Predictability of Dry Season Reforecasts over the Tropical South American Region

Atmospheric conditions during the dry season of the South American monsoon are instrumental in the initiation of convection during the wet season and are strongly correlated to SSTs within the Atlantic Warm Pool. Subsequently, accurate seasonal prediction of temperature and rainfall during the dry season has the potential to improve our understanding of and the predictability of these variables during future seasons. In this study, we review the fidelity of South American dry season (June-July-August) reforecasts from one global climate model (GCM), and one downscaled regional climate model (RCM). Additionally, we evaluate a second integration of the RCM that uses a bias correction method called anomaly nesting, which is designed to remove the bias of the GCM before the downscaling process is performed. The models are integrated for seven dry seasons (2001–2007), and each season consists of six ensemble members. For this study, we focus on two primary regions: the Amazon River Basin (ARB) and the subtropical region (ST). There are three objectives of this research. The first is to locate regions of model bias for two-meter air temperature and for precipitation within the ARB and the ST using NCEP Climate Forecast System Reanalysis (CFSR) as a comparison dataset. The second is to evaluate the predictability of above normal, normal, and below normal occurrences of the two variables using potential predictability ratios and calculations of the area under the relative operative characteristic (ROC) curve (AUC). Through this analysis we should be able to determine whether downscaling or anomaly nesting improve upon the skill of the GCM. Lastly we wish to evaluate how the three models depict land-atmosphere interactions during the dry season and compare their results with results from CFSR. The models produced the largest biases of both variables over elevated terrain and within the Intertropical Convergence Zone (ITCZ). However, neither of these locations significantly impacts the ARB or the ST. Signal-to-noise ratios show that the ARB exhibits more potential predictability than the ST and that temperature exhibits more potential predictability than precipitation. AUCs confirm that temperature is more skillfully predicted than precipitation and that the models exhibit more skill in the ARB than in the ST. AUCs show that the downscaled and the downscaled with anomaly nesting integrations display more skill than the GCM integration, particularly in the ARB. Lastly, we find conflicting results between the models and CFSR regarding how the land and the atmosphere interact during the dry season. However, a full moisture budget analysis is needed to completely resolve land-atmosphere feedbacks and that is beyond the scope of this study.