基于最小面积准则的天气图等值线相似算法

为了进一步提高天气形势的相似预报精度,基于最近邻非参数回归方法的思想,提出了一种基于最小面积准则的天气图等值线相似识别方法。该方法利用有限区域内2条等值线所围成的面积作为一种新的相似性测度,将相似性测度从一维的“距离”提高到二维的“面积”,并用数学方法证明了其能够同时兼顾形相似和值相似。试验表明,与传统的相似算法相比,该算法不但操作简单方便,而且对天气图中等值线的相似识别具有更好的效果,便于在基层气象台站推广使用。

Min-area-based contour similarity algorithm of synoptic chart

To improve the accuracy of the similarity calculation, a novel similarity algorithm was presented in the paper to identify the contours in the synoptic chart, which is based on the principle of area minimum. The complicated contour similarity calculation was simplified by sorting the enclosed area of the contour lines. The proposed method is suitable for value similarity and shape similarity, which is proven by the mathematical way. Compared with the similarity discrete value method, the better results can be obtained by the proposed method in the test. The presented method can be implemented with the good performance in an easy way, and can be widely applied to most forecasting mission.