Using Multiple Scale Space-Time Patterns in Variance-Based Global Sensitivity Analysis for Spatially Explicit...

Area of Science:

• Computational epidemiology
• Ecological modeling
• Complex systems analysis

Background:

• Sensitivity analysis (SA) is vital for agent-based models (ABMs), especially spatially explicit ones, to address challenges in model specification and parameterization.
• Evaluating spatially explicit ABMs often involves comparing spatial or spatio-temporal patterns, but understanding parameter influence on observed mismatches requires further investigation.
• Existing methods pay less attention to quantifying the impact of specific parameter values on the discrepancy between ABM outcomes and real-world observations.

Purpose of the Study:

• To propose and demonstrate a method for assessing parameter influence in spatially explicit agent-based models (ABMs) using multiple scale space-time patterns within a variance-based global sensitivity analysis (GSA) framework.
• To identify which input factors significantly affect the model's ability to replicate observed spatio-temporal disease patterns.
• To highlight the importance of parameterization for achieving accurate model outputs across different space-time scales.

Main Methods:

• Application of variance-based global sensitivity analysis (GSA) to a spatially explicit agent-based model (ABM) simulating vector-borne disease transmission.
• Utilizing multiple scale space-time patterns derived from model outputs for the sensitivity analysis.
• Input factors analyzed included environmental (introduction rates), agent-environment interaction (herd immunity, mosquito density), and agent state transition (mosquito extrinsic incubation period) parameters.

Main Results:

• Parameters related to agent-environment interactions (herd immunity, mosquito population density) were found to have a substantial impact on the model's ability to reproduce observed patterns.
• The magnitude of parameter influence varied significantly across different space-time scales.
• Time-dependent sensitivity to parameter values was observed, emphasizing the dynamic nature of parameter importance in spatially explicit ABMs.

Conclusions:

• The proposed GSA approach using multiple scale space-time patterns effectively identifies critical input factors for parameterization in spatially explicit ABMs.
• Parameters governing agent-environment interactions are particularly influential in reproducing observed disease dynamics.
• Accurate parameterization, guided by sensitivity analysis across various space-time scales, is essential for developing ABMs that reliably mimic real-world patterns.