Human-Wildlife Conflict Modeling and Mitigation: Integrating Agent-Based Simulations, Green Security Games, and Climate Change Analysis

Abstract

Human-Wildlife Conflict (HWC) represents one of the most pressing challenges in biodiversity conservation, particularly in a changing climate, and requires innovative strategies for effective management. This thesis develops and applies Agent-Based Models (ABMs) to understand complex socio-ecological systems experiencing HWC, focusing on two key species: (I) Asian Elephants (Elephas maximus) in the Periyar-Agasthyamalai Complex (PAC) of the Western Ghats and (II) Saltwater Crocodiles (Crocodylus porosus) in the Andaman and Nicobar Islands (ANI). ABMs provide a flexible framework that captures individual differences, social structures, and decision-making processes, making them particularly well-suited for modeling the heterogeneity inherent in these conflict scenarios. Given the distinct ecological contexts and conflict dynamics of these two species, the thesis is structured into two parts, each addressing unique research objectives tailored to the specific challenges of the Human-Elephant Conflict and the Human-Crocodile Conflict problems. Furthermore, ABMs are integrated with a novel green security game formulation to learn effective mitigation strategies for the Human- Elephant Conflict problem. The impact of climate change on the management strategies of the crocodile population is analyzed to recommend science-based data-driven policy for conservation. PART-I Habitat Suitability Analysis for Asian Elephants in India: The first part of the thesis focuses on Human-Elephant Conflict (HEC). We first developed a Random Forest model to estimate the species distribution of Asian elephants in India and examine inter-annual and intraannual spatiotemporal variability in suitable habitats. Using climatic variables, topographic vii viii conditions, and satellite-derived metrics (land use/land cover, net primary productivity, leaf area index, and normalized difference vegetation index) as predictors, alongside species sighting data from the Global Biodiversity Information Reserve, we found that seasonal reductions in suitable habitat may explain observed elephant migration patterns. Alarmingly, the total available suitable habitat area has declined, further exacerbating HEC. ABM of Elephant Crop Raid Dynamics: Building on this foundation, we developed a spatially explicit ABM for Seethathode, Kerala, India, where HEC is a recurring concern. Unlike existing models that only consider food scarcity as a conflict driver, our ABM incorporates a broader range of factors that influence elephant movement: crop habituation, risk-taking behavior, seasonality, and thermoregulation requirements. The prototype ABM was developed to simulate interactions between humans and solitary bull elephants, addressing two main challenges: the complex behavior of elephants and insufficient movement data from the region. Using data from the extensive literature survey, expert insights, and field surveys, we created a behavior model that incorporates crop habituation, thermoregulation, and aggression. To develop the movement model, we designed a four-step calibration method to adapt relocation data from radio-tagged elephants in Indonesia to the model domain. The ABM’s structure, including assumptions, submodels, and data usage, is detailed following the Overview, Design concepts, Details (ODD) protocol. Emergent Conflict Dynamics and Crop Raid Patterns: The ABM simulates various food availability scenarios to study elephant behavior and environmental impacts on space use and conflict patterns, successfully reproducing observed movement patterns and revealing the emergence of HEC hotspots within the study area. The simulation results indicate that wet months increase conflict and that thermoregulation significantly influences elephant movements and crop raiding patterns, with starvation and crop habituation intensifying these patterns. This prototype ABM represents an initial model for developing a decision-support system in wildlife management that will be further enhanced with layers of complexity in various dimensions. ix Water Availability and Effect on Conflict Dynamics: We then investigated the spatial dynamics of HEC under different water availability scenarios. The main objective was to examine how artificial water sources (water holes) and natural water sources (rivers and streams) affect the spatial distribution of elephants and crop-raiding incidents within the study area. Numerical experimental results emphasize the role of water availability in the evolution of the elephant trajectories. Our findings suggest that crop raiding is not only a foraging behavior but also occurs opportunistically as elephants move into human settlements to access water sources. We also studied the spatial scales at which the ABM generates biologically plausible trajectories by investigating the elephants’ frequency of visits to water sources. This study offers a valuable framework for understanding the dynamics of HEC and implies that effective conservation and conflict mitigation strategies could depend on strategic water management. Adaptive Security Game for HEC Mitigation: Finally, we framed HEC as a challenging variant of a green security game where elephants strategically target crops and water sources, while defenders must allocate scarce patrol resources to protect forest-agricultural boundaries. Unlike typical security game settings, HEC involves adaptive opponents with uncertain behavior and significant observability limitations. We adapted the Follow-the-Perturbed Leader with Uniform Exploration (FPL-UE) algorithm for HEC mitigation, making three key contributions to online green security games: (1) reformulating the defender’s problem for adversarial settings with partial observability, where elephant strategies remain largely unknown; (2) developing a dynamic mechanism for learning and updating rewards and penalties associated with covered and uncovered boundary patches in real-time as elephants adapt to guard deployments; and (3) validating our approach using the calibrated ABM, demonstrating convergence properties against multiple adversarial models. This work presents a first-of-its-kind game-theoretic solution verified against realistic opponent adaptation in the human-wildlife conflict domain, extending security game theory to ecological adversaries with emergent learning behaviors and opening new research directions in adaptive resource allocation under model uncertainty. x PART-II ABM of Crocodile Population Dynamics: The second part of the thesis focuses on Human- Crocodile Conflict (HCC). HCC represents a significant wildlife management issue in the ANI, where saltwater crocodiles are responsible for more human deaths and injuries than any other crocodilian species. The spatial overlap between human populations and optimal crocodile habitats creates a concentrated conflict zone requiring urgent management strategies. The ANI provides an ideal habitat for saltwater crocodiles due to its unique geographical and ecological characteristics. The growing ANI crocodile population may potentially worsen HCC in the coming years. We developed a spatially explicit prototype territory dynamics ABM that integrates territoriality, dominance hierarchies, site fidelity, and temperature-dependent sex determination to project the population of C. porosus from 1975 to 2100. The model evaluates demographic trajectories under varying habitat availability, nesting site constraints, and climate scenarios (SSP1-2.6, SSP5-8.5). The ABM represents demographic processes without extensive parameterization requirements typical of population matrix models, which is significant given that crocodile population census surveys are massive undertakings with inherent difficulties in accurately quantifying size classes in wild populations. The ABM also incorporates temperature-dependent sex determination, where ambient and nest temperatures influence hatchling sex ratios, thereby skewing populations toward females. We investigated how these factors influence population dynamics and their potential contribution to future HCC. The ABM’s structure, including assumptions, submodels, and data usage, is detailed following the ODD protocol. Emergent Demographic Trends and Recruitment Constraints: Simulations reveal that nesting site availability, not aquatic habitat, functions as the primary recruitment bottleneck. Populations with limited nesting sites (<100) stabilize significantly below environmental carrying capacity, creating false demographic stability that masks latent growth potential. Under extreme warming (SSP5-8.5), temperature-dependent sex determination progressively shifts sex ratios toward males by 2100, simultaneously exacerbating conflict risk while depleting female reproductive reserves. This study marks a significant methodological leap for crocodilian xi ecology, representing one of the first population models specifically tailored to the unique demographic and environmental constraints of small localized populations such as those found in the South Andaman Islands. Critical Harvest Thresholds and Policy Recommendations: The final component of the thesis focuses on policy recommendations for managing HCC using our ABM. Historically, successful HCC mitigation strategies have controlled population size through adult culling or egg harvesting. Other methods, such as removing and relocating problematic adults, have proven ineffective as removed individuals either return upon re-release or are replaced by other dominant males. Our ABM effectively captures these aspects of territoriality and space use. Harvest simulations identify critical sustainability thresholds: annual removal rates exceeding 15% for adults or 5% for the total population (excluding hatchlings) trigger collapse once subadult reserves deplete. Notably, adult removal produces counterintuitive recruitment pulses as subadults rapidly occupy vacated territories, temporarily masking long-term population decline. Comparative analysis with matrix models demonstrates that the ABM’s compensatory dynamics emerge mechanistically from observable spatial variables (territory distributions, nesting sites) rather than abstract density-dependence parameters lacking empirical calibration. This framework advances ecological informatics by transforming static demographic models into spatially explicit decision-support tools, providing the computational foundation for adaptive management strategies balancing species conservation with community safety in human-wildlife conflict zones.