Seasonal forecasting of groundwater levels in Europe using Pastas time series models and ECMWF SEAS5 forecasts

Groundwater level monitoring is an important way for water resource managers to obtain information on the state of the groundwater system and make informed decisions. In many countries around Europe the right to abstract groundwater (e.g., for drinking water or irrigation purposes) is bound to observed groundwater levels. In particular during and after periods of drought such rights to abstract groundwater may be temporarily denied. As climate change is expected to increase the frequency and intensity of hydrological extremes, severe drought events become more likely, potentially increasing the gap between groundwater demand and supply. An early warning system of a potential groundwater drought could help water managers make informed decisions in advance, to try and counteract the effects of drought. In this study we investigate the use of seasonal forecasts from the ECMWF SEAS5 system to forecast groundwater levels around Europe. The groundwater levels are simulated using a non-linear time series model using impulse response functions as implemented in Pastas (https://github.com/pastas/pastas). Forecasts are compared to groundwater level simulations based on historic meteorological data from the E-OBS database. The methods are tested on 10 long-term (30 years) groundwater level time series. The use of the Standardized Groundwater Index (SGI) is tested to assess the forecast quality and communicate results with decision makers. Bias-correction of the SEAS5 forecasts is found to be necessary to forecast groundwater levels at this local scale. Preliminary results show that the forecast quality depends on the memory effect of the groundwater system, which can be characterized by the auto-correlation of the time series. In addition, it is found that the groundwater levels forecasts have smaller ranges in spring then in the winter months. This may be explained by the fact that groundwater levels in spring are more dependent on evaporation than on precipitation and that forecast of the first are better than those of the latter. The results from this study may be used to improve early warning systems that forecast groundwater droughts.