Minimizing Biases in Radar Precipitation Estimates
MetadataShow full item record
The demand for real-time drought information in recent years led to the development of a suite of objective drought indicators that relies on the high-resolution Stage IV precipitation estimates that are produced each day by the National Weather Service in near real-time. The drawback to using the Stage IV dataset for this purpose is the presence of numerous biases in the estimates, which lead to erroneous assessments of drought conditions. Among the types of biases in the Stage IV dataset are 1. Underestimation of precipitation due to beam blockage. 2. Range-dependent errors that originating from the measurement of reflectivity above the surface. 3. Mean-field biases resulting from radar calibration and measurement errors. A three stage bias correction procedure is developed and evaluated for minimizing the biases, methods used to produce an improved, bias-adjusted Stage IV precipitation dataset. The original Stage IV data are initially corrected by a beam blockage identification procedure and Kriging interpolation to replace the precipitation values in grid cells affected by blockage. Next, range-dependent and mean field biases are identified and corrected by use of a statistical model based on the vertical profile of reflectivity in mixed-phase precipitating systems. The last bias quantification procedure estimates and removes a two-dimensional field of residual biases using available gauges as an assumed unbiased estimate of the ground truth. Data withholding testing showed the bias-adjusted Stage IV dataset to have a significant reduction in the overall bias relative to the original Stage IV precipitation dataset. This includes a reduction in the overall bias at each of the three major steps. The bias-adjusted Stage IV dataset will be utilized in the drought indicators to enable a better objective assessment of real-time drought conditions.
McRoberts, Douglas B (2014). Minimizing Biases in Radar Precipitation Estimates. Doctoral dissertation, Texas A & M University. Available electronically from