In West Africa, a crop yicld forecasting model, DHC, is being used by the AGRHYMET Regional Center (ARC) and its national components since 1988 to help anticipate food crisises that might occur as a result of crop failure due to drought. The DHC model simulates crop water balance and relates crop yields to a water requirements satisfaction index. Although it gives a good indication of expected yields in years and locations with bad rainfall distribution, underestimations have been noticed when water supply was adequate. This study was conducted in order to evaluate the performance of a more elaborated crop model, SARRA-H. For model adaptation, agronomie trials werc conducted in 2002 and 2003 at ARC using the IIKP (90 days), ZATIB (105 days) and MTDO (photoperiod sensitive) millet varieties sown at 2 dates and at 2 levels of nitrogen fertilization (0 and 100 kg ha -1 of urca). For model evaluation, agronomie surveys with final yield measurements were conducted from 2004 to 2008 in 10 villages and 30 farmer plots per village in the squared degree area of Niamey, Niger. The results of the adaptation show that the SARRA-H model reproduces well the above ground biomass dynamics and final grain yield in ail sowing date and nitrogen fertilization treatment when water supply was adequate. With on-farm data, small differences were found between observed and simulated yields. Depending on the year, the results are equivalent or better than those obtained with DHC. Considering this performance and its ability to take into account other factors (soil fertility) and other crops (sorghum, maize, upland rice). we recommend the SARR-H model food use in security early warning at the regional leved in West Africa.