| description abstract | Renard recently reported (Monthly Weather Review, July 1968) on the development of a numerical scheme for predicting the motion of tropical storms for periods up to three days. An extension of the forecast scheme, as presented here, may be described as a two-step process. First, numerical geostrophic steering of the cyclone center is accomplished using Fleet Numerical Weather Central's analyses and prognoses of smoothed isobaric height fields, called SR fields. Next, a statistical correction for vector bias in the numerical steering computation is used selectively in an attempt to enhance the accuracy of the forecast track of the storm. The bias modification is dependent solely on the peculiarities of recent history 12- and 24-hr forecasts in relation to the storm's actual trajectory. Forecasts for intervals up to 72 hr, generated from 1967 Atlantic operational storm positions, are compared to results from the previous experimental forecasts for 1965 using best-track positions of Atlantic storms.Results indicate that the numerical scheme shows skill in relation to the official forecast accuracy for both 1965 and 1967, as documented by Fleet Weather Facility, Jacksonville, Fla. In 1967, the average relative improvement over official forecasts, using 700-mb prognostic SR fields for steering, ranges from 52% for 12-hr forecasts (90 cases) to 9% for the 48-hr estimates (129 cases).The paper includes a discussion of various forecast modes and selective modification schemes as well as stratification of error statistics by geographical area, storm track and stage. | |
