In a context of severe water scarcity and high uncertainty of available irrigation water, the sustainable management not only of irrigation, but also of nutrient inputs to the crop becomes necessary. The main objectives were to determine the long-term effect of water and nutrient restrictions by using several plant water status indicators of yield and quality of flat peach, and to develop predictive models for irrigation scheduling based on seasonal water availability. Three treatments were tested: (i) control (CTL), irrigated at ~100% of crop evapotranspiration (ET c ) during the entire season; (ii) a first regulated deficit irrigation treatment (RDI 1 ), irrigated as the CTL, except during the second fruit growth stage (~70% ET c ) and late post-harvest (~50% ETc), with the same fertilization as CTL; (iii) and a second RDI treatment (RDI 2 ), irrigated as RDI 1 , but with a reduction in N-P-K fertilizer units proportionally to the reduction of post-harvest irrigation. An average weekly reduction of 43 and 109 m 3 ha –1 during the first and second deficit periods, respectively, allowed water savings of 33.6% with respect to the CTL. Irrigation water use efficiency increased by 45% and reached values of around 4.16 kg m –3 , and nutrient use efficiency increased by 46% for N and P, and 34% for K. From the data obtained, it was possible to train and validate a machine learning algorithm with the week of the year and reference evapotranspiration as predictor variables, to estimate the weekly irrigation needed to satisfy 100% of the ET c , and to know which RDI could be applied during the non-critical periods identified. From the second season onwards, trees from the RDI 2 treatment showed a higher water stress integral than those with only reduced irrigation, showing that the stem water potential is not only sensitive to water deficit, but also to reduced nutrient application.