Fig 1.
Study area.
Fig 2.
Annual accumulated rainfall in the study area.
Fig 3.
Proposed framework for rainfall estimation.
Table 1.
Rainfall thresholds used to classify rainfall classes.
Table 2.
List of hyperparameters used for model training.
Fig 4.
Number of rainfall samples in the input datasets of the models. (a) M1, (b) M2, (c) M3, (d) M4.
Table 3.
Basic metrics for evaluating classification models.
Table 4.
Basic metrics for evaluating regression models.
Fig 5.
Variation of BT values of IR bands with respect to rainfall.
(a) WVB, (b) IRB, (c) I2B, (d) 2B14-I2B-IRB.
Fig 6.
Rain/no-rain classification maps from the proposed model and radar.
(a) rain/no-rain classification maps from model M1; (b) classification maps of rain/no-rain-bearing clouds based on CM1; (c) combined rain/no-rain classification maps from M1+CM1; (d) reference rain/no-rain classification maps from radar observations.
Table 5.
List of feature subsets used for model training.
Table 6.
Number of samples across rainfall intervals before and after applying the RR technique.
Table 7.
Hyperparameter optimization.
Fig 7.
Variation of BT values from IR bands for rainfall rates .
(a) WVB, (b) I4B, (c) IRB, (d) B14.
Fig 8.
Classification maps of low-intensity rain and high-intensity rain from the proposed model and radar.
(a) Classification of low-intensity rain and high-intensity rain from model M2; (b) Classification of clouds bearing low-intensity and clouds bearing high-intensity rain based on CM2; (c) Combined classification results from ; (d) Radar-based classification of low-intensity rain and high-intensity rain.
Fig 9.
Rainfall classification maps of the proposed product and radar.
(a) Classification of small rain and moderate rain by model M3, applied to areas identified as low-intensity rain; (b) Classification of heavy rain and very heavy rain by model M4, applied to areas identified as high-intensity rain; (c) The final rainfall classification map of the proposed rainfall product; (d) Reference rainfall classification map from radar.
Fig 10.
Classification performance of rainfall products.
The black solid line represents the CSI, while the blue dashed line indicates the BIAS.
Table 8.
Rain classification performance of the proposed product and radar.
Fig 11.
Rain/no-rain classification maps of the rainfall products.
Table 9.
Classification performance of rainfall products compared to radar in five rainfall events.
Fig 12.
Regression performance of rainfall estimation: (a) CC, (b) mKGE, (c) MAE, (d) RMSE.
Table 10.
Regression performance comparison with radar.
Fig 13.
Detailed rainfall maps of the rainfall products.
Table 11.
RMSE values of the different rainfall products compared with radar in the five rainfall events.
Table 12.
Rainfall regression performance of the rainfall products using the different classification architectures.
Table 13.
Performance of rainfall estimation using the different frameworks.
Fig 14.
Classification results of models M1, M2, M3, and M4, with balanced and imbalanced data.
Table 14.
Performance of rainfall estimation using the different input feature sets.
Table 15.
Performance of rainfall estimation using the different input feature sets.