The precipitation in January–February 2022 in southern China was the second-largest amount since 1961. Anthropogenic influence reduced the likelihood of extreme events like 2022 by about 50% (55%) in HadGEM3 (CMIP6).

According to HadGEM3 (CMIP6) models, anthropogenic forcing reduced the probability of 2022-like June mean precipitation by about 32% (15%) and increased 5-day rainfall extreme probability by about 1.8 (1.3) times.

The research and application of convolutional neural networks (CNNs) on statistical downscaling have been hampered by the fact that deep learning is highly dependent on sample size and is considered to be a black-box model. Therefore, a CNN model with transfer learning (CNN-TL) is proposed to study the pre-rainy season precipitation of South China. First, an augmented monthly dataset is created by sliding a fixed-length window over the daily circulation field and precipitation data for the entire year. Next, a base CNN network is pretrained on the augmented dataset, and then the network parameters are tuned on the actual monthly dataset from South China. Finally, guided backpropagation is conducted to obtain the distribution regions of the key features and explain the net. The coefficient of determination (R²) and root mean square error (RMSE) show that the CNN-TL model has higher explanatory power and better fitting performance than the feature extraction-based random forest (RF). Compared with the base CNN, the transfer learning approach can improve the explanatory power of the model by 10.29% and reduce the average RMSE by 6.82%. In addition, the interpretation results of the model show that the critical regions are primarily South China and its surrounding areas, including the Indochinese Peninsula, the Bay of Bengal, and the South China Sea. Furthermore, the ablation experiments and composite analysis illustrate that these regions are very important. The research and application of convolutional neural networks (CNNs) on statistical downscaling have been hampered by the fact that deep learning is highly dependent on sample size and is considered to be a black-box model. Therefore, a CNN model with transfer learning (CNN-TL) is proposed to study the pre-rainy season precipitation of South China. First, an augmented monthly dataset is created by sliding a fixed-length window over the daily circulation field and precipitation data for the entire year. Next, a base CNN network is pretrained on the augmented dataset, and then the network parameters are tuned on the actual monthly dataset from South China. Finally, guided backpropagation is conducted to obtain the distribution regions of the key features and explain the net. The coefficient of determination (R²) and root mean square error (RMSE) show that the CNN-TL model has higher explanatory power and better fitting performance than the feature extraction-based random forest (RF). Compared with the base CNN, the transfer learning approach can improve the explanatory power of the model by 10.29% and reduce the average RMSE by 6.82%. In addition, the interpretation results of the model show that the critical regions are primarily South China and its surrounding areas, including the Indochinese Peninsula, the Bay of Bengal, and the South China Sea. Furthermore, the ablation experiments and composite analysis illustrate that these regions are very important.

The first rainy season in South China (SC) is a concentrated precipitation season, however, it is unclear that in such concentrated precipitation season, characteristics of heavy precipitation concentration degree in SC during the past 50 years. Based on daily precipitation observation data of 192 stations in SC during the first rainy season (April-June) for 1965–2016, the study investigated and analyzed the temporal and spatial variations of the heavy precipitation concentration index (Q index for short). Results showed that the intensity of heavy precipitation over most regions of SC exhibited an increasing trend. The multi-year mean Q index of the heavy precipitation at different stations varied between 0.3 and 0.6, with an average value of 0.375. The heavy precipitation has a concentration trend in the SC inland and western mountain areas and a dispersion trend in the SC coastal area. Notably, both heavy precipitation and the Q index in most inland and western mountain areas showed an increasing trend, denoting a higher flooding risk there. On the other hand, for the duration of heavy precipitation, the 2-day duration of heavy rainfalls had an increasing trend in inland, western Mountain and East Guangdong. The 3-day duration of rainfalls tended to increase steadily in coastal areas and East Guangzhou, and the 4-day duration of rainfalls increased only in the coastal region. In conclusion, the shorter duration heavy precipitation decreased, while the longer ones increased especially in the coast and East Guangdong, which can contribute to the decrease in the Q index but can not denote a decrease in flooding risks. The possible cause may be attributed to the long-duration heavy rainfall that happened in the West Mountain, East Guangdong and Southeast Coast. On the other hand, the frequency and intensity of ongoing heavy precipitation events in coastal cities of South China have increased significantly.