The impacts of three different water stress-timing patterns for three levels of seasonal applied water on production were evaluated in mature almond trees [Prunus dulcis (Mill.) Webb cv. Nonpareil] grown under high-evaporative demand conditions in the southern San Joaquin Valley of California. The stress timing patterns involved biasing water deficits to the pre-harvest or post-harvest periods in addition to uniform deficit irrigation for the entire season, referred to as A–C patterns. The three levels of water availability were 55, 70, and 85% of potential seasonal evapotranspiration (ETc) equivalent to 580, 720, and 860 mm of applied water per season, respectively. Treatments were imposed over four seasons. Predawn leaf water potential was used as the stress indicator and approached −4.0 MPa with the A pattern at the lowest applied water level and −3.5 MPa with the B pattern at the same irrigation level. For every level of applied water, kernel weight at harvest was significantly reduced in the A pattern relative to the B and C patterns. At harvest, the most severe reduction in kernel dry weight relative to the control (17%) occurred in 580A, while there were 11% reductions in 580B and 580C. At the 860 mm level, only the A pattern dry kernel weight was less than the control. Moreover, the A patterns for all irrigation levels had lower kernel percentages than for the B and C patterns, indicating the greater sensitivity of kernel growth relative to shell growth in the regulated deficit irrigation (RDI) scenarios that biased the stress toward pre-harvest. The B stress patterns had a strong negative impact on fruit load relative to the A patterns at the 580 and 720 mm levels of applied water. No differences in crop load relative to the control were observed among the A and C regimes for all three levels of applied water. Nut load tended to increase during the experiment with 580A and 720A while it decreased with time with the B patterns for the same irrigation levels. We believe that the lower fruit loads involve stress during flower bud differentiation, which occurs mid-August–September in this cultivar and location, quite late in the season relative to other fruit and nut crops. The most successful stress timing pattern in terms of yield (the integrator of fruit size and load) was C, which avoided the large swings in tree stress observed with A and B. The onset of hull splitting was delayed by the severe pre-harvest stress in 580A while being accelerated by the milder stress of 720A. Spider mite levels were unaffected by the RDI. Canopy size was reduced with the A patterns at all irrigation levels. This occurred without any concomitant reduction in fruit load, resulting in higher fruiting densities (305 and 283 nuts/m2 of orchard floor shaded area in 580A and 720A, respectively, vs. 214 nuts/m2 in the control). Coupling the higher fruiting densities and smaller canopy sizes with higher tree planting densities offers growers the possibility of increasing yields while consuming less water. Maintaining more compact canopies with RDI rather than pruning would also lessen the amount of wood requiring disposal, thereby moderating air quality degradation resulting from burning. It must be emphasized that the scenario we outline—increasing kernel yields while using less water due to stress-related higher fruiting densities—requires that the smaller canopies be maintained by RDI, not pruning.