Page 90 - 《高原气象》2026年第3期
P. 90
高 原 气 象 45 卷
698
图3 CBMA U-Net、 Attention U-Net、 U-Net、 R(Z )-Z和R(Z )-L估计分钟雨量的CC (a)、 R MSE (b)、 R MAE (c)、 R (d)箱线图
H
H
MB
箱体上下边为上下四分位数,箱内横线为中位数,虚线两端横线为极值点,离散点为异常值
Fig. 3 Box plots of CC (a), R MSE (b), R MAE (c) and R (d) for minute rainfall estimated by CBAM U-Net, Attention
MB
U-Net, U-Net, R(Z )-Z and R(Z )-L. The upper and lower edges of the box represent the upper and lower quartiles,
H
H
the horizontal line inside indicates the median, the horizontal lines at the end of the dashed line
indicate the extreme points, and the discrete points represent the outliers
表5 不同分钟雨量阈值下分类评估指标统计
Table 5 Statistics of classification evaluation metrics for different minute rainfall thresholds
分钟雨量阈值 评估 模型
-1
/ (mm·min ) 指标 CBAM U-Net Attention U-Net U-Net R(Z )-Z R(Z )-L
H H
0. 10 CSI 0. 894 0. 846 0. 860 0. 820 0. 842
POD 0. 992 0. 946 0. 952 0. 890 0. 922
FAR 0. 100 0. 112 0. 101 0. 088 0. 094
0. 34 CSI 0. 655 0. 609 0. 537 0. 492 0. 555
POD 0. 915 0. 849 0. 650 0. 608 0. 713
FAR 0. 302 0. 316 0. 243 0. 280 0. 285
0. 67 CSI 0. 485 0. 408 0. 160 0. 244 0. 298
POD 0. 643 0. 608 0. 176 0. 331 0. 437
FAR 0. 336 0. 447 0. 373 0. 518 0. 517
1. 17 CSI 0. 181 0. 084 0. 008 0. 074 0. 092
POD 0. 204 0. 100 0. 008 0. 130 0. 198
FAR 0. 385 0. 650 0. 538 0. 852 0. 854

