CCA Prediction for Eastern Africa Rainfall in Oct-Nov-Dec 1998 At one Month Lead-Time

contributed by Wassila Thiaw and Anthony Barnston

Climate Prediction Center/NCEP/NWS/NOAA, Washington DC, 20233

1. Methodology

CCA is a multivariate regression model that relates patterns in the predictor fields to patterns in the predictand field. The prediction scheme is designed such that four consecutive 3-month predictor periods are followed by a lead time and then a single 3-month predictand, or target period. The predictor SST data were derived from a combination of the COADS data and more recent OI data. The predictand Eastern Africa rainfall data come from the gridded global rainfall data set developed by M. Hulme at 2.5^olat. by 3.75^olon. resolution, resulting in 32 points in the Eastern Africa region. For the Oct.-Nov.-Dec. 1998 East Africa rainfall prediction discussed above, the predictor data are the global SST anomaly field over the four 3-month periods of Sep.-Oct.-Nov. 1997, Dec.-Jan.-Feb. 1997-98, Mar.-Apr.-May, and Jun.-Jul.-Aug. 1998. Using data from 1955-96, relationships between the prior years SST anomaly evolution and the target years Oct.-Nov.-Dec. East Africa rainfall anomaly patterns are linearly modeled by the CCA. The predictor SST data for the current forecast are then projected onto the preferred relationships derived from the past years, and a forecast for Oct.-Nov.-Dec. 1998 developed. Here the lead time is 1 month, because the latest predictor data used are those of August 1998, preceding the beginning of the target period by 1 month. The diagnostic data produced by CCA indicate that expected skill is low to modest in predicting Oct.-Nov.-Dec. East Africa rainfall. Correlations between observed and predicted rainfall ranged between 0.1 and 0.4. A cross validation design is used in obtaining these skill estimates, where each year of the time series is held out of the developmental data set in turn, and then used as the forecast target.

2. Outlook

Eastern Africa is defined here as the region between 10N and 10S; 25-50E. In this forecast, quasi-global SSTs between 40S and 60N at a resolution of 10 by 10 lat.-lon. are used to predict East African rainfall. The predictions (Figure 1) are expressed in terms of departures from climatological probabilities of 3 equi-probable categories of below, near, and above normal rainfall. The climatological probability of each category is 0.333. Positive (negative) departures from the climatological probability indicates a shift toward wet (dry) extremes. For instance a departure of 0.05 from the climatological probability indicates probabilities of 0.38 and 0.28 for above and below normal rainfall, respectively. For the sake of simplicity, the probability for the normal category remains at 0.33. When skill is low, climatology (denoted with 0) is suggested. The diagnostic data produced by CCA indicate that the current moderately cold episode of ENSO has a weak influence on Oct.-Nov.-Dec. East African rainfall. However, cold SSTs over the equatorial central Pacific tend to be associated with below normal rainfall over East Africa, particularly south of 5S. The prediction for Oct.-Nov.-Dec. 1998 calls for slightly higher than climatological probabilities for below normal rainfall along the coast from Tanzania northward to Somalia, and across Kenya and southern Sudan. Near-to-above normal rainfall is expected over southern Uganda. Climatology is indicated elsewhere.

Figure 1. The CCA-based rainfall probability anomaly forecast for Eastern Africa for Oct.-Nov.-Dec. 1998. Probability anomalies (X100) are with respect to the "above normal" rainfall tercile: "2" indicates probabilities of .313, .333, .353 for the below, near and above normal terciles, respectively; "-7" indicates .403, .333, .263; "N2" and "0" indicates .333, .333, .333 (i.e., climatological probabilities, or no useful forecast in information available).