Author(s): Jesus Casado-Rodriguez; Manuel Del Jesus; Salvador Navas

Linked Author(s): MANUEL DEL JESUS, Salvador Navas

Keywords: Calibration; Remote sensing; Hydrological model

Abstract: Parameter identifiability is a known issue in conceptual hydrological models, which originates from the amount of model parameters and the lack of variety in the target variables used during calibration and causes uncertainty about the representativeness of the optimised model. In this context, remote sensing emerges as a key source of information that allows us to calibrate spatial processes in the water cycle. In this study, we combine the use of remotely sensed data in the calibration of spatial surface processes, such as snow and evapotranspiration, with a novel procedure that calibrates discharge by applying flow disaggregation techniques. The objective is to improve the identifiability of the calibrated model parameters. State of the art hydrological models often require a vast number of parameters to reproduce the processes in the water cycle. This is a result of modelling an increasing number of processes, which requires a larger model flexibility. Calibrating such models becomes a complicated task, both in terms of computational effort and parameter identifiability. The concept of parameter identifiability (or equifinality) states that multiple parameterizations of a model, representing diverse behaviors of the catchment, are indistinguishable in terms of model performance. This fact limits the applicability of the model to posterior studies such as climate change impacts. A way of reducing equifinality is to calibrate several target variables, instead of simply discharge. Following this thread, remote sensing becomes an invaluable source of information for calibrating processes in the water cycle for which traditional information sources are scarce, such as snow or vegetation. Examples of the use of remotely sensed data for calibrating hydrological models are abundant in the literature . Since remotely sensed data can only supply information about surface processes, there remains equifinality in the identification of subsurface processes, namely the distribution of discharge in surface runoff, interflow and groundwater flow. To tackle this uncertainty without the need for further information, we developed a sequential calibration method based on hydrograph separation techniques. We apply this methodology to the Deva river basin, a mountainous catchment in the Picos de Europa National Park (Northern Spain), whose water cycle we simulate using the hydrological model TETIS, a conceptual, distributed model widely used in Spain. The proposed calibration procedure uses first remotely sensed products from MODIS to fit the snow and vegetations processes, and then fits the parameters controlling discharge in a sequence from those affecting quick, flow and total discharge. The objective of this study is to combine the information of remotely sensed data and that obtained from hydrograph disaggregation in a sequential calibration process in order to improve the identifiability of a hydrological model.

DOI: https://doi.org/10.3850/IAHR-39WC252171192022174

Year: 2022