Land Use and Land Cover Change in the Yellow River Basin from 1980 to 2015 and Its Impact on the Ecosystem Services

: Land use and land cover change is an important driving force for changes in ecosystem services. We deﬁned several important human-induced land cover change processes such as Ecological Restoration Project, Cropland Expansion, Land Degradation, and Urbanization by the land use/land cover transition matrix method. We studied human-induced land cover changes in the Yellow River Basin from 1980 to 2015 and evaluated its impact on ecosystem service values by the beneﬁt transfer method and elasticity coefﬁcient. The results show that the cumulative area of human-induced land cover change reaches 65.71 million ha from 1980 to 2015, which is close to the total area of the Yellow River Basin. Before 2000, Ecological Restoration Project was the most important human-induced land cover change process. However, due to the large amount of cropland expansion and land degradation, the area of natural vegetation was reduced and the ecosystem value declined. Since 2000, due to the implementation of the “Grain for Green” program, the natural vegetation of upstream area and midstream area of Yellow River Basin has been signiﬁcantly improved. This implies that under an appropriate policy framework, a small amount of human-induced land cover change can also improve ecosystem services signiﬁcantly.


Introduction
Land use and land cover (LULC) change is an important indicator that characterizes the impact of human activities on the natural ecosystem [1]. It is the link between human social-economic activities and natural ecological processes [2], and it is also an important content of global climate change and environmental change research [2][3][4]. LULC change is of great significance for in-depth study of biosphere-atmosphere interaction, biodiversity, land surface radiative forcing, biogeochemical cycles, and sustainability of resources and environment [2,3,5].
LULC change is one of the important factors driving changes in ecosystem services provision [6,7]. Ecosystem services (ES) are the benefits that humans obtain from nature through direct or indirect approaches [8,9], and are the core contents of regional sustainable development and human well-beings [10][11][12][13][14]. By changing important ecological processes such as energy exchange, water cycle, soil erosion and deposition, and biogeochemical cycle of the ecosystem, LULC change can affect the provision of ES [15]. Furthermore, there are significant differences in the impact of different LULC change on the ES provision [16,17]. For example, ecosystem protection and ecological restoration projects will increase the key ES such as soil and water conservation, water yield, and nutrient cycling [18], whereas the fragmentation of landscapes caused by urbanization will reduce ES [19].
Ecosystem service value (ESV) assessment is one of the important tools to carry out ecosystem service evaluation [8,20,21]. The monetary value evaluation of ecosystem services has greatly promoted the development of research related to ecosystem services, including raising public awareness of the importance of ecosystem services to society [22,23], safeguard the benefits and interests of different stakeholders [24], guiding decision-making and policy-making [14], and developing knowledge about ecosystem management [25]. It can also provide incentives for ecosystem protection and restoration, such as payments for ecosystem services programs [26][27][28].
There have been many studies about the impact of LULC change on ES [29][30][31], and most of them focus on assessing the overall sensitivity of ES to LULC change [32,33], or analyzing the driving process of LULC changes affecting ES based on temperature, precipitation, population, and economy [34][35][36]. However, most studies only regard LULC changes as the result of land policy or human activities, and fail to distinguish between natural and human-induced processes in land cover changes. Land use (LU) is the use of land by humans for economic and social purposes [14], while land cover (LC) refers to the physical and biological cover over the surface of land including water, vegetation, bare soil, and artificial structures [37]. Although land use and land cover are intimately linked and are often studied as a whole [38], land cover change is influenced both by natural and human-induced processes. The natural processes that affect LC changes mainly include climate change and wildfires [39,40], while human-induced processes that affect land cover include agricultural expansion, water conservancy construction, urbanization, and ecological restoration projects. Therefore, in order to better understand the impact of land policy or human activities on ecosystem services, it is necessary to distinguish between natural and man-made processes in land cover change. Since few ES studies distinguish between human-induced LC changes and nature-process LC changes, and quantitatively evaluate the impact of human-induced LC changes on ES, our understanding of the impact of human-induced LC changes on ES still needs to be strengthened.
The Yellow River Basin (YRB) is an important ecological barrier in northern China, including the Qinghai-Tibet Plateau, the Loess Plateau, and the North China Plain [41]. The ecosystem of the YRB has undergone major changes due to frequent human activities [19,42,43], which in turn affected the ES provision and social economy and development in the YRB [36,44,45]. Investigating the impact of human-induced LC changes on ES in the YRB has important significance for ecological environmental protection and sustainable development policy making in the YRB. The objectives of this paper are to (1) investigate the land use/land cover (LULC) trend and human-induced LC changes in the YRB from 1980 to 2015, (2) assess the ES monetary values (ESV) of the YRB based on the benefit transfer method from 1980 to 2015, and (3) evaluate the impact of LULC changes on ESV.

Study Area
The Yellow River originates from the northern foot of the Bayan Har Mountain on the Qinghai-Tibet Plateau, with a total length of 5464 km. It is the second longest river in China with a basin area of 79,500,000 ha. The YRB is located in 96 • E-119 • E and 32 • N-42 • N, with a length of about 1900 km from east to west and a width of about 1100 km from north to south. (Figure 1a), the topography of the YRB is high in the west and low in the east (Figure 1b). The annual average temperature of the YRB is −13.1-15.3 • C (Figure 1c), and the annual average precipitation of the YRB is 116.2~1038.7 mm (Figure 1d), with a decreasing trend from southeast to northwest. The Yellow River basin (YRB) can generally be divided into four watersheds: the source area (SA), the upstream area (UA), the midstream area (MA), and the downstream areas (DA).

Data Sources
The land use and land cover (LULC) data come from the Resource and Environmental Science Data Center of the Chinese Academy of Sciences (RESDC) (http://www.resdc. cn accessed on 20 August 2021) [1,[46][47][48]. These data are based on Landsat Thematic Mapper/Enhanced Thematic Mapper Plus remote sensing images from different periods and are generated through visual interpretation based on national field surveys. LULC data were used in this study from the year 1980,1990,1995,2000,2005,2010, and 2015. The spatial resolution of LULC data used was 1 km × 1 km, which is resampled from the original dataset (30 m × 30 m) by nearest neighbor method. According to field surveys and records, the overall evaluation accuracy of the LULC primary types is over 94.3%, and the accuracy of the LULC secondary types is over 91.2% [1,[46][47][48], which can meet the data requirements of this study. The LULC types of original data were classified into 6 primary categories (i.e., cropland, forestland, grassland, water area, construction land, unused land) and 24 secondary categories. By overlaying the LULC data and Vegetation Map of China (1:1,000,000), the LULC types are reclassified into seven primary categories including cropland, forest, grassland, wetland, bare land, urban land, and water area and 16 secondary categories (Table A1 in Appendix A). The boundary, Vegetation Map of China (1:1,000,000), temperature and precipitation data of the YRB also come from RESDC (http:// www.resdc.cn accessed on 20 August 2021), where the temperature and precipitation data are obtained by the thin film spline interpolation method using ANUSPLIN software [49], with a spatial resolution of 1 km × 1 km. The DEM data come from the Geospatial Data Cloud Platform of the Computer Network Information Center of the Chinese Academy of Sciences (http://www.gscloud.cn accessed on 20 August 2021), with a spatial resolution of 90 m × 90 m, and the nearest neighbor method was used to resample the data into the resolution of 1 km × 1 km.

LULC Change Transition Analysis
The transition matrix method was used to detect the conversion information of LULC change during the study period. The general form of the LULC change transfer matrix is, where s represents the area (ha); n represents the number of LULC types before and after the transfer; i, j (i, j = 1,2, . . . , n) represent the LULC types before and after the transfer respectively; s ij represents the area (ha) of land type i converted to land type j.

Estimation of Human-Induced LC Changes
In this study, LULC change was divided as human-induced LC changes and naturalprocess LC changes. Land cover or ecosystem transfer and change detection method [50] is used for LULC changes attribution analysis (Table 1). Different natural and humaninduced land cover transition processes are defined through LULC change transition analysis described in Section 2.3. For example, cropland expansion (CE) is defined as the LC transition process from natural vegetation/desert/barren/ice to cropland; urbanization (UR) is defined as the LC transition process from natural vegetation/desert/barren/ice to urban; Ecological Restoration Project (ERP) is defined as the LC transition process from desert/desert/ice to natural vegetation; land degradation (LD) is defined as the LC transition process from natural vegetation/farmland/urban to desert/barren/ice. The details of LC changes definitions can be found in Figure A1. A summary of land policies implemented in the YRB can be found in Table A7. Ecological restoration projects (ERP), water conservancy construction (WCC), urbanization (UR), and land degradation (LD) are considered as human-induced LC change processes. It should be noted that, drawing on the concept of "human influence on land" proposed by Zalles et al. [5], human-induced LC change only considers human production and life processes that directly affect the land, which is different from human occupation of net primary production [51] and human footprint [52], and did not consider factors such as population density and spatial distance [53] as an indicator of human activity.

Estimation of ESV
The Benefit Transfer Method refers to applying existing values or information from one or multiple previously established study sites to other unstudied sites with similar characteristics where the site-specific primary valuation is lacking [30,54]. The method proposed by Xie et al. [21] was used to evaluate the ecosystem service value (ESV) of the Yellow River Basin. The formula is as follows, where ESV is the value of ecosystem services, in US dollar (USD); A k is the area of the k-th LULC type, in ha; VC k is the equivalent coefficient of the k-th ESV, in USD ha −1 .
In the method proposed by Xie et al. [21], the VC k was estimated as the product of the standard equivalent factor and the equivalent coefficients. The standard equivalent factor, which serves as the benchmark for other ecosystem services, refers to the value provided by the natural ecological components of a certain ecosystem per unit land area. Based on the labor value theory [55], Xie et al. [21] take the average net benefit excluding the cost value of human input as the standard equivalent factor of the Chinese ecosystem and estimate the value of the standard equivalent coefficient of 503.2 USD ha −1 in 2010. Equivalent coefficient is the weight coefficient of the service value of each ecosystem. Through a professional questionnaire survey of 500 Chinese ecologists and scholars in related disciplines, Xie et al. [21,56] proposed an ESV equivalent coefficient table. In the ESV equivalent coefficient table, the ecosystem includes six primary types (i.e., cropland, forest, grassland, wetland, wasteland, and waters) and 14 secondary types. Ecosystem services include four primary classifications (i.e., provisioning services, regulating services, habitat services, and culture and amenity services) [57] and 11 secondary classifications ( Table 2). The method proposed by Xie et al. [21] has been used to calculate the monetary value of china's ecosystem services, and researchers have used this method to calculate the monetary value of ecosystem services in the YRB. Therefore, it is feasible to apply this method to study the monetary value of ecosystem services in the YRB.

Elasticity-Sensitivity Analysis
Elasticity coefficient (E), a concept from economics, was used to study response of ESV to LULC changes [30], and its formula is as follows, where ESV is the total estimated value of all ES, the unit is USD; ∆A k is the land transfer area of k-th LULC type, the unit is ha; A k is the land area of the k-th LULC type, the unit is ha; i and j represent the beginning and end of the study period, respectively; k represents LULC category. Elasticity (E) is a measure about percentage change in ESV as a result of percentage change in LULC. If a small area conversion resulted in a significant ESV change, then the elasticity is large. According to Jiang et al. [30], elasticity (E) can be divided into three categories, 0 < E < 0.5, 0.5 < E < 1, and E > 1, indicating that response of ESV to LULC change is inelastic, elastic, and very elastic. In 1980, the area of grassland, cropland, and woodland in the YRB is about 61.36 million hectares (Mha), accounting for 86.76% of the total area of the basin; the area of barren land, wetland, and water areas is about 7.88 Mha, accounting for 11.14% of the total area of the basin; and the area of urban land is about 1.49 Mha, accounting for 2.10% of the total area of the basin (  Table A2. Compared with 1980, the area of urban land, water bodies, and urban lands in the YRB increased by 45.10%, 3.44%, and 2.58% in 2015, while wetland, bare land, grassland, and farmland decreased by 7.18%, 2.27%, 1.31%, and 1.29%, respectively. However, only the average annual change rate (ACR) of urban land area continued to increase (Table A3), with the ACR reach its maximum value in 2010-2015 period (0.0673 Mha/a), while other LU/LC trends showed significant reversals or fluctuations, for example, the LU/LC with the largest variation in the ACR is grassland, and its maximum and minimum ACR values occurred in 1990-1995 period (0.1540 Mha/a) and 1995-2000 period (−0.1896 Mha/a), respectively. Details of average annual LULC change rate in YRB from 1980 to 2015 can be found in Table A3.

Watershed Scale
In 1980, the LULC in SA and UA of the YRB was dominated by grasslands, with an area of 8.14 Mha (75.60% of SA) and 14.78 Mha (54.51% of UA), respectively ( Figure 2). The LULC in MA and DA of the YRB was dominated by cropland, with an area of 11.88 Mha (39.84% of MA) and 1.87 Mha (67.81% of DA), respectively. Compared with 1980, the grassland area in the SA and UA decreased by 0.72% and 2.09% respectively in 2015, the area of cropland in the MA decreased by 2.70%, and the area of cropland in the DA increased by 0.61%. Details of LULC areas in SA, UA, MA, and DA from 1980 to 2015 can be found in Table A2.
In the SA of the YRB, all the LULC trends showed fluctuations (Table A2), where the LULC with the largest variation in ACR is barren land, followed by grassland; in the UA of the YRB, all the LULC trends also showed fluctuations, where the LULC with the largest variation in ACR is cropland, followed by grassland; in the MA of YRB, the urban land area continued to increase from 1980 to 2015, with the ACR reached its maximum value in 2010-2015 period (0.0313 Mha/a), while other LULC trends showed significant reversals or fluctuations, and LULC with the largest variation in ACR is grassland, followed by forest; in the DA of the YRB, the urban land area also continued to increase from 1980 to 2015, with the ACR reached its maximum value in 2000-2005 period, while other LULC trends showed significant reversals or fluctuations, and LULC with the largest variation in the ACR is grassland, followed by cropland. Details of average annual LULC change rate in SA, UA, MA, and DA from 1980 to 2015 can be found in Table A3.  Table A5.    Table A5.

Watershed Scale
In 1980, the ESV in the SA was ranked as (Figure 4) regulating services > habitat services > provisioning services > cultural and amenity service; the ESV in the UA was ranked as regulating services > habitat services > cultural and amenity service > provisioning services; the ESV in the MA was ranked as regulating services > habitat services > cultural and amenity service > provisioning services; the ESV in the DA was ranked as regulating services > habitat services > cultural and amenity service > provisioning services. Compared with 1980, the total amount of ESV in the SA, UA, MA, and DA of YRB decreased by 0.62%, 0.85%, 0.09%, and 11.29% respectively, reaching 873.  Table A5.

Implications for Sustainable Development
Humans can significantly alter environments through the conversion of natural vegetation, cropland, reservoir, urban, and other land uses [5]. This process can change the regional climate by adjusting the surface energy, water balance [58,59], and greenhouse gas emissions [60], and it can also lead to a decline in biodiversity through the loss, modification, and fragmentation of habitats [61], and ultimately affect ES and human well-being [6,7].
In the past 35 years, the cumulative area of human-induced LUCC in the YRB has reached 65.71 Mha (i.e., cumulative areas in Figure 3), which is close to the total area of YRB; the ESV of the YRB is between 387.71-394.25 billion USD (Figure 4), which is close to the ESV of the Tibet Autonomous Region in China [21]. Since the 1970s, vegetation restoration and reconstruction has been the important measure for ecological protection and soil erosion control in the YRB [62]. The area of ERP in the YRB in 1980-2000 period was 27.77 Mha (Figure 3), but the area of natural vegetation such as woodland, grassland, and wetland in the YRB declined from 1980 to 2000 (Figure 1), resulting in a decline in ESV (Figure 4). This contradiction shows that ERP has not reversed the decline of natural vegetation and ES in the YRB. The reason is that FR reduces the area of natural vegetation [63]. For a long time, agricultural activities have been an important human activity in the YRB in response to population growth [45,63]. From 1980 to 2000, the area of FR in the YRB reached 25.58 Mha. A large amount of natural vegetation has been replaced with cropland, coupled with the land degradation (about 6.58 Mha) caused by unreasonable human activities such as overexploitation of [64] and overuse of water resources [65]. Eventually, the area of natural vegetation in the YRB decreased from 1980 to 2000.
Since the year 2000, benefit from the implementation of the "grain for green program" (Table A7), the natural vegetation of the YRB, especially the natural vegetation in the UA and MA (i.e., the Loess Plateau) has been significantly improved [66,67]. The area of cropland in the YRB decreased by 0.51 Mha from 2000 to 2015, the area of forest land increased by 0.27 Mha (Figure 1), and ESV increased by 2.84 billion USD (Figure 4). It is worth noting that the area of ERP in 2000-2015 period was only 0.57 Mha, and the elasticity coefficient of ESV to LULC change in 2000-2015 period increased significantly compared with that in the 1980-2000 period ( Figure 5). This shows that under the guidance of inappropriate policies, a large amount of LUCC may be harmful to the ecosystem, while under the guidance of appropriate policies, a small amount of LULC change may be beneficial to the ecosystem.
In addition, UR is also an important human-induced LULC change process in YRB, and the area of UR gradually increased from 1980 to 2015 (Figure 1). It is foreseeable that the UR process will become the most important LULC change process in the YRB in the future, and the ecological security issues that may arise from this process need more attention [65,68].

Uncertainties in ESV Assessment
ESV assessment is usually regarded as an important tool for sustainable land development decision-making and trade-offs [20,69,70]. In this study, the benefit transfer method proposed by Xie et al. [21] was used to calculate the ESV of the YRB. However, recent studies have calculated that the ESV in the YRB is 264.42-266.4 billion USD [71] or 78.17-96.39 billion USD [72]. There is a big difference between these results. It is worth noting that, in these studies, different versions of the value equivalent method [56,73] were used to calculate ESV, and different methods were used to revise the key coefficients [71,72]. We also tried to add revisions to the CPI index in the calculation of ESV to obtain comparable ESV values in different periods ( Figure A2). Based on this calculation, the ESV of the YRB increased from US$80.53 billion to US$452.43 billion from 1980 to 2015. However, this result to a large extent only reflects the influence of economic factors and conceals the impact of LULC changes on ESV. Therefore, the result was not used in the study of ESV on LULC changes in the YRB. This shows that the benefit transfer method has huge uncertainties in calculating ESV [74]. Some studies have conducted sensitivity-analysis based on elasticity-sensitivity coefficient to study the impact of changes in ES coefficients on ESV assessment [75,76]. However, Aschonitis et al. [77] pointed out the elasticity-sensitivity coefficient should be used for ranking the importance of land use and land cover (LULC) changes rather than assessing the robustness and sensitivity of the ES coefficients.
ES trade-off refers to the situation where one type of ES increases while the other type of ES decreases [78]. In the current ES research community, ES trade-offs are often regarded as key indicators that guide land policy and management to support the provision of ES [79]. It is generally accepted that all decisions that involve trade-offs involve valuation, whether implicitly or explicitly [8,22]. Due to the ability to examine multiple ESs at the same time, ESV assessment has been used in many studies to study trade-offs between different ESs [80,81]. Since the ES trade-offs can be studied from the perspective of relative value, this can reduce the uncertainty of ESV to a certain extent [79]. However, affected by the temporal and spatial scales [79] and landscape history [82], the uncertainty of ES trade-offs are not well understood.
It is generally accepted that when ESV assessment is used to raise public awareness rather than specific policy guidance, lower confidence is acceptable [55,62]. However, it has been reported that about 82% of ecosystem services mapping studies have cited decision-making purposes [63]. Therefore, it is necessary to construct an uncertainty evaluation framework to increase confidence in ESV assessments and trade-offs to support decision-making in the future.

Limitations
Human-induced land cover (LC) changes are only defined by land cover (LC) transition, but the impact of different categories and intensities of land use on land cover was not considered. For example, the grazing area and grazing intensity will significantly affect the grassland, but it does not necessarily cause the grassland transfer to other land cover type. This will be one of the important research directions for future LULC changes.
Time-series land use and land cover (LULC) datasets are used to analyze the response of ecosystem services to LULC changes in this study. However, due to the coarse temporal resolution (≥5 years), the time lags of LULC changes to ecosystem services (ES) are unable to be detected. The time lags of ES's responses to LULC changes is the key direction of future land change sciences.

Conclusions
Land use and land cover (LULC) transition matrix and benefit transfer methods were used to study LULC changes in the Yellow River Basin (YRB) and their impact on ecosystem services values (ESV). From 1980 to 2015, the area of urban land, water bodies, and urban lands in the YRB increased by 45.10%, 3.44%, and 2.58%, while wetland, bare land, grassland, and farmland decreased by 7.18%, 2.27%, 1.31%, and 1.29%, respectively. The LULC change area in the YRB showed an increasing trend before the year 2000 and a decreasing trend after the year 2000. From 1990 to 2015, the human-induced LC change dominates in YRB, and the natural-process LC in the YRB accounted for less than 25% of the total LULC change areas. The proportion of ecological restoration projects (ERP) areas was the highest before the year 2010, the proportion of urbanization (UR) was the highest in 2010-2015. From 1980 to 2015, the ESV of the YRB decreased by 0.94%, valued at 390.55 billion USD, and the elasticity of ESV changes in response to LUCC in the YRB is always an inelastic process. We found that under the guidance of inappropriate policies, a large amount of LUCC may be harmful to the ecosystem, while under the guidance of appropriate policies, a small amount of LULC change may be beneficial to the ecosystem.   Control the expansion of sandy/desertified land, and mitigate wind erosion of sand/soil and dust storms in northern China via forest plantation, mountain closure, and sandy area regeneration.
2 Natural Forest Conservation Program 1998-2020 Increase the area of cultivated land and revenues via consolidation (reorganizing and merging fragmented and underused land), reclamation, constructing high-quality cropland, and improving land use and management. 3 Grain for Green Program 1999-2020 Prevent soil erosion, mitigate flooding, store carbon, and improve livelihoods by increasing forest and grassland cover on cropped hillslopes and converting cropland, barren hills and wasteland to forest.

4
Beijing-Tianjin Sand Source Control Engineering

2001-2022
Reduce desertification and dust storms, and improve the environment in the Beijing/Tianjin area via reforestation, grassland management, and water conservation, relocating affected people and establishing basic governance of desertified lands. Raise rural quality of life, incomes and food security through land reform, land management, ecological construction, agricultural infrastructure and industry development, and production/efficiency gains using science and technology. 6 National Land Consolidation Program. 1997-2020 Increase the area of cultivated land and revenues via consolidation (reorganizing and merging fragmented and underused land), reclamation, constructing high-quality cropland, and improving land use and management.