The effects of non‐stationarity on SPI for operational drought monitoring in Europe

Abstract

AbstractIt is a good practice to follow common guidelines in the computation of Standardized Precipitation Index (SPI) data sets as part of operational drought monitoring systems. In the European Drought Observatory (https://edo.jrc.ec.europa.eu/), reference statistics are computed following the World Meteorological Organization Guidelines on the Calculation of Climate Normals, where a definition of the reference period for monitoring applications is introduced as the most recent 30‐year period finishing in a year ending with 0. In this study, the temporal consistency of the SPI time series computed using the fifth‐generation European Centre for Medium‐range Weather Forecast reanalyses model precipitation data set is tested over Europe to quantify the effect of the transition from one baseline period (1981–2010) to another (1991–2020) and to evaluate the capability of these static baselines to reproduce the behaviour of non‐stationary SPIs (nSPIs). The results of the comparison suggest that the threshold commonly used to identify droughts (SPI = −1) is only marginally affected by the change in reference period (mean absolute deviation, MAD = 0.15 ± 0.1) for short‐term SPI, whereas larger differences (MAD up to 0.6) can be observed over certain areas (i.e., Southern Italy and Eastern Europe) for longer accumulation periods (i.e., SPI‐9 or SPI‐12). Examples show that changes in drought classification from extremely dry (SPI < −2) to moderately dry (SPI < −1) are not uncommon, which may lead to misinterpretation by users. Finally, analyses against nSPI highlight an overall good correspondence between stationary and nSPIs, even if both static baselines displayed difficulties in reliably capturing the magnitude of nSPI for the entire 10‐year period for which they should be used. In this regard, it has been demonstrated that more spatially uniform results can be achieved with 5‐year updates, with a good matching (MAD <0.25) for SPI‐1 and an acceptable matching (MAD < 0.50) for SPI‐12 over more than 80% of Europe.