CCA Forecast for Sahel Rainfall in Jul-Aug-Sep 1997

contributed by Wassila Thiao and Anthony Barnston

Climate Prediction Center, NOAA, Camp Springs, Maryland

Severe and recurrent rainfall deficits across the African continent during the past two to three decades have been detrimental to the economy of the African nations. Thus, policy makers and funding agencies often face tough challenges to make relief plans. There clearly is a need for forecasts of short-term climate fluctuations, such as for seasonal total rainfall one or more seasons in advance. The African Desk, established at the Climate Prediction Center (CPC) of the NWS/NCEP, has been experimenting with African seasonal forecasting in collaboration with the CPC. While numerical approaches are being considered, work so far has focused more on statistical methods. Here we apply canonical correlation analysis (CCA) to produce an experimental forecast for rainfall anomalies in the Sahel region of northern tropical Africa (10-25N, 18W-45E) for the Jul-Aug-Sep 1997 period. The boreal summer months comprise the climato-logically rainy season in the Sahel, as this is when the highest temperatures and atmospheric moisture content (i.e. the ITCZ) occur at these tropical regions of the Northern Hemisphere.

The CCA method is a multivariate regression that relates patterns in predictor fields to patterns in the predictand field. The prediction design used here is the same as that of the CCA used as one of the tools for operational climate prediction in the U.S. (Barnston 1994), based on earlier work of Barnett and Preisendorfer (1987). Four consecutive 3-month predictor periods are followed by a lead time and then a single 3-month predictand, or target, period. Forecast skill experiments have indicated that the global SST field serves best as a predictor. While additional fields such as upper air geopotential height, tropical low-level wind or outgoing longwave radiation might well enhance skill farther, data sets of these fields do not extend far enough into the past to satisfy the CCA's need for a long-term (e.g. at least 25-year) data record from which to identify the dominant relationships. The predictor and predictand data sets used here begin in 1955. For the 1997 Sahel rainfall prediction shown below, the predictor data are the global SST anomaly field over the four 3-month periods of Jun-Jul-Aug 1996, Sep-Oct-Nov 1996, and Dec-Jan-Feb 1996-97, and Mar-Apr-May 1997. Using data from 1955-96, relationships between the prior year's SST anomaly evolution and the target year's Jul-Aug-Sep Sahel 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 1997 boreal summer developed. Here the lead time is 1 month, because the latest predictor data used are those of May 1997, preceding the beginning of the target period by 1 month.

The predictor SST data were derived from a combination of the COADS data (Slutz et al. 1985) and more recent OI data (Smith et al. 1996). The predictand Sahel rainfall data come from the gridded global rainfall data set developed by M. Hulme (Hulme 1994), at 2.5 by 3.75 resolution, resulting in 73 points in the Sahel. A rainfall data set consisting of individual stations has also been tested, with results shown in Thiao et al. (1996) and Barnston et al. (1996). While skill results are roughly similar between the two rainfall data sets for the Sahel because of the sufficient station data density, the gridded data tend to show higher skill in parts of Africa having sparser data. This may be because the gridded data use stations that have gaps during some periods, while the station data set com-pletely excludes such stations. The data smoothing in-herent in forming grid point values may also be helpful.

The diagnostic data produced by CCA indicate that expected skill is modest to moderate in predicting Jul-Aug-Sep precipitation at 1-month lead, with average region wide correlation skill of 0.27, and 0.60 or higher at several locations. A cross-validation design is used in obtaining these skill estimates, where each year is held out of the developmental data set in turn, and then used as the forecast target. As shown in Thiao et al. (1996) and Barnston et al. (1996), the expected area-average skill in forecasting the Sahel decreases slowly from the low 0.30s as the lead time is increased from 0 to 4 months, more rapidly from 4 to 7 months lead, and then remains near 0.2 for 7 months out to a year lead. This has favorable implications for moderately long lead forecasts such as the one shown 3 months ago in the March 1997 issue of this Bulletin.

The spatial loading patterns of the leading CCA modes suggest two major sources of skill in the SST field, each expressed as a separate CCA mode. The first is an interdecadal trend toward warmer SST in the Indian Ocean, the extratropical South Atlantic and the eastern tropical and Southern Hemisphere Pacific. Warming of the SST in those regions has been associated with a decrease in the Jul-Aug-Sep Sahel precipitation, especially between the early 1970s and the mid-1980s. The central north Pacific SST has tended to cool during this same period. The participation of the eastern tropical Pacific and oppositely-signed central North Pacific SST in this predictor pattern implies a presence of ENSO in the relationship with Sahel rainfall: during warm (cold) episodes rainfall is lighter (heavier). However, ENSO does not appear to be the dominating aspect of the relationship. During the past two years we have a pattern somewhat opposite of the recent trend; i.e. without the anomalous SST pattern that has typified the most recent 15-20 years as a whole. The occurrence of a cold ENSO episode has undoubtedly contributed to this condition. However, another CCA run predicting SST for boreal summer 1997 shows a return to an SST pattern more typical of the most recent 1 to 2 decades, including a warm ENSO condition. This expectation, implicit in the Sahel rainfall CCA, is helping to push the current Sahel rainfall forecast toward slightly drier than normal conditions for some sectors for summer 1997.

Another source of skill for Jul-Aug-Sep rainfall at 1-month lead is an interhemispheric difference in SST anomaly, particularly in the Atlantic. Warm Southern Hemisphere SST tends to be associated with below normal Sahel rainfall. This year the northern (southern) subtropical Atlantic has positive (negative) SST anomalies, leading to a rainier forecast for specific sectors of the Sahel. Although There are additional modes (totaling 6) included in the CCA and the weaker modes add incrementally to the cross-validated skill, the physical relationships encompassed by those modes are multiple and complex, and thus difficult to identify.

The resulting forecast for Jul-Aug-Sep 1997 is shown in Fig. 1. Mainly near to slightly above normal rainfall is predicted for the western and central Sahel, while near to somewhat below normal rainfall is predicted for the eastern Sahel. Figure 2 shows the geographical distribution of expected skill for the forecast, expressed as a temporal correlation coefficient between forecasts and observations over the historical period of study. While the regional mean correlation skill is 0.27, estimated skills exceed 0.60 in central Mali, at the 3-way border of Mali/Senegal /Guinea, and at the southern border of Chad/Sudan. The field significance of the skill map, computed using Monte Carlo randomizations (indicating the probability that this skill map could have occurred by chance), is less than 0.001; i.e. there is virtually no chance that these levels of skill over this region are accidental.

Where the skill is estimated at 0.50 or higher, confidence in this forecast can be regarded as at least moderate. Where skill is below 0.50 but at least 0.30, confidence is modest but the forecast is still usable. The lower the expected skill, the more the CCA tends to damp forecast amplitudes toward the mean. When skill is at least moderate, a forecast for near-average conditions (denoted with "N" in Fig. 1) reflects neutral or contradictory predictor patterns rather than a damping toward the mean due to low skill. In this case, a shift in the probability distribution away from the two extremes, toward normal, is indicated, just as a shift toward the dry or wet extreme is indicated by a positive or negative forecast without an "N".

Note: The anomalies presented here are expressed relative to the 1955-95 base period. Because of secular changes in SST and Sahel rainfall, this may lead to conflicting anomalies with those cited in studies that use different base periods.

Barnston, A.G., W. Thiao and V. Kumar, 1996: Long-lead forecasts of seasonal precipitation in Africa using CCA. Wea. Forecasting, 11, 506-520.

Hulme, M., 1994: Validation of large-scale precipitation fields in general circulation models. In Global Precipitation and Climate Change, M. Desbois and F. Desalmand, Ed., NATO ASI Series, Springer-Verlag, Berlin, 466 pp.

Slutz, R., S.J. Lubler, J.D. Hiscox, S.D. Woodruff, R.J. Jenne, D.H. Joseph, P.M. Steurer, and J.D. Elius, 1985: Comprehensive Ocean Atmosphere Data Set. NOAA, Boulder, CO, 268 pp. [Available from Climate Research Program, ERL, R/E/AR6, 325 Broadway, Boulder, CO 80303.]

Smith, T.M., R.W. Reynolds, and C.F. Ropelewski, 1994: Optimal averaging of seasonal sea surface temperatures and associated confidence intervals (1860-1989). J. Climate, 7, 949-964.

Thiao, W., A.G. Barnston and V. Kumar, 1996: Teleconnections and seasonal rainfall prediction in Africa. Proceedings of the 20th Annual Climate Diagnostics Workshop, Seattle, Washington, October 23-27, 1995, 413-416.

Fig. 1. The CCA-based rainfall probability anomaly forecast for the Sahel region of northern tropical Africa for Jul-Aug-Sep 1997. 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. An "N" refers to an enhanced "near normal" probability: e.g., "N2" indicates .323, .343, .323. A "0" indicates .333, .333, .333 (i.e., climatological probabilities, or no useful forecast information available).

Fig. 2. Geographical distribution of the cross-validated estimate of expected skill for the rainfall forecast shown in Fig. 1. Skill is expressed as a correlation between predictions and observations over an extended (~40-year) period.

[Previous Article] [Next Article]