Convective cloud machine learning identification method based on satellite cloud picture

Abstract

A convective cloud machine learning identification method based on a satellite cloud picture comprises the following steps: 1, acquiring a cloud top brightness temperature historical time period and current monitoring data, and acquiring ground rainfall in the same historical time period; 2, interpolating the historical time period and the current TBB data to the grids of the ground precipitation field in the historical time period through bicubic spline interpolation; 3, generating a historical period brightness temperature index library and a current brightness temperature index library; 4, selecting a rainfall threshold value; 5, calculating mutual information of Y and X; 6, obtaining a nonlinear relation between Y and corresponding parameters through a support vector machine technology; 7, calculating mutual information of Z and X; 8, obtaining a nonlinear relation between Z and corresponding parameters through a support vector machine technology; 9, estimating the current probability of rainfall; 10, estimating the probability of existence or absence of convective rainfall at present; and 11, determining whether convective cloud exists. According to the method, the machine learning technology is utilized, the characteristics of the convection cloud on different channel satellite cloud pictures are quantitatively recognized, and the real-time recognition technology of the convection cloud is developed.

Description

technical field [0001] The invention relates to the technical field of satellite cloud image analysis, in particular to a method for machine learning identification of convective clouds based on satellite cloud images. Background technique [0002] Severe convective cloud clusters are often accompanied by catastrophic events such as strong winds, heavy rains, and huge waves, which can not only cause significant casualties, but also bring huge economic losses. Overland convective cloud monitoring can be achieved through ground automatic weather stations, ground-based radars, satellites, etc., while in the vast ocean, satellites are the primary means of convective cloud monitoring. Early studies mainly relied on the analysis of cloud shape, cloud coverage area and other visual appearance features on satellite cloud images to identify convective clouds. With the continuous upgrading and development of meteorological satellites, later studies mainly quantified the identificatio...

AI Technical Summary

Problems solved by technology

In the process of real-time processing of satellite cloud images, it is very difficult to select a definite threshold, which seriously restricts the real-time monitoring of convective clouds

Images