Abstract
A lightning–precipitation relationship (LPR) is studied at high temporal and spatial resolution (5 min and 5 km). As a proof of concept of these methods, precipitation data are retrieved from the National Severe Storms Laboratory (NSSL) NMQ product for southern Arizona and western Texas while lightning data are provided by the National Lightning Detection Network (NLDN). A spatial- and time-invariant (STI) linear model that considers spatial neighbors and time lags is proposed. A data denial analysis is performed over Midland, Texas (a region with good sensor coverage), with this STI model. The LPR is unchanged and essentially equal, regardless of the domain (denial or complete) used to obtain the STI model coefficients. It is argued that precipitation can be estimated over regions with poor sensor coverage (i.e., southern Arizona) by calibrating the LPR over well-covered domains that are experiencing similar storm conditions. To obtain a lightning-estimated precipitation that dynamically updates the model coefficients in time, a Kalman filter is applied to the STI model. The correlation between the observed and estimated precipitation is statistically significant for both models, but the Kalman filter provides a better precipitation estimation. The best demonstration of this application is a heavy-precipitation, high-frequency lightning event in southern Arizona over a region with poor radar and rain gauge coverage. By calibrating the Kalman filter over a data-covered domain, the lightning-estimated precipitation is considerably greater than that estimated by radar alone. Therefore, for regions where both rain gauge and radar data are compromised, lightning provides a viable alternative for improving QPE.
© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
