 0.85 p_soy_500m -0.05 0.53 0.00 0.97 -0.04 0.56 -0.27 0.00 0.09 0.24 -0.10 0.17 -0.08 0.26 -0.09 0.21 -0.02 0.83 -0.05 0.52 -0.04 0.56 0.00 0.98 p_soy_1000m -0.03 0.63 -0.01 0.87 0.01 0.84 -0.25 0.00 0.12 0.12 -0.12 0.10 -0.11 0.12 -0.12 0.10 -0.01 0.94 -0.07 0.36 -0.06 0.43 -0.02 0.81 p_corn_250m -0.10 0.16 0.06 0.40 -0.01 0.86 -0.31 0.00 0.16 0.03 -0.29 0.00 0.02 0.82 -0.32 0.00 -0.10 0.16 -0.10 0.16 -0.09 0.23 -0.07 0.34 p_corn_500m -0.11 0.15 0.01 0.85 0.00 0.95 -0.31 0.00 0.17 0.02 -0.25 0.00 0.00 0.96 -0.30 0.00 -0.10 0.16 -0.10 0.19 -0.08 0.29 -0.08 0.29 p_corn_1000m -0.09 0.23 0.01 0.90 -0.02 0.83 -0.29 0.00 0.13 0.07 -0.24 0.00 -0.03 0.64 -0.30 0.00 -0.12 0.09 -0.12 0.12 -0.10 0.19 -0.10 0.18 p_rice_250m 0.05 0.54 -0.04 0.60 -0.02 0.78 0.49 0.00 -0.18 0.01 0.26 0.00 -0.04 0.55 0.21 0.00 -0.02 0.83 0.10 0.18 0.10 0.19 0.01 0.84 p_rice_500m 0.08 0.30 0.00 0.98 -0.05 0.53 0.50 0.00 -0.18 0.02 0.30 0.00 -0.06 0.45 0.23 0.00 -0.04 0.63 0.08 0.25 0.09 0.22 0.02 0.83 p_rice_1000m 0.08 0.25 0.03 0.70 -0.05 0.53 0.50 0.00 -0.17 0.02 0.31 0.00 -0.06 0.39 0.23 0.00 -0.02 0.76 0.12 0.10 0.13 0.07 0.05 0.47 120 Table 38: Pearson’s correlation values between CWM for the ground nesting trait/migration status and landscape structure metrics. Positive and negative correlation values >|0.1| are highlighted in green and red, respectively. P-values ≤0.1 are highlighted in blue. R.Ground_Nester p.Ground_Nester R.Migration_Route p.Migration_Route R.Summer p.Summer R.Winter p.Winter R.Resident p.Resident R.Scarce p.Scarce msidi_crop_250m 0.01 0.93 -0.05 0.50 NA NA 0.13 0.09 0.03 0.70 0.06 0.41 msidi_crop_500m 0.09 0.20 -0.07 0.34 NA NA 0.18 0.01 0.13 0.07 0.07 0.32 msidi_crop_1000m 0.14 0.05 -0.07 0.34 NA NA 0.16 0.03 0.18 0.01 0.05 0.48 mn_patch_250m 0.06 0.44 0.06 0.41 NA NA 0.01 0.88 0.09 0.22 -0.05 0.49 mn_patch_500m 0.00 0.96 0.05 0.48 NA NA -0.04 0.55 0.03 0.66 -0.05 0.53 mn_patch_1000m 0.02 0.82 0.00 0.96 NA NA -0.06 0.45 0.00 0.95 -0.07 0.31 ai_250m 0.16 0.03 0.06 0.44 NA NA -0.02 0.79 0.19 0.01 -0.09 0.20 ai_500m 0.12 0.10 0.11 0.15 NA NA -0.07 0.37 0.15 0.05 -0.14 0.05 ai_1000m 0.13 0.09 0.05 0.53 NA NA -0.05 0.53 0.11 0.12 -0.10 0.19 te_250m -0.13 0.07 -0.04 0.57 NA NA 0.04 0.62 -0.16 0.03 0.10 0.18 te_500m -0.10 0.19 -0.10 0.19 NA NA 0.08 0.26 -0.12 0.10 0.14 0.05 te_1000m -0.11 0.13 -0.05 0.51 NA NA 0.06 0.39 -0.10 0.18 0.10 0.17 ed_250m -0.13 0.07 -0.04 0.57 NA NA 0.04 0.61 -0.16 0.03 0.10 0.18 ed_500m -0.10 0.19 -0.09 0.20 NA NA 0.08 0.27 -0.12 0.11 0.14 0.05 ed_1000m -0.11 0.14 -0.05 0.51 NA NA 0.06 0.40 -0.10 0.19 0.10 0.17 p_nat_250m 0.05 0.50 0.15 0.04 NA NA 0.05 0.51 -0.02 0.82 0.03 0.68 p_nat_500m 0.12 0.11 0.13 0.07 NA NA 0.03 0.73 0.00 0.98 0.02 0.77 p_nat_1000m 0.19 0.01 0.16 0.03 NA NA 0.03 0.67 0.07 0.31 -0.02 0.80 p_soy_250m 0.32 0.00 -0.13 0.07 NA NA 0.25 0.00 0.33 0.00 -0.02 0.83 p_soy_500m 0.25 0.00 -0.19 0.01 NA NA 0.25 0.00 0.27 0.00 -0.01 0.92 p_soy_1000m 0.18 0.01 -0.23 0.00 NA NA 0.25 0.00 0.20 0.01 0.00 0.97 p_corn_250m -0.21 0.00 -0.38 0.00 NA NA 0.06 0.42 -0.23 0.00 -0.07 0.36 p_corn_500m -0.20 0.01 -0.34 0.00 NA NA 0.08 0.25 -0.19 0.01 -0.06 0.39 p_corn_1000m -0.21 0.00 -0.34 0.00 NA NA 0.09 0.21 -0.18 0.01 -0.07 0.37 p_rice_250m -0.09 0.20 0.31 0.00 NA NA -0.27 0.00 -0.05 0.52 -0.02 0.74 p_rice_500m -0.07 0.35 0.31 0.00 NA NA -0.26 0.00 -0.02 0.76 0.00 0.97 p_rice_1000m -0.04 0.54 0.30 0.00 NA NA -0.27 0.00 -0.01 0.89 0.03 0.71 121 CHAPTER 5: MODELING BIRD OCCUPANCY IN A CHANGING AGROECOSYSTEM 122 Abstract International food trade drives global land use change, with consequences for wildlife that utilize agroecosystems. Most studies on trade-driven land use change focus on conversion of natural habitats to cropland in countries that produce and export agricultural goods. Importing countries, however, undergo more indirect land use shifts. Imports have substantial effects on domestic crop prices, sometimes resulting in crop switching and abandonment by farmers in the importing country and altering the agricultural landscape. To explore these drivers shaping the landscape of agroecosystems, we analyzed relationships between bird species-level occupancy and agricultural landscape metrics in a region heavily impact by global trade. Our results indicated that increased crop diversity significantly increased occupancy of many birds at taxonomic and functional levels. Crop diversity was particularly important for less commonly surveyed functional groups. Percentage of natural landcover was not as important as expected, while metrics related to landscape configuration had few significant impacts on occupancy. Increases in rice area had fewer negative impacts to bird occupancy compared to increases in corn and soybean. Furthermore, we found more bird groups were negatively impacted by soybean area than corn, though this was more apparent at the taxonomic level. By combining remotely sensed landscape characteristics with bird diversity observations, this study provides a better understanding of how global trade and land use change may ultimately impact agroecosystems. 123 5.1 Introduction 5.1.1 Background Environmental sustainability and food security are two of the greatest global challenges in modern times (Berry et al., 2014; Foley et al., 2005; Godfray & Garnett, 2014; Willett et al., 2019). Up until the mid-20th century, food demands were met by local producers but are now increasingly satisfied through international trade (Porkka et al., 2013; Sandström et al., 2014; V. H. Smith & Glauber, 2020; Tarrant, 1985). Recent research indicates cropland conversions are occurring in importing countries in response to international trade (J. Sun et al., 2015; van Vliet et al., 2015). Farmer land-use decision making is affected by multiple factors such as climate, resources, and crop prices, among others (Lutz, 1998; S. A. Wood et al., 2014). As global trade increases, farmers with smaller operations struggle to compete when markets are flooded with cheap imports (Tilt, 2008). Crop price plays a large role in determining what small-scale farmers choose to grow and whether they plant a single crop type or diversify (Di Falco & Perrings, 2005; Michler & Josephson, 2017; D. Wood & Lenné, 1999). Additionally, as small-scale farmers in developing countries become integrated in a global commodity market, they must consider risk and market uncertainty more than ever before (Hao, 2010). Responses of farmers to global markets contribute to socioeconomic drivers that ultimately shape landscapes of agroecosystems. Research that links global trade, changes in landcover/use, and local biodiversity in the same context is lacking. Due to challenges associated with linking social and ecological data, many studies default to measuring landcover change alone or using a species area relationship to evaluate biodiversity impacts (Chaudhary & Brooks, 2019; Estrada et al., 2019; Newbold et al., 2015; Sala et al., 2000). While informative at broad scales, this approach often over-simplifies 124 the complex relationships species have with their habitat at the local scale (Souza et al., 2015). Moreover, many studies on landcover change focus on destruction of “natural” habitat to assess impacts on biodiversity. Subsequently, the biodiversity potential of farmland is often overlooked (Norris, 2008). The level of biodiversity that can be sustained in a predominantly agrarian region depends on farming practices that shape the landscape. For example, soil management (tilling), marginal land care (mowing), and chemical inputs (fertilizer and pesticide application) can all impact suitability of farmland for both animal and plant species (Billeter et al., 2008; Geiger et al., 2010; Humbert et al., 2012; Sapkota et al., 2012). It is also understood that landscape heterogeneity, in general, supports high levels of biodiversity (Fahrig et al., 2011; MacArthur & MacArthur, 1961; Tews et al., 2004) -- though too much heterogeneity can have negative impacts due to fragmentation or inadequate patch sizes of habitat (Allouche et al., 2012). Studies in agriculture have found heterogeneity across all landcover classes, including semi-natural, natural, built-up, and other features can positively effect biodiversity (Benton et al., 2003; Firbank et al., 2008; M.-B. Lee & Martin, 2017). However, those measures of heterogeneity are often outside the scope of conservation strategies that could be employed to increased farmland biodiversity. In terms of landscape structure, farmers have control on the types of crops they grow, where they grow them on their land, and their management of marginal land/field edges (Aguilar et al., 2015; Hayden et al., 2021; Jupiter, 2020, 2020; Preissel et al., 2017). Unless they decide to take land out of production (which may not be ideal for either the farmer’s bottom line or food security (Garibaldi et al., 2017; Kremen & Merenlender, 2018; Tscharntke et al., 2012)), their control over other aspects of the landscape is restricted. However, if relationships between biodiversity and actionable landscape metrics (e.g., percentage of natural area to inform land 125 sparing amounts, crop diversity to inform how many crops to grow in a given area) were understood, they could be more easily translated to conservation strategies focused on farmers. 5.1.2 Research Goal The goal of this study is to assess relationships between landscape structure (i.e., composition and configuration) and bird diversity in a region highly impacted by global trade. Specifically, this study looked at relationships between bird occupancy and landscape metrics using a multi-species occupancy modeling approach and assessed occupancy of taxonomic and functional bird groups. By gathering empirical wildlife data and linking it with landscape metrics, this chapter bridges the gap between landcover and a metric of local biodiversity that often exists in CHANS research and places it into context of the global interconnected CHANS encompassing the soybean trade. 5.2 Methods 5.2.1 Bird Data In the summer of 2017 and 2018, we collected bird biodiversity data at 207 different points across Heilongjiang. These geolocated data were collected using a point count method (Bibby, 2000; Hutto, 1986). All birds seen or heard were recorded within a 50m (2017) or 100m (2018) radius for three, three-minute intervals. The use of temporal intervals allows for the calculation of detection probabilities (Zipkin et al. 2010). If individuals were not identifiable to the species, they were put into a non-specific category (e.g., crow spp, swallow spp). We also evaluated bird occurrence from a functional perspective by grouping each family into 15 different groups based on morphological and life history traits (see chapter 3). The composition of the functional groups was a mixture of single-family groups as well those that combined 126 several similar families. Each group was also given an alias to aid in the interpretation of the results. 5.2.2 Landscape Metrics To portray landcover around each bird survey point, landscape metrics were derived from classified landcover rasters (see chapters 3-4). Nineteen samples were removed from this analysis due to uncertainty in the land cover classification, resulting in a final sample size of 188. All metrics were calculated using the R package landscapemetrics at three different extents: 250m, 500m, and 1000m (Hesselbarth et al., 2019). We calculated seven landscape metrics to use as predictors of bird occupancy in our model (Table 39). The landscape-level metrics we selected for our model included the modified Simpson’s diversity index of crop cover classes (msidi_crop), mean patch size (mn_patch), and aggregation index (ai). At the class level, we chose to include percent natural area, percent soybean, percent corn, and percent rice. These metrics were chosen based on previous work as well as our research goal of evaluating both landscape composition and configuration (Gil-Tena et al., 2015; M. B. Lee & Goodale, 2018; D. Li et al., 2020; Liao et al., 2020; C. Wood et al., 2017). To capture the largest range of variation for each metric, the extent which demonstrated the most variability (i.e., largest distance between end points of a boxplot) was identified for each metric. The correlations between each metric were also calculated to ensure the predictors in the model were not severely colinear (i.e., correlation > 0.7). 127 Table 39: Landscape metrics selected for model. The ‘Reclass’ input refers the landcover raster that merged crop classes (soybean, corn, rice, and other) into one ‘cropland’ class and the natural classes (forest, grassland, wetland, and water) into one ‘natural’ class. Metric Unit Level Type Input MSIDI none landscape diversity Crop classes only AREA_MN hectares landscape area & edge All classes AI percentage landscape aggregation All classes PLAND_Natural percentage class area & edge Reclass PLAND_Corn percentage class area & edge All classes PLAND_Soy percentage class area & edge All classes PLAND_Rice percentage class area & edge All classes 5.2.3 Occurrence Model To assess relationship between landscape structure and bird diversity, we utilized a multispecies hierarchical occupancy model (M.-B. Lee & Martin, 2017; Zipkin et al., 2010) (Zipkin et al. 2010). The model estimates occupancy using a Bayesian approach while accounting for imperfect detection. All variables were standardized prior to running the model. Since the radius for point counts was increased from 50m to 100m in 2018, the detection and occurrence means were also calculated separately to account for unequal effort between the years (MacKenzie, 2018). Our model is structured as a typical hierarchical occupancy model with two connected regression models estimating the ecological and detection processes separately (Kéry & Schaub, 2012). The occupancy state for species i at point j is a binary variable where a species occurs at point 𝑗 when 𝑂𝑐𝑐𝑖𝑗 = 1 and 𝑂𝑐𝑐𝑖𝑗 = 0, otherwise. The occurrence state ( 1.1 ) was modeled as the outcome of a Bernoulli random variable where 𝑝𝑂𝑐𝑐𝑖𝑗 is the probability species 𝑖 occurs at point 𝑗. The analysis of the model used data augmentation to allow for estimations of the number of families (or groups) that were unobserved during sampling. As described in Royle et al. 2007 and Zipkin et al. 2010, this is achieved by modifying the occurrence process with wi (Royle, 128 Dorazio, and Link 2007; Zipkin et al. 2010). For the taxonomic and functional level models, the dataset was augmented with 30 and 15 zeroes, respectively. 𝑂𝑐𝑐𝑖𝑗 ~ 𝐵𝑒𝑟𝑛(𝑝𝑂𝑐𝑐𝑖𝑗 ∗ 𝑤𝑖 ) ( 1.1 ) We modeled the detection state similarly ( 1.2 ), with detection being determined by detection probability for species 𝑖 at point 𝑗 during interval 𝑘 and occupancy state (i.e., whether or not species 𝑖 is actually at point 𝑗). 𝐷𝑒𝑡(𝑖, 𝑗, 𝑘)~ 𝐵𝑒𝑟𝑛(𝑝𝐷𝑒𝑡𝑖𝑗𝑘 ∗ 𝑂𝑐𝑐𝑖𝑗 ) ( 1.2 ) Since we expected occurrence and detection probabilities to vary by species, landscape structure, and survey characteristics, a logit link function was used to incorporate these effects. We modeled occurrence probability ( 1.3 ) using a linear combination of crop diversity, aggregation index, mean patch size, percent natural landcover, percent rain-fed crop cover, and percent rice cover as well as longitude to account for underlying species distributions that differed from east to west. The 𝑂𝑐𝑐17𝑖 term is the occurrence probability for species 𝑖, when all other covariates are equal to their mean. The 𝑂𝑐𝑐18𝑖 term exists to modify occurrence probability for samples collected in 2018 (𝐼𝑛𝑑𝑗 = 1). logit(pOccij) = 𝑂𝑐𝑐17𝑖 + 𝑂𝑐𝑐18𝑖 𝐼𝑛𝑑𝑗 + 𝑀𝑆𝐼𝐷𝐼𝑐𝑟𝑜𝑝𝑗 𝛽1𝑖 + 𝑀𝑁𝑝𝑎𝑡𝑐ℎ𝑗 𝛽2𝑖 + 𝐴𝐼𝑗 𝛽3𝑖 + 𝑃𝐿𝐴𝑁𝐷𝑛𝑎𝑡𝑗 𝛽4𝑖 + 𝑃𝐿𝐴𝑁𝐷𝑠𝑜𝑦𝑗 𝛽5𝑖 + 𝑃𝐿𝐴𝑁𝐷𝑐𝑜𝑟𝑛𝑗 𝛽6𝑖 + ( 1.3 ) 𝑃𝐿𝐴𝑁𝐷𝑟𝑖𝑐𝑒𝑗 𝛽7𝑖 + 𝐿𝑂𝑁𝐺𝑗 𝛽8𝑖 For detection probability ( 1.4 ), we included the time of the sample (calculated as the number of minutes after sunrise) as a predictor in addition to the 𝐷𝑒𝑡17𝑖 and 𝐷𝑒𝑡18𝑖 terms. logit(pDetijk) = 𝐷𝑒𝑡17𝑖 + 𝐷𝑒𝑡18𝑖 𝐼𝑛𝑑𝑗 + 𝑠𝑢𝑛𝑟𝑖𝑠𝑒𝑗𝑘 𝛼𝑖 ( 1.4 ) An important characteristic of this modeling approach links the species-level processes to a community-level component, where all species parameters are drawn from a shared distribution. This allows for the species-specific parameters to be treated as random effects that 129 are governed by community hyperparameters. These hyperparameters were assumed to be normally distributed with a parameter-specific mean of µ and standard deviation of σ. Similarly, µ and σ were also specified for all species-level parameters. We conducted a Bayesian analysis of the model using the programs R and WinBUGS (Lunn et al., 2000; R Core Team, 2021). We defined independent, diffuse prior distributions for the community hyperparameters (see appendix for model code) and ran the model with 3 chains of 75,000 iterations after a 75,000 burn-in and thinned by 10. Model convergence was evaluated using the Gelman diagnostic (Brooks & Gelman, 1998). 5.3 Results 5.3.1 Model data A total of 2444 bird observations were made across all 188 point-count locations and intervals, spanning 13 orders and 30 families (Table 40). The majority of species were passerines from the Hirundinidae (swallow), Acrocephalidae (reed warbler), Corvidae (crow/magpie), Passeridae (old world sparrow), and Emberizidae (bunting) families (Table 40). At the species level, 46 unique species were identified. The most frequent group in our sample (group 9) consisted of swallows and reed tits (swallow/reedtit) followed by group 1 (reedwarb/flycather) and group 7 (crow/magpie) (Table 42). 130 Table 40: Frequency of bird order and family in sample Order # Family # Passeriformes 2080 Hirundinidae 727 Columbiformes 104 Acrocephalidae 468 Charadriiformes 60 Corvidae 321 Falconiformes 59 Passeridae 284 Pelicaniformes 43 Emberizidae 118 Anseriformes 28 Columbidae 104 Cuculiformes 28 Falconidae 59 Gruiformes 11 Remizidae 53 Galliformes 9 Laniidae 42 Phoenicopteriformes 4 Ardeidae 41 Piciformes 4 Charadriidae 32 Bucerotiformes 3 Anatidae 28 Coraciiformes 1 Cuculidae 28 Laridae 28 Oriolidae 13 Rallidae 11 Muscicapidae 10 Panuridae 10 Motacillidae 9 Paridae 9 Phasianidae 9 Turdidae 8 Picidae 4 Podicipedidae 4 Sturnidae 3 Troglodytidae 3 Upupidae 3 Campephagidae 2 Phalacrocoracidae 2 Alcedinidae 1 131 Table 41: Functional group composition and species frequency Group Spp Family Order Freq 1 reed_warbler Acrocephalidae Passeriformes 351 1 black_browed_reed_warbler Acrocephalidae Passeriformes 106 1 thick-billed_warbler Acrocephalidae Passeriformes 6 1 oriental_reed_warbler Acrocephalidae Passeriformes 5 1 stonechat Muscicapidae Passeriformes 10 2 common_kingfisher Alcedinidae Coraciiformes 1 3 duck_spp Anatidae Anseriformes 28 3 bittern_spp Ardeidae Pelicaniformes 18 3 great_egret Ardeidae Pelicaniformes 10 3 gray_heron Ardeidae Pelicaniformes 7 3 purple_heron Ardeidae Pelicaniformes 5 3 striated_heron Ardeidae Pelicaniformes 1 4 ashy_minivet Campephagidae Passeriformes 2 4 common_cuckoo Cuculidae Cuculiformes 28 4 black_naped_oriole Oriolidae Passeriformes 13 4 thrush_spp Turdidae Passeriformes 7 4 siberian_thrush Turdidae Passeriformes 1 5 shorebird_spp Charadriidae Charadriiformes 23 5 northern_lapwing Charadriidae Charadriiformes 9 6 dove_spp Columbidae Columbiformes 46 6 pigeon Columbidae Columbiformes 45 6 spotted_dove Columbidae Columbiformes 10 6 oriental_turtle_dove Columbidae Columbiformes 3 6 bunting_spp Emberizidae Passeriformes 76 6 chestnut-earred_bunting Emberizidae Passeriformes 24 6 black_faced_bunting Emberizidae Passeriformes 10 6 meadow_bunting Emberizidae Passeriformes 7 6 reed_bunting Emberizidae Passeriformes 1 6 ring-neck_pheasant Phasianidae Galliformes 9 7 eurasian_magpie Corvidae Passeriformes 308 7 crow_spp Corvidae Passeriformes 12 7 eurasian_jay Corvidae Passeriformes 1 8 eurasian_kestrel Falconidae Falconiformes 36 8 amur_falcon Falconidae Falconiformes 11 8 eurasian_hobby Falconidae Falconiformes 7 8 falcon_spp Falconidae Falconiformes 5 8 brown_shrike Laniidae Passeriformes 33 8 gray_shrike Laniidae Passeriformes 5 8 shrike_spp Laniidae Passeriformes 4 9 swallow_spp Hirundinidae Passeriformes 685 9 red-rumped_swallow Hirundinidae Passeriformes 26 9 barn_swallow Hirundinidae Passeriformes 16 9 bearded_reedling Panuridae Passeriformes 10 9 chinese_penduline_tit Remizidae Passeriformes 53 10 white_wing_tern Laridae Charadriiformes 25 10 tern_spp Laridae Charadriiformes 3 10 great_cormorant Phalacrocoracidae Pelicaniformes 2 10 little_grebe Podicipedidae Phoenicopteriformes 4 11 white_wagtail Motacillidae Passeriformes 4 11 pipit_spp Motacillidae Passeriformes 3 11 olive-backed_pipit Motacillidae Passeriformes 2 12 tit_spp Paridae Passeriformes 9 12 eurasian_tree_sparrow Passeridae Passeriformes 284 12 eurasian_wren Troglodytidae Passeriformes 3 13 woodpecker_spp Picidae Piciformes 4 14 common_moorhen Rallidae Gruiformes 8 14 common_coot Rallidae Gruiformes 2 14 crake_spp Rallidae Gruiformes 1 15 common_starling Sturnidae Passeriformes 3 15 common_hoopoe Upupidae Bucerotiformes 3 132 Table 42: Functional group frequency and group aliases. Group Aliases Group Freq swallow/reedtit group_9 790 reedwarb/flycatcher group_1 478 crow/magpie group_7 321 sparrow/tit group_12 296 dove/bunting/pheasant group_6 231 falcon/shrike group_8 101 heron/duck group_3 69 thrush/cuckoo/oriole group_4 51 tern/grebe group_10 34 shorebird/lapwing group_5 32 rail group_14 11 pipit/wagtail group_11 9 hoopoe/starling group_15 6 woodpecker group_13 4 kingfisher group_2 1 As expected, variation of each landscape metric varied among extents (Table 43). Two metrics had the greatest spread at the largest extent (mean patch and rice percentage). Crop diversity, aggregation, percent corn maximized variation at 500m and the percentage of natural landcover and soybean variation peaked at the smallest extent (250m). For the most part, none of the metrics were highly correlated, however, aggregation and mean patch size were moderately correlated (R = 0..57) as well as corn and rice area (R = -0.59) (Figure 41). Table 43: Data variation for each metric across all extents (250m, 500m, and 1000m). spread 250m 500m 1000m metric (standardized) max 250m 500m 1000m max min max min max min msidi_crop 500 3.35 3.58 3.37 2.43 -0.92 2.53 -1.06 2.07 -1.29 mn_patch 1000 1.36 1.44 3.47 0.48 -0.89 0.69 -0.75 2.24 -1.23 ai 500 4.20 4.42 3.94 1.61 -2.58 2.05 -2.37 1.81 -2.13 pland_natural 250 3.05 2.32 2.53 2.12 -0.93 1.43 -0.89 1.62 -0.91 pland_soybean 250 3.36 2.69 3.17 2.33 -1.03 1.82 -0.88 2.34 -0.83 pland_corn 500 3.38 3.39 3.31 1.84 -1.54 2.09 -1.31 2.15 -1.16 pland_rice 1000 2.95 3.04 3.11 1.43 -1.52 1.73 -1.31 2.02 -1.09 133 Figure 41: Correlogram for metrics selected as model predictors. 134 5.3.2 Taxonomic model results At the community-level, all landscape parameters had 90% Bayesian credible intervals (BCIs) that contained zero (Table 44). Crop diversity (msidi_crop) was the only variable indicating marginal significance, suggesting that greater crop diversity corresponded to higher bird occurrence probabilities (Table 44). Table 44: Estimates of hyperparameters for the taxonomic occurrence model parameter mean sd lower upper occurrence 17 -5.04 2.49 -7.53 -2.56 occurrence 18 0.42 0.93 -0.50 1.35 detection 17 0.15 0.65 -0.50 0.80 detection 18 0.15 0.35 -0.20 0.50 msidi_crop 0.25 0.32 -0.06 0.57 mn_patch -0.07 0.34 -0.41 0.26 ai -0.08 0.39 -0.47 0.31 pland_nat 0.19 0.36 -0.17 0.54 pland_soy -0.34 0.46 -0.80 0.12 pland_corn -0.28 0.47 -0.75 0.18 pland_rice -0.16 0.53 -0.68 0.37 long -0.10 0.65 -0.74 0.55 sunrise -0.04 0.43 -0.47 0.38 The number of bird families with significant and nearly significant coefficient estimates (i.e., BCI crossed zero by <0.1) varied across the different landscape metrics. At the family level, higher crop diversity significantly associated with higher occupancy probabilities for 5 of the 30 families (Figure 43). Additionally, all but one family (Laridae) had positive mean estimates and several additional families had 90% credible intervals that crossed 0 by <0.1 (e.g., Remizidae, Charadriidae, Figure 43). Only one family, Hirundinidae, showed significant effects of mean patch size (positive) and no families were significant for aggregation (Figure 44). The means for patch size were mostly negative. For the aggregation parameter, family means were distributed more equally 135 between positive and negative effects (Figure 44). Panuridae and Anatidae were nearly significant, with a positive mean estimate. Figure 42: Number of bird families with significant and nearly significant coefficient estimate means (green, blue = positive, red, orange = negative) for each landscape metric: crop diversity (crop div), natural area (nat), rice area (rice), mean patch size (mn patch), aggregation of patches (ai), corn area (corn), and soybean area (soy). 136 crop diversity estimate Figure 43: Dot plot showing the family-specific mean estimates with 90% Bayesian credible intervals for crop diversity. Five families had positive estimates with BCIs that did not cross 0 (Corvidae, Ardeidae, Anatidae, Rallidae, Paridae) 137 mean patch size estimate aggregation index estimate Figure 44: Dot plots showing the family-specific mean estimates with 90% credible intervals for mean patch area and aggregation index. 138 For the percent natural landcover area metric, Panuridae, Troglodytidae, and Phasianidae were the only families significantly more likely to occur as natural landcover increased, though most of the non-significant groups also had positive mean estimates (Figure 45). Some families had negative means (e.g., Corvidae and Campephagidae) but were not significant. For the crop percent area metrics, the majority of families exhibited negative associations (Figure 45). Soybean area was significantly and negatively associated with Hirundinidae, Corvidae, Passeridae, Remizidae, Ardeidae, and Laridae families. The Corvidae family was significantly, negatively related to corn planted area with Anatidae and Falconidae being nearly significant and negative. Rice paddy area was positively associated with Corvidae occurrence (Figure 45). 139 percent natural estimate percent soybean estimate percent corn estimate percent rice estimate Figure 45: Dot plots showing the family-specific mean estimates with 90% Bayesian credible intervals for percent area natural landcover (forest, wetland, grassland), soybean, corn, and rice paddy. 140 Longitude was included in the occurrence model as a covariate to address potential underlying regional differences of species distribution between eastern and western halves of the province. Families more likely to occur in the western half of the province include Musciapidae, Falconidae, Sturnidae, Columbidae, Campephagidae, Lariidae, and Rallidae. Those occurring more in the eastern side include Anatidae and Corvidae. The minutes after sunrise covariate for detection was only significant for species in the Columbidae family (Figure 46). No estimates for detection in 2018 have BCI’s that exclude zero, implying that the increase in radius between the two years did not impact detection significantly (Figure 47). Some of the families that occurred more often in 2018 due to the increased sample radius include Ardeidae, Hirundidae, Campephagidae, and Rallidae. 141 longitude estimate minutes after sunrise estimate Figure 46: Dot plots showing the family-specific mean estimates with 90% credible intervals for longitude and minutes after sunrise, which were used as covariates in the occurrence and detection models respectively. 142 occupancy_17 estimate occupancy_18 estimate detection_17 estimate detection_18 estimate Figure 47: Dot plots showing the family-specific mean estimates with 90% credible intervals for occurrence and detection. 143 Most of the most abundant bird families (>30 observations) had occupancy probabilities higher than the mean across all crop diversity levels. Corvidae, Columbidae, and Ardeidae demonstrated the largest increases in occupancy probability as crop diversity increased (Figure 48). The remaining five families (11 total) were less impacted by crop diversity and had relationships similar to the community mean. Hirundinidae demonstrated a substantial positive relationship with mean patch size. Most families had a neutral or slightly positive relationship with percent natural area, except Corvidae which decreased in occupancy probability as natural area increased. For the percent crop area metrics, most families had a negative relationship as the landscape becomes more dominated by a given class, especially in the case of corn and soybean area (all negative). The negative relationship between soybean area and occurrence was significant for 5 of the most common families while corn was only significant for one. Rice area had negative, non-significant relationships with Columbidae and Ardeidae occupancy and a significant, positive relationship in the case of Corvidae (Figure 48). 144 Figure 48: All landscape metric relationships with occurrence probability for the most abundant families (>30 observations) 145 5.3.3 Functional model results For the functional model, positive values of crop diversity (MSIDI_crop) corresponded to higher occurrence probabilities at the community-level, compared to the taxonomic model, which had a credible interval that crossed zero (Table 45). All other landscape covariates had 90% credible intervals that contained zero. Table 45: Estimates of hyperparameters for the functional occurrence model parameter mean sd lower upper occurrence 17 -3.33 2.54 -5.88 -0.79 occurrence 18 0.43 0.84 -0.41 1.26 detection 17 0.27 0.60 -0.33 0.86 detection 18 0.24 0.36 -0.12 0.59 msidi_crop 0.35 0.34 0.01 0.69 mn_patch -0.09 0.37 -0.46 0.27 ai -0.08 0.48 -0.56 0.40 pland_nat 0.16 0.35 -0.19 0.51 pland_soybean -0.34 0.51 -0.85 0.16 pland_corn -0.26 0.49 -0.76 0.23 pland_rice -0.02 0.55 -0.57 0.53 long -0.08 0.45 -0.53 0.37 sunrise -0.13 0.44 -0.57 0.31 The number of bird groups with significant and nearly significant coefficient estimates (i.e., BCI crossed zero by <0.1) varied across the different landscape metrics (Figure 49). At the group level, higher crop diversity was significantly associated with higher occupancy probabilities for 8 of the 15 groups (Figure 50). Additionally, all groups had positive mean estimates and several additional families had BCIs that barely crossed zero (e.g., reedwarb/flycatcher, woodpecker). Only the dove/bunting/pheasant group was significantly impacted by patch size. The majority of the groups had mean estimates that were negative, with the woodpecker and shorebird/lapwing groups being nearly significant (Figure 51). The shorebird/lapwing group was significantly, positively impacted by aggregation index with the 146 crow/magpie and hoopoe/starling groups being nearly significant with negative estimates (Figure 51). Figure 49: Number of bird groups with significant and nearly significant coefficient estimate means (green, blue = positive, red, orange = negative) for each landscape metric: crop diversity (crop div), natural area (nat), rice area (rice), mean patch size (mn patch), aggregation of patches (ai), corn area (corn), and soybean area (soy). 147 crop diversity estimate Figure 50: Dot plot showing the group-specific mean estimates with 90% Bayesian credible intervals for crop diversity. Six groups had positive estimates with BCIs that did not cross 0 (crow/magpie, shorebird/lapwing, rail, pipit/wagtail, hoopoe/starling, kingfisher) 148 mean patch size estimate aggregation index estimate Figure 51: Dot plots showing the group-specific mean estimates with 90% credible intervals for mean patch area and aggregation index. 149 For the percent natural landcover area metric, the falcon/shrike and shorebird/lapwing groups were significantly more likely to occur as natural landcover increased (Figure 52). For soybean and corn area, the majority of groups exhibited negative associations. Soybean area was significantly and negatively associated with the crow/magpie, dove/bunting/pheasant, and kingfisher groups with the thrush/cuckoo/oriole and falcon/shrike groups being nearly significant (Figure 52). Corn area was negatively related to the thrush/cuckoo/oriole and crow/magpie groups with the falcon/shrike also being nearly significant (Figure 52). Rice paddy area was only positively related to the crow/magpie group (Figure 52). 150 percent natural estimate percent soybean estimate percent corn estimate percent rice estimate Figure 52: Dot plots showing the group-specific mean estimates with 90% Bayesian credible intervals for percent area natural landcover (forest, wetland, grassland), soybean, corn, and rice paddy. 151 Groups that were more likely to occur in the western half of the province included the woodpecker, swallow/reedtit, falcon/shrike, and hoopoe/starling groups (Figure 53). Those occurring more in the eastern side include the shorebird/lapwing, crow/magpie, dove/bunting/pheasant, and kingfisher groups. The minutes after sunrise covariate for detection was only significant for the kingfisher (negative) and woodpecker groups (positive) (Figure 53). longitude estimate mins after sunrise estimate Figure 53: Dot plots showing the group-specific mean estimates with 90% credible intervals for longitude and minutes after sunrise, which were used as covariates in the occurrence and detection models respectively. 152 As with the taxonomic model, the mean estimates for the occurrence and detection parameters demonstrate that the difference between the two years was most significant for occurrence (Figure 54). Some of the groups that occurred more often in 2018 due to the increased sample radius include the hoopoe/starling , kingfisher, dove/bunting/pheasant, , swallow/reedtit , and crow/magpie groups (Figure 54). occupancy_17 estimate occupancy_18 estimate detection_17 estimate detection_18 estimate Figure 54: Dot plots showing the group-specific mean estimates with 90% credible intervals for occurrence and detection. 153 Finally, we examined the relationship between occurrence probability for each functional group and the landscape metrics and found differing effects (Figure 55). Looking at the most abundant bird groups (>30), half of the groups had occupancy probabilities higher than the mean across all crop diversity levels. Crow/magpie and shorebird/lapwing in particular had the largest increases in occupancy probability as crop diversity increased. The dove/bunting/pheasant group occupancy increased with mean patch size and the other groups were negatively impacted. The aggregation index significantly increased occupancy for shorebird/lapwings, while the other groups had negative relationships. Percent natural landcover positively impacted the occupancy of the shorebird/lapwing, falcon/shrike, and dove/bunting/pheasant groups while negatively impacting the crow/magpie and swallow/reedtit groups. The relationships with soybean area and corn were similar, with the shorebird/lapwing group increasing in occupancy more with soybean percentage. Increasing rice area decreased occupancy probability for the swallow/reedtit, shorebird/lapwing, and falcon/shrike and increased for the dove/bunting/pheasant and crow/magpie groups (though the rice parameter was only significant for the crow/magpie group). 154 Figure 55: All landscape metric relationships with occurrence probability for the most abundant groups (>30 observations) 155 5.4 Discussion & Conclusion We found that different bird families and functional groups have different relationships to landscape structure. Our results show that crop diversity (MSIDI of corn, rice, soybean, and ‘other crop’) had the most substantial impact on bird diversity. Higher crop diversity increased mean occupancy for nearly all bird families and functional groups and was significant for many of the rarer families and functional groups, particularly for the latter. Of the five groups that had fewer than 30 observations, four had significant, positive parameter estimates for crop diversity (rail, pipit/wagtail, hoopoe/starling, kingfisher). We saw differences between our taxonomic and functional groupings when modeling occupancy. For crop diversity, a majority of the functional groups were significantly more likely to occur as crop diversity increased compared to the taxonomic model. This disparity between taxonomic and functional highlights that fact that landscape-biodiversity relationships are both complex and context-dependent as well as heavily influenced by the method used to define functional diversity (Hevia et al., 2016; M.-B. Lee & Martin, 2017). Several studies corroborate our finding that a positive relationship between crop diversity and biodiversity exists (M. B. Lee & Goodale, 2018; D. Li et al., 2020; Lindsay et al., 2013) yet others found no effect in agricultural systems (Fahrig et al., 2015; Martin et al., 2020; Redlich et al., 2018). We also found the percent of natural area (i.e., forest, grassland, and wetland) was not nearly as important as we expected. The amount of natural landcover was significant in some cases, but the effect size was often similar to or smaller than the other percent area metrics. A few groups even had negative estimates for natural area (e.g., Corvids). Only a handful of bird groups had significantly positive parameter estimates/credible intervals. This may be due to an under sampling of natural habitat, though range of percent natural area (250m) was 0.027-100%. 156 The metrics in our model that were more configurational in nature (mean patch area & aggregation) were not as significant for the bird groups in our dataset. This is in contrast with several studies that report mean patch size as a significant predictor of biodiversity in agricultural landscapes (Fahrig et al., 2015; Martin et al., 2020). However, it should be noted that we limited the potential effect of patch size by using a maximum extent of 1000m. It is possible there would be a significant effect if mean patch size were calculated at a larger extent. All significant parameter estimates for percent corn and soybean were negative, while rice area only had a positive, significant estimate for Corvidae and the crow/magpie group. At the taxonomic level, soybean area had more significantly negative estimates (six families) compared to corn (one family). Furthermore, the six families that had significant, negative coefficient estimates for soybean area were some of the most commonly observed in our sample (i.e., >28 observations). The disparity between soybean and corn’s impact lessened when looking at functional group occupancy. Soybean was negatively associated with two groups and corn with one, as well as one shared between the two (crow/magpie). A possible explanation for this difference is that the lack of perching opportunities for birds in soybean fields. For example, shrikes (Laridae family) were often observed perching on corn stalks or trees while scanning for prey. That is not possible in a landscape dominated by soybean fields. Given that rice area had no negative parameter estimates, this implies landscapes completely dominated by rain-fed crops are substantially less habitable than those dominated by paddy rice, potentially due to some of the more biodiversity-friendly practices related to rice cultivation (e.g., flooding for weed control, rice-fish cultures)(Amira et al., 2018; Y. Sun et al., 2015; C. Wood et al., 2017). Further validation of this model would enhance its predictive ability 157 as well as provide insight for agroecosystems with similar landscape characteristics (i.e., paddy rice, high heterogeneity). Overall, we found that different crop types support different bird families and groups and that, in general, landscapes that have high crop diversity (in terms of both number of crops and relative abundance) maximized occupancy for the most bird species. This suggests that monocultures, even at the relatively fine scale of 500m, decrease a landscape’s ability to support bird biodiversity. Therefore, in the context of global trade impacts, if a region that was only growing soybeans converted some land to different crops, then there may have been a benefit to bird biodiversity. However, we observed that the more highly impacted region had less crop diversity after experiencing soybean area loss. Given the prevalence of small-scale farming in Heilongjiang, it is more likely that soybean patches within a crop mosaic (with many crops and heterogenous crop field sizes) were converted to more profitable crops, thus decreasing overall crop diversity. In addition to decreasing occupancy for all bird families/groups, this likely had an outsized, negative impact on rarer bird functional groups (which may have consequences for the ecosystem as whole). That being said, if soybean area was converted to rice, the detrimental impacts to biodiversity may have been less severe compared to corn as rice had no significantly negative parameter estimates for occupancy. As these results show, cultivated land is more than capable of supporting a wide array of bird species. Furthermore, while many of the species in this agroecosystem are considered common, many are experiencing decreasing populations. For example, the Northern Lapwing (Vanellus vanellus) is a common sight in Heilongjiang landscapes growing rice, corn, and soybeans and is considered 'near threatened' by the IUCN red list. Bird conservation in this 158 region does not necessarily need to focus on increasing natural habitat, but rather, maintaining the heterogeneity of the landscape, including the diversity of crops grown, at a relatively fine scale. Challenges to the idea that farmland cannot be wildlife habitat are essential for the future sustainability of global food systems and biodiversity, in China and across the globe (Kremen & Merenlender, 2018; L. Li et al., 2020). As farmers become more involved in global markets, their practices often scale-up to reduce costs and maximize yields, resulting in more monocultural landscapes. Preserving the crop diversity of farmland, especially in the case of a landscape that combines rain-fed crops and paddy rice, will be vital for many of the bird species that rely on crop mosaics. 159 APPENDIX 160 Model Code for WinBUGS: model{ # Define prior distributions for community-level model parameters omega ~ dunif(0,1) # grand means mean ~ dunif(0,1) mu.u0 <- log(mean) - log(1-mean) mean2 ~ dunif(0,1) mu.u1 <- log(mean2) - log(1-mean2) mean3 ~ dunif(0,1) mu.v0 <- log(mean3) - log(1-mean3) mean4 ~ dunif(0,1) mu.v1 <- log(mean3) - log(1-mean3) mu.a1 ~ dnorm(0, 0.001) # MSIDI_CROP mu.a2 ~ dnorm(0, 0.001) # MN_PATCH_ALLCLASS mu.a3 ~ dnorm(0, 0.001) # AI_ALLCLASS mu.a4 ~ dnorm(0, 0.001) # PLAND_NATURAL mu.a5 ~ dnorm(0, 0.001) # PLAND_SOYBEAN mu.a6 ~ dnorm(0, 0.001) # PLAND_CORN mu.a7 ~ dnorm(0, 0.001) # PLAND_RICE mu.a8 ~ dnorm(0, 0.001) # LONGITUDE mu.b1 ~ dnorm(0, 0.001) # SUNRISE tau.u0 ~ dgamma(0.1,0.1) tau.u1 ~ dgamma(0.1,0.1) tau.v0 ~ dgamma(0.1,0.1) tau.v1 ~ dgamma(0.1,0.1) tau.a1 ~ dgamma(0.1,0.1) tau.a2 ~ dgamma(0.1,0.1) tau.a3 ~ dgamma(0.1,0.1) tau.a4 ~ dgamma(0.1,0.1) tau.a5 ~ dgamma(0.1,0.1) tau.a6 ~ dgamma(0.1,0.1) tau.a7 ~ dgamma(0.1,0.1) tau.a8 ~ dgamma(0.1,0.1) tau.b1 ~ dgamma(0.1,0.1) # Species loop (i) for (i in 1:(n+nzeroes)) { # Create priors for species i from the community level prior distributions w[i] ~ dbern(omega) u0[i] ~ dnorm(mu.u0, tau.u0) u1[i] ~ dnorm(mu.u1, tau.u1) v0[i] ~ dnorm(mu.v0, tau.v0) v1[i] ~ dnorm(mu.v1, tau.v1) 161 a1[i] ~ dnorm(mu.a1, tau.a1) a2[i] ~ dnorm(mu.a2, tau.a2) a3[i] ~ dnorm(mu.a3, tau.a3) a4[i] ~ dnorm(mu.a4, tau.a4) a5[i] ~ dnorm(mu.a5, tau.a5) a6[i] ~ dnorm(mu.a6, tau.a6) a7[i] ~ dnorm(mu.a7, tau.a7) a8[i] ~ dnorm(mu.a8, tau.a8) b1[i] ~ dnorm(mu.b1, tau.b1) ## Likelihood ## # Create a loop to estimate the Z matrix (true occurrence for species i at point j) # Point loop (j) for (j in 1:J) { logit(psi[j,i]) <- u0[i] + u1[i] * Year[j] + a1[i]*std.MSIDI_CROP[j] + a2[i]*std.MNPATCH_ALLCLASS[j] + a3[i]*std.AI_ALLCLASS[j] + a4[i]*std.PLAND_NATURAL[j] + a5[i]*std.PLAND_SOYBEAN[j] + a6[i]*std.PLAND_CORN[j] + a7[i]*std.PLAND_RICE[j] + a8[i]*std.long[j] mu.psi[j,i] <- psi[j,i]*w[i] Z[j,i] ~ dbern(mu.psi[j,i]) #Create a loop to estimate detection for species i at point k during sampling period k for (k in 1:K[j]) { logit(p[j,k,i]) <- v0[i] + v1[i] * Year[j] + b1[i]*std.sunrise[j,k] mu.p[j,k,i] <- p[j,k,i]*Z[j,i] X[j,k,i] ~ dbern(mu.p[j,k,i]) } #k } #j } #i # Sum all species observed (n) and unobserved species (n0) to find the # total estimated richness n0 <- sum(w[(n+1):(n+nzeroes)]) N <- n + n0 # Create a loop to determine point level richness estimates for the # whole community and for subsets or assemblages of interest. for(j in 1:J){ Nsite[j]<- inprod(Z[j,1:(n+nzeroes)],w[1:(n+nzeroes)]) } #j 162 CHAPTER 6: SYNTHESIS 163 This dissertation is part of a larger project aimed at investigating the socioeconomic and environmental impacts of global soybean trade across distances using the telecoupling framework. Broadly, the work in this dissertation explores biodiversity impacts of telecoupling. Furthermore, this study goes beyond quantifying environmental impacts solely as land use change with assumed biodiversity implications by translating landscape change due to global trade to explicit consequences for bird diversity. In chapter two, a systematic review of the distant drivers of global biodiversity change was conducted. After conducting full-text analyses of 131 studies and synthesizing impacts of ten types of distant drivers and seven taxonomic groups, negative impacts on biodiversity were found to be the most frequent outcome reported. Some instances of beneficial and species composition change were observed as well. The most common type of drivers studied were trade and tourism. Despite containing some of the world’s most important biodiversity hotspots, substantially fewer studies were conducted in Africa and Oceania, compared to other developed continents. This synthesis highlights the need for more biodiversity research that empirically assesses biodiversity change in the context of globalization. Chapter three reports on the process of creating the landcover maps used to inform the landscape variables used in later chapters. Studies that use landscape characteristics to explain biodiversity patterns are only as good as the landcover maps input. Of the landcover maps available to researchers, many have significant limitations (coarse spatial resolution, limited temporal and geographic availability, class schemes that are too generalized) in addition to the inherent inaccuracies common to all remotely sensed products. The resulting landcover map had a fine enough spatial resolution to capture the heterogeneity of the farming landscape in 164 Heilongjiang and achieved a high level of accuracy by utilizing the large library of remotely sensed data and computing power of Google Earth Engine. The fourth chapter contributes to the scientific community’s understanding of how landscape structure mediates the biodiversity of agroecosystems and places these relationships in the context of global trade impacts. In order to assess biodiversity and landscape relationships at the functional level, life history and morphological traits were collected for each bird family observed in this study, allowing for the calculation of community functional metrics as well as the creation of functional groups. After analyzing the relationships of several landscape metrics with taxonomic and functional bird community diversity metrics, significant correlations between landscape structure and biodiversity were found. Crop diversity, natural landcover, and edge metrics were positively correlated with richness while aggregation and rain-fed crop (i.e., corn and soybean) area were both negatively related. No relationship with mean patch size was found, contrary to other studies’ findings. This disagreement is likely due to inherent differences between agroecosystems that primarily cultivate rain-fed crops (i.e., corn, soybean) and those that farm paddy rice. In the context of global trade impact, crop diversity appears to be negatively impacted as farmers convert soybean fields to other crops, such as corn and rice. However, growing paddy rice versus a crop such as corn may be beneficial for functional diversity as different bird groups and community traits are associated with flooded rice paddies. As growing food security challenges necessitate the use of land for food production, understanding how to design these landscapes to bolster biodiversity will be necessary for the health of ecosystems across the world. The fifth and final chapter models the relationship between landscape structure and group-specific occupancy at both taxonomic and functional levels. Using a hierarchical modeling 165 structure and Bayesian analysis, estimates of landscape metric effects were obtained for each group and at the community-level. Overall, crop diversity significantly increased occupancy of birds for taxonomic and functional groups. However, the percentage of natural landcover was not as important as expected, and metrics related to landscape configuration (i.e., mean patch size, aggregation index) had fewer significant impacts on occupancy. The effects of rice, corn, and soybean area differed as well. Increases in rice area were not detrimental to bird occupancy whereas increases in corn and soybean area had significant negative effects for a number of bird families and functional groups. Soybean area in particular exhibited more significant negative associations with bird families, which suggests that conversion away from soybean due to global trade in a soybean monoculture may have been beneficial for some bird species, especially when those fields were converted to paddy rice. However, in the typical heterogeneous landscape of Heilongjiang farmland, it is more likely that the conversion of soybean fields to another crop decreased the crop diversity of the landscape, which was a much more important contributor to bird occupancy. By combining remotely sensed landscape characteristics with biodiversity observations, this chapter provides a better understanding of how global trade and land use change may ultimately impact agroecosystems and can help inform farmland biodiversity conservation strategies. 166 REFERENCES 167 REFERENCES Aguilar, J., Gramig, G. G., Hendrickson, J. R., Archer, D. W., Forcella, F., & Liebig, M. A. (2015). Crop Species Diversity Changes in the United States: 1978–2012. PLOS ONE, 10(8), e0136580. https://doi.org/10.1371/journal.pone.0136580 Allouche, O., Kalyuzhny, M., Moreno-Rueda, G., Pizarro, M., & Kadmon, R. (2012). Area– heterogeneity tradeoff and the diversity of ecological communities. Proceedings of the National Academy of Sciences, 109(43), 17495–17500. https://doi.org/10.1073/pnas.1208652109 Alroy, J. (2015). Current extinction rates of reptiles and amphibians. Proceedings of the National Academy of Sciences of the United States of America, 112(42), 13003–13008. https://doi.org/10/f7vhh7 Amira, N., Rinalfi, T., & Azhar, B. (2018). Effects of intensive rice production practices on avian biodiversity in Southeast Asian managed wetlands. Wetlands Ecology and Management, 26(5), 865–877. https://doi.org/10/gfd2tf Arizaga, J., Unamuno, E., Clarabuch, O., & Azkona, A. (2013). The impact of an invasive exotic bush on the stopover ecology of migrant passerines. Animal Biodiversity and Conservation, 36(1), 1–11. https://doi.org/10/gnn658 Arthington, A. H., Dulvy, N. K., Gladstone, W., & Winfield, I. J. (2016). Fish conservation in freshwater and marine realms: Status, threats and management. Aquatic Conservation: Marine and Freshwater Ecosystems, 26(5), 838–857. https://doi.org/10/f9cf9c Assédé, E. S. P., Djagoun, C. A. M. S., Azihou, F. A., Gogan, Y. S. C., Kouton, M. D., Adomou, A. C., Geldenhuys, C. J., Chirwa, P. W., & Sinsin, B. (2018). Efficiency of conservation areas to protect orchid species in Benin, West Africa. South African Journal of Botany, 116, 230–237. https://doi.org/10/gdnsfs Austin, J. E., Morrison, K. L., & Harris, J. T., editors. (2018). Cranes_and_agriculture_web_2018.pdf. Internatoinal Crane Foundation. https://savingcranes.org/wp- content/uploads/2018/10/cranes_and_agriculture_web_2018.pdf#page=64 Balduino, G., Marques, T., & Bessa, E. (2017). Fish assemblages from touristic and non-visited headwater streams differ. Environmental Biology of Fishes, 100(5), 575–584. https://doi.org/10/f96q2n Balmer, B., McDonald, T., Hornsby, F., Adams, J., Allen, J., Barleycorn, A., Clarke, P., Cush, C., Honaker, A., McHugh, K., Speakman, T., Wells, R., & Schwacke, L. (2018). Long- term trends in a northern Gulf of Mexico common bottlenose dolphin (Tursiops truncatus) population in the wake of the Deepwater Horizon oil spill. Journal of Cetacean Research and Management, 18, 1–9. 168 Barragán, F., Moreno, C. E., Escobar, F., Halffter, G., & Navarrete, D. (2011). Negative Impacts of Human Land Use on Dung Beetle Functional Diversity. PLOS ONE, 6(3), e17976. https://doi.org/10.1371/journal.pone.0017976 Bek, D., Binns, T., Blokker, T., Mcewan, C., & Hughes, A. (2017). A High Road to Sustainability? Wildflower Harvesting, Ethical Trade and Social Upgrading in South Africa’s Western Cape. Journal of Agrarian Change, 17(3), 459–479. https://doi.org/10/gnn66r Belgiu, M., & Drǎguţ, L. (2014). Comparing supervised and unsupervised multiresolution segmentation approaches for extracting buildings from very high resolution imagery. ISPRS Journal of Photogrammetry and Remote Sensing, 96, 67–75. https://doi.org/10.1016/j.isprsjprs.2014.07.002 Benton, T. G., Vickery, J. A., & Wilson, J. D. (2003). Farmland biodiversity: Is habitat heterogeneity the key? Trends in Ecology and Evolution, 18(4), 182–188. https://doi.org/10/d9fvcn Berry, E. M., Dernini, S., Burlingame, B., Meybeck, A., & Conforti, P. (2014). Food security and sustainability: Can one exist without the other? Food Security and Sustainability, 10. Bibby, C. J. (Ed.). (2000). Bird census techniques (2nd ed). Academic. Billerman, S. M., Keeney, B. K., Rodewald, P. G., & Schulenberg, T. S. (Editors). (2020). Birds of the World. Cornell Laboratory of Ornithology, Ithaca, NY, USA. https://birdsoftheworld.org/bow/home Billeter, R., Liira, J., Bailey, D., Bugter, R., Arens, P., Augenstein, I., Aviron, S., Baudry, J., Bukacek, R., Burel, F., Cerny, M., De Blust, G., De Cock, R., Diekötter, T., Dietz, H., Dirksen, J., Dormann, C., Durka, W., Frenzel, M., … Edwards, P. J. (2008). Indicators for biodiversity in agricultural landscapes: A pan-European study. Journal of Applied Ecology, 45(1), 141–150. https://doi.org/10/czfj26 Birkhofer, K., Andersson, G. K. S., Bengtsson, J., Bommarco, R., Dänhardt, J., Ekbom, B., Ekroos, J., Hahn, T., Hedlund, K., Jönsson, A. M., Lindborg, R., Olsson, O., Rader, R., Rusch, A., Stjernman, M., Williams, A., & Smith, H. G. (2018). Relationships between multiple biodiversity components and ecosystem services along a landscape complexity gradient. Biological Conservation, 218(November 2017), 247–253. https://doi.org/10/gczvfp Blaix, C., & Moonen, A.-C. (2020). Structural field margin characteristics affect the functional traits of herbaceous vegetation. PLOS ONE, 15(9), e0238916. https://doi.org/10.1371/journal.pone.0238916 Böhm, M., Collen, B., Baillie, J. E. M., Bowles, P., Chanson, J., Cox, N., Hammerson, G., Hoffmann, M., Livingstone, S. R., Ram, M., Rhodin, A. G. J., Stuart, S. N., van Dijk, P. P., Young, B. E., Afuang, L. E., Aghasyan, A., García, A., Aguilar, C., Ajtic, R., … Zug, 169 G. (2013). The conservation status of the world’s reptiles. Biological Conservation, 157, 372–385. https://doi.org/10/f4s8jr Borrelli, P., Robinson, D. A., Panagos, P., Lugato, E., Yang, J. E., Alewell, C., Wuepper, D., Montanarella, L., & Ballabio, C. (2020). Land use and climate change impacts on global soil erosion by water (2015-2070). Proceedings of the National Academy of Sciences of the United States of America, 117(36), 21994–22001. https://doi.org/10/gkx8b2 Brandt, J. S., & Buckley, R. C. (2018). A global systematic review of empirical evidence of ecotourism impacts on forests in biodiversity hotspots. Current Opinion in Environmental Sustainability, 32, 112–118. https://doi.org/10/gfcpv6 Brazil, M. (2009). Birds of East Asia: China, Taiwan, Korea, Japan, and Russia. Princeton University Press. Bren d’Amour, C., Reitsma, F., Baiocchi, G., Barthel, S., Güneralp, B., Erb, K.-H., Haberl, H., Creutzig, F., & Seto, K. C. (2017). Future urban land expansion and implications for global croplands. Proceedings of the National Academy of Sciences, 201606036. https://doi.org/10/f3t38d Brooks, S. P., & Gelman, A. (1998). General methods for monitoring convergence of iterative simulations)? In Journal of Computational and Graphical Statistics (Vol. 7, Issue 4, pp. 434–455). https://doi.org/10.1080/10618600.1998.10474787 Cadotte, M. W., Carscadden, K., & Mirotchnick, N. (2011). Beyond species: Functional diversity and the maintenance of ecological processes and services. Journal of Applied Ecology, 48(5), 1079–1087. https://doi.org/10/bvz3zq Carrasco, L. R., Chan, J., McGrath, F. L., & Nghiem, L. T. P. (2017). Biodiversity conservation in a telecoupled world. Ecology and Society, 22(3). https://doi.org/10/gcgpk4 Ceballos, G., Ehrlich, P. R., Barnosky, A. D., García, A., Pringle, R. M., & Palmer, T. M. (2015). Accelerated modern human-induced species losses: Entering the sixth mass extinction. Science Advances, 1(5), e1400253. https://doi.org/10/5mm Chan, K. M. A., Boyd, D. R., Gould, R. K., Jetzkowitz, J., Liu, J., Muraca, B., Naidoo, R., Olmsted, P., Satterfield, T., Selomane, O., Singh, G. G., Sumaila, R., Ngo, H. T., Boedhihartono, A. K., Agard, J., de Aguiar, A. P. D., Armenteras, D., Balint, L., Barrington-Leigh, C., … Brondízio, E. S. (2020). Levers and leverage points for pathways to sustainability. People and Nature, 2(3), 693–717. https://doi.org/10/gh8p34 Chaudhary, A., & Brooks, T. M. (2019). National Consumption and Global Trade Impacts on Biodiversity. World Development, 121, 178–187. https://doi.org/10.1016/j.worlddev.2017.10.012 Chaudhary, A., & Kastner, T. (2016). Land use biodiversity impacts embodied in international food trade. Global Environmental Change, 38, 195–204. https://doi.org/10/f8mt4t 170 Chen, H., Meng, F., Yu, Z., & Tan, Y. (2022). Spatial–temporal characteristics and influencing factors of farmland expansion in different agricultural regions of Heilongjiang Province, China. Land Use Policy, 115, 106007. https://doi.org/10.1016/j.landusepol.2022.106007 Convention on Biological Diversity. (2020). Convention on Biological Diversity. https://www.cbd.int/conferences/post2020 Coppes, J., Ehrlacher, J., Thiel, D., Suchant, R., & Braunisch, V. (2017). Outdoor recreation causes effective habitat reduction in capercaillie Tetrao urogallus: A major threat for geographically restricted populations. Journal of Avian Biology, 48(12), 1583–1594. https://doi.org/10/gct3dv Cuvin-Aralar, M. L. A. (2016). Impacts of aquaculture on fish biodiversity in the freshwater lake Laguna de Bay, Philippines. Lakes and Reservoirs: Research and Management, 21(1), 31–39. https://doi.org/10/gnn68c Czortek, P., Delimat, A., Dyderski, M. K., Zięba, A., Jagodziński, A. M., & Jaroszewicz, B. (2018). Climate change, tourism and historical grazing influence the distribution of Carex lachenalii Schkuhr – A rare arctic-alpine species in the Tatra Mts. Science of the Total Environment, 618, 1628–1637. https://doi.org/10/gc6ckx Dale, V. H., Parish, E., Kline, K. L., & Tobin, E. (2017). How is wood-based pellet production affecting forest conditions in the southeastern United States? Forest Ecology and Management, 396, 143–149. https://doi.org/10/gbg5z2 Darwall, W. R. T., & Freyhof, J. (2015). Lost fishes, who is counting? The extent of the threat to freshwater fish biodiversity. In G. P. Closs, J. D. Olden, & M. Krkosek (Eds.), Conservation of Freshwater Fishes (pp. 1–36). Cambridge University Press. https://doi.org/10.1017/cbo9781139627085.002 De Fraiture, C., Cai, X., Amarasinghe, U., Rosegrant, M., & Molden, D. (2004). Does International Cereal Trade Save Water? The Impact of Virtual Water Trade on Global Water Use. In Comprehensive Assessment Research Report (Vol. 4). Del Giudice, D., Matli, V. R. R., & Obenour, D. R. (2020). Bayesian mechanistic modeling characterizes Gulf of Mexico hypoxia: 1968–2016 and future scenarios. Ecological Applications, 30(2), e02032. https://doi.org/10/gnn67w Di Falco, S., & Perrings, C. (2005). Crop biodiversity, risk management and the implications of agricultural assistance. Ecological Economics, 55(4), 459–466. https://doi.org/10.1016/j.ecolecon.2004.12.005 Díaz, S., Settele, J., Brondízio, E. S., Ngo, H. T., Agard, J., Arneth, A., Balvanera, P., Brauman, K. A., Butchart, S. H. M., Chan, K. M. A., Lucas, A. G., Ichii, K., Liu, J., Subramanian, S. M., Midgley, G. F., Miloslavich, P., Molnár, Z., Obura, D., Pfaff, A., … Zayas, C. N. (2019). Pervasive human-driven decline of life on Earth points to the need for transformative change. Science, 366(6471). https://doi.org/10/ggfg6n 171 Dorninger, C., von Wehrden, H., Krausmann, F., Bruckner, M., Feng, K., Hubacek, K., Erb, K. H., & Abson, D. J. (2021). The effect of industrialization and globalization on domestic land-use: A global resource footprint perspective. Global Environmental Change, 69, 102311. https://doi.org/10/gkq9t2 Dou, Y., da Silva, R. F. B., Yang, H., & Liu, J. (2018). Spillover effect offsets the conservation effort in the Amazon. Journal of Geographical Sciences, 28(11), 1715–1732. https://doi.org/10/gnn64d Duflot, R., Georges, R., Ernoult, A., & Burel, F. (2014). Landscape heterogeneity as an ecological filter of species traits. Acta Oecologica, 56, 19–26. https://doi.org/10.1016/j.actao.2014.01.004 Duro, D. C., Franklin, S. E., & Dubé, M. G. (2012). A comparison of pixel-based and object- based image analysis with selected machine learning algorithms for the classification of agricultural landscapes using SPOT-5 HRG imagery. Remote Sensing of Environment, 118, 259–272. https://doi.org/10.1016/j.rse.2011.11.020 Edie Stewart M., Jablonski David, & Valentine James W. (2018). Contrasting responses of functional diversity to major losses in taxonomic diversity. Proceedings of the National Academy of Sciences, 115(4), 732–737. https://doi.org/10.1073/pnas.1717636115 Estrada, A., Garber, P. A., & Chaudhary, A. (2019). Expanding global commodities trade and consumption place the world’s primates at risk of extinction. PeerJ, 7, e7068. https://doi.org/10.7717/peerj.7068 European Space Agency. (2017). Sentinel-2. http://www.esa.int/Our_Activities/Observing_the_Earth/Copernicus/Sentinel-2 Fahrig, L. (2001). How much habitat is enough? Biological Conservation, 100(1), 65–74. https://doi.org/10/c87xvd Fahrig, L., Baudry, J., Brotons, L., Burel, F. G., Crist, T. O., Fuller, R. J., Sirami, C., Siriwardena, G. M., & Martin, J. L. (2011). Functional landscape heterogeneity and animal biodiversity in agricultural landscapes. Ecology Letters, 14(2), 101–112. https://doi.org/10/bq85mt Fahrig, L., Girard, J., Duro, D., Pasher, J., Smith, A., Javorek, S., King, D., Lindsay, K. F., Mitchell, S., & Tischendorf, L. (2015). Farmlands with smaller crop fields have higher within-field biodiversity. Agriculture, Ecosystems and Environment, 200, 219–234. https://doi.org/10/f6zkkd Fahrig, L., & Nuttle, W. K. (2005). Population Ecology in Spatially Heterogeneous Environments. In G. M. Lovett, M. G. Turner, C. G. Jones, & K. C. Weathers (Eds.), Ecosystem Function in Heterogeneous Landscapes (pp. 95–118). Springer. https://doi.org/10.1007/0-387-24091-8_6 172 FAOSTAT Database. (2017). Food and Agriculture Organization of the United Nations. FAO; FAO. Farris, E., Pisanu, S., Ceccherelli, G., & Filigheddu, R. (2013). Human trampling effects on Mediterranean coastal dune plants. Plant Biosystems, 147(4), 1043–1051. https://doi.org/10/gnn67z Farwell, L. S., Wood, P. B., Sheehan, J., & George, G. A. (2016). Shale gas development effects on the songbird community in a central Appalachian forest. Biological Conservation, 201, 78–91. https://doi.org/10/f88bvq Fearnside, P. M. (2002). Soybean cultivation as a threat to the environment in Brazil. Environmental Conservation, 28(01), 23–38. https://doi.org/10/dwmm8v Firbank, L. G., Petit, S., Smart, S., Blain, A., & Fuller, R. J. (2008). Assessing the impacts of agricultural intensification on biodiversity: A British perspective. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1492), 777–787. https://doi.org/10/c8fhg8 Fischer, C., Gayer, C., Kurucz, K., Riesch, F., Tscharntke, T., & Batáry, P. (2018). Ecosystem services and disservices provided by small rodents in arable fields: Effects of local and landscape management. Journal of Applied Ecology, 55(2), 548–558. https://doi.org/10/gc357j Fitzherbert, E. B., Struebig, M. J., Morel, A., Danielsen, F., Brühl, C. A., Donald, P. F., & Phalan, B. (2008). How will oil palm expansion affect biodiversity? Trends in Ecology and Evolution, 23(10), 538–545. https://doi.org/10/dhgqbk Flecks, M., Weinsheimer, F., Böhme, W., Chenga, J., Lötters, S., & Rödder, D. (2012). Watching extinction happen: The dramatic population decline of the critically endangered Tanzanian Turquoise Dwarf Gecko, Lygodactylus williamsi. Salamandra, 48(1), 12–20. Flynn, D. F. B., Gogol-Prokurat, M., Nogeire, T., Molinari, N., Richers, B. T., Lin, B. B., Simpson, N., Mayfield, M. M., & DeClerck, F. (2009). Loss of functional diversity under land use intensification across multiple taxa. Ecology Letters, 12(1), 22–33. https://doi.org/10.1111/j.1461-0248.2008.01255.x Foley, J. A., DeFries, R., Asner, G. P., Barford, C., Bonan, G., Carpenter, S. R., Chapin, F. S., Coe, M. T., Daily, G. C., Gibbs, H. K., Helkowski, J. H., Holloway, T., Howard, E. A., Kucharik, C. J., Monfreda, C., Patz, J. A., Prentice, I. C., Ramankutty, N., & Snyder, P. K. (2005). Global Consequences of Land Use. Science, 309(5734). https://doi.org/10/dqd2tm Forman, R. T. T., & Godron, M. (1981). Patches and Structural Components for a Landscape Ecology. BioScience, 31(10), 733–740. https://doi.org/10.2307/1308780 173 Frazier, A. E., & Kedron, P. (2017). Landscape Metrics: Past Progress and Future Directions. Current Landscape Ecology Reports, 2(3), 63–72. https://doi.org/10.1007/s40823-017- 0026-0 Fukami, T., Bellingham, P. J., Peltzer, D. A., & Walker, L. R. (2013). Non-Native Plants Disrupt Dual Promotion of Native Alpha and Beta Diversity. Folia Geobotanica, 48(3), 319–333. https://doi.org/10/f5c7ht Gale, F. (2015). Development of China’s Feed Industry and Demand for Imported Commodities. 22. Gale, F., Valdes, C., & Ash, M. (2019). Interdependence of China, United States, and Brazil in Soybean Trade (p. 48) [Economic Research Service]. United States Department of Agriculture. Gamero, A., Brotons, L., Brunner, A., Foppen, R., Fornasari, L., Gregory, R. D., Herrando, S., Hořák, D., Jiguet, F., Kmecl, P., Lehikoinen, A., Lindström, Å., Paquet, J. Y., Reif, J., Sirkiä, P. M., Škorpilová, J., van Strien, A., Szép, T., Telenský, T., … Voříšek, P. (2017). Tracking Progress Toward EU Biodiversity Strategy Targets: EU Policy Effects in Preserving its Common Farmland Birds. Conservation Letters, 10(4), 394–401. https://doi.org/10/f9scx3 Gao, B. (1996). NDWI—A normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sensing of Environment, 58(3), 257–266. https://doi.org/10.1016/S0034-4257(96)00067-3 Garibaldi, L. A., Gemmill-Herren, B., D’Annolfo, R., Graeub, B. E., Cunningham, S. A., & Breeze, T. D. (2017). Farming Approaches for Greater Biodiversity, Livelihoods, and Food Security. Trends in Ecology & Evolution, 32(1), 68–80. https://doi.org/10.1016/j.tree.2016.10.001 Geedicke, I., Oldeland, J., & Leishman, M. R. (2018). Urban stormwater run-off promotes compression of saltmarshes by freshwater plants and mangrove forests. Science of the Total Environment, 637–638, 137–144. https://doi.org/10/gnn68f Geiger, F., Bengtsson, J., Berendse, F., Weisser, W. W., Emmerson, M., Morales, M. B., Ceryngier, P., Liira, J., Tscharntke, T., Winqvist, C., Eggers, S., Bommarco, R., Pärt, T., Bretagnolle, V., Plantegenest, M., Clement, L. W., Dennis, C., Palmer, C., Oñate, J. J., … Inchausti, P. (2010). Persistent negative effects of pesticides on biodiversity and biological control potential on European farmland. Basic and Applied Ecology, 11(2), 97–105. https://doi.org/10.1016/j.baae.2009.12.001 Gelder, B. K., Kaleita, A. L., & Cruse, R. M. (2009). Estimating Mean Field Residue Cover on Midwestern Soils Using Satellite Imagery. Agronomy Journal, 101(3), 635–643. https://doi.org/10.2134/agronj2007.0249 Gil-Tena, A., De C??ceres, M., Ernoult, A., Butet, A., Brotons, L., & Burel, F. (2015). Agricultural landscape composition as a driver of farmland bird diversity in Brittany 174 (NW France). Agriculture, Ecosystems and Environment, 205, 79–89. https://doi.org/10/f699fc Gislason, P. O., Benediktsson, J. A., & Sveinsson, J. R. (2006). Random forests for land cover classification. Pattern Recognition Letters, 27(4), 294–300. https://doi.org/10.1016/j.patrec.2005.08.011 Godfray, H. C. J., & Garnett, T. (2014). Food security and sustainable intensification. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 369(1639), 20120273. https://doi.org/10.1098/rstb.2012.0273 Green, J. M. H., Croft, S. A., Durán, A. P., Balmford, A. P., Burgess, N. D., Fick, S., Gardner, T. A., Godar, J., Suavet, C., Virah-Sawmy, M., Young, L. E., & West, C. D. (2019). Linking global drivers of agricultural trade to on-the-ground impacts on biodiversity. Proceedings of the National Academy of Sciences of the United States of America, 116(46), 23202– 23208. https://doi.org/10/gnb9h5 Habel, J. C., Rasche, L., Schneider, U. A., Engler, J. O., Schmid, E., Rödder, D., Meyer, S. T., Trapp, N., Sos del Diego, R., Eggermont, H., Lens, L., & Stork, N. E. (2019). Final countdown for biodiversity hotspots. Conservation Letters, 12(6). https://doi.org/10/gjsmbk Hao, A. (2010). Uncertainty, risk aversion and risk management in agriculture, farmer; uncertainty; risk aversion; risk management. Agriculture and Agricultural Science Procedia, 1, 152–156. https://doi.org/10.1016/j.aaspro.2010.09.018 Harris, J. B. C., Tingley, M. W., Hua, F., Yong, D. L., Adeney, J. M., Lee, T. M., Marthy, W., Prawiradilaga, D. M., Sekercioglu, C. H., Suyadi, C. H., Winarni, N., & Wilcove, D. S. (2017). Measuring the impact of the pet trade on Indonesian birds. Conservation Biology, 31(2), 394–405. https://doi.org/10/f9nw84 Hayden, M. T., Mattimoe, R., & Jack, L. (2021). Sensemaking and the influencing factors on farmer decision-making. Journal of Rural Studies, 84, 31–44. https://doi.org/10.1016/j.jrurstud.2021.03.007 Headey, D. (2011). Rethinking the global food crisis: The role of trade shocks. Food Policy, 36(2), 136–146. https://doi.org/10/dj5x7z Herzog, F., Lüscher, G., Arndorfer, M., Bogers, M., Balázs, K., Bunce, R. G. H., Dennis, P., Falusi, E., Friedel, J. K., Geijzendorffer, I. R., Gomiero, T., Jeanneret, P., Moreno, G., Oschatz, M. L., Paoletti, M. G., Sarthou, J. P., Stoyanova, S., Szerencsits, E., Wolfrum, S., … Bailey, D. (2017). European farm scale habitat descriptors for the evaluation of biodiversity. Ecological Indicators, 77, 205–217. https://doi.org/10.1016/j.ecolind.2017.01.010 Hesselbarth, M. H. K., Sciaini, M., With, K. A., Wiegand, K., & Nowosad, J. (2019). landscapemetrics: An open-source R tool to calculate landscape metrics. Ecography, 42, 1648–1657. https://doi.org/10.1111/ecog.04617 175 Hevia, V., P. Carmona, C., Azcárate, F., Torralba, M., Alcorlo, P., Ariño, R., Lozano, J., Castro- Cobo, S., & González, J. (2016). Effects of land use on taxonomic and functional diversity: A cross-taxon analysis in a Mediterranean landscape. Oecologia, 181. https://doi.org/10.1007/s00442-015-3512-2 Hooper, D. U., Adair, E. C., Cardinale, B. J., Byrnes, J. E. K., Hungate, B. A., Matulich, K. L., Gonzalez, A., Duffy, J. E., Gamfeldt, L., & Connor, M. I. (2012). A global synthesis reveals biodiversity loss as a major driver of ecosystem change. Nature, 486(7401), 105– 108. https://doi.org/10/f23wvj Huhta, E., & Sulkava, P. (2014). The impact of nature-based tourism on bird communities: A case study in Pallas-Yllästunturi National Park. Environmental Management, 53(5), 1005–1014. https://doi.org/10/ggzxp6 Hull, V., & Liu, J. (2018). Telecoupling: A new frontier for global sustainability. Ecology and Society, 23(4). https://doi.org/10/ghr6g8 Humbert, J.-Y., Pellet, J., Buri, P., & Arlettaz, R. (2012). Does delaying the first mowing date benefit biodiversity in meadowland? Environmental Evidence, 1(1), 9. https://doi.org/10.1186/2047-2382-1-9 Hutto, R. L. (1986). A Fixed-Radius Point Count Method for Nonbreeding and Breeding Season Use Author ( s ): Richard L. Hutto , Sandra M. Pletschet and Paul Hendricks Published by: American Ornithological Society Stable URL : http://www.jstor.org/stable/4087132 REFERENCES. Society, American Ornithological Auk, The, 103(3), 593–602. Hymowitz, T. (1970). On the Domestication of the Soybean. Economic Botany, 24(4), 408–421. https://doi.org/10/fq7845 Hymowitz, T., Newell, C. A. (1981). Taxonomy of the Genus Glycine, Domestication and Uses of Soybeans. Economic Botany, 35(August 1980), 272–288. https://doi.org/10.1038/nmeth.1449.Fast Hyun, P. S., Ok, K. J., Jae, J. K., & Bae, L. D. (2013). A study on the role of habitats of a tidal land reclamation area for winter migrants in Korea. Irrigation and Drainage, 62(S1), 44– 51. https://doi.org/10/gnn68d Jacoboski, L. I., & Hartz, S. M. (2020). Using functional diversity and taxonomic diversity to assess effects of afforestation of grassland on bird communities. Perspectives in Ecology and Conservation, 18(2), 103–108. https://doi.org/10.1016/j.pecon.2020.04.001 Jupiter, K. (2020). The function of open-field farming – managing time, work and space. Landscape History, 41(1), 69–98. https://doi.org/10.1080/01433768.2020.1753984 Kang, J., Zhang, H., Yang, H., & Zhang, L. (2018). Support Vector Machine Classification of Crop Lands Using Sentinel-2 Imagery (p. 6). https://doi.org/10.1109/Agro- Geoinformatics.2018.8476101 176 Kéry, M., & Schaub, M. (2012). Bayesian population analysis using WinBUGS: A hierarchical perspective (1st ed). Academic Press. Kleijn, D., Cherkaoui, I., Goedhart, P. W., van der Hout, J., & Lammertsma, D. (2014). Waterbirds increase more rapidly in Ramsar-designated wetlands than in unprotected wetlands. Journal of Applied Ecology, 51(2), 289–298. https://doi.org/10/gnn673 Krawczyk, E., Hedman, H., Pafilis, P., Bergen, K., & Foufopoulos, J. (2019). Effects of touristic development on Mediterranean island wildlife. Landscape Ecology, 34(11), 2719–2734. https://doi.org/10/gnn68g Kremen, C., & Merenlender, A. M. (2018). Landscapes that work for biodiversity and people. Science, 362(6412). https://doi.org/10/gfhzfq Kujawa, K., Wuczyński, A., Dajdok, Z., & Grzesiak, W. (2020). Effect of Habitat Structure and Crop Diversity on Common and Threatened Birds Breeding in Semi-Natural Field Margins. Acta Ornithologica, 54(2), 181–199. https://doi.org/10.3161/00016454AO2019.54.2.005 Lajeunesse, M. J. (2016). Facilitating systematic reviews, data extraction and meta-analysis with the metagear package for R. Methods in Ecology and Evolution, 7(3), 323–330. https://doi.org/10/f8gsbd Laliberté, E., & Legendre, P. (2010). A distance-based framework for measuring functional diversity from multiple traits. Ecology, 91(1), 299–305. https://doi.org/10/d6rqtj Laliberté, E., Legendre, P., & Shipley, B. (2014). FD: measuring functional diversity from multiple traits, and other tools for functional ecology [Manual]. Lambin, E. F., & Meyfroidt, P. (2011). Global land use change, economic globalization, and the looming land scarcity. Proceedings of the National Academy of Sciences of the United States of America, 108(9), 3465–3472. https://doi.org/10/cxqb7q Lasne, E., Sabatié, M. R., Jeannot, N., & Cucherousset, J. (2015). The Effects of DAM Removal on River Colonization by Sea Lamprey Petromyzon Marinus. River Research and Applications, 31(7), 904–911. https://doi.org/10/f7qkzr Lausch, A., Blaschke, T., Haase, D., Herzog, F., Syrbe, R.-U., Tischendorf, L., & Walz, U. (2015). Understanding and quantifying landscape structure – A review on relevant process characteristics, data models and landscape metrics. Ecological Modelling, 295, 31–41. https://doi.org/10/ghv99z Lavorel, S., Grigulis, K., McIntyre, S., Williams, N. S. G., Garden, D., Dorrough, J., Berman, S., Quétier, F., Thébault, A., & Bonis, A. (2008). Assessing functional diversity in the field – methodology matters! Functional Ecology, 22(1), 134–147. https://doi.org/10.1111/j.1365-2435.2007.01339.x 177 le Polain de Waroux, Y., Garrett, R. D., Graesser, J., Nolte, C., White, C., & Lambin, E. F. (2017). The Restructuring of South American Soy and Beef Production and Trade Under Changing Environmental Regulations. World Development. https://doi.org/10/ch9f Lebourgeois, V., Dupuy, S., Vintrou, É., Ameline, M., Butler, S., & Bégué, A. (2017). A combined random forest and OBIA classification scheme for mapping smallholder agriculture at different nomenclature levels using multisource data (simulated Sentinel-2 time series, VHRS and DEM). Remote Sensing, 9(3), 1–20. https://doi.org/10.3390/rs9030259 Lee, M. B., & Goodale, E. (2018). Crop heterogeneity and non-crop vegetation can enhance avian diversity in a tropical agricultural landscape in southern China. Agriculture, Ecosystems and Environment, 265, 254–263. https://doi.org/10/gd74qh Lee, M.-B., & Martin, J. A. (2017). Avian Species and Functional Diversity in Agricultural Landscapes: Does Landscape Heterogeneity Matter? PLoS ONE, 12(1), 1–21. https://doi.org/10.1371/ Leff, B., Ramankutty, N., & Foley, J. A. (2004). Geographic distribution of major crops across the world. Global Biogeochemical Cycles, 18(1), n/a-n/a. https://doi.org/10/bnhfkk Lenzen, M., Moran, D., Kanemoto, K., Foran, B., Lobefaro, L., & Geschke, A. (2012). International trade drives biodiversity threats in developing nations. Nature, 486(7401), 109–112. https://doi.org/10/f32b9h Li, D., Lee, M. B., Xiao, W., Tang, J., & Zhang, Z. (2020). Noncrop features and heterogeneity mediate overwintering bird diversity in agricultural landscapes of southwest China. Ecology and Evolution, 10(12), 5815–5828. https://doi.org/10/gnn675 Li, L., Hu, R., Huang, J., Bürgi, M., Zhu, Z., Zhong, J., & Lü, Z. (2020). A farmland biodiversity strategy is needed for China. Nature Ecology and Evolution, 4(6), 772–774. https://doi.org/10/gh4w28 Li, L., Yang, T., Liu, R., Redden, B., Maalouf, F., & Zong, X. (2017). Food legume production in China. Crop Journal, 5(2), 115–126. https://doi.org/10/gm2nrz Li, M., Ma, L., Blaschke, T., Cheng, L., & Tiede, D. (2016). A systematic comparison of different object-based classification techniques using high spatial resolution imagery in agricultural environments. International Journal of Applied Earth Observation and Geoinformation, 49, 87–98. https://doi.org/10.1016/j.jag.2016.01.011 Li, W., He, S., Cheng, X., & Zhang, M. (2021). Functional diversity outperforms taxonomic diversity in revealing short-term trampling effects. Scientific Reports, 11(1), 18889. https://doi.org/10.1038/s41598-021-98372-3 Li, X., Yu, L., Peng, D., & Gong, P. (2021). A large-scale, long time-series (1984‒2020) of soybean mapping with phenological features: Heilongjiang Province as a test case. 178 International Journal of Remote Sensing, 42(19), 7332–7356. https://doi.org/10.1080/01431161.2021.1957177 Liao, J., Liao, T., He, X., Zhang, T., Li, D., Luo, X., Wu, Y., & Ran, J. (2020). The effects of agricultural landscape composition and heterogeneity on bird diversity and community structure in the Chengdu Plain, China. Global Ecology and Conservation, 24. https://doi.org/10/gnn674 Lima, A. C., Assis, J., Sayanda, D., Sabino, J., & Oliveira, R. F. (2014). Impact of ecotourism on the fish fauna of Bonito region (Mato Grosso do Sul State, Brazil): Ecological, behavioural and physiological measures. Neotropical Ichthyology, 12(1), 133–143. https://doi.org/10/f54dh7 Lima, A. C., Sayanda, D., Agostinho, C. S., Machado, A. L., Soares, A. M. V. M., & Monaghan, K. A. (2018). Using a trait-based approach to measure the impact of dam closure in fish communities of a Neotropical River. Ecology of Freshwater Fish, 27(1), 408–420. https://doi.org/10/gcq7hf Lindsay, K. E., KIRK, D. A., BERGIN, T. M., BEST, L. B., SIFNEOS, J. C., SMITH, J., & Sifneos, J. C. (2013). Farmland Heterogeneity Benefits Birds in American Mid-west Watersheds. American Midland Naturalist, 170(1), 121–143. https://doi.org/10.1674/0003-0031-170.1.121 Liu, C., Chen, Z., Shao, Y., Chen, J., Hasi, T., & Pan, H. (2019). Research advances of SAR remote sensing for agriculture applications: A review. Journal of Integrative Agriculture, 18(3), 506–525. https://doi.org/10.1016/S2095-3119(18)62016-7 Liu, J. (2020). Telecoupling. In D. Richardson, N. Castree, M. F. Goodchild, A. Kobayashi, W. Liu, & R. A. Marston (Eds.), International Encyclopedia of Geography (pp. 1–8). Wiley. https://doi.org/10.1002/9781118786352.wbieg2049 Liu, J. (2021). Consumption patterns and biodiversity. Programme of the Royal Society, 1–16. 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. F., Naylor, R., Ouyang, Z., Polenske, K. R., Reenberg, A., de Miranda Rocha, G., Simmons, C. S., … Zhu, C. (2013). Framing sustainability in a telecoupled world. Ecology and Society, 18(2), 26. https://doi.org/10/gbdc2h Liu, J., Yang, W., & Li, S. (2016). Framing ecosystem services in the telecoupled Anthropocene. Frontiers in Ecology and the Environment, 14(1), 27–36. https://doi.org/10/f78wg8 Liu, Y., Duan, M., & Yu, Z. (2013). Agricultural landscapes and biodiversity in China. Agriculture, Ecosystems and Environment, 166, 46–54. https://doi.org/10/cxwtfr Lunn, D. J., Thomas, A., Best, N., & Spiegelhalter, D. (2000). WinBUGS – a Bayesian modelling framework: Concepts, structure, and extensibility. Statistics and Computing, 10, 325–337. https://doi.org/10.1023/A:1008929526011 179 Lutz, E. (1998). Agriculture and the Environment—Perspectives on Sustainable Rural Development. In Agriculture and the Environment—Perspectives on Sustainable Rural Development (Vol. 1). https://doi.org/10.1017/S1355770X01250281 MacArthur, R. H., & MacArthur, J. W. (1961). On Bird Species Diversity. Ecology, 42(3), 594– 598. https://doi.org/10.2307/1932254 MacArthur, R. H., & Wilson, E. O. (1967). The theory of island biogeography. Princeton University Press. Mace, G. M., Norris, K., & Fitter, A. H. (2012). Biodiversity and ecosystem services: A multilayered relationship. Trends in Ecology and Evolution, 27(1), 19–26. https://doi.org/10/fb9h7m MacKenzie, D. I. (2018). Occupancy estimation and modeling: Inferring patterns and dynamics of species occurrence (Second edition). Academic Press, an imprint of Elsevier. Mantelatto, M. C., Silva, A. G. da, Louzada, T. dos S., McFadden, C. S., & Creed, J. C. (2018). Invasion of aquarium origin soft corals on a tropical rocky reef in the southwest Atlantic, Brazil. Marine Pollution Bulletin, 130, 84–94. https://doi.org/10/gdvg8j Marquardt, S. G., Guindon, M., Wilting, H. C., Steinmann, Z. J. N., Sim, S., Kulak, M., & Huijbregts, M. A. J. (2019). Consumption-based biodiversity footprints – Do different indicators yield different results? Ecological Indicators, 103, 461–470. https://doi.org/10/gm5nr8 Marques, A., Martins, I. S., Kastner, T., Plutzar, C., Theurl, M. C., Eisenmenger, N., Huijbregts, M. A. J., Wood, R., Stadler, K., Bruckner, M., Canelas, J., Hilbers, J. P., Tukker, A., Erb, K., & Pereira, H. M. (2019). Increasing impacts of land use on biodiversity and carbon sequestration driven by population and economic growth. Nature Ecology and Evolution, 3(4), 628–637. https://doi.org/10/c3fj Martin, A. E., Collins, S. J., Crowe, S., Girard, J., Naujokaitis-Lewis, I., Smith, A. C., Lindsay, K., Mitchell, S., & Fahrig, L. (2020). Effects of farmland heterogeneity on biodiversity are similar to—Or even larger than—The effects of farming practices. Agriculture, Ecosystems & Environment, 288, 106698. https://doi.org/10.1016/j.agee.2019.106698 Martín, C. A., Martínez, C., Bautista, L. M., & Martín, B. (2012). Population increase of the great bustard otis tarda in its main distribution area in relation to changes in farming practices. Ardeola, 59(1), 31–42. https://doi.org/10/f4mv84 Mason, N. W. H., Mouillot, D., Lee, W. G., Wilson, J. B., & Setälä, H. (2005). Functional Richness, Functional Evenness and Functional Divergence: The Primary Components of Functional Diversity. Oikos, 111(1), 112–118. https://doi.org/10/cmwgjt Mcgill, B., Enquist, B., Weiher, E., & Westoby, M. (2006). Rebuilding community ecology from functional traits. Trends in Ecology & Evolution, 21(4), 178–185. https://doi.org/10.1016/j.tree.2006.02.002 180 Meichtry-Stier, K. S., Duplain, J., Lanz, M., Lugrin, B., & Birrer, S. (2018). The importance of size, location, and vegetation composition of perennial fallows for farmland birds. Ecology and Evolution, 8(18), 9270–9281. https://doi.org/10.1002/ece3.4420 Meyer, O. L., Du Preez, L., Bonneau, E., Héritier, L., Franch, M., Valdeón, A., Sadaoui, A., Kechemir-Issad, N., Palacios, C., & Verneau, O. (2015). Parasite host-switching from the invasive american red-eared slider, trachemys scripta elegans, to the native mediterranean pond turtle, mauremys leprosa, in natural environments. Aquatic Invasions, 10(1), 79–91. https://doi.org/10/gnn68h Michler, J. D., & Josephson, A. L. (2017). To Specialize or Diversify: Agricultural Diversity and Poverty Dynamics in Ethiopia. World Development, 89, 214–226. https://doi.org/10.1016/j.worlddev.2016.08.011 Moher, D., Liberati, A., Tetzlaff, J., & Altman, D. G. (2010). Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. International Journal of Surgery, 8(5), 336–341. https://doi.org/10/d969xc Mokotjomela, T. M., & Nombewu, N. (2020). Potential benefits associated with implementation of the national biodiversity economy strategy in the Eastern Cape Province, South Africa. South African Geographical Journal, 102(2), 190–208. https://doi.org/10/gnn677 Moraga, C. A., Funes, M. C., Pizarro, J. C., Briceño, C., & Novaro, A. J. (2015). Effects of livestock on guanaco Lama guanicoe density, movements and habitat selection in a forest-grassland mosaic in Tierra del Fuego, Chile. Oryx, 49(1), 30–41. https://doi.org/10/f6tjr2 Morelli, F., Benedetti, Y., & Šímová, P. (2018). Landscape metrics as indicators of avian diversity and community measures. Ecological Indicators, 90, 132–141. https://doi.org/10/gnn68m Morris, E. K., Caruso, T., Buscot, F., Fischer, M., Hancock, C., Maier, T. S., Meiners, T., Müller, C., Obermaier, E., Prati, D., Socher, S. A., Sonnemann, I., Wäschke, N., Wubet, T., Wurst, S., & Rillig, M. C. (2014). Choosing and using diversity indices: Insights for ecological applications from the German Biodiversity Exploratories. Ecology and Evolution, 4(18), 3514–3524. https://doi.org/10.1002/ece3.1155 Muldavin, J. S. S. (1997). Environmental Degradation in Heilongjiang: Policy Reform and Agrarian Dynamics in China’s New Hybrid Economy. Annals of the Association of American Geographers, 87(4), 579–613. https://doi.org/10/c8w575 Newbold, T., Hudson, L. N., Hill, S. L. L., Contu, S., Lysenko, I., Senior, R. A., Börger, L., Bennett, D. J., Choimes, A., Collen, B., Day, J., De Palma, A., Díaz, S., Echeverria- Londoño, S., Edgar, M. J., Feldman, A., Garon, M., Harrison, M. L. K., Alhusseini, T., … Purvis, A. (2015). Global effects of land use on local terrestrial biodiversity. Nature, 520(7545), 45–50. https://doi.org/10.1038/nature14324 181 Nimubona, A. D. (2012). Pollution Policy and Trade Liberalization of Environmental Goods. Environmental and Resource Economics, 53(3), 323–346. https://doi.org/10/f4g9jx Nishijima, S., Furukawa, T., Kadoya, T., Ishihama, F., Kastner, T., Matsuda, H., & Kaneko, N. (2016). Evaluating the impacts of wood production and trade on bird extinction risks. Ecological Indicators, 71, 368–376. https://doi.org/10/f9cnhb Nobre, R. L. G., Caliman, A., Guariento, R. D., Bozelli, R. L., & Carneiro, L. S. (2019). Effects of the introduction of an omnivorous fish on the biodiversity and functioning of an upland Amazonian lake. Acta Amazonica, 49(3), 221–231. https://doi.org/10/gnn67x Norris, K. (2008). Agriculture and biodiversity conservation: Opportunity knocks. Conservation Letters, 1(1), 2–11. https://doi.org/10/ddqw46 Odorico, P. D., Carr, J. A., Laio, F., Ridolfi, L., & Vandoni, S. (2014). Earth ’ s Future Feeding humanity through global food trade Earth ’ s Future. 458–469. https://doi.org/10.1002/2014EF000250.Abstract Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D., Minchin, P. R., O’Hara, R. B., Simpson, G. L., Solymos, P., Stevens, M. H. H., Szoecs, E., & Wagner, H. (2020). vegan: Community ecology package [Manual]. https://CRAN.R- project.org/package=vegan Oldekop, J. A., Sims, K. R. E., Whittingham, M. J., & Agrawal, A. (2018). An upside to globalization: International outmigration drives reforestation in Nepal. Global Environmental Change, 52(June), 66–74. https://doi.org/10/gfpzvr Oliver, T. H., Heard, M. S., Isaac, N. J. B., Roy, D. B., Procter, D., Eigenbrod, F., Freckleton, R., Hector, A., Orme, C. D. L., Petchey, O. L., Proença, V., Raffaelli, D., Suttle, K. B., Mace, G. M., Martín-López, B., Woodcock, B. A., & Bullock, J. M. (2015). Biodiversity and Resilience of Ecosystem Functions. Trends in Ecology and Evolution, 30(11), 673– 684. https://doi.org/10/f7zf26 Olson, D. M., Dinerstein, E., Wikramanayake, E. D., Burgess, N. D., Powell, G. V. N., Underwood, E. C., D’amico, J. A., Itoua, I., Strand, H. E., Morrison, J. C., Loucks, C. J., Allnutt, T. F., Ricketts, T. H., Kura, Y., Lamoreux, J. F., Wettengel, W. W., Hedao, P., & Kassem, K. R. (2001). Terrestrial Ecoregions of the World: A New Map of Life on Earth. BioScience, 51(11), 933. https://doi.org/10/c635xt Orynbaikyzy, A., Gessner, U., & Conrad, C. (2019). Crop type classification using a combination of optical and radar remote sensing data: A review. International Journal of Remote Sensing, 40(17), 6553–6595. https://doi.org/10.1080/01431161.2019.1569791 Osterman, L. E., Poore, R. Z., Swarzenski, P. W., Senn, D. B., & DiMarco, S. F. (2009). The 20th-century development and expansion of Louisiana shelf hypoxia, Gulf of Mexico. Geo-Marine Letters, 29(6), 405–414. https://doi.org/10/cft48k 182 Pejchar, L., Clough, Y., Ekroos, J., Nicholas, K. A., Olsson, O., Ram, D., Tschumi, M., & Smith, H. G. (2018). Net Effects of Birds in Agroecosystems. BioScience, 68(11), 896–904. https://doi.org/10/ghpwbm Perfecto, I., & Vandermeer, J. (2010). The agroecological matrix as alternative to the land- sparing/agriculture intensification model. Proceedings of the National Academy of Sciences of the United States of America, 107(13), 5786–5791. https://doi.org/10/dspjcx Petchey, O. L., & Gaston, K. J. (2006). Functional diversity: Back to basics and looking forward. Ecology Letters, 9(6), 741–758. https://doi.org/10.1111/j.1461-0248.2006.00924.x Pimm, S. L., Jenkins, C. N., Abell, R., Brooks, T. M., Gittleman, J. L., Joppa, L. N., Raven, P. H., Roberts, C. M., & Sexton, J. O. (2014). The biodiversity of species and their rates of extinction, distribution, and protection. Science, 344(6187), 1246752. https://doi.org/10/szx Pimm, S., Raven, P., Peterson, A., Şekercioǧlu, Ç. H., & Ehrlich, P. R. (2006). Human impacts on the rates of recent, present, and future bird extinctions. Proceedings of the National Academy of Sciences of the United States of America, 103(29), 10941–10946. https://doi.org/10/fnf786 Porkka, M., Kummu, M., Siebert, S., & Varis, O. (2013). From Food Insufficiency towards Trade Dependency: A Historical Analysis of Global Food Availability. PLOS ONE, 8(12), e82714. https://doi.org/10.1371/journal.pone.0082714 Powers, R. P., & Jetz, W. (2019). Global habitat loss and extinction risk of terrestrial vertebrates under future land-use-change scenarios. Nature Climate Change, 9(4), 323–329. https://doi.org/10/ggdtkg Preissel, S., Zander, P., & Knierim, A. (2017). Sustaining Farming on Marginal Land: Farmers’ Convictions, Motivations and Strategies in Northeastern Germany. Sociologia Ruralis, 57(S1), 682–708. https://doi.org/10.1111/soru.12168 Prell, C., Sun, L., Feng, K., He, J., & Hubacek, K. (2017). Uncovering the spatially distant feedback loops of global trade: A network and input-output approach. Science of The Total Environment, 586(16), 401–408. https://doi.org/10/f94248 Prugh, L. R., Hodges, K. E., Sinclair, A. R. E., & Brashares, J. S. (2008). Effect of habitat area and isolation on fragmented animal populations. Proceedings of the National Academy of Sciences, 105(52), 20770–20775. https://doi.org/10/dzb4wp Qi, J., Marsett, R., Heilman, P., Biedenbender, S., Moran, S., Goodrich, D., & Weltz, M. (2002). RANGES improves satellite-based information and land cover assessments in Southwest United States. Eos, 83(51), 601+605-606. Scopus. https://doi.org/10.1029/2002EO000411 R Core Team. (2021). R: A language and environment for statistical computing [Manual]. https://www.R-project.org/ 183 Redlich, S., Martin, E. A., Wende, B., & Steffan-Dewenter, I. (2018). Landscape heterogeneity rather than crop diversity mediates bird diversity in agricultural landscapes. PLOS ONE, 13(8), e0200438. https://doi.org/10/gn7jpn Reenberg, A., & Fenger, N. A. (2011). Globalizing land use transitions: The soybean acceleration. Geografisk Tidsskrift-Danish Journal of Geography, 111(1), 85–92. https://doi.org/10/gnn63x Reif, J., & Vermouzek, Z. (2019). Collapse of farmland bird populations in an Eastern European country following its EU accession. Conservation Letters, 12(1). https://doi.org/10/gf69ng Remis, M. J., & Jost Robinson, C. A. (2012). Reductions in Primate Abundance and Diversity in a Multiuse Protected Area: Synergistic Impacts of Hunting and Logging in a Congo Basin Forest. American Journal of Primatology, 74(7), 602–612. https://doi.org/10/f34rwd Remsen, J. V. Jr., & Robinson, S. K. (1990). A classification scheme for foraging behavor of birds in terrestrial habitats. Studies in Avian Biology, 13, 144–160. Reyes-Martínez, M. J., Ruíz-Delgado, M. C., Sánchez-Moyano, J. E., & García-García, F. J. (2015). Response of intertidal sandy-beach macrofauna to human trampling: An urban vs. Natural beach system approach. Marine Environmental Research, 103, 36–45. https://doi.org/10/f6xv4t Robinson, T., & Pozzi, F. (2011). Mapping Supply and Demand for Animal-Source Foods to 2030 (Animal Production and Health Working Paper). FAO. Rouse, W., & Haas, R. H. (1974). MONITORING VEGETATION SYSTEMS IN THE GREAT PLAINS WITH ERTS. 9. Rutledge, D. (2003). Landscape indices as measures of the effects of fragmentation: Can pattern reflect process ? DOC Science Internal Series 98, 1–27. https://doi.org/10/fgj8bv Ruxton, G. D., & Hansell, M. H. (2011). Fishing with a Bait or Lure: A Brief Review of the Cognitive Issues. Ethology, 117(1), 1–9. https://doi.org/10.1111/j.1439- 0310.2010.01848.x Sala, O. E., Stuart Chapin, F., III, Armesto, J. J., Berlow, E., Bloomfield, J., Dirzo, R., Huber- Sanwald, E., Huenneke, L. F., Jackson, R. B., Kinzig, A., Leemans, R., Lodge, D. M., Mooney, H. A., Oesterheld, M., Poff, N. L., Sykes, M. T., Walker, B. H., Walker, M., & Wall, D. H. (2000). Global Biodiversity Scenarios for the Year 2100. Science, 287(5459), 1770–1774. https://doi.org/10.1126/science.287.5459.1770 Sanderson, F. J., Pople, R. G., Ieronymidou, C., Burfield, I. J., Gregory, R. D., Willis, S. G., Howard, C., Stephens, P. A., Beresford, A. E., & Donald, P. F. (2016). Assessing the Performance of EU Nature Legislation in Protecting Target Bird Species in an Era of Climate Change. Conservation Letters, 9(3), 172–180. https://doi.org/10/f8tmdk 184 Sandström, V., Saikku, L., Antikainen, R., Sokka, L., & Kauppi, P. (2014). Changing impact of import and export on agricultural land use: The case of Finland 1961-2007. Agriculture, Ecosystems and Environment, 188, 163–168. https://doi.org/10/gnn63q Santos, G. S., Burgos, D. C., Lira, S. M. A., & Schwamborn, R. (2015). The Impact of Trampling on Reef Macrobenthos in Northeastern Brazil: How Effective are Current Conservation Strategies? Environmental Management, 56(4), 847–858. https://doi.org/10/f7qh5b Sá-Oliveira, J. C., Hawes, J. E., Isaac-Nahum, V. J., & Peres, C. A. (2015). Upstream and downstream responses of fish assemblages to an eastern Amazonian hydroelectric dam. Freshwater Biology, 60(10), 2037–2050. https://doi.org/10/f7rgff Sapkota, T. B., Mazzoncini, M., Bàrberi, P., Antichi, D., & Silvestri, N. (2012). Fifteen years of no till increase soil organic matter, microbial biomass and arthropod diversity in cover crop-based arable cropping systems. Agronomy for Sustainable Development, 32(4), 853– 863. https://doi.org/10.1007/s13593-011-0079-0 Sarmento, V. C., & Santos, P. J. P. (2012). Trampling on coral reefs: Tourism effects on harpacticoid copepods. Coral Reefs, 31(1), 135–146. https://doi.org/10/cfnv85 Sato, C. F., Wood, J. T., & Lindenmayer, D. B. (2013). The Effects of Winter Recreation on Alpine and Subalpine Fauna: A Systematic Review and Meta-Analysis. PLoS ONE, 8(5), e64282. https://doi.org/10/f4w6xj Schaffer-Smith, D., Tomscha, S. A., Jarvis, K. J., Maguire, D. Y., Treglia, M. L., & Liu, J. (2018). Network analysis as a tool for quantifying the dynamics of metacoupled systems: An example using global soybean trade. Ecology and Society, 23(4), art3. https://doi.org/10.5751/ES-10460-230403 Scheele, B. C., Pasmans, F., Skerratt, L. F., Berger, L., Martel, A., Beukema, W., Acevedo, A. A., Burrowes, P. A., Carvalho, T., Catenazzi, A., De La Riva, I., Fisher, M. C., Flechas, S. V., Foster, C. N., Frías-Álvarez, P., Garner, T. W. J., Gratwicke, B., Guayasamin, J. M., Hirschfeld, M., … Canessa, S. (2019). Amphibian fungal panzootic causes catastrophic and ongoing loss of biodiversity. Science, 363(6434), 1459–1463. https://doi.org/10/gfxkrt Selier, S. A. J., Slotow, R., & Di Minin, E. (2016). The influence of socioeconomic factors on the densities of high-value cross-border species, the African elephant. PeerJ, 2016(10). https://doi.org/10/gnn68b Seshadri, K. S., & Ganesh, T. (2011). Faunal mortality on roads due to religious tourism across time and space in protected areas: A case study from south India. Forest Ecology and Management, 262(9), 1713–1721. https://doi.org/10/d5nsx6 Shochat, E., Lerman, S. B., Anderies, J. M., Warren, P. S., Faeth, S. H., & Nilon, C. H. (2010). Invasion, competition, and biodiversity loss in Urban ecosystems. BioScience, 60(3), 199–208. https://doi.org/10/bn8w69 185 Sieges, M. L., Smolinsky, J. A., Baldwin, M. J., Barrow, W. C., Randall, L. A., & Buler, J. J. (2014). Assessment of bird response to the migratory bird habitat initiative using weather-surveillance radar. Southeastern Naturalist, 13(1), G36–G65. https://doi.org/10/gnn672 Šilc, U., Dajić Stevanović, Z., Ibraliu, A., Luković, M., & Stešević, D. (2016). Human impact on sandy beach vegetation along the southeastern Adriatic coast. Biologia (Poland), 71(8), 865–874. https://doi.org/10/gnn66v Singh, D., Slik, J. W. F., Jeon, Y. S., Tomlinson, K. W., Yang, X., Wang, J., Kerfahi, D., Porazinska, D. L., & Adams, J. M. (2019). Tropical forest conversion to rubber plantation affects soil micro- & mesofaunal community & diversity. Scientific Reports, 9(1). https://doi.org/10/gjrk6g Smith, O. M., Kennedy, C. M., Echeverri, A., Karp, D. S., Latimer, C. E., Taylor, J. M., Wilson- Rankin, E. E., Owen, J. P., & Snyder, W. E. (2021). Complex landscapes stabilize farm bird communities and their expected ecosystem services. Journal of Applied Ecology, n/a(n/a). https://doi.org/10/gn7jpm Smith, V. H., & Glauber, J. W. (2020). Trade, policy, and food security. Agricultural Economics, 51(1), 159–171. https://doi.org/10.1111/agec.12547 Sodhi, N. S., Bickford, D., Diesmos, A. C., Lee, T. M., Koh, L. P., Brook, B. W., Sekercioglu, C. H., & Bradshaw, C. J. A. (2008). Measuring the meltdown: Drivers of global amphibian extinction and decline. PLoS ONE, 3(2), e1636. https://doi.org/10/bnhvbf Souza, D. M., Teixeira, R. F. M., & Ostermann, O. P. (2015). Assessing biodiversity loss due to land use with Life Cycle Assessment: Are we there yet? Global Change Biology, 21(1), 32–47. https://doi.org/10.1111/gcb.12709 Stehfest, E., van Zeist, W. J., Valin, H., Havlik, P., Popp, A., Kyle, P., Tabeau, A., Mason- D’Croz, D., Hasegawa, T., Bodirsky, B. L., Calvin, K., Doelman, J. C., Fujimori, S., Humpenöder, F., Lotze-Campen, H., van Meijl, H., & Wiebe, K. (2019). Key determinants of global land-use projections. Nature Communications, 10(1), 2166. https://doi.org/10/gh6qxz Sudmanns, M., Tiede, D., Augustin, H., & Lang, S. (2020). Assessing global Sentinel-2 coverage dynamics and data availability for operational Earth observation (EO) applications using the EO-Compass. International Journal of Digital Earth, 13(7), 768–784. https://doi.org/10.1080/17538947.2019.1572799 Sun, J., Mooney, H., Wu, W., Tang, H., Tong, Y., Xu, Z., Huang, B., & Cheng, Y. (2018). Importing food damages domestic environment: Evidence from global soybean trade. https://doi.org/10/gdms6g Sun, J., Wu, W., Tang, H., & Liu, J. (2015). Spatiotemporal patterns of non-genetically modified crops in the era of expansion of genetically modified food. Scientific Reports, 5, 14180. https://doi.org/10/f7rvgm 186 Sun, Y., Wang, T., Skidmore, A. K., Wang, Q., & Ding, C. (2015). Decline of traditional rice farming constrains the recovery of the endangered Asian crested ibis (Nipponia nippon). Ambio, 44(8), 803–814. https://doi.org/10/f7z4w7 Survey Office of the National Bureau of Statistics in Heilongjiang. (2013). Heilongjiang Provincial Bureau of Statistics. In Heilongjiang Statistical Yearbook. China Statistics Press. Tanentzap, A. J., Lamb, A., Walker, S., Farmer, A., Valenzuela, E., & Kremen, C. (2015). Resolving Conflicts between Agriculture and the Natural Environment. PLOS Biology, 13(9), e1002242. https://doi.org/10/f7tns8 Tarrant, J. (1985). A Review of International Food Trade. Progress in Human Geography, 9(2), 235–254. https://doi.org/10.1177/030913258500900204 Tews, J., Brose, U., Grimm, V., Tielbörger, K., Wichmann, M. C., Schwager, M., & Jeltsch, F. (2004). Animal species diversity driven by habitat heterogeneity/diversity: The importance of keystone structures. Journal of Biogeography, 31(1), 79–92. https://doi.org/10.1046/j.0305-0270.2003.00994.x Thomas, E. H., Brittingham, M. C., & Stoleson, S. H. (2014). Conventional oil and gas development alters forest songbird communities. Journal of Wildlife Management, 78(2), 293–306. https://doi.org/10/f5t5q3 Tilman, D., Balzer, C., Hill, J., & Befort, B. L. (2011). Global food demand and the sustainable intensification of agriculture. Proceedings of the National Academy of Sciences of the United States of America, 108(50), 20260–20264. https://doi.org/10/ds75t3 Tilt, B. (2008). Smallholders and the “household responsibility system”: Adapting to institutional change in Chinese agriculture. Human Ecology, 36(2), 189–199. https://doi.org/10.1007/s10745-007-9127-4 Tscharntke, T., Clough, Y., Wanger, T. C., Jackson, L., Motzke, I., Perfecto, I., Vandermeer, J., & Whitbread, A. (2012). Global food security, biodiversity conservation and the future of agricultural intensification. Biological Conservation, 151(1), 53–59. https://doi.org/10/f34kx3 U.S. Census Bureau, U. S. C. B. (2019). International Data Base. USGS US Dept of the Interior. (2017). Landsat Missions. Van Bocxlaer, B., & Albrecht, C. (2015). Ecosystem change and establishment of an invasive snail alter gastropod communities in long-lived Lake Malawi. Hydrobiologia, 744(1), 307–316. https://doi.org/10/f6s76m Van Deventer, A. P., Ward, A. D., Gowda, P. M., & Lyon, J. G. (1997). Using thematic mapper data to identify contrasting soil plains and tillage practices. Photogrammetric Engineering and Remote Sensing, 63(1), 87–93. Scopus. 187 van Klink, R., Bowler, D. E., Gongalsky, K. B., Swengel, A. B., Gentile, A., & Chase, J. M. (2020). Meta-analysis reveals declines in terrestrial but increases in freshwater insect abundances. Science, 368(6489), 417–420. https://doi.org/10/ghhmwc van Vliet, J., de Groot, H. L. F., Rietveld, P., & Verburg, P. H. (2015). Manifestations and underlying drivers of agricultural land use change in Europe. Landscape and Urban Planning, 133, 24–36. https://doi.org/10.1016/j.landurbplan.2014.09.001 Vanwey, L. K., Richards, P. D., Lathuillière, M. J., Johnson, M. S., Galford, G. L., Garrett, R. D., Carlson, K. M., Rueda, X., Henders, S., Persson, M., Kastner, T., Godar, J., Suavet, C., Gardner, T. A., Furumo, P. R., Aide, T. M., Zaks, D. P. M., Barford, C. C., Ramankutty, N., & Lambin, E. F. (2013). Globalization’s unexpected impact on soybean production in South America: Linkages between preferences for non-genetically modified crops, eco-certifications, and land use. Environ. Res. Lett, 8, 44055–11. https://doi.org/10/ghgb58 Vasseur, C., Joannon, A., Aviron, S., Burel, F., Meynard, J. M., & Baudry, J. (2013). The cropping systems mosaic: How does the hidden heterogeneity of agricultural landscapes drive arthropod populations? Agriculture, Ecosystems and Environment, 166, 3–14. https://doi.org/10/fx6bjm Villéger, Mason, & Mouillot. (2008). New multidimensional functional diversity indices for a multifaceted framwork in functional ecology. Ecology, 89(8), 2290–2301. https://doi.org/10/djg9r7 Vogels, M. F. A., de Jong, S. M., Sterk, G., Douma, H., & Addink, E. A. (2019). Spatio-temporal patterns of smallholder irrigated agriculture in the horn of Africa using GEOBIA and Sentinel-2 imagery. Remote Sensing, 11(2). https://doi.org/10.3390/rs11020143 Wake, D. B., & Vredenburg, V. T. (2008). Are we in the midst of the sixth mass extinction? A view from the world of amphibians. Proceedings of the National Academy of Sciences of the United States of America, 105(SUPPL. 1), 11466–11473. https://doi.org/10/ftcq74 Waldron, A., Miller, D. C., Redding, D., Mooers, A., Kuhn, T. S., Nibbelink, N., Roberts, J. T., Tobias, J. A., & Gittleman, J. L. (2017). Reductions in global biodiversity loss predicted from conservation spending. Nature, 551(7680), 364–367. https://doi.org/10/gcf5nv Walz, U., & Wellisch, D. (1997). Is free trade in the interest of exporting countries when there is ecological dumping? Journal of Public Economics, 66(2), 275–291. https://doi.org/10/bfz75t Ward, J. H. (1963). Hierarchical Grouping to Optimize an Objective Function. Journal of the American Statistical Association, 58(301), 236–244. https://doi.org/10.1080/01621459.1963.10500845 Weather and Climate. (2016). https://weather-and-climate.com/average-monthly-Rainfall- Temperature-Sunshine,harbin,China 188 Weinzettel, J., Hertwich, E. G., Peters, G. P., Steen-Olsen, K., & Galli, A. (2013). Affluence drives the global displacement of land use. Global Environmental Change, 23(2), 433– 438. https://doi.org/10.1016/j.gloenvcha.2012.12.010 Wiens, J. A. (1976). Population Responses to Patchy Environments. Annual Review of Ecology and Systematics, 7(1), 81–120. https://doi.org/10/czttbc Willett, W., Rockström, J., Loken, B., Springmann, M., Lang, T., Vermeulen, S., Garnett, T., Tilman, D., DeClerck, F., Wood, A., Jonell, M., Clark, M., Gordon, L. J., Fanzo, J., Hawkes, C., Zurayk, R., Rivera, J. A., Vries, W. D., Sibanda, L. M., … Murray, C. J. L. (2019). Food in the Anthropocene: The EAT–Lancet Commission on healthy diets from sustainable food systems. The Lancet, 393(10170), 447–492. https://doi.org/10.1016/S0140-6736(18)31788-4 Wilting, H. C., Schipper, A. M., Bakkenes, M., Meijer, J. R., & Huijbregts, M. A. J. (2017). Quantifying Biodiversity Losses Due to Human Consumption: A Global-Scale Footprint Analysis. Environmental Science and Technology, 51(6), 3298–3306. https://doi.org/10/gm5nrf Wolf, I. D., Hagenloh, G., & Croft, D. B. (2013). Vegetation moderates impacts of tourism usage on bird communities along roads and hiking trails. Journal of Environmental Management, 129, 224–234. https://doi.org/10/f5kkjp Wood, C., Qiao, Y., Li, P., Ding, P., Lu, B., & Xi, Y. (2017). Implications of Rice Agriculture for Wild Birds in China Author ( s ): Chris Wood , Yi Qiao , Peng Li , Ping Ding , Baozhong Lu and Yongmei Xi Source: Waterbirds: The International Journal of Waterbird Biology , Vol. 33 , Special Publication 1: Ecolog. 33(2010), 30–43. Wood, D., & Lenné, J. M. (1999). Agrobiodiversity: Characterization, Utilitization, and Management. CABI Pub. Wood, S. A., Jina, A. S., Jain, M., Kristjanson, P., & DeFries, R. S. (2014). Smallholder farmer cropping decisions related to climate variability across multiple regions. Global Environmental Change, 25(1), 163–172. https://doi.org/10.1016/j.gloenvcha.2013.12.011 Wood, S. A., Karp, D. S., DeClerck, F., Kremen, C., Naeem, S., & Palm, C. A. (2015). Functional traits in agriculture: Agrobiodiversity and ecosystem services. Trends in Ecology and Evolution, 30(9), 531–539. https://doi.org/10.1016/j.tree.2015.06.013 Xu, Z., Li, Y., Chau, S. N., Dietz, T., Li, C., Wan, L., Zhang, J., Zhang, L., Li, Y., Chung, M. G., & Liu, J. (2020). Impacts of international trade on global sustainable development. Nature Sustainability, 3(11), 964–971. https://doi.org/10/gg4vzm Yang, Z., Tang, H., Tao, J., & Zhao, N. (2017). The effect of cascaded huge dams on the downstream movement of Coreius guichenoti (Sauvage & Dabry de Thiersant, 1874) in the upper Yangtze River. Environmental Biology of Fishes, 100(11), 1507–1516. https://doi.org/10/gcmwgk 189 Yao, G., Hertel, T. W., & Taheripour, F. (2018). Economic drivers of telecoupling and terrestrial carbon fluxes in the global soybean complex. Global Environmental Change, 50(November 2017), 190–200. https://doi.org/10/gds4mg You, N., Dong, J., Huang, J., Du, G., Zhang, G., He, Y., Yang, T., Di, Y., & Xiao, X. (2021). The 10-m crop type maps in Northeast China during 2017–2019. Scientific Data, 8(1). https://doi.org/10/ghzp75 Zhang, H., Kang, J., Xu, X., & Zhang, L. (2020). Accessing the temporal and spectral features in crop type mapping using multi-temporal Sentinel-2 imagery: A case study of Yi’an County, Heilongjiang province, China. Computers and Electronics in Agriculture, 176. https://doi.org/10/gh7tg2 Zhong, L., Hu, L., Yu, L., Gong, P., & Biging, G. S. (2016). Automated mapping of soybean and corn using phenology. ISPRS Journal of Photogrammetry and Remote Sensing, 119, 151– 164. https://doi.org/10/gd8dqd Zhou, Y., Buesching, C. D., Newman, C., Kaneko, Y., Xie, Z., & Macdonald, D. W. (2013). Balancing the benefits of ecotourism and development: The effects of visitor trail-use on mammals in a Protected Area in rapidly developing China. Biological Conservation, 165, 18–24. https://doi.org/10/f48tzm Zipkin, E. F., Andrew Royle, J., Dawson, D. K., & Bates, S. (2010). Multi-species occurrence models to evaluate the effects of conservation and management actions. Biological Conservation, 143(2), 479–484. https://doi.org/10/dgw3c3 190