Pattern analysis and recalibration of a perfectly forced atmospheric general circulation model

Abstract

Empirical techniques are developed to adjust dynamic model forecasts on the seasonal time scale for southern African summer rainfall. The techniques, called perfect prognosis and model output statistics (MOS), are utilized to statistically "recalibrate" general circulation model (GCM) large-scale fields to three equi-probable rainfall categories for December to February. The recalibration is applied to a GCM experiment where simultaneously observed sea-surface temperature (SST) fields serve as the lower boundary forcing, referred to as the simulation mode experiment. Cross-validation sensitivity tests are first performed over a 28-year climate period to design an optimal canonical correlation analysis (CCA) model for each of the two recalibration methods. After considering several potential predictor fields, the 700 hPa geopotential height field is selected as the single predictor field in the two sets of statistical equations that are subsequently used to produce recalibrated rainfall simulations over a 1 a-year independent test period. Patterns analysis of the predictor and predictand fields suggests that anomalously low (high) 700 hPa geopotential heights over the subcontinent are associated with wet (dry) conditions over land, an association that is supported by observational evidence of rain (drought) producing systems. Additionally, the dominant mode of the recalibration equations is associated with the EI Nino/Southern Oscillation (ENSO) phenomenon. Somewhat higher retro-active skill levels are found using the MOS technique, but the computationally less intensive perfect prognosis technique should also be able to produce usable seasonal rainfall forecasts over southern Africa in an operational forecast environment hampered by the lack of computing resources.