Localized Drought Early Warning using In-situ Groundwater Sensors

As drought risk continues to increase globally, there is an urgent demand for improved Early Warning Systems (EWSs) that enable earlier, risk mitigating actions. This is evident in Somalia, where estimated fatalities from the 2011 drought exceed 250,000, and a further drought in 2016/17 required humanitarian assistance for 6 million people. However, current early actions are delayed by the short lead time of satellite-based drought forecasts, and the coarse spatial resolution of satellite-data leaves uncertainties over where to prioritize interventions. The increasing affordability and feasibility of deploying in-situ, automatic sensors are an opportunity to provide much needed data on drought exposure. At Imperial College London, we are assessing the potential of high-frequency groundwater monitoring of abstraction wells to improve drought indexing in Maroodi Jeex, Somalia. A modified version of the AquiMod groundwater model has been developed and calibrated using 15-minute interval groundwater level observations collected from three wells. We find that high-frequency well data enables calibration with a relatively short (< 1-year) time-series of observed groundwater levels, addressing the technical challenge of collecting long time-series of groundwater data in Somalia. We estimate time-variable abstraction yield at the three sites from the modelled water levels, using a yield index calculated from abstraction drawdown behaviors in the observed monitoring data. Our findings show that a combination of meteorological forecasts and in-situ data can generate real-time simulations of abstraction yields and local water availability during a drought event. Such technologies have considerable potential to improve drought management, by identifying and forecasting areas of greatest drought exposure. This exposure information has potential applications informing drought EWSs, as a clear mapping of spatial drought intensity can guide subsequent risk-mitigating actions.