Based on MODIS NDVI and ERA5-Land reanalysis data for 2000–2022, we estimated net ecosystem productivity (NEP) at the eco-functional-zone scale in the Yellow River Basin (YRB)with an improved Carnegie-Ames-Stanford Approach (CASA) model. Theil–Sen slope estimation, Mann–Kendall testing, coefficient-of-variation (CV) stability assessment and Pearson correlation analysis were applied to quantify spatiotemporal variations in vegetation NEP and its response to climate change and extreme events. The main findings are: (1) Over the past 23 years basin-wide NEP exhibited a significant upward trend (β = 8.36 g C m⁻² a⁻¹) with a pronounced “low in the northwest, high in the southeast” spatial pattern; areas of high inter-annual variability (CV > 25%) were concentrated in desert regions. (2) NEP was positively correlated with temperature (TEM) and precipitation (PRE) but negatively correlated with vapor pressure deficit (VPD); correlations with solar radiation (RADI) varied among eco-functional zones. (3) Both the frequency of extreme-high-temperature days and that of extreme drought days are significantly and negatively correlated with regional vegetation NEP, with NEP in arid zones exhibiting markedly stronger sensitivity to these extremes than in humid zones. These findings furnish a scientific basis for carbon-accounting, eco-compensation policy design and climate-adaptation strategies across the YRB.