Categorical forecast of precipitation anomaly using the standardized precipitation index SPI

Abstract

In the paper, the verification of forecasts of precipitation conditions measured by the standardized precipitation index (SPI) is presented. For the verification of categorical forecasts, a contingency table was used. Standard verification measures were used for the SPI value forecast. The 30-day SPI, moved every 10 days by 10 days, was calculated in 2013-2015 from April to September on the basis of precipitation data from 35 meteorological stations in Poland. Predictions of the 30-day SPI were created in which precipitation was forecasted for the next 10 days (the SPI 10-day forecast) and 20 days (the SPI 20-day forecast). For both the 10 and 20 days, the forecasts were skewed towards drier categories at the expense of wet categories. There was a good agreement between observed and 10-day forecast categories of precipitation. Less agreement is obtained for 20-day forecasts-these forecasts evidently "over-dry" the assessment of precipitation anomalies. The 10-day SPI value forecast accuracy is very good or good depending on the performance measure, whereas accuracy of the 20-day forecast is unsatisfactory. Both for the SPI categorical and the SPI value forecast, the 10-day SPI forecast is trustworthy and the 20-day forecast should be accepted with reservation and used with caution.