Stage-specific construction of crop water stress index based on canopy–air temperature difference for potato water stress assessment in an arid region

In precision irrigation, canopy temperature (Tc) is a well-known indicator of crop water stress. However, translating this signal into reliable and actionable irrigation decisions for potato production in arid and semi-arid regions remains a challenge. This difficulty largely arises from the limited availability of systematic, growth-stage-specific quantitative evidence on the stability of crop water stress index (CWSI) baselines, irrigation-trigger thresholds, and their relationships to plant water status and yield responses. To address these gaps, a two-year field experiment (2023–2024) was conducted in the Yinshan region of Inner Mongolia, China, using five irrigation levels (W0, rainfed; W1–W4, 75–300 mm). The objective was to develop and validate a stage-specific, CWSI-based framework for monitoring potato water stress. Based on Tc observations (n = 7297 in 2023; n = 13,145 in 2024), the 10:00–16:00 interval was identified as the optimal diagnostic window, offering strong treatment discrimination and low variability. Within this window, the non-water-stressed baseline (NWSB) was defined as the linear relationship between the canopy–air temperature difference (ΔT = Tc − Ta) and vapor pressure deficit, and it exhibited clear stage dependence. Separate NWSB models were established for tuber initiation and tuber bulking. The upper baseline (UL) under extreme drought conditions was 3.70 °C during tuber initiation and 3.55 °C during tuber bulking. CWSI accounted for variation in soil water content (R 2 = 0.746 for 0–40 cm at tuber initiation; R 2 = 0.670 for 0–60 cm at tuber bulking), effectively tracked plant water status (R 2 = 0.76–0.80), and was negatively correlated with C i , T r , g s , and P n (R 2 = 0.645–0.771). A segmented yield-response model revealed a lower irrigation-trigger threshold during tuber bulking ( ψ = 0.278) compared with that during tuber initiation ( ψ = 0.364), with yield losses above these thresholds being well captured (R 2 > 0.83). Overall, these findings provide quantitative, stage-specific thresholds to guide CWSI-based irrigation scheduling in arid and semi-arid potato systems.