Abstract
In order to predict the daily rain amount at Dakar at 1–5-day lead times, 65 thermodynamical and dynamical indices are computed at each grid point for the area 15°S–30°N, 30°W–30°E. The data used are NCEP–NCAR reanalyses and daily rainfall obtained by averaging over 21 rain gauges near Dakar, for 23 Augusts (1968–90). At each lead time and each grid point, a Pearson product–moment correlation coefficient r is computed between each index and the rainfall over 17 Augusts (1968–84). Predictive regression equations are developed including the 65 indices taken at the grid points where their r value is at a maximum. The prediction skill is tested over six Augusts (1985–90). The variance (R2) explained is 42% for the 1-day lead time, it decreases slowly up to the 4-day lead time (35%), and it is 28% at the 5-day lead time. The skill is better than when climatological data are used to predict rain amount. Among the predictors that appear most frequently in the predicting equations are lifting condensation level, vorticity at 700 hPa, humidity at 925 hPa, the total water vapor flux in the monsoon layer, and water vapor meridional flux in the 600–300-hPa layer.
Corresponding author address: Dr. Abdoulaye Deme, LMD, Ecole Polytechnique, F91128 Palaiseau Cedex, France. Email: defelice@lmd.polytechnique.fr
