A Study On Refined Forecast Of Cloud Cover In Two Representative Stations Of Hulu Island

Cloud is a common weather phenomenon, which plays an important role in weather development and climate change. The change of cloud has a direct influence on residents’ daily life, economic production and military activity. And forecast of cloud cover has always been a difficulty. Hulu Island is an industrial, tourist and military city. So study of refined forecast of cloud cover in local representative stations has a significance for life service and military meteorological support.In this paper, Xingcheng and Suizhong are selected as representative stations of Hulu Island. Based on the data of conventional meteorological stations and prediction products from T639 model, interannual variation and annual variance of total cloud cover and low cloud cover are educed, main meteorological factors that affect the cloud cover are analyzed, and time refined quantitative forecast models are built. The main conclusions are as follows:(1) The interannual variation of cloud cover from 1964 to 2013 shows that total cloud cover has a significant decreasing trend while low cloud cover has an increasing trend(only Suizhong is significant). The precipitation in two stations show an indistinctive decreasing trend during the 50 years, which reflects that local precipitation efficiency of low cloud cover has dropped. And contribution of heavy rain to Xingcheng precipitation shows a significant decreasing trend, others are indistinctive. The annual cloud cover reach the peak in summer(always in July) and the valley in winter(always in December).(2) The direct factors that affect cloud cover are water vapor, vertical velocity, and atmospheric instability. Humidity of the whole layer, atmospheric precipitable water and relative humidity of the whole layer, which are related to water vapor, are significantly correlated with cloud cover. Average vertical velocity from 850 hPa to 400 hPa mandatory level is significant, too. K-index and geopotential instability index, which are related to atmospheric instability are also significant. Factors about intensity of weather system which include intensity of trough and vorticity at 500 hPa level, has an indirect effect on cloud cover by affecting factors mentioned before.(3) The final forecast factors are obtained by principal component analysis method which can extract the main information. The forecast models of total cloud cover and low cloud cover in two stations are established by stepwise regression and Support Vector Machine. The forecast results show that the average absolute error of stepwise regression model of total cloud cover is less than 30 percent, and the forecast accuracy is about 70%. While the average absolute error of SVM model is 20 percent, and the forecast accuracy is higher than 70%. Case studies show that both of two regression models can forecast the total cloud cover and the transform of cloudy day and sunny day well. In addition, SVM model has a slightly better forecast effect.In conclusion, the paper do a primary study on refined forecast of cloud cover in two representative stations of Hulu Island, the research method may provide guidance for quantitative explaining and application of numerical forecast products, and the results may also provide reference for local meteorological service.