Radar is a powerful tool for observing the atmosphere, and many derived products are useful to forecasters and researchers for a variety of purposes. Radar quality control is a vital step in the processing progression to ensure accurate products downstream. In particular, quantitative precipitation estimation (QPE) is highly dependent on the accuracy and quality of the radar data. There are many techniques for quality control already in existence, including the National Severe Storm Laboratory's Quality Control Neural Network (QCNN) system, which is being used operationally for the National Mosaic and Multi-Sensor QPE (NMQ) system.

The national WSR-88D radar network is currently being upgraded to dual-polarization technology. We believe that the new dual-polarization products, in conjunction with the legacy products, will greatly improve our ability to differentiate non-meteorological targets from hydrometeors. We intend to employ this capacity in order to remove non-hydrometeor returns from the reflectivity data while preserving hydrometeorological returns to aid in improving QPE. In particular, polarimetric variables, such as Correlation Coefficient (ρHV) and Differential Reflectivity (ZDR) offer significant improvements over legacy products for discriminating between non-meteorological and hydrometeorological echoes. In this project, the NSSL's QCNN was retrained to incorporate these polarimetric variables in an attempt to automatically identify and remove non-hydrometeor echoes, including biological echoes and anomalous propagation, from radar data. Early results have shown a significant increase in the ability to detect and remove biological echoes as well as anomalous propagation compared to the legacy QCNN. The retrained QCNN was too aggressive in removing weak reflectivities and struggled with strong reflectivity gradients. In addition, a range issue was noted where polarimetric data was no longer present at 300km from the radar site. Further research is being done into resolving these issues.