Introduction

Ecosystem services (ESs) are those life-support goods and services received directly or indirectly through the structure, processes, and functions of ecosystems and whose formation, supply, and distribution are profoundly influenced by land use (Xie et al. 2015). Land creates a large amount of ecosystem service value (ESVs) for humans, and the majority of social and economic activities use it as a carrier. Moreover, land plays a fundamental role in food security and ecological safety, holding a supply value (provision of food, raw materials, and water resources), regulating value (purification of the landscape, gases, climate, and hydrology), supporting value (soils, nutrient cycling, and biodiversity), and cultural value (aesthetic landscape; Costanza et al. 1997). With population growth and rapid urbanization (Fan et al. 2022), land use patterns are gradually transforming toward semi-natural and semi-artificial ecosystems (Cumming et al. 2014; Li et al. 2016), and the resulting changes in material and energy flows affect the ability of ecosystems to provide services. Therefore, unsuitable land use will endanger regional ecological security and even limit urban sustainable development (Huang et al. 2021). With the construction of large-scale, long time series and the diffusion of high-intensity human activities, watersheds provide an important material base and ecological services for human survival (Xie et al. 2022). At the same time, as the population gathers in river basins and the scale of a city expands, the demand for ESs such as food supply, water supply, air purification, water containment, and cultural recreation will continue to increase. The HRB (Huaihe River basin) is located in the north–south climate transition zone of China. It is an important ecological barrier, as well as a significant coal and energy base in eastern China, and is a representative case of China for the purposes of this study. The ecological environment and socio-economic development of this region are particularly sensitive to land use changes; hence, it is important to study the land use changes in this basin. In this context, the study of land use impacts on ESV changes in watersheds under multiple scenarios is conducive to the construction of a new pattern of spatial development and protection of watersheds and provides technical support and scientific information.

Changes in land use/land cover directly affect ESVs (Gashaw et al. 2018). Therefore, the estimation of ESVs based on land use changes is the most direct and effective method to quantify the loss of EVS over time. The spatial pattern of ESVs reflects the spatial and temporal evolution of regional ecosystems, which is important to identify regional ecosystem service problems, maintain balanced regional ecological development, and promote regional sustainable development. ESV is a quantitative estimate of the capacity of ESs and can be assessed in two main ways: monetization and energetic valuation (Li 2019). The monetary form of ESV can be easily understood by people and used by decision-makers; it can effectively assist spatial planning, ecological control, and ecological restoration and is widely used (Crossman and Bryan 2009; Crossman et al. 2011; Groot et al. 2012). ESVs are the basis to describe the spatiotemporal evolution of ESs; hence, their accurate accounting is critical. At present, several scholars in China and abroad have generated a wealth of research results on the effects of land use change on ESVs (Assefa et al. 2021; Hu et al. 2022; Song et al. 2017; Tan et al. 2020). In 1997, Constanza et al. developed ESV coefficients for different land use types based on an assessment of global ESVs, which laid the foundation for related studies. Xie et al. (2008) established a base equivalence scale for ESVs in different regions of China, which has been used by a wide range of scholars. ESVs have been explored in relation to three main topics: definition and classification (Braat 2012), value assessment (Xie et al. 2017), and the relationship between ecosystems and social development (Garcia et al. 2018). Specifically, several studies focused on the underlying theory, driving mechanisms, and spatiotemporal variability of ESs, covering national, regional, provincial, municipal, and county scales. In this study, we considered the special situations of different years and regions, continuously improved the adjustment of value coefficients, conducted an assessment of the spatiotemporal variation patterns of regional ESV under different scales of land use changes, and proposed optimized temporal sequencing schemes and spatial development strategies and differences in ESV under different scenario simulations.

Land use patterns can affect ESVs. With the advancement of research, scholars have tried to analyze future land use changes and their impacts on regional ESVs to develop sustainable land use measures Land use simulation is a current research hotspot, and related studies have been conducted using CLUE-S (Niu et al. 2021), SLEUTH (Niu et al. 2021), SD (Bing et al. 2016), FLUS (Liu et al. 2021), and CA-Markov (Guzman et al. 2020), among others. The Patch-generating land use simulation (PLUS; Liang et al. 2021), as a new land use simulation model, has higher simulation accuracy than the commonly used prediction models, and its results can better support policies and sustainable development. Xie et al. (2022) found that ESVs were assessed based on general scenarios and single scales, and comparisons of ecological conservation, cultivated land conservation, and multi-scenario ESVs have been relatively weak. Previous studies focused on a single scale based on macro- (Ye et al. 2021) or micro-scale grid cells (Gao et al. 2021), thereby lacking multi-scale coupled exploration. For example, Liu et al. (2019) assessed the impact of urbanization-induced land use change on ESs under a general scenario and developed an ESV matrix of land use shifts; while Arowolo et al. (2018) assessed ESV changes in response to land use dynamics in Nigeria based on a graphical use shift approach. However, ESVs under a single scenario cannot reveal the differences between natural development and ESVs under the constraints of other policy scenarios, which hinders their ability to optimize land use structure and measure the trade-offs between socio-economic development and ecological conservation.

Therefore, this paper simulated use changes in the HRB under four scenarios of natural development, ecological development, cultivated land development, and urban development to 2030. Five periods of land use data were used from 2000, 2005, 2010, 2015, and 2020, with the help of the PLUS model. The spatial distribution of ESVs at different scales was explored by combining municipal, county, and grid scales. On this basis, a hotspot analysis was performed to assess the spatial distribution of ESVs under different preferences and the degree of aggregation; moreover, the contribution of land use transformation to ESV change was introduced to explore the influence of land use transformation on ESVs. The results provide a scientific basis for optimizing land use structure and environmental protection in the HRB.

Materials and methods

Overview of the study area

The Huaihe River is one of the seven major rivers in China, originating at the northern foot of Taibai Peak in Tongbai Mountain, Henan Province, and flowing from west to east through the four provinces of Henan, Hubei, Anhui, and Jiangsu, with a total length of about 1000 km. The HRB (30°55′–36°36′N, 111°55′–121°25′E) spans 47 prefectures in the five provinces of Henan, Anhui, Hubei, Jiangsu, and Shandong. It has an area of 480,000 km2, with an average annual temperature of 11–16 °C and an average annual precipitation of about 927 mm. With a dense population, fertile land, abundant resources, and convenient transportation, it is an important base for grain production, energy and mineral extraction, and manufacturing in China (Fig. 1).

Fig. 1
figure 1

Geographical location of the study area: (a) location of the HRB in China; (b) location of administrative cities within the HRB; (c) elevation distribution of the HRB

Full size image

Data sources

Total crop coverage, production, and yields for 2000, 2005, 2010, 2015, and 2020 were obtained from the statistical yearbooks of each municipality in the study area and the National Compilation of Agricultural Product Cost Craft Resources for 2000–2020. Data such as constraints used to describe land use conversion (Table 1) were employed to determine meta-cellular conversion rules.

Table 1 Description of driver factor data
Full size table

Research framework

In this study, we analyzed the spatial and temporal evolution of land use in the HRB from 2000 to 2020. We simulated scenarios for inertia, ecological land, cultivated land, and urban development in 2030 based on the PLUS model combining five-dimensional drivers. We quantified the ESVs and ESVs granularity of land use at municipal, county, and grid scales. Then, we explored the spatial distribution of ESVs at multiple scales based on hotspot analysis and introduced the degree of contribution of land use conversion to ESVs within a specific framework (Fig. 2).

Fig. 2
figure 2

Research framework

Full size image

Research methods

Ecosystem service value (ESV) estimation

In this paper, the equivalence factor method (Table 2) established by Xie et al. (2008) was used to estimate the ESVs in the HRB for five periods from 2000 to 2020, based on the actual land types of the study area. Changes in ESVs were analyzed over multiple years.

  1. (1)

    Determination of the economic value of one standard unit of ESs

Table 2 Ecosystem service value equivalent per unit area
Full size table

The economic value of natural cereal produced by 1 ha of plowland was considered an equivalent value of the ecosystem. The equivalent value of a standard ES is worth seven times the economic value of grain produced per unit area in that year. Given the availability of statistical data, the economic value of the main food crops (i.e., wheat, corn, and soybean) in the urban areas of the study region from 2000 to 2020 was selected for correction, and their average value was calculated using Eq. (1). The ESV of one standard unit in the study area was 45,010.88613 CNY·hm−2a−1.

$${E}_{a}=\frac{1}{7}\sum_{i=1}^{n}\frac{{m}_{i}{p}_{i}{q}_{i}}{M}$$
(1)

where Ea is the economic value of one standard ecosystem service unit in CNY·hm−2a−1, i is the type of food crop, mi is the average price of the i-th food crop in the study area in CNY/kg, Pi is the yield of the i-th food crop in kg/hm2, qi is the planted area of the i-th food crop in hm2, and M is the total planted area of the food crop in hm2.

  1. (2)

    Calculation of the ESV

The ESV of the study area was calculated as follows:

$$ESV=\sum_{i=1}^{n}{A}_{i}\times V{C}_{i}$$
(2)
$$V{C}_{i}=\sum_{j=1}^{k}E{C}_{j}\times {E}_{a}$$
(3)

where ESV is ecosystem service value in CNY/year, i is the land use type, j is the ecosystem service type, Ai is area of type I land use in hm2, VCi is the ESV per unit area of the i-th land use type in CNY and of the i-th food crop in the study area in CNY·hm−2a−1, and ECj is the j-th ESV equivalent of a given land use type. Because the ESV indicator of construction land was 0 or negative, it was not considered in this paper. K is the number of ecosystem service types; and Ea is the economic value of one standard ecosystem service unit in CNY·hm−2a−1.

  1. (3)

    Correction based on the biomass factor of the study area

Because of the large area of the HRB, the wide distribution of vegetation cover, and the large differences between the ecosystem and other regions in China, the ESV coefficients needed further revision. In this paper, we referred to the ecosystem biomass correction factor proposed by Xie et al. (2005) for each province in the study area and selected the average value of 1.42 to calculate each ESV coefficient for different land use types (Table 3).

Table 3 Ecosystem service value equivalent per unit area in the HRB
Full size table

ESV under scale effects

Based on the exploration of ESV coupling at multiple scales, ESV over different scales can be better analyzed from a spatial perspective. This paper conducted a multi-spatial scale ESV relationship study based on municipal administrative units, county administrative units, and 7.5-km grid cells and used multi-scale cold hotspot spatial clustering to describe the ESV spatial distribution.

ESV sensitivity analysis

Ecological sensitivity (CS) is an important indicator of the degree of dependence of ESV on the value factor. CS was calculated by adjusting the value equivalents of each category upward and downward by 50%, respectively, as follows:

$$CS=\frac{\left({ESV}_{j}-{ESV}_{i}\right)/{ESV}_{i}}{\left({VC}_{jk}-{VC}_{ik}\right)/{VC}_{ik}}$$
(4)

where the sensitivity index is denoted by CS. If CS > 1, it means that ESV is elastic to VC, and if CS < 1, it means that ESV is inelastic to VC. ESV represents the total value of ecosystem services; VCk is the ESV coefficient of land in category K; and j and i represent the ESV before and after the value coefficient adjustment case, respectively.

PLUS model

The PLUS model is based on raster data. It is a patch-generated land use change simulation model coupled with a new land expansion analysis strategy (LEAS) and a CA model based on multiple types of random patch seeds (CARS). PLUS can describe the factors influencing land use change across categories with higher simulation accuracy. This model uses a random forest algorithm to obtain the development probability of each class by extracting land use expansion in period 2 and then simulates and predicts the future land use based on a CA model with multi-class random patch seeds.

Kappa coefficient and FOM coefficient accuracy verification

To determine the accuracy of the simulated land use data, 2020 land use data were simulated based on the 2000 and 2010 base period land use data. Actual and simulated data accuracy was verified using Kappa and FOM (Figure of Merit) coefficients. The FOM index for cell-level agreement has been widely used to quantify the accuracy of land change models (Pontius Jr and Millones 2011; Pontius et al. 2008). The 2020 Kappa coefficient was 0.68, and the FOM was 0.05. The formulas employed are as follows:

$$Kappa=\frac{{\sum }_{i=1}^{k}{n}_{ii}-{\sum }_{i=1}^{k}{n}_{i}+{n}_{+i}}{{n}^{2}-{\sum }_{i=1}^{k}{n}_{i}+{n}_{+i}}$$
(5)

where n is the total number of validation pixels, nii is the number of correctly classified pixels in class i, ni is the total number of classified pixels in class i, n+i is the total number of reference pixels in class i, and k is the number of classes.

$$Fom=\frac{B}{A+B+C+D}$$
(6)

where A represents the error area caused by actual land use change and is predicted to be unchanged, B represents the area that was accurately predicted, C represents the error area caused by prediction of the wrong land use type, and D represents the error area when no change occurred but it was predicted.

Different scenario settings

Based on the Markov module in the PLUS model, four different scenarios of land use types in the HRB in 2030 were constructed. More in detail, the inertia development scenario was based on current land use data for 2020 and predicted the area and spatial distribution of each land use type in 2030. Considering the further promotion of policies related to urban development and county construction in the HRB, the urban development scenario was set to increase the probability of transferring cultivated land, forest, and grassland to construction land by 20% and reduce the probability of transferring construction land to landscape types other than cultivated land by 30%. Considering the protection of ecological patterns in the HRB, in the ecological protection scenario, the probability of transferring cultivated land and forest to construction land was reduced by 40%, while the probability of transferring unused, water, and grassland to construction land was reduced by 20%, and the probability of transferring construction land to forest was increased by 20%. Considering that the HRB is an important grain base in China, in the cultivated land protection scenario, the probability of transferring forest, grassland, water, and unused land to cultivated land was increased by 20%, while the probability of transferring construction land to cultivated land was increased by 30%, and the probability of transferring cultivated land to other land types was reduced by 40%. Under the different scenarios, water, nature reserves, construction land, and cultivated land were used as constraints to limit their arbitrary conversions.

Land use simulation drivers

Referring to previous studies (Sun et al. 2022; Jiang et al. 2021; Aytac 2022), a total of 18 drivers were selected in the PLUS module to simulate land use type changes in five dimensions: topographic, climatic, social, accessibility, and natural (Fig. 3).

Fig. 3
figure 3

Drivers of land use evolution

Full size image

ESV hotspot analysis

Hot spot analysis (Getis-Ord \({G}_{i}^{*}\); Getis and Ord 1992) is commonly used to identify the spatial distribution of cold and hot spot areas. In this paper, hot spot analysis reflected whether ESVs in the HRB had high-value clustering (hot spot) or low-value clustering (cold spot), to determine where clusters occurred in space. The statistical significance of \({G}_{i}^{*}\) was tested using standardized Z values, where a positive and higher Z value indicated tighter clustering for high values (hot spots) and a negative and lower Z value indicated tighter clustering for low values (cold spots).

$${G}_{i}^{*}=\frac{\sum\limits_{j=1}^{n}{w}_{ij}{x}_{j}}{\sum\limits_{j=1}^{n}{x}_{j}}$$
(7)
$$Z\left({G}_{i}^{*}\right)=\frac{{\sum }_{j=1}^{n}{w}_{ij}{x}_{j}-\overline{X}{\sum }_{j=1}^{n}{w}_{ij}}{\sqrt{\frac{n{\sum }_{j=1}^{n}{w}_{ij}^{2}-{\left({\sum }_{j=1}^{n}{w}_{ij}\right)}^{2}}{\left(n-1\right)}}}$$
(8)
$$S=\sqrt{\frac{1}{n-1}\sum_{j=1}^{n}{x}_{j}^{2}-{\left(\overline{X }\right)}^{2}}$$
(9)

where \({G}_{i}^{*}\) is the aggregation index for patch i. wij is the spatial weight between raster i and j; if the distance between raster i and raster j is within the specified range, wij = 1, otherwise wij = 0. n is the total number of patches; \(\overline{X }\) is the mean value of all plaques in the space; and S is the standard deviation of all patch attribute values. The clustering characteristics of the low-value (cold spots) and high-value (hot spots) areas were determined by the Z-values. The ESV cold- and hot-spot partitioning was carried out with reference to previous studies (Zhao et al. 2022; Table 4).

Table 4 ESV cold and hot spot partitioning
Full size table

The contribution of land use conversion to ESV

In this paper, the effect of land use change on ESV was analyzed following Zhang et al. (2020), and the contribution of land use conversion to ESV change was calculated as follows:

$${ESCI}_{ij}=\frac{\Delta {Q}_{ij}}{\Delta {S}_{ij}}\times {P}_{ij}$$
(10)

where \({ESCI}_{ij}\) is the contribution of land use conversion to ESV change, \(\Delta {Q}_{ij}\) quantifies the amount of ESV change from the conversion of land use i to land use j, \(\Delta {S}_{ij}\) is the area of land use i converted to land use j, and \({P}_{ij}\) is the proportion of the area of land use i converted to land use j in the total converted area. When ESCI > 0, the contribution is positive and increases as ESCI increases.

Results

Analysis of land use time series change

From 2000 to 2020, the main land use types in the HRB were cultivated land, construction land, and forest (Fig. 4). Cultivated land was the most dominant land type, accounting for 77.80% of the total study area, followed by construction land, accounting for 11.56%, and forest, accounting for 6.35%. Each land type in the district was relatively concentrated; cultivated land was distributed across all directions, while construction land was mainly concentrated in the central and northern parts. Forest was concentrated in the southwest and grassland in the north and west. Except for large lakes where water was concentrated, other water bodies were scattered in patches. Unused land was concentrated in the northern fringe area. Changes in land use type area mainly showed a continuous decrease in cultivated land, grassland, and unused land, a continuous increase in construction land and forest, and an increase followed by a decrease in water. The cultivated land area decreased by 28,344.69 km2, the grassland area decreased by 2087.81 km2, the unused land area decreased by 581.88 km2, the water body area increased by 1511.875 km2, and the construction land area increased by 26,914.56 km2. From 2000 to 2010, the cultivated land area changed significantly, decreasing by 15,139.94 km2. The most significant change in construction land area was an increase of 14,630.19 km2 from 2010 to 2020 (Table 5).

Fig. 4
figure 4

Land use in the study area from 2000 to 2020

Full size image
Table 5 Area of land use types (unit: km2)
Full size table

Multi-scale ESV estimation and variability analysis

Land use change can cause regional ESV changes. The ESV in the HRB was 222,019 × 1012 CNY in 2000, 235,015 × 1012 CNY in 2005, 234,419 × 1012 CNY in 2010, 229,885 × 1012 CNY in 2015, and 224,759 × 1012 CNY in 2020, following an overall pattern of a first increase, followed by a decrease. Over the 20 years investigated, the largest change in ESV for water was 0.86 × 1016 CNY. Cultivated land contributed the most to ESV, with a decrease of 0.73 × 1016 CNY over 20 years. In parallel, forest ESV increased by 0.31 × 1016 CNY, grassland ESV decreased by 0.16 × 1016 CNY, and the ESV of unused land changed the least, decreasing by 409,098.86 CNY.

The total ESV in the HRB was calculated using the ArcGIS 10.3 spatial statistics tool and was classified into five major categories using the natural breakpoint method: low, lower, medium, medium–high, and high values (Fig. 5). Municipal scale, high-value ESVs were located mainly in the south and east, and low-value ESVs were distributed the research area around. At the county scale, ESVs were mainly distributed with low and lower values, with less distribution of medium–high and high values. The low and lower values were mainly distributed in the northwest and west, whereas medium–high and high values were mainly distributed in the south, and medium values were distributed in the central and northern parts of the region. At grid scale, low and lower values occupied most of the area. The median zone was mainly located in the south. Medium–high and high-value grids were clustered and distributed in strips in the middle of the region.

Fig. 5
figure 5

ESV at multiple scales in the HRB, 2000–2020

Full size image

The distribution of land use ESV granularity data classes was not consistent across years at different scales (Fig. 6). Municipal scales were largely unchanged except for 2015, when low-value areas dominated. At the county scale, the distribution was more regular across the years, with mainly low-value areas, followed by lower values. At the grid scale, the variability was higher in 2010 and 2020, lower values dominating in 2010 and 2020.

Fig. 6
figure 6

Number of multi-scale ESV particle sizes in the HRB in 2000–2020: a municipal scale; b county scale; c grid scale

Full size image

Hot spot analysis of ESVs under land use change

The \({G}_{i}^{*}\) hotspot analysis tool was used to explore the spatial clustering characteristics of ESVs at different scales by mapping ESV cold hotspots for the five periods from 2000 to 2020 at three scales: municipal, county, and grid (Fig. 7). ESV spatial distribution characteristics remained unchanged at the same scale, and there were significant differences in the spatial distribution at different scales. At the municipal scale, very significant cold spots and significant cold spots were clustered in the western and northern fringe areas, and cold spots were mainly clustered in the central part. Hot spots, significant hot spots, and very significant hot spots were mainly clustered in the eastern and northeastern edges and parts of the southern region, while the non-significant areas were more dispersed. The cold hot spots had a similar spatial distribution at the county and municipal scales; however, there was high variability at the grid scale. Very significant cold spots were found in scattered patches; significant cold spots and cold spots were concentrated in the northwestern and central regions; non-significant regions were more scattered in distribution; and very significant hot spots were mainly located in the eastern, western, and northwestern fringes and central part of the region, with more significant distribution characteristics.

Fig. 7
figure 7

Multi-scale ESV cold hotspots in the HRB, 2000–2020

Full size image

As can be seen in Fig. 8, the largest percentage of non-significant granularity was found at the municipal and county scales. The cold spot particle size was followed by the significant cold spot particle size at the municipal and county scales. At the grid scale, the significant cold spot particle size was predominant, followed by the non-significant particle size.

Fig. 8
figure 8

Number of multi-scale ESV cold and hot spot granularity in the HRB, 2000–2020: a municipal scale; b county scale; c grid scale

Full size image

Land use change and ESV analysis under multiple scenarios

Multi-scenario land use prediction simulation

Using 2020 as the base period, the land use and expansion changes in the study area in 2010 and 2020 were simulated for 2030 in the HRB under four scenarios by setting different constraints (Fig. 9). The statistics showed a decrease of 12,763 km2 in cultivated land under the inertial development scenario compared to the year 2020. Forest area decreased by 301.188 km2, grassland area decreased by 619.063 km2, water area decreased by 121.188 km2, and the areas of unused land and construction land increased by 164.625 km2 and 13,639.188 km2, respectively. Under the ecological development scenario, forest and construction land areas increased by 652.813 km2 and 7,792 km2, respectively. Under the cultivated land development scenario, cultivated land area increased significantly by 20,206.563 km2. Under the urban development scenario, there was a significant increase in the area of land for construction of 16,634.375 km2.

Fig. 9
figure 9

Simulation of multi-scenario land use projections for 2030

Full size image

The spatial distribution (Fig. 9) under inertial development (Fig. 9a) exhibited a sharp expansion of construction land, spreading around in a star-like pattern and encroaching on most cultivated land. The central and northern parts received significantly more land for construction, while the southern part had a large area of forest, which was less affected by construction land. Under the cultivated land development scenario (Fig. 9b), there was a dramatic expansion in the area of cultivated land, a large reduction in construction land and forest, and a significant reduction in the area of forest in the south. Under the urban development scenario (Fig. 9c), there was a corresponding encroachment of cultivated land area within the expansion of construction land, and complementary area changes between forest, grassland, water, and unused land. Under the ecological development scenario (Fig. 9d), forest area expanded but was still small compared to the total study area, making forestry construction relatively severe.

In relation to land use shifts under multiple scenarios for 2020–2030 (Fig. 10), under inertia development, the total area transferred was 50,430 km2, mainly from cultivated land to construction land. Some forest and water areas were transferred out to building sites, resulting in a reduction of water and forest. Under the ecological development scenario, the total transferred area was 46,584.13 km2, of which 52.82% of the cultivated land was transferred out and 43.92% of the construction land was transferred in. Cultivated land was developed with a total transferred area of 37,682.56 km2, and 70.46% of cultivated land transferred in. Under the urban development scenario, the total transferred area was 56,352.06 km2, including 58.93% of cultivated land and 24.16% of construction land that was transferred out, and other land types that shifted to a small extent.

Fig. 10
figure 10

Multi-scenario land use shifts 2020–2030: a inertia; b cultivated land; c construction land; d ecology

Full size image

Simulated ESV variability at multiple scales

The ESV of the HRB in 2020 was 224,759 × 1012 CNY. ESV declined under all four scenarios, mainly due to construction land expansion (Fig. 11). ESVs under the inertial development, ecological development, cultivated land development, and urban development scenarios were 219,977 × 1012 CNY, 218,098 × 1012 CNY, 219,757 × 1012 CNY, and 213,985 × 1012 CNY, respectively. Compared to 2020, in 2030 the inertial development scenario ESV declined by 4,782 × 1012 CNY, the ecological development scenario declined by 6,661 × 1012 CNY, the cultivated land development scenario declined by 5,002 × 1012 CNY, and the urban development scenario declined by 10,774 × 1012 CNY. The highest ESV for water was found under the inertial development scenario, at 9.29 × 1016 CNY. The inertial development scenario ESV in 2030 saw a decrease of 687.94 × 1012 CNY compared to 2020 and the highest decrease of cultivated land, 3,271.17 × 1012 CNY, under different scenarios, compared to each category of ESV. Under the four scenarios, the highest ESV was 2.927 × 1013 CNY under the ecological development scenario, the highest ESV was 8.79 × 1016 CNY under the cultivated land scenario, and the largest decrease in ESV was 10.774 × 1012 CNY under the urban development scenario.

Fig. 11
figure 11

Multi-scale ESV in the HRB under different scenarios

Full size image

Changes in the spatial granularity of land use data under different scenarios and scales altered the number of ESV classes (Fig. 12). There was particle size variability when comparing the municipal, county, and grid scale ESV particle sizes. County and grid scale ESV particle size variability was similar, with low and medium-value areas dominating. At the municipal scale, the median value area was dominant under the inertial development, ecological development, and cultivated land development scenarios, and the low and lower value areas were dominant under the urban development scenario, with small fluctuations in the median, medium–high, and high ESV granularity.

Fig. 12
figure 12

Number of multi-scale ESV particle sizes under different scenarios in 2030: a municipal scale; b county scale; c grid scale

Full size image

Cold and hot spot analysis of ESV under multiple scenarios and scales

Based on the ESVs under four different development scenarios in 2030, the spatial distribution of ESV cold and hot spots at different scales was determined using the \({G}_{i}^{*}\) hot spot analysis tool (Fig. 13). Spatially, ESV cold and hot spots were similar at the same scale, with cold spots mainly distributed in the west and central part and hot spots mainly distributed in the east and northeast. There was greater variability at different scales. At the municipal scale, highly significant cold spots, significant cold spots, and cold spots were mainly distributed in the southwest direction. Significant hot spots and highly significant hot spots were mainly distributed in the southeast direction, and non-significant spots were distributed in the north and south directions. At the county scale, highly significant cold spots, significant cold spots, and cold spots were mainly clustered in the west. Hot spots, significant hot spots, and highly significant hot spots were mainly concentrated in the east and south. A small area of highly significant cold spots and hot spots was clustered in the center. At the grid scale, the zone was dominated by significant cold spots and by non-significant and highly significant hot spots. The highly significant cold spots were scattered in a grid distributed in the northwestern fringe, and the highly significant hot spots were mainly distributed in the southwest, south, and east, except for a strip in the center.

Fig. 13
figure 13

Distribution of ESV multi-scale cold and hot spots under different scenarios in 2030

Full size image

The distribution of the number of hot and cold spots at the municipal and county scales was relatively similar in rank. There were mainly non-significant cold and hot spots, whereas the grid scale was dominated by significant cold spots (Fig. 14). Cold and hot spots under the cultivated land development scenario were not consistent in their spatial clustering characteristics with those in the other three development scenarios. In fact, the non-significant data distribution for cultivated land in the former scenario was lower than that of the other three scenarios at the municipal scale, although there were significantly more hot spots. The number of significant and non-significant cold spots in both the cultivated land development scenarios was lower than in the other three development scenarios at the county scale. At the grid scale, the four development scenarios were mainly clustered around significant cold points, and the highly significant cold points were the least diffused. The cultivated land development scenario had a relatively small percentage of significant cold points and a relatively large percentage of non-significant points, compared to the other three development scenarios.

Fig. 14
figure 14

Number of ESV multi-scale granularity for different scenarios in 2030: a municipal scale; b county scale; c grid scale

Full size image

ESV sensitivity analysis

The sensitivity index values of the value coefficients of land use types in the Huaihe River basin for each year from 2000 to 2030 ranged from 0 to 0.5 (Table 6). In various years, several values of the ESVs sensitivity index CS were lower than 1 for different land types, with low variation inside a same year. This indicated that the ESV in the study area was somewhat inelastic relative to VC and relatively stable. Overall, the land use types in the study area were ranked from highest to lowest in terms of sensitivity index as follows: cultivated land > water > grassland > forest > unused land. The highest sensitivity index value among these land types was 0.4368 for cultivated land in 2000, indicating a 1% reduction in the overall ecological value coefficient, along with a 0.4368% reduction in the ESVs. This was followed by water and forest, due to the relatively large VC of ecosystem service coefficients of these two land types. After 2015, the lowest value of the sensitivity index of 0.0001 was observed for unused land; this was because the area of this category was small and the service value per unit area was low, and therefore it did not have a significant impact on the total value of the study area. Therefore, the selected coefficients VC of ecosystem service values had little effect on the total ESVs of ESs in the HRB, and the results of the study were fully credible.

Table 6 ESV sensitivity factors for the period 2000–2030
Full size table

Analysis of the contribution of land use conversion to ESV

Land use change had different effects on the ESV across different years (Fig. 15). In the period 2000–2010, the conversion of cultivated land to water had a significant positive impact on ESV, while from 2010 to 2020, the largest positive impact on ESV was observed from the conversion of water to cultivated land. The most significant positive impact on ESV from the transfer of cultivated land and water to each other was in the 2020–2030 inertial development scenario. The conversion of all land types to unused land had a small impact on ESV, and construction land had a large positive impact on ESV for both cultivated land and forest.

Fig. 15
figure 15

Contribution of land use change to ESV (2000–2030)

Full size image

Discussion

Impact of land use change on ESV

The HRB is an important grain and energy base in China, and the conflict between environmental protection and economic development has always been prominent. This study analyzed five periods of historical land use data from 2000 to 2020, simulated future land use development changes under four development scenarios to 2030, assessed the land use changes and ESV in the HRB, and analyzed the spatial distribution of ESV at different scales. The results of this study provide a basis for future land use structure optimization and sustainable development goals. In the past, coal and thermal power resources in the HRB have been used to meet the energy needs of the Chinese government, entailing a serious disturbance of land use and vegetation destruction. However, modern coal production in China has now shifted from the middle and eastern regions to the west (Wang et al. 2022), and the intensity of coal mining has weakened (Hao et al. 2019). With the rapid development of urbanization and industrialization, construction land continues to encroach on other land. ESV first increased, and then decreased from 2000 to 2020, mainly due to the huge increase in ESV from water and forests, especially coal mining subsidence that created large areas of sunken water (He et al. 2020; Yang et al. 2021), which offset the decrease in ESV due to the loss of cultivated land and other land types. The loss of ESV under multiple scenarios for 2020–2030 was mainly caused by the expansion of construction land (Maimaiti et al. 2022), especially under the cultivated land development scenario where, although the area of cultivated land, water, and construction land increased, this was not sufficient to offset the decrease in ESV due to the decrease in forest area.

Overall ESV trends were aggregated at different scales. ESVs were localized between the same years at the municipal-county-grid scale, and especially at the grid scale, where a grid of low-value areas occupied large areas. This was mainly due to the large area of cultivated land, although the ESV generated by cultivated land was low. This was consistent with the conclusion that cultivated land had a negative effect on ESV (Arowolo A O.et al. 2018). In particular, the fragmentation of cultivated land caused by large-scale and high-intensity cultivated land development (Yu et al. 2017) caused a decrease in soil fertility and the degradation of large areas of cultivated land, which reduced crop yields and decreased cultivated land ESV. In the cultivated land development scenario, the relative increase in cultivated land area under cultivated land policy protection led to a large reduction in forest, which reduced overall ESV. This was consistent with the results of Fenta et al. (2020), who found that cultivated land-forest conversion led to a decrease in forest ESV.

The overall spatial clustering of ESVs at different scales in a same year became increasingly focused from the municipal to the county and grid scale and top-down. Highly significant hotspots gradually expanded and radiated from the edges toward the interior, with a significant trend from the southeast toward the center. At smaller scales, the regional problem was more focused (Chen et al. 2022), which has important implications for regional conservation and restoration. Under the 2030 cultivated land development scenario at the county scale, there was lower aggregation of highly significant cold spots, while at the grid scale, there was more aggregation of highly significant cold spots. The impact of human activities on ESV was mainly reflected at the county and grid levels (Zhang et al. 2020); therefore, the systematic explanation of spatial heterogeneity at the grid and regional scales needs to be further explored (Pan et al. 2020). By introducing the degree of contribution of land use conversion to ESV, the impact of the conversion among categories on ESV was further explored, with the conversion of cultivated land to water contributing the most to ESV in 2000–2010, 2000–2020, and 2020–2030 under the ecological development scenarios and in 2020–2030 under the urban development scenario. The highest contribution of the transfer of water to cultivated land was in 2010–2020. In the inertial development scenario for 2020–2030, the highest impact on ESV was observed from the conversion of cultivated land to water and vice versa, while in the 2020–2030 cultivated land development scenario, the highest impact was from the conversion of cultivated land to water and the second highest from the conversion of water to cultivated land. This was related to the size of the area and of the unit ESV values transferred from cultivated land and water; moreover, the conversion between cultivated land and water indicated increasing disturbance of land use by human activities (Li et al. 2022). Apart from that, the change in unused land had the least impact on ESV; this depended on the size of the transferred area, the low base area of unused land in the study area, and the relatively low area between the other land types transferred.

Limitations and uncertainties

In this study, ESV was estimated due to the lack of statistical data in some county statistical yearbooks. Only 47 municipalities in the study area were considered without county data; moreover, ESV provided by urban greening and ecological parks was ignored, and the ESV generated by construction land was not evaluated. Limited by the land use classification data, this study only macro-classified land use types into six categories; this needs to be further developed in the future using more refined classification data and needs to be verified by actual survey data. In addition, while the PLUS model is a data-driven forecasting model that makes predictions based on historical data, the actual situation may be influenced by other factors. Lastly, further research on the internal drivers of ESs and other service types within cities is needed in the future.

Management and implications for land use and ecosystems in the HRB

Based on the land use-driven change assessment of ESV, the ecological impacts of different land use types were identified. These provide support to decision-makers to optimize land use structure and achieve the sustainable development goals. The rapid urbanization in the southern region of the study area accelerated the transfer of forest to other land types, leading to a continuous decline in forest ESV. Because the southern region is an ecologically protected area, it is necessary to reasonably delineate the urban development boundary (Cao et al. 2021), limit the disorderly expansion of urban open space, actively build ecological corridors, stabilize the ecosystem structure (Andersson et al. 2019), and strictly observe the ecological protection red line. Although the ESVs for large open water and collapsed water areas formed by coal mining subsidence were the highest in the region, further water expansion prevention and control measures are still needed because natural land use transfer is irreversible. In addition, farmland ecosystems represented the largest area encroached upon during urbanization; therefore, the red line of 1.8 billion acres of cultivated land protection needs to be strictly observed (Zhou et al. 2021). Development should accommodate increasing ESV under urbanization and socio-economic development goals. By setting up four scenarios for land use simulation and analysis, we optimized land use structure, improved land utilization, and provided important references for future ecological construction and land use expansion in the basin. The selection of corresponding scenarios in combination with the development strategy and actual development needs of the HRB is of great significance to achieve regional ecological-social-economic harmony and stability.

Conclusions

  1. (1)

    From 2000 to 2020, land use patterns in the HRB were mainly cultivated land, construction land, and forest. Of these, cultivated land, grassland, and unused land continued to decline, while construction land and forest increased, and water first increased and then decreased. Under the multiple scenarios for 2020–2030, except for the cultivated land development scenario, in the other three scenarios, the area of cultivated land significantly decreased, the area of construction land increased, and there was low variability among other land types.

  2. (2)

    From 2000 to 2020, ESV in the HRB first increased and then decreased. From 2020 to 2030, the four scenarios showed ESV declines, with the lowest ESV under the urban development scenario. ESV at different study scales was spatially variable, with high-value areas concentrated in the eastern and central parts of the region and low-value areas concentrated in the western and northern parts of the region. As the study scale declined, the granularity of low-value areas increased, while that of high-value areas decreased.

  3. (3)

    From the multi-scenario ESV hotspot analysis for 2000–2030, highly significant hot spots and significant hot spots were mainly distributed in the south, eastern fringe regions, and some central regions. Highly significant cold spots and significant cold spots were mainly distributed in the west, northwest fringe regions, and some central regions. The municipal and county scales were mainly dominated by non-significant types, and the grid was dominated by significant cold spots.

  4. (4)

    Interconversion between cultivated land and water contributed the most to land use ESV, while the conversion of unused land contributed the least to ESV.