A Barotropic Model for Operational Prediction of Tracks of Tropical Storms

In response to early success in barotropic prediction of selected storm tracks, we undertook development of a system for operational application. The main effort was in devising a statistical method for automated analysis of the flow averaged in depth from 1000 to 100 mb. Initially the storm was represented in the flow pattern in a parametric way. Numerical calculation was carried out over a grid area from the Equator to 55°N latitude and from 36.5° to 123.5°W longitude, with a mesh length of 154 km. It was necessary to modify the prediction equation to prevent spurious retrogression of the long waves, despite the restricted grid area. It was also necessary to employ truncation-error control in the evaluation of Jacobians to avoid a spuriously slow predicted speed for the tropical storm, despite the small mesh length. The method of representing the storm in the initial flow pattern was found unsatisfactory. After considerable experimentation, we now require the pattern of the initial streamfunction over the region of storm influence to represent exactly the sum of a steering flow, derived from the best concurrent estimate of storm motion, and an idealized vortex described in terms of three variable parameters. The present model appears to have modest skill with respect to persistence and climatology in forecasts for the ranges from 24 to 72 h. Prominent present sources of error in barotropic forecasting are discussed, with recommendation for their mitigation, suggesting an achievable accuracy of 75 n mi at 24 h and 300 n mi at 72 h. Further reduction of errors would require baroclinic models.