Accurate estimation of the spatial distribution of rainfall is critical to successfully model hydrologic processes. Rainfall distributions are typically estimated by assuming a spatial geometry tied to point rain gage observations using, for example, Thiessen polygons, inverse distance squared weighting, or statistical Kriging techniques. Unfortunately, the spatial distributions inferred by these approaches have little connection with how rain actually falls. From a modeling perspective, these techniques too often place the wrong rain at the wrong place at the wrong time.

In recent years, improvements in technology have made radar a viable tool to improve the estimation of rainfall between the gages. Radar provides a high resolution view of the variability of rain falling over a region. Unfortunately, radar by itself has not proven to be a consistent estimator of the actual rainfall amounts.

The strength of a rain gage network is its ability to consistently estimate rain falling on a number of discrete points. Its weakness is the network’s inability to estimate rain falling between the gages. On the other hand, radar’s strength is its ability to see between the gages but radar lacks the consistency in estimating rainfall at a specific point.

By merging rain data from a gage network and rain data derived from radar, hydrologists can take advantage of the strengths of each measurement system while minimizing their respective weaknesses. Essentially, a radar image is used as an areal template for the spatial distribution of rainfall. The rain gage data are used to scale the areal template. The net result is a gage-adjusted radar rainfall data set that combines the spatial characteristics of the radar image with the scaling information from the gages.

A specific adjustment algorithm was developed and applied to radar rainfall estimates over the state of Florida. The original radar rainfall estimates were extracted from a national dataset composed of a mosaic of rainfall estimates from the WSR-88D radar sites across the country. The data were smoothed in order to minimize discontinuities caused by ground clutter, anomalous propagation, and other processes inherent in the radar rainfall dataset. Due to the sheer size of the state of Florida and the varying climates and weather characteristics found across the state, a modified version of a spatial adjustment algorithm originally discussed in Brandes (1975) was used instead of a uniform radar bias adjustment. The spatially variable bias adjustment is calculated by determining the relative effect that each gage has on a given radar pixel, weighted primarily by distance. The resulting adjusted radar dataset consists of a field of rainfall estimates that closely match the rainfall volumes observed at the gage locations, while still maintaining smoothly varying spatial characteristics.