-C.W. Landsea, W.M. Gray, P.W. Mielke, Jr., K.J. Berry -Department of Atmospheric Science and Department of Statistics -Colorado State University -Internet: landsea@downdry.atmos.colostate.edu -1 December, 1994The rainy season in North Africa's Sahel occurs almost exclusively during the months of June through September when the ITCZ reaches its farthest northward extension. The Sahel is defined here as the North African region between 10 and 20N. It is this area that has experienced numerous devastating droughts within the last three decades. This report provides a verification of Sahelian seasonal rainfall forecasts issued in early January and early June 1994 as well as presenting an extended range forecast for next year's June to September seasonal forecast for the Sahel based upon data available through late November.
Because of rainfall variability within the region, a homogeneous index of precipitation should not be utilized for the entire Sahel. Instead, three smaller subregions are organized within which precipitation shows similar within season behavior. These regions are the West, Central, and East Sahel. The West Sahel extends from the Atlantic coast to 6W including portions of the countries of Mauritania, Senegal, Gambia, Guinea-Bissau, Guinea, and Mali. The Central Sahel is the region from 6W to 26E and includes parts of Mali, Burkina Faso, Ghana, Togo, Benin, Niger, Nigeria, Cameron, Chad, Central African Republic, and Sudan. The East Sahel reaches from 26E to the Red Sea and is composed of parts of Sudan, Ethiopia, Eritrea, and Djibouti. Note that the Central Sahel is by far the largest of the three regions, comprising well over one half of the entire Sahel.
VERIFICATION OF FORECASTS FOR 1994
Forecasts for the Sahel were issued by the authors on 4 January (Landsea et al. 1994a) and an update released on 7 June (Landsea et al. 1994b). (Copies of these are available from the lead author upon request.) The following are the "official" forecasts of June to September rainfall that we issued on 4 January and 7 June. Note that a standardized deviation value of 0 would indicate normal conditions with respect to the 1950-1991 time frame, a positive value meaning above normal rainfall, a negative value meaning below normal. Also indicated is what quintile the forecast would be within - VERY DRY/DRY/NEUTRAL/WET/VERY WET:
4 January Forecast 7 June Forecast
West Sahel: -0.10 S.D. (NEUTRAL) -0.70 S.D. (DRY)
Central Sahel: -0.20 S.D. (DRY) +0.10 S.D. (WET)
East Sahel: -0.25 S.D. (NEUTRAL) +0.30 S.D. (WET)
Thus the forecasts originally in early January were calling for near normal to slightly drier than normal. The updated forecast in early June called for wetter than normal conditions in the Central and East Sahel, but for contrasting drier than normal conditions in the West Sahel. However, none of these forecasts called for extremely wet or dry conditions to occur. Note that for the last 25 years almost three-quarters have been under DRY to VERY DRY conditions for the Sahel as a whole.
Based upon data received to date (we gratefully acknowledge the data provided by Rich Tinker at the U.S. Climate Analysis Center and Tom Ross at the U.S. National Climate Data Center), our analysis documents that the Sahel region experienced wet conditions in the Central and East and near normal rainfall in the Western Sahel. The following table provides a month- by-month breakdown of the rainfall variability:
Region June July August September | Jun.-Sep.
-----------------------------------------------------------------------------
West -0.29 -0.29 -0.22 -0.03 | -0.24 (NEUTRAL)
(19/4/7) (16/5/8) (16/4/9) (15,5,9) | (13,5,9)
Central -0.36 +0.13 +0.72 +0.90 | +0.62 (WET)
(25/8/8) (15/14/12) (7,9,23) (5,7,28) | (7,11,21)
East -0.32 -0.36 -0.47 +0.38 | +0.39 (WET)
(3,3,0) (9,1,3) (11,2,2) (7,3,5) | (2,1,3)
The top numbers indicate the rainfall anomaly above (positive value) or
below (negative value) the 1950 to 1991 long term average expressed as the
mean number of standard deviations above or below normal based on the mean
anomaly for all available stations. The lower numbers indicate the number
of stations reporting in the region that were below normal/near normal/or
above normal, respectively. Note that the low numbers of stations available
for the East Sahel in June (and thus June to September) make the value a bit
suspect as to its representativeness of the whole region.
Thus the early January forecast correctly forecast the West Sahel rainfall, was one quintile too dry for the East Sahel, and two quintiles too dry for the Central Sahel. The early June forecast correctly altered the forecasts in the Central and East Sahel regions toward the observed values (no quintile errors). The early June West Sahel forecast, while also correctly making the early January forecast drier, went one quintile too dry. To make comparisons versus other years, the following are June to September periods that were most similar to rainfall amounts occurring in 1994:
West Sahel -0.24: 1971:-0.19, 1963:-0.20, 1970:-0.32, 1981:-0.37 Central Sahel +0.62: 1955:+0.57, 1962:+0.59, 1959:+0.70, 1961:+0.72 East Sahel +0.39: 1952:+0.27, 1967:+0.29, 1958:+0.42, 1980:+0.51Two important conclusions from 1994 are that 1) this forecast appears to have been able to anticipate the wetter conditions observed this year than have been seen in the previous years and 2) there is a distinct need to forecast and verify for the three regions separately because of the large differences that occur in observed rainfall. Also of note is that the Western Sahel observed rainfall has a better correspondence to the low amount of Atlantic hurricane activity than the wetter Central Sahel. In Landsea and Gray (1992), we identified that dry years in the Western Sahel usually are accompanied by quiet Atlantic hurricane seasons. For comparison, here are the observed values of June to September rainfall and days of intense hurricane activity from 1988-1994:
1988 1989 1990 1991 1992 1993 | 1994
Intense Hurricane Days: 8.0 10.8 1.0 1.2 3.2 0.8 | 0.0
West Sahel: +0.30 +0.32 -0.83 -0.76 -0.96 -0.66 | -0.24
Central Sahel: +0.11 -0.14 -0.94 -0.18 -0.45 -0.57 | +0.62
East Sahel: +0.52 -0.41 -0.96 -0.56 -0.97 -0.71 | +0.39
STATISTICAL METHODOLOGY AND 1995 FORECAST
The forecasting techniques are detailed in full in Landsea et al. (1993) and Gray et al. (1992, 1993, 1994). This statistical model utilizes the years 1950 to 1991 in a cross-validated (or jackknife) procedure that allows the pseudo-independent testing of individual years to check the skill of the model (for example, the year 1968 was tested by using the years 1950-1967 and 1969-1991 as the developmental data set). We use a Least Absolute Deviations (LAD) regression instead of the traditional Ordinary Least Squares (OLS) multiple regression. LAD is selected over OLS in that LAD is based upon minimizing the absolute differences between predicted and observed instead of the square of that value. Thus outliers do not overly influence the prediction equations. Five predictors are constructed via the LAD regression in a linear model to make the forecast. These five predictors that are suggested to provide long lead skill are the stratospheric Quasi-Biennial Oscillation (QBO) and previous North African rainfall data.
The individual predictors and their values to be used as input for the 1995 forecast are the following:
GROUP 1: U winds at 50 mb (extrapolated for Sep 1995) = 0 m/s
U winds at 30 mb (extrapolated for Sep 1995) = -4 m/s
|U50-U30| (extrapolated for Sep 1995) = 4 m/s
GROUP 2: Aug-Sep 1994 West Sahel rainfall = +0.08 S.D.
Aug-Nov 1994 Gulf of Guinea rainfall = +0.22 S.D.
Given these values for the predictors based upon information available to us at CSU by the end of November, the statistical model predicts values of:
West Sahel ........ +0.34 S.D. (WET QUINTILE) Central Sahel ..... +0.17 S.D. (WET QUINTILE) East Sahel ........ +0.02 S.D. (NEUTRAL QUINTILE)
These moderately wet to normal values of rainfall are due to 1) very favorable conditions for rainfall due to strong west phase of the QBO and small shear between 30 and 50mb; 2) near normal conditions of Aug. - Sep. Western Sahel rainfall indicating near normal rainfall conditions for the Sahel in 1995 based upon persistence; and 3) slightly wetter than normal Aug. - Nov. rainfall along the Gulf of Guinea which suggests that the continental evaporation/evapotranspiration moisture source will be providing slightly more moisture than normal enhancing Sahelian rainfall. Thus, in combination, the regions are forecast for near normal to wetter than normal June to September rainfall in 1995. One crucial item for the forecast is what El Nino/Southern Oscillation (ENSO) phase will be present next summer and fall. If an El Nino reappears, then while the Central Sahel tends to be relatively insensitive to ENSO changes, both the West and East Sahel can be driven toward drier conditions than are indicated here. On the other hand, if a strong cold phase of ENSO becomes present, then the West and East Sahel may show wetter conditions than suggested here. Thus the ENSO parameters will have to be closely monitored over the next several months.
If these forecasts verify closely, then the West Sahel will have rainfall amounts very similar to 1988 and 1989. These two years, not coincidentally, were the last two years with large amounts of Atlantic intense hurricane activity in the deep tropics (south of 25N). The Central Sahel forecast is similar to favorably wet rainfall amounts occurring in 1988, 1965, 1960, and 1951. The East Sahel forecast is similar to near normal rainfall amounts that occurred in 1979, 1974, 1969, and 1955.
The amount of skill indicated in the hindcast testing of the five predictors during the years 1950 to 1991 is 56% of the variability in the West Sahel, 56% in the Central, and 40% in the East. Note that simple use of only persistence provides just 25%, 32%, and 14% for the three regions, respectively. Recent tests (Mielke et al. 1994), however, have suggested that the true skill that would be found in independent data may be on the order of 45% for the West and Central Sahel and 30% for the East Sahel.
An updated forecast will be provided in early June at the beginning of the Sahel rainy season, which will use updated QBO, ENSO, and African land surface data. A verification of this report will be done by late November of 1995.
REFERENCES
Gray, W.M., C.W. Landsea, P.W. Mielke, Jr., K.J. Berry, 1992: Predicting Atlantic seasonal hurricane activity 6-11 months in advance. _Wea. Forecasting_, 7, 440-455.
Gray, W.M., C.W. Landsea, P.W. Mielke, Jr., and K.J. Berry, 1993: Predicting Atlantic basin seasonal tropical cyclone activity by 1 August. _Wea. Forecasting_, 8, 73-86.
Gray, W.M., C.W. Landsea, P.W. Mielke, Jr., and K.J. Berry, 1994: Predicting Atlantic basin seasonal tropical cyclone activity by 1 June. _Wea. Forecasting_, 9, 103-115.
Landsea, C.W., and W.M. Gray, 1992: The strong association between western Sahel monsoon rainfall and intense Atlantic hurricanes. _J. Climate_, 435-453.
Landsea, C.W., W.M. Gray, P.W. Mielke, Jr., K.J. Berry, 1993: Predictability of seasonal Sahelian rainfall by 1 December of the previous year and 1 June of the current year. _Preprints of the 20th Conference on Hurricanes and Tropical Meteorology_, San Antonio, AMS, 473-476.
Landsea, C.W., W.M. Gray, P.W. Mielke, Jr., K.J. Berry, 1994a: _Forecast for 1994 June to September rainfall for North Africa's Sahel_, CSU Dept. of Atmospheric Science paper released on 4 January, 1994, 2 pp.
Landsea, C.W., W.M. Gray, P.W. Mielke, Jr., K.J. Berry, 1994b: _Forecast for 1994 June to September rainfall for North Africa's Sahel_, CSU Dept. of Atmospheric Science paper released on 7 June, 1994, 3 pp.
Mielke, Jr., P.W., K.J. Berry, C.W. Landsea, and W.M. Gray, 1994: Simulated effects of sample size, amount and severity of contamination, and informative predictors on LAD and LSD regression model agreement measures. Submitted to _Wea. Forecasting_, 38pp.
***************************************************************************** Chris Landsea Voice: (303) 491-8605 Department of Atmospheric Science Fax: (303) 491-8449 Colorado State University Internet: Fort Collins, CO 80523 landsea@downdry.atmos.colostate.edu ***************************************************************************** From: landsea@typhoon.uucp (Chris Landsea) Subject: Africa's Sahel: 1994 rainfall verification and 1995 forecast Date: 3 Dec 1994 05:24:01 GMT Message-ID: 3bovdh$2v17@yuma.ACNS.ColoState.EDU