<p>Dryland vegetation underpins ecosystem services and livelihoods. Understanding the influencing factors of its dynamics is critical for effective restoration and degradation risk reduction. Most assessments still rely on unvalidated vegetation indices, assume monotonic trends over a single period, and use coarse attribution approaches that blur the respective roles of climate, soil–water conditions, and land use. This paper verifies NDVI (Normalized Vegetation Index) and kNDVI (Kernel Normalized Vegetation Index) using unmanned aerial vehicle (UAV) observation data. The temporal and spatial changes of vegetation in Inner Mongolia from 2000 to 2024 and the driving mechanisms of climate-soil-groundwater and land use were analyzed by using the sequence Mann-Kendall mutation test, the trend analysis of Theil-Sen (Theil-Sen) + MK (Mann-Kendall), and the Hurst index, pixel-wise correlations and a Geodetector model. Main findings: (1) compared with NDVI, kNDVI better identifies low-cover/poor-growth areas; (2) vegetation shows a fluctuating upward trend (slope ≈ 0.0034 yr⁻¹) with a mean kNDVI of 0.255, and a northeast-to-southwest decline in greenness with peaks in Hulunbuir; (3) vegetation conditions improved over 77.29% of the region (mainly in the northeast) and degraded over 22.71% (chiefly central–eastern); Theil–Sen slope estimator combined with the Hurst exponent indicates kNDVI is likely to increase over most areas, with ~ 10.65% showing a declining tendency; (4) groundwater depth and precipitation are the principal natural drivers of interannual fluctuations, with groundwater showing the strongest association (up to <i>r</i> = 0.95, <i>p</i> &lt; 0.01). In contrast, spatial heterogeneity is mainly shaped by soil nutrients, land use, and topography, among which total nitrogen provides the highest explanatory power (q = 0.41). Overall, the results underscore the central role of groundwater and soil conditions, calling for restoration strategies that integrate water management and land-use planning.</p>