Dianchi Lake basin, located in Yunnan-Guizhou Plateau in western China, is a typical plateau basin with remarkable ecological sensitivity and vulnerability (Zhang H. et al., 2013a). Specifically, it has a subtropical monsoon humid climate with an annual temperature of 15°C and an average annual rainfall of 935 mm. The whole Dianchi Lake basin is within the administrative area of Kunming city, covering an area of about 2909.41 km². It is made up of nine county-level administrative districts, including the main metropolitan region of Kunming: Wuhua, Guandu, Panlong, Xishan, and Chenggong (Zhao et al., 2012; Zhang H. et al., 2013a) (Figure 1). As the economic, political and cultural core of Yunnan Province, the Dianchi Lake basin only accounts for 14% of Kunming’s land area, but it has 52% of this city’s population and 70% of its economic aggregate (Li et al., 2012; Wang et al., 2020). The economic development of the whole basin is closely related to the economic development of Kunming city. The land use types of Dianchi Lake basin are complex and diverse, including farmland, forest land, grassland, water and built-up land, of which 27% are farmland, 26% are built-up land, and 11% are water and wetland (Table 1).

However, the development intensity of Dianchi Lake basin is more than 15%, and available land and water resources are limited and scarce, with low environmental capacity (Wang et al., 2020). As the region with the highest degree of urbanization in Yunnan Province, it cannot ignore the realistic damage and potential threat caused by urbanization to watershed ecosystem health (Xu, 2011). Since the 1980s, rapid urbanization of Kunming city has caused serious disturbance and damage to the natural hydrological cycle of Dianchi Lake basin, mainly in the following aspects: the degradation of river ecosystem function, water pollution, especially the increasingly serious eutrophication of Dianchi Lake water, the reduction of ecological land area, serious soil and water loss, and the sharp decrease of the density and curvature of major rivers (Lei and Xu., 2018). At present, the urbanization process has a very prominent impact on the Dianchi Lake basin, especially the water environment deterioration and eutrophication in Dianchi Lake that has threatened the ecosystem health status of the whole basin and restricted the sustainable development of the lakeside social economy (Xu, 2011; Xie et al., 2015). Therefore, Dianchi Lake basin has highly urgent and representative in regional development, outstanding ecological vulnerability, rapid urbanization, and dramatic land-use/land-cover change, which can offer practical guiding significance for multiscale ecosystem health assessment (Lei and Xu., 2018).

The basic data of this study include the following. 1) Land use data. We used Landsat TM/OLI remote sensing images (2018) with a spatial resolution of 30 m. The images were taken on cloudless days from February to March. On this basis, according to the national standard Classification of Land Use Status, the landscape types of the study area were divided into six categories: forest land, farmland, water, grassland, built-up land and unused land. These classification results were further corrected by manual visual interpretation using background data from field investigations. Finally, it ensures that the Kappa coefficient of classification results is greater than 85%, which meets the application accuracy requirements. 2) DEM data with resolution of 30 m. All these data were downloaded from the geospatial data cloud (www.gscloud.cn). All spatial data were processed and mapped using ENVI (version 5.1) and ArcGIS (version 10.2) software.

The purpose of this study was to determine the appropriate scale for ecosystem health assessment in Dianchi Lake basin based on landscape pattern index and VOR model. Three steps are included in this study. The first was to ensure the appropriate grain size for the landscape pattern index. The second was to evaluate ecosystem health at different scales based on VOR model. The third was to compare and identify the most appropriate evaluation scale based on the results of multiscale assessment. The specific technology roadmap was shown in Figure 2.

As the grain size increased, the values of landscape pattern indices changed and showed different trends, which indicated that landscape pattern indices were affected by the change of spatial grain size to a certain extent (Table 6). Compared with SHAPE-CV and LSI, other landscape pattern indices varied more slightly. SHAPE-CV and LSI changed obviously and fluctuated greatly, indicating the existence of particle size effect. The reason was that with the increase of spatial grain size, smaller plaques merged into peripheral plaques, which led to the rapid reduction of small plaques, and the reduction of small plaques led to the regularity of plaque shape, which was manifested as the decrease of LSI and the fluctuation of SHAPE-CV. When the spatial grain size was 150 m, the two landscape pattern indices, SHAPE-CV and LSI, fluctuated obviously. Therefore, 150 m was chosen as the appropriate spatial grain size for this study.

The results of multiscale ecosystem health assessment are shown in Table 7. There were differences in ecosystem health index values, health level and area at three different scales in Dianchi Lake basin, and the order of ecosystem health index was as follows: quadrat scale > administrative scale > ecological scale (Table 7).

The overlay analysis of ecosystem health results at three scales (Table 8) showed that: 1) For the quadrat scale and administrative scale, there were two overlapping regions of healthy and sub-healthy grades, and the overlap area of sub-healthy accounted for the largest proportion (29.61%). 2) For the quadrat scale and ecological scale, there were three overlapping regions of healthy, sub-healthy and unhealthy grades, and the overlap area of sub-healthy accounted for the largest proportion (38.26%). 3) For the administrative scale and ecological scale, there were two overlapping regions of healthy and sub-healthy grades, and the overlap area of sub-healthy accounted for the largest proportion (43.03%). 4) For the three scales, there were two overlapping regions of healthy and sub-healthy grades, with the total overlapped area accounting for 34.59%.

Table 8 further showed that, the overlap area of ecological health regions at the ecological scale and quadrat scale was the largest based on the overlapping results of ecological health regions at all levels, which was 60.16%. The mean values of the overall ecosystem health of these two scales were in the range of 0.4–0.6, with the most similar health evaluation results. Although it was easy to control the accuracy of ecological health analysis by dividing ecological health regions with quadrat scale, its disadvantage was that it could not ensure the ecological integrity of the study region, especially for the plateau watershed with a high degree of terrain fragmentation. On the contrary, taking the sub-watershed as the evaluation unit could ensure the ecosystem integrity of the study area to the greatest extent. Therefore, it was more reasonable to use the ecological scale than the quadrat scale to analyze the ecosystem health in this study area.

Establishing an adequate index system is the basis and key to evaluating ecosystem health objectively and accurately. However, there is no universally adopted index system in ecosystem health assessment (Yang et al., 2017). The main reasons are as follows. Firstly, the establishment of the index system involves many subjects, and it is necessary to have enough understanding of the evaluation system. Secondly, different indicators should be selected to evaluate ecosystem health at different scales for different types of ecosystems. Therefore, it is impossible to generalize the index system for evaluating ecosystem health (Gao et al., 2010). Comprehensively considering the actual situation of land use in the Dianchi Lake basin, this study used the landscape pattern theory of landscape ecology to understand the ecosystem health of the basin and constructed the ecosystem health evaluation index system consisting of ecosystem vitality, organization and resilience. The evaluation results of the three scales coincided well, which verified the scientific and operational nature of this evaluation system. The evaluation index system constructed in this study enriches the existing technical method system of ecosystem health evaluation.

Some studies also believe that the landscape pattern index quantitatively describes the quantity and pattern of land cover and the morphological structure of surface vegetation, which can reflect the impact of human activities on watershed ecosystem health and is an appropriate indicator for watershed ecosystem health evaluation (Xiao et al., 2003). Because the regional ecosystem is composed of multiple landscapes, the ecosystem service function of the region is maintained dynamically by landscape structure and function, and the different adjacent relationships will inevitably affect the service function. Accordingly, landscape pattern plays an important role in ecosystem health, and landscape pattern index can be used as a suitable indicator for regional ecosystem health assessment (Ren et al., 2000; Fang et al., 2012). In addition, some studies also consider other indicators for ecosystem health assessment (Gao et al., 2010). For example, Gao et al. (2010) chose the landscape pattern index but also selected social and economic statistics data and ecological and environmental factors data from surface monitoring to construct an indicator system of ecosystem health evaluation.

There are three kinds of evaluation units for ecosystem health assessment. Firstly, the spatial grid is taken as the evaluation unit (Jiang et al., 2021). Its advantage is that regional spatial differences can be expressed in more detail, while its disadvantage is that the spatial resolutions of multi-source data are mostly different, and data resampling operation is required to obtain the same spatial resolution (Jiang et al., 2021). Secondly, as an evaluation unit, the administrative unit is more common and widely-used (Lu et al., 2010). Since the monitoring projects and data statistics of China’s social economy, environment, forestry, agriculture, water conservancy and other departments are conducted based on administrative units, the selection of administrative areas for evaluation units will facilitate data acquisition, ensure the integrity of the evaluated ecosystem, and promote the application of evaluation results (Gao et al., 2010). Lastly, the sub-watershed is taken as the evaluation unit (Wu et al., 2015; Ma et al., 2019). The sub-watershed is a complete geomorphic unit with ecological integrity and can reduce the interference of some overly detailed information (Xie et al., 2015).

The results of this study indicated that the most suitable scale for evaluating ecosystem health in the plateau lake basin was the ecological scale, that is, the sub-watershed was the evaluation unit. This is consistent with the research of some scholars (Luo et al., 2003). For example, Luo et al. (2003) also believed that the study of watershed ecosystem health should take into special consideration the drainage system of the watershed. Due to the continuous characteristics of rivers, the health damage of tributaries often affects the health of the mainstream and even the whole watershed. Therefore, the continuity and integrity of tributaries should be fully considered when selecting the scale of watershed ecosystem health research (Luo et al., 2003).

The results of this study showed that using the VOR model is a good way to evaluating the ecosystem health status of the Dianchi Lake basin: at the quadrat, administrative and ecological scales, the spatial distribution of the comprehensive evaluation results of ecosystem health are relatively consistent, and the overall evaluation results are relatively similar. However, in this study, the ecosystem health evaluation index system based on the VOR model highlighted the characteristics of the natural ecosystem of the watershed and ignored the urbanization pressure and social and economic development pressure of the watershed as a typical socio-economic-natural complex ecosystem.

In the future, to further study the ecological health of watershed, the main exploration contents are: 1) To build a new interdisciplinary and multiscale health research framework, collect data by “3S” technology, and update the health indicators timely; 2) The evaluation units are classified according to the main driving forces affecting the ecological process of sub-watershed units, and then the evaluation index system is set up by zoning classification. Additionally, the connotation, evaluation process and application of watershed ecosystem health have not been unified, and need to be further analyzed and discussed.