The reference evapotranspiration (ET₀) is an integrated climatic variable from which many crop models derive simulated crop yields. In most of these models, different equations are parameterized leaving the choice of the equation to the user. However, the impact of the choice of the ET₀ equations on crop yield prediction has been little studied. The present study proposes a sensitivity analysis of the impact of the choice of the ET₀ equation on simulated millet yields using SARRA‐H crop model over 12 Senegalese stations representative of the Sudano‐Sahelian climate conditions of West Africa. Priestley‐Taylor, a modified Priestley‐Taylor and Hargreaves equations lead to simulated yields up to 19% than those calculated using the Penman‐Monteith equation. Despite high biases in wind speed, among the tested methods, the Penman‐Monteith method remains the most robust to derive ET₀ and yield over the major part of Senegal, Hargreaves equation being more appropriated under dry climates. The choice of ET₀ formulation introduces uncertainties representing 8% of baseline yield regardless of precipitation changes; for wet conditions these uncertainties approach 30% of the overall climate change impact. The choice of ET₀ equation is increasingly important, with local temperature changes out to 4 °C, while extreme changes above 6 °C depend less on the ET₀ equation.