The Impact of Land Use and Landscape Pattern on Ecosystem Services in the Dongting Lake Region, China

Abstract

Ecosystem services (ES) are directly affected by land use and land cover changes (LUCC); however, the impacts of extended period LUCC on ES are poorly explored. Here, we mapped the 1998–2019 annual land use and land cover in the Dongting Lake Region (China) and explored the spatiotemporal evolution of LUCC and landscape patterns (i.e., composition, shape, and aggregation) and their relationship with ES, including carbon storage, gross primary production (GPP), water conservation capacity, and crop yield in the region. The results showed a significant increase in forest areas and impervious surfaces and a decrease in croplands and bare lands with spatial heterogeneity. Carbon storage was strongly correlated with forest, cropland, waterbody, impervious surface, and bare land, and there was a nonlinear relationship between landscape patterns and ES. The trade-offs and synergies (correlations) among ES varied considerably, with crop yield being significantly synergistic with carbon stocks, GPP, or GPP with carbon stocks. This study revealed the nonlinear relationship between landscape patterns and ES, and the mechanism of landscape characteristics on ES. The findings can provide scientific support for regional land use planning, ES regulation, and landscape optimization in the lake region.

1. Introduction

Ecosystems have suffered unprecedented impacts due to the pursuit of economic development and the lack of protection of ecosystems in the past decades [1,2]. Environmental problems have become increasingly serious on a global scale [3], such as the weakening of the carbon sequestration capacities of forests [4], the change of suitable areas for crop cultivation [5], and the frequent occurrence of extreme weather [6,7]. The Millennium Ecosystem Assessment report stated that more than 60% of global ecosystems experienced varying degrees of ecological degradation. This loss and degradation of ecosystem functions would have significant impacts on human well-being and pose a direct threat to regional and even global ecological security [8,9].

Land use and land cover change (LUCC) as well as landscape pattern evolution are the most direct ways to influence ecosystem processes and services [10,11]. Rapid changes in global land cover and landscape patterns have occurred over the past decades. For example, in 2020 the Food and Agriculture Organization of the United Nations (FAO) reported that the area of global forest had declined by nearly 100 million hectares in 20 years, indicating a steady decline in forest area despite a lower rate of decline than in the past. Global cities were expanding continuously and rapidly in the 21st century, with the building area increasing from 239,000 km² in 2000 to 519,800 km² in 2020, an expansion of 117.49% [12].

Ecosystem services (ES) are critically important for human well-being, yet they face numerous natural and anthropogenic threats [2,13]. ES can be classified into provisioning services (e.g., providing food and water), regulating services (e.g., controlling floods and diseases), supporting services necessary to maintain other types of services (e.g., nutrient cycling to sustain a living environment), and cultural services (e.g., spiritual recreation and cultural benefits) [14]. Sustainable, healthy and harmonious ecosystem services are the basis for maintaining regional ecological sustainability and the normal functioning of natural ecosystems and socio-economic activities [15,16]. Many studies have been conducted on ecosystem services. For example, Holdren and Ehrlich [17] introduced the services that natural ecosystems provide to humans and explored the anthropogenic impact on ecosystem service functions. Daily [18] provided a detailed description of the connotation, definition, and classification of ecosystem service functions. Costanza, et al. [19] described the methodology of ecosystem services valuation and estimated the economic value of global ecosystem service functions. Furthermore, ecosystem assessment models based on remote sensing and geographic information system (GIS) were widely applied to the assessment of ecosystem services at large regional scales [20].

Different types of land cover can provide different ES [21] and, correspondingly, ES changes feedback to land use patterns through ecological processes. In recent years, the dynamics of macro-scale ecosystem services have been widely expressed with the increase in research on the effects of LUCC on ecosystem services [22]. Numerous studies have shown that the changes in ecosystem services caused by LUCC were extensive and profound [23]. These changes produced significant impacts on biological diversity [24], provisioning services [25], regulating services [26], cultural services [27], etc. Since the late 1990s, the Chinese government has taken a series of measures to restore and reconstruct degraded ecosystems, such as the construction of the “Three Norths” protective forest system [28], the return of fields to lakes, and the return of farmland to forestry or grass. Therefore, LUCC, land use patterns, ES, and ecological security have formed an interlinkage and interaction nexus.

Although existing studies of LUCC on ESs have gradually increased, most of them are conducted on one-time cross-sections or interval time cross-sections, that is, from static perspectives. They ignored the impacts of extended periods of LUCC on ES, which may result in wrong guiding policies. Moreover, it is difficult to explain the driving mechanisms of ES when the historical changes in driving forces were not considered. In addition, how landscape patterns affect ecosystems was rarely explored. Different types of LUCC significantly impacted the quality and quantity of ecosystem services driven by human activities. Through rational land resource use and management [29], the structure and function of the ecosystem can be maintained in a healthy and balanced state. Additionally, such management can provide stable, balanced, and abundant natural resources for the sustainable development of human society, economy, and agriculture, thus maintaining the long-term coordinated development of natural-social-economic relations [30].

The Dongting Lake connects the Yangtze River, serves as an important water storage site and provides the ecosystem services of the Yangtze River basin [31]. Recently, the over-exploitation of ecosystems and natural resources has led to serious ecological and environmental implications in the lake region, such as the degradation of wetlands, forests, soil, and water conservation functions. Consequently, exploring the impact of LUCC on ES would contribute to a comprehensive understanding of the evolution of ecosystem functions in the Dongting Lake region. In this study, we aim to explore the spatiotemporal evolution characteristics of LUCC, landscape patterns, and ES in the lake region from 1988–2019, and to analyze the synergies and trade-offs among and the impacts of LUCC and landscape patterns on ES.

2. Materials and Methods

2.1. Study Area

The Dongting Lake (28°30′–30°20′N, 110°40′–113°10′E), the second largest freshwater lake in China, is located in the middle reach of the Yangtze River (Figure 1). The Lake provides an important ecological capacity for flood regulation and water conservation. The study area (the Dongting Lake Region, DTLR) covers the Dongting Lake District of the Hunan Province, including 12 districts and counties: Huarong County, Yueyang City, Yueyang County, Miluo City, Xiangyin County, Yiyang City, Ziyang District, Yuanjiang City, Nan County, Anxiang County, Hanshou County, and Changde City. The total area of DTLR is around 20,000 km². The average annual temperature is about 16 °C, and the average annual rainfall is about 1200 mm. Low mountains and hills in the east, south, and west of the region have an average elevation of fewer than 60 m, presenting a variety of geomorphology and landforms [32]. As the largest of the fresh lakes in this region and a globally important wetland for freshwater aquatic biodiversity, Dongting Lake has undergone drastic changes in recent history. The water surface area of Dongting Lake decreased by ∼17% after the impoundment of the Three Gorges Dam (TGD) in 2003, triggering the divergent change of ecosystem services [33,34].

2.2. Data Source and Data Processing

A critical issue in monitoring long-term changes in land use/cover is the appropriate selection of remote sensing data [34]. In this work, long-term remotely sensed images collected by Landsat instruments, including MSS, TM, ETM + and OLI, were obtained from Landsat Collection 2 Surface Reflectance Product through the Google Earth Engine (GEE) platform. All data were atmospherically rectified to eliminate errors from atmospheric scattering, absorption and reflection, which provided direct access to detect changes in the earth’s surface among multiple images of the same area [35].

In addition, datasets reflecting different bioclimatic and topographic features were collected as auxiliary data based on the GEE cloud platform to provide a broader range of land use classification features. The elevation, slope, and aspect data were obtained from the Shuttle Radar Topography Mission (SRTM) [36] data (https://lpdaac.usgs.gov/products/srtmgl1v003/, accessed on 15 April 2023). Annual mean temperature and precipitation were obtained from the WorldClim V1 Bioclim dataset (GEE). Imaging of water bodies was obtained from the JRC Global Surface Water Mapping Layers, v1.3 dataset (GEE).

2.3. Data Analyses

Land cover types were classified based on Google Earth high-resolution images, and a total of 3801 points were randomly collected as training samples. The land cover was classified as forest, shrub, grassland, wetland, cropland, water, impervious land, and bare land.

2.3.1. LUCC Analysis

We used an improved Continuous Change Detection and Classification (CCDC) algorithm to detect and map annual land cover [37]. CCDC uses a robust iterative reweighted least squares (RIRLS) approach to iteratively fit a dynamic time series model of land cover that includes seasonality, trends (gradual change), and interruptions (abrupt change) [37]. The equation is as follows:

$$\widehat{f}{\left(i,x\right)}_{OLS}=a_{i}x+b_i+{\displaystyle \sum }_{n=1}^3(c_{n,i}\cos {\epsilon }_{n}x+d_{n,i}sin{\epsilon }_{n}x)$$

(1)

where the intercept value and slope of the i-th landsat band are represented by $b_i$ and $a_i$. The coefficients $c_{n,i}$ and $d_{n,i}$ are used to estimate the intra-annual variation caused by the physical and seasonal differences of the i-th Landsat band.

We randomly selected 15,480 samples for accuracy assessment. The samples were divided into two groups, with 80% of them used for model training and the remaining 20% for model validation. The accuracy of the products was calculated by a confusion matrix, including User’s Accuracy, Producer’s Accuracy, and Overall Accuracy [38].

2.3.2. Calculation of Landscape Patterns

More than thirty landscape metrics were calculated in a 5 km-wide moving window based on the above land cover at 30 m using the “landscape metrics” package [39] in the R environment. Then, three of them (i.e., percentage of landscape, PLAND; fractal dimension index, FRAC; and contiguity index, COHESION) were selected to depict the landscape composition and configuration (

Table 1. Landscape metrics used in this study.

Abbreviation	Category	Description
PLAND	area and edge metric	percentage of landscape: the percentage of the landscape belonging to a given class
FRAC	shape metric	fractal dimension index: based on the patch perimeter and the patch area and describing the patch complexity
COHESION	aggregation metric	contiguity index: quantifies the connectivity of habitat as perceived by organisms dispersing in binary landscapes

). PLAND is the most widely used landscape metric to quantify the composition of landscape [40,41]; FRAC quantifies the configuration of landscape [42,43]; and COHESION is commonly used to quantify the connectivity of habitat [44].

$$PLAND=P_i=\frac{{\sum }_{j=1}^n{a}_{ij}}{A}\times 100$$

(2)

where P_i means the proportion of the specific landscape; a_ij means the area of patch j landscape i; and A means the total area of the specific landscape.

$$\begin{array}{c}FRAC=\frac{2\times \ln \left(0.25\times P_{ij}\right)}{\ln A_{ij}}\end{array}$$

(3)

where the P_ij means the perimeter of patch j landscape i, and A_ij means the area of patch j landscape i.

$$\begin{array}{c}COHESION=\left|1-\frac{{{\displaystyle \sum }}_{j=1}^n{P}_{ij}^{\ast }}{{{\displaystyle \sum }}_{j=1}^n{P}_{ij}^{\ast }\sqrt{A_{ij}^{\ast }}}\right|\times {\left|1-\frac{1}{\sqrt{Z}}\right|}^{-1}\times \left(100\right)\end{array}$$

(4)

where P_ij* means the perimeter of patch j landscape i in terms of the number of pixel surfaces; A_ij* means the area of patch j landscape I in terms of pixel number; and Z means the total pixel number for the specific landscape.

2.3.3. Calculation of Ecosystem Services

Given the need for socio-economic development and environmental protection in the region, we utilized carbon storage, GPP, water conservation, and crop yield for ES estimation.

(1)

Carbon storage

Based on land covers and their respective carbon storage [45,46,47,48,49], a regional multi-ecosystem above-ground carbon stock calculation model was established. The model assumed that the average age of forests in 1988 was 5, 10, 15, 20, 25, and 30 years, respectively, and then calculated the ecosystem carbon storage; subsequently, the data of multiple starting years (six layers per year) were averaged to obtain the ecosystem carbon sequestration from 1988–2019.

(2)

GPP

Gross primary production (GPP) was conducted with the concept of light use efficiency (LUE), and vegetation productivity was calculated from the photosynthetically active radiation absorbed by vegetation and LUE [50,51]. The specific calculation equation is as follows:

$$\begin{array}{c}GPP=PAR\times FPAR\times {\epsilon }_{max}\times TMI{N}_{scalar}\times VP{D}_{scalar}\end{array}$$

(5)

where $PAR$ is light and effective radiation (MJ/m²); $FPAR$ is the proportion of solar radiation absorbed by the vegetation canopy; ${\epsilon }_{max}$ is the maximum light energy use efficiency. $TMI{N}_{scalar}$ and $VP{D}_{scalar}$ are expressed as the limits of temperature and moisture environmental factors, respectively, and $TMI{N}_{scalar}$ is calculated from the linear relationship between $TMI{N}_{min}$ (temperature at ε = 0) and $TMI{N}_{max}$ (temperature at ε = ${\epsilon }_{max}$), and $VP{D}_{scalar}$ is obtained from the linear relationship between $VP{D}_{min}$ (saturated water vapor pressure difference at ε = 0) and $VP{D}_{max}$ (saturated water vapor pressure difference at ε = ${\epsilon }_{max}$).

(3)

Water conservation capacity

The water conservation capacity is calculated through the water balance equation, which considers the different ecosystem types as a “black box” and focuses on the input and output of water. The difference between precipitation and evapotranspiration and other consumption is the total water quantity (TQ) [52]. The formula is as follows [53]:

$$\mathrm{TQ}={\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{j}}\left(P_i-R_i-E{T}_i\right)\times A_i$$

(6)

where, $P_i$ is the rainfall of land class i; $R_i$ is the surface runoff of land class i, calculated from the surface runoff coefficient and annual rainfall; $E{T}_i$ is the evapotranspiration of land class i; and $A_i$ is the area of land class i.

Surface runoff coefficient (α) is the ratio of surface runoff (R) to rainfall (P), which to a certain extent reflects the capacity of the ecosystem water holding capacity. Its standard coefficient table is generally used to calculate surface runoff in the region. The formula is as follows:

$$\mathrm{R}=\mathrm{P}\times \left(\frac{\mathsf{\alpha }}{100}\right)$$

(7)

(4)

Crop yield

The crop yield was calculated using the following function based on the GPP product.

$$\begin{array}{c}Y=GPP\times AR\times HI\times RS\end{array}$$

(8)

where the Y means the crop yield; GPP is the gross primary production calculated above; AR is the proportion of autotrophic respiration in GPP (0.53 in this study) [54]; HI is the harvest index, a standard metric that means the proportion of aboveground crop yield to plant economic yield; RS is the ratio of root to shoot, indicating the ratio of underground yield to aboveground yield. HI and RS are different with crop types [55], and the percentage of each crop planted is based on the local statistical yearbook.

2.3.4. Trade-Offs and Synergies among Ecosystem Services

The correlation coefficient is used to identify spatial trade-offs and synergistic effects of ES including carbon storage, GPP, water conservation capacity, and crop yield. The correlation coefficient (Res) is calculated as follows:

$$R_{es}=\frac{{\sum }_{i=1}^n\left(E{S}_i-\overline{E{S}_i}\right)\left(E{S}_j-\overline{E{S}_j}\right)}{\sqrt{{\sum }_{i=1}^n{\left(E{S}_i-\overline{E{S}_i}\right)}^2 {\sum }_{j=1}^n{\left(E{S}_j-\overline{E{S}_j}\right)}^2}}$$

(9)

where $E{S}_i$ is the value of ecosystem service of class i; $\overline{E{S}_i}$ is the mean value of ecosystem service of class i; $E{S}_j$ is the value of ecosystem service of class j; and $\overline{E{S}_j}$ is the mean value of ecosystem service of class j.

2.3.5. Impacts of LUCC and Landscape Patterns on Ecosystem Services

The Pearson correlation coefficients (r value) were used to quantify the relationships between LUCC and ES. The generalized additivity model (GAM) was used to explore the relationship between landscape patterns and ES [56]. All fitted models were adjusted with R² > 0.6 and p < 0.05 (statistically significant). This analysis was performed in the R environment [57] using the “mgcv” package.

3. Results

3.1. Changes in Land Use and Land Cover

The average overall accuracy of the maps was 90.5%, the producer’s accuracy was 93.3%, and the user’s accuracy was 90.6%, suggesting that the classification was satisfactory. LUCC experienced a large change since 1988 (Figure 2). The area of forest increased by 618.66 km² over the past 30 years, with an average annual increase of 18.07 km²/year⁻¹ (

Table 2. Land use in Dongting Lake Region during 1988-2019 (km²).

Year	Forest	Shrubland	Grassland	Wetland	Cropland	Waterbody	Imperious	Bareland
1988	4518.88	351.71	2.97	1839.46	9127.57	2431.50	1138.22	540.11
1995	4534.69	330.19	1.13	1836.52	9189.68	2452.00	1142.07	464.44
2003	4619.41	330.59	1.22	1910.33	9803.62	2456.85	1166.33	382.21
2011	4754.04	326.28	2.38	1948.63	9047.43	2352.30	1204.63	311.89
2019	5137.54	306.05	2.43	1836.37	8758.43	2363.33	1341.56	176.05
1988–1995	15.81	−21.52	−1.84	−2.94	62.11	20.50	3.85	−75.67
1995–2003	84.72	0.40	0.09	73.81	613.94	4.85	24.26	−82.23
2003–2011	134.63	−4.31	1.16	38.30	−756.19	−104.55	38.30	−70.32
2011–2019	383.50	−20.23	0.05	−112.26	−289.00	11.03	136.93	−135.84
1988–2019	618.66	−45.66	−0.54	−3.09	−369.14	−68.17	203.34	−364.06

). The proportion of forests increased from 22.65% (1988) to 25.79% (2019) with a ratio of 0.09% year⁻¹. The area of shrubland decreased by 45.66 km², with an average annual decrease of 1.47 km² year⁻¹ (0.42% year⁻¹). The area of grassland decreased overall, showing an obvious decrease from 1988 to 1995 and then increases until 2005. The wetland area first increased and then decreased sharply by 112.26 km² from 2011 to 2019. The cropland area increased initially and then decreased, with an overall decrease of 369.14 km². The waterbody area showed an obvious phase change. For instance, it increased at a rate of 3.97 km² year⁻¹ from 1988 to 1995, decreased at a rate of 13.75 km² year⁻¹ from 2003 to 2011, and finally increased slowly at a rate of 1.44 km² year⁻¹ after 2011. The area of the impervious surface increased linearly by 5.21 km² year⁻¹, with its proportion increasing from 5.71% (1988) to 6.73% (2019). The area of bare land decreased linearly by 10.70 km² year⁻¹, from 2.77% (1988) to 0.88% (2019).

Figure 3 shows the different LUCC types in the recent 30 years. The forest was mainly converted from cropland and bare land (Figure 3a). From 2000 to 2009, most of the increase in the forest came from bare land, and cropland became the largest contributor after 2010. The shrubland was mainly converted from the forest and bare land (Figure 3b). From 1994 to 1996, the shrubland showed rapid transformation into forest. After 2000, shrubland mainly came from bare land, while cropland transferred to the forest. The grassland showed feeble transformation with wetland, cropland, and forest (Figure 3c), but the changes in grassland were highly variable in different periods. There was a main transformation between wetland and waterbody, cropland and forest (Figure 3d). Waterbody areas were mainly converted to wetland before 2006, and then the occurred after 2011. The cropland tended to convert to waterbody and forests (Figure 3e). In the recent ten years, cropland dramatically decreased and was mainly converted to forests, waterbody, and impervious surfaces. Additionally, the forest was the main transformation target from 2016 to 2018. The waterbody showed frequent transformation with cropland and wetland (Figure 3f). The impervious surface was mainly converted from forest and cropland (Figure 3g). The bare land was mainly converted from the impervious surface, and converted to the forest (Figure 3h).

3.2. Evolution of Landscape Patterns

The forest PLAND was higher on the edge of the study area, and the most rapidly changed areas were scattered in the east (Figure 4 and Figure S1). Similarly, the distributions of annual mean PLAND of shrubland, grassland, impervious surface, and bare land were higher in the edge of this region, and their annual change rates were mainly concentrated in the east. The PLAND of wetlands increased significantly in the central lake region, while waterbody was mainly distributed in the west of the lake. Meanwhile, the decrease of waterbody PLAND was located in the east. Finally, the high PLAND of cropland was mainly distributed in the northwest region.

The FRAC (i.e., the fractal dimension index of a specific landscape) of the forest was higher in the east and southwest parts and lower in the central part, while its increase was mainly in the central region (Figure 5 and Figure S2). The higher FRAC of both the waterbody and the impervious surface was mainly in the northwest region, and the change of FRAC was scattered across the study region. In contrast, the FRAC of cropland was higher in the surrounding region (mainly in the southwest part), and the increase rate was also higher in this region. Finally, the FRAC of bare land was higher in the east and lower in the west region, and the change rate of bare land FRAC also showed a similar spatial pattern: decrease in the east and increase in the west.

The COHESION of forest and cropland increased sharply in the central region (near the lake region), while the increases in COHESION index of shrubland, wetland and waterbody are scattered throughout the whole study region. In addition, the increase of impervious surface COHESION index is mainly in the southeast part, but in the western part for bare land (Figure 6 and Figure S3).

3.3. Ecosystem Service Function and Trade-Offs

Carbon storage, GPP, water conservation, and crop yield have increased generally (Figure 7 and Figure S4). The average annual carbon storage was 1669 kg C ha year⁻¹ with an annual growth rate of 32 kg C ha year⁻¹ from 1988 to 2019. The inter-annual variation of total carbon stocks (TCS) of different land covers was significantly different (Figure 7a). Forests showed the highest TCS, increasing from 1.27 million tons (1988) to nearly 3 million tons (2019). The wetland carbon stocks increased from about 0.64 million tons (1990) to nearly 0.68 million tons (2010), and then rapidly declined to lower than 0.64 million tons (2019). The carbon stocks of cropland declined from about 0.0337 million tons (1990) to 0.032 million tons (2019).

The average annual GPP was 2.81 g C m⁻² day⁻¹, and the annual growth rate was 0.0366 g C m⁻² year⁻¹ (Figure 7b). Cropland contained the highest total GPP, and its total GPP increased from 27,000 tons of carbon per day (2000) to 32,000 tons of carbon per day (2008). In general, the total GPP of the forest increased while that of shrubland and cropland decreased over time.

From 2000 to 2015, the average annual water conservation capacity was 9400 mm, with an annual change rate of 3.27 mm year⁻¹. The water conservation capacity of different land covers was different with an overall growth trend (Figure 7c). Cropland showed the highest capacity of 100 billion tons/year, followed by forest (50 billion tons/year), waterbody (40 billion tons/year), wetland (30 billion tons/year), impervious surface (15 billion tons/year), shrubland (3 billion tons/year), and bare land (3 billion tons/year). For all land covers except grassland and bare land, the annual capacity increased from 2000 to 2003, decreased in 2005, and then continued to rise until 2015.

The crop yield changed dramatically in the study period, with an annual mean crop yield of about 7.8 million tons (Figure 7d). There were three obvious time points of the sharp decline in crop yield: 2001–2003, 2008–2009, and 2018–2019.

The trade-offs and synergies (correlations) among different ES varied considerably (Figure 8). Overall, water conservation and crop yield showed weak synergies with carbon storage, and there was a weak synergistic relationship between water conservation and crop yield. The relationships between GPP and water conservation varied geographically, synergistic in the central and trade-off in the surrounding regions (Figure 8f). GPP was significantly synergistic with carbon storage and crop yield in the most regions (Figure 8b,c).

3.4. Impact of LUCC and Landscape Patterns on Ecosystem Services

There were some significant relationships between ES and land uses (Figure 9), particularly carbon storage (R² > 0.8, p < 0.05) including forest (0.94), cropland (−0.88), waterbody (−0.88), impervious surface (0.89), and bare land (−0.95). The correlation between water conservation capacity and the land cover area was weak. In terms of different land covers, there was a strong correlation between forest, grassland, wetland, cropland, water body, and different ES.

There was a non-linear relationship between landscape patterns and ES with large variations (Figure 10). The carbon storage of forests and wetlands decreased with the increase of PLAND when PLAND was low (PLAND < 5%), and subsequently (PLAND > 5%), as carbon sequestration increases with the increase of PLAND. The crop yield also changed nonlinearity with the PLAND of cropland. When the PLAND of cropland was higher than 75%, crop yield decreased as cropland PLAND increased (Figure 10f). Carbon storage decreased with the increase of FRAC when FRAC was lower than 1.012 and afterward increased with FRAC. Similar nonlinearity was also found in the relationship between carbon storage and FRAC of both forest and cropland (Figure 10a,c), as well as wetland COHESION (Figure 10b).

4. Discussion

The current study employed extended periods of Landsat images and the up-to-date CCDC algorithm to monitor the LUCC of the DTLR continuously from 1990 to 2019. Then the corresponding landscape patterns and ecosystem services were evaluated. We further analyzed the impacts of LUCC and landscape pattern changes on ecosystem services. This is the first study that reveals the driving mechanisms of ecosystem services impacted by the long-time LUCC and landscape patterns changes in the Dongting Lake region, which is informative for regional landscape optimization and sustainable development.

The results showed that drastic LUCC occurred in the DTLR in the past 30 years (Figure 2 and Figure 3), which was attributed to local socio-economic development and ecological protection policies. During the period of the 12th Five-Year Plan (2011–2015), the central government authorized the Dongting Lake region as the key area of the Yangtze River Basin for one of the “Ten ecological barriers.” In 2012, cycle economy and sustainable development strategies in Pan-Dongting Lake Economic Circle were included in the national development strategy, and its eco-environmental sustainability has been of increasing concern [58]. For example, the rapid increase of impervious surface areas mainly occurred around urban areas and in the plain areas of the lake region, which could be attributed to the fact that this region is the core of the Yangtze River Economic Belt and has undergone rapid urbanization [31,38]. Meanwhile, the area of cropland declined obviously throughout the period mainly in the plain areas around the lake region, while the area of forest and waterbody increased simultaneously. This trade-off link could be explained by the strict land use and ecological protection policy implemented in the area [59,60].

Regarding waterbody and wetlands, the dam and reservoir constructions, land reclamation, and ecological restoration could change the cycle of surface water and wetland cover [61]. In the DTLR, the area of waterbody has decreased gradually since 2003, resulting from the impoundment of the Three Gorges Dam [62,63]. The land reclamation or waterbody into other land covers resulted in the loss or shrinking of the waterbody and the rapid increases in cropland between 2000 and 2010. These processes together brought about the decline of wetlands and the shrinkage of the waterbody. In addition, the continued development of urbanization, industrialization, and population in the DTLR rapidly increased, leading to waterbody shrinkage. However, after the catastrophic flood in 1998, the government implemented the “Returning Farmland to Lake” program to prevent floods by promoting the conversion of farmland to natural lakes in the Yangtze River basin [60]. In Hunan, a project in Dongting Lake called “4350” (i.e., restoring the lake to its original size of 4350 km² in 1949) was implemented after 1998 to prevent floods and restore the wetland [59]. This program systematically returned parts of cropland to wetlands, forests and grasslands, which increased the waterbody to some extent. The land reclamation was reversed in some areas to restore wetlands and other aquatic systems.

This study confirmed that the evolution of composition (e.g., PLAND), agglomeration degree (e.g., COHESION), and fragmentation degree (e.g., FRAC) showed obvious spatial heterogeneity. For example, the composition and agglomeration degree of the wetland landscape increased in DTLR and its surrounding areas and decreased in the marginal areas (Figure 6), indicating the effectiveness of regional wetland protection measures, especially in the core region. Considering that wetland landscape patterns have a strong relationship with flood regimes in Dongting Lake [64], the results were of significance for regional flood risk analysis. In addition, the degree of forest fragmentation decreased and then increased, the landscape structure tended to be complex, and the overall degree of aggregation was enhanced, indicating the significant benefits of the project of returning farmland to the forest in the study area [65].

The changes in each landscape metric were closely related to the implementation of regional policies [66]. With the economic development, the area of the impervious surface showed a rapid expansion, resulting in the rise of landscape fragmentation, especially for the regions around the Lake. The PLAND of the impervious surface increased rapidly due to urbanization development. In the recent 30 years, driven by economic development and ecological restoration, the aggregation of forest landscape patches showed a decrease, suggesting the transition of forest and farmland. At the same time, the urbanization of DTLR increased rapidly. Although the area of cropland continued to decline, the intensification showed an increasing trend.

We also found that carbon storage, GPP, water conservation, and crop yield overall increased (Figure 7). Forest carbon storage increased annually, which was mainly attributed to the increase of reforestation and conversion of cropland to forests, which also led to higher total GPP. However, carbon storage in shrubland, grassland, and croplands declined, which was caused by the decline in the area of these land covers. Water conservation has also improved annually since 2007, also due to the impacts of man-made measures such as returning farmland to forest, vigorous ecological restoration, and protection projects. Crop yields showed an upward trend. An important reason for this increase was the mechanization and intensive management of agriculture in DTLR. Despite the decrease in planted area, the continuous improvement in scientific farming techniques (e.g., variety optimization) and management (e.g., intensive farming and nutrient management) contributed to the continuous increase in crop yield. In general, it can be seen that the ecosystem service capacity of DTLR was increasing, and reasonable anthropogenic regulation and measures were conducive to the sustainable development of DTLR [67].

The trade-offs and synergies between different ecosystem services varied considerably. There was a certain synergistic relationship between crop yield and water conservation, and the overall correlation was positive, but the correlation coefficient (r value) was small, possibly because crop yield was mainly influenced by agricultural measures such as irrigation [68]. Crop yield and GPP/carbon storage showed a clear synergistic relationship, which was mainly determined by the close relationship between C3 and C4 plants and GPP/carbon stock [69,70]. The correlation between carbon stock and water conservation was weak, but the overall synergistic effect was mainly because water conservation directly benefited vegetation and crops, indirectly affecting the growth of carbon stock. There was a strong synergistic relationship between GPP and water conservation in the surrounding forest regions, while there was a clear trade-off in the center of lake regions. This relationship was due to the synergistic relationship between vegetation and water conservation, as the increase in water conservation capacity indicated that vegetation and water resources were sufficient and conducive to vegetation growth. In the center lake region, the increase in water region was accompanied by a sharp decrease in croplands, which led to a decrease in GPP and thus a trade-off relationship [71,72]. The interrelationships among ecosystem services could synergistically optimize multiple services and promote sustainable management of regional ecosystems.

A strong nonlinear relationship was represented between landscape patterns and ES (Figure 10), indicating that the benefits and consequences of ecological services from landscape regulation were not linear but depended upon the landscape structure. This dynamic was also reflected by the variations in effectiveness (i.e., sensitivity or derivative of a function) and directionality (elevated or reduced). For example, COHESION index reached a certain value (e.g., COHESION = 75) when carbon storage was at the minimum and then started to increase. This relationship indicates that peak carbon sequestration was influenced by the size and timing of the patch index [41]. The nonlinear relationship between water conservation and landscape patterns in forests was also observed, mainly for forests located in the mountainous areas far from the central water and wetland areas (Figure 10). The interannual variability of the relationship between water conservation and landscape patterns in forests reflected the time-lagged nature of water conservation, to some extent. We found that the regulation of wetland landscape composition on regional water conservation showed a U-shaped response curve (Figure 10). In general, the landscape pattern-ecological service nonlinear relationships revealed they could help to precisely enhance regional ecosystem services and provide scientific references for regional governance measures such as land use planning and landscape pattern regulation policies in the region.

The approach proposed in this study can quantitatively characterize the ecosystem service indicators that can be derived from remotely sensed data. However, this study was based on LUCC, which was limited by the resolution of remote sensing images and the algorithms, inevitably introducing some errors to the process of classifying the land cover and land use in study area. In addition, the indexes employed in this study do not fully reflect the overall level of ESs, and only several natural services were included. Furthermore, social and cultural services were not considered due to the lack of data available in study areas such as the fisheries, nursery products, and ecotourism. Therefore, efforts should be made to investigate the factors affecting ES change, and more detailed data should be used jointly to assess ES changes more comprehensively.

5. Conclusions

This study employs LANDSAT data and the CCDC algorithm to continuously monitor the Dongting Lake region over an extended period, revealing the spatiotemporal evolution process of land use/cover change and landscape patterns. The study further clarifies the temporal legacy effect of spatial location, irreversibility, and controllability.

Over the past 30 years, we have observed a clear trend of ecological restoration in the study area, which is mainly concentrated in the surrounding mountains and hilly areas. The comprehensive restoration of forests reflects the effects of shrub succession/transformation to forests and the conversion of cultivated land to forests. The process of urbanization continues to advance, mainly occurring in the surrounding urban areas and lake plain areas. The cultivated land area continues to decrease while intensification increases and is mainly distributed in the plain areas around the lake area. In addition, the non-linear relationship between landscape patterns and ecosystem services in the Dongting Lake region indicates that the benefits and consequences of ecological services generated by landscape regulation are not linear but have significant landscape structure dependence, which is reflected in changes in effectiveness and directionality.

The impact mechanism of LUCC on landscape patterns and ES was explored to provide a scientific basis for the formulation of ecological environment management policies and land use strategic planning in the Dongting Lake region. It is of great practical and strategic significance for promoting the ecological construction and protection of the Dongting Lake region and promoting the high-quality development of the ecological environment in the Yangtze River Economic Belt. The feasible framework developed in this study for evaluating and exploring the impact of land use change on ecosystem services can provide reasonable references and a scientific basis for ecological protection and construction in other regions around the world.