Objectives Net primary productivity (NPP) of vegetation serves as a key indicator for monitoring the carbon cycle process of regional terrestrial ecosystem and regional carbon sources and sinks. Estimating NPP of vegetation,revealing its evolutionary characteristics, and clarifying its responses to natural and human factors are crucial for the sustainable development of regional ecosystem.

Methods This paper takes Yinchuan City as the study area, employs Carnegie-Ames-Stanford approach (CASA), and combined with sub-meter high-resolution remote sensing images, climate, land use and other data, the NPP of vegetation in different seasons from 2015 to 2021 in the study area are estimated and its temporal and spatial evolution characteristics are analyzed. Gradient boosting decision tree (GBDT) model is used to reveal the nonlinear response and threshold effect of natural factors, human activities and land use on NPP of vegetation.

Results The NPP of vegetation in Yinchuan City ranges from 0 to 753.951 gC·m^-2·a^-1, with high-value areas concentrated in the agricultural irrigation zones along the Yellow River. The integration of high-resolution images in NPP estimation yieldes the detailed spatial distribution maps at both field and urban block scales, while enhancing estimation accuracy. The seasonal variation trend of vegetation NPP in Yinchuan City aligns with vegetation phenology and crop growth cycle. From 2015 to 2021, vegetation NPP in the third quarter shows an overall upward trend. However, urban expansion-led cultivated land occupation causes a significant local NPP decline, whereas vegetation NPP in ecological protection areas increases remarkably. Natural factors, including normalized difference vegetation index (NDVI), soil organic carbon density, elevation, and rainfall, exert significant threshold effects on NPP. Among these factors, NDVI contributes the most relatively. If NDVI exceeds 0.6, it no longer significantly promotes vegetation NPP. The NPP threshold of dry land is higher than that of paddy fields. When urban green space accounts for a certain proportion in construction land, such land exerts a positive effect on NPP. Human activities within a certain scope can promote vegetation NPP to some extent. Once the population exceeds the threshold, a larger population leads to lower ecological carrying capacity and consequently reduces vegetation NPP.

Conclusions In this paper, we use high-resolution images to estimate vegetation NPP, which enhance the spatial refinement of vegetation NPP distribution. Vegetation NPP in Yinchuan City exhibits distinct seasonal characteristics, and the seasonal variation trend is consistent with vegetation phenology and crop growth cycles. Natural factors, land use, and human activities in Yinchuan City all exert significant impacts on vegetation carbon sinks, with notable threshold effects. Notably, natural factors contribute more relatively to vegetation NPP than the other two types of factors.