Keywords
Sensitivity Analysis, Spatial Modelling, Agent-Based Modelling, Land Use Change, Lakes
Location
Session C1: Compexity, Sensitivity, and Uncertainty Issues in Integrated Environmental Models
Start Date
16-6-2014 9:00 AM
End Date
16-6-2014 10:20 AM
Abstract
Models of coupled landscape and aquatic systems (CLAS) are prone to input uncertainties that vary over space. To address this challenge, we employ a comprehensive model evaluation that: [1] quantifies the variability of model results (uncertainty analysis), and [2] decomposes this variability based on the relative contribution of inputs to identify major drivers in the model (sensitivity analysis). Our study simulates how agricultural land conversion from active to fallow lands reduces nutrient loading to lakes. We employ an agent-based model of farmer decision making coupled with a spatially-explicit biophysical lake model. A number of model inputs are uncertain including: variables reflecting farmer decision making, maps that represent the environmental benefits of land conservation, and variables that drive nutrient concentrations in CLAS. To be useful for policy analysis, the model requires simplification. To this end, we employ variance-based sensitivity analysis. We run the model multiple times to generate a distribution of lake total phosphorus concentration (TP) and evaluate the variability of TP using two spatial scales. First, the sensitivity analysis is run at a regional scale, at which the TP values from all lakes are lumped into a scalar calculated for the entire study area (aggregate analysis). Second, the sensitivity analysis is run at a lake scale focusing on TP values for individual lakes (fine-scale analysis). The aggregate analysis identifies the most critical components affecting the overall uncertainty of regional TP. The fine-scale analysis identifies the most crucial components affecting uncertainty of TP in individual lakes. A comparison of results from both scales provides useful insights for model simplification.
Included in
Civil Engineering Commons, Data Storage Systems Commons, Environmental Engineering Commons, Hydraulic Engineering Commons, Other Civil and Environmental Engineering Commons
Multiscale Spatial Sensitivity Analysis for Agent-Based Modelling of Coupled Landscape and Aquatic Systems
Session C1: Compexity, Sensitivity, and Uncertainty Issues in Integrated Environmental Models
Models of coupled landscape and aquatic systems (CLAS) are prone to input uncertainties that vary over space. To address this challenge, we employ a comprehensive model evaluation that: [1] quantifies the variability of model results (uncertainty analysis), and [2] decomposes this variability based on the relative contribution of inputs to identify major drivers in the model (sensitivity analysis). Our study simulates how agricultural land conversion from active to fallow lands reduces nutrient loading to lakes. We employ an agent-based model of farmer decision making coupled with a spatially-explicit biophysical lake model. A number of model inputs are uncertain including: variables reflecting farmer decision making, maps that represent the environmental benefits of land conservation, and variables that drive nutrient concentrations in CLAS. To be useful for policy analysis, the model requires simplification. To this end, we employ variance-based sensitivity analysis. We run the model multiple times to generate a distribution of lake total phosphorus concentration (TP) and evaluate the variability of TP using two spatial scales. First, the sensitivity analysis is run at a regional scale, at which the TP values from all lakes are lumped into a scalar calculated for the entire study area (aggregate analysis). Second, the sensitivity analysis is run at a lake scale focusing on TP values for individual lakes (fine-scale analysis). The aggregate analysis identifies the most critical components affecting the overall uncertainty of regional TP. The fine-scale analysis identifies the most crucial components affecting uncertainty of TP in individual lakes. A comparison of results from both scales provides useful insights for model simplification.
