Evolution of Urban Ecosystem Service Value and a Scenario Analysis Based on Land Utilization Changes: A Case Study of Hangzhou, China

Abstract

Simulating the change in ecosystem service values (ESVs) caused by land use changes in metropolitan areas under multiple scenarios is of great significance to ensure regional ecological security and sustainable urban development. This study assessed the variations in land use and ESV in the main urban area of Hangzhou, China, from 2000 to 2020. A total of four future land use scenarios for 2030 were created using the cellular automata–Markov prediction model and ESVs were calculated for four future scenarios. The results are as follows: (1) Arable land and construction land were the most drastic types of land use changes in the main urban area of Hangzhou during the study period. From 2000 to 2020, construction land in the main urban area of Hangzhou expanded rapidly, with an increase of 46,916.82 hm², while the cultivated land area decreased significantly by 38,396.43 hm². (2) The ESV in the study area continuously declined from 2000 to 2020, with high-value ecosystem service areas predominantly found in forest areas and water areas. (3) The values of ecosystem services under the simulated future scenarios of natural development, rapid development, planned development, and ecological development were CNY 15.053, 14.525, 16.690, and 16.799 billion, respectively. The planned development and ecological development scenarios effectively ensure a high ESV. The results indicated that balancing various types of land use is essential to guarantee ecological security in urban development. Moreover, urban development and construction should be undertaken in areas with a low ESV. This forecasting study can serve as a key reference for policy makers regarding the urban landscape sustainability of Hangzhou City. The integrated simulation method of land use and ESV proposed in this study can shed light on the urban spatial layout and spatial regulation in urban land use planning.

1. Introduction

Due to a large amount of human construction activities, land use is transformed under natural and anthropogenic influences, which ultimately has a significant impact on regional ecosystems and ecosystem services (ESs) [1]. Protecting natural ecosystems and enhancing their ecosystem services have become pressing challenges globally [2]. Over the past few decades, driven by population growth, rapid urbanization, and increasing demand from economic development, humans have altered ecosystems more dramatically and extensively than ever before, resulting in large and largely irreversible ES losses. Of all human activities, changes in land use/land cover are most associated with changes in the availability of multiple ESs [3]. The rapid expansion of urban land has posed enormous challenges to the ecosystem [4]. Accelerating urbanization has caused ecological problems, such as degradation of ecological quality, destruction of biological habitats, and fragmentation of landscapes, which have become the main obstacles to the sustainable development of cities [5,6,7]. This phenomenon occurs worldwide, especially in China, where rapid urbanization over the past two decades has compounded the problem [8]. In particular, the rapid social and economic development of major urban agglomerations such as the Yangtze River Delta in China in recent years has resulted in dramatic land use changes, which have had an unprecedented impact on urban ecosystem health [9]. Ecosystem management is evolving into an application-oriented, multidisciplinary science [10]. Therefore, understanding the linkages between land use and ecosystem services has important implications for the scientific community and policy makers, as well as urban planning authorities.

Ecosystem services refers to the relevant functions provided by various ecosystems for human survival and activities [11]. Costanza et al. (1997) divided ecosystem services into provisioning services, regulating services, cultural and tourism services, and other services and measured the values of ecological service functions on a global scale [12]. The development of spatial information technology, such as remote sensing, has further pushed forward research on ecosystem services, and the evaluation method of coupling land use changes and ecosystem service value (ESV) has been widely used [13,14,15]. Several studies have examined the effects of land use changes on the value of services of different ecosystems, such as studies on highlands [16], coastal areas [17,18,19], and urban landscapes [17,20,21,22,23]. At present, an increasing number of scholars at home and abroad are studying the impact of land use changes on ESV, and ecosystem service valuation methods are widely used in urban planning.

Early research has mainly focused on the impact of current or past land use changes on ESV on a regional scale, while there are few studies on future space regulation and development based on land use change and the evaluation of ESV. Some studies in recent years have gradually analyzed changes in ecosystem services and spatial differentiation by simulating changes in future land use patterns and creating development scenarios [24,25,26,27,28]. Cities are drastically affected by human activities, and simulating land use changes in cities is more elaborate and complex than in natural ecosystems [29]. Nevertheless, a multi-scenario simulation approach can effectively integrate land use and ecosystem service changes to guide urban planning [30,31]. Against this background, simulating the changes in the future land use patterns and exploring the spatial response of its ecosystem services from different development scenarios are conducive to constructing a landscape development pattern and provide an effective approach to regulating sustainable urban development.

However, in related studies, the prediction of land use change is mostly based on the evolution of natural laws, and the revision of equivalent value factors per unit area only considered price and grain yield, without fully studying the impact of different influencing factors on land use changes and ESV [32,33]. Consequently, further investigation is necessary to establish a more holistic understanding of these complex relationships. Given that, we have chosen the main urban area of Hangzhou as the study area. We have selected geographical, water, location, demographic, and landscape factors as the influencing factors. Subsequently, we aim to forecast the evolution of land utilization under various simulated scenarios. Furthermore, our study refines the equivalence factor method [34,35] by considering the local biomass levels of Zhejiang Province, and we incorporate an assessment of the ESV for construction land. Through the application of multi-source data and the revision of ESV assessment models, our research is capable of more accurately predicting future land use changes and assessing the local ESV within the study area.

Hangzhou, the capital city of Zhejiang Province in China, ranks as the second-largest metropolis in the Yangtze River Delta region, following Shanghai. Over the past two decades, Hangzhou was one of the twenty fastest-growing metropolitan areas in China [36,37], and its rapid economic and population growth has triggered massive amount of urban construction and development. Furthermore, the uneven spatial land development and intense land use competition in Hangzhou have undermined the sustainable supply of land use functions and exacerbated conflicts about land use functions [38]. Therefore, the area is well suited to simulate urban development scenarios to study the links between land use and changes in ecosystem services.

The objectives of this study are the following: (1) to characterize land use dynamics and patterns in Hangzhou from 2000 to 2020; (2) to map out the spatial and temporal patterns of ESV in response to land use changes; and (3) to examine the relationships between land use and ESV changes under different urban development scenarios. This study constructed a suitability atlas of land use driving factors from the perspectives of geography, water sources, location, population economy, and landscape based on the random forest regression model. We used the CA–Markov model to simulate the four scenarios of urban natural development, ecological development, planned development, and rapid development in 2030. On this basis, the urban spatial changes and ESV were simulated to provide a reference for the decision-making process of land use planning in Hangzhou and other regions and to promote regional sustainable development.

2. Materials and Methods

2.1. Overview of the Study Area

Hangzhou is the capital of Zhejiang Province in China (118°20′–120°37′ E, 29°11′–30°34′ N). This study focuses on the Hangzhou metropolitan area, which is the second largest metropolis in the Yangtze River Basin, China [36]. From 2000 to 2020, the population of Hangzhou increased from 6.2158 million to 11.965 million, and its gross domestic product (GDP) surged from CNY 138.256 billion to CNY 1.610583 trillion. With the rapid growth in the population and economy, the coordinated development of urban land expansion and ecological security has become a key issue to be considered [39]. To explore the impact of Hangzhou’s urban spatial expansion on the ecosystem, this work studied the main urban area of Hangzhou (Figure 1), including Xihu District, Yuhang District, Shangcheng District, Gongshu District, Qiantang District, Xiaoshan District, Linping District, and Binjiang District, based on 2020 data. Hangzhou faces challenges brought by significant human–land conflicts and ecological risks, making it an ideal place for research that will lead to significant practical implications.

2.2. Data Sources

The data used in this study were uniformly resampled to 30 m. The land use remote sensing image data of Hangzhou in 2000, 2010, and 2020 came from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/ (accessed on 13 August 2022)), including six Class-I land use types, namely cultivated land, forest land, grassland, water area, construction land, and unused land. Digital elevation data and slope data came from the Geospatial Data Cloud (http://www.gscloud.cn/ (accessed on 26 September 2022)). The MOSID NDVI dataset was downloaded from NASAs official website (https://ladsweb.modaps.eosdis.nasa.gov/ (accessed on 21 August 2022)). The data on road networks, rivers, and lakes were obtained from the National Geomatic Center of China (https://www.ngcc.cn/ngcc/ (accessed on 28 September 2022)). VIIRS/NPPYE night band data were acquired from the official website of the National Oceanic and Atmospheric Administration (https://www.ngdc.noaa.gov/ (accessed on 28 September 2022)). GDP data came from the Resources and Environmental Sciences and Data Center of the Chinese Academy of Sciences. The prices of staple food crops in the study area in 2020 came from the Zhejiang Provincial Food and Strategic Reserves Administration. Other data came from the Hangzhou Statistical Yearbook.

2.3. Research Methods and Data Preprocessing

2.3.1. Dynamic Degree of Land Use

The dynamic degree of land use is based on the conversion between different land use types. It analyzes the degree of land use change and the direction of conversion and includes the dynamic degrees of single land use and comprehensive land use [40]. The dynamic degree of single land use reflects the change in the area of a certain land use type in the research area within a certain time period, and its formula is as follows:

$$\mathrm{K}=\frac{{\mathrm{A}}_2-{\mathrm{A}}_1}{{\mathrm{A}}_1}\times \frac{1}{{\mathrm{T}}_2-{\mathrm{T}}_1}\times 100\%$$

(1)

where K is the dynamic degree of single land use; ${\mathrm{A}}_1$ and ${\mathrm{A}}_2$ are the areas of a certain land use type in the research area at the beginning and the end of the research period, respectively; and ${\mathrm{T}}_2-{\mathrm{T}}_1$ indicates the research period for a certain land use.

2.3.2. CA–Markov Model

In this study, the CA–Markov model was used to simulate and predict land use in the study area. The cellular automata (CA) model focuses on the simulation of the spatial changes in land use and the Markov chain (Markov) can effectively calculate the change in area of a certain land use type. The CA–Markov model, which combines the advantages of the two, can accurately predict the land use type, area, and spatial location [41].

The cellular automata model (CA) is based on transformation rules, and the cells carry out the interaction between local cells according to the transformation rules, thereby guiding the global cells to change, and the formula is:

$${\mathrm{S}}_{\mathrm{t}+1}=\mathrm{f}({\mathrm{S}}_{\mathrm{t}},\mathrm{N})$$

(2)

where S is the set of states of the cell, t and t + 1 are the two time points before and after, respectively, N is the neighborhood size of the cell, and $\mathrm{f}$ is the transition rule between cells.

The Markov chain model (Markov) can calculate the probability and quantity of land use transitions based on the information statistics of the land use state before and after the land. The formula is:

$${\mathrm{N}}_{\mathrm{t}+1}={\mathrm{N}}_{\mathrm{t}}\times {\mathrm{P}}_{\mathrm{ij}}$$

(3)

where ${\mathrm{N}}_{\mathrm{t}+1}$ is the land use state at the time t + 1, ${\mathrm{N}}_{\mathrm{t}}$ is the land use state at the time t, and ${\mathrm{P}}_{\mathrm{ij}}$ is the land use transfer matrix.

2.3.3. Analysis of Driving Factors of Land Use Change Based on Random Forest Regression

Random forest is an ensemble algorithm of decision trees. Compared with other statistical models, it is characterized by the ability to process data with a large sample size and evaluate the importance of sample variables [42]. Its principle is as follows: take the land use change as the spatially dependent variable, select the driving factors as the spatially independent variables, determine the contribution of each spatially independent variable, and sort the importance of the driving factors to explain the role of the driving factors on land use change. Taking the driving factor x_a as a variable, the formula of its characteristic weight value ${\mathrm{score}}_{\mathrm{a}}$ is

$${\mathrm{score}}_{\mathrm{a}}=\left[\sum _{\mathrm{b}=1}^{\mathrm{n}}\left({\mathrm{MSE}}_{\mathrm{b}}-{\mathrm{MSE}}_{\mathrm{ab}}\right)/\mathrm{M}\right]/{\mathrm{S}}_{\mathrm{E}}$$

(4)

where a is the number of selected driving factor variables, M is the number of samples, ${\mathrm{MSE}}_{\mathrm{b}}$ is the mean square error of the ${\mathrm{b}}_{\mathrm{th}}$ sample, and ${\mathrm{S}}_{\mathrm{E}}$ is the standard error.

In ArcGIS 10.8, the land use data of 2010 and 2020 were intersected to obtain a land use change map of Hangzhou in the past 10 years. The parcels with changed land use were assigned a value of 1, and the parcels with unchanged land use were assigned a value of 0. In this study, a total of 16 indicators from the 5 dimensions of geography, water source, location, population economy, and landscape, were selected as the land use driving factors (

Table 1. Driving factors of land use change.

Factor	Indicator	Indicator Content
Natural conditions	Geographical factors	Elevation, slope, and soil texture (clay content, silt content, and sand content)
	Water factor	Distance to water sources
Human disturbance	Location factors	Distance to subway stations, distance to a railway, distance to the highway, distance to a first-tier road, distance to a second-tier road, distance to the administrative center at the provincial and municipal levels, and distance to the administrative center at the township and subdistrict levels
	Demographic factors	Low light intensity at night and poor GDP
Environmental response conditions	Landscape factor	Vegetation coverage (fvc)

). The raster data atlas of each factor was normalized (Figure 2) to calculate the sum of the assigned value of land use change and the assigned value of each driving factor in a unit grid. A total of 35,983 sets of sample data were obtained, and the random forest model was used to analyze the feature weights.

2.3.4. Setting of the Multi-Scenario Development Model

The multi-criteria evaluation (MCE) module in IDRISI software creates a suitability atlas by setting several suitability images, and its evaluation criteria are the conditional factors of land use type conversion, including restrictive and constrained conditional factors. In this study, the MCE module was used to create a suitability atlas based on the driving factors of land use change in Table 1. With the 2020 land use map as the image of the base year, the land use images of the study area in 2030 under the four scenarios of ecological development, natural development, rapid development, and planned development were simulated (

Table 2. Scenario development mode setting.

Scenario Mode	Scenario Settings
Natural development	Land use evolves based on historical patterns and land is transformed according to the suitability standards set in the basic suitability atlas based on the probability of conversion between various land uses from 2010 to 2020.
Ecological development	Using the ecologically safe zone as a restrictive area, strict control measures are implemented to limit land development within the ecologically safe zone. Land conversion within the ecologically safe zone, except for ecological use, is also restricted.
Rapid development	As urban expansion maintains a high speed, the probability of construction land expansion increases. Economic factors are given an important role in driving land transformation, by increasing the weight of land suitability indicators such as GDP growth and nighttime lighting.
Planned development	The development model is based on overall urban planning, which restricts the expansion of construction land beyond the urban development boundary and reduces the probability of conversion of construction land to ecological use. The development of various types of land relies on urban planning to increase the probability of transforming unused land into farmland, forest land, grassland, water bodies, and developed land.

). To understand the spatial distribution of ESV more intuitively, a 300 m × 300 m square fishing net was created in ArcGIS to calculate the sum of ESV in each grid unit and generate the spatial distribution map of ESV in the study area in the base year (2020) and in 2030 under the four simulated scenarios.

2.3.5. Calculation of the ESV

According to the Hangzhou Statistical Yearbook, the annual output of staple food crops in Hangzhou in 2020 was 508,643 tons, the sown area was 90,840 hm², and the average purchase price of food staples was 2.52 CNY/kg. The economic value of grain crops in the farmland ecosystem per unit area in Hangzhou was 2015.76 CNY·hm². Its formula is as follows:

$${\mathrm{E}}_{\mathrm{n}}=\frac{1}{7}\sum _{\mathrm{i}=1}^{\mathrm{n}}\frac{{\mathrm{q}}_{\mathrm{i}}{\mathrm{p}}_{\mathrm{i}}}{\mathrm{M}}$$

(5)

where ${\mathrm{E}}_{\mathrm{n}}$ is the economic value of food production services provided by the farmland ecosystem per unit area of the study area (CNY/hm²), n is the type of staple food crops in the study area, ${\mathrm{q}}_{\mathrm{i}}$ is the price of crop i (CNY/kg), ${\mathrm{p}}_{\mathrm{i}}$ is the total output of crop i (kg), and M is the total area of the n types of crops (hm²).

The equivalence factor method assumes that each unit area of the ecosystem acts as a functional unit that provides ecosystem services and products [34]. Drawing on a previous study about the value assignment of land equivalent factors in Zhejiang Province, this study revised the ESV equivalent table. By considering the economic value of grain crops per unit area within the farmland ecosystem and the biomass factor of Zhejiang Province [43,44], we obtained the table of coefficients for ESV in Hangzhou (

Table 3. Ecological service value per unit area of an ecosystem (CNY/hm²) in the main urban area of Hangzhou City.

Primary Classification	Secondary Classification	Cultivated Land	Forestland	Grassland	Water Area	Construction Land	Unused Land
Supply services	Food production	3920.27	895.79	827.80	2838.17	20.16	17.75
	Raw material production	869.19	2057.69	1218.07	815.98	0.00	53.20
	Water supply	−4629.80	1064.34	674.08	29,410.73	−15,138.35	35.50
Conditioning services	Gas regulation	3157.49	6767.29	4280.92	2731.77	−4878.15	230.61
	Climate regulation	1649.72	20,248.73	11,317.29	8124.30	0.00	177.41
	Environmental purification	478.92	5933.61	3736.93	19,689.95	−4958.77	727.28
	Hydrological regulation	5303.88	13,250.82	8289.89	362,720.70	0.00	425.72
Support services	Soil conservation	1844.83	8239.61	5215.18	3299.39	40.32	266.06
	Maintenance of nutrient cycling	549.92	629.73	402.08	248.36	0.00	17.75
	Maintenance of biodiversity	603.13	7503.47	4742.14	9046.74	685.37	248.36
Cultural services	Provision of aesthetic landscapes	266.06	3290.54	2093.14	6705.24	20.16	106.45
Total ESV		14,013.60	69,881.62	42,797.52	445,631.34	−24,209.26	2306.09

).

3. Results

3.1. Characteristics of Land Use Evolution in the Study Area from 2000 to 2020

The overall land use in the study area from 2000 to 2020 showed a significant decrease in cultivated land and a rapid expansion of construction land (

Table 4. Land use area proportion and dynamics of various types of land use in the study area from 2000 to 2020.

Land Use Types		Cultivated Land	Forestland	Grassland	Water Area	Construction Land	Unused Land
2000	Area/hm2	174,196.17	83,215.80	1509.39	34,014.78	40,308.03	221.94
	Percentage/%	52.24	24.95	0.45	10.20	12.09	0.07
2010	Area/hm2	157,813.20	82,359.99	1485.45	26,835.66	64,489.23	482.58
	Percentage/%	47.33	24.70	0.45	8.05	19.34	0.14
2020	Area/hm2	135,799.74	81,964.62	1553.58	26,445.24	87,224.85	478.08
	Percentage/%	40.72	24.58	0.47	7.93	26.16	0.14
Area change from 2000 to 2020/hm2		−38,396.43	−1251.18	44.19	−7569.54	46,916.82	256.14
Rate of area change from 2000 to 2020/%		−22.04%	−1.50%	2.93%	−22.25%	116.40%	115.41%
Dynamic of single land use/%		−1.10%	−0.08%	0.15%	−1.11%	5.82%	5.77%

). The land use change in the study area was concentrated and significant around the urban construction land in the central part (Figure 3). The terrain in the south and west was dominated by mountains and forests, and the changes were discrete and insignificant. A large number of water areas in the east were converted into cultivated land, where the changes were significant. Amongst all land use types, cultivated land accounted for the largest proportion, and its land use area decreased by as much as 38,396.43 hm², with a reduction rate of 22.04%. The area of forestland declined gently, with a reduction rate of 1.5%. Grassland accounted for the second smallest proportion (after unused land), and its total area fluctuated at approximately 1500 hm², showing modest growth with fluctuations. The reduction in water areas, 7569.54 hm², was second only to that in cultivated land amongst all land use types, with a reduction rate comparable to that of cultivated land. The area of construction land in Hangzhou increased from 40,308.03 hm² in 2000 to 87,224.85 hm² in 2020, with a growth rate of 116.4%. The area of unused land experienced a relatively large increase from 2000 to 2010 and remained basically unchanged from 2010 to 2020.

3.2. Characteristics of Variation in the ESV in the Study Area from 2000 to 2020

From 2000 to 2020, the ESV in the study area decreased by CNY 51.32 × 10⁸, with the decrease being more significant from 2000 to 2010 compared to from 2010 to 2020. As can be seen in

Table 5. Changes in the ESV of various types of land use in the study area from 2000 to 2020.

Land Use Type		Cultivated Land	Forestland	Grassland	Water Area	Construction Land	Unused Land	Total
2000	ESV/CNY 108	24.41	58.15	0.65	151.58	−9.76	0.0051	225.04
	Proportion/%	10.85	25.84	0.29	67.36	−4.34	0.0023	100.00
2010	ESV/CNY 108	22.12	57.55	0.64	119.59	−15.61	0.0111	184.29
	Proportion/%	12.00	31.23	0.34	64.89	−8.47	0.0060	100.000
2020	ESV/CNY 108	19.03	57.28	0.66	117.85	−21.12	0.0110	173.72
	Proportion/%	10.95	32.97	0.38	67.84	−12.16	0.0063	100.00
ESV variation/CNY 108		−5.38	−0.87	0.02	−33.73	−11.36	0.0059	−51.32

, these data indicate that forest land and water contributed the most to the ESV in the study area. Grassland and unused land, given their relatively small proportions, exhibited relatively lower ESVs in the study area (Table 5). With urban expansion, the large increase in construction land also led to an increasingly clear negative effect on the ESV. The proportions of the ESV of cultivated land and forest land, despite their reduction in area, increased in the overall ecosystem. The ESV of the water area decreased from CNY 151.58 × 10⁸ to CNY 117.85 × 10⁸ over the two decades, resulting in a significant decline in the overall ESV in the study area.

3.3. Analysis of Land Use Evolution and Ecological Value under Multi-Scenario Simulations

3.3.1. Analysis of Driving Factors of Land Use

This study, based on the random forest model, quantitatively analyzed the importance of each driving factor to the land use change in the study and used the characteristic weight as the weight of the suitability map of suitability factors. The land use change in the study area was greatly affected by the water factor (distance to water sources), location factors (distance to the administrative center at the provincial and municipal levels, distance to the administrative center at the township and subdistrict levels, distance to first-tier roads, and distance to subway stations) and demographic and economic factors (GDP growth) (

Table 6. Influence weights of various driving factors on land use change.

Indicator	Indicator Content	Cultivated Land	Forest Land	Grassland	Water Area	Construction Land	Unused Land	Total
Geographical factors	Clay content	6.00%	6.30%	—	2.60%	6.60%	19.60%	6.50%
	Silt content	4.50%	5.70%	—	1.90%	3.30%	13.50%	3.30%
	Sand content	4.30%	6.00%	—	1.60%	3.00%	13.90%	3.90%
	Slope	3.70%	3.00%	—	8.90%	3.20%	4.20%	3.60%
	Elevation	3.90%	4.50%	—	5.70%	3.10%	2.00%	3.20%
Water factor	Distance to water sources	10.50%	6.10%	19.30%	8.60%	8.40%	2.20%	8.30%
Location factors	Distance to the administrative centre at the provincial and municipal levels	5.80%	10.10%	14.70%	7.60%	8.50%	8.20%	10.40%
	Distance to the administrative centre at the township and subdistrict levels	4.80%	8.40%	29.50%	3.80%	8.50%	5.90%	7.10%
	Distance to the highway	12.30%	10.20%	4.00%	7.10%	7.10%	4.20%	6.80%
	Distance to first-tier roads	6.60%	5.80%	—	6.70%	8.90%	2.60%	9.60%
	Distance to second-tier roads	11.10%	2.90%	13.20%	7.30%	3.70%	5.10%	3.50%
	Distance to a railway	5.30%	3.70%	13.80%	14.70%	8.70%	4.40%	6.50%
	Distance to subway stations	6.10%	6.10%	1.30%	4.20%	8.50%	9.30%	9.10%
Demographic factors	Poor GDP	8.60%	7.50%	—	4.90%	9.00%	—	7.50%
	Low light intensity at night	3.10%	3.50%	1.30%	5.50%	5.50%	1.20%	5.80%
Landscape factor	Vegetation coverage (fvc)	3.40%	10.20%	3.00%	8.80%	4.10%	3.80%	4.90%

). These six driving factors accounted for 52% of the total weight. To be precise, the distance to the administrative center, the distance to first-tier roads, and the distance to the subway station are the most important factors, as each shows a higher weight of influence than the water factor.

3.3.2. Model Accuracy Verification

With the land use data in 2010 as the base year data, parameters, including the determined suitability atlas and the conversion probability, were set in the CA–Markov model to obtain the simulated land use results in 2020. The CrossTab module in IDIRISI software was used to compare the actual 2020 land use map with the simulated 2020 land use results. The kappa value of the model was 0.9231, indicating the high accuracy of the model simulation, which enabled it to accurately predict future land use.

3.3.3. Characteristics of the ESV in the Study Area under Multiple Scenarios

For the ESV calculated based on the equivalent factor, the type and area of land use are the key factors in determining the value (

Table 7. Area and ESV of various land use types in the study area in 2030 under four scenarios.

Scenario Mode	Area and ESV	Arable Land	Forest Land	Grassland	Water Land	Build-Up Land	Unused Land	Total
The base year	Area/hm2	135,799.74	81,964.62	1553.58	26,445.24	87,224.85	478.08	333,466.11
	ESV/CNY 108	19.03	57.28	0.66	117.85	−21.12	0.011	173.71
Natural development scenario	Area/hm2	116,340.84	74,744.82	3284.28	24,290.64	114,281.55	523.98	333,466.11
	ESV/CNY 108	16.30	52.23	1.41	108.25	−27.67	0.012	150.53
Economic development scenario	Area/hm2	111,105.54	73,929.42	1990.98	23,945.04	122,062.95	432.18	333,466.11
	ESV/CNY 108	15.57	51.66	0.85	106.71	−29.55	0.010	145.25
Planned development scenario	Area/hm2	134,139.24	77,309.82	2004.48	26,005.14	93,709.62	297.81	333,466.11
	ESV/CNY 108	18.80	54.03	0.86	115.89	−22.69	0.007	166.90
Ecological development scenario	Area/hm2	127,717.02	82,808.91	3307.50	25,467.84	93,760.74	404.10	333,466.11
	ESV/CNY 108	17.90	57.87	1.42	113.49	−22.70	0.009	167.99

and Figure 4). In general, the reduction in ESV will be more noticeable under the natural and rapid development scenarios, with a 13.34% reduction for the former and a 16.38% reduction for the latter. Construction land, by occupying a large amount of cultivated land under the natural development scenario, will increase by 27,056.7 hm² compared with the 2020 data. Under the rapid development scenario, even more cultivated land and ecological land will be occupied by construction land, allowing the construction land area to reach 122,062.95 hm².

The spatial distributions of ESV under four simulated scenarios in the base year (2020) and 2030 were mapped using the natural breakpoint method to allocate values into six value intervals (Figure 5). From the spatial distribution of ecological service values in the study area in the base year (2020), the ESV in the study area was generally low in the middle and high in the south and west, which was highly consistent with the distribution of urban land development in space. Areas with a high ESV were mainly distributed in forests in the western and southern regions and watersheds in the central and eastern regions. The central region was the core area of urbanization development in the study area, and because the space dominated by construction land exerted a negative effect on the ESV, the ESV of this region was relatively low.

Compared with the ESV in the base year, under the natural development scenario, the overall ESV will decrease. The area of the Grade I region will increase significantly, that of the Grade IV region will shrink considerably following its original boundary, and the band-shaped central area of the Grade VI region will also decrease. A comparison of the spatial distribution of ESV under the four simulated scenarios showed that the overall ESV under the ecological and planned development scenarios would be significantly higher than those under the natural and rapid development scenarios. The ecological development scenario will contribute to an increase in Grade IV regions in the west and south and will effectively constrain the area of Grade I regions in the central part compared to the natural development scenario. The reason for this is that with forest land and cultivated land under protection, the area of forest land will increase slightly. The overall ESV under the planned development scenario will experience a similar variation to that under the ecological development scenario. Its reduction rate will decline compared with the base year, which will be a result of the effective restriction that will be placed on the expansion of urban construction land and the consequently decreased encroachment on forest land and cultivated land.

4. Discussion

4.1. Change in Land Use and Ecosystem Services

In this study, the greatest changes in land use in the main urban area of Hangzhou from 2000 to 2020 were the rapid expansion of urban construction land and the sharp reduction in cultivated land and water areas. Similar studies have also shown that transforming cultivated land into construction land is the main mode of land use transformation in urban areas [45,46,47]. This is mainly due to China’s stricter control of ecological regions, the fact that cultivated land is often an easily overlooked type of land (except for permanent basic farmland), and the fact that the expansion of urban construction land is generally based on sacrificing cultivated land. Areas with the most dramatic land use changes were distributed on both sides of the Qiantang River (Figure 3). This indicates that Hangzhou has seen the greatest development intensity in urban areas along large rivers over the past 20 years. In the process of urban development, the river system often plays a significant role in shaping the spatial structure of cities. As a result, urban development strategies that emphasize river-supported development and cross-river expansion are relatively common. The recent development of Hangzhou City aligns with this phenomenon. This study also examined the factors affecting land use changes; the weights of slope and elevation in geographical factors were relatively low, which is contrary to other studies [36,48,49,50]. This discrepancy may be attributed to the dominance of urban construction land expansion in the study area, as well as land use changes primarily occurring in plains characterized by gentle slopes and low elevations.

Additionally, the shift in ESV is similar to the historical land use change, exhibiting a continuous decline from 2000 to 2020. This aligns with the findings of Xia et al., who posited that the value of urban ecosystem services in rapidly urbanizing areas tends to exhibit a persistent downward trend [51]. A previous study in Hangzhou has also shown that habitat quality in the Hangzhou metropolitan area is continuing to decline [38]. We also found that areas with a high ESV were mainly concentrated in the western and southern forest areas, as well as the Qiantang River Basin. This finding suggests that areas of high value ecosystem services are mainly located in the peripheries of cities. This is similar to the results of other studies that found the greatest decline in habitat quality in downtown areas [38] and found that the urban ecosystem health significantly deteriorated [19]. Changes in the ESV are a direct reflection of land use changes, and in the context of rapid urbanization, dramatic land use changes have become a major factor in weakening and destroying the value of urban ecosystem services [52].

Simultaneously, we found that the decline in the water ESV was the main contributor to the decrease in overall ESV, and the negative effect of the expansion of construction land on the ecosystem service value was significant. Our findings are consistent with those of Tan et al., underscoring the crucial role that water areas play in ecological and land use changes [53]. Land use types such as water bodies and wetlands are not only vital for hydrological regulation and water resource supply but also serve as a critical foundation for sustainable urban development [54,55]. However, their significance can be easily overlooked during urban development, highlighting the importance of effective land use management of water-related areas. Our results are consistent with previous studies, which also found that urbanization induces significant land cover and ecosystem service changes [36,48,49,50]. Obviously, water bodies and woodlands are often the main sources of value for ecosystem services in urban areas [33]. Consequently, it is necessary to strengthen the protection of ecological lands, such as forest land and waters, to promote sustainable development in the process of rapid urbanization.

4.2. Land Use and Ecosystem Services under Multi-Scenario Simulations

In this study, the CA–Markov model was used to simulate the land use changes in Hangzhou in 2030 under four scenarios, planned development, ecological development, natural development, and rapid development, and the corresponding ESVs were predicted. On this basis, the spatial distribution and clustering characteristics of the ESV were analyzed. The accuracy of simulating urban expansion can be significantly improved by selecting the weights of multiple source factors based on random forest regression compared to the traditional CA–Markov model based on natural evolution. With an accuracy of 92.31%, our model surpasses other similar models that have an accuracy of 80% [47]. Urban expansion is often complex and dynamic [56]; it is essential to consider location, socio-economic status, and ecological environment as multiple factors. A limitation of this study is that the multi-source data employed may not be universally applicable to all research areas. Future studies could further refine and select appropriate factors to address this constraint.

Based on a CA–Markov multi-scenario simulation, the future land use pattern (2030) under four scenarios remains largely consistent with the base period (2020); the differences in future urban land use spatial development are mainly reflected in local areas. This phenomenon is due to the relatively saturated land development and stable urban spatial structure in the main urban area of Hangzhou. Scholars have also found similar results in the simulation of land use in Nanping City, China [57]. Our results also showed that the planned development scenario and the ecological development scenario could effectively guarantee a high ESV. In terms of the numeric value, the decreases under both scenarios are less than 4%, which can meet the needs of urban development whilst protecting the ecological environment. Related studies in Wuhan and Shenzhen also found that the ecological development scenario is one of the best options for urban development [49,58], indicating that controlling important ecological areas from an ecological priority perspective can effectively improve the overall ecological quality level of the city. In terms of the spatial distribution of the values, compared with the natural development scenario, low-value areas will shrink significantly and areas with sharp value changes will also decrease. The ecological development scenario can even slightly increase the ESVs of forest land and grassland areas.

Our results indicate that the overall ESV will also decrease under the natural development scenario. The area of urban construction land will increase sharply, and agglomeration will become noticeable. A study in Wuhan also found that under a natural development scenario, built-up land will continue to grow significantly in the future while ecological land will decrease [31]. Interestingly, under natural development, besides the expansion of urban construction land, the grassland area would also increase to some extent. A previous study suggested that the natural development scenario could protect the ecological environment and ensure urban–rural development. However, our research results contradict this notion, as we found that following the natural development trend would lead to significant damage to the ecological environment [59]. This may be the result of different development conditions and urban development strategies in the two different cities. There are also studies that have found that if the current pace of urbanization development is maintained, other types of land use besides urban and grassland areas will see a decrease in area in the future [47]. The influence of the central city will further expand to a larger area, and the expansion of construction land will encroach on the original cultivated land and ecological land. In addition, the shrinking of the Qiantang River on both sides in the middle of the study area will substantially aggravate the decline in the ESV. Under the rapid development scenario, urban expansion, such as the expansion of construction land and the fragmentation of cultivated land, causes direct and indirect losses in natural forest areas [59,60]. Hence, the overall ESV will be largely lower than those under the other scenarios, and the value of ecological land will decrease significantly.

Under the guidance of rational planning and coordination, such as the planned development scenario and the ecological development scenario, the urban construction land will be effectively managed to ensure the spatial pattern of ecological security within the study area and promote the sustainable development of the city. Our research offers valuable insights, particularly for addressing issues arising from rapid urbanization in developing countries such as India [61] and Thailand [62]. However, each country needs to develop suitable indicators and corresponding ecological assessment equivalence tables based on local and regional characteristics and development conditions. In addition, besides the ESV indicators, recent studies on the spatial correlation between ecosystem health (ESH) and human disturbance have provided insights into sustainable urban landscape development [63,64,65]. Ecosystem health indicators may serve as suitable metrics for further evaluating future urban development scenarios.

5. Conclusions

This study explored the evolution of land use in the main urban area of Hangzhou over the past twenty years and its impact on ecosystem services in the context of rapid urbanization. We assessed the importance of multi-source land use driving factors using the random forest model. The rapid expansion of urban construction land and the sharp decrease in water area were found to be the main factors contributing to the decline in the overall ESV. During urban development, a careful balance of various land use types should be considered. When protecting and cultivating ecological land, attention should also be given to developing agricultural land. Urban development and construction should be focused in areas with a low ESV. These research findings can play a crucial role in land use planning decisions by balancing and coordinating urban growth with sustainable ecological development. However, urban expansion is influenced by numerous factors, and the selection of factors for land expansion in this paper is limited by the available data acquisition methods, which can be improved and optimized in further research.

The CA–Markov model was employed to simulate land use changes under four scenarios: natural development, ecological development, planned development, and rapid development. The selection of land expansion driving factors based on multi-source data and random forest regression can accurately simulate the future development of the city. Moreover, we evaluated the ESV under different urban development scenarios. The planned development and ecological development scenarios can effectively ensure a high ESV. This interesting finding suggests a trade-off between urbanization and ecological land use under conscious ecological development planning. Urban development does not always come at the expense of the ecological environment. The results of this study will provide scientific guidance to government departments and local decision makers in future land use planning.