The Sanders Barotropic Tropical Cyclone Track Prediction Model (SANBAR)

Abstract

Abstract Sanders designed a barotropic tropical cyclone (TC) track prediction model for the North Atlantic TC basin that became known as the Sanders barotropic (SANBAR) model. It predicted the streamfunction of the deeplayer mean winds (tropical circulation vertically averaged from 1000 to 100 hPa) that represents the vertically averaged tropical circulations. Originally, the wind input for the operational objective analysis (OA) consisted of winds measured by radiosondes and 44 bogus winds provided by analysis at the National Hurricane Center (NHC), which corresponded to the vertically averaged flow over sparsely observed tropical, subtropical, and midlatitude oceanic regions. The model covered a fixed regional area and had a grid size of ~ 154 km. It estimated the initial storm motion solely on the basis of the large-scale flow from the OA, not taking into account the observed storm motion. During 1970, the SANBAR model became the first dynamical TC track model to be run operationally at NHC. Track forecasts of SANBAR were verified from the 1971 TC season when track model verifications began at NHC until its retirement after the 1989 Atlantic TC season. The average annual SANBAR forecast track errors were verified relative to Climatology and Persistence (CLIPER), the standard no-skill track forecast. Comparison with CLIPER determines the skill of track forecast methods. Verifications are presented for two different versions of the SANBAR model system used operationally during 1973–84 and 1985–89. In homogeneous comparisons (i.e., includes only forecasts for the same initial times) for the former period, SANBAR's track forecasts were slightly better than CLIPER at 24–48-h forecast intervals; however, from 1985 to 1989 the average SANBAR track forecast errors from 24–72 h were ~10% more skillful than homogeneous CLIPER track forecasts.