翻译：As shown in Fig 5 for all secondary sub-series their SE and RMSE values are much lower and R2 are much greater than their counterparts of S1 As a result VMD increased R2 greatly from 016 to 067 and

如图5所示，对于所有的次级子系列，它们的SE和RMSE值比S1的对应值要低得多，而R2值则要高得多。因此，VMD将广州的S1的R2值从0.16大大提高到0.67，并将RMSE值从19.0降低到12.4。这清楚地表明，VMD可以进一步减少S1的不规则性，从而便于机器学习模型进行特征识别和信息提取，从而改善MDA8 O3的预测性能。