A three-year experiment on early nectarine (Prunus persica L. Batsch cv. Flanoba) trees was carried out with the aim of increasing water use efficiency through applying a sustained irrigation treatment, in a commercial orchard in southern Spain. Experiments compared irrigation scheduling using conventional micrometeorology (110% of crop evapotranspiration, ETc) as a control treatment (TCTL), a treatment based on the normal practice of the farmer (TFARMER) and a regulated deficit irrigation treatment (TRDI), which involved irrigating the crop at the same level as the control (TCTL) during the critical periods of the first year (second rapid fruit growth period and 2 months after harvest) and at 60% TCTL during postharvest. In the last two years (2010 and 2011), the irrigation was scheduled to maintain the signal intensity (SI) of the maximum daily shrinkage of the trunk (MDS, SI = MDSTRDI/MDSTCTL) at different water stress levels depending on the phenological stage SI = 1.0 (non-water stress) and SI = 1.4 (moderate water stress). Most of the time that irrigation scheduling was based on MDS SI, this parameter varied only slightly around the pre-established threshold values. The information given by the stem diameter sensors and stem water potential (Ψstem) gave −1.5 MPa and MDS SI 1.5 as threshold values not to be exceeded during postharvest, since MDS and Ψstem SI values were only linear down to 1.5. The water saved amounted to 17, 15 and 37% of the amount used in the control in the three seasons, respectively. In contrast, the TFARMER treatment applied more water (about 20 and 5% more than TCTL) during the first 2 years, and 10% less than TCTL during the third season.