Statistics downscaling method based on SVM algorithm

Abstract

The invention relates to a statistics downscaling method based on the SVM algorithm. The statistics downscaling method includes the following steps that firstly, large-scale weather forecast factors are screened and representative factors are obtained; secondly, a wet day state and a dry day state of existing day-by-day rainfall data are judged, the data are processed, an SVM is used for constructing a classification relation between a raining state and the corresponding forecast factor at first, and a statistics relation on raining in wet days and meteorological factors is constructed; thirdly, an existing classification relation is used for classifying rainfall states in the incoming days through the weather forecast factors; fourthly, an existing statistics relation is used for fitting rainfall data on incoming rainy days through the weather forecast factors; fifthly, the data are restored and a forecast rainfall sequence is obtained. The statistics downscaling method based on the SVM algorithm has the advantages that effects of the statistics downscaling method are superior to those of a traditional statistics downscaling method and advantages are prominent when downsizing is carried out on the temperature; the operation volume is little, the method is convenient to use and specialized in processing a large volume of data of multiple batches, rainy days and dry days can be distinguished and higher precision is obtained; random values are added into rainfall and the fitting precision of rainstorms is higher.

Description

technical field [0001] The invention relates to a statistical downscaling method, in particular to a statistical downscaling method based on a support vector machine algorithm. Background technique [0002] The downscaling method is to degrade the large-scale predictors to a small-scale area through a series of processing, so that it can match the scale input by the hydrological model. Downscaling methods are mainly divided into dynamic downscaling and statistical downscaling. Compared with dynamic downscaling, statistical downscaling method requires less calculation, and the construction of model algorithm is relatively simple, and there are many models with more flexible forms to choose from. [0003] The existing statistical downscaling methods have the following deficiencies: (1) The regression accuracy is not enough, the fitting ability is not strong, and the amount of calculation is not streamlined; (2) With a small number of samples, it is difficult for traditional s...

AI Technical Summary

Problems solved by technology

[0003] The existing statistical downscaling methods have the following deficiencies: (1) The regression accuracy is not enough, the fitting ability is not strong, and the amount of calculation is not streamlined; (2) With a small number of samples, it is difficult for traditional statistical methods to achieve ideal regression results

Images