Spatial Depiction of Drought Variability from Merged Recent and Historical Data

Construction of spatially-detailed representations of historical drought variability is challenging. Exclusive use of long-term records eliminates the spatial information available in recent, more plentiful data. The use of all available data can lead to temporal inhomogeneities as stations come and go. Conventional spatial analysis involves interpolation that can artificially underestimate extremes such as drought.

We have applied inverse weighting of squared differences to estimate missing COOP data and obtain a data set free of temporal homogeneities in its mean values. To correct for suppression of extreme values for the purpose of drought estimation, we assume a gamma distribution for the monthly precipitation data and apply variance inflation to force the interpolated data to have the same variance as the actual data. The result is a data set that, while suboptimal for estimating the actual precipitation at any given location, is well-suited for comparing the relative prevalence of drought over multidecadal time periods.

We use this data set to create detailed maps of relative drought frequency, using the Standardized Precipitation Index drought index, across all decades of the 20th century. The maps show the regionalization of the Dust Bowl droughts, the 1950s drought, and other droughts of historic interest.