Hydrological consistency in modeling - Improving model structure and parameter estimation with temporal diagnostic analyses

点击翻译按钮获取中文摘要

Catchment models reproduce the hydrological processes and their temporal dynamic in a catchment. Seeing models as learning tools in contrast to application studies, these models are used for an improved understanding of processes and their representation in models. In a consistent model concept, the hydrological processes are appropriately represented in the model structure. Each model parameter is used for the related process and its dominant periods coincide with the temporal process patterns. Temporal diagnostic analyses investigate the relation of model structure and real-world processes in a high temporal resolution. In this project, temporal variations in dominant processes are detected by a temporal analysis of parameter sensitivity for each model time step. The temporal analysis of parameter sensitivities is related to total discharge and to the corresponding process to identify temporal process patterns. In combination with a temporal analysis of model performance, dominant parameters in phases of poor model performance are detected. Based on this, failures in model components are derived. Temporal diagnostic information is also included in the model calibration. For the estimation of the parameter values, only the periods of high sensitivity of this parameter are used.This project aims to develop new methods to achieve hydrologic consistency in model structure and calibration based on the temporal patterns of dominant processes. The main research objectives aim to achieve hydrologic consistency1) in the temporal process patterns by analyzing the temporal variations in dominant model parameters in relation to the corresponding hydrological process2) in the model structure by detecting structural model failures with a joined temporal analysis of parameter sensitivity and model performance3) in the estimation of model parameters by developing a new innovative calibration method based on periods of high parameter sensitivity.