Seasonal forecasting of groundwater levels in principal aquifers of the United Kingdom

• Published in: Journal of hydrology (Amsterdam) 2015-11, Vol.530, p.815-828
• Main Authors: Mackay, J.D., Jackson, C.R., Brookshaw, A., Scaife, A.A., Cook, J., Ward, R.S.
• Format: Article
• Language: English
• Subjects: AquiMod; Climatology; Ensemble forecasting; GloSea5; Groundwater; Groundwater level forecasting; Groundwater levels; Hydrology; Mathematical models; Rainfall; Seasonal forecasting; Skills; Weather forecasting
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2015.10.018

•We forecast groundwater levels 3months into the future for 21 boreholes in the UK.•We use GloSea5 seasonal rainfall forecasts to drive a conceptual groundwater model.•The forecasts consistently show more skill than a persistence forecasting approach.•The forecasts are not able to capture extreme groundwater level events.•Sensitivity to (skill derived from) rainfall forecasts is highly site specific. To date, the majority of hydrological forecasting studies have focussed on using medium-range (3–15days) weather forecasts to drive hydrological models and make predictions of future river flows. With recent developments in seasonal (1–3months) weather forecast skill, such as those from the latest version of the UK Met Office global seasonal forecast system (GloSea5), there is now an opportunity to use similar methodologies to forecast groundwater levels in more slowly responding aquifers on seasonal timescales. This study uses seasonal rainfall forecasts and a lumped groundwater model to simulate groundwater levels at 21 locations in the United Kingdom up to three months into the future. The results indicate that the forecasts have skill; outperforming a persistence forecast and demonstrating reliability, resolution and discrimination. However, there is currently little to gain from using seasonal rainfall forecasts over using site climatology for this type of application. Furthermore, the forecasts are not able to capture extreme groundwater levels, primarily because of inadequacies in the driving rainfall forecasts. The findings also show that the origin of forecast skill, be it from the meteorological input, groundwater model or initial condition, is site specific and related to the groundwater response characteristics to rainfall and antecedent hydro-meteorological conditions.