基於衛星數據的大數據中雨預報方法研究

RESEARCH ON BIG DATA MODERATE RAINFALL FORECAST METHOD BASED ON SATELLITE DATA

10.6652/JoCICHE.202406_36(4).0007

李夢靜(Mengjing Li)；楊富榮(Furong Yang)；崔晨風(Chenfeng Cui)

矩陣匹配 ； 圖像相似 ； 中值降水 ； 降水預報 ； matrix matching ； matrix similarity ； precipitation ； precipitation forecast

中國土木水利工程學刊

36卷4期（2024 / 06 / 01）

387 - 393

繁體中文;英文

中雨的降水強度適中，時間相對較長，對改善環境、促進農作物生長等都具有重要作用。目前常見的降水預報方法中，對於中雨的預報仍存在範圍界定模糊、預報精度不夠高等問題。本文採用的矩陣匹配方法，基於1951～2017年的大量衛星數據，填補了中值降雨預報方法研究的空缺，綜合考慮各種氣象要素的影響，並且可以更高精度地預報中雨。結果表明：隨著氣象要素數據的增加，相似指標值的範圍以及取值越來越小，以河南省鄭州市為例，最終結果有100%的d值小於0.27779，59.24%的d值小於0.09522；且與傳統線性回歸方法相比，RMSE減少約30%，ETS評分增加0.09。因此，利用矩陣匹配法預報中雨，對於工農業生產等都具有重要的意義。

The precipitation intensity of the moderate rainfall is moderate and the duration is relatively long, which plays an important role in improving the environment and promoting the growth of crops. At present, there are still some problems in the prediction of moderate rain, such as fuzzy definition of the range and not high enough accuracy. The matrix matching method adopted in this paper, based on a large number of satellite data from 1951 to 2017, fills the gap in the research of moderate rainfall forecast methods, takes the influence of various meteorological elements into account comprehensively, and can forecast moderate rainfall with higher accuracy. The results show that with the increase of meteorological element data, the range and value of similar index values become smaller and smaller. Taking Zhengzhou City of Henan Province as an example, 100% of the d value is less than 0.27779, and 59.24% of the d value is less than 0.09522., and compared with the traditional linear regression method, RMSE decreased by about 30% and ETS score increased by 0.09. Therefore, using matrix matching method to forecast moderate rain is of great significance to industrial and agricultural production.