Verifying Regional Tropical Cyclone Model Forecasts using Synthetic Satellite Imagery

While regional tropical cyclone (TC) models are traditionally evaluated by the verification of track and intensity forecasts, these methods are limited by the lack of in-situ ground truth in TCs located far from land. These traditional metrics also to fail address the real or modeled TC structure, which is routinely observed by microwave instruments. This study addresses these shortcomings by systematically validating TC structure forecasts from the 2012 operational Hurricane Weather Research and Forecast (HWRF) model, through the comparison of newly developed synthetic microwave imagery with real microwave observations. Results from this analysis indicate that the HWRF shows some skill in forecasting the general convective organization of TCs in the Atlantic and Eastern Pacific basins. These results will be used to improve model forecasts of TC structure by identifying conditions in which the HWRF performs well or poorly. The results will also be used to improve the interpretation of HWRF forecasts in real time by forecasters at the National Hurricane Center.

Using the Community Radiative Transfer Model (CRTM), the Environmental Modeling Center (EMC) produces several simulated satellite products from the operational HWRF model. This study focuses on one of those products, the synthetic SSMIS 91GHz microwave brightness temperature. The 91GHz channel was selected because real instruments operating at that frequency have relatively high effective resolution, closer to the 3km grid spacing of the HWRF, than lower frequencies. Horizontal and vertical polarization brightness temperatures are converted to color composite images, equivalent to the color composites produced by the Navy Research Laboratory for real TCs. This conversion minimizes the influence of various assumptions, such as a constant synthetic instrument viewing angle, that are made during the generation of synthetic brightness temperatures. The conversion to color composite imagery also minimizes the effect of varying instrument resolution. The synthetic composite images are verified against color composite ice-scattering channel microwave images from the SSMIS, SSMI, TRMM, and AMSU instruments. The extent to which an eyewall and a primary band are present in real and synthetic imagery is quantified by measuring the radial coverage of the eyewall, expressed as a fraction, and the length of the primary band, using a Dvorak Log-10 spiral. The resulting database is used to verify 24, 48, and 72 hour forecasts, using percentage-based metrics such as reliability diagrams. The database is also used to asses situations in which the model performed particularly well or poorly. This information will help forecasters identify such situations in the future.