Spatially distributed model calibration of a highly managed hydrological system using remote sensing-derived ET data

Kurzfassung

Calibrating a spatially distributed hydrological model of a region with complex and highly diverse agricultural settings is a difficult task. It becomes even more challenging, if major parts of the hydrology are human controlled, e.g. through irrigation agriculture. In this study, a Soil and Water Assessment Tool (SWAT) was set up and calibrated for an entirely managed meso-scale irrigation scheme, where irrigation channels are the only water-courses and no natural streams exist. An automated and spatially distributed model calibration was applied, implementing the global Dynamically Dimensioned Search Algorithm (DDS), and using Kling-Gupta Efficiency (KGE) as objective function. To account for the spatio-temporal heterogeneity of water balance fluxes in the calibration process, remote sensing-derived evapotranspiration (ET) estimates with a spatial resolution of 1 km × 1 km, were used as observational data. Results showed that the automated and spatially distributed model calibration was able to increase the model ET simulation performance from 0.04 to 0.27 KGE. However, a strong land-use dependent goodness-of-fit, ranging from approx. −0.60 KGE to 0.60 KGE for individual land-use types, revealed that the information, extracted from remote sensing-derived ET estimates used in this study, is lacking the necessary spatial heterogeneity. An unsatisfactory spatial comparison between simulated ET and ET derived from remote sensing estimates showed to inhibit better calibration results for the study area, which is characterized by small-scale and diverse cropping patterns. Using a modified remote sensing-derived ET product, which accounts for land-use class dependent ET characteristics, improved the overall calibration results to 0.40 KGE. A cross-validation could indicate a successful application of this automated and spatially distributed model calibration approach, using the modified ET product.