An ET-Based Two-Phase Method for the Calibration and Application of Distributed Hydrological Models

Abstract

Parameter calibration is a core process for the application of distributed hydrological models, which depends strongly on river runoff records. However, the sparse distribution of hydrological gauging stations has largely limited the application of hydrological modeling. Remote sensing evapotranspiration (ET) data open a new avenue for the calibration of hydrological models. In this study, we used MOD16 ET, a data product from the moderate-resolution imaging spectroradiometer (MODIS), to calibrate a hydrological model in a two-phased approach. A simulated runoff time series was created with the Soil and Water Assessment Tool (SWAT) model that was automatically calibrated for ET in the first phase. The simulated runoff time series was then used in the second phase as the “observed” variable for a semiautomatic calibration in which runoff composition was further refined based on a comprehensive understanding of the local hydrological cycle. The results indicate that the output of simulated ET to MOD16 ET was acceptable and produced a reasonable runoff time series compared to the actual observed runoff, yet the composition of runoff was unsatisfactory for the different runoff components (e.g., surface runoff and base flow) after only automatic calibration. The runoff components were improved by semimanual calibration of the runoff sensitive parameters and resulted in a well calibrated distributed hydrological model. The increasing availability and improving quality of global ET datasets provide an alternative way to calibrate hydrological models, especially in ungauged meso- and large-scale basins. Even though the runoff gauging series is not a prerequisite for the method, some in situ observations are still valuable for the optimization of water cycle components.