Global Sensitivity Analysis for Improved Understanding of Mechanisms Underlying Spatial Simulation Models of Coupled Human and Natural Systems

Keywords: model uncertainty, sensitivity analysis, agent-based modeling, model drivers
Abstract Type: Paper Abstract

Abstract

The paper demonstrates the practical implications of spatial sensitivity indices, obtained through output variance decomposition, for understanding the inner workings of spatial models. We utilize an agent-based model that simulates the process of farmers’ enrollment in a land conservation reserve program. This model generates land use change maps and is instrumental in understanding the enrollment process. The results are evaluated using the total-effect sensitivity index, which captures individual and higher-order contributions to model output variance. A sensitivity map is created for each factor (both biophysical and social), and these maps are then combined through an overly. This process is crucial in identifying the most influential factors affecting the enrollment decision at a specific location and demonstrating the spatial heterogeneity of factor influence. The presented methodology provides a robust approach to elucidating the interplay of factors driving variations in model outputs, particularly in spatially explicit coupled human and natural systems models, where heterogeneity is a crucial characteristic. The framework allows for a deeper understanding of the model mechanisms and, by extension, the dynamic processes in the modeled system.