Objective  In recent years, rapid urbanization has dramatically changed the landscape and made it present a trend of fragmentation and homogenization, which has led to a series of landscape and environmental problems, making the scientific management of landscape changes for sustainable development an urgent problem to be solved. Landscape health assessment can diagnose ecological problems and is an important tool for landscape planning. Landscape characters are the most convenient and effective characterization object of landscape change, and are an important tool to manage landscape change. In response to the shortage of current landscape health evaluation research, this research tries to introduce landscape character identification as a framework to build a set of landscape health evaluation methods at the regional scale. The organic combination of landscape health assessment and landscape character identification can help realize scientific and effective landscape management at the regional scale. On the one hand, it can help ensure the expression and continuity of regional landscape, alleviate landscape change problems such as erosion of natural landscape, and promote the integration and development of multifunctional landscape; on the other hand, it can realize the spatial positioning of ecological problems at the landscape level, clarify the stress factors of landscape change, and then propose differentiated landscape management strategies to improve the efficiency of landscape management.

Method  This research is conducted in the Qinba Mountain Area in northwest Hubei Province, which is the "ecological green heart" of China featuring the concentration of human-land conflicts. The research covers Zhuxi, Zhushan and Fang counties in Shiyan City, Baokang and Nanzhang counties in Xiangyang City, Yuan'an County in Yichang City, and Shennongjia Forest Area in Hubei Province. The research firstly conducts the identification of landscape character types (LCTs) and landscape character areas (LCAs), with four landscape character elements of elevation, undulation, land cover and vegetation type applicable to mountainous areas being screened, and landscape character types in the research area being identified by the clustering method with a spatial accuracy of 500 m×500 m. After that, the research achieves the transformation of fragmented LCTs into continuous LCAs by combining automatic machine elimination with a manual decoding method based on the Gestalt archiving principle, and identifies and simplifies the naming process of key characters for each character unit. Subsequently, the research introduces the pressure-state-response (PSR) model to, with landscape health as the core, and LCA as the evaluation unit, to construct a landscape health evaluation system in the Qinba Mountain Area in northwest Hubei Province. From the three aspects of pressure on landscape character change, composition and pattern state of landscape character and policy response, the research selects seven indicators in total, including the proportion of construction land, road network density, proportion of arable land and construction land with a slope of above 25°, vegetation cover, Shannon's diversity index of LCTs, CONTAG of LCTs and ecological protection red line ratio. To eliminate subjective and objective bias, the hierarchical analysis method is combined with the entropy value method for the calculation of combined weights to determine the index weights, followed by the calculation of the landscape health composite index for each LCA.

Results  The research identifies totally 30 LCTs and 26 LCAs in the research area, which are classified into five landscape health levels (unhealthy, subhealthy, relatively healthy, healthy and very healthy) based on the results of landscape health evaluation. The results show that the integrated index of landscape health of each unit ranges from 0.40 to 0.75, and the average integrated index of landscape health is 0.61. The landscape health of the Qinba Mountain Area in northwest Hubei Province is at a relatively healthy level overall, which performs better in the west than in the east, and performs best in the southwest. Specifically, 14.4% of the LCAs in the research area are unhealthy, scattered in various counties; 23.0% are subhealthy, mainly concentrated in Baokang County, and partially distributed in Zhuxi County and the eastern part of Shennongjia Forest Area; 23.8% are relatively healthy, distributed in Shennongjia Forest Area, southeast Fang County, south-central Nanzhang County and west Yuanan County; 24.6% are healthy, mainly in Zhushan County and north Fang County; 14.2% are very healthy, concentrated in Zhuxi County, Zhushan County, south Fang County and Shennongjia Forest Area.

Conclusion  Finally, the research analyzes the single-factor assessment results of pressure, state and response layers to explore the coercive factors affecting landscape health and summarize the risks faced by some representative LCAs. Taking problems as the guide, referring to relevant guiding documents and combining with the current situation of regional natural resources and industries, the research adopts the hierarchical management mode of "one district, one policy" according to the idea of "strategic guidance-problem diagnosis-key optimization", and propose the remediation strategy for representative areas. Specifically, for unhealthy and subhealthy areas, the focus should be put on ecosystem restoration; for relatively healthy areas, the principle of "focusing on maintenance with restoration as supplement" can be adopted; for the healthy and very healthy areas, the management mode of "concurrent implementation of protection and utilization" or "key protection" should be strictly followed.