Determining the potential distribution of highly invasive plants in the Carpathian Mountains of Ukraine: a species distribution modeling approach under different climate-land- use scenarios and possible implications for natural-resource management Maria Simpson Independent Graduate Project Report – December 2011 Johns Hopkins University – Zanvyl Krieger School of Arts and Sciences Advanced Academic Programs – Environmental Sciences and Policy Supervisor: Dr. Bohdan Prots (WWF Danube-Carpathian Programme, Ukraine and State Museum of Natural History, National Academy of Sciences, Lviv, Ukraine) M. Simpson – Contents TABLE OF CONTENTS ACKNOWLEDGEMENTS i LIST OF FIGURES ii-iv LIST OF TABLES v ABSTRACT iv 1. INTRODUCTION 1-3 2. THE UKRAINIAN CARPATHIANS 3 2.1 Geomorphology 3-5 2.2 Climate and hydrology 5-8 2.3 Vegetation 8-9 2.4 Land use and economic development 9-12 2.5 Importance of the UA Carpathians within the Carpathian Ecoregion 12 3. HIGHLY INVASIVE PLANTS IN THE UKRAINIAN CARPATHIANS 12 3.1 Invasiveness of non-native invasive plant species (NIPS) 12 3.1.1 Assessment of the invasiveness of NIPS in the UA Carpathians: 12-13 3.1.2 Ecology of NIPS: 13-17 3.1.3 Impacts of NIPS on ecosystems and humans: 17-19 3.2 Introduction pathways into and within the UA Carpathians 19 3.2.1 Human development and land use and NIPS: 19-22 3.2.2 Climate change and spread of NIPS: 23-25 4. SPECIES DISTRIBUTION MODELING 25 4.1 Ecological concepts 25 4.1.1 The ecological niche: 25-26 4.1.2 Equilibrium vs. non-equilibrium: 26-27 4.2 Modeling approaches 28-30 5. METHODS 30 5.1 Data requirements 30 5.1.1 Study area: 30-31 M. Simpson – Contents 5.1.2 Species presence data: 31-32 5.1.3 Predictor variables: 32-35 5.2 The modeling algorithms 35 5.2.1 Maximum Entropy – Maxent: 35-37 5.2.2 Biodiversity Modeling - BIOMOD: 37-39 5.3 Model evaluation 40-41 5.4 Variable importance 41 5.5 Climatic and land-use projections 41-42 5.6 Visual comparison of results 43 6. RESULTS 43 6.1 Accuracy of model predictions 43-44 6.2 Suitability predictions for current conditions 44-47 6.3 Importance of predictor variables 47-51 6.4 Potential species distribution under future projections 51-59 6.5 Maxent vs. BIOMOD 59-63 7. GENERAL IMPLICATIONS OF STUDY FOR NATURAL-RESOURCE MANAGEMENT 63-74 8. CONCLUSIONS 74-77 9. REFERENCES 78-85 APPENDICES Appendix A: Glossary – Definitions of Scientific Terms A1-A3 Appendix B: Ecology of the 11 Most Aggressive Non-Native Invasive Plant Species in the UA Carpathians A4-A18 Appendix C: Overview of BIOMOD Functions A19-A23 Appendix D: BIOMOD Best Model Based on AUC Values A24-A29 Appendix E: Predictions for Current Condition A30-A38 Appendix F: Measures of Variable Importance in Maxent and BIOMOD A39-A55 Appendix G: Projections for 2050 and 2100 Based on Change Scenarios A56-A91 M. Simpson – Acknowledgements ACKNOWLEDGEMENTS The project which resulted in this report serves as my final graduate project for the Master of Science degree in Environmental Sciences and Policy at the Johns Hopkins University. It was carried out under supervision of the World Wildlife Fund (WWF) Danube-Carpathian Programme. Therefore, I thank, first and foremost, Dr. Bohdan Prots, Programme Coordinator of the WWF in Lviv and my direct project supervisor, for coordinating this unique and important study. Furthermore, my dearest thanks go out to the entire team at WWF Danube-Carpathian Programme and the State Museum of Natural History in Lviv for their hospitality, expertise, and inspiration they provided through their relentless determination to protect the natural heritage of the Carpathian Mountains. I also thank Dr. Ivan Kruhlov, Dr. Tobias Kuemmerle, and Dr. Patrick Hostert for allowing me to use their grid and vector maps on the Ukrainian Carpathians. Furthermore, I am grateful for the advice I received from Dr. Kruhlov and Dr. Yuriy Andreychuk on how to properly manipulate these datasets to get the GIS layers required for this study and for the expert opinion on species distribution modeling I received from Dr. Tobias Kuemmerle. Lastly, my gratitude goes to my dearest friend Isabel Marie Mitchell who patiently and competently reviewed my report drafts. i M. Simpson – List of Figures LIST OF FIGURES Main Text Figure 1: Physical map of the study area (Ukrainian Carpathians) within the Carpathian Mts. in Eastern Europe 4 Figure 2: Overview of A – geomorphological types and B – terrain/elevation within the Ukrainian Carpathians 4 Figure 3: Overview of A - main climatic zones and B – river networks of the Ukrainian Carpathians 7 Figure 4: Settlement locations and sizes in km2 in Western Ukraine 10 Figure 5: Urban-industrial development, tourism, and major transport routes in the Carpathians 11 Figure 6: Simplified depiction of the Enemy Release Hypothesis 14 Figure 7: Heracleum sosnowskyi invasion in species-rich meadow in the Ukrainian Carpathians 18 Figure 8: Cumulative impact of plant invasion 19 Figure 9: "Impacts of global change on invasion and associated processes" 25 Figure 10: Species Distribution Model (SDM) building process 29 Figure 11: Illustration of density distributions derived from two covariates (i.e., predictor variables) at presence locations and background samples 36 Figure 12: A schematic overview of the modeling processes occurring in BIOMOD 38 Figure 13: Confusion matrix 40 Figure 14: Measures of accuracy derived from the confusion matrix 40 Figure 15: Relative suitability for establishment of Acer negundo in the Ukrainian Carpathians determined by Maxent and BIOMOD best model (GMB) 44 Figure 16: Proportion of total area within the Ukrainian Carpathians predicted as suitable for establishment of invasive species with Maxent and BIOMOD best models 45 Figure 17: Overlap of Maxent and BIOMOD binary predictions of invasive plant distributions in the Ukrainian Carpathians 46 Figure 18: Spatial correlation between density of roads (polylines) and settlements (points) in the Ukrainian Carpathians and the areas predicted by Maxent to be highly suitable for all modeled invasive plant species 47 Figure 19: Maxent - contribution of predictor variables per invasive species for distribution models within the Ukrainian Carpathians 48 ii M. Simpson – List of Figures Figure 20: BIOMOD - contribution of predictor variables per invasive species for distribution models within the Ukrainian Carpathians 48 Figure 21: Maxent - contribution of predictor variables to model gain per invasive species for distribution models within the Ukrainian Carpathians 49 Figure 22: Boxplot of range change of suitable habitats for invasive plant species in the Ukrainian Carpathians across future change projections 52 Figure 23: Boxplot of range change of suitable habitats for invasive plant species in the Ukrainian Carpathians across study species 53 Figure 24: Spatial pattern of model predictions of suitable habitats in the Ukrainian Carpathians for Acer negundo and Echinocystis lobata based on current conditions and projections (to scenario iv) in BIOMOD 54 Figure 25: Maxent projections for scenarios i and ii (2050) of potential distribution of Robinia pseudoacacia in the Ukrainian Carpathians 55 Figure 26: Maxent projections for scenarios iii and iv (2100) of potential distribution of Robinia pseudoacacia in the Ukrainian Carpathians 55 Figure 27: Maxent projections for scenarios i and ii (2050) of potential distribution of Impatiens glandulifera in the Ukrainian Carpathians 56 Figure 28: Maxent projections for scenarios iii and iv (2100) of potential distribution of Impatiens glandulifera in the Ukrainian Carpathians 56 Figure 29: Overlap of binary predictions (at least 7 invasive study species) of suitable habitats for invasion within the Ukrainian Carpathians as projected by Maxent and BIOMOD under future climate and land-use change scenarios 58 Figure 30: Binary Maxent predictions of suitable habitat within the Ukrainian Carpathians for Helianthus tuberosus by 2100 59 Figure 31: Binary BIOMOD predictions of suitable habitat within the Ukrainian Carpathians for Helianthus tuberosus by 2100 60 Figure 32: Probability of presence of Acer negundo in the Ukrainian Carpathians as a function of maxtwarm (in °C x 10) while the remaining five predictor variables are held at their average value 61 Figure 33: BIOMOD-GBM projections of the potential distribution of Heracleum sosnowskyi in the Ukrainian Carpathians by 2100 62 Figure 34: The range of potential impact of non-native invasive plants on biodiversity in the Ukrainian Carpathians under current environmental conditions 64 iii M. Simpson – List of Figures Figure 35: The range of potential impact of non-native invasive plants on biodiversity in the Ukrainian Carpathians by 2100 assuming climate change and high economic development/ insufficient environmental protection 65 Figure 36: Suitable areas for establishment of Heracleum sosnowskyi populations in the Uzhanskyi National Nature Park (NNP) based on logistic modeling results in Maxent 67 Figure 37: Suitable areas for establishment of Heracleum sosnowskyi populations in the Uzhanskyi National Nature Park (NNP) based on binary modeling predictions in BIOMOD 68 Figure 38: Density of settlements and roads and suitable areas for establishment of Heracleum sosnowskyi populations in the Uzhanskyi National Nature Park (NNP) 69 Figure 39: The range of potential impact of highly invasive plants on biodiversity in the UA Carpathians (with a focus on the Transcarpathians) by 2050 assuming climate change and high economic development/insufficient environmental protection 72 Appendices