Climate data and emission scenarios.

To predict the distribution of future climate forcings, this study uses two Shared Socioeconomic Pathways (SSPs) from the latest CMIP6 updates: RCP 4.5 and RCP 8.5. RCPs are quantitative descriptions of alternative environmental development pathways for the 21st century, combining climate change projections with assumptions about mitigation or adaptation policies to generate comprehensive climate change scenarios. The RCP 4.5 scenario (4.5 W/m²) represents a moderate path for greenhouse gas emissions. In this scenario, environmental systems face some degree of degradation, and climate protection measures are assumed to be implemented [29]. The RCP 8.5 scenario (8.5 W/m²), on the other hand, represents the extreme case of rapid socioeconomic development, leading to the highest path for future greenhouse gas emissions [30]. In this scenario, global markets become increasingly integrated, fostering innovation and technological progress. However, socioeconomic development heavily relies on intensified exploitation of fossil fuel resources, with a higher proportion of coal, and an energy-intensive lifestyle continues globally [31]. This study selects RCP 4.5 and RCP 8.5 because they represent the moderate and worst-case scenarios for greenhouse gas emissions, respectively.

Land use change scenarios.

MOP is an open and flexible method, which mainly consists of three parts: decision variables, objective function, and constraints, and takes into account the manager’s expectations of relevant planning by setting proper objective optimization functions and constraints [32]. To investigate the trade-offs/synergies between ESs on the basis of simulated LUC under different scenarios, three LC scenarios of NIS, EPS and EDS were constructed herein in the light of the characteristics of LUC and compliance to the requirements of macro-policy regulation.

1. NIS: in this this scenario, the demand on LC by 2030 followed the same trend as the historical data, is not affected by policies, and serves as a reference baseline for other scenarios. The amount of various types of LC in 2030 under this scenario was calculated using the LUC matrix (2010–2020), which was generated by Markov model, and with various types of LC in 2020 as the baseline.
2. EPS: this scenario indicates that priority is given to the protection of ecological land (woodlands, grasslands, waters). In this scenario, the ecological benefits provided by each type of LC are maximized through planning and management, and the function for assessing ecological benefits is expressed as follows.

(1)

In equation (1): f(x) denotes the total ecological benefit of all types of LC, ESV_i denotes the ecological benefit per hectare of the type i land use (in million CNY), and x_i denotes the area of the type i LC. In this paper, the economic value was obtained by correcting the economic value generated by the yield per unit area of the main crops in CYUA, with reference to the revised Equivalent Value per Unit Area of ESs in China by Xie et.al as the basis [33]. The objective optimization function of the EPS is: .

1. EDS: this scenario represents a great effort on economic development effort, so as to maximize the economic benefits to humans from various types of land, and the function for assessing economic benefits is given in equation (2) below.

(2)

In equation (2), f(x) denotes the total economic benefit, EC_i denotes the economic benefit per hectare of type i LC (in million CNY), and x_i denotes the area of type i LC. The average land economic value is sourced from the Yunnan Statistical Yearbook, where the total value of output for forestry, animal husbandry, and fishery corresponds to the economic benefits achieved by woodland, cropland, grassland, and waters, respectively. The GDP of the secondary and tertiary sectors corresponds to the economic benefits of built-up land, and the unused land was set based on the previous study (Ling et al., 2024). The objective optimization function of the EDS scenario is: .

The constraints are as follows.

1. 1). Total land area constraint: It is necessary to ensure that the total area of all land use types remains unchanged. As shown in equation (3) below.

(3)

1. 2). Total population constraint(P): this constraint ensures that the total population living on the land should not exceed the land’s carrying capacity for human activities. The upper limit of population was set according to the policy laid down in the 《Central Yunnan Urban Agglomeration Development Plan》(https://www.yn.gov.cn/), and the existing population is the lower limit. Therefore, the mathematical expression for the constraint on the total population can be written as equation (4).

(4)

In the formula, P represents the total population of the study area, measured in ten thousand people.

1. 3). Cropland area constraint: the food yield of farmland should be at least the food demand of the population. Herein, the per capita food demand was assumed as 141.2 kg; the food self-sufficiency rate is 114.11 percent; the food production per hectare of the cropland was 4316.24 kg; the proportion of crops planted was 67 percent; and the multiple cropping index is 1.11. The cropland constraint can be written as follows:

(5)

1. 4). Forest area constraint: The area of forest land in the CYUC showed a downtrend from 2000 to 2020, which was expected to continue in the future. According to the ‘Yunnan Province Ecological Civilisation Construction Leading Plan (2021-2025)’ and the ‘Development Plan for the Central Yunnan Urban Agglomeration (CYUC)’, the local government scheduled to implement more ecological protection plans. Herein, using the woodland area in 2020 as the baseline, the land area (2020) ±1% was determined as the upper and lower bounds, respectively; and then the constraint can be written as:

(6)

1. 5). Grassland area constraint: a downtrend of grassland area was identified during the study, so using the grassland area projected by the NIS scenario as a baseline, the projected area ±1% was determined as the upper and lower bounds, respectively; and then the constraint can be written as:

(7)

1. 6). Water area constraint: Large areas of water in CYUA have been reclaimed for urbanization and crop production. To prevent further shrinkage of lake area, the local government has carried out the Dianzhong Diversion Project for water resources protection. As a result, the water area will be on an upward trend in the future. Therefore, the water area in 2020 was set as the lower bound of constraint, and the water area in 2030, which was calculated according to the natural growth rate of the water area from 2015 to 2020, was considered as the upper bound of the constraint, then the constraint can be written as:

(8)

1. 7). Built-up land constraint: the built-up land should satisfy the population demand and the per capita building land in CYUA was 0.0109 ha. Therefore, the constraint can be written as:

(9)

However, the built-up area was projected to be 332,112.951 hectares after 2020 and 367,072.209 in 2030, based on the current trend of growth in built-up land in CYUA. Then the constraint can be detailed as:

(10)

1. 8). Unused land area constraint: since CYUA has been reclaiming unused land for long time due to socio-economic development, it is expected a smaller area of unused land in 2030 than that of 16,049.16 ha in 2020. However, since CYUA is a plateau zone, the utilization rate of unused land is low. Therefore, the unutilized land will not be less than 15849.3 ha. Then, the constraint can be written as:

(11)

1. 9). Other constraints: in view of the ecological protection measures proposed in the ‘Yunnan Province Ecological Civilisation Construction Leading Plan (2021-2025)’, the study concluded that the area of ecological land (woodland land, grassland, and water) in CYUA will be at least 75% of the total area by 2030. Therefore, the constraints can also be written as:

(12)

PLUS model.

As a raster data-based meta cellular automata (CA) model, the PLUS model can be used for simulation of patch-scale change in LULC [34]. The model integrates a land expansion analysis strategy-based rule mining framework (LEAS) and a multi-type random patch seeding-based CA model (CARS). Using the random forest classification (RFC) method, the proper probability for various types of LULC expansion can be obtained by mining the driving factors. It consists of two modules: (1) a rule mining framework based on Land Expansion Analysis Strategy (LEAS); (2) a Cellular Automata (CA) model based on multi-type random patch seeding (CARS). The steps are mainly as follows.

1. By using LEAS, the change patches of different land use types over a period can be analyzed to obtain the transition rules for all land use types. The transition rules for each land use type are mined and transformed into a binary classification problem, which allows for the calculation of the change probability and inertia probability for each land use type. Additionally, LEAS can employ the Random Forest Classification (RFC) algorithm to explore the relationship between the growth of different land use types and multiple driving factors.
2. The CARS module is a Cellular Automata (CA) model based on multi-type random patch seeding. The CA model is a scenario-driven land use simulation tool that integrates both “top-down” (i.e., overall land use demand) and “bottom-up” (i.e., regional land use competition) effects. During the simulation, land use demand influences land use competition through adaptive coefficients, thereby driving the land use amount to meet future demand (obtained through the MOP model in this study).

LULC-RCP multi-scenario model.

According to the needs of the study, six modelling scenarios, as shown in Table 2, were established by coupling the future climate models (RCP4.5, RCP8.5) and the three land use scenario data (NIS, EPS, EDS) in 2030. Among them: in Scenario 1, the data on LC under NIS scenario and the climate of RCP4.5 scenario in 2030were used as input data; in Scenario 2, the data on LC under NIS scenario and the climate of RCP8.5 scenario in 2030 were used as input data; in Scenario 3, the data on LC under EPS scenario and the climate of RCP4.5 scenario in 2030 were used as input data; in Scenario 4, the data on LC under EPS scenario and the climate of RCP8.5 scenario in 2030 were used as input data; in Scenario 5, the data on LC under EDS scenario and the climate of RCP4.5 scenario in 2030 were used as input data; and in Scenario 6, the data on LC under EDS scenario and the climate of RCP8.5 scenario in 2030 were used as input data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Future LULC-RCP multi-scenario models.

https://doi.org/10.1371/journal.pone.0324015.t002

1. WY.

The WY of the waters characterizes the regional water supply capacity, and during the experiment, in the InVEST model, the WY module was used for WY evaluation. The WY module is on the basis of the law of water balance, taking into account the factors of vegetation cover, climate, and soil texture, etc., and the primary formula is as follows:

(13)

Where: denotes the W of each raster x (mm); and denote the actual annual evapotranspiration and annual precipitation of each raster x (mm), respectively [17].

2. CS.

In the InVEST model, the CS module divides the CS of EC into four parts: above-ground CS, below-ground CS, soil CS and dead organic matter CS. Using the module, the CS in the watershed was calculated herein with the following equation:

(14)

Where: denotes the total CS (t/hm²); denotes the above-ground CS (t/hm²); denotes the below-ground CS (t/hm²); denotes the soil CS (t/hm²); and C_dead_ denotes the dead organic matter CS (t/hm²). Herein, the data on carbon density for the various types of LC was sourced from the National Ecosystem Science Data Centre and certain literature [28,35,36]. The carbon density data were sourced from studies on various regions in China, thus the selected literatures should have a study area as close to or similar to that in this study as possible to prevent significant error due to data gap. In addition, carbon density varies depending on many factors like climate, soil properties and LC. Therefore, appropriate corrections should be made in accordance with the types of LC and the climate characteristics in different periods in CYUA. It has been reported a positive correlation of carbon density with annual precipitation and a weak correlation between carbon density and average annual temperature [37]. The specific calculation process refers to equations (3)–(9) in the literature [38], and the carbon densities in each period are shown in Annex 1.

3. HQ.

In the InVEST model, the HQ module has been employed for calculating the HQ of upper basin of the Yangtze River, using the following formula. In this paper, cropland and built-up land were used as habitat stress factors, and the values of the required parameters were assigned with reference to the existing literature [16].

(15)

(16)

where denotes the HQ of raster x; denotes the total threat level of raster x; K and Z denote scaling factors; denotes the HQ of LULC type j; R denotes the coercion factor; denotes the number of rasters occupied by the stress factor r; denotes the weight of the stress factor, within a range of 0–1; denotes the value of the coercion factor for raster y; denotes the coercion () of the raster y to the level of stress for raster x; denotes the level of accessibility for raster x; and denotes the sensitivity of type j habitat to the stress factor r.

4. SC.

Herein, the erosion model of Revised Universal Soil Loss Equation was used to estimate soil conservation in CYUC, using the following formula [39]:

(17)

Where: A denotes the soil conservation per unit area; R denotes the factor of rainfall erosivity; K denotes the factor of soil erodibility; LS denotes the factor of slope length; C denotes the factor of vegetation cover and management; P denotes the factor of soil and water conservation measures.

Spearman correlation analysis.

To detect the correlation between the ESs on the time scale, Spearman’s correlation analysis and significance test were performed based on the R platform using the ‘Hmisc’ package [40]. Positive and negative correlations indicate synergies and trade-offs between ESs, respectively; and the larger the absolute value of the correlation coefficient, the greater the relationship. The calculation formula is as follows [41]:

(18)

Where, represents the Spearman rank correlation coefficient, which indicates the correlation between variables and , with a value range of [−1, 1]. , represents positive correlation, , represents negative correlation, , represents no correlation. and represent the values of variables and for the i-th data point, where i = 1,2,…,n. and are the sample means of variables and , respectively.

Geographically Weighted Regression (GWR).

The ESs trade-offs/synergies were quantified spatially using GWR, and the formula is as follows [42]:

(19)

Where, denotes the geographic coordinates of the i^th raster; denotes the measured value of i^th dependent variable Y at the spatial location; denotes the measured value of the independent variable dataset at the spatial location; k denotes the total number of independent variables; denotes the constant at ; denotes the value of continuous function at the i^th grid point; and denotes the random error.

MOP-based forecasts of LC demand.

The analysis of LC data in CYUA between 2000 and 2020 indicated that the general trend of LUC in CYUA during the 20 years was a reduction in the area of cropland, woodland and grasslands, and a rise in the area of waters and built-up land, and a stable area of unused land (Supplementary Material, Appendix1 S1 Fig and Fig 3). According to the calculations following the aforementioned equations (1)–(13), the area of each type of land use under various scenarios in 2030 was obtained and compared with those in 2020, and it was found that: cropland decreases under the NIS and EPS scenarios, and increased in the EDS scenario; woodland land decreased under the NIS and EDS scenarios, and increased under the EPS scenario; grassland decreased to varying degrees under all three scenarios, with the most obvious decrease up to 69799.8 ha under the EDS scenario; waters increases in all three scenarios; and built-up land showed a similar trend to that of waters, with highest increase of 100826.6 ha under the EDS scenario, followed by an increase of 83346.93 ha under the NIS scenario. All the data is detailed in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Area of different types of land use (hm²) in CYUA in 2030 under different scenarios.

https://doi.org/10.1371/journal.pone.0324015.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Land use transfer matrix for CYUA, 2000-2020 (hm²).

https://doi.org/10.1371/journal.pone.0324015.g003

Multi-scenario simulation of LC based on MOP-PLUS model.

With various data from multiple sources, including data on LC, topography and socio-economy in CYUC in 2000, 2010 and 2020, the simulation of land use in CYUA in 2020 was performed by the PLUS model, and then a comparison between the simulation data (Supplementary Material, Appendix1 S2b Fig.)and the actual data in 2020(Supplementary Material, Appendix1 S2a Fig.) was carried out. The simulation data showed an overall accuracy of 91.97% and a kappa coefficient of 0.876, indicating that the model meets the accuracy requirements of this study.

Based on the land use maps from 2000 to 2020, we predicted the land use spatial distribution for 2030 under the NIS, EPS, and EDS scenarios (Fig 4). It is found that the dominated type of land use in CYUA under all the three scenarios is woodland, which is largely distributed in the highland mountainous regions in the western and southern areas, and is an important ecological barrier in Yunnan Province, with a proportion of approximately 49.08% of the total area; followed by grassland with an area proportion of approximately 26.84%, which is largely distributed in Kunming City, Qujing City, and Honghe Prefecture; cropland is largely distributed in blocks in the plains of CYUA, with an area proportion of approximately 20.23%; waters are mainly dominated by the six highland lakes distributed in the center of CYUA; built-up land is mainly centered in Kunming City, and the rest is distributed in the form of blocks in the urban centres and their surroundings; and the unused land accounts for a very small proportion of 0.2%, which is mainly dominated by the high altitude snow-covered mountains. Furthermore, the simulations of LC under the various scenarios indicated an increase in built-up land and waters in 2030 in comparison to that in 2020; specifically, the expansion of built-up land is mainly due to encroaching on cropland and forests, and the expansion of waters is due to spreading centered around the lakes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Spatial distribution of different land use scenarios for CYUA 2030.

https://doi.org/10.1371/journal.pone.0324015.g004

Analysis of changes in CYUA ESs for multi-scenario simulations.

The MOP-PLUS-RCP and InVEST models were coupled to predict the future (2030) changes of WY, HQ, SR, and CS services in the CYUA under six different scenarios, and the results are provided in Table 4 and Fig 5 and described as follows:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Number of simulated ESs in CYUA in 2030under various scenarios.

https://doi.org/10.1371/journal.pone.0324015.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Spatial distribution of ESs in CYUA ESs in 2030 under various scenarios.

https://doi.org/10.1371/journal.pone.0324015.g005

1. (1). WY and characteristics of its spatial distribution: under different scenarios, Scenario 6 was projected to have the highest WY (5312.13 m³/ha), followed by Scenario 2 (5300.90 m³/ha) and Scenario 4 (5288.22 m³/ha). For climate scenarios, the WY of RCP8.5 is higher than that of RCP4.5, which may be related to the high precipitation and low temperature of RCP8.5. For land use scenarios, the WY of EDS scenario was larger than that of the EPS and NIS2 scenarios, and the main explanation to this is the increase in cropland under the EDS scenario, which mitigates the impacts of reduced woodlands and grasslands. With respect to spatial distribution, it was roughly the same under the same climate scenario. For RCP4.5 (in Fig 5 S1, S3 and S5), the WY showed a trapezoidal distribution, with gradual increase from north to south; for RCP8.5 (in Fig 5 S2, S4 and S6), the distribution of WY was low in the northwestern region, and high in the eastern and southern regions. It should be noted that the lower reaches of the Yuan River in the south were always the main water-producing area in all scenarios.
2. (2). HQ and characteristics of its spatial distribution: From the average HQ of various scenarios, the effect of land use on HQ was greater than that of climate. Therefore, HQ showed no significant change under different climate scenarios. While for HQ under LC scenarios, it was ranked from high to low as follows: EPS (0.610)> NIS (0.606)> EDS (0.604), all of which are of high HQ, and this is closely correlated with the high forest coverage rate of the region. Therefore, due to more attention paid to ecological protection, which resulted in increased woodland, the EPS scenario showed the best HQ. With respect to spatial distribution, all six scenarios showed a same pattern of high in the western region and low in the centre. Changes in HQ were largely concentrated in the center, where is densely populated and highly urbanised, indicating that HQ was directly affected by the increase in built-up land.
3. (3). SR and characteristics of its spatial distribution: among the various scenarios, S4 showed the best SR, followed by S2 and S6. Compared with that in 2020, the SR of all the six scenarios has increased, with the EPS scenario showing the most significant increase. With respect to spatial distribution, the high SR was mainly located in the lower reaches of the Yuan River in the south, while the rest of the study area showed different low values, especially in the central area, with a downtrend from west to east. This indicates that the high vegetation cover in the south plays a positive role in soil conservation, while in the west, due to the high altitude difference and dense steep slopes, the soil and water conservation capacity is compromised even if there is a high vegetation cover. Although the terrain in the central region is relatively flat, most of the region is an urban area under development, so the soil conservation capacity is low, indicating that continued urban expansion has led to encroachment on ecological land.
4. (4). CS and characteristics of its spatial distribution: the order of CS under different scenarios is as follows: S4 > S2 > S6 > S3 > S1 > S5. It can be seen that the CS capacity under the RCP8.5 is significantly greater than that under the RCP4.5, and that the EPS scenario has the most prominent CS capacity among the various LC scenarios. From the perspective of their spatial distributions, the CS capacity and HQ of S6 showed a highly consistent overall pattern of high in the west and low in the centre. In addition, the CS reduction mainly occurs in the periphery of the central cities. The above findings suggest that urban development in the future will occupy a large number of lands with high CS, such as agricultural land and forests, resulting in decreased CS in the region.

Changes in ES trade-offs/synergies under different scenarios.

Using the Spearman correlation, the trade-offs/synergies between ESs under various scenarios was analyzed to explore the changes in relationship between ESs under various scenarios in 2020 and 2030. As shown in Table 5, the results indicated a positive correlation between WY-SR, HQ-SR, HQ-CS and SR-CS, reflecting a synergistic relationship, and a negative correlation between WY-HQ and WY-CS, indicating a trade-off relationship, in both 2020 and 2030 across different scenarios. Compared to 2020, under the RCP4.5 climate model (S1, S3 and S5), the trade-off between WY-HQ and WY-CS in 2030 was weakened, while the synergistic relationship between WY-SR was strengthened, particularly in S5, where the increase was approximately 0.22. The synergistic effects between CS-SR and CS-HQ are weakening and increasing, respectively; under the RCP8.5 climate model (S2, S4 and S6), the synergistic relationship between CS-SR and CS-HQ were weakened and increased, respectively, which are in exactly the opposite way to that of the RCP4.5 model. It is noteworthy that the HQ-SR synergies showed different degrees of weakening in all six future scenarios. Under the same climate scenarios, the trade-offs between WY-HQ and WY-CS and the synergies between WY-SR, HQ-CS, and SR-CS were higher in the EDS scenario than in the EPS and NIS scenarios, and the HQ-SR synergy was highest in the EPS scenario.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. ES trade-offs/synergies under various scenarios.

https://doi.org/10.1371/journal.pone.0324015.t005

For the purpose of further elucidating the significant spatial heterogeneity of ESs trade-offs/synergies due to geographic differences in highland mountainous regions, the spatial layout of such trade-offs/synergies was obtained at the raster scale using GWR analysis. As presented in Fig 6, the results indicated that the trade-offs/synergies between four pairs of ESs were consistent in spatial distribution under all the six scenarios; SR-HQ spatial effects showed a weak synergy in general, but showed a weak trade-off in the highly urbanised region in the central area; SR-WY showed a synergy in areas distributed in blocks in Qujing City, Chuxiong City, and Honghe Prefecture, and showed a trade-off relationship in areas mainly distributed in Yuxi City; SR-CS showed a weak synergistic relationship in the whole area excepting the southwestern region, in where it showed a weak trade-off relationship; HQ-WY showed a strong synergy in areas largely located in the north and the south, and a weak trade-off relationship in areas mainly distributed in the west and the northeast and particularly evident in the central region; HQ-CS showed a weak synergy in the western and northeastern regions, and a strong synergistic relationship in the rest; WY-CS showed a weak synergy in the northern and northeastern regions and a relatively strong synergy in the south of Honghe Prefecture, and a weak trade-off in the southwest of Yuxi City, Kunming City, and the east of Qujing City.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Spatial distribution of ES trade-offs/synergies under different scenarios.

https://doi.org/10.1371/journal.pone.0324015.g006