Development of improved fertilizer recommendations entirely based on field experiments is time-consuming and costly. We employed a combination of two simulation models and selected field data to develop Alternative Fertilizer Recommendations (AFR) for irrigated rice (Oryza sativa L.) in Bagré, Burkina Faso. Existing fertilizer recommendations are 82 kg N ha^-1 (wet season) or 105 kg N ha^-1 (dry season), 31 kg P ha^-1, and 30 kg K ha^-1. The model RIDEV was used to improve timing of sowing date to avoid cold-induced sterility and timing of N fertilizer applications. The model FERRIZ was used to determine AFR, based on estimations of indigenous nutrient supply for N, P, and K; yield potential (Ypot); internal N, P, and K efficiency of rice; fertilizer N, P, and K recovery fractions; and fertilizer and rice prices. Simulations suggested decreasing P and K doses to 21 kg P ha^-1 and 20 kg K ha^-1 but increasing the N dose to 116 kg N ha^-1 in the wet season (Ypot = 8 t ha^-1) and to 139 kg N ha^-1 in the dry season (Ypot = 9 t ha^-1). Alternative fertilizer recommendations keep the P balance neutral, but a negative K balance was tolerated based on the high soil K supply. Compared with existing recommendations, yield gains of up to 0.5 t ha^-1 were simulated at equal costs. These yield gains were more than confirmed in farmers' fields during four consecutive growing seasons. Alternative fertilizer recommendations increased gross returns above fertilizer costs by an average of about US $ 160 per season compared with both farmers' practice and existing recommendations. Keywords: Burkina Faso, simulation modeling, irrigated rice, nutrient management.