Evaluation of different precipitation and potential evapotranspiration time series for hydrological modeling in Australian catchments

For hydrological modeling in snow-free catchments, precipitation (P) and potential evapotranspiration (E_pot) are the two key input time series. There are different methods to observe, calculate and interpolate these time series. In the Australian large sample data set for hydrological modeling (CAMELS-AUS, Catchment Attributes and Meteorology for Large-sample Studies)  with data for 222 catchments, two different time series for P and seven different time series for E_pot are provided. Here, we address the open question of which data should be used as input to an hydrological model.

Our basic assumption is that the most suitable combination of P and E_pot is the one that results in the best model performances in terms of runoff simulations. For this we first tested the differences between the different input time series. Secondly, we conducted a thorough comparison of the 14 possible combinations of P and E_pot time series. First analyses show that the differences between the two P time series are relatively minor, whereas the seven E_pot time series differ more substantially from each other, especially in terms of seasonality and magnitude. Despite these differences, preliminary modeling results show that for the majority of the catchments there is no significant difference in model performance between the model calibrations carried out for each of the 14 different P/E_pot combinations, suggesting that the model has a certain capability to compensate for differences in the input data by adapting its (soil) parameters. However, for some of the catchments there is a clear trend between the mean E_pot and the corresponding model performance. Characterizing and further investigating these catchments can help to gain insight in the impact of different input data on the model performance, as well as to provide general recommendations that can help the user of a hydrological model to make an informed choice when it comes to the selection of the input data.