Real-time calibration of WSR-88D precipitation estimates

The advent of the WSR-88D weather radar network has ushered in a new era in hydrological and hydroclimatological monitoring. Until their development, precipitation estimates usually were limited to sparse gage networks that often were inadequate in resolving the spatial variability associated with storm-scale precipitation. Radars, by contrast, provide a high-resolution representation of the spatial distribution of precipitation with sufficient fidelity, given the 4 km by 4 km resolution of the Digital Precipitation Array (DPA), to resolve even most small-scale convective systems.

Many studies, however, have demonstrated that the WSR-88D weather radars tend to underestimate precipitation during heavy rainfall events (e.g., convective showers and tropical storm events), sometimes by as much as a factor of two or three (see Legates, 1999). By contrast, gage data, when adjusted for gage measurement biases, represent relatively accurate point estimates of precipitation. Given these inherent limitations of both raingage networks and the WSR-88D radars when used alone, a distinct need clearly exists to provide precipitation estimates from a combined radar-raingage approach.

Several approaches have been developed to provide a statistical calibration of radar precipitation estimates using raingage observations. They range from simple bias computation approaches to more complex schemes that involve subjective identification of areas where different reflectivity-to-rainfall relationships (Z-R) are applied. Simple bias computation approaches are purely statistical and do not attempt to model the sources of bias whereas many of the complex schemes are not amenable to real-time calibration owing to their need for constant user intervention.

Our goal has been to provide a robust, objective approach to calibrating radar-based precipitation estimates in real-time. The radar precipitation product is obtained from the Level III Digital Precipitation Array (DPA) that is part of the real-time distribution of radar products from the nationwide network of WSR-88D weather radars. To exploit the advantages of radar estimates and gage measurements, our radar-gage composite product uses the DPA to provide the spatial "footprint" of the storm while gage data are used to calibrate the radar estimate. Our calibration procedure is based on a model of the potential biases that lead to uncertainties in radar precipitation estimates. Thus, the calibration is a more physically-based approach than some methods and it does not require subjective decisions and human intervention. Consequently, this provides a more accurate, higher resolution precipitation field than can be afforded by the use of gage or radar estimates alone.

In this paper, we will present the method and its results throughout several regions of the country. Spatial scales range from a single radar umbrella to a composite area represented by more than twenty radar sites. Applications will focus on both extreme precipitation events as well as monitoring of normal conditions. An assessment of the degree of uncertainty that still remains in the combined radar-raingage product is computed and adaptations for future consideration also will be presented.