Abstract Interest in region-specific assessments of droughts and the need to optimise water resources planning and allocation on a local scale via additional investments in water infrastructures is emerging as novel management initiatives to build drought resilience. In this study, a novel two-step regularization procedure that combines statistical rotation with support vector machine regression (SVMR) is employed to assess and identify hydrological regions in Brazil associated with global climate teleconnection patterns (e.g., ENSO, PDO, etc.). To enhance realistic drought impact assessments, region-specific attributes of drought and the climate modes associated with its variability and characteristics are studied using standardised precipitation index (SPI) and reanalysis data (MERRA). Compared to other regions, results show that drought variability and its occurrence are relatively higher in the extreme north, north-east, and south of Brazil. The predominance of extreme drought events shows that more than 50% of Brazil was affected by the 1998/1999 drought while areas under droughts in recent times fluctuated between 25% in 2012 and 70% in 2015. Results also show significant association of ENSO (e.g., R2 = 28%) and PDO (e.g., R2 = 18%) with drought indicators in several climatic hotspots. The synthesis of climate modes as predictors of droughts in the SVMR scheme highlights the influence and importance of the Pacific and Atlantic oceans on drought evolutions in Brazil. The MERRA-derived drought indicator extracted this influence better (e.g., r = 0.72) than the SPI and appears to be a more suitable drought metric to understand the impacts of global climate on extreme events in the region.

中文翻译：

---

巴西气候影响的水文热点：两步正则化方法

摘要 对特定区域干旱评估的兴趣以及通过对水基础设施进行额外投资来优化当地水资源规划和分配的必要性正在成为建立抗旱能力的新型管理举措。在这项研究中，一种新的两步正则化程序将统计旋转与支持向量机回归 (SVMR) 相结合，用于评估和识别巴西与全球气候遥相关模式（例如，ENSO、PDO 等）相关的水文区域。为了增强现实的干旱影响评估，使用标准化降水指数 (SPI) 和再分析数据 (MERRA) 研究干旱的特定区域属性以及与其变异性和特征相关的气候模式。与其他地区相比，结果表明，巴西极北部、东北部和南部的干旱变异性及其发生率相对较高。极端干旱事件的主导地位表明，巴西 50% 以上的地区受到 1998/1999 年干旱的影响，而近期干旱地区在 2012 年的 25% 和 2015 年的 70% 之间波动。结果还显示了 ENSO（例如, R2 = 28%) 和 PDO（例如，R2 = 18%）以及几个气候热点的干旱指标。SVMR 方案中作为干旱预测因子的气候模式的综合突出了太平洋和大西洋对巴西干旱演变的影响和重要性。MERRA 衍生的干旱指标更好地提取了这种影响（例如，r = 0。