Statistical recalibration of GCM forecasts over southern Africa using model output statistics

Publication Status is "Submitted" Or "In Press: 
LDEO Publication: 
Publication Type: 
Year of Publication: 
Journal Title: 
Journal of Climate
Journal Date: 
Place Published: 
Tertiary Title: 
Section / Start page: 
ISBN Number: 
ISSN Number: 
Short Title: 
Accession Number: 
LDEO Publication Number: 
Call Number: 

A technique for producing regional rainfall forecasts for southern Africa is developed that statistically maps or "recalibrates'' large-scale circulation features produced by the ECHAM3.6 general circulation model (GCM) to observed regional rainfall for the December-February (DJF) season. The recalibration technique, model output statistics (MOS), relates archived records of GCM fields to observed DJF rainfall through a set of canonical correlation analysis (CCA) equations. After screening several potential predictor fields, the 850-hPa geopotential height field is selected as the single predictor field in the CCA equations that is subsequently used to produce MOS-recalibrated rainfall patterns. The recalibrated forecasts outscore area-averaged GCM-simulated rainfall anomalies, as well as forecasts produced using a simple linear forecast model. The MOS recalibration is applied to two sets of GCM experiments: for the "simulation'' experiment, simultaneous observed sea surface temperature (SST) serves as the lower boundary forcing; for the "hindcast'' experiment, the prescribed SSTs are obtained by persisting the previous month's SST anomaly through the forecast period. Pattern analyses performed on the predictor-predictand pairs confirm a robust relationship between the GCM 850-hPa height fields and the rainfall fields. The structure and variability of the large-scale circulation is well characterized by the GCM in both simulation and hindcast mode. Measures of retroactive skill for a 9-yr independent period (1991/92-1999/2000) using the hindcast MOS are obtained for both deterministic and probabilistic forecasts, suggesting that a probabilistic representation of MOS forecasts is potentially more valuable. Finally, MOS is employed to investigate its potential to downscale the GCM large-scale circulation to more specific forecasts of land surface characteristics such as streamflow.


576LRTimes Cited:21Cited References Count:48