Extended Abstract (88K)3.11
Statistical models for lightning prediction using Canadian Lightning Detection Network observations
William R. Burrows, AES, Downsview, ON, Canada
Since February 1998 the Canadian Lightning Detection Network (CLDN) provides continuous lightning detection over all Canada to about 65°N in the west and 55°N in the east. Coverage is melded with the U.S. network. Previous information on thunderstorm occurrence was available only from manned surface stations and a few provincial networks with limited coverage. CLDN flash data was analyzed on equal area squares with 20 km sides to understand the "climatological" characteristics of lightning occurrence in Canada and adjacent United States. A complex pattern was revealed, showing strong latitudinal, seasonal, and diurnal dependencies, and significant influences by local elevated terrain features and major land-water boundaries. At this time work on the statistical lightning prediction models is just underway. Models will be built for several climatological regions identified from climatological lightning occurrence patterns. In statistical models for prediction the predictand is matched with several potential predictors derived from output of the Canadian Meteorological Center (CMC) numerical weather prediction model at grid points. Resolution is 22 km, going to 16 km later this year. The predictand is lightning report density, formed by giving each flash report a weight of 1 if it is within 0-10 km of a grid point and a decreasing weight 1-0 if it is within 10-20 km. Among the potential predictors are net CAPE, convective stability indices, helicity, elevation, binary land-water designation, predicted flash rate (Price and Rind, 1992), cloud height, tropopause height, rainfall, precipitable water (total and above 700 mb), geopotential height difference in 3 layers, wet bulb potential temperature, 500 mb temperature, 700 mb vertical motion, and time derivatives of several of these. Types of statistical models are not yet established, but a suite of hybrid models rule-based and regression or neural network models is currently envisioned. Available results will be shown at the conference.
Session 3, weather forecasting
Thursday, 17 January 2002, 8:30 AM-3:30 PM
Previous paper Next paper