Scholarlink

<p>Summer precipitation over South China exhibits pronounced interannual variability, leading to frequent droughts and floods. Skillful prediction of precipitation anomalies remains highly challenging. To address the prediction difficulties of summer precipitation in South China, this study utilizes slow feature analysis (SFA) to directly extract slow-varying driving force signals from observed precipitation series as predictors. By combining empirical orthogonal function analysis and multiple linear regression methods, we develop a new statistical model for predicting summer precipitation over South China. During the independent validation period from 2011 to 2022, the SFA-based climate prediction (SCP) method achieved a mean temporal correlation coefficient of 0.36, substantially higher than those of existing dynamical models (0.13 for BCC-CSM1.1 and 0.08 for CFSv2) and a hybrid dynamical-statistical model (0.19 for FODAS). Moreover, in 10 out of 12 years of independent forecasts, SCP successfully predicts the sign of regional mean precipitation anomaly percentage over South China. The correlation coefficient between SCP and observations is 0.62, compared with 0.45, 0.28, and 0.17 for FODAS, BCC-CSM1.1, and CFSv2, respectively. These results indicate that SCP outperforms two dynamical models (BCC-CSM1.1 and CFSv2) and the hybrid dynamical-statistical model (FODAS) in predicting summer precipitation over South China. The proposed method provides a new pathway to improve seasonal prediction skill and provides key scientific and technological support for meteorological disaster prevention, mitigation, and decision-making services.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A new method for predicting summer precipitation over South China based on slow feature analysis

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Siyi Wang,
  • Jingyan Guo,
  • Wenping He

摘要

Summer precipitation over South China exhibits pronounced interannual variability, leading to frequent droughts and floods. Skillful prediction of precipitation anomalies remains highly challenging. To address the prediction difficulties of summer precipitation in South China, this study utilizes slow feature analysis (SFA) to directly extract slow-varying driving force signals from observed precipitation series as predictors. By combining empirical orthogonal function analysis and multiple linear regression methods, we develop a new statistical model for predicting summer precipitation over South China. During the independent validation period from 2011 to 2022, the SFA-based climate prediction (SCP) method achieved a mean temporal correlation coefficient of 0.36, substantially higher than those of existing dynamical models (0.13 for BCC-CSM1.1 and 0.08 for CFSv2) and a hybrid dynamical-statistical model (0.19 for FODAS). Moreover, in 10 out of 12 years of independent forecasts, SCP successfully predicts the sign of regional mean precipitation anomaly percentage over South China. The correlation coefficient between SCP and observations is 0.62, compared with 0.45, 0.28, and 0.17 for FODAS, BCC-CSM1.1, and CFSv2, respectively. These results indicate that SCP outperforms two dynamical models (BCC-CSM1.1 and CFSv2) and the hybrid dynamical-statistical model (FODAS) in predicting summer precipitation over South China. The proposed method provides a new pathway to improve seasonal prediction skill and provides key scientific and technological support for meteorological disaster prevention, mitigation, and decision-making services.