Improving evapotranspiration processes in distrubing hydrological models using Remote Sensing derived ET products.

Abstract:

Evapotranspiration is the main process in hydrology (on average around 60%), though has not received as much attention in the evaluation and calibration of hydrological models. In this study, Remote Sensing (RS) derived Evapotranspiration (ET) is used to improve the spatially distributed processes of ET of SWAT model application in the upper Mara basin (Kenya) and the Blue Nile basin (Ethiopia). The RS derived ET data is obtained from recently compiled global datasets (continuously monthly data at 1 km resolution from MOD16NBI,SSEBop,ALEXI,CMRSET models) and from regionally applied Energy Balance Models (for several cloud free days). The RS-RT data is used in different forms:

Method 1) to evaluate spatially distributed evapotransiration model results

Method 2) to calibrate the evotranspiration processes in hydrological model

Method 3) to bias-correct the evapotranpiration in hydrological model during simulation after changing the SWAT codes

An inter-comparison of the RS-ET products shows that at present there is a significant bias, but at the same time an agreement on the spatial variability of ET. The ensemble mean of different ET products seems the most realistic estimation and was further used in this study.

The results show that:

Method 1) the spatially mapped evapotranspiration of hydrological models shows clear differences when compared to RS derived evapotranspiration (low correlations). Especially evapotranspiration in forested areas is strongly underestimated compared to other land covers.

Method 2) Calibration allows to improve the correlations between the RS and hydrological model results to some extent.

Method 3) Bias-corrections are efficient in producing (sesonal or annual) evapotranspiration maps from hydrological models which are very similar to the patterns obtained from RS data.

Though the bias-correction is very efficient, it is advised to improve the model results by better representing the ET processes by improved plant/crop computations, improved agricultural management practices or by providing improved meteorological data.