The Global Drought Hazard project is a collection of spatial raster datasets that provide access to statistical extreme value analyses of the Standardised Precipitation Evapotranspiration Index (6-month SPEI) to identify high risk areas on a global scale.
The source data are monthly spatial raster datasets from the Climatology and Climate Services Laboratory's Global SPEI Database for the period from January 1902 to December 2018. Each raster dataset has a spatial resolution of about 0.5 degrees, the values of which we converted into time series for each grid cell. These time series are used to model the number of months above selected SPEI thresholds (e.g. SPEI -1.5 or lower) for a range of return periods (e.g. 100 years) via Poisson-Generalized Pareto Point Process models.
Since extreme value analyses are highly influenced by non-stationarity and time dependence of the data, time-varying parameters to deal with non-stationarity, declustering to reduce tail dependence, and spatiotemporally weighted error term corrections were incorporated into the modelling process for each cell. The modelling results for each return period are projected bilinearly onto a spatial grid with a higher resolution of about 0.0083 degrees for publication.