Radar reflectivity data can be quality-controlled using just the radar moments, and several techniques have been proposed to do this. It is possible to use texture features as inputs to a neural network to discriminate between precipitating radar echoes, and echoes that correspond to clear-air return, ground clutter or anamalous propagation. The texture feature neural network that was recently developed at the National Severe Storms Laboratory performs much better at this discrimination than other radar-moment-based quality control methods discussed in the literature.

However none of the radar-only techniques can discriminate betwen shallow precipitation and spatially smooth clear-air return. The radar-only techniques also have problems removing some biological targets, chaff and terrain-induced ground clutter far away from the radar.

We show how the use of satellite infrared channel data and surface observations can help the radar quality problem in these situations. There are several practical issues related to using satellite and surface data, however, mostly having to do with the low spatial and temporal resolution of the non-radar observations. We describe the considerations in assimilating infrared data from satellites and surface observations from mesoscale models especially with regard to temporal resolution. We then describe using the assimilated grid to remove clutter and non-precipitating targets from radar reflectivity data. Based on archived data sets, we show that such quality control works, and would be useful, if the surface observations can be received in near-realtime.