Research On Precipitation Forecast Of GPS Precipitable Water Vapor Based On EMD And RBF

Weather plays an important role in national economic construction while the changes of water vapor in atmospheric plays an important role in weather. In order to provide important theoretical basis and practical basis for the prevention of heavy rain and other disasters, the prediction of the changes in water vapor need to be more precision. Further more, the peak value of the change in GPS precipitable water generally ahead of the actual precipitation 1h ~ 2hs, so the GPS can be used as an important indicator of short-term rainfall precipitation forecasts.Due to the non-stationary and non-liner characteristics of the GPS precipitable water, the combination of empirical mode decomposition and RBF neural network method in the GPS and meteorological factors precipitable water data about Qinhuangdao to do prediction.Though the weather data at July and August in Qinhuangdao to inverse the GPS precipitable water, then compared with the actual precipitation water, a strong relevance was found. The probability of precipitation after a peak of GPS precipitable water 1h ~ 2hs is 75%, it can be judged that, relative to the actual precipitable water the changes of GPS precipitable water precipitation have some advance. have some advance.When decompose the EMD with the noise signal in Matlab, the selected neural network continuation method and Waveform matching continuation method to solve the problem of the endpoint get a good result. The results of precipitation forecast indicate that: to use the GPS PWV and weather factors as predictors to establish RBF neural network has the ability to forecast precipitation. Directly use RBF neural network model to forecast the precipitation, the accuracy rate in an hour will be 56.3%, and 46.5% in two hours. In addition, the application of EMD-RBF forecasting model to forecast the precipitation, the accuracy rate in an hour is 82.9%, and 91.5% in two hours.