Abstract:Ecological health assessment is of great significance for understanding regionally ecological health status and promoting regional sustainable development. How to automatically screen out important indicators that reflect the characteristics of ecosystems is a key issue for quantitative assessment of ecological health. Based on the pressure-state-response (PSR) framework and Ecological Hierarchy Network (EHN), this paper establishes a cross-link between the feature layer and indicator layer through literature research and causal analysis, and builds a "network" indicator system for ecological health assessment. On the basis of ensuring the completeness of the indicator system, by combining the objective calculation of candidate components weights with principal component analysis and entropy weight method, an automatic screening model of evaluation indicators is constructed based on the target optimization theory. The ecological health comprehensive indexes (EHCI) of 2001-2017 in the Kaikong River Basin of Xinjiang were calculated based on the selected indicators, then spatial differentiation and time variation characteristics of EHCI were analyzed. The results showed that, using the established indicator automatic screening model, the eco-health evaluation index of the Kaikong River Basin automatically selected 17 indicators from 31 candidate indicators, and expressed 85.98% of the information with 54.8% of the indicators. The selected indicators have been widely used in the relevant literature, and the use frequency of the selected indicators was all above 20%. The frequency percentages of normalized difference vegetation index (NDVI), annual precipitation and fractional vegetation coverage (FVC) were all over 50%, indicating the rationality of the indicator automatic screening model. The EHCI's spatial distribution in the Kaikong river basin was significantly different, generally higher in the northwest and lower in the southeast. The ecological health in the southeastern and central oasis areas was poor, and the northwest valley and the two oasis regions in the central region was good. In the past 17 years, the overall ecological quality of the river basin has improved. The area of significant improvement was 10.26%, mainly distributing in the Peacock River Oasis, far higher than the 1.61% of significant degradation. The overall ecological health improvement in the Kaikong river basin indicates that regionally ecological comprehensive management has achieved good outcomes.