Atmospheric drought, defined here as days on which vapor pressure deficit (VPD) exceeds its 90th percentile, is usually studied in water-limited regions yet threatens vegetation where water is abundant. Using solar-induced chlorophyll fluorescence (SIF) from the contiguous SIF (CSIF) dataset, MODIS gross primary productivity (GPP), and meteorology in the Nanling Mountains, China, we quantify the dominant climatic signal and vegetation response. Extreme hot days showed no trend, whereas extreme drought days rose nearly tenfold, from 10.6 to 101.8 d yr⁻¹, partly reflecting a 2002–2003 humidity step; both VPD tails widened. Hydrological drought (via the standardized precipitation evapotranspiration index, SPEI-3) showed no regional trend, and the two trend fields were uncorrelated. A counterfactual reconstruction validated the record; humidity alone contributed +1.06 d yr⁻¹ (46% of the rise), temperature alone +0.12 (not significant), and the remainder reflects their joint covariation. Path analysis showed associations consistent with drought suppressing GPP mainly via SIF (indirect effect −0.174; direct effect not significant). Canopy greenness (near-infrared reflectance of vegetation, NIRv) was unresponsive in dry months, topsoil moisture showed no trend, and MODIS GPP moved in the opposite direction to CSIF in dry months (+2.42% versus −2.33%); both declined in the 2022 flash drought, CSIF five times more (−9.6% versus −1.9%).

The Qinling Mountains form a major geographical boundary and climatic transition zone in China. This study derived the kernel Normalized Difference Vegetation Index (kNDVI) from Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI data and evaluated it against conventional NDVI and gross primary productivity (GPP) in the Shaanxi section of the Qinling Mountains for 2000–2025, using trend, partial-correlation, variance-partitioning and geographical-detector analyses. (1) Vegetation improved overall: annual kNDVI rose from 0.593 to 0.668 at 0.0027 yr⁻¹ (R² = 0.90), and 84.55% of the area increased significantly. (2) kNDVI was only slightly better correlated with GPP than NDVI (pixel-wise r = 0.566 versus 0.563; difference 0.0026, 95% CI 0.0025–0.0027), and no pixel differed significantly after false-discovery-rate correction, so local superiority cannot be claimed. (3) Temperature and precipitation promoted growth whereas sunshine duration inhibited it; the three factors uniquely explained 44.3%, 39.2% and 15.9% of the climatically explained variance of kNDVI, and temperature alone 74.2% for GPP. (4) After significance masking, 60.08% of the kNDVI grid and 33.94% of the GPP grid had no significant dominant driver; temperature dominated 64.96% of the area for GPP, whereas precipitation dominated kNDVI in the west and temperature in the east. Annual climate explains only about one-fifth to one-quarter of inter-annual variability, leaving a large, spatially coherent residual trend.