[Previous Article] [Next Article]
CCA Forecast for Southern African Rainfall in Jan-Feb-Mar 1998
contributed by Wassila Thiaw* and Anthony Barnston
Climate Prediction Center, NOAA, Camp Springs, Maryland
*Name has been changed from Thiao to Thiaw
Because of the severity and recurrence of rainfall deficits across the African continent in recent
decades, policy makers and funding agencies have faced tough challenges to make relief plans.
There clearly is a need for long-lead forecasts of seasonal rainfall anomaly fluctuations. The
African Desk, established at the Cli-mate Prediction Center (CPC) of the NWS/NCEP, has been
developing African seasonal forecasting capability in collaboration with the CPC. While numerical
ap-proaches 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 southern Africa (10-35S, 10-50E) for the Jan-Feb-Mar 1998 rainy season.
CCA is a multivariate regression that relates predictor field patterns to predictand field patterns.
In the prediction design used here, four consecutive 3-month predictor periods are followed by a
lead time and then a single 3-month predictand (target) period. Fore-cast skill experiments
indicate that the global SST serves best as a predictor. While other fields might fur-ther enhance
skill, their data sets for the Tropics and Southern Hemisphere are only now in the process of
extending far enough into the past to provide a long-term history for CCA. The SST and rainfall
data sets used here begin in 1955. For this forecast, SST pre-dictors span the four 3-month
periods of DJF, MAM, JJA, and SON 1997. Historical relationships between the prior year's SST
anomaly evolution and the target year's Jan-Feb-Mar Southern Africa rainfall anomaly patterns are
linearly modeled by the CCA. The current year's SSTs are then projected onto these relationships
and a forecast for Jan-Feb-Mar 1998 is developed. The lead time of 1 month is the time by which
the latest predictor data (Nov) precede the target period onset.
The predictor SST data were derived from a com-bination of the COADS data and more recent
OI data (Smith et al. 1994). The predictand rainfall data come from the Hulme (1994) gridded
global rainfall data, providing 58 grid points in Southern Africa. These rainfall data yield slightly
higher skills than those using station rainfalls, the latter shown in Thiao et al. (1996) and Barnston
et al. (1996).
The CCA diagnostics indicate that expected skill averages 0.23 over southern Africa for the
Jan-Feb-Mar precipitation forecast (at 1-month lead) shown below, with highest skill near 0.56 in
Namibia. Cross-validation using the sequential "1-year held out" method is used in obtaining the
skill estimates.
While 6 CCA modes are used in this prediction, the leading 2 or 3 modes account for most of the
skill. The CCA diagnostics indicate that the ENSO state is the most important factor, as El Niño
tends to be associated with negative rainfall anomalies in much of southern Africa. A nearly
equally important factor here is the globaltropical SST anomaly (emphasizing the Indian Ocean,
but also including the ENSO state). On both counts, the current El Niño is likely to result in
suppressed rainfall in most of Southern Africa in JFM.
The forecast shown below (Fig. 1) is expressed as probability anomalies with respect to
climatological probabilities for three equiprobable categories: below, near, and above normal (B,
N, and A, respectively). The climatological probability of each category is 1/3. The predictand
grid points are shown by the locations of the numbers on the map. The forecast gives the
deviation from these probabilities for Jan-Feb-Mar 1998, with respect to the above normal class.
For example, a -0.08 probability anomaly (shown as -8) implies probabilities of (B, N and A) of
0.413, 0.333 and 0.253 compared to climatological probabilities of o.333, 0.333, 0.333. In some
cases the probability of the near normal category may be elevated at the expense of the two outer
categories. Probabilistic forecasts convey the degree of uncertainty, and assume that the users
have some idea of the rainfall amount intervals defined by the bottom, middle and top thirds of the
distribution. These thirds may be roughly asso-ciated with drought, normal, or plentiful rainfall
cases.
Figure 1 broadly indicates dry conditions for Jan-Feb-Mar 1998 for much of southern Africa.
Even with only modest to moderate skills at many of the grid points, the strength of the predictor
signal for this forecast (the strong ENSO condition) produces a relatively strong tilt of the odds
toward dryness over the region. The prediction favors below normal rainfall in the western
half of southern Africa, including Namibia, Botswana, Angola, and Zambia. Low skills for
the eastern part of the regions, including Zimbabwe, Mozambique and South Africa, prevented a
definitive forecast from being issued there, resulting in a climatology forecast.
Statistically, while some weakening is possible, the ENSO conditions found in
November-December are likely to persist through at least January-February, implying a
continuation of current ENSO conditions into the rainy season for southern Africa.
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.
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 forecasts for the southern African region for
Jan-Feb-Mar 1998, made at 1 month lead. Values shown are deviations (X100) from
climatological probabilities for the above normal category. For example, 3 implies (B,N,A)
probabilities of 0.303, 0.333, 0.363 (a weak tilt toward wetness), while -7 implies 0.403, 0.333,
0.263 (a somewhat stronger tilt toward drought). See text for further details.