Using Gauge-Adjusted Radar Rainfall Data

Historically, analysis of weather impacts on water demand have required that historical precipitation throughout a study region be estimated from data at specific points (stations) within or even outside that region. Historical precipitation data at stations are usually freely available from agencies such as NOAA, but estimates of precipitation between stations are inevitably subject to errors of unknown magnitude simply due to the spatial heterogeneity of precipitation itself. This heterogeneity, when severe, can obfuscate demand-weather analyses at all but the coarsest spatial and temporal scales.

This paper describes the application of Gauge-Adjusted Radar Rainfall (GARR) data, a typically proprietary type of historical precipitation data, for higher-precision weather-demand analyses. GARR data are developed by first maintaining a long archive of historical radar images, each a high-resolution grid of pixels covering a geographic area. These images, which are familiar sights on weather reports, are at a resolution as small as 1-km by 1-km and are updated as frequently as every 15 minutes. These images are then compared to historical precipitation data at stations, calibrating the strength of precipitation indicated by radar to actual measurements. The resulting GARR data product contains spatial distributions of precipitation across an area that are more precise and accurate than could ever be possible using station data alone, potentially affording weather/demand analyses with lower measurement errors at higher resolutions than achievable in the past.

This paper will demonstrate the application of GARR data for demand/weather analysis using Tampa Bay Water. This GARR data, which was purchased by and subsequently obtained from the Southwest Florida Water Management District, covers most of the western Florida peninsula for the period January 1, 1995 to December 31, 2014 at 2-km by 2-km resolution and 15 minute intervals. Procedures for applying GARR data will be outlined, sample analyses will be shown, and weather/demand relationships using GARR data will be compared with similar analyses using station data alone. Possible future applications of the data will be discussed as well.