Автоматическая генерация программ для графических процессоров по непроцедурным спецификациям

Abstract

The mesoscale convective system (MCS) cinematic and evolution routinely derived from realtime radar reflectivity Z data are used to develop 3 step short-range forecasting technique. The first forecasting step is estimation of MCS translation velocity. It is proposed that translation can be estimated by robust interactive procedure, identifying and tracking of “conservative meso-β fragments”, or by automatic clustering algorithms. The second step is nowcasting when entire Z field is extrapolated by translation to forecast time ~0,5-1 hours. The third step is forecast verification and MCS diagnostic when prognostic and real-time images of Z are coupled to one composite image, denominated as life-cycle display (LCD). The construction of LCD is adapted to use both in forecast verification in terms of FAR, POD and CSI and to display 3 type conventionally outlined MCS areas where: 1) ongoing convection occur; 2) dissipating convection/stratiform transformation and 3) precipitation was recently ceased. As poor extrapolation forecasts associated, mainly, with growth and decay of storms in the forecast period, life-cycle composite images provide valuable information about of developing, dissipating areas and areas prohibited to new convective development.