High-quality ecological environments are vital for sustainable agro-pastoral development. This study evaluated the spatiotemporal dynamics of ecological environment quality (EEQ) in Zalait Banner from 2000 to 2022 using the Remote Sensing Ecological Index (RSEI) and explore the correlation between various factors and EEQ via Geodetector. Results show a fluctuating upward RSEI trend over 22 years. EEQ hit a low in 2004, with “poor” areas peaking at 30.77%, followed by a significant recovery between 2009 and 2013. Spatially, the region exhibits a “high in the west/northeast, low in the central-south/southeast” pattern. Notably, the central-south region, despite early recovery, has shown continuous deterioration since 2009, requiring urgent remediation. Geodetector analysis revealed that land surface temperature (LST) is the dominant single factor (q = 0.87) influencing EEQ, followed by land use/cover (LULC) and air temperature. Interaction analysis indicates that the synergy between RSEI’s four components (NDVI, WET, NDBSI, and LST) provides the highest explanatory power, while socioeconomic factors (GDP, population) and topography show weaker effects. These findings could provide a scientific basis for local ecological management, with future research planned for the Qinghai–Tibet Plateau.

This study introduced a pre-training method for machine learning-based climate prediction models. The method leverages the advantage of climate events (some theoretical knowledge) to address their limitation (small sample size). It consists of the following steps: generating artificial samples via composite analysis of high and low anomaly events, pre-training predictive models with these samples, and selecting an optimal pre-trained model that most closely matches the observational training set from numerous repeated experiments with only the model’s random number seeds being varied. Sensitivity experiments demonstrate that this pre-training method not only substantially improves predictive skills but also significantly reduces prediction instability. This simple and practical pre-training method is applicable not only to the climate prediction events in this study but also to all climate events for which composite analysis is applicable.