Sensitivity and identifiability analyses of parameters for water quality modeling of subtropical reservoirs

Abstract

Sensitivity and identifiability analyses are a standard practice in modeling applications to investigate the relative importance of model components that control the system's behavior. In this study, both analyses were implemented to identify the most influential parameters in a coupled hydrodynamic-biogeochemical model applied for three subtropical reservoirs. The one-dimensional General Lake Model coupled to Aquatic EcoDynamics (GLM-AED) was used to simulate the dynamics of dissolved oxygen, total phosphorus, nitrate, ammonium, and chlorophyll-a. Results reveal consistent sensitivity patterns between reservoirs, especially for a subset of temperature multipliers affecting dissolved oxygen and nutrients. In contrast, the sensitivity of chlorophyll-a is highly site-specific. Additionally, the majority of parameters are medium or high sensitive, which indicates the need for a calibration procedure to improve model accuracy. The analyses provide a detailed understanding of the governing ecosystem dynamics as a step forward to model identifiability and guidance for future model calibration.