Droughts in Amazonia: Spatiotemporal Variability, Teleconnections, and Seasonal Predictions. Issue 12 (29th December 2017)

• Record Type: Journal Article
• Authors: Lima, Carlos H. R.; AghaKouchak, Amir
• Is Part Of: Water resources research. Volume 53:Issue 12(2017)
• Page Start: 10824
• Page End: 10840
• Publication Date: 2017-12-29
• Subjects: drought -- prediction -- Amazonia -- sparse canonical correlation analysis; Hydrology -- Periodicals; 333.91
• DOI: 10.1002/2016WR020086
• Languages: English
• ISSNs: 0043-1397

Abstract

Abstract: Most Amazonia drought studies have focused on rainfall deficits and their impact on river discharges, while the analysis of other important driver variables, such as temperature and soil moisture, has attracted less attention. Here we try to better understand the spatiotemporal dynamics of Amazonia droughts and associated climate teleconnections as characterized by the Palmer Drought Severity Index (PDSI), which integrates information from rainfall deficit, temperature anomalies, and soil moisture capacity. The results reveal that Amazonia droughts are most related to one dominant pattern across the entire region, followed by two seesaw kind of patterns: north‐south and east‐west. The main two modes are correlated with sea surface temperature (SST) anomalies in the tropical Pacific and Atlantic oceans. The teleconnections associated with global SST are then used to build a seasonal forecast model for PDSI over Amazonia based on predictors obtained from a sparse canonical correlation analysis approach. A unique feature of the presented drought prediction method is using only a few number of predictors to avoid excessive noise in the predictor space. Cross‐validated results show correlations between observed and predicted spatial average PDSI up to 0.60 and 0.45 for lead times of 5 and 9 months, respectively. To the best of our knowledge, this is the first study in the region that, based on cross‐validation results, leads to appreciable forecast skills for lead times beyond 4 months. This is a step forward in better understanding the dynamics of Amazonia droughts and improving risk assessment and management, through improved drought forecasting. Key Points: Spatiotemporal dynamics and teleconnections associated with Amazonia droughts are investigated based on the PDSI indices A drought forecast model for Amazonia is developed and tested based on the global SST field and sparse canonical correlation analysis This is the first study in the region that, based on cross‐validation, leads to appreciable forecast skills for lead times beyond 4 months