Climate-Driven Vegetation Stress in Cotton and Wheat Zones of Sindh, Pakistan: A Field-Validated Assessment Using Sentinel-2 Imagery and NASA POWER Climate Data

Introduction: Agricultural productivity in Sindh, Pakistan, is increasingly threatened by climate variability. Crop-specific studies linking climate to vegetation stress remain scarce, and satellite assessments are rarely anchored to field data.

Methods: Sentinel-2 imagery was processed in Google Earth Engine with Cloud Score+ masking (cs_cdf ≥ 0.6) and seasonal median compositing over the cotton (Kharif 2023) and wheat (Rabi 2022–23) seasons. NDVI and EVI were computed at 10 m and extracted at 9,801 active-canopy ground-truth sites (5,099 cotton; 4,702 wheat) from the national 2022–23 Agricultural Ground-Truthing Survey, spanning 22 districts of Sindh. Seasonal temperature, precipitation and solar radiation were obtained from NASA POWER. Because NASA POWER resolves climate only at approximately 0.5°, correlations were computed at the climate-grid scale (28 cotton, 38 wheat cells) with bootstrap confidence intervals and repeated as partial correlations controlling for latitude.

Results: Wheat vigour declined across the regional gradient (EVI versus mean air temperature r = −0.63, 95% CI −0.80 to −0.41; Spearman ρ = −0.57; p < 0.001, significant after Bonferroni correction). Cotton showed no comparable decline, instead exhibiting a positive association of lower magnitude and stability (NDVI versus maximum temperature r = 0.58; ρ = 0.24). However, mean temperature is nearly collinear with latitude in Sindh (r = −0.91), and the wheat association does not survive latitude control (partial r = −0.22, p = 0.18); it is therefore reported as a spatial gradient that this design cannot attribute to thermal forcing alone. Precipitation was not significant for either crop. Label consistency was quantified rather than assumed: 94.3% of cotton and 94.9% of wheat sites exceeded NDVI 0.4 in their labelled season, with seasonal peaks at the expected phenological timing (cotton: August, Cohen's d = 5.03; wheat: February, d = 2.86).

Conclusions: Climate–vigour sensitivity in Sindh is crop- and season-dependent, and the absence of a precipitation signal is consistent with canal irrigation partially decoupling vigour from short-term climate. Regional temperature gradients in irrigated systems are readily confounded with co-varying agro-environmental gradients, a caution for studies inferring climate sensitivity from spatial correlation alone.