Effects of On-Farm Pond Storage on Irrigation Timing, Kernel Number and Water Use of Drip-Irrigated Spring Wheat Under Canal Rotation

In arid oases of Northwest China, canal water is delivered on a rotation calendar that does not follow crop demand. On-farm ponds can store canal water, but their benefits are usually measured against canal-timed border irrigation, which confounds application method, timing and amount. A two-season field experiment (2023–2024) on a calcareous silt loam in Qitai, Xinjiang, compared canal-timed border irrigation (CB), canal-timed drip irrigation (CD), pond-supplied drip irrigation on demand (PD) and PD with 70% of the depth applied outside the jointing–milk window (PDR). The canal deliveries were 14–17 d apart, whereas readily available soil water was depleted in 8.5–9.2 d (9.8–11.1 d even if all rainfall had been effective). The canal-timed regimes accumulated 12–16 stress days (root-zone depletion above readily available water) between jointing and the milk stage, against less than 2 d under pond supply. Relative to CD, PD increased kernels per spike by 9–15%, grain yield by 12–14% (0.15–0.19 g more grain per spike) and evapotranspiration (ET) by 28 mm, without changing spike number or kernel weight. Drip irrigation on canal dates did not affect yield and reduced irrigation by 106 mm, mainly through less estimated drainage. PDR diverted 79 mm (19%) less water than CB, but its ET was not lower (+9 mm; 95% CI: −3 to 20 mm). Pond storage increased yield by improving irrigation timing, but consumptive water use did not decline, and pond evaporation added 5.0–5.4 mm (1.4% of PD irrigation). In similar conditions, ponds should be sized to bridge the delivery interval between jointing and the milk stage, and reduced diversion should not be reported as water saving unless ET falls.