On the use of 3-D, dual-polarimetric radar data within the Current Icing Product

The Current Icing Product (CIP) is a real-time nowcast capability for In-Flight Icing Probability, Supercooled Large Drop (SLD) Potential and Icing Severity, which currently runs over the CONtiguous US (CONUS). The current operational version of CIP utilizes 2D, low-level composite radar reflectivity data, which is the maximum reflectivity in the lowest 4 km of the vertical column.

The overarching goal of this research is to utilize 3D, dual-polarimetric radar data to improve the performance of NCAR’s prototype CIP (P-CIP). One approach taken is to use the output from the Radar Icing Algorithm (RadIA), which was developed as a stand-alone icing detection algorithm using the National Weather Service Weather Surveillance-1988 Doppler (WSR-88D) radar data in its native coordinates. However, due to operational computing constraints, running RadIA on each of the individual WSR-88D radars within the CONUS is not currently feasible. Therefore, the current design is for RadIA to be computed on the Multi-Radar Multi-Sensor (MRMS) mosaic 3D radar data and then interpolated to the P-CIP grid.

Several challenges to implementing this approach have arisen. The first is that the computation of RadIA on the MRMS grid instead of native resolution radar data results in compromises in performance. Several inputs are computed from the spatial standard deviation of the radar measurements. Much of the information content of these standard deviations is lost in the interpolation to the coarser grid, degrading their usefulness. Second, it was discovered that the MRMS mosaic often exhibits a time lag between the reflectivity and dual-pol fields. Thus, the reflectivity field and the dual-pol fields that are given the same time stamp may actually originate from data collected at different times – up to about 10 minutes. If the echoes are moving, the result of the time lag is a spatial lag. Since our approach is to use the combined reflectivity and dual-polarimetric radar data as inputs into RadIA and P-CIP, the time-lag errors will have substantial and unpredictable negative impacts on performance.

The MRMS time-lag errors have been identified in both the MRMS data generated for the In-Cloud Icing and Large-Drop Experiment (ICICLE) using a replay capability and the real-time operational MRMS fields. The MRMS developers at National Oceanic and Atmospheric Administration are aware of these issues and have already fixed the ICICLE replay data and have plans to fix the operational data as well.

This study evaluates the skill of the RadIA inputs, called feature fields, computed from the MRMS data to differentiate in-flight icing conditions from ice only using the ICICLE research aircraft data and the MRMS replay. Also, we search for additional feature fields that have good skill and can be used to improve the performance of RadIA and resultant icing nowcasts from the P-CIP.

This paper will provide a progress update on the efforts to improve the performance of RadIA computed on MRMS data.

This research is in response to requirements and funding by the Federal Aviation Administration (FAA). The views expressed are those of the authors and do not necessarily represent the official policy or position of the FAA.