Introduction

Although seasonal forecast applications are still in an early stage of development there is now enough collective experience from research efforts around the world to induce some meaningful considerations.

In particular, for West Africa, whose economy is mainly dependent on agricultural sector, the possibility of having seasonal predictions for farm level and food early warning system applications is very useful. Due in part to its interdisciplinary nature, the literature on agricultural applications of seasonal forecasts is scattered. A few collected works (Sivakumar 2000; IRI 2000) cover efforts across countries.

Concerning meteorological seasonal predictions in the following pages, some of the main methods and data over the world are summarized:

■ At the Hadley Centre of UK Met Office outlooks for temperature and rainfall up to 6 months ahead for all regions of the globe, updated shortly after the middle of each month are available. These forecasts are based on an empirical model using multiple linear regression (MLR) and linear discriminant analysis (LDA) and using as input sea surface temperature anomalies (SSTAs) representing interhemispheric contrast, and tropical Atlantic and El Nino-Southern Oscillation (ENSO) signals (Folland et al. 1991).

■ At the Climate Prediction Center (CPC) at NCEP a canonical correlation analysis is conducted for the African continent using quasi-global SST data for the area 40° S to 60° N (Barnston et al.1996).

■ At the Prediction Group of the Colorado State University an empirical method is developed using as inputs previous Sahel precipitation, tropical North and South Atlantic SSTs, Pacific SSTs and Quasi-Biennial Oscillation (QBO) (Landsea et al. 1993).

Concerning Numerical methods, studies are going on at the following centers:

■ United Kingdom Met Office: an ensemble of runs by UKMO's Third Generation Coupled Ocean-Atmosphere GCM (HadAM3) are used with forced SSTAs assumed to be persistent through the forecast period.

■ European Centre for Medium-Range Weather Forecasting (ECMWF): a fully coupled ocean-atmosphere model produces global predictions each month for precipitation and rainfall. Atmospheric and land surface conditions come from ECMWF operational systems. Other oceanic thermal input data come from established observation networks (Stockdale et al. 1998).

■ International Research Institute for Climate Prediction (IRI) at Columbia University: a multi-ensemble approach using different AGCMs generates global forecasts; Pacific SSTA forecast (from the NCEP model) and other SSTAs from statistical techniques are used. Much information can be found at http://iri.ldgo.columbia.edu/ climate/forecast/.

Obviously skills for these long-range predictions are substantially lower than for the more familiar shorter-range predictions; predictive skill and detailed comparisons between the numerical methods and statistical ones are rare (Goddard et al. 2001). For West Africa, the seasonal rainfall forecasting skill of statistical methods (in particular the Climate Prediction Center method using CCA) is notable (CLIVAR 1998) and is largely due to the persistence of successive rainfall anomalies.

Anyway Goddard et al. (2001) point out that different techniques have different performances according to the region of interest and that end user interactions play a part. For this reason, regional climate outlook forums are needed to keep in touch policy makers, funding agencies, and users of climate information. Basically the role of climate forum is to discuss the current state of the global and regional climate, to produce a consensus seasonal forecast in the region in question and to develop a mitigation plan based on the seasonal outlook.

Was this article helpful?

0 0

Post a comment