Accuracy of motion-compensated NEXRAD precipitation
June 23, 2003
A number of Federal Aviation Administration (FAA) aviation weather systems utilize Next Generation Weather Radar (NEXRAD) precipitation products including the Integrated Terminal Weather System (ITWS), Corridor Integrated Weather System (CIWS), Medium Intensity Airport Weather System (MIAWS), and the Weather and Radar Processor (WARP). The precipitation products from a NEXRAD [e.g., base reflectivity, composite reflectivity (CR), and vertical integrated liquid (VIL)] are generally only updated once with each NEXRAD volume scan, nominally at 5-6 minute intervals. Hence, the indicated position of storms may not correspond to the actual position due to movement of the storms since the last NEXRAD product update. This latency is particularly a concern in terminal applications such as MIAWS, which use the NEXRAD precipitation product to provide time critical information on moderate and heavy precipitation impacts on the final approach and departure corridors and runways. In order to provide a more accurate depiction, the MIAWS precipitation map is updated (advected) every 30 seconds based on the motion of the storms. The CIWS system performs a similar advection of NEXRAD data before mosaicing the precipitation products from individual NEXRADs. In both cases, motion vectors used for advection are generated by spatial cross-correlation of two consecutive precipitation maps (Chornoboy et al., 1994). This report addresses the accuracy of the advected precipitation map as compared to the current NEXRAD precipitation map using seven MIAWS cases from the Memphis, TN testbed and Jackson, MS prototype. We find that the advected precipitation product is significantly more accurate at providing a depiction of the current intensity of the storms as a fbnction of location. Without advection, the precipitation product from successive NEXRAD volume scans differs by at least one VIP level for over 47.5% of the one square kilometer pixels and has VIP level differences of two levels or more for 6.9% of the pixels in cases where both products had precipitation in a location. The advected precipitation product differs by one or more levels in only 17.2% of the pixels and a VIP level difference of two or more levels is observed in only 1.6% of the pixels. The percentage of cells in which there is precipitation in one map and no precipitation in the other is reduced from over 22% to less than 11% by use of advection. The analysis approach utilized did not quantitatively determine the relative importance of storm growth and decay over the period of the volume scan versus errors in storm motion estimation in causing the differences between the advected precipitation field and the current precipitation field.