Modelling Past and Future Urban Ecosystem Health Assessment —A Case Study of Kunming

Rapid urbanisation leads to increasing conflict in the human-environment relationship. The health of urban ecosystems is deteriorating and this will directly harmcommunity health and wellbeing. This paper used Kunming, the capital city of Yunnan Province, China as a case study. A health assessment model for the urban ecosystem of Kunming was built using 25 indicators reflecting five measures: driving force, pressure, state, impact and response. We calculated the indicator values in 2006, 2011 and 2016 with remote sensing and statistical data. We used the entire-array-polygon method to draw polygon graphs and calculate the overall indicator values of the three periods, based on the standardised values of all indicators. All the indicator values were below 0.25, showing that the urban ecosystem was assessed as unhealthy. On the basis of the past health assessment model, we applied a grey system forecasting method to predict the future health of the urban ecosystem.

4 assessment indicators are relatively unified. In particular, the indicators for assessment models, such as the driving force-pressure-state-impact-response (DPSIR) framework , which was modified from the pressure-state-response model (Xu et al. 2017), have been commonly used in the assessments of urban ecological safety and land use status.
Compared with other models, the DPSIR model can clearly show the relationship between various factors in a complex system. It is capable of reflecting the relationship between human activities and the ecological environment. In terms of its assessment factors, the DPSIR model considers both the impact of human factors and the current characteristics of the natural environment (Xiong, 2020). This model is, therefore, applicable for establishing an indicator system for the relatively complex and large-scale urban ecosystem of Kunming, China, used in the present study.
The methods of urban ecosystem health assessment include analytic hierarchy process (Chen et al. 2015), fuzzy mathematics (Guan et al. 2014), grey forecasting (Xing and Zhao 2019) and energy analysis . These methods are all characterised by the determination of indicator weights and, hence, they are highly subjective; their results tend to be non-objective and incomplete. In contrast, the entire-array-polygon method overcomes the limitations of the multivariate statistical methods used commonly in previous studies. This method can eliminate the dimension effects of data with different units, displaying the distinctive, objective and visible characteristics of an assessed object . We applied the entire-array-polygon method in the present study.

Study procedure
In this study, we embedded the DPSIR model into an urban ecosystem health assessment in Kunming, and conducted a dynamic analysis based on the changes in the values of 25 specific indicators representing five measures of the ecosystem (driving force, pressure, state, impact, and response). Following quantitative calculation using the model, we used the entire-array-polygon method to eliminate the effect of unit dimension and subjectivity on the model results. Then, we assessed the overall health level of the urban ecosystem during the study period based on the standardised data. According to the dynamic changes of the indicator values coupled with the actual situation of Kunming, we explored the causes of the observed changes. Finally, we adopted a grey system forecasting model to predict the health of the urban ecosystem in Kunming over the next 10 years and proposed rehabilitation measures based on the prediction results.

The driving force-pressure-state-impact-response (DPSIR) model
The DPSIR model is an environmental health information framework proposed by the European Environment Agency in 1998 ( Karen et al. 2010). This assessment model can fully reflect the human-environment relationship, not only indicating the impact of changes in the external factors (such as economy and society) on the environment, but also demonstrating the feedback of the resulting ecological state to society (Wang et al. 2020). The DPSIR model is both systematic and comprehensive , wherein D (driving force) represents the initial cause of changes to the existing state or the beginning of the event, namely the potential factors that may drive changes in the ecological environment; P 6 (pressure) represents the direct cause of state changes, namely the factors that directly affect the ecosystem, such as human production and living activities; S (state) represents the appearance and state of the ecological environment under the impacts of the driving force and pressure factors; I (impact) refers to the reaction of the ecological environment to human society under the pressure; and R (response) represents the measures and means directly or indirectly implemented to restore the ecosystem to the stable state before the pressure .

Entire-array-polygon method
On the basis of the assessment indicator model, the entire-array-polygon method creates a central n-sided polygon with the upper limits of n indicators after standardisation as radii. In total, n indicators can form (n-1)/2 different irregular central n-sided polygons. The overall indicator value is calculated based on the relationships between the standardised values of different indicators after standardisation with their maximum and minimum values (Zhou et al. 2012), as shown in Eq. (1): where Si and Sj are the standardised values of the i-th and j-th indicators, respectively; n is the number of indicators; and S is the overall indicator value. The entire-array-polygon method can effectively prevent the effects of human subjectivity during the determination of indicator weights in assessment studies, allowing for objective and quantitative assessment of single and overall indicators (Cheng et al. 2013). It has been widely used and has achieved satisfactory results in environmental and resource studies. However, this method has not been applied in the assessment of urban health.

Grey system forecasting method
The urban ecosystem, as a composite economic-social-ecological system, has typical grey system characteristics, with some information being unclear (Quan et al. 2020). Here, we used the grey system forecasting method to predict the health state of the urban ecosystem in Kunming, China. Because the original data were relatively scattered, ordinary model simulations seldom passed the relevant accuracy tests, and it was difficult to provide exact prediction values. Therefore, we adopted interval forecasting to give a range of future changes, as shown in Eqs. (2-7): is the original sequence, and its 1-AGO ; n is the number of original data.
then the upper-bound function ) ( t n f s + and lower-bound function where t = 1,2,...m, m is the number of predicted data.
The inverse accumulated generating operation of The inverse accumulated generating operation of ) 0 ( u X is also deduced, where min max u s and ∆ ∆ are the upper and lower bounds of the predicted values, respectively.

Study area
The study area consists of the six districts and eight counties of Kunming City, which is located between 102°10'-103°40' E and 24°23'-26°22' N ( Fig. 1). Kunming was a key land hub connecting Southeast and South Asia on the ancient Southern Silk Road during the Western Han Dynasty, and it is also the starting city of the Bangladesh-China-India-Myanmar Economic Corridor, with an important geographical location in the construction of the Belt and Road initiative (Zhu 2020). By the end of 2019, Kunming had a permanent population of 6.67 million. Because of the large urban population and insufficient supply of social public service resources, the fragile ecological environment is under great pressure. Meteorological hazards such as forest fires, lightning strikes and droughts also occur frequently. Furthermore, as urban expansion continues, the ecological connectivity of rivers, grasslands and forest belts in the city has been interrupted. For example, the rivers feeding into lakes such as Dianchi and Yangzonghai often dry up, while the problem of water pollution in the lakes is critical.

Construction of the model
To ensure that the indicators of different factors could accurately and completely reflect the actions and reactions of the ecosystem to the social, economic and natural conditions of the study area, we initially selected a total of 43 indicators from the different factors to design a questionnaire. The selection of the indicators was based on the assessment indicators used in previous studies, combined with the social and natural characteristics of the present study 11 area. Seventy questionnaires were sent to experts at the Central South University and Nanjing University, with 56 forms returned. Through data analysis using the Analytic Hierarchy Process software, we ranked the relative importance of various factors corresponding to the indicators. On the basis of data availability, we finally selected 25 indicators to construct the DPSIR model (Fig. 3).

Fig. 3
Diagram of the driving force-pressure-state-impact-response (DPSIR) model constructed for urban ecosystem health assessment Table 1 Description and calculation method for the indicators for urban ecosystem health assessment Afforestation area in the study area of the year

The indicator system for urban ecosystem health assessment in Kunming
The image part with relationship ID rId47 was not found in the file.
The results of the indicators for the urban ecosystem health assessment in Kunming based on the constructed model and the data from Kunming across three different periods (2006, 2011 and 2016) are shown in Table 2.

Assessment of urban ecosystem health in Kunming
We assess the indicators for the DPSIR model of Kunming using the entire-array-polygon method. The specific steps included standardising the indicator values, drawing the polygon graphs and interpreting the assessment results.

Standardising the indicator values
To eliminate the dimension effects of different indicators, we standardised all the indicators using the hyperbolic function in the R environment . The results are summarised in Table 3.

Entire-array-polygon drawings of single indicators and analysis results
Using the standardisation results of the indicator values (Table 7), we drew the polygon graphs of single indicators in each factor layer (Fig. 4). Taking into account the variation in the proportion of area affected by natural disasters pressure factor, it can be seen that when Kunming, as a plateau city surrounded by mountains, suffered from drought, the irregular distribution of water led to a complex landscape spatial structure. Overall, the landscape pattern indices suggested that the urbanisation process was accelerating and urban construction land tended to aggregate in Kunming, which strongly disturbed the landscape elements.

4) Social and economic impacts can guide the healthy trends of the urban ecosystem
During the study period, both the GDP per capita and economic density in Kunming displayed an evident upward trend over time, indicating that the urban economy was developing rapidly and that economic benefits were improving, based on ecosystem services.
The per capita disposable income of urban residents increased continuously, but the growth rate declined over time, which was consistent with the characteristics of the environmental

Calculation of the overall index value and assessment of the health level
The criteria for assessing the health state of the urban ecosystem are listed in Table 4. The overall indicator value was calculated from the standardised values of the different indicators in each year using Eq. (1). The overall assessment results for the health state of the urban ecosystem in Kunming during the study period are summarised in Table 5.

Prediction of the health state of the urban ecosystem in Kunming
On the basis of the overall indicator values of the three periods in Kunming, we used grey system forecasting theory to predict the health state of the urban ecosystem in Kunming over the next 10 years. The calculation was conducted using Eqs. (2-7), and the detailed procedure is not described again here. We obtained the following result: The findings indicate that if the current state continues to evolve, the predicted values of 2021 and 2026 are both within the range [0.1799-0.2432]; that is, the urban ecosystem will continue to be in an unhealthy state. Without taking strong measures, the health of the urban ecosystem will fail to improve and may continue to deteriorate in Kunming.

Discussion
The results showed that: (1) The DPSIR model successfully reflected urban, natural, and socio-economic characteristics, and clearly showed the causal relationship between the state of urban ecosystem health and the five influencing factors. Thus, this model provided an effective tool to determine the causal relationship between urban development and ecological environment problems (Xiong,2020). The selected indicators reflected the characteristics of Kunming as a plateau city surrounded by mountains, including fragile ecosystems, frequent natural disasters, and poor ecological restoration potential. Therefore, these indicators could comprehensively and systematically reflect the health state of the urban ecosystem in Kunming. The model showed that the driving force factor initiated the changes in the entire ecosystem, and its impact was the most critical underlying cause of an unhealthy urban ecosystem. In terms of pressure, the assessment indicators were the manifestation of the driving force indicators, and they reflected the direct cause of changes in the ecosystem. The The health of the urban ecosystem can be restored through controlling the driving force, reducing the pressure, improving the state, reducing the impact, and promoting the response.
(2) The entire-array-polygon method was combined with the DPSIR model to standardise indicators with different unit dimensions. The graphical presentation of entire-array-polygons allows for visual comparison of the dynamic changes between various indicators and factors.
In this way, the overall changes of various factors and indicators in each period were displayed, which gave clear quantitative results for the research and facilitated the comparison of the final results over different periods (Zhang et al.2016). Finally, relatively accurate results of the overall assessment of urban ecosystem health were obtained, which have universal applicability. (3) Given the incomplete future information, the grey system forecasting method was used to predict the future health level of the urban ecosystem, based on the above-mentioned multi-period overall indicator values for the current health state of the urban ecosystem. The results provided an early warning for the health state of the urban ecosystem in Kunming, which will continue to be poor if the current trajectory is maintained.
Thus, associated rehabilitation measures were proposed in a targeted manner, based on science and pertinent to the main problems (Liu et al. 2017 ).
Additionally, this was a quantitative study assessing the overall ecosystem health in Kunming, but we did not look at the spatial distribution of ecosystem health state in counties and cities within Kunming. Exploring the distribution of the health state within the study area can better show the relationships between neighbouring sub-areas within the region. Such research can also help researchers to analyse more detailed humanistic and socioeconomic relationships within the scope of the study, and it is useful to enhance the pertinence of the strategies for improving the health state of the urban ecosystem. In future studies, we should consider investigating the spatial distribution patterns of ecosystem health within the study area.
Note: All drawings were prepared by Yang Liu and Xue Huang. China under Grant No. 52078222. We thank Leonie Seabrook, PhD, from Liwen Bianji, Edanz Group China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.
Availability of data and material: We utilized both remote sensing image data and basic data. For secondary data, we duly referred to the statistical data published by the Kunming Municipal Government. Among them, the remote sensing image data is the landsat TM satellite image map of Kunming City with a resolution of 30m. The data comes from the geospatial data cloud. The basic data includes geographic data such as roads, rivers, and administrative divisions of Kunming in three stages, as well as socioeconomic data such as population and humanities in each study year.

Comflicts of interest/competing interests:
The authors confirm that no conflict of interest are associated in conducting the research and in preparing the manuscript. Personal or other relationships with other people or organizations have no influence on the research.