To investigate the long-term patterns of variation and driving factors of vegetation vitality in Changting County in the context of soil erosion control in the red soil region of southern China, a long-term vegetation vitality dataset was constructed from Landsat imagery from 2000 to 2020. The integrated topographic correction (ITC) model was applied to mitigate the impact of complex mountainous terrain, thereby enhancing the spectral consistency and classification accuracy of the imagery. High-precision identification of major vegetation types was achieved by integrating the random forest classification method. Furthermore, the Theil-Sen trend analysis and the Mann-Kendall significance test were employed to reveal the spatiotemporal evolution characteristics of vegetation vitality in Changting County from 2000 to 2020. The geodetector was utilized to analyze the dominant driving forces and interaction mechanisms underlying its spatial heterogeneity. The results indicate that among the four vegetation types, broad-leaved forests exhibited the highest vitality, followed by coniferous forests and bamboo forests with similar values, while shrub-grass communities showed the lowest vitality. All vegetation types demonstrated a fluctuating upward trend. The trends in key treatment areas and non-key treatment areas aligned with the overall county pattern. However, the vitality advantage of broad-leaved forests was more pronounced in key treatment areas, and regional differences in vegetation vitality were observed. Vegetation vitality across the county improved significantly, with areas showing a highly significant increase accounting for 52.17%. The recovery rate in key treatment areas exceeded that in non-key treatment areas. Topographic factors and vegetation type were the primary drivers of vegetation vitality change, while the influence of human activity factors has gradually increased, reflecting the dynamic interplay between ecological governance and socio-economic development.

In the report, in order to improve the accuracy of vegetation classification under rugged terrain conditions, the Integrated Topographic Correction (ITC) model was used to correct the topography of Landsat remote sensing images of Fuzhou City in 2014 and 2023, and the Random Forest (RF) algorithm and the Recursive Feature Elimination (RFE) algorithm were used for feature selection to construct an optimal feature subset that eliminates the impact of terrain. Ultimately, a Random Forest classifier was used for vegetation classification. The Rate of Change was used to elucidate the degree of dynamic change of each vegetation type in Fuzhou City from 2014 to 2023. The driving factors behind vegetation changes were explored. The results indicated that ITC model effectively restored the spectral data of self and cast shadows to the level of sunny areas. After correction, the overall accuracy and Kappa coefficient of vegetation classification were significantly improved. From 2014 to 2023, the total vegetation area in Fuzhou City showed a decreasing trend, with a land-use dynamic change rate at −0.71%. Factor detection revealed that the driving factors of vegetation spatial changes at different stages are significant different, however, temperature, soil type, and nighttime light brightness are the key influencing factors. Interaction detection showed that the factors exhibited dual-factor enhancement or nonlinear enhancement interactions across all years, which suggested that the interactions among the factors further accelerated the spatial changes of vegetation.