Introduction

Ecosystem services (ESs) encompass a wide array of benefits derived by humans from ecosystems (Daily and Matson 2008), including tangible material product supply and intangible service provision (Costanza et al. 1997; Daily 1997), and serve as a crucial bridge for coupling social-ecological systems. The supply of ESs characterizes the capacity of natural ecosystems to provide ecosystem goods and services within a given time period (Villamagna et al. 2013), whereas the demand of ESs refers to the consumption and utilization of goods and services by the social system, primarily humans (Wolff et al. 2015; Burkhard et al. 2012). The two together constitute a dynamic flow of ESs from ecosystems to the social system (Mouchet et al. 2014; Schröter et al. 2014). With the acceleration of urbanization, more and more natural ecosystems are disturbed by human activities, and the growing conflict makes it necessary to study the relationship between urbanization and ESs, which is the basis for the sustainable development of socio-ecological systems (Larondelle and Haase 2013). Many previous studies have explored the impact of land urbanization on ESs supply. As the extensive process of urbanization has led to the conversion of natural land cover types, such as woodlands and grasslands, into developed land (Deng et al. 2009; Nagasawa et al. 2015), which has significantly altered biogeochemical cycle and led to the changes in ecosystem structure and function, and accordingly diminished the supply of ESs and escalated ecological risks (Fu et al. 2015; Delphin et al. 2016; Li et al. 2023). Moreover, rapid economic development and population growth due to urbanization further exacerbate the demand for ESs within the social system and presenting a formidable challenge to urban sustainable development (Wang et al. 2021).

However, previous studies have predominantly focused on the impact of land urbanization on the supply of ESs, few studies have explored the demand of ESs considering the demographic and economic urbanization, which signifies feedback from the social system and the absence of a closed-loop flow in the social-ecological system. By investigating ESs comprehensively through the lens of both supply and demand, we can gain deeper insights into ecosystem carrying capacity and the consequences of anthropogenic disturbances (Zhang et al. 2021). Additionally, the relationship between urbanization and ESs supply may vary when using different urbanization indicators that measure the different aspects of urbanization (Wang et al. 2019). For instance, ESs supply has negative correlation with urbanization in terms of land use/cover (Tao et al. 2018; González-García et al. 2020) (Peng et al. 2017), but an inverted U-shape relationship when urbanization is measured by population and economic density (Wan et al. 2015). The varied results highlight that it’s vital to evaluate the multifaceted impacts of urbanization on ecosystem services from an integrated urbanization perspective (Zhang et al. 2021).

The impact of urbanization on ESs may also vary along the urban–rural gradient, and such variation has important implications for urban sustainability (Gren and Andersson 2018). Previous studies mostly focused on the difference between urban and non-urban, with little attention to that between the old and the new urban areas. In newly urbanized areas, urbanization can either negatively affect ESs due to the conversion of large amounts of natural or agricultural land to built-up land, or more effectively protect natural lands with the implementation of sustainable technologies than that in the old urban areas (Wu 2013; Zhou et al. 2022). Understanding on the difference and similarity of the impacts of urbanization on ESs between old and new urban areas can help urban planners and policymakers to achieve more sustainable planning and design (Hou et al. 2015).

The ChangZhuTan (CZT) urban agglomeration, designated as an experimental zone for a resource-saving and environment-friendly society, the so-called "Two-oriented Society" initiative in China, is currently experiencing rapid urbanization, accompanied by dramatic landscape transformations and degradation of ecosystem services. In light of these challenges, this study took the CZT urban agglomeration as case study area and sought to address two key questions: (1) How does multidimensional urbanization impact the supply and demand of ESs within urban agglomerations? (2) How do the impacts differ across old, new and non-urban areas? To address these two questions, we first assessed the supply and demand of four types of ESs, namely food production, water retention, carbon sequestration, and habitat quality from 2000 to 2020. We then evaluated the impacts of urbanization on ESs supply and demand across old, new and non-urban areas. Results from this study can provide important insights for the integration of urban and rural areas and sustainable development of urban agglomeration.

Materials and methods

Study area and data sources

Covering an area of 28,000 square kilometers, the Changsha-Zhuzhou-Xiangtan (CZT) urban agglomeration comprises three prefecture level cities: Changsha, Zhuzhou, and Xiangtan, and is the political, economic, and cultural center of Hunan Province (Fig. 1). The CZT urban agglomeration is characterized by a subtropical monsoon climate, with four distinct seasons and uniform distribution of rainfall and heat throughout the year. The region has an average temperature range of 16–19℃ and an annual rainfall range of 1230-1700 mm. According to the seventh national census, the urbanization rate of the CZT urban agglomeration is 80.9%, with Changsha having an urbanization rate of 82.60%, Zhuzhou 71.26%, and Xiangtan 64.37%. It is currently in a stage of rapid development of urbanization and is implementing the new route of "Integrated Urban–rural Development" in line with national policies, making it an experimental zone for a resource-saving and environment-friendly society, the so-called "Two-oriented Society" initiative in China. As such, it provides a unique opportunity to study the response of ecological services to urbanization, offering valuable insight for promoting coordinated and sustainable development throughout the country.

Fig. 1
figure 1

Location of the CZT urban agglomeration

Full size image

There are three types of data used in this study, the urban–rural boundary data, natural environment data and socio-economic data.

Urban–rural boundary data

In our study, we divided the study areas into three parts based on the pattern of impervious surfaces between 2000 and 2018, namely old urban areas, new urban areas, and non-urban areas. Specifically, urban areas developed before the year 2000 (land cover of this area was impervious surface in 2000) was defined as the old urban areas, urban areas developed between 2000 and 2018 (land cover of this area was transformed into the imperious surface from 2000 to 2018) were defined as the new urban areas, and the rest of the areas were defined as the non-urban areas. (Fig. 2). We used the Global Urban Boundaries dataset based on 30 m global artificial impervious area (GAIA) data (Li et al. 2020), which spans 1990–2018.

Fig. 2
figure 2

The old, new and non-urban areas of the CZT urban agglomeration

Full size image

Natural environment data

This study used land cover data for 2000 and 2020, with the 2000 land cover data from the project “Survey and Assessment of National Ecosystem Changes Between 2000 and 2010” (https://www.ecosystem.csdb.cn/ecosys/index.jsp). It is divided into six categories: forest, grass, water, farmland, developed land and barren land, of which developed land is consisting of settlement and transportation land. The Map2020 was updated by using the approach of integrating the updating/backdating method and an object-based image analysis (Yu et al. 2016), including changed objects detection, rule sets building, and extensive manual interpretation to identify changed areas (Yu et al. 2019).

NDVI and DEM are from USGS (https://lpdaac.usgs.gov/); NPP is from The Global Land Surface Satellite (GLASS) product suite, generated at Beijing Normal University (Liang et al. 2021); meteorological data like temperature and precipitation are from National Earth System Science Data Center (http://www.geodata.cn/) and potential evapotranspiration data is from National Tibetan Plateau Data Center (http://data.tpdc.ac.cn/).

Socio-economic data

Rapid urbanization has brought the conversion of land use/cover, including the conversion of natural landscape into construction land, and therefore affected the supply of ESs (Delphin et al. 2016). Meanwhile, the ESs demand of social system is directly dependent on population growth and economic development (Wang et al. 2021). Therefore, this study selected three indicators of land, population and economy to quantify urbanization (Peng et al. 2017; Su et al. 2012). Among them, land urbanization refers to construction land proportion (CLP). Population urbanization is determined by population density (PD, 100 people/km2), which can be obtained from Worldpop (https://www.worldpop.org). Economic urbanization is measured by gross national product density (GDPD, 104 yuan/km2), which is spatialized by dividing the gross national product (GDP) of each district and county by the area of each city and county, where the GDP data are from Changsha, Zhuzhou, and Xiangtan Statistical Yearbook. All data were rasterized to create a visual representation, as shown in Fig. 3. The rest of the socio-economic data required for calculating ecosystem services, such as grain production and water consumption, were obtained from the statistical yearbooks or water resources bulletins of Hunan Province or three prefecture-level cities. All data were resampled to 1 km in Arcgis 10.5.

Fig. 3
figure 3

The construction land proportion, population density and GDP density of the CZT urban agglomeration

Full size image

ESs supply and demand measurements

The selection of ecosystem services in this study was based on four main criteria: (1) alignment with the categorization framework/system of the Millennium Ecosystem Assessment, which includes supporting, regulating, provisioning, and cultural services (Hassan et al. 2005); (2) closely related to urbanization and human well-being (Ouyang et al. 2016); (3) based on the localization context: in the national ecosystem service function importance characterization, CZT is located in the very important area of water conservation and soil conservation, and in the national ecological function zoning, CZT mainly has the ecological regulating function of agricultural products provision and biodiversity protection; (4) data availability.

Based on the above principles, we choose food production (provisioning service), water conservation and carbon sequestration (regulating service), habitat quality (supporting service). Food production and water conservation reflect the necessities for human survival and are closely related to population urbanization, carbon sequestration can measure the impact of urbanization on carbon emissions in the context of "carbon peak and carbon neutrality". Habitat quality reflects the maintenance of biodiversity by ecological conservation and was therefore chosen to represent support services. The supply and demand for each service was calculated as follows:

Food production

Food production service is one of the main services provided by agro-ecosystems and is the basis for human survival and stable socio-economic development (Li et al. 2016a). In this study, the supply of food production services is defined as the sum of the yields of grain, oilseeds, vegetables and melons that can be provided by farmland in the unit area (1 km*1 km grid) of each prefecture-level city (Wang et al. 2019; Xie et al. 2018), which is calculated by the following formula:

$${S}_{FP}=A\times \left({P}_{G}+{P}_{O}+{P}_{V}\right)$$

where SFP is the supply of food production services per unit area in t/ km2; A is the area of arable land in each raster in km2/ km2; and PG, PO, and PV are the yields of grains, oilseeds, vegetables and melons in t/ km2, respectively, and the data were obtained from the Hunan Provincial Statistical Yearbooks of 2001, 2011 and 2021.

The demand for food production services in this study is calculated by the product of per capita food consumption multiplied by population density (Deng et al. 2021). The formula is as follows:

$${D}_{FP}={\rho }_{POP}\times {D}_{PC}$$

where DFP is the demand for food production per unit area in kg/km2, ρPOP is the population density per raster in people/km2. DPC is the per capita consumption of food in kg/people, and the data were obtained from the China Statistical Yearbook, of which 206 kg/people in 2000 and 141.2 kg/people in 2020.

Water retention

Hydrological services (for instance, water yield and water retention) are the basis for realizing other services such as soil generation, carbon sequestration and recreation (Brauman et al. 2007). Compared with water yield services, water retention services are more representative of the ecosystem's capacity to store water from rainfall, as its service supply is calculated based on the principle of water balance: the amount of water that ecosystems can store is the amount of water supply minus the amount of runoff (Sharps et al. 2017).

$${S}_{WR}={S}_{WY}-R$$

where SWR is the supply of water retention services per unit area, in t/km2; SWY is water yield per unit area, in t/km2, and R is runoff, in t/km2.

The water yield supply is calculated based on the Budyko coupled hydrothermal equilibrium, i.e., all water except evapotranspiration reaches the watershed outlet, which in this study was calculated by the water yield module in InVEST (Brauman et al. 2007). The runoff is the product of the precipitation and the runoff coefficient.

$${S}_{WY}=(1-\frac{AE{T}_{x}}{{P}_{x}})\times {P}_{x}$$

where AETx is the annual actual evapotranspiration for pixel x, and Px is the annual precipitation on pixel x.

The demand of water retention service is evaluated by the total amount of water consumption per unit area, that is, the water consumption of agriculture, industry, residential life (including urban public) and ecological environment is allocated to the corresponding land use type, and the sum of water consumption of each grid is calculated. In this study, agricultural water consumption in prefecture-level cities was allocated to farmland, the sum of industrial, residential and urban public water consumption was allocated to developed land, and ecological environmental water consumption was allocated to forest, grass, water and barren land according to the weight of area proportion to achieve spatialization of service demand.

$${D}_{WR}={D}_{Agricultural}+{D}_{Industrial}+{D}_{Domestic}+{D}_{Ecological}$$

where DWR is the demand for water conservation services per unit area, in m3/km2; DAgricultural, DIndustrial, DDomestic, DEcological refer to the consumption of water for agriculture, industry, domestic and ecological environment per unit area, in m3/km2, and the data on water consumption for different purposes were taken from Water Resources Bulletin of Hunan.

Carbon sequestration

Based on the principle of photosynthesis, we used NPP data to determine the amount of carbon sequestration, that is, for producing every kilogram of dry matter, 1.63 kg of CO2 must have been sequestrated (Li and Zhou 2016). The formula is shown below:

$${S}_{C{S}_{x}}=1.63\times NP{P}_{x}$$

where SCSx denotes the quantity of carbon sequestration for pixel x, and NPPx stands for the amount of net primary productivity for pixel x.

The demand for carbon sequestration services is expressed using the sum of carbon dioxide emissions from each sector. In this study, the agricultural carbon dioxide emissions from agriculture in each prefecture-level city are allocated to farmland, and the sum of industrial and domestic carbon dioxide emissions are allocated to developed land, with the formula shown below:

$${D}_{CS}={D}_{Agricultural}+{D}_{Industrial}+{D}_{Domestic}$$

where DCS is the demand for carbon sequestration services per unit area, in t/km2; DAgricultural, DIndustrial and DDomestic are the CO2 emissions per unit area from agriculture, industry, and domestic use, respectively, in t/km2, and were derived from the China City CO2 Emission Dataset (Because of missing data for 2000, 2005 data were used to replace 2000).

Habitat quality

Habitat quality can be a sideways response to biodiversity, and in general, it is degraded as the intensity of nearby land-use increases (McKinney 2002). The supply of this service is calculated using the InVEST model, which is expressed by assessing the extent of various habitat types or vegetation types in a given area and the degree of degradation of each of these types, and which requires inputs such as data on the type of land use, the sensitivity of the various types of land use to the different sources of threat, and the spatial data on the distribution and density of each source of threat.

To calculate the demand for habitat quality, we use the average level of habitat quality in the study area as the demand standard, and then the demand is the difference between the demand standard and its supply of habitat quality (Shi et al. 2020), with the following equation:

$${D}_{HQst}=\frac{\sum\limits_{x=1}^{n}{S}_{HQx}}{A}$$
$${D}_{HQx}=\left\{\begin{array}{cc}0,& {S}_{HQx}\ge {D}_{HQst}\\ {D}_{HQst}-{S}_{HQx},& {S}_{HQx}<{D}_{HQst}\end{array}\right.$$

where DHQst is the standard for habitat quality demand, indicating the average supply level of habitat quality in the study area. Meanwhile, SHQst refers to the amount of habitat quality service supplied for pixel x. DHQx and SHQx represent the demand and supply for habitat quality service for pixel x, respectively.

Ecosystem services supply–demand ratio (ESDR)

In this study, the ecosystem services supply–demand ratio (ESDR) is used to reflect the deficit, surplus and equilibrium of ecosystem services supply and demand (Li et al. 2016b), with the following equation:

$$ESDR=\frac{S-D}{\left({S}_{max}+{D}_{max}\right)/2}$$

where S represents the supply of a service, D represents the demand of a service, Smax and Dmax represent the maximum value of the supply and demand for the service respectively. ESDR < 0 is deficit, ESDR > 0 is surplus, and ESDR = 0 indicates in a balanced state.

Spearman's rank correlation and OLS linear regression

Spearman's rank correlation is used to calculate the correlation between the urbanization indicators and the supply and demand of ESs.

The ordinary least squares based one-way linear regression is used in this study to detect the positive and negative impacts of urbanization on different ESDRs. And the population density and the GDP density taken as logarithms to maintain dimensional unity in this study. The formula is shown below:

$$y={\beta }_{0}+{\beta }_{1}x+\varepsilon$$

where y is the ESDR for each ES, x is the urbanization indicator, β0 and β1 are coefficients, and ε is the residual. The p-value for a hypothesis test whose null hypothesis is that the slope is zero, using Wald Test with t-distribution of the test statistic. And R2 reflects the goodness of the fit. Greater the R2, better the fit.

Random forest

Sensitivity analysis can identify sensitivity factors from various independent variables that have a significant effect on the dependent variable to indicate the contribution of the independent variable to the accuracy of the model or the importance of the independent variable to the model (Gong et al. 2006). Random Forest (RF) is used to qualify the sensitivity analysis in this study by evaluating the significance of each independent variable. It is a machine learning algorithm based on decision trees that can be used for classification and regression (Breiman 2001). The RF regression model builds multiple decision trees that are unrelated to each other by randomly extracting samples and features, and obtains predictions in a parallel fashion. Each decision tree can produce a prediction result from the extracted samples and features, and the regression prediction result of the whole forest is obtained by averaging the results of all trees together. Because of its high accuracy and resistance to overfitting, the random forest regression model can be used for sensitivity analysis by assessing the importance of each independent variable, and has been widely used in the fields of ecology, remote sensing, and social sciences (Li et al. 2019; Zhang et al. 2017).

Specifically, we chose three urbanization indicators (CLP, PD, GDPD) as the independent variables and ESDRs as the dependent variable. The RF Regressor function in the Scikit-learn module is called in Python 3.10 to perform the simulation, and the dataset is randomly divided into a training set used to build the model, with a sample size of 75% of the total dataset, and a test set used to evaluate the model performance, with a sample size of 25% of the total dataset to find the optimal combination of parameters and then perform the sensitivity analysis. Finally, the explained variance score (EVS) is used in this study to express the contribution of model accuracy, with the following formulation:

$$explained\_variance\;\left(y,\widehat y\right)=1-\frac{Var\left\{y,\widehat y\right\}}{Var\left\{y\right\}}$$

A higher EVS indicates that the model predicts more accurately, the best possible score is 1.0, lower values are worse.

Results

Spatial and temporal pattern of ESs

The supply and demand of four ESs all exhibited significant spatial heterogeneity both in 2000 and 2020 within the CZT urban agglomeration. High-value supply areas of ESs were primarily situated in the northeastern and southwestern non-urban areas. In contrast, high-value demand areas of ESs were concentrated in the old urban areas, indicating substantial deficits in ESDRs (Fig. 4). Low-value supply areas, high-value demand areas, and ESDR deficit areas expanding significantly over the two decades, as depicted in Fig. 4.

Fig. 4
figure 4

Spatial distribution of ESs (a) Supply; (b) Demand and (c) ESDR in 2000 and 2020

Full size image

Between 2000 and 2020, the supply of carbon sequestration and food production services increased by 30% and 17%, respectively, while the supply of water retention and habitat quality services decreased by 22.7% and 2.9%, respectively. Zhuzhou exhibited the highest total supply and had a higher increasing trend compared to Changsha and Xiangtan. On the other hand, the demand for water retention, carbon sequestration, and habitat quality increased by 35.2%, 14.2%, and 7.1%, respectively, while the demand for food production decreased by 19.9%. Meanwhile, the total demand and its increased rate in Zhuzhou exceeded Changsha and Xiangtan during the study period. Regarding ESDRs, habitat quality, water retention, and food production showed surplus supply and demand, while carbon sequestration experienced a deficit, albeit to a lesser extent. Deficit areas expanded outward from the center of the CZT urban agglomeration between 2000 and 2020. Additionally, the supply–demand ratios for each ES exhibited a clear downward trend. At the city level, Zhuzhou boasted the highest ESDR for each service, followed by Changsha and Xiangtan.

Differences in changes of ESDRs across old, new and non-urban areas

The changes of ESDRS for four ESs had significantly spatial heterogeneity across old, new and non-urban areas from 2000 to 2020 (Fig. 5). For the CZT urban agglomeration scale, the ESDR for water retention notably declined in old urban areas among the three urbanization areas, primarily due to increased industrial and domestic water demand in these regions. In old urban areas, there was a increase in the ESDR for carbon sequestration, while the ESDR for food production remained stable. In new urban areas, the ESDRs for habitat quality and water retention decreased, while those for food production and carbon sequestration remained stable. Within non-urban areas, the ESDR for water retention experienced an increase, whereas ESDRs for other services remained stable (Fig. 5a). In terms of city scale, total ESDRs in Zhuzhou was highest, followed by Changsha and Xiangtan. In addition, the changes in ESDR for each ES in Changsha and Zhuzhou across different urbanization areas were similar to those in the CZT urban agglomeration (Fig. 5b and c). In contrast, Xiangtan experienced the greatest decreases in ESDRs, both in new and old urban areas (Fig. 5d). Overall, new urban areas displayed the lowest total ESDR, followed by old and non-urban areas.

Fig. 5
figure 5

Differences in changes of ESDRs across old, new and non-urban areas in (a) the CZT urban agglomeration; (b) Changsha; (c) Zhuzhou; (d) Xiangtan

Full size image

Impact of urbanization on ESs and ESDRs

The results of the Spearman's correlation analysis between urbanization and ESs show that the impacts of urbanization on food production was not significant on both the supply and demand sides. On the supply side, the three urbanization indicators had a significant negative impact on water retention, carbon sequestration and habitat quality, with the degree of impact being CLP > PD > GDP. On the demand side, there was a significant positive impact on water retention and habitat quality, whereas urbanization had a significant negative impact on carbon sequestration. The degree of impact of urbanization on the demand for these three services is PD > CLP > GDP (Fig. 6).

Fig. 6
figure 6

Correlation coefficients and significance of urbanization and ESs supply and demand

Full size image

The results of the linear regression showed that urbanization demonstrated predominantly negative impacts on ESDRs across the CZT urban agglomeration. The new urban areas and the non-urban areas, both experiencing rapid urbanization, were notably more susceptible to those negative effects when compared to old urban areas. The degree of impact appears most pronounced in non-urban areas, followed by new urban areas, and least in old urban areas. Moreover, the results showed that the extent of impact remained relatively consistent from 2000 to 2020. (See Fig. S1 for 2000 results and Fig. 7 for 2020 results).

Fig. 7
figure 7

Impacts of urbanization on ESDRs in the CZT in 2020

Full size image

In old urban areas, both PD and GDP showed positive influences on ESDRs for carbon sequestration and water retention, with a notably more significant positive effect on carbon sequestration (p<0.05). However, they showed negative impacts on ESDRs for food production and habitat quality, with the negative linear effect on food production being more significant (p<0.05, R2=0.52). In addition, the CLP negatively affected ESDRs for water retention and habitat quality, with the highest fit for habitat quality (p<0.05, R2=0.46), showcasing a non-significant negative impact on food production and a weak, yet non-significant positive impact on carbon sequestration (p>0.05). In new urban areas, different urbanization indicators had varying impacts on ESDRs for each service. PD displayed a substantial significant negative impact on ESDR for food production but showcased a negative yet statistically insignificant effect on ESDR for carbon sequestration. Meanwhile, it revealed the most significant and positive impact on ESDR for habitat quality and a positive insignificant effect on ESDR for water retention. On the other hand, CLP significantly affected ESDRs for habitat quality (p<0.05, R2=0.50) and water retention (p<0.05, R2=0.19) negatively, exhibited an extremely slight and statistically insignificant negative impact on ESDR for food production, and an insignificant positive effect on ESDR for carbon sequestration. GDP demonstrated a significant positive effect on ESDR for carbon sequestration and an insignificant impact on ESDRs for all other services. In non-urban areas, all three urbanization indicators showed significantly negative impacts on all types of ESDRs, and had highly similar impacts on the same ESDR. Such negative impacts were found to be the greatest, with almost the largest R2, in non-urban areas compared to both old and new urban areas. Analyzing individual ESDRs, urbanization displayed the most substantial negative effects on habitat quality and water retention, with food production and carbon sequestration following closely behind. Three urbanization areas were distinguished in Fig. 7, the lines are orange for old urban areas, green for new urban areas, and blue for non-urban areas. Detailed results of the regression models are available in Table S1.

Sensitivity of ESDRs to urbanization

The results of the RF regression model indicated a consistently high and stable explained variance score, signifying a strong sensitivity of ESDRs to urbanization. However, the extent of sensitivity varied among different ESDRs and urbanization indicators. Specifically, the ESDR for food production exhibited the highest sensitivity to PD, while ESDRs for water retention and habitat quality were most sensitive to the CLP. From 2000 to 2020, the ESDR of carbon sequestration increased notably in sensitivity to GDP, whereas the sensitivity of other ESDRs to urbanization remained relatively stable generally (Fig. S2). Furthermore, the findings from the perspective of three types of urbanization areas demonstrated that ESDRs were sensitive to GDP density, and construction land proportion in both old and new urban areas, however, construction land proportion was found to be the most sensitive indicator for ESDRs in non-urban areas. These results are illustrated in Fig. 8, where the top three rows represent the outcomes for (a) old urban areas, (b) new urban areas, and (c) non-urban areas, respectively. The horizontal axis in each graph signifies the significance of the different urbanization indicator, and the percentage displayed in each figure corresponds to the explained variance scores (EVS).

Fig. 8
figure 8

Sensitivity of ESDRs to urbanization in 2020

Full size image

Discussion

Mechanism behind urbanization impacts on ESDRs

Taking the CZT urban agglomeration as a case study, we investigated the impacts of urbanization on the ESs supply and demand across old, new and non-urban areas. In addition to the varied impact of urbanization on ESs supply and demand between urban and non-urban areas, we also found there were significant differences between new and old urban areas within the cities. Specifically, we discovered that new urban areas, which are undergoing rapid transformation, carried higher economic and population densities than non-urban areas, faced greater negative impact of urbanization on the ESs supply and demand than old urban areas. Therefore, it is important to focus on improving urban planning and implementing sustainable technologies in new urban areas to address these challenges. This research contributes to the expanding field of ESs research and offers scientific support for integrated urban and rural development within the CZT urban agglomeration, promoting sustainable development and optimal management of urban socio-ecological systems.

In our analysis, the ESs supply and demand demonstrated spatial mismatch across the CZT urban agglomeration. The areas with high supply values were predominantly located in the non-urban regions, while high demand areas were centered in the new and old urban areas, often corresponding to areas with marked changes and deficits in ESDRs. This trend aligns with the research by Xin and colleagues, who found that in the Fujian Delta, significant alterations in the ESs supply and demand were concentrated in the urban core areas and economic development zones (Xin et al. 2021). Additionally, the results of the relationship between urbanization and the ESs show that the impacts of urbanization on food production was not significant on both the supply and demand, which can be attributed to the food needs to be redistributed across regions for the purpose of food security (Ercsey-Ravasz et al. 2012; Konar et al. 2018). As for ESs supply, the three urbanization indicators had a significantly negative impact on other three ESs. For the demand, the increase in the degree of urbanization increased the demand of water retention and habitat quality services, whereas the demand for carbon sequestration services decreased with the increase in the degree of urbanization, which is due to the implementation of "energy saving and emission reduction" policies and regulations leading to the reduction of industrial carbon emissions, as well as the increase in the use of cleaner energies, all of which leads to a decrease in carbon sequestration demand.

In addition, new urban areas, characterized by higher population and economic densities than non-urban areas, are currently undergoing a phase of rapid transformation. However, urbanization had negative impacts on the ESDRs of these burgeoning regions, underscoring the need for focused attention. These areas present an opportunity to enhance urban planning and integrate sustainable technologies, which could mitigate the negative impacts of urbanization. Intriguingly, certain findings in the old urban areas exhibit a positive impact, which can be supported by previous studies that it is attributed to the improved and optimized internal ecological space in old urban areas as urbanization levels increase (Li et al. 2016a; Zhou et al. 2022). Furthermore, the sensitivity response of ESDRs to urbanization revealed that both old and new urban areas were highly sensitive to PD, CLP and GDPD within the CZT urban agglomeration, while non-urban areas exhibited the highest sensitivity to CLP. Combining the results of the impact of urbanization on ESDRs, we can infer that the internal renewal of old urban areas or land construction in new urban areas increases or decreases the supply of ESs, which made the ESDRs sensitive to CLP. Additionally, the rapid change of PD and GDPD affected the demand for ESs by the social system, so the ESDRs were sensitive to all three urbanization indicators in these two regions. On the other hand, in non-urban areas that are dominated by natural landscape, development was still dominated by land urbanization, making ESDRs in this region most sensitive to CLP. This aligns with Hou et al.'s findings (Hou et al. 2023) where land use/land cover changes were the primary contributors to ESs supply and demand alterations in 19 urban agglomerations in China. Based on this, our suggestion is to emphasize coordinated sustainable development of all three factors in both new and old urban areas, and to focus on rational planning of built-up areas expansion in non-urban areas, to ensure the future sustainable development of urban agglomerations.

Moreover, our study analyzed the different impacts of urbanization on ESs supply and demand at both the city and urban agglomeration scales under the ecological framework of “pattern, process and scale” (Fu et al. 2011). At the city scale, the ES supply and demand are more localized, significantly influenced by specific municipal policies, infrastructure development, and local environmental conditions. For instance, urban greening initiatives play a pivotal role in shaping both the quality and quantity of ESs supplied within the city, concurrently, the density and distribution of urban populations are instrumental in determining the demand for ESs (Wu et al. 2019; Quaranta et al. 2021). At the urban agglomeration scale, assessing the cumulative impacts of urbanization across cities highlights the need for regional cooperation and planning to manage ESs effectively. The integration of transportation networks, shared resource management, and coordinated land-use policies are essential for balancing the ESs supply–demand dynamics across the entire agglomeration. Understanding these dynamics at both scales is crucial for developing policies that promote sustainable social-ecological system development.

Uncertainties and limitations

There are several limitations in our study that warrant further exploration in future research. To begin, with regards to the selection of ESs, the services chosen for this study fall under the four primary categories of provisioning, supporting, and regulating services, while cultural services are not included. Cultural services are predominantly provided by natural land types such as woodlands and water bodies (commonly referred to as blue-green infrastructure in city), encompassing activities like outdoor recreation, leisure and tourism (Zhou et al. 2018). These cultural services are closely linked to the physical and spiritual well-being of urban residents (Gulickx et al. 2013). In future studies, considering the cultural services can enhance the comprehensiveness and completeness of the quantitative ESs framework. By analyzing the changes in ESs supply and demand in the region as a whole and the results of urbanization's impact on service supply and demand under different urbanization levels, we provide scientific support for policy recommendations to promote synergistic development of urbanization and ESs in regions with varying levels of urbanization.

Besides, green infrastructure serves as a significant source of ESs supply, offering various benefits such as cooling (Li et al. 2022; Zhou et al. 2021) and the enhancement of air quality (Yu et al. 2022) within urban areas. It's worth noting that studies have revealed that using low-resolution data can lead to underestimations in green space assessment (Qian et al. 2020). Hence, it is advisable to incorporate high-resolution land cover data, which not only enables a more accurate identification of diverse land uses but also facilitates the measurement of urban ESs at a finer scale. Regarding the selection of urbanization indicators, we opted for three key metrics: construction land proportion, population density, and economic density. In future research, it is recommended to provide a more comprehensive depiction of urbanization by considering various aspects, such as urban transportation, the quality of urban green spaces, and their accessibility, among other relevant factors.

Moreover, we examined the influence of urbanization on ESs across the old, new and non-urban areas. It's worth noting that the definition of urban–rural boundaries remains a subject of ongoing debate (Gonçalves et al. 2017). In addition to the global artificial impervious area (GAIA) data employed in our study for the three-tier division of different urbanization levels, it is also possible to draw on methods of classifying the urban–rural gradient, which include using concentric buffer zones (Guo et al. 2021; Wang et al. 2022) or combining a hexagonal sampling approach with cluster analysis (Xu et al. 2020), etc. Future research could investigate and compare the differences in the impacts of urbanization on ESs under various methods for the division of different urbanization levels, which has rarely been explored yet. Furthermore, in modeling the relationship between urbanization and ESs, we chose the random forest model based on decision tree. In future research, it is advisable to explore a variety of machine learning methods, such as support vector machines (SVM), artificial neural networks (ANN) and other machine learning methods, from which we can filter out the optimal method for modeling complex nonlinear relationships.

Conclusion

Our study quantified the differences in the impact of urbanization on the ESs supply and demand in old, new, and non-urban areas, and the sensitivity of ESs supply and demand to urbanization We found: (1) Urban expansion resulted in expanded areas with low ESs supply but high demand, and thereby increased deficit in ESs, extending outward from the urban edges for all types of ESs from 2000 to 2020. New urban areas had the lowest ecosystem services supply–demand ratio (ESDR), followed by old and non-urban areas. (1) Urban expansion resulted in expanded areas with low ESs supply but high demand, and thereby increased deficit in ESs, extending outward from the urban edges for all types of ESs from 2000 to 2020. New urban areas had the lowest ecosystem services supply–demand ratio (ESDR), followed by old and non-urban areas. (2) The proportional cover of developed land, population density, and GDP density generally had negative associations with the ESDRs. This negative effect was particularly prominent in non-urban areas. But it is worth noting that all three urbanization indicators had positive impacts on the ESDRs of carbon sequestration and water retention in the old urban areas. (3) The sensitivity of ESDRs to urbanization indicators differs in old, new, and non-urban areas, indicating the necessity of place-based strategies for landscape management. ESDRs were sensitive to GDP density, and construction land proportion in both old and new urban areas. However, construction land proportion was found to be the most sensitive indicator for ESDRs in non-urban areas. Results from this study can enhance our understanding on the impacts of urbanization on ESs, and provide insights for landscape sustainability in an increasingly urbanizing planet.