Estimation and Prediction of Water Conservation Capacity Based on PLUS–InVEST Model: A Case Study of Baicheng City, China

Abstract

As an important ecosystem service, water conservation is influenced by land use related to human activities. In this study, we first evaluated spatial and temporal changes in water conservation in Baicheng City, western Jilin Province, from 2000 to 2020. Then, we identified three different scenarios: the natural development scenario (NDS), cropland protection scenario (CPS), and ecological protection scenario (EPS). We coupled the Patch-generating Land Use Simulation (PLUS) and Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) models to predict the distribution of land use types and water conservation in Baicheng City under these scenarios for 2030. The results showed the following: (1) The average water conservation in Baicheng City from 2000 to 2020 was 7.08 mm. (2) Areas with higher water conservation were distributed in the northwest and northeast, while lower water conservation areas were distributed in the central and southwest of Baicheng City. (3) The simulation results of the future pattern of land use show an increasing water conservation trend in all three scenarios. Compared with the other two scenarios, the ecological protection scenario is the most suitable option for the current development planning of Baicheng City. Under the ecological protection scenario (EPS), ecological land is strictly protected, the area of agricultural land increases to some extent, and the overall structure of changes in land use becomes more rational. This study provides a reference for land resource allocation and ecosystem conservation.

1. Introduction

Ecosystem services refer to all benefits that humans obtain from the natural environment [1]. Water conservation is a fundamental and indispensable ecosystem regulating service, defined as the capacity of ecosystems to retain and regulate precipitation [2]. It plays a crucial role in intercepting and storing rainfall, enhancing soil infiltration, maintaining soil moisture, replenishing groundwater, regulating runoff, and ultimately increasing the availability of water resources [3]. Water resources are fundamental to human survival, social development, and economic growth. Worldwide, the ecology and environmental issues related to water have been influenced by many factors over the past few decades, including population growth, rising living standards, the expansion of irrigated agriculture, and climate change [4,5,6]. In China, water resource problems are widespread, with many areas facing growing pressure on their water resources due to water pollution, water deficit, and uneven water resource allocation [7,8,9]. Regarding the allocation of water resources, agriculture consumes more water than other fields [10], with its water consumption accounting for over 70% of total freshwater use [11]. In particular, the northeastern region of China serves as a major grain-producing area, thereby accelerating water resource depletion [10]. With the continued expansion of agricultural production, groundwater reserves have been drastically declining, further intensifying water stress. Water resources have become an increasingly critical limiting factor for sustainable regional development. Therefore, assessing and enhancing the water conservation capacity of major grain-producing regions is crucial and necessary.

Regional land use and land cover change can affect water conservation [12]. Land use change caused by human activity directly and indirectly affects the characteristics and processes of terrestrial and aquatic ecosystems [13]. Land cover change alters precipitation retention, evapotranspiration, infiltration, and groundwater recharge [14]—which changes the hydrological cycle processes that regulate water yield [15]. These factors can be influenced by regional water conservation [16,17] and the spatial distribution of water resources.

In recent years, drought and heavy rainfall have occurred alternately in western Jilin. In particular, the total amount and frequency of heavy rainfall have increased significantly. Compared to previous years, precipitation in northwestern Jilin in 2020 increased by 50%, even reaching up to double the usual amount. Moreover, the precipitation in winter was twice as much as the average, there were rare snowstorms, and the daily cooling exceeded the historical extreme. Therefore, it is crucial to enhance ecosystem adaptations and responses to extreme weather events.

The water balance theory [18] can be used to assess water conservation in large regions. Water ecosystem services have been assessed using various models, such as the Soil and Water Assessment Tool (SWAT) model, the Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model [19], and the Artificial Intelligence for Ecosystem Services (ARIES) model [20]. Water yield is a crucial indicator for measuring water conservation [21]. The water yield module in the InVEST model is based on the principle of water balance [22], which estimates regional water yield by inputting data on precipitation, potential evapotranspiration, and land use. The InVEST model can be combined with ArcGIS 10.8 [23]. This tool has a simple structure, convenient parameter acquisition, and provides an intuitive spatial representation of results, making it suitable for large-scale hydrological simulations [24]. The Patch-generating Land Use Simulation (PLUS) model [25] simulates and predicts land use patterns under different natural and social scenarios [26]. Compared with the CA–Markov model [27], the Conversion of Land Use and its Effects at Small regional extent (CLUE-S) model [28], and the Future Land Use Simulation (FLUS) model [29], the PLUS model integrates a land expansion analysis strategy (LEAS) and a cellular automata model utilizing multiple types of randomly seeded patches (CARS) [30]. This can better explore the causes of land use change and simulate the changes at the patch level of multiple land use types [23]. Therefore, the PLUS model can accurately simulate spatial and temporal patterns for various land use types. It can provide baseline data for land resource planning [31].

The Western Jilin is located in the Agro-Pastoral Transitional Zone of Northeast China [32]. It is not only important for ecological security barriers, but also for grain production, as it bears the important responsibility of maintaining national food security. However, the area is facing multiple challenges, including a shortage of water resources, degraded ecosystems, desertification, and salinization. It is an arduous task to coordinate the economy and society with population, resources and environment. To align with the Baicheng City Territorial Spatial Plan (2021–2035) and the goal of ecological city construction, this study set up three land use simulation scenarios: natural development (NDS), cropland protection (CPS), and ecological protection (EPS). Focusing on the challenges of ecological protection pressure and water shortages in western Jilin Province, we evaluated the water conservation service of Baicheng City and predicted its water conservation in 2030 under the three scenarios. Specifically, we aim to: (1) quantitatively assess water conservation in Baicheng City, (2) analyze its temporal and spatial variations, and (3) identify the influencing factors. The results provide a basis for improving water ecosystem services, allocating land for ecosystem conservation, and promoting sustainable development of the region.

2. Materials and Methods

2.1. Study Area

Baicheng City (121°38′–124°4′ E, 44°02′–46°18′ N) is located in the northwest of Jilin Province (Figure 1). It includes the western Nenjiang Plain and the eastern Horqin Grassland [33]. The city covers a land area of 25,800 km², including five counties and urban areas: Zhenlai, Taonan, Da’an, Taobei, and Tongyu. Baicheng City has an undulating topography, which ranges from low mountains and hills to plains from northwest to southeast. The climate of Baicheng City is temperate semi-arid and semi-humid [34], with significant continental monsoon characteristics, simultaneous rain and heat, and significant precipitation variability, with an average annual precipitation below 450 mm in most areas [35]. The main soil types in Baicheng City are light chernozem soils, meadow soils, aeolian soils, saline soils, and alkaline soils [36]. The vegetation types is mainly grassland, including meadow grassland, steppe, and salinized meadow-steppe [35].

2.2. Data Preparation

The data used in this study, along with their sources, are listed in

Table 1. Main data sources in this study.

Data Type	The Name of Data	Year	The Source of the Data
Basic data	Land use data	2000, 2010, 2020	Resource and Environment Science Data Platform of Chinese Academy of Sciences (http://www.resdc.cn/, accessed on 10 January 2025)
Natural conditions	Daily temperature	2000–2020	National Meteorological Information Center of China (http://data.cma.cn/, accessed on 11 January 2025)
	Daily precipitation	2000–2020	National Meteorological Information Center of China (http://data.cma.cn/, accessed on 11 January 2025)
	Potential evapotranspiration	2000–2020	National Meteorological Information Center of China (http://data.cma.cn/, accessed on 11 January 2025)
	Soil type	-	China Soil Data Set (1:1 million)
	Plant available water content	-	One global AWC raster is provided by ISRIC, (https://data.isric.org:443/geonetwork/srv/eng/catalog.search, accessed on 10 January 2025).
	Elevation	-	Geospatial data cloud (https://www.gscloud.cn, accessed on 10 January 2025)
Accessibility factor	Distance from main water, distance from main first-class road, distance from main second-class road, distance from main third-class road	2010, 2020	Resource and Environment Science Data Platform of Chinese Academy of Sciences (http://www.resdc.cn/, accessed on 15 January 2025)
Socio-economic data	Population density, gross domestic product (GDP)	2010, 2020	Resource and Environment Science Data Platform of Chinese Academy of Sciences (http://www.resdc.cn/, accessed on 15 January 2025)

. All raster data were resampled to a spatial resolution of 30 × 30 m and projected to the GCS_WGS_1984 UTM Zone 51N coordinate system. The distances to the main roads and water were obtained through a Euclidean distance analysis of vector data in ArcGIS 10.8. The average precipitation was based on daily precipitation data from meteorological stations. The average potential evapotranspiration was calculated using the ETo Calculator 3.2 software based on wind speed, sunshine hours, temperature, and humidity data from meteorological stations. Spatial distributions of precipitation and potential evapotranspiration were then generated using kriging interpolation in ArcGIS 10.8. Additionally, ArcGIS 10.8 was used to classify land use into seven categories: cropland, forestland, grassland, wetland, water area, construction land, and unused land.

The InVEST model required the biophysical coefficient of each land use (

Table 2. Biophysical table used for the InVEST water yield model.

Lucode	LULC_Desc	LULC_Veg	Kc	Root_Depth (mm)
1	Cropland	1	0.65	350
2	Forestland	1	0.95	3000
3	Grassland	1	0.65	500
4	Water	0	1	1
5	Construction land	0	0.3	1
6	Unused land	0	0.3	1
7	Wetland	1	0.65	400

Notes: Lucode: land use code for each type. LULC_desc: names of different land use types. LULC_veg: land use attribute category. Kc: plant evapotranspiration coefficient for each land use type. Root depth: the maximum root depth for each land use type.

). The root depth was set by reference data from the InVEST model documents [37]. Kc was set based on the reference values of the evapotranspiration coefficient of the FAO [38]. The topographic index was calculated based on DEM and soil depth using the spatial analysis tools in ArcGIS 10.8. The velocity coefficient was obtained from relevant studies and the InVEST model manual [39]. The saturated soil hydraulic conductivity was calculated using the SPAW 6.02.75 software based on soil texture data. All data were resampled to a 30 m resolution.

2.3. Methods

2.3.1. PLUS Model

The PLUS model is a land use change simulation model based on patch data. It integrates the rule mining strategy and the patch generation mechanism, which can be used to explore the driving factors of land expansion and predict the patch-level evolution of land use landscapes [40].

(1)

Model input

We selected 10 natural and socio-economic data, including slope, elevation, temperature, precipitation, population density, GDP, distance from waters, distance from first-class roads, distance from second-class roads, and distance from third-class roads as the driving factors (Figure 2), obtained the contribution degree of each land type expansion and generated the corresponding development probability.

(2)

Scenario and Parameters

To explore the future development status of the region under different development goals, we established three scenarios: the natural development scenario (NDS), the cropland protection scenario (CPS), and the ecological protection scenario (EPS), with reference to the Territorial Spatial Plan of Baicheng City (2021–2035) [41]. The description of the three scenarios was as follows:

By modifying the land use transition probabilities of the Markov chain under different scenarios, the projected area of each land use type was derived [42]. Referring to the relevant studies, we set this adjustment between 30% and 50% [43].

Based on the historical development trend of land use change from 2010 to 2020, the NDS does not impose extra restrictions on land use, except that the conversion of water areas to other land types was limited [44].

The CPS is assumed that croplands will be protected. Baicheng serves as an important grain-producing area. It strictly implements the cropland protection policy and promotes both the expansion and improvement of cropland efficiency. Therefore, the conversion of cropland to other land use types was restricted in the Markov model transition matrix, with the probability of unused land converting to cropland increased by 30% and the probability of cropland converting to other land uses reduced by 50%.

The EPS is assumed that the natural ecology of Baicheng City will be preserved. Therefore, the protection of ecological land is strengthened, and all developmental and construction activities are strictly prohibited on such land to maintain ecosystem stability. Transition probabilities were adjusted as follows: the probability of unused land converting to forestland and grassland was increased by 30%, whereas the probabilities of forestland and grassland, and of cropland converting to construction land were reduced by 50% and 30%, respectively.

For each development scenario, a cost matrix was established to represent the transitions between different land use types. The cost matrix was represented using 0 and 1, where 0 indicated that conversion between two land use types was not allowed, and 1 indicated that conversion between them was allowed (

Table 3. Cost matrix settings. A, B, C, D, E, F, and G represent the Farmland, Forestland, Grassland, Water area, Construction land, Unused land, and Wetland, respectively.

	NDS							CPS							EPS
	A	B	C	D	E	F	G	A	B	C	D	E	F	G	A	B	C	D	E	F	G
A	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
B	1	1	1	1	1	1	1	0	1	0	0	0	0	1	0	1	0	0	0	0	0
C	1	1	1	1	1	1	1	1	1	1	0	0	0	1	0	1	1	0	0	0	1
D	1	1	1	1	1	1	1	1	1	1	1	0	0	1	0	1	1	1	0	0	1
E	0	0	0	0	1	0	0	1	1	1	1	1	1	1	1	1	1	1	1	1	1
F	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
G	1	1	1	1	1	1	1	0	1	1	0	0	0	1	0	1	1	0	0	0	1

).

(3)

Accuracy verification

The Kappa coefficient is a method used to assess the consistency of remote sensing image classifications by comparing classified results with land use data using an error matrix [45]. It represents the ratio of classification accuracy to random classification accuracy and is a commonly used indicator to evaluate the accuracy of classification:

$${\mathrm{OA}}_{\mathrm{O}}={\displaystyle \frac{\sum _{\mathrm{k}=1}^{\mathrm{n}}{\mathrm{OA}}_{\mathrm{kk}}}{\mathrm{N}}},$$

(1)

$$\mathrm{Kappa}={\displaystyle \frac{{\mathrm{OA}}_{\mathrm{o}}-{\mathrm{OA}}_{\mathrm{E}}}{\left(1-{\mathrm{OA}}_{\mathrm{E}}\right)}},$$

(2)

where OA_O is the overall accuracy (OA) classification, indicating the proportion of the simulation results that is consistent with the land use for each random sample, OA_E is the proportion of agreement expected by chance, n is the total number of land use types, N is the total number of samples, OA_kk is the number of correctly classified samples for the k-th land use types.

This study used land use data and driving factors from 2000 and 2010 to simulate land use in 2020 (Figure 3). By comparing the spatial prediction value of land use in 2020 with the actual value of land use in 2020, the Kappa coefficient was 0.89 and the OA was 0.92, indicating that the accuracy of the model simulation results were highly accurate [46]. Based on this, the land use under different scenarios in 2030 can be predicted based on the land use data in 2020.

2.3.2. InVEST Model

This study calculated water yield using the water yield module of the InVEST model. The module is calculated based on precipitation, potential evapotranspiration, land use, root depth, and plant available water content.

Z was an important seasonal constant to input into the water InVEST model. It reflected the regional precipitation pattern and seasonal factors associated with hydrogeological characteristics, with a value ranging from 1 to 30. Z was calibrated based on the total amount of surface water resources reported in the Baicheng City Water Resources bulletin of from 2000 to 2020. According to this bulletin, the average amount of surface water in Baicheng City was 20.52 × 10⁸ m³. When the seasonal constant Z value was 2.7, the multi-year average water yield simulated by the InVEST model was 20.78 × 10⁸ m³, which closely matched the measured amount of surface water. Therefore, a Z value of 2.7 was selected because it produced results closest to the water resources bulletin, and the simulation effect of the InVEST model is optimal. To eliminate the influence of interannual variability in meteorological data on water yield, multi-year average precipitation (2000–2020) and evapotranspiration data were used in the InVEST model when estimating the water yield for 2030.

(1)

The InVEST water yield model

The annual water yield module of the InVEST model is based on the Budyko hydrothermal coupling equilibrium hypothesis. The model sets the total annual precipitation of the grid unit minus the total actual evapotranspiration of the year as the total annual water yield:

$$\mathrm{Y}\left(\mathrm{x}\right)=\left(1-{\displaystyle \frac{\mathrm{AET}\left(\mathrm{x}\right)}{\mathrm{P}\left(\mathrm{x}\right)}}\right)\times \mathrm{P}\left(\mathrm{x}\right),$$

(3)

where Y(x) is the annual water yield of grid unit x, AET(x) is the annual actual evapotranspiration of grid unit x, and P(x) is the annual precipitation amount of grid unit x:

$${\displaystyle \frac{\mathrm{AET}\left(\mathrm{x}\right)}{\mathrm{P}\left(\mathrm{x}\right)}}=1+{\displaystyle \frac{\mathrm{PET}\left(\mathrm{x}\right)}{\mathrm{P}\left(\mathrm{x}\right)}}-{\left[1+\left\{{\displaystyle \frac{\mathrm{PET}{\left(\mathrm{x}\right)}^{\mathsf{\omega }}}{\mathrm{P}\left(\mathrm{x}\right)}}\right\}\right]}^{1/\mathsf{\omega }},$$

(4)

where PET(x) is the potential evapotranspiration and ω(x) is a nonphysical empirical fitting parameter that characterizes the natural climatic-soil properties of the catchments:

$$\omega \left(\mathrm{x}\right)=\mathrm{Z}{\displaystyle \frac{\mathrm{AWC}\left(\mathrm{x}\right)}{\mathrm{P}\left(\mathrm{x}\right)}}+1.25,$$

(5)

where Z is an empirical constant that captures the local precipitation pattern and hydrogeological characteristics. In this study, we improved the model by repeatedly verifying the simulated water yield and the measured water resources to obtain the Z value. AWC (x) refers to the volumetric (mm) plant available water content on pixel x, which is estimated as the product of the plant available water capacity (PAWC) and the minimum of root restricting layer depth and vegetation rooting depth:

$$\mathrm{AWC}\left(\mathrm{x}\right)=\min \left(\mathrm{Rest}\mathrm{layer}\mathrm{depth},\mathrm{root}\mathrm{depth}\right)\times \mathrm{PAWC},$$

where PAWC is the plant available water capacity and rest layer depth is the depth of root restricted layer, often expressed as the depth at which 95% of root biomass of the plant.

(2)

Calculation of water conservation

After the water yield is calculated according to the InVEST model. Based on the topographic index (TI), the velocity coefficient (Velocity) and the saturated soil hydraulic conductivity (Ksat) are used to correct the water yield to obtain the water conservation:

$$\mathrm{Retention}=\mathrm{M}\mathrm{in}\left(1,{\displaystyle \frac{249}{\mathrm{Velocity}}}\right)\times \mathrm{M}\mathrm{in}\left(1,{\displaystyle \frac{0.9\times \mathrm{TI}}{3}}\right)\times \mathrm{M}\mathrm{in}\left(1,{\displaystyle \frac{\mathrm{Ksat}}{300}}\right)\times \mathrm{Yield}.$$

(6)

2.3.3. Spatial Relevance Evaluation Based on Grid

Spatial autocorrelation reflects the degree of similarity of things in similar positions, which is divided into global autocorrelation and local autocorrelation. Global spatial autocorrelation reflects the overall spatial characteristics of geographic phenomena and is generally measured using the Global Moran’s I. The values of Global Moran’s I range from −1 to 1: values between −1 and 0 indicate spatial negative correlation, while values between 0 and 1 indicate spatial positive correlation. Local spatial autocorrelation reflects the degree of spatial correlation in local areas and is commonly measured using the Local Moran’s I. The results of local Moran’s I can be classified into five types: high-high (HH), low-low (LL), high-low (HL), low-high (LH), and insignificant. Local Moran’s I are visualized by the map for local indicators of spatial association (LISA map) [47].

ArcGIS 10.8 was used to generate a 1 km × 1 km grid of water conservation in the study area. GeoDa 1.20 was used to obtain the global Moran’s I, local Moran’s I, and LISA aggregation graphs of the water conservation grid:

$$\mathrm{I}={\displaystyle \frac{\mathrm{n}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\omega }_{\mathrm{ij}}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}\left)\right({\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{x}})}{{\sum _{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\omega }_{\mathrm{ij}}}},$$

(7)

$${\mathrm{I}}_{\mathrm{i}}={\displaystyle \frac{\mathrm{n}\left({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}\right){\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\omega }_{\mathrm{ij}}\left({\mathrm{x}}_{\mathrm{j}}-\overline{\mathrm{x}}\right)}{{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2}},$$

(8)

$$\overline{\mathrm{x}}={\displaystyle \frac{1}{\mathrm{n}}}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{x}}_{\mathrm{i}},$$

(9)

where I and I_i are global Moran’s I and local Moran’s I respectively; n and ω_ij are the number of spatial units and the weight matrix, respectively; x_i and x_j are the attribute value of spatial unit i and j, respectively; and $\overline{\mathrm{x}}$ is the mean value of all spatial units.

3. Results

3.1. Spatial-Temporal Changes in Land Use in Baicheng City from 2000 to 2020

From 2000 to 2020, cropland remained the dominant land use type in Baicheng, the proportion of which increased from 44.81% to 47.32%. Construction land accounted for the smallest proportion (Figure 4). Over the past two decades, the areas of cropland, forestland, and construction land increased by 647.55 km², 210.61 km², and 149 km², respectively. In contrast, the areas of other land use types decreased, with the largest reduction observed in grassland, followed by unused land.

The land use changes in Baicheng City from 2000 to 2020 were analyzed using a land use transfer matrix. As shown in Figure 5, the total transferred area was 5069.95 km², accounting for 19.59% of the total land use area, indicating that changes in the land use pattern were minor. From 2000 to 2020, cropland increased by 647.55 km², primarily transferred from grassland, wetland, and unused land, accounting for 3.32%, 1.13% and 1.30% of the total area, respectively. Forestland increased by 210.61 km², mainly transferred from cropland, representing 1.13% of the total area. Grassland decreased by 489.07 km², primarily converted to cropland. Wetland decreased by 140.54 km², mainly converted to cropland, grassland, and unused land. Construction land increased by 149 km², mainly converted from cropland, accounting for 0.65% of the total area. Unused land decreased by 188.78 km², primarily converted to cropland.

3.2. Multi-Scenario Simulation of LUCC

We used the PLUS model to simulate land use changes in Baicheng City in 2030 under three different scenarios (Figure 4). In the prediction, cropland accounted for the largest proportion, followed by unused land, grassland, wetland, forestland, water area, and construction land. Cropland was mainly distributed in the western part of Baicheng City, particularly in Taobei District and Taonan District. Forestland, grassland, and wetland were scattered across the northwest, north, south, and east. Construction land was concentrated in the west, mainly in Taobei District. Under all three scenarios, the areas of cropland, forestland, water area, and construction land were predicted to increase, whereas the areas of grassland, unused land, and wetland were predicted to decrease by 2030.

Comparison of land use under different scenarios (Figure 5):

Under the NDS, which did not include policy constraints, substantial areas of ecological land were predicted to be converted to cropland and construction land with economic development. Between 2020 and 2030, the areas of cropland, forestland, water bodies, and construction land were predicted to increase by 310.47, 45.92, 53.46, and 41.62 km², respectively. Conversely, grassland, unused land, and wetland were predicted to decrease by 87.17, 241.49, and 122.81 km², respectively.

Under the CPS, the area of cropland, forestland, water bodies, and construction land were predicted to increase by 502.54, 12.75, 33.18, and 5.57 km², respectively. Conversely, the area of grassland, unused land, and wetland were predicted to decrease by 114.62, 310.07, and 129.35 km², respectively.

Under the EPS, cropland was predicted to increase by 144.94 km², and forestland increased the most, by 153.02 km². The water bodies and construction land increased by 82.64 and 5.3 km² respectively. Unused land showed the largest decrease of 277.33 km², followed by grassland and wetland, which decreased by 74.48 and 34.09 km², respectively.

3.3. Spatial and Temporal Variation Pattern of Water Yield

From 2000 to 2020, the water yield of Baicheng City fluctuated considerably but showed an overall increasing trend (Figure 6). The average annual water yield was 80.31 mm, with a maximum of 155.4 mm in 2020 and a minimum of 34.86 mm in 2001. By administrative division, Zhenlai County had the highest average annual water yield (100.34 mm), followed by Taobei District (91.20 mm), Da’an City (89.84 mm), and Taonan City (78.85 mm), while Tongyu County had the lowest (61.21 mm). In 2030, the water yield simulated under NDS, CPS, and EPS was 80.92, 81.51, and 80.51 mm, representing increase of 0.61, 1.2, and 0.2 mm, respectively.

Both the average annual water yield (2000–2020) and simulated water yields for 2030 were higher values in the north of Baicheng City and lower in the south (Figure 7). Areas with higher water yield were primarily located in the Nenjiang River basin (northeast Baicheng) and the low mountains regions of the northwest. Areas with lower water yield were mainly distributed in the central and southwestern plains of Baicheng City. The eastern part of Zhenlai County, the northern part of Da’an City, and the northwestern part of Taonan City had high water yields. In contrast, the eastern part of Taonan City and the western part of Tongyu County exhibited lower water yields.

3.4. Spatial and Temporal Variation Pattern of Water Conservation

From 2000 to 2020, the average annual water conservation in Baicheng City was 7.08 mm, with the lowest value in 2001 (3.43 mm) and the highest in 2020 (11.61 mm) (Figure 6). By administrative division, the annual average water conservation followed the order of Taonan City (8.91 mm) > Da’an City (8.68 mm) > Zhenlai County (6.76 mm) > Tongyu County (5.67 mm) > Taobei District (5.64 mm). The water conservation predicted for Baicheng City under NDS, CPS, and EPS was 7.09, 7.12, and 7.11 mm, respectively, representing increase of 0.01, 0.04, and 0.03 mm compared with 2020.

According to Figure 8, areas with high water conservation were mainly distributed in the northwestern and northeastern parts of Baicheng City, including the upstream areas of the Tao’er River and Jiaoliu River, Xianghai Nature Reserve Area, Nenjiangwan-Chagan Lake-Longnuma-Niuxinjiu Wetland Area, and other wetlands and grasslands with high vegetation cover (Figure 8). Areas with lower water conservation were concentrated in the central and southwestern regions. The central part includes the central urban area, which contains a high amount of construction land. Here, the ground is generally impermeable, such as cement and asphalt, with a low capacity for water conservation. The southwestern region is characterized by gentle terrain and extensive cropland. In this region, long-term cultivation has resulted in soil compaction, reduced permeability, increased surface runoff, and a weakened water conservation capacity.

Land use influences temperature, precipitation, evapotranspiration, soil properties, and vegetation type, and therefore affects water conservation (

Table 4. The total amount of water conservation by land use type under different scenarios (unit: 10⁵ m³).

Different Land Simulation Scenarios	Land Use
	Cropland	Forestland	Grassland	Wetland	Construction Land	Unused Land
Average	1080.80	158.51	318.06	172.80	23.02	73.86
NDS	1105.09	163.34	308.43	163.90	23.96	70.89
CPS	1088.89	175.35	310.53	171.14	23.17	70.70
EPS	1120.24	160.13	305.53	163.17	23.19	70.35

). From 2000 to 2020, grasslands had the highest water conservation (10.87 mm), followed by forestland (10.30 mm), cropland (8.83 mm), wetland (7.76 mm), constructed land (2.52 mm), and unused land (1.61 mm) (Figure 9).

Compared with the 2000–2020 average, total water conservation under NDS in 2030 was predicted to increase by 8.58 × 10⁵ m³, with increases in cropland (24.29 × 10⁵ m³) and forestland (4.83 × 10⁵ m³), and decreases in grassland (9.63 × 10⁵ m³) and wetland (8.90 × 10⁵ m³).

Under EPS, total water conservation was predicted to increase by 12.74 × 10⁵ m³. Specifically, with increased in cropland (8.07 × 10⁵ m³) and forestland (16.84 × 10⁵ m³), and decreases in grassland (7.53 × 10⁵ m³) and wetland (1.66 × 10⁵ m³).

Under CPS, total water conservation was projected to increase by 15.57 × 10⁵ m³, with increases in cropland (39.44 × 10⁵ m³) and forestland (1.62 × 10⁵ m³), and decreases in grassland (12.53 × 10⁵ m³) and wetland (9.63 × 10⁵ m³).

3.5. Spatial Correlation of Water Conservation

The global Moran’s I of the average water conservation capacity of Baicheng City was 0.553, while the values under the NDS, CPS, and EPS were 0.601, 0.601, and 0.589, respectively, indicating positive spatial correlations. Local spatial autocorrelations of water conservation in Baicheng City (Figure 10), as shown by the LISA aggregation graphs, mainly manifested as high–high, low–low, and non-significant agglomeration types, with only a few areas exhibiting high–low and low–high agglomeration types. These results indicated that the spatial distribution of water conservation capacity was highly clustered.

High–high agglomeration indicated an area with high water conservation surrounded by other areas with similarly high values, while low–low agglomeration indicated the opposite. High–high agglomerations were distributed in the northeast and northwest of Baicheng City, consistent with the distribution of high water conservation areas identified previously. Low–low agglomerations were distributed in the central and southern areas, where cropland and unused land were dominant, and water conservation was relatively low.

Low–high agglomerations indicated areas with low water conservation surrounded by high value areas, whereas high–low agglomerations indicate areas with high water conservation surrounded by low value areas. These two types of agglomerations were scattered and occurred primarily in the transitional areas of different land use types.

We used pixel statistics to determine the percentage of each agglomeration type. From 2000 to 2020, the high–high agglomeration type accounted for 11.78% of the study area, the low–low agglomeration type accounted for 18.27%, and the low–high and high–low agglomeration types accounted for only 2%. In 2030 under NDS, CPS, and EPS, the high–high agglomeration type accounted for 15.27%, 15.78%, and 15.70%, respectively, which was an increase of 3.48%, 4.00% and 3.92%, respectively, compared to 2000–2020. The low–low agglomeration type accounted for 17.68%, 17.76%, and 17.66%, respectively, which was a decrease of 0.59%, 0.51%, and 0.62%, respectively, compared to 2000–2020.

4. Discussion

We analyzed the spatial and temporal changes in land use and water conservation in Baicheng city from 2000 to 2020. Then we coupled the PLUS and InVEST models to predict the future land use and water conservation patterns.

4.1. Spatial-Temporal Pattern and Influencing Factors of Water Conservation Service Function in Baicheng City

Water conservation is an important component of ecosystem services. The regional water conservation function is influenced by both natural factors, such as precipitation and evapotranspiration, and human factors including land use changes. Climate change was the direct driving force affecting regional water conservation [48].

Rainfall and evapotranspiration are directly related to the water–heat balance, which regulates the global terrestrial water cycle [3,49]. Correlation analysis between water conservation, precipitation and potential evapotranspiration indicated that climate factors have a significant influence on water conservation [50]. Water conservation is greater in areas with higher precipitation, while is reduced in areas with less precipitation [51]. In regions with high evapotranspiration, water resources evaporated into the atmosphere, thereby reducing water availability [52]. In this study, water conservation and precipitation in Baicheng City from 2000 to 2020 were positively correlated (correlation coefficient: 0.67), with significant correlation observed in 89.12% of the study area (Figure 11). However, in areas dominated by bare land, a negative correlation was observed. In these areas, low vegetation cover allowed precipitation to flow over the surface, leading to soil erosion and low water conservation. Overall, the spatial distribution of water conservation closely resembled that of precipitation.

In addition to climate factors, water conservation is also influenced by land use change [53]. Land use change alters the supply of ecosystem services by affecting key ecological processes, including soil erosion, energy exchange, water cycling, and carbon and nitrogen cycles. Changes in land cover impact the soil water-holding capacity [48]. The highest average water conservation capacity of different land use types in Baicheng City from 2000 to 2020 was founded in grassland, followed by forestland. This result is consistent with previous research findings [39,41]. In areas with dry climates, the growth of forest vegetation in environments with low precipitation accelerates soil moisture consumption. In addition, the canopy, shrub-grass layer, and leaf litter layer of the forest intercept and store precipitation. When precipitation is low, it is intercepted by tree branches and ground litter and subsequently evaporates; therefore, forested areas have lower water-holding capacities than grasslands [54]. Jia et al. demonstrated that although the water conservation capacity of forest land is better than that of grassland, the greater area covered by grasslands results in the largest total water conservation, making them more beneficial for ecological restoration projects [55]. Grassland can reduce the impact of rainfall on the soil surface, slow down surface runoff, and increase the soil infiltration rate. The leaf area index of grassland is small, transpiration is low, and water retention capacity is high [48]. However, overgrazing can exacerbate soil compaction and erosion, decrease soil saturated hydraulic conductivity, thereby reducing soil water-holding capacity. Therefore, it is necessary to optimize grazing patterns, balance ecological and economic benefits, and enhance the water-holding capacity of grassland.

4.2. Land Use Change and Water Conservation Prediction Under Different Scenarios

We established three scenarios (NDS, CPS, and EPS) to predict land use and water conservation in 2030. The land use scenarios used in the current study differed from those used by Liu et al. [43,56]. Baicheng City is dominated by agriculture, so we set the Cropland protection scenario but not the economic development scenarios. Under NDS, changes in different land uses were not constrained, leading to the continuous expansion of cropland and construction land, which utilized previously unused land and encroached upon ecological land. Under CPS, croplands were protected; the conversion of cropland to other land types was limited, and unused land was converted to cropland. Cropland exhibited a marked expansion trend, with most of the newly added cropland concentrated in resource-scarce areas such as the eastern part of Baicheng. Compared with 2020, cropland increased by 502.54 km², of which 353.39 km² (about 70%) resulted from the conversion of unused land. The expansion of cropland can increased the water conservation capacity of the study area [57]. Since cropland had a much higher water conservation capacity than unused land, the water conservation capacity under the CPS scenario was consequently higher. CPS resulted in the highest total water conservation in 2030, followed by EPS and then NDS. Although the CPS exhibited the highest total water conservation, most of the unused land in Baicheng consists of saline-alkali and bare land, which required substantial time and financial investment for reclamation. Therefore, large-scale conversion of unused land to agricultural land is relatively difficult to implement. In the CPS, the focus is on economic benefits, while ecological development is largely overlooked.

Under EPS, the focus was on ecological benefits. Ecosystems such as forestlands, grasslands, wetlands, and water areas tend to be protected. The area of forestland increases significantly, and wetland and grassland were not converted to other land types, which effectively improves vegetation coverage. The EPS simulated land use transitions under the premise of protecting cropland and permanent basic farmland, and the area of agricultural land remained unchanged compared to 2020. Compared with the CPS and NDS scenarios, forestland, grassland, and wetland exhibited the highest average water depth under the EPS, indicating that the water conservation function of ecological land performed best under this scenario. Studies have shown that policies promoting the conversion of farmland to forests could reduce farmers’ incomes in some areas of northern China [58]. The EPS poses no threat to farmers’ incomes, while the natural landscape can be leveraged to develop tourism, and the value of the natural ecosystem is fully acknowledged.

Overall, compared with the other two scenarios, EPS protects ecological land while controlling the expansion of agricultural land [41]. Therefore, EPS achieves the goal of land and space planning (ecological priority and maintaining ecosystem stability) in Baicheng City, resulting in improved water conservation [56]. The difference in water conservation among the three scenarios in 2030 is small because the PLUS model incorporates constraint factors that limit the extent of land use change within the short-term period [59]. Although the overall differences are numerically small, the results indicate that even minor land use adjustments may affect ecosystem water conservation services [60,61]. In water scarce regions, even slight improvements in water conservation can generate cumulative benefits at the basin scale over time.

4.3. Suggestions for Improving the Water Conservation Function

The results of this study showed that the CPS scenario produced the highest total water conservation, primarily because the conversion of large areas of unused saline–alkali land into cropland substantially increased the overall water conservation. In contrast, under the EPS scenario, forestland, grassland, and wetland exhibited higher water conservation depth per unit area, but their total contribution was limited due to their relatively small extent. This suggested that both expanding productive cropland and enhancing the quality of existing ecological land were crucial for improving water conservation in Baicheng City. The government should implement saline–alkali land improvement projects to rehabilitate suitable unused saline–alkali land. In addition, salt-tolerant crop varieties should be promoted to convert the improved land into productive cropland moderately. The water conservation capacity of existing forests, grasslands, and wetlands should also be enhanced through ecological restoration measures such as soil improvement, vegetation restoration, and salinization control. Although ecological restoration has been carried out, studies reported that the water conservation capacity did not improve significantly, likely due to the inappropriate selection of plant varieties, which resulted in high mortality rates [62,63]. The arid and semi-arid regions of China were not suitable for growing plants with high water demands, these plants could alter regional hydrological processes and exacerbate desertification [64]. Additionally, large-scale ecological projects and afforestation activities have consumed excessive water resources and increased pressure on water availability in arid areas [65]. For successful ecological restoration, suitable plant species need to be chosen, planted in an appropriate pattern, and managed [63].

Due to the heterogenous distribution of land use types, the spatial distribution of water resources in Baicheng City was uneven. Therefore, we recommend strengthening water resource regulations, improving land resource allocation, and promoting the integrated restoration of forests and grasslands so that forests, grasslands, and wetlands serve as an ecological safety barrier for agricultural ecosystems.

4.4. Limitations of This Study

In this study, we simulated land use change in Baicheng City to provide a reference for decision-making. However, we only considered three scenarios. The social and economic development goals of Baicheng City were not adequately considered, which may limit the effectiveness of the results. Moreover, the input data for the InVEST model, such as root depth and crop coefficient, were based on previous studies and local conditions, which could affect the accuracy of the results. The InVEST model is relatively simple and does not adequately capture processes such as groundwater dynamics. In addition, the transition probabilities in the PLUS model involve a certain degree of subjectivity. Furthermore, the differences in outcomes across different scenarios are relatively small, which raises concerns regarding their statistical significance. Ecosystems in ecologically fragile areas provide a wide range of services, which involve trade-offs and synergies and are influenced by multiple factors. Future work will extend the analysis to long-term time series, investigate additional ecosystem services in Baicheng, and examine their interrelationships from multiple perspectives.

5. Conclusions

In this study, we combined the PLUS model and the InVEST model to evaluate water conservation in Baicheng City for the periods 2000–2020 and 2020-2030. The results show the following:

(1)

From 2000 to 2020, the average annual water conservation capacity of Baicheng City was 7.08 mm, which was greatly affected by climatic factors. The highest water conservation capacity occurred in 2020, and the lowest in 2001. Water conservation levels were high in the northwest and northeast of Baicheng City, and low in the central and southwest regions. Areas with high vegetation cover, such as wetlands and grasslands, exhibited higher water conservation capacity.

(2)

Land use in Baicheng City in 2030 was simulated under three scenarios. Under NDS and CPS, croplands increased substantially. There was a slight increase in croplands under EPS, and a greater increase in forestlands and water areas.

(3)

The multi-scenario simulation results revealed the potential changes in water conservation under different land use policies. The EPS achieved a relative balance between cropland and ecological land, making it the most suitable option for Baicheng City’s current development planning.

This study provides a reference for land resource allocation and the protection of ecosystems. Compared with NDS and CPS, EPS increased the proportion of ecological land and limited the expansion of construction land, while protecting cropland and permanent basic farmland, which aligns with the development strategy proposed in Baicheng City Territorial Spatial Plan (2021–2035). Baicheng City must take effective measures to improve environmental quality and protect ecosystems. Since land use change impacts water conservation capacity, the appropriate allocation of land resources is essential for improving ecosystem services. Rational land use planning and balancing food security with ecological security are the main challenges facing Baicheng City.