Evaluation of potential evapotranspiration assessment methods for hydrological modelling with SWAT in the Densu river basin in Ghana.

Abstract

Accurate estimation of evapotranspiration (ET) is required for a good estimation of available water for use in any catchment as ET constitutes major means by which water is lost in any catchment. SWAT model uses climate data in estimating potential evapotranspiration (PET). The PET together with other parameters is then used in estimating actual evapotranspiration (AET). SWAT model provides three different methods in estimating PET which are Penman–Monteith (PM), Hargreaves (H) and Priestly–Taylor (PT). These three PET methods were evaluated in a limited available spatial and temporal data Densu river basin to assess their impact on the resulted model water balance. The missing data in the 8 climate stations that were used in this study were estimated using a WXGEN weather generator. The results showed that the accuracy of the water balance from the SWAT model was defined by how well the PET method selected for the SWAT was able to estimate an accurate spatial and temporal distributed PET in the basin. A discharge at Nsawam, a town within the basin, was used for the model calibration and validation. Hargreaves PET method had the best Nash–Sutcliffe model efficiency coefficient (NSE) values of 0.70 for the calibration and 0.74 for validation period, while the PM PET method had the worse NSE value of 0.66 for calibration and 0.62 for validation. These results were attributed to the limited available data within the basin. The discrepancies in the resultant water balance model could be attributed to the three different PET methods used in relation to the climate data that were available for the calibration and validation of the SWAT model.