Summary:

Numerical weather forecasts have reached the acuraccy and reliability level that their results can be used as a replacement for the measurements of weather parameters when the availability of the latter is limited. Here we present the application of the COAMPS numerical weather forecast model to prediction of the potential evapotranspiration ET0. ET0 is computed using the data from the model. Also a hybrid model with part of the data coming from the model and part of the data from the measurements was used. Additionally a machine learning methods were used to improve model skill. The results show that application of the simulated data gives very good agreement of the predicted ET0 with that computed using measurement data. Hybrid models are slightly better than the purely simulation-based and machine learning allows for further improvement od the ET0 models.