Spatial and temporal changes of ecosystem service value and its influencing mechanism in the Yangtze River Delta urban agglomeration

Abstract

As the mainstream and trend of urban development in China, deeply exploring the spatiotemporal patterns and influencing mechanisms of ecosystem service value in the Yangtze River Delta urban agglomeration is of great significance for achieving sustainable development goals in urban agglomerations. This paper uses the normalized difference vegetation index and net primary productivity as dynamic adjustment factors to measure the ecosystem service value of the Yangtze River Delta urban agglomeration and analyze its spatiotemporal evolution characteristics. Furthermore, a panel quantile regression model is constructed to explore the response differences of ecosystem service value at different levels to various influencing factors. The results show that: (1) From 2006 to 2020, the ecosystem service value of the Yangtze River Delta urban agglomeration decreased by 37.086 billion yuan, with high-value areas mainly concentrated in the southern part of the urban agglomeration. (2) The value structure of various land type ecosystems and primary ecosystem sub-services in the Yangtze River Delta urban agglomeration is stable. (3) The number of grid units with reduced ecosystem service value is continuously increasing, mainly distributed in the eastern coastal areas. (4) The degree of interference of various types of land on ecosystem service value varies, and the response of ecosystem service value at different levels to the same influencing factor also shows heterogeneity. In summary, exploring the spatiotemporal patterns of ecosystem service value in the Yangtze River Delta urban agglomeration and analyzing its influencing mechanisms is conducive to adjusting the intensity of human utilization and protection methods of ecosystems, which is of great significance for enhancing the value of ecosystem products in urban agglomerations.

Introduction

Ecosystems play a crucial role in providing humans with essential biological resources, including food and raw materials. Additionally, they also provide vital services for climate regulation, hydrology management, gas regulation, and waste disposal from human activities¹, serve as a fundamental cornerstone for ecological security, human living environments, and sustainable economic and social development². According to the United Nations Environment Programme's 2021 Annual Report, humanity still faces a significant journey in addressing the global crisis and improving the “human-earth relationship”. Promoting harmony between people and nature has become increasingly crucial. There is a global consensus on the importance of protecting and enhancing the ecology to ensure a healthy living environment for humans³. The conflict between the degradation of ecosystem service capacity and the growing demands of human beings is intensifying. The increasingly complex relationship between humans and land has brought the research on Ecosystem service value (ESV) to the forefront of the academic community^4,5,6. China's urbanization level continues to rise due to rapid economic development in recent years^7,8,9. However, this trend is accompanied by the occupation of significant amounts of arable land, wetlands, and forestland^10,11. Changes in ecological land use patterns pose a serious threat to the balance of urban ecological environments¹². Serve as the primary form and crucial driver of urbanization in China, urban agglomeration experience frequent changes in ecological land patterns during urbanization and exhibit a significant positive spatial spillover effect on urban land urbanization. This implies that an increase in the urbanization level of central cities will drive a corresponding increase in the urbanization level of adjacent cities¹³. As a new form of urbanization, the construction of urban agglomerations not only promotes regional economic development but also has a profound impact on the local ESV¹⁴. The Yangtze River Delta (YRD) urban agglomeration, known as the most economically developed and highly urbanized region in China¹⁵. The rapid urban expansion and economic growth have attracted widespread attention¹⁶. The significant increase in urbanization level is accompanied by a substantial consumption of land and water resources, placing enormous pressure and challenges on the ecological environment, which is critical to the region's sustainable development and the quality of life for its residents¹⁷. The unique position of the YRD urban agglomeration, along with its core role in the economic pattern of China and even the world, provides an ideal case for studying the spatiotemporal pattern evolution and influence mechanisms of ESV in rapidly urbanizing areas.

The concept of ESV was first proposed by Daily (1997) and gives a numerical indicator based on monetary units, which is a quantitative estimate of the capacity of ecosystem services¹⁸. Costanza et al.¹⁹ first summarized the functional types of ecosystem services and quantified the global natural asset value, laying the foundation for quantitative estimation of ESV. Since then, academics have conducted extensive research on ecosystem services. In terms of calculating ESV, the approach is based on two main dimensions: economic value and material quality. From the economic value per spective, there are the equivalent factor method²⁰ and the remote sensing model method²¹ respectively. The material quality assessment²² method is the perspective based on the material quality. In addition, some researches have also evaluated regional ESV by Meta-regression using the results of the available literature²³, which is a rigorous statistical approach with strong objectivity²⁴. Li et al.²⁵ conducted an assessment of China's terrestrial ESV from the perspective of economic value, finding a downward trend over time in the country's terrestrial ESV. Additionally, they observed spatial heterogeneity in ESV changes. Chen et al.²⁶ assessed the ESV of the urban agglomeration in the middle reaches of the Yangtze River and found that forests contributed the largest proportion to the ESV, primarily providing hydrological regulation services. They also noted that the ESV in urban areas and plains was significantly lower than in other regions. Hou et al.²⁷ discovered that the ESV of Xi’an in China increased along the urban–rural gradient. Scholars have extensively studied the temporal and spatial characteristics of ESV at various administrative scales, which could aid regional ecosystem management. However, to manage different ecosystems precisely, assessments need to be conducted on the ESV of specific ecosystems. Song et al.²⁸ found that in the wetlands of the Northeast China region, the reduction in natural wetland area was mainly converted to farmland, resulting in an overall increase in ESV. Kibria et al.²⁹. conducted research on Cambodia's national forest park and identified that the value of this forest ecosystem was underestimated and faced threats from population pressure and illegal extraction.

The assessment of ESV is a way to identify ecological problems and is not sufficient for optimal regional ecosystem service management. Therefore, it is particularly important to study the influence mechanism of ESV in order to solve regional ecological problems. The research methods on exploring the impact mechanism of ESV are mainly divided into whether to consider the spatial correlation of variables. To be specific, the methods that consider the spatial correlation of variables include geographic weighted regression^30,31 and geographic detector³². The methods that do not consider the spatial correlation of variables mainly include grey correlation analysis³³ and logistic regression³⁴. It has been shown that the factors affecting ESV can be summarized as both natural environmental and human factors³⁵ with natural factors as direct drivers and human factors as indirect drivers³⁶. Natural factors mainly include vegetation growth, climate conditions and terrain conditions³⁷. Vegetation can affect ESV by maintaining soil, sequestration of carbon and other ecological services³⁸. For example, climatic factors represented by thermal and water conditions (temperature, rainfall)³⁹, topographic relief, elevation⁴⁰, and so on, mainly affect ESV by influencing vegetation growth. Human factors include land use, population density and economic density⁴¹. With the development of cities, population agglomeration and urban expansion⁴² have changed the pattern of regional land use⁴³ as well as the accompanying socio-economic activities⁴⁴, which indirectly affect ESV by generating pollutants⁴⁵. Natural and human factors are intertwined, and human beings will selectively and systematically develop land use according to natural conditions⁴⁶. However, inappropriate over-exploitation⁴⁷ and excessive population density⁴⁸ can exceed the carrying capacity of the ecosystem for human activities, resulting in serious consequences such as environmental damage (soil erosion, desertification) and ecological collapse (For example, the ecosystem service capacity is declining, and the service capacity of food provision, climate regulation and hydrological regulation is insufficient). Therefore, the investigation of ESV impact mechanisms can help humans improve the efficiency of ecosystem management and contribute to the improvement of human-land relations. Therefore, the exploration of ESV impact mechanisms can help humans to decode regional ecological problems, improve the efficiency of ecosystem management, as well as contribute to the improvement of human-earth relations.

In summary, there is an abundance of research results on the estimation of ESV and its influencing mechanisms, which lays a solid theoretical and methodological foundation for this study. By taking the Yangtze River Delta urban agglomeration as the entry point, this paper will analyze the characteristics and influencing mechanisms of ESV’s spatiotemporal pattern evolution. This can provide insight for good ecosystem management in urban agglomeration across China. Additionally, it offers theoretical guidance for promoting sustainable socio-economic development and is crucial for formulating effective ecological conservation and urban planning policies, contributing to the sustainable development of the region. It also has the potential to point out new paths for the synergistic coexistence of economic development and ecological environment protection in other regions globally. The possible contributions of this paper include: (1) Calculating ESV by subdividing grid units and analyzing the ESV spatiotemporal pattern evolution from multiple perspectives, which is beneficial for more refined ecosystem management. (2) Using normalized difference vegetation index (NDVI) and net primary productivity (NPP) as dynamic adjustment factors in the ESV accounting, making the assessment results more consistent with regional characteristics. (3) Building a panel quantile regression model and utilizing “segmented effects” to reveal the underlying mechanisms of ESV influence in depth, emphasizing the heterogeneity of various influencing factors at different ESV levels on its dynamic changing trend, providing a new perspective for the study of ESV influence mechanisms.

Study area and data sources

Study area overview

The YRD urban agglomeration is located in the downstream region of the Yangtze River in China, with a total area of 211,700 km², occupying approximately 2.2% of China's land area (Fig. 1). It is composed of 26 cities including Shanghai, Nanjing, Hangzhou, and Hefei. The YRD urban agglomeration is low-lying in the south and high in the north, with an elevation below 10 m and scattered remnants of some solitary mountains. The YRD urban agglomeration is the region with the highest density of river network in China, with a predominantly subtropical monsoon climate and a significant increase in average annual temperature and average annual maximum and minimum temperatures in recent years. The YRD urban agglomeration has maintained high growth since 2006. According to the data of the municipal statistics bureau, the average urbanization rate of the YRD urban agglomeration reached 75.01% by 2020. As a core region, the YRD urban agglomeration had a GDP of over 20 trillion yuan in 2020, concentrating one-fourth of China's total economic output within its 2.2% regional area.

Figure 1

Source: Based on the standard map GS(2019)1825 from the Standard Map Service website of the Ministry of Natural Resources of China (http://bzdt.ch.mnr.gov.cn/), it was clipped and produced using Arcgis10.2 software. The base map boundaries have not been modified, same as below.

Location map of the Yangtze River Delta urban agglomeration.

Data sources

Since China proposed in 2006 to take urban agglomeration as the main form of advancing urbanization, this study chooses the period from 2006 to 2020 to investigate the spatiotemporal pattern evolution of ESV since the construction of urban agglomeration. The data used in this paper are mainly divided into two categories: remote sensing data and regional statistical data, the specific types and sources of which are shown in Table 1. All remote sensing data are processed through geographical operations such as mask extraction, resampling, grid projection, and spatial statistics using ARCGIS 10.2 software. To meet the needs of this study and based on the current standards of “Land Use Status Classification” issued by the Ministry of Land and Resources of China in 2017, the land use types are reclassified into six categories: farmland, forest, grassland, water bodies, desert, and construction land. This classification provides the input data for calculating the ESV of the YRD urban agglomeration. To improve the spatial differentiation resolution and the reliability of quantile regression, this paper studies the spatiotemporal pattern evolution of ESV in the YRD urban agglomeration at a grid scale. Previous research experience suggests that if the grid assessment unit is close to the resolution of the land use data, i.e., the unit is too small, it will destroy the relative consistency of the adjacent units⁴⁹; while if the grid assessment unit is close to the scale of administrative divisions, i.e., the unit is too large, it will greatly ignore the spatial differences of ESV⁵⁰. Taking into account the actual situation of the study area and after repeated experiments, this paper divides the YRD urban agglomeration into 5 km × 5 km grids, resulting in 8764 basic assessment units for related research on ESV in the Yangtze River Delta urban agglomeration.

Table 1 Data sources.

Research methodology

This paper unfolds around the research approach of "posing questions—analyzing problems—solving problems," primarily focusing on analyzing the spatio-temporal pattern evolution and its influencing mechanisms of the ESV of the YRD urban agglomeration from 2006 to 2020. Consequently, it proposes effective ecological conservation and urban planning policies to promote regional sustainable development. The research framework is shown in Fig. 2.

Figure 2

Research framework.

Remote sensing quantitative ESV accounting model

Building upon the previous work, Xie et al.⁵² derived an ESV equivalent factor table specific to China through a combination of literature research, expert knowledge, and model operation. In this study, we adopt this table as the fundamental basis for calculating the ESV of the YRD urban agglomeration⁵². Additionally, taking into account the influence of social and economic factors, we estimated the average market value of grain yield per unit area in the YRD urban agglomeration. This was done by calculating the arithmetic average of the sown area and yield of major crops, such as corn, wheat, and rice, in Shanghai, Zhejiang, Jiangsu, and Anhui provinces within the YRD urban agglomeration for the years 2006, 2011, 2016, and 2020. We also considered the annual prices in the Chinese agricultural market for the year 2020. This paper conducted a preliminary adjustment of the ESV coefficients for the YRD urban agglomeration based on the definition of the ESV equivalent factor. Specifically, this adjustment assumes that one seventh of the average market value of grain yield in the YRD urban agglomeration for a given year represents the economic value of one equivalent factor⁵³. The calculation formula is shown in the followed equation:

$$V=\frac{1}{7}\sum_{i=1}^{n}\frac{{s}_{i}{p}_{i}{q}_{i}}{M}$$

(1)

$${V}_{j}=V{\beta }_{j}$$

(2)

Let $V$ represent the economic value of one equivalent factor. Here, $i$ refers to the crop type, while $n$ represents the total number of crops $n=3$. The planting area of the $i$th crop is denoted as ${s}_{i}$, the annual average market price as ${p}_{i}$, and the yield per unit area as $q$. The total planted area of the $n$ crops is represented by $m$. Additionally, ${V}_{j}$ represents the value coefficient of ecosystem service for category $j$, and ${\beta }_{j}$ represents the current value of type $j$ ecosystem services. The calculated market value of grain output per unit area in the YRD urban agglomeration is 17317.02 yuan/hm². Consequently, using Eq. (1), the economic value of per unit equivalent in the YRD urban agglomeration is determined to be 2473.86 yuan/hm². The ESV coefficients per unit area of the YRD urban agglomeration were derived from Eq. (2) (Table 2).

Table 2 Coefficients of ecosystem service value per unit area in the Yangtze River Delta urban agglomeration (yuan hm⁻² a⁻¹).

The same land use type will serve significantly different ecosystem services due to the difference of vegetation density in each land use type⁵⁴. In addition, the net carbon sequestration by green plants in different regions through photosynthesis represents the carbon sequestration capacity of surface vegetation as well as the differences in regional ecosystem service capacity⁵⁵. Therefore, this paper uses the NDVI to reflect the plant growth status in the region and the NPP of vegetation to reflect the carbon sequestration capacity of regional vegetation, and revises the ESV of the study area to make the evaluation results more in line with regional characteristics and ensure the credibility of the results. Finally, the formulas for calculating the ESV for each land category, each sub-service, and the ESV per unit area for each city are obtained:

$${ESV}_{i}=\sum_{m=1}^{a}\sum_{k=1}^{b}\sum_{n=1}^{c}{S}_{mn}{V}_{mk}{R}_{n}$$

(3)

$${ESV}_{im}=\sum_{k=1}^{b}\sum_{n=1}^{c}{S}_{mn}{V}_{mk}{R}_{n}$$

(4)

$${ESV}_{ik}=\sum_{m=1}^{a}\sum_{n=1}^{c}{S}_{mn}{V}_{mk}{R}_{n}$$

(5)

$${AESV}_{i}=\frac{{ESV}_{i}}{\sum_{m=1}^{a}\sum_{n=1}^{d}{S}_{mn}}$$

(6)

Here, ${EVS}_{i}$ represents the ecosystem services value of the i-th city, and the ESV value of the YRD urban agglomeration is the cumulative value of ESV of the 26 cities in the region; ${ESV}_{im}$ represents the ESV of the mth land type of the ith city; ${ESV}_{ik}$ represents the value of the kth type of sub-service in the ith city.${AESV}_{i}$ represents the ESV per unit area of the $i$th city, reflecting the abundance of natural capital in the city. ${S}_{mn}$ represents the area of the $m$th type of ecosystem in the $n$th grid unit, ${V}_{mk}$ represents the value coefficient of the $k$ th ecosystem service of the $m$th type of ecosystem; ${R}_{n}$ is the dynamic adjustment factor of ESV calculated based on the spatiotemporal variation of NDVI and NPP for the $n$th grid unit. m represents the type of ecosystem, and since the ESV coefficient of construction land is 0, this paper does not consider the ESV of construction land, so $a=5$; k represents the type of ecosystem service, $b=9$; $n$ represents the grid unit number, and the value of $c$ is determined based on the size of the area of each city region, with a total of 8764 grid units in the YRD urban agglomeration.

The formula for the dynamic adjustment factor ${R}_{n}$ is:

$${R}_{n}=\frac{1}{2}(\frac{{NPP}_{n}}{{NPP}_{mean}}+\frac{{f}_{n}}{{f}_{mean}})$$

(7)

In the above equation,

$$f=\frac{NDVI-{NDVI}_{soil}}{{NDVI}_{veg}-{NDVI}_{soil}}$$

(8)

In the equation, ${NPP}_{mean}$ and ${f}_{mean}$ are the mean values of $NPP$ and vegetation cover $f$; ${NPP}_{n}$ and ${f}_{n}$ are the $NPP$ and $f$ of the $n$th grid unit; the calculation of vegetation cover $f$ is obtained using the pixel-based binary model, where ${NDVI}_{veg}$ and ${NDVI}_{mean}$ represent the $NDVI$ values of pixels that are completely covered by vegetation and pixels that have no vegetation cover, respectively. These values can be replaced by the maximum and minimum $NDVI$ values of grid units within the region. Other variables are the same as above.

Cold hotspot analysis model

The cold-hot spot analysis method is used to identify high and low value areas in different spatial locations, that is, to explore the spatial distribution characteristics of hotspots and cold spots of ESV ecological resilience⁵⁶. The calculation formula is as follows:

$${G}_{i}^{*}=\frac{{\sum }_{i=1}^{n}{\sum }_{j=1}^{n}{w}_{ij}{x}_{i}{x}_{j}}{{\sum }_{i=1}^{n}{\sum }_{j=1}^{n}{x}_{i}{x}_{j}}$$

(9)

Equation (7) is normalized.

$${Z(G}_{i}^{*})=\frac{{G}_{i}^{*}-E({G}_{i}^{*})}{\sqrt{Var({G}_{i}^{*})}}$$

(10)

where ${W}_{ij}$ is the spatial weight matrix. ${X}_{i}$ and ${X}_{j}$ represent ESV of $i$ and $j$, respectively. $E\left({G}_{i}^{*}\right)$ and $Var({G}_{i}^{*})$ are the mathematical expectation and coefficient of variation of ${G}_{i}^{*}$, respectively. If $Z({G}_{i}^{*})$ is significantly positive, it indicates that the ESV around this area is high and it belong to the hotspot area. If $Z({G}_{i}^{*})$ is significantly negative, it indicates that the ESV around this area is low and it belong to the cold pot area.

Land use intensity index

Land use intensity reflects the degree of human interference on land resources and becomes an important driving force of environmental change. Different land use types represent the characteristics of different degrees of land use, which is an important indicator of sustainable land use⁵⁷. According to Chen⁵⁸,this paper assigns values to the intensity of human use of various types of land and calculates the land use intensity index for each research unit. The specific values are: 4 for construction land; 3 for farmland; 2 for forestland, grassland and water area; 1 for desert. The calculation formula is as follows.

$${LUI}_{j}=100\times \sum_{i=1}^{n}\frac{{S}_{i}}{\sum_{i=1}^{n}{S}_{i}}{D}_{i}$$

(11)

where ${LUI}_{j}$ represents the land use intensity index of the $j$ th grid cell. ${S}_{i}$ is the area of category $i$ land use type. ${D}_{i}$ is the assigned value of land use intensity of category $i$.

Contribution rate of ESV change

The data form of land use types could not be used in the panel quantile regression model. However, considering its most direct and dominant influence on the ESV⁵⁹, the contribution of ecosystem services of each type of land is applied in this paper to analyze the interference extent of land use types with changes in the ESV. The contribution rate of ESV change can reveal the main influencing factors of ESV change in each city of the YRD urban agglomeration. It signifies the proportion of ESV change for different land use types in each city to the total ESV change of the city during the study period⁶⁰, with the following formula:

$${F}_{i}=\frac{\left|\Delta {ESV}_{i}\right|}{\sum_{i=1}^{6}\left|\Delta {ESV}_{i}\right|}$$

(12)

where ${F}_{i}$ is the contribution rate of category $i$ land to the change in ESV of urban agglomeration during the study period. $i$ is the land use type. $\Delta {ESV}_{i}$ is the change in ESV of the $i$th land during the study period.

Variable selection

The selection of influencing factors variables is based on the above review and analysis of influencing factors of ESV. The intensity of human activities on land use indirectly affects the natural ecosystem. In this paper, LUI was used as a variable to influence the ESV. Average annual rainfall (AAR) and average annual temperature (AAT) were selected as climate factor variables affecting the value of ecosystem services. Population density (PD) and economic density are the main socio-economic factors, which can reflect the overall characteristics of the development and utilization level of regional land resources. PD is considered to be the main driver of land use change⁶¹. Economic density reflects land use efficiency and regional social production development intensity⁶². Both of them affect the ESV by influencing the utilization of land resources. Therefore, the two variables were selected as the socio-economic factors affecting the ESV. Under certain conditions, night light brightness (NLB) can be used as a proxy variable of GDP, so this paper uses the regional average NLB as a proxy variable of economic density⁶³. The topographic factors do not vary significantly over the years and are affected by the deficiencies of the year of monitoring data, which could not meet the data conditions of the panel regression. Therefore, topographic factors are not included in this paper.

Panel quantile regression model

The panel quantile regression model has the advantages of both the panel model and the quantile regression model, providing more information and weaker collinearity between variables. Moreover, it has strong resistance to estimation in the presence of non-normal distribution or outliers, and can provide more complete information and clearer explanation of the heterogeneity of the response of independent variables under different distributions of dependent variables⁶⁴. This method differs from the traditional least squares regression as it minimizes the weighted absolute residuals to fit conditions at various quantiles, thereby revealing the effect of the independent variables on the dependent variable across different distributions⁶⁵. By modeling different quantiles, it is possible to analyze the effect of the independent variables at various quantile values of the dependent variable, which helps to unveil the variable sensitivity and response patterns under different conditions⁶⁶. Panel data has both time and cross-sectional properties, greatly increasing the sample size of data, and can improve the reliability and accuracy of regression results. Due to the large amount of data and non-normal distribution with outliers in the study of ESV impact mechanisms at the grid scale, this paper overcomes the non-normal distribution and outlier problems based on the characteristics of the data, using panel data of 8764 grid units for four periods in 2006, 2011, 2016, and 2020, and selects the panel quantile regression model to explore the “segmentation effect” of the response of ESV in the YRD urban agglomeration to various influencing factors. The commonly used panel regression models include fixed effects and random effects. The Hausman test shows that the fixed effects model is more suitable for this study. The basic model is as follows:

$${ESV}_{kt}=c+{\alpha }_{1}{LUI}_{kt}+{\alpha }_{2}{AAR}_{kt}+{\alpha }_{3}{AAT}_{kt}+{\alpha }_{4}{PD}_{kt}+{\alpha }_{5}{NLB}_{kt}+{\mu }_{k}+{\tau }_{kt}$$

(13)

where ${ESV}_{kt}$ is the explained variable; $c$ is the constant term; $\alpha$ is the regression coefficient; ${LUI}_{kt}$ is the land use intensity index; ${AAR}_{kt}$ is the average annual rainfall; ${AAT}_{kt}$ is the average annual temperature; ${PD}_{kt}$ is the population density; ${NLB}_{kt}$ is the night light brightness; The value of ${\mu }_{k}$ does not change over time as an individual effect; ${\tau }_{kt}$ is the disturbance term.

Panel fixed effects model studies the influence of each explanatory variable $X$ on the conditional expectation $E(ESV|X)$ of ${ESV}_{kt}$. This is a mean reversion⁶⁷. In the case of asymmetric distribution of $ESV|X$, the regression results only reflect part of the relationship between $X$ and $ESV$. Quantile regression model can be used to explore the linear relationship between $X$ and $ESVESV$ quantiles, which can depict the conditional distribution in more detail. Therefore, regression results that are not affected by extreme values are more robust⁶⁸. The corresponding regression coefficient estimates can be obtained for ${\varvec{X}}$ and ${\varvec{ESV}}$ at different quartiles. To sum up, this paper uses panel quantile regression model to study the influence mechanism of ESV. The model design is as follows:

$$Z_{{ESV_{kt} }} \left( {\beta {|}X_{kt} } \right) = c\left( \beta \right) + \alpha_{1} \left( \beta \right)LUI_{kt} + \alpha_{2} \left( \beta \right)AAP_{kt} + \alpha_{3} \left( \beta \right)AAT_{kt} + \alpha_{4} \left( \beta \right)PD_{kt} + \alpha_{5} \left( \beta \right)NLB_{kt} + \mu_{k} \left( \beta \right) + \tau_{kt}$$

(14)

where ${\varvec{\beta}}$ is the quantile, ${\varvec{\beta}} \in \left( {0,1} \right)$; ${\varvec{Z}}_{{{\varvec{ESV}}_{{{\varvec{kt}}}} }}$ is the ${\varvec{\beta}}$ conditional quantile of ESV; ${\varvec{c}}\left( {\varvec{\beta}} \right)$ is the constant term at the ${\varvec{\beta}}$ quartile; ${\varvec{\alpha}}\left( {\varvec{\beta}} \right)$ is the influence coefficient at the quartile of ${\varvec{\beta}}$; ${\varvec{\mu}}_{{\varvec{k}}} \left( {\varvec{\beta}} \right)$ is the individual effect at the quartile of ${\varvec{\beta}}$, and other variables are the same as above.

Results analysis

Spatial and temporal patterns of ESV evolution in YRD urban agglomeration

According to Table 2 and Eq. (4), the ESV of different land categories in each city were measured by grid cells using the raster calculator tool of ARCGIS, and then summed up using the spatial statistics tool to obtain the ESV of each land category in the YRD urban agglomeration (Table 3). The total ESV in the YRD urban agglomeration was 8220.23, 8189.37, 8016.77 and 784.937 billion yuan in 2006, 2011, 2016 and 2020 respectively, showing a monotonically decreasing trend, but the average annual decline rate gradually stabilized. According to the land use data, the construction land of the YRD urban agglomeration that does not generate ESV has expanded by 9343.18 km² since 2006. The area of land types generating ESV have all decreased, leading to a decline in ESV and degradation of ecosystem service capacity in the YRD urban agglomeration. It can be seen that the spatial urbanization of the YRD urban agglomeration has brought enormous pressure to the construction of ecological civilization, and the contradiction between the degradation of ecosystem service capacity and the increasing human demand has become increasingly intense. The structure of ESV in the YRD urban agglomeration is stable, mainly dominated by forestland, farmland and water area. These three together generate more than 98% of the overall value, and their respective proportion do not change significantly over time. During this study period, the average annual decline rate of ESV generated by farmland and desert ecosystems was gradually decreasing, while that generated by grassland and water area ecosystems was gradually increasing. Stress on grassland and water area ecosystems in the YRD urban agglomeration is gradually increasing. Therefore, efforts to protect grassland and water area ecosystems should be strengthened.

Table 3 Ecosystem service value and the structure of ecosystem service value in the Yangtze River Delta urban agglomeration.

Similarly, according to Table 2 and Eq. (5), the value of each sub-service for each city is obtained, and then aggregated to yield the overall value of each ecosystem sub-service for the YRD urban agglomeration. (Table 4). Overall, the value structure of Tier 1 subservices is stable, mainly providing adjustment services and support services, which together account for more than 82% of the overall value, with insignificant changes over time. Ecosystems regulate the pollution caused by human activities and maintain the stable operation of ecosystems through hydrological regulation, climate regulation, soil conservation and biodiversity maintenance. Among them, the capacity of supply and support service has been continuously degraded, however, the degradation rate has slowed down and the trend has been controlled since 2016. The value of regulating services and cultural services is also declining, and their average annual decline rate is expanding. The ability of ecosystems to play a role in gas regulation, hydrological regulation, climate regulation, and provision of aesthetic landscapes is continuously degrading, which may lead to a series of ecological and environmental problems in the YRD urban agglomeration, such as intensified climate change, water pollution, air pollution, ultimately resulting in the degradation of the living environment.

Table 4 Value and structure of ecosystem sub-services in Yangtze River Delta urban agglomeration.

By combining Eqs. (3) and (6), the per unit area ESV for each city can be calculated. To visually analyze the composition of the ESV and the level of natural capital abundance in each city, a dual y-axis graph can be drawn as shown in Fig. 3. The composition of ESV for each city is represented by a bar chart, with different colors indicating the ESV of different land categories, and their values corresponding to the y-axis on the left, unit is billion yuan. The per unit area ESV for each city is represented by a line graph, with its values corresponding to the y-axis on the right, unit is ten thousand yuan /km². At the same time, since the ESV generated by grasslands and deserts accounts for a very small proportion, it is not easy to show it in the bar chart. Therefore, we take Figure a as an example and zoom in on it to illustrate the ESV generated by grasslands and deserts. Overall, the ESV and ESV per unit area in Hangzhou are the highest, which are 97.082 billion yuan and 5.7764 million yuan /km² respectively by 2020. Hangzhou has the strongest ecosystem service capacity and the highest degree of natural capital abundance. Although Zhoushan City has the lowest ESV, it has the highest ESV per unit area of 4.1546 million yuan/km², indicating a high degree of natural capital abundance. During the study period (2006–2020), only Anqing city had a positive growth trend of ESV, while the rest of cities had a negative growth. The cities of Shanghai, Zhoushan, Changzhou, Suzhou, Jiaxing and Zhenjiang suffered serious degradation of ecosystem service capacity, and their ESV decreased by more than 10% per year compared with 2006. According to data from the regional statistical bureaus, the level of urbanization in these cities has increased significantly since 2006, which easily shows that rapidly urbanizing areas inevitably come at the cost of destroying ecosystem services. According to the analysis of land category, the urban ESV was mainly generated by farmland, forestland and water area ecosystems, while grassland and desert ecosystems ESV accounted for a small proportion. Anqing, Nantong, Hefei, Chuzhou and Yancheng, as the main food producers in the YRD urban agglomeration, have high values of farmland ecosystem services.

Figure 3

Yangtze River Delta urban agglomeration ecosystem service value and its structure.

The ESV of the grid cells of the YRD urban agglomeration were precisely spatially located, and its spatio-temporal evolution characteristics were analyzed (Fig. 4). For ease of analysis, classification is carried out at intervals of 40 million yuan. The numbers in parentheses in the legend represent the quantity of grid cells, and the same applies below. Firstly, overall, ESV in the YRD urban agglomeration presents the characteristics of spatial imbalance and time instability. The spatial divergence between high and low values of ESV is obvious, and the north–south divergence of ESV becomes more obvious as time goes on. The number of high-value grid cells gradually decreases and the number of low-value grid cells gradually increases over time, and the low-value space gradually spreads to the south of the YRD urban agglomeration. Second, counting statistical analysis of grid cells with different value levels. The number of grid cells with ESV less than 40 million yuan is increasing year by year, while the number of grids with ESV greater than 80 million RMB is decreasing year by year, and there are constantly new grid cells with decreasing ESV. Third, looking at the spatial distribution of grid units with different value levels across cities. The low-value spatial units with ESV less than 80 million yuan are mainly distributed in the northern cities of the YRD urban agglomeration. The low-value space is spread out from the center of Shanghai, Hefei, Nanjing and Hangzhou. High-value spaces worth more than 160 million yuan are mainly distributed in the southern cities of the urban agglomeration, forming a solid ecological barrier for the YRD urban agglomeration.

Figure 4

Spatial and temporal pattern evolution of ecosystem service value in Yangtze River Delta urban agglomeration.

The spatial agglomeration characteristics of ESV in the YRD urban agglomeration from 2006 to 2020 were further explored through a cold hotspot analysis model (Fig. 5). Here, “HS” stands for hotspots, “CS” stands for coldspots, and “99”, “95”, “90” respectively represent significance levels of 99%, 95%, and 90%. “NS” indicates non-significant. Counting statistical analysis shows that half of the regional ESV spatial agglomeration characteristics are not significant from 2006 to 2020, and the insignificant areas are on the rise. The number and spatial distribution of agglomeration areas are more stable, and the number of high-value agglomeration areas is higher than the number of low-value agglomeration areas. High-value agglomerations are mainly located in the eastern coastal regions, while low-value agglomerations are mainly located in the northeastern part of the YRD urban agglomeration and the eastern coastal region. The high cold spot areas spatially overlap with the central city areas with high urbanization levels. It can be found from Fig. 5 that the cold spot region in northeast China, namely the agglomeration area of low ESV value, decreases year by year and transforms from the cold spot region to the insignificant region, indicating that the ecosystem service capacity of this region has changed significantly and gradually transformed from contiguous distribution to discrete distribution. Figures 3 and 4 both show that the ESV of the eastern coast and the northern region along the Yangtze River in the YRD urban agglomeration is relatively low, which has become the frontier position of urbanization construction. The southern region of the urban agglomeration builds the logistics support area of the YRD urban agglomeration.

Figure 5

Spatial agglomeration characteristics of ecosystem service value in Yangtze River Delta urban agglomeration.

In the same way, spatial positioning was carried out for the grid cells with ESV changes in the YRD urban agglomeration (Fig. 6). Analysis of their spatial and temporal pattern evolution enables accurate monitoring to identify the changing status of ESV in the region. The spatial units of ESV changes in each research stage have obvious spatial differentiation and are discretely distributed in urban agglomeration. From Fig. 6a, b, and c, the number of grid cells with increasing (decreasing) ESV decreases (increases) with time. The percentage of the number of areas with decreasing ESV in each specific phase is 61.86%, 79.01%, and 82.98%, showing a gradual increasing trend. From Fig. 6d, the spatial percentage of ESV reduced is 82.78%, which mainly concentrated in the contiguous areas centered on Shanghai, Nanjing, Suzhou and Hangzhou, as well as the central part of Jinhua and the central part of Hefei. The spatial proportion of units with a increase in ESV is only 17.22%, which is much smaller than the number of units with a decrease in ESV. The above analysis further explains the reasons for the decrease of ESV in the YRD urban agglomeration from a spatial perspective, and the spatial positioning of the units with changed ESV can provide a reference for local governments to make decisions on targeted protection of regional ecology.

Figure 6

Spatial and temporal evolution of ecosystem service value changes in the Yangtze River Delta urban agglomeration.

Degree of disturbance to changes in ESVs by land type

According to Eq. (12), the contribution rates of ESV changes in various land categories in each city to the overall ESV change are calculated, thereby revealing the main influencing factors of ESV change in the cities of the YRD urban agglomeration (Fig. 7). The overall impact of each land ecosystem on ESV change in the same city varies from period to period; the impact of each land on ESV change in different cities in the same period also varies. In this paper, the contribution rate of all types of land to ESV change exceeding 50% is defined as the main influencing factor of ESV change in the city. During the study period, the farmland ecosystem was identified as the primary influencing factor for ESV variations in Taizhou, Wuhu, Jinhua, Xuancheng, Suzhou, Anqing, and Shaoxing. Forestland ecosystem was identified as the primary influencing factor for ESV variations in Chizhou, Nantong, Nanjing, Taizhou, Chuzhou, Yangzhou, Zhenjiang, Changzhou, Hefei and Ma 'anshan. Water area ecosystem was identified as the primary influencing factor for ESV variations in Tongling, Shanghai, Wuxi and Yancheng. From Fig. 6, it can be observed that the degree of interference from agricultural ecosystems on ESV shows an increasing trend in Ma'anshan, Huzhou, and Jinhua. The disturbance degree of forestland ecosystem to ESV in Xuancheng City and Changzhou city also showed an increasing trend. The disturbance of water area ecosystems to ESV changes in the majority of cities showed an oscillating upward or strict upward trend, and even the disturbance to ESV changes in Shaoxing, Zhoushan, Xuancheng, Taizhou, Tongling and Wuxi cities has reached 90% during 2016–2020. Based on the above results, local governments should strengthen the monitoring and management of various kinds of ecosystems to ensure the stability of regional ecosystems and enhance the ability of regional ecosystem services.

Figure 7

Contribution rate of ecosystem service value change in the Yangtze River Delta urban agglomeration.

Segmentation effect of influencing factors of ESV

Quantile regression can identify the degree of response of different levels of ESV to each influencing factor, i.e., the "segmentation effect" of various influencing factors on different levels of ESV, which can enable us to understand the influence mechanism of ESV in a deeper level. To precisely identify this "segmentation effect", nine quartiles from 0.1 to 0.9 are selected in the model for regression, and the results reveal the changes in the elasticity coefficients of each influencing factor in the conditional distribution of ESV in the YRD urban agglomeration (Table 5). The baseline regression results show that LUI, AAR, AAT, PD, and NLB which represents economic density, have significant effects on ESV in the YRD urban agglomeration. The regression coefficients of LUI and AAR are significantly negative, indicating that land development and rainfall have a significant constraining effect on the enhancement of ecosystem service capacity in the YRD urban agglomeration. The regression coefficients of AAT, PD and NLB, which represent economic activities, are significantly positive, indicating that the heat conditions provided by high temperature for plant growth, human activities and appropriate economic activities have obvious positive promoting effects on the improvement of ecological environment and the promotion of ESV capacity.

Table 5 Panel quantile regression results of ecosystem service value influencing factors in Yangtze River Delta urban agglomeration.

The following analyzes the "segmentation effect" of the impact factors on the influence of ESV at different levels. First, observing the quantile regression results for LUI, the regression coefficients for each quantile are significantly negative, indicating that the reduction in ESV is coerced by the exploitation of land at different ESV levels. The absolute value of the regression coefficient at the 0.1 quantile was significantly smaller than the rest of the quantile, indicating that areas with low levels of ESV were less stressed by land development at the same land use intensity. Second, the regression coefficients of AAR at different quartiles are significantly negative, probably due to the sufficient water resources and rainfall in the YRD urban agglomeration itself, which implies that increased rainfall will instead reduce ESV⁶⁹. The overall stress effect on ESVs in the YRD urban agglomeration was characterized by an "inverted U-shape", with the 0.5 quantile as the boundary. At the 0.1 to 0.5 quantile, the lower the ESV, the stronger the stress effect of rainfall on the region, while the result was opposite at the 0.5 to 0.9 quantile. Third, the regression coefficients of AAT were not significant at the 0.6, 0.7 and 0.8 quartiles, while the rest of the quartiles were significantly positive. The promotion effect of average temperature on ESV showed a U-shaped characteristic of marginal decrease at first and then marginal increase, with a cut-off at the 0.6 quantile. Fourth, the regression coefficients of both PD and NLB were insignificant only at the 0.1 quantile, and the rest of the quantile regression coefficients were significantly positive at the 1% level and monotonically increasing with quantile. This indicates that in the natural context, if land is not developed and utilized, certain population and economic activities in the region may provide aesthetic landscape value by improving the cultural services of the ecosystem, and thus enhancing ecosystem services.

Discussion

Ecosystem provides important ecological resources for human production and life, serves as a bridge and link for the harmonious development of man and nature⁶⁹. Ecological civilization construction is the fundamental plan for China's development, and the contradiction between the degradation of ecosystem services and the ever-increasing needs of human beings is becoming increasingly intense. The quantitative accounting of ecosystem service capacity of urban agglomeration in monetary form and the spatial identification of ESV change status translate ecological and environmental issues into economic issues that is easily understandable for the public⁷⁰, which helps the general public to understand ecological issues intuitively and aids in the monitoring of regional ecosystem dynamics⁷¹. The ESV was divided into different levels by quantile, and the heterogeneity of the dynamic change trend of the same influencing factor under different ESV levels was emphasized, so as to obtain a more comprehensive explanation of ESV in the affected region. It can also provide reference for local governments to manage regional ecosystems, solve ecological problems, and target policies. This article takes the YRD urban agglomeration as the research object, divides the study area into detailed grid units, and revises the ESV assessment model again with NDVI and NPP. It analyzes the spatiotemporal pattern evolution of the ESV in the YRD urban agglomeration from different perspectives. Finally, it applies panel quantile regression to analyze the mechanisms influencing ESV. The following discussions are conducted:

regarding the discussion of research methods

Costanza et al. accounted for the value of global ecosystem services and natural capital¹⁹, which quantified the measurement of ESV, but their results also contain biases and are not entirely applicable to ESV research in China. Xie Gaodi and others, based on Costanza et al.'s research, derived ESV equivalent factor tables suitable for China, which this article uses as the basis for measuring the ESV of the YRD urban agglomeration⁵². Based on this, the economic value per equivalent unit was first revised using the yield, planting area, and market prices of the main crops in the YRD urban agglomeration. Then, on the basis of previous work using NDVI to reflect the vegetation growth status within a region⁷², NPP was added to reflect the carbon sequestration capacity of regional vegetation for a second revision. The revised model with added NPP can measure ecosystem productivity more accurately. More importantly, NPP directly reflects the ability of ecosystems to convert CO2 into biomass, which can to some extent reflect the carbon sink potential of the region. This is of great significance in the assessment of regional ESV under China's "dual carbon" goals and conforms to the development concept that "lucid waters and lush mountains are invaluable assets." The ESV assessment model, after two rounds of revisions, can ensure that the ESV assessment results conform to regional characteristics and the characteristics of the era. It can provide decision-making references for local governments to manage regional ecosystems, solve ecological problems, and target policies. Additionally, this paper uses the panel quantile regression model to explore the mechanisms influencing ESV. Previous studies often used geodetectors⁷³, geographic weighted regression models⁷⁴, piecewise linear regression⁷⁵, etc., to examine the mechanisms influencing ESV, ignoring the differential response levels of different ESV levels to influencing factors. This paper effectively solves this problem by using the panel quantile regression model. By dividing ESV into different levels using quantiles, it emphasizes the heterogeneity in the dynamic trends of the same influencing factor at different ESV levels, leading to a more comprehensive and scientific understanding of regional ESV influences, providing a new perspective for the study of ESV mechanisms.

In terms of the discussion of research findings

From 2006 to 2020, the ESV of the YRD urban agglomeration continuously declined, which is consistent with the results of Ding et al.⁷⁶, confirming the reliability of the calculations in this study⁷⁷. Moreover, the grid cells with decreased ESV have primarily expanded from coastal cities to inland areas and from urban built-up areas to the periphery. This indicates that rapid urbanization and high-intensity land development are the main reasons for the ESV decline in the YRD urban agglomeration⁷⁸. According to the ESV grid distribution map, cities such as Hangzhou and Jinhua in the southern part of the urban agglomeration have higher ESVs, because the dominant land use type in the southern part is forest land. The study finds that the ESV of forests accounts for a relatively large proportion, which explains the higher ESV in the cities of the southern urban agglomeration. This also shows that the ESV of cities is closely related to their natural resource endowment, and during the practice of ecological civilization construction, cities should carry out targeted restoration and protection based on their local natural resource endowment. Observing the degree of disturbance of each land type on ESV changes, it's apparent that farmland, forest, and water ecosystems are the main influencing factors of ESV changes in most cities. This also suggests that in the process of ESV decline, a considerable amount of these ecosystems have been destroyed⁷⁹. In the future, protection and restoration of the ecosystems of farmland, forests, and water areas should be strengthened in all cities, and the boundaries of urban development should be controlled⁸⁰. Especially in shrinking cities, land development should be stopped or restricted, and existing built-up areas should be used rationally for urbanization construction⁸¹. The segmented effects of each influencing factor on ESV suggest that areas with higher ESV levels have a stronger negative response to land use intensity. This might be because these areas, with high ESV, initially lacked land use and development, and once they undergo land use and development, the damage to ESV is extremely severe⁸². This also tells us that when developing land, we should as much as possible adopt a contiguous development model instead of an enclave-style development model.

Future research prospects

There are also some limitations in this paper. Firstly, the accounting model of ESV has room for improvement, and the dynamic factors of spatial and temporal changes are not selected comprehensively enough, and future studies can consider more human economy, regional natural conditions and other factors. Secondly, due to the limitations of data availability and method models, the research on the influence mechanism of ESV is not comprehensive enough. For example, soil conditions, light and other factors are not considered in the analysis. In the future, new research methods can be actively explored to carry out more completed research on these aspects.

Conclusions and policy recommendations

Conclusion

Based on the multi-source remote sensing data, this paper combines the equivalent factor method and the quantitative remote sensing method, and integrates the NDVI and NPP factors that represent the regional habitat quality with spatial and temporal changes, to measure the ESV of the YRD urban agglomeration at the grid scale and analyze its spatial agglomeration and spatio-temporal evolution characteristics. The contribution rate model was used to explore the interference of each land use type on ESV changes. A panel quantile regression model is used to reveal the degree of difference in the response of different spatial regions of ESV to the influencing factors at different levels. The following conclusions were obtained.

First, from 2006 to 2020, the ESV of the YRD urban agglomeration continuously decreased, with a total reduction of 37.086 billion yuan. The ESV declines were more severe in Shanghai, Zhoushan, Changzhou, Suzhou, Jiaxing, and Zhenjiang.

Second, the value structure of various land-type ecosystems and primary ecosystem services in the YRD urban agglomeration is stable. The ESV is mainly generated from farmland, forest land, and water bodies, primarily providing regulation services and support services.

Third, the number of grid cells with reduced ESV is continuously increasing, mainly distributed in the eastern coastal areas. From 2006 to 2020, over 80% of the grid cells experienced a decrease in ESV, primarily located in contiguous areas centered around Shanghai, Nanjing, Suzhou, and Hangzhou.

Fourth, the degree of disturbance to ESV varies among different land types, and there is heterogeneity in the response of ESV levels to the same influencing factors. Farmland, forest land, and water bodies have a greater degree of disturbance to the ESV of the urban agglomeration. LUI exerts more stress on areas with higher ESV levels, while population and economic activities within the region have a greater promoting effect on these high ESV areas. AAR and AAT have stress and promoting effects on the ESV of the YRD urban agglomeration, respectively, exhibiting "inverted U-shaped" and "U-shaped" characteristics.

Policy recommendations

In this paper, through the study of ESV in the YRD urban agglomeration, the following policy recommendations are obtained.

Firstly, under the background of integrated urban agglomeration construction, it is crucial to establish an integrated ecosystem management mechanism centered on urban agglomerations. Facing the severe fact that many cities are experiencing a decline in ESV, it is important for urban agglomerations to coordinate and jointly demarcate ecological land use red lines with local natural resource departments that are adapted to the sustainable economic and social development of the urban agglomeration, especially the red lines for construction land use. The same applies to other urban agglomerations; they need to establish urban development boundaries and ecological land use red lines, clearly defining areas that are off-limits to development and the boundaries of ecological resources that can be used sustainably. This can ensure that the pursuit of economic development does not excessively sacrifice the health and service capacity of ecosystems.

Secondly, in the urbanization construction of urban agglomerations, the concept "to protect in development, and to develop in protection" should be practiced. Implementing ecosystem protection measures such as "compensation for occupation"—where for every piece of ecological land occupied by urbanization construction, measures such as artificial afforestation and lake creation are taken to maintain the total ESV. Actively construct urban green spaces to achieve ecological urbanization. Promoting rooftop greening, city parks, green belts, and other urban greening projects can not only enhance the ecological functions of cities but also improve residents' quality of life and the city's attractiveness.

Thirdly, the management of forest, water, and farmland ecosystems should be highly emphasized. Since the ecosystem services of the YRD urban agglomeration are primarily provided by forests, water, and farmlands, increased protection and restoration of these ecosystems is necessary. Scientific management methods and restoration technologies, such as ecological agricultural practices and wetland restoration, can be applied to enhance the natural resilience and service functions of these ecosystems. This is critical for increasing the regional ecosystem service capacity and ensuring the self-regulation ability of the regional ecosystem, maintaining stability while meeting human demands.

Fourthly, it is essential to improve the land-use efficiency of urbanized areas. According to the exploration of factors affecting ESV, in any area, land development behavior greatly damages the ecosystem service capacity. Therefore, during the urbanization process, contiguous development should be adopted to avoid “enclave-style” development as much as possible and reduce the damage to ecologically vulnerable areas. Optimize the city layout and land use to ensure that development activities are coordinated with ecological protection goals. Promote land-intensive use and contiguous development strategies to minimize the negative impact of urban expansion on ecosystems.