An assessment of the value of

The Met Office is conducting research into the effects of sea surface temperatures and other climatic variables on tropical rainfall. As part of this research, experimental forecasts of seasonal rainfall for the Sahel (region 1 in figure 1a) have been made for each year from 1986 onwards. Since 1992, forecasts of seasonal rainfall have also been made for a slightly redefined rectangular Sahel (region 2, 15W to 37.5E and 12.5N to 17.5N), for an area south of the Sahel (region 3, 7.5W to 33.75E, 10N to 12.5N), and for an area extending further south to the coast (region 4, approximately 7.5W to 7.5E, 5N to 10N). The four regions are labelled in figure 1a.

The forecast period for regions 2-4 is July-September. For region 1, annual rainfall is predicted, though most of the rain in this region falls during July-September. Rainfall predictions are expressed using 5 (quint) categories which are equi-probable over 1961-1990 and presented in terms of probabilities for each category and a deterministic “best estimate” category. The 5 quints are referred to as Very Dry, Dry, Average, Wet and Very Wet. The boundaries of the quint categories are defined (as percentages of the 1961-1990 average) in Table 1.

TABLE 1

REGION	VERY-DRY / DRY	DRY/ AVERAGE	AVERAGE /WET	WET/ VERY WET
1	75	97	109	121
2	81	93	102	117
3	88	99	104	112
4	82	94	106	115

2. Forecast method

Forecasts are generated using a combination of empirical/statistical predictions and output from the Met Office’s coupled ocean-atmosphere global seasonal prediction model (known as GloSea). The issued forecast is a weighted average of forecasts from the empirical and dynamical methods – where weighting is determined according to the skill of the method. In addition a persistence forecast (last year’s observed seasonal rainfall) is also included in the weighted average.

2.1 Statistical methods

A number of statistical methods are employed each using as input observed May and June Sea Surface Temperature (SST) anomaly patterns. Further details of the methods are documented in Folland et al.,1991, Journal of Forecasting , Vol.10,21-56.

The SST predictor indices used represent both global-scale anomaly patterns and patterns for a number of regions influential on North Africa. The most important regional patterns are those found in the tropical Pacific and Atlantic regions; the most important global-scale pattern is the contrast in SST anomaly between the northern and southern hemispheres.

The statistical best estimate forecasts are produced by linear regression with SST indices as predictors. Statistical probability forecasts are calculated from the same SST indices using linear discriminant analysis.

2.2 Dynamical Forecasts (GloSea)

The dynamical forecast was produced using the Met Office coupled ocean-atmosphere seasonal prediction model, GloSea - a version of the Hadley Centre climate model (HadCM3). GloSea is run out to 6-months ahead in an ensemble of 40 individual forecasts each initialised with atmospheric conditions and slightly different perturbations of oceanic conditions observed at the beginning of June. Further information about dynamical ensemble forecasts at the Met Office can be found on the Met Office’s website at http://www.metoffice.com/research/index.html

The GloSea forecast is expressed in deterministic and probability format for the 5 quint categories. The deterministic and probabilistic forecasts are calibrated according to past forecast performance using a set of retrospective GloSea forecasts for 1959-2000 (produced as part of the EU DEMETER project see www.ecmwf.int/research/demeter).

2.3 Forecast combination

The forecasts obtained using the statistical methods, GloSea GCM and persistence are weighted to reflect the skill of the different methods. The ratio of weights for the statistical dynamical and persistence forecasts are shown in table 2 . Persistence is not used for the region 4 forecast, as persistence skill is negligible for this region. Dynamical skill is somewhat higher for region 4 than for the other regions hence the higher weights.

Last year, the average category was observed in region 2, and the wet category in regions 1,3 and 4.

TABLE 2: FORECAST WEIGHTS

Region	Statistical	Dynamical	Persistence
1	0.59	0.24	0.17
2	0.56	0.28	0.16
3	0.58	0.26	0.16
4	0.51	0.49	0.00

3. FORECAST

Influence of current SST patterns

SST patterns are similar to those observed in April prior to the previous forecast. Below average SST in SE Atlantic currently favours below average rainfall in the Guinea coast region (region 4) but above average rainfall further north in the Sahel (regions 1 and 2). SST is pre-dominantly below average in the southern Extra-tropics and above average in the northern extratropics. This interhemispheric contrast favours above average rainfall in regions 1, 2 and 3.

Weighted average best estimate forecasts are shown as percentages of the 1961-1990 average in figure 1a. In figure 1b, the forecasts are expressed as percentage standardised units (i.e. standardised values of +100, 0, -100 indicate rainfalls 1 standard deviation above average, average and 1 standard deviation below average respectively) relative to 1961-1990 (NB. 1901-1980 climatology is used for region 1 for compatibility with previous publications by the Met Office and to define the quint category in figure 1c.

This year the statistical, dynamical and persistence forecasts are in good agreement. The averaged statistical forecasts, dynamical forecasts and persistence all favour above average rainfall in regions 1 and 2 and the averaged statistical forecasts, dynamical forecasts favour below average rainfall in region 4. For the intermediate region, region 3, near average rainfall is predicted with the dynamical forecast being drier than the statistical or persistence forecasts.

3.1 Deterministic Forecasts (Fig.1 a-c)

For region 1 and 2 the Very Wet category is favoured, for region 3 the Average Category is favoured, and for region 4 the Very Dry Category is favoured.

3.2 Probability Forecasts (Fig.1 f-j)

The Probability forecasts tell a similar story to the deterministic forecasts. For region 1 and 2 the Very Wet category is most probable and for region 4 the Very Dry category is most probable. For region 3 probabilities indicate little bias towards any category which is not inconsistent with the “Average” deterministic forecast .

3.3. Forecast track record (Fig.1 d-e)

Estimates of the skill of these weighted forecasts are presented in figure1d as correlations between trial forecasts and observed rainfall for the period 1959-2001. This assessment period is slightly different to periods used previously (1951 or earlier to 2000) and is the period for which DEMETER project GloSea retrospective forecasts are available. Unfortunately this new period does not include most of the 1950s when the region went through a very wet period and hence the skill appears lower than before. However the correlations still exceed the 5% significance level for all 4 regions. The Relative Operating Characteristic (ROC) skill in figure 1e is a measure of the performance of the probability forecasts over the period 1959-2001. ROC scores above 60% are considered to indicate significant (5% level) skill.

4. Forecast summary

Our Overall best estimate forecasts are:

Region 1: VERY WET

Region 2: VERY WET

Region 3: AVERAGE

Region 4: VERY DRY

For the 4 regions, our choices of overall best estimate are the same as the deterministic forecast categories which were in good agreement with the probability forecasts.

Note: The Very Wet Category has not been observed in Regions 1 and 2 since 1999 and neither the Dry or the Very Dry category have been observed in region 4 since 2001.

REFERENCES

Folland, C.K., Owen, J., Ward, M.N and Colman, A.W. 1991: Prediction of seasonal rainfall in the Sahel region using empirical and dynamical methods. Journal of Forecasting, 10, 21-56.

Folland, C.K., Parker, D.E., Colman, A.W. and Washington,R. 1999: Large scale modes of Ocean Surface Temperature since the late nineteenth century. In Beyond El Nino, decadal and Interdecadal variability. Ed. A Navarra, Springer pp 75-102.

Nicholson, S.E. 1985: Sub-Saharan rainfall 1981-84. J. Clim. Appl. Met., 24, pp 1388-1391.

Combined Statistical/Dynamical Forecast of 2004 Season Rainfall in the Sahel and Other Regions of Tropical North Africa using Observed Ocean and Atmosphere Information from up to June 2004

Met Office, Exeter, UK

3. FORECAST