Calibrating spatially distributed hydrological models is complex due to the lack of reliable data, uncertainty in representing the physical features of a river catchment, and the implementation of hydrological processes in a simulation model. In this paper, an innovative approach is presented which incorporates remote sensing derived evapotranspiration in the calibration of the Soil and Water Assessment Tool (SWAT) in a catchment of the Krishna basin in southern India. The Gauss–Marquardt–Levenberg algorithm is implemented to optimise different combination of land use, soil, groundwater, and meteorological model parameters. In the best performing optimisation, the r2 between monthly sub-basin simulated and measured actual evapotranspiration (ETact) was increased from 0.40 to 0.81. ETact was more sensitive to the groundwater and meteorological parameters than the soil and land use parameters. Traditional calibration on a limited number of discharge stations lumps all hydrological processes together and chances on the equifinality problem are larger. In this study we have shown this problem can be constrained by using spatially distributed observations with a monthly temporal resolution. At a spatial resolution below the sub-basin level further study is required to fine-tune the calibration procedure.