loading page

NEXRAD weather radar coverage and National Weather Service warning performance
  • Daniel Melendez-Alvira,
  • Kate Abshire,
  • John Sokich
Daniel Melendez-Alvira
NOAA/NWS/Science and Technology Integration Office

Corresponding Author:[email protected]

Author Profile
Kate Abshire
NOAA/NWS/Analysis, Forecast and Support Office
Author Profile
John Sokich
NOAA/NWS/Congressional Affairs
Author Profile

Abstract

Over 33,000 flash flood events recorded by the NOAA National Weather Service (NWS) between 2006 and 2017, after polygon warnings replaced county-wide warnings, are analyzed in terms of NEXRAD (Next Generation Weather Radar) radar coverage in two ways. Events are binned in two large categories according to whether the starting point of an event falls inside or outside the range implied by the 6 kft AGL bottom-of-the-beam altitude, accounting for beam clutter and beam obstructions. Separately, events are classified according to the percentage surface area covered under the 6 kft beam height for each NWS Weather Forecast Office (WFO) or state. Emphasis is on unwarned events associated with fatalities, injuries and property damage and their relationship to WFO-area NEXRAD coverage. Significant property damage is defined in terms of the mean damage per event of about $411k. There is parity in the ratio of events warned or unwarned inside the inside/outside distance bins except for injury-event ratios. The linear fits to the traditional metrics of warned ratio (WR), false alarm ratio (FAR) and critical success index (CSI) are statistically significant and improve with more radar coverage. A stronger sensitivity is found for the WR as a function of the total number of events. New metrics are also evaluated, such as the mean flash flood partial event warning (PEW) and the ratios of unwarned to all events per WFO have statistically significant variations with radar coverage, improving with higher coverage. This is unlike the linear fits to the ratios of unwarned fatal-to-all-fatal events, which is marginally significant. Unwarned fatal-to-all-fatal, unwarned injury-to-all events and unwarned significant damage-to-nonzero damage event ratios are both not statistically significant to the 95% level of significance. The large scatter in the data results in very low coefficients of determination, and is interpreted to mean that radar coverage alone is not necessarily the primary driver though there is sensitivity to variations in such coverage.