Sugarcane yield is influenced by many abiotic factors such as radiation, temperature, water availability and crop management. In order to understand the variability of cane yield over years and to help improve management, it is necessary to disentangle the different factors. In this study, we used a modelling approach to quantify the annual variability of cane yield in Réunion over 16 years and compared it to cane yield obtained by sugar companies. Réunion presents contrasting climates with both irrigated and rainfed areas. We used the growth model MOSICAS, combined with spatial databases (sugarcane field boundaries, soil type, meteorological data, irrigation management), to predict cane yield at field and regional scales under two scenarios: a non-water stress scenario and a scenario under standard water availability (rainfed or irrigated). The average yield gap (i.e. difference between potential cane yield and actual yield) across Réunion was 95 t/ha, of which 52% was due to water deficit. A minor proportion (<5%) of cane yield variability over years could be explained by global radiation and temperature variability, while water deficit explained up to ~25% of temporal sugarcane variability. On the contrary, spatial cane yield variability across Réunion was strongly influenced by cane cultivar (20%) and global radiation (15%). Nevertheless, a large part of the yield gap resulted from factors not taken into account in our modelling approach (e.g. fertilization practices, weed pressure or extreme events such as storms). This study highlights water deficit as a primary climatic limiting factor to yield gap in both irrigated and rainfed areas. Nevertheless, the yield gap due to crop management has to be further investigated to better understand the main causes of limited cane yield.