Spatio-temporal modelling of the status of groundwater droughts

An empirical (geo)statistical modelling scheme is developed to address the challenges of modelling the severity and distribution of groundwater droughts given their spatially and temporally heterogeneous nature and given typically highly irregular groundwater level observations in space and time. The scheme is tested using GWL measurements from 948 observation boreholes across the Chalk aquifer (UK) to estimate monthly groundwater drought status from 1960 to 2013. For each borehole, monthly GWLs are simulated using empirical mixed models where the fixed effects are based on applying an impulse response function to the local monthly precipitation. Modelled GWLs are standardised using the Standardised Groundwater Index (SGI) and the monthly SGI values interpolated across the aquifer to produce spatially distributed monthly maps of SGI drought status for 54 years for the Chalk. The mixed models include fewer parameters than comparable lumped parameter groundwater models while explaining a similar proportion (more than 65%) of GWL variation. In addition, the empirical modelling approach enables confidence bounds on the predicted GWLs and SGI values to be estimated without the need for prior information about catchment or aquifer parameters. The results of the modelling scheme are illustrated for three major episodes of multi-annual drought (1975-1976; 1988-1992; 2011-2012). They agree with previous documented analyses of the groundwater droughts while providing for the first time a systematic, spatially coherent characterisation of the events. The scheme is amenable to use in near real time to provide short term forecasts of groundwater drought status if suitable forecasts of the driving meteorology are available.