Uncertainty analysis for landscape models used for coastal planning. (5th August 2021)

• Record Type: Journal Article
• Authors: Meselhe, E.M.; White, E.D.; Wang, Y.; Reed, D.J.
• Is Part Of: Estuarine, coastal and shelf science. Volume 256(2021)
• Publication Date: 2021-08-05
• Subjects: Uncertainty -- Downscaling -- Ecosystem model -- Coastal -- Planning -- Wetlands; Estuarine oceanography -- Periodicals; Coasts -- Periodicals; Estuarine biology -- Periodicals; Seashore biology -- Periodicals; Coasts; Estuarine biology; Estuarine oceanography; Seashore biology; Periodicals; 551.461805
• DOI: 10.1016/j.ecss.2021.107371
• Languages: English
• ISSNs: 0272-7714

Abstract

Abstract: In long-term coastal planning, it is important to consider the effects of uncertainties on predicted outcomes. The approach proposed here provides a framework to perform uncertainty analysis for landscape models used for planning-level efforts. The approach is presented through an ecosystem Integrated Compartment Model (ICM) applied to Coastal Louisiana, USA. The ICM includes components for hydrology, water quality, morphology, vegetation, barrier islands, and habitat suitability indices. The framework quantifies the magnitude of the uncertainty in key model output driven by uncertainties in critical model variables. The approach is based on perturbations applied to model variables that directly influence the model output of interest. The magnitude of the perturbations was guided by the ICM calibration errors. The model variables examined include water level, salinity, wetland types, suspended mineral sediment concentration, and organic loading. The perturbations were initially applied to individual model variables in separate experiments to identify which variable would significantly influence key model outputs. The uncertainty range resulting from linearly adding the uncertainty of the individual perturbations was compared to the outcome of a set of composite experiments designed to examine the interdependency among the uncertainty of the model variables. The comparison showed that the uncertainty range resulting from the composite experiments set was wider than the linearly added uncertainty bracket. This outcome demonstrates that interdependency among model variables is important. Overall, this approach provides valuable insights on the uncertainties associated with predictions made by large scale landscape models for coastal and deltaic environments. Highlights: Uncertainty analyses identify model parameters that affect the key model output. Non-linear uncertainty analyses highlight interdependency among model parameters. Uncertainties may not grow from one module to another within landscape models. Uncertainty analysis needed to evaluate confidence in landscape model performance.