The vegetation types on the Qinghai-Tibet Plateau are complex and diverse, and the ecological environment is fragile and sensitive. It is very important to study the dynamic changes in vegetation and the main factors related to these changes to grasp the present state of the regional ecosystem, maintain the balance of the ecosystem and promote the sustainable development of the ecosystem. Therefore, this paper is based on SPOT/VEG NDVI (normalized difference vegetation index) data, land use data, topographic data and meteorological data from 1999 to 2019. The spatiotemporal variation in the NDVI over the Tibetan Plateau in the last 21 years and its response to different driving factors were investigated by using the Sen slope method, Mann–Kendall mutation test, partial correlation analysis and geographical detector method. The results showed that (1) the vegetation coverage on the Qinghai-Tibet Plateau showed an increasing trend from 1999 to 2019, with increases in approximately 67.00% of the plateau area. (2) The spatial differences in vegetation coverage were large; notably, low-density vegetation areas decreased obviously, moderate-density vegetation areas accounted for approximately 50% of the total area, high-density vegetation areas were the least common, and the overall growth rate was significant. (3) The NDVI was positively correlated with temperature and precipitation, and a positive correlation was observed in more than 66% of the region. (4) The order of the influence of single driving factors on the NDVI was as follows: precipitation > soil type > altitude > temperature > gradient > slope > population density > GDP. (5) The effect of multiple factors was significantly higher than that of single driving factors, with a notable nonlinear influence. The interactions between meteorological factors, such as precipitation, and topographic factors, such as altitude, were important, with a q value over 0.79.

Developing short-term forecasting model for global atmospheric temperature and wind in the near space is crucial for understanding atmospheric dynamics and supporting human activities in this region. While numerical models have been extensively developed, deep learning techniques have recently shown promise in improving atmospheric forecasting accuracy. In this study, convolutional long short-term memory (ConvLSTM) and convolutional gated recurrent unit (ConvGRU) neural networks were applied to build for short-term global-scale forecasting model of atmospheric temperature and wind in the `near space based on the MERRA-2 reanalysis dataset from 2010-2022. The model results showed that the ConvGRU model outperforms the ConvLSTM model in the short-term forecast results. The ConvGRU model achieved a root mean square error in the first three hours of approximately 1.8 K for temperature predictions, and errors of 4.2 m/s and 3.8 m/s for eastward and northward wind predictions, respectively.