Reliability of meteorological drought indices for predicting soil moisture droughts

Abstract. Meteorological drought indices based on precipitation and/or evaporation are commonly used to detect the presence, severity and duration of soil moisture droughts. However, it is debatable whether droughts can be adequately characterised using only precipitation and/evaporation, or whether more physical based methods using soil water deficits and pressures is necessary. To address this question, the performances of two commonly used meteorological drought indices, the Standard Precipitation Index (SPI) and the Reconnaissance Drought Index (RDI), are evaluated against soil moisture droughts identified using a physically based soil water model. Our analysis is based on three sites in Eastern Australia, each representing specific soil-climate conditions. Drought duration and severity were estimated using SPI and RDI and soil water pressure data were simulated with Hydrus-1D. The performance of the two drought indices was measured in terms of their correlation with simulated monthly minimum soil water pressures, and their ability to estimate the frequency with which the simulated pressure drops below threshold values. There was a significant correlation between the two drought indices (SPI and RDI) and the monthly minimum soil water pressure. Failure rate (FR) and false alarm rate (FAR) of drought indices detect soil moisture drought reasonably well (FR and FAR is <50 %) for both drought indices (SPI and RDI) and soil depths (5 cm and 30 cm) (except Melbourne). Overall SPI performs better (except shallow soils in Bourke) than RDI. However an uncertainty of the FR and FAR in the soil water retention curve is always higher than the uncertainties of drought indices. The complexity and the uncertainty in the model encourage to use the simple drought indices, however the model provide physically relevant soil water pressure values which are species specific for plants.