Prediction Of Rainy Season Precipitation In China Based On Dynamical And Analogical Skills

Because of the difficulty and uncertainty of short-term climate prediction, the research of short-term climate prediction in China no longer simply focuses on the application of the physical statistical forecasting methods or numerical model prediction method, but also attaches importance to the combination of the two methods. In view of this, based on the basic principles of dynamic-statistics and the short-term climate prediction mode CGCM of National Climate Center, we construct a variety of dynamic-statistics prediction schemes from the angle of the dynamic characteristics, static characteristics and abnormal strong signal of the ambient field of earlier stage. We not only carry out the dynamic-statistics prediction test which is integrated by several schemes, but also develop forecasting performance assessment method as well as droughts and abnormal level reliability calculation method. Thus dynamics methods and statistical methods are combined effectively and learn from each other. Through this way we can avoid substantial modifications on the dynamic framework of numerical model and get the same prediction effect at the relatively small expense, historical statistics are made full use of to correct the systematic error generated from dynamics method in the integration process. Finally, based on the platform, we set up the short-term climate objective prediction system in which dynamic and statistical are combined effectively to promote the improvement of prediction accuracy rate and provide better service to government’s disaster prevention and mitigation decision-making. The main conclusions are as follows:1. The ideal test results of dynamic-statistics prediction show that the information of similar error field in the historical data is quite enough for the prediction error field of precipitation models of the flood season in the various regions. The precipitation characteristics in every year can be find similar years from the historical data. The best similar year is selected by a scientific similar criterion to forecasting precipitation by dynamic-statistics prediction method very well. However, maximum theoretical forecasts exist in each region, for example, under ideal conditions, the maximum ACC of regions is about0.8and the forecasting results is the best when the number of the ideal similar years is four.2.Based on the idea of using the historical-analogue information to revise the prediction errors of National Climate Centre numerical business model, For North China, based on analysis data of the CMAP from1983to2009,40pieces of climate indices from NOAA,27years of the season prediction model results from1983to2009and74pieces of circulation characteristics materials, using the method of combining data analysis and numerical simulation of diagnostic tests, taking the advantage of the prediction error of the key information of similar model from the historical data, by identifying key factors, optimizing allocation of the different factors of different forecasting years, to establish specific multi-factor dynamic optimal portfolios to revise prediction errors in different periods of the power statistical model in North China, to construct early environmental factors similar to field multiple objective criteria, to develop new technology of revising prediction errors from the power-statistical model based on dynamic optimal combination of multi-factor, to improve the prediction effect in the summer precipitation in North China and the forecasting skills. Results of independent sample return of2005-2009shows that, the score of similarity revised method has improved significantly comparing to the score of Systematic revised method.3.We studied the correlations between the interactions of prophase key factors and the precipitation of rainy season in China and found the key atmospheric circulation predictor. According to the predictors which are abnormal in prophase, the authors compressed the dimensions of the factors to select the similar years through EOF analysis. Furthermore, a new dynamical analogue prediction scheme is constructed, which is based on the anomalous signals of prophase environment field. Analyses show that there is a good corresponding relationship between precipitation in North China and numbers of atmospheric circulation factors which are abnormal in prophase. The authors developed a comprehensive scheme to revise prediction errors of numerical model combined with the abnormal factors scheme and the optimal dynamic multi-factor scheme.Through the diagnostic analysis, the authors found that the comprehensive scheme has a good adaptability. Results of independent sample return during2003-2009show that the anomaly correlation coefficient (ACC) score has increased from0.38to0.61.The similarity revised method has further improved the prediction capacity of numerical model and has a good application prospect for summer precipitation in North China.4.Based on dynamic characteristics, we established an objective and quantitative forecasting scheme for summer precipitation in China, which was based on finding the similar evolutions of sea surface temperature (SST) in key areas. Different from the multi-factor scheme which has been founded with an eye to the static state of prophase environmental fields, the new scheme is focused on the dynamic state and is attempting to find the analogous year from the history according to the evolution of prophase SST. Using the observational data of precipitation and SST in recent decades, we have screened out the critical areas, which had a great impact on the summer precipitation of China. Similarity coefficient was used as the criterion to show the similarity degree of evolutions and then we calculated the best threshold value through forecasting verification. Result of independent sample return of2002-2011shows that it has a good consistency between predictions and observations and the accuracy is promoted significantly compared with the system revised prediction. The predictive ability for abnormal precipitation has been improved, making up the deficiency of multi-factor scheme.The new scheme provided us a feasible method to increase the accuracy of objective and quantitative forecasting system.5.We apply the dynamical analogue prediction scheme to predict the subtropical anticyclone in western Pacific.The key prophase predictors for forecasting subtropical anticyclone in summer were separated from the atmospheric circulation factors through correlation analysis and cross validating the anomaly correlation coefficients (ACC).Then we have made the independent sample return test of2003-2010, the results show that the optimal dynamic multi-factor schemes can help the numerical model improving the accuracy of prediction. Based on this, we extract the two typical indexes (western ridge point and ridge line index) from the prediction of subtropical anticyclone, which can represent the characteristics of subtropical anticyclone.Then we project the two indexes in the two-dimensional plane and associate this work with the first part study of statistical classification of subtropical anticyclone.Furthermore, we get the summer precipitation distribution type of forecasting years correspond to the type of subtropical anticyclone.The result shows that the distributions of precipitation correspond to the projection of the type of subtropical anticyclone are consistent with the observations, which demonstrate the rationality of this type of classification of the subtropical anticyclone and precipitation distribution. Based on this study we could carry our point of forecasting the monsoon precipitation through the objective and quantitative prediction of subtropical anticyclone so as to provide a possible scheme for improving the monsoon precipitation prediction skill.6.The results of multi-scheme prediction tests indicate that the forecasting effect of the dynamic preferable integrated scheme is the most stable and the forecasting techniques are the highest. Specific to a given year, the integrated forecasting techniques are not necessarily higher than the best single forecasting scheme, but it is always better than the worst forecasting scheme, so that the annual forecasting effect can be maintained at a relatively stable level and the great fluctuation of forecasting effect can be avoided effectively, thereby the credibility of the prediction is enhanced.7.Through the statistics of the forecast results of summer precipitation in many years and the errors,we find that the distribution of the model prediction errors satisfies the Gaussian distribution. Based on the Gaussian distribution characteristics, prediction ability of mode for summer precipitation in China can be analyzed and compared, for example, viewing from the position of the error distribution, the more the normal distribution is satisfied, the better the forecast performance is:viewing from the shape of the fitting, the "slenderer" distribution pattern is, the better the forecast performance is. The forecast performance of the dynamic-statistics scheme is significantly better than that of system error correction scheme of the model. The dynamic-statistics optimal combination of factors revise is corrected, relative to error distribution pattern of the system revised forecast:(1) improvement of the amplitude (the deviation decreases);(2) improvement of the displacement (the drift of the system correction is revised). The interdecadal distribution characteristics of forecast error frequency are very consistent and there are almost no changes. The maximum prediction errors frequency are all around0, which is indicate that the distributions of the model forecast errors are relatively stable and there is no obvious interdecadal variability.8. Based on the statistical characteristics of the prediction error, measure of the credibility of model forecast results is came up to quantitatively assess credibility of the numerical model forecast results in different regions. Putting the summer precipitation in2012as an example, we analyse the credibility of actual observations, prediction and drought or flood level. Comparing actual observations and prediction, we find that there is a good consistency where the credibility of them is larger, while there are relatively many bad predictions in the region where the credibility of them is smaller. What’s more, compared with other reliability test method, the credibility of the abnormal precipitation forecast by this method is more accurate, showing the effectiveness of the method.9.The FODAS is constructed initially and used to actually predict and inspect in2009-2012. In terms of seasonal forecast, there is a good predictive effect on summer precipitation forecast in2009-2012. The four-year average forecast score of the FODAS is72.8and the average ACC is0.16. It is obvious that the forecast results of the FODAS are better than that of numerical model and there is a good forecast capability for abnormal precipitation by using the FODAS. Moreover, the FODAS effectively forecasts the winter climate trends of2012as well. The PS score of winter temperature forecast in2012is92and PS score of precipitation is82. At the same time, from the perspective of the application effect of month forecast, there are also a certain degree of forecast skill and a forecast capability for abnormal precipitation. But as a whole, the month forecast effect is not better than the seasonal prediction effect, it is not stable enough and there is greater volatility. The FODAS makes a good performance in the temperature and precipitation forecast of summer and winter. Importantly, the value of the FODAS to be applied and promote is in the beginning of another and there is a big space for the FODAS to be developed, improved and promoted.