关键词: 植被变化归因, 残差趋势法, 人类干扰, 空间基准, 迭代

Abstract: Residual trend method is an important method for attributing vegetation changes. The performance of this method depends on the ability of vegetation-climate relationship model to avoid the disturbance from signals of human activities effects (referred to as human disturbance). The fundamental way to suppress human disturbance is to seek modeling reference, and to ensure the degree of freedom of spatial reference is far greater than that of the temporal reference. Previous vegetation-climate relationship model was limited by the fact that only temporal reference could be used in the traditional pixel-by-pixel modeling approach. We broke through the pixel-by-pixel vegetation-climate relationship model and constructed a spatially integrated vegetation-climate relationship model. Within the new model, we developed an iterative scheme for selecting spatial reference, which help improve residual trend method. We further analyzed the vegetation changes in China from 2003 to 2022 with this new model. Results showed that the enhanced vegetation index in China showed an overall increasing trend from 2003 to 2022, with a growth rate 0.002·a^-1. Vegetation distribution showed significant spatial differences, which was bounded by the Heihe-Tengchong line. The eastern region showed significant and extremely significant improvement, accounting for 61.5% of the area covered by vegetation. Vegetation in the western region showed insignificant improvement and degradation, accounting for 36.5%. The remaining 2% area showed significant and extremely significant vegetation degradation. Human factors dominated such vegetation changes in China, with an average contribution of 87.9%. The contribution rates of human factors to both vegetation improvement and degradation areas exceeded 85%. The implementation of ecological protection policies, the improvement of agricultural management and the transformation of socio and economic development patterns were the main reasons for promoting vegetation improvement in most regions of China. Overgrazing and rapid urbanization led to the vegetation degradation in parts of the northern, eastern and central regions. The vegetation-climate relationship model constructed by residual trend method based on spatial reference outperformed the traditional residual trend method in prediction accuracy, which was more precise in quantifying the relative roles of climate and human factors. Moreover, the new model effectively avoided overestimation of the influence of climate factors and reduced human disturbance to a certain extent.

Key words: attribution of vegetation change, residual trend method, human disturbance, spatial reference, iteration