Improved precipitation nowcasting algorithm using a high-resolution NWP model and national radar mosaic

Precipitation patterns can be reasonably well forecast in the short-term using extrapolation techniques to propagate radar reflectivity. However, without convective initiation, growth or decay provided by dynamic mesoscale processes, the forecast has limited skill after 1 or 2 hours. McGill University has created a nowcasting algorithm which combines an extrapolation technique with a fully dynamic Numerical Weather Prediction (NWP) model, to produce an enhanced precipitation nowcast out to 10 hours.

The Adjustment of Rain from Models with Radar (ARMOR) algorithm was developed at McGill University to combine the latest radar data with a model forecast of precipitation to produce a phase and intensity-corrected forecast with skill greater than the NWP model out to six hours. The ARMOR algorithm has been licensed and implemented at Weather Decision Technologies, Inc (WDT).

The WDT implementation of ARMOR uses a special version of the Weather Research and Forecasting (WRF) model at 10-km resolution over North America as its background, taking advantage of WRF's ability to predict mesoscale precipitation patterns that occur over the 0-12 hour timeframe. ARMOR combines WRF 15-minute precipitation accumulations with 1-km resolution North American low altitude radar mosaics. WRF spatial phase errors and precipitation intensity correction are derived by comparing the past radar mosaics with the model forecasts over the last hour. The spatial phase errors are expressed as linear functions of time at each model pixel. Corrections are applied at each time step to produce the ARMOR forecast.

This presentation will explain the ARMOR algorithm and its MAPLE heritage, provide samples of QPF output along with real-time radar verification and discuss the observed skill of ARMOR compared with the techniques of pure extrapolation or dynamic NWP models.