Statistical Correction of Radar Data

First results of radar based climatologies have emerged over the last couple of years, as data sets of appropriate extent are becoming available. Not unexpectedly, however, data quality turns out to be one of the major problems in such studies. Besides calibration drifts or maintanance jumps in the time series of radar data, slight systematic deviations among pixels of one individual radar image may have escaped (or even been introduced by) the basic correction algorithms as unconspicuous and negligible.

Even small systematic deviations may seriously bias areal climatology and should be removed. The approach chosen here is to render them visible by means of statistics, considering frequencies instead of intensities. The underlying assumption is that regions may be found where light precipitation is expected to be equally distributed. The frequency of certain threshold exceedances has been calculated for individual radars over 7 years of data. The bright band becomes apparent statistically, and the median of the frequency shows a pronounced dependence on the beam height over ground. A linear regression model has then been established to quantify the artificial decrease. The result may be used to correct climatological analyses by a statistical correction. Correction of individual radar images is not proposed. The correction algorithm is suitable for climatological or statistical analyses with a temporal resolution of at least one year; this limits the possible application. "Suspicious" regions or pixels such as shading spokes or clutter remnants have been disregarded in the statistical analysis but corrected for by an interpolation scheme. The positive effect of the correction has been verified by gauge statistics.