Research On Precipitation Forecasts In Nanjing

Rainfall is one of the primary forecasted factors, it has significant importance to predict precisely and timely, which can greatly reduce its damage. Nanjing located in the East Asian monsoon region and the lower reaches of Yangtze River, which can represent the main precipitation features in this area. In this paper, precipitation probability forecasts (PPF) are generated for 24h summer rainfall in Nanjing through stepwise regression and logistic regression methods. The regression equation uses the operational T213 products, observed rainfall data and auto weather station observation data in Nanjing station during 2004-2006. Meanwhile, the factors of synoptic scale and meso-and micro-scale causing precipitation are analyzed. A conceptual rainfall model has been established based on thermodynamic and statistical relationship among the selected predictors in the regression equation. It was found that:abundant water vapor accumulation of the middle and low layer, prepared the ground for the occurrence of the rainfall; the unstable stratification structure with low-level convergence and high-level divergence induced vertical convection, the transportation of the low layer leaning south air current to water vapor and the convergence climbing movement was of great benefit to the production and maintenance of the rainfall; vertical ascent motion provided dynamical condition for the production of the rainfall; The good configuration of warm advection in lower troposphere and cold advection in upper troposphere are provided the suitable thermal environment. It can also show the importance of the physics fields.The results from the logistic regression method are much better than that using regression estimation of event probability (REEP) in Threat Score (TS). In order to further improve PPF, optimal threshold has been applied to discriminate rain and no-rain events, depending on sample climatological frequency. Attributes diagrams are also used to establish the detail information of rainfall forecasts. The results show that logistic regression is superior to REEP in resolution and the reliability of lower probability. The main reason is that logistic regression raised the intensityof low altitude convergence and high altitude divergence.Meso-and micro-scale circulations are helpful to the occurrence of rainstorms and other severe convective weather under the cooperation of the favorable synoptic circulation. Nanjing city may be influenced by river-land breeze, slope and valley wind circulation and city cluster heat island circulation. Therefore, we take river-land breeze for an example to discuss the effect of local circulations, and the results are better in higher probability than before.