Spatiotemporal evolution and influencing factors of carbon stock in the water receiving areas from the perspective of carbon neutrality

Abstract

Water resources of water transfer projects are not only used to solve the water scarcity problem in the water-receiving area but also to change the regional carbon absorption capacity. Using the water-receiving area of the Jiangsu-Shandong section of the East Route of the South-to-North Water Diversion Project (ER-SNWDP) of China as a case study, this study explored the dynamic variation in carbon stocks in response to water diversion project in the context of carbon neutrality. The results showed that (1) After the ER-SNWDP came into operation, there was a trend of growth in water area. Based on multi-scenario simulation, under the ER-SNWDP scenario, built-up land expansion would be curbed, forest and grassland reductions would be alleviated, and water areas would increase significantly compared to the natural variation scenario. (2) Due to the implementation of the project, the research area had better carbon sequestration capacity. Under the natural variation scenario from 2015 to 2025, the carbon stock would decrease by 1228.35 × 10⁴ t. However, under the ER-SNWDP scenario, there would be an increase of 262.84 × 10⁴ t. In addition, the water resource allocation of ER-SNWDP may affect the spatial distribution of carbon stocks. In the northeast region, particularly in the Binzhou and Dongying areas with large water transfer volumes, the increase in carbon stocks was significant, and the center of gravity of increase also tended to tilt these areas. (3) Land use had the highest explanatory power and driving force for spatial variation in carbon stocks. According to the results of the interaction factor analysis, the strongest interaction factor after 2005 was “land use ∩ nighttime lights”, indicating that the interaction between socio-economic factors and land use factors gradually amplified the impact on the spatial variation of carbon stocks. This study provides a scientific basis for future land use planning, promotes the rational and optimal allocation of water and carbon resources, and provides a prospective reference for water resources to cope with climate change and achieve carbon neutrality.

Introduction

With the increasing interference of climate change and human activities on natural systems, many countries have proposed the goal of achieving carbon neutrality, which has become a global consensus^1,2. Water resources are the material basis for the survival and continuation of human beings and are key for socioeconomic development and ecological environmental protection. Water resources are the key to implementing the carbon neutrality goal, regulating the relationship between people and nature, and supporting sustainable development. The water resources of water transfer projects are not only used to solve the problem of water scarcity in water-receiving areas but also to change the regional ecosystem function and carbon cycle model³, which play an important role in improving the carbon absorption capacity and resilience of the regional ecosystem⁴. Owing to the different water allocation mechanisms and strategies along the water transfer route, the carbon sink forms of water resources brought about by the project are different on spatial-temporal scales⁵, which makes the problems of water resource allocation and carbon neutrality more prominent.

Ecosystem services and land use and land cover (LULC) are contradictory units with mutual restriction and influence. Land use and cover changes (LUCC) alter the function and value of ecosystem services, which in turn affects the purpose and mode of LULC⁶. LUCC is the process of acquiring material, ability, and value from land⁷ and combines different land use needs and types. The system dynamics model, Markov model, cellular automata (CA), gray linear programming, multi-agent, fuzzy multi-objective optimization models^8,9, and other methods are often used to predict LUCC trends. In addition, multiple models are often coupled¹⁰. For example, the CLUE-S and system dynamics models were coupled to build a spatial-temporal pattern simulation model of LULC and dynamic simulation was carried out in different modes, such as reference, ecological, and economic mode¹¹. The CA-Markov model, which has been widely used in recent years, combines the spatial simulation ability of the CA model and the quantity change prediction ability of the Markov model¹². This model improved the prediction accuracy of land-use conversion and effectively simulated the spatial change in land-use patterns^13,14.

LUCC is a key factor affecting terrestrial carbon stocks and their dynamics¹⁵. The mutual transfer of various land use types will change the carbon density, thus affecting the total carbon stock of an area¹⁶. In recent years, under the influence of frequent human activities, the increase in built-up land has become a major factor leading to the decrease in carbon stocks^17,18. Evaluation of the spatial and temporal evolution characteristics of carbon stocks based on LUCC has gradually become a widely used method. This method can map the spatial distribution and dynamic changes in carbon stocks and reflect the relationship between LUCC and carbon stocks¹⁹. Sample-site inventories, model simulations, and other methods are often used to study carbon stocks in various ecosystems^20–22. The InVEST model is relatively mature and is extensively utilized among many methods that use model simulation²³. Water and carbon cycles are coupled through vegetation²⁴ and terrestrial ecosystems are largely restricted by water resources^25,26. At present, there have been many studies on the carbon stocks of terrestrial ecosystems, but few on the carbon stocks of terrestrial ecosystems under the influence of water resources. In addition, the impact of inter-basin water resources on the carbon stocks of terrestrial ecosystems remains unclear.

The Jiangsu-Shandong section of the East Route of the South-to-North Water Diversion Project (ER-SNWDP) in China contains many important energy and chemical production bases and agricultural production bases along its route. Since the project began operation in December 2013, many benefits have been realized to maintain and enhance human livelihoods and quality of life. Large- and medium-sized cities in the Jiangsu-Shandong section of ER-SNWDP generally have dense populations, high water consumption, and high carbon emissions. These cities are transitioning from rapid economic growth to high-quality development. There is an urgent need for the development of “green GDP” and the promotion of “carbon reduction by water” to analyze the spatio-temporal changes of water resources to LUCC and carbon stocks in water-receiving areas.

Thus, our study combines carbon neutrality with water resource management. Based on the LUCC of the ER-SNWDP from 2005 to 2020, the CA-Markov model was applied to predict the land use and cover spatial patterns of 2025 under the natural variation scenario (S1) and ER-SNWDP scenario (S2). Based on the corrected carbon density values of the regional temperature and precipitation, the InVEST model carbon module was used to simulate carbon stocks in the study area and predict carbon stocks under the two scenarios. The Geo-detector was utilized to assess the influence of different driving factors, identifying the primary elements that impact variations in carbon stocks within the study area. The results of this study can accelerate the formation of ecological functional zones for water resources, promote rational and optimal allocation of water and carbon resources, and provide a scientific reference for achieving carbon neutrality and high-quality regional development.

Materials and methods

Study area

The research object was the water-receiving area of the Jiangsu-Shandong section in the ER-SNWDP of China, which covers 17 prefecture-level cities, 18 county-level cities, and 43 counties, as shown in Fig. 1. From the perspective of emission reduction (carbon sources), the water-receiving area belongs to “energy basin,” and the cities along the route are densely populated with many large and medium-sized cities, which has a very important strategic position in the economic and social development of China²⁷. Considering carbon sequestration (carbon sink), the water-receiving area of the ER-SNWDP has numerous vital water conservation areas and key ecological functional zones with important ecological functions, such as water conservation, soil and water conservation, and maintenance of biodiversity. Wetlands, gardens, cultivated land, and forests are rich and important carbon sinks and storage areas. The ER-SNWDP has been in operation since 2013 and its allocation mechanism and management have been relatively mature. The water resources of the ER-SNWDP play an important role in enhancing the carbon absorption capacity and ecosystem resilience of the water-receiving area.

Fig. 1.

Location of the study area: (a) Jiangsu-Shandong section of the ER-SNWDP in China; (b) cities along the route (Software: ArcMap 10.8, https://www.esri.com).

Data sources

The Data sources are as follows: (1) LULC data were provided by the Resource and Environment Science and Data Center of the Chinese Academy of Science (http://www.resdc.cn), with a resolution of 30 m. (2) Digital Elevation Model (DEM) data was derived from ASTER GDEM, with a resolution of 30 m. (3) Road data was provided by Open Street Map (http://www.openstreetmap.org). (4) The data of water diversion quantity was from the Shandong Water Resources Bulletin of China (http://wr.shandong.gov.cn/zwgk_319/fdzdgknr/tjsj/szygb/) and Jiangsu Water Resources Bulletin of China (https://jswater.jiangsu.gov.cn/col/col84437/index.html). (5) Carbon density data was from field investigations (Fig. 2) and “A dataset of carbon density in Chinese terrestrial ecosystems (2010s)” (www.csdata.org), and combined with relevant experimental test data of the research group. The sampling points were located in the Jiangsu section of ER-SNWDP. The sampling points of vegetation carbon density were arranged according to the principle of representativeness of LULC, soil carbon density sample points were arranged according to the principle of evenly covering each land use type. (6) Soil type data was obtained from the 1:1 million Soil Database of China. (7) Precipitation and temperature grid data were sourced from the National Earth System Science Data Center, with a spatial resolution of 1 km. (8) Population density data was obtained from WorldPop, with a spatial resolution of 1 km. (9) Nighttime light data was sourced from the National Data Center for the Qinghai-Tibet Plateau²⁸, with a spatial resolution of 1 km. (10) Socioeconomic data was obtained from the Statistical Yearbooks of Jiangsu and Shandong provinces.

Fig. 2.

Distribution of Sampling Points in Jiangsu Section (Software: ArcMap 10.8, https://www.esri.com).

Methods

Prediction of LUCC

By employing DEM, slope, and transportation factors, which have a significant influence on LUCC, a suitability atlas was created. The official inauguration of the ER-SNWDP on November 15, 2013, the closest year 2015 was designated as the demarcation point. Two distinct scenarios were formulated to simulate LUCC.

(1) Natural variation scenario (S1): LULC data for 2005 and 2010 within the study area were selected as the base and terminal years, respectively, and no limiting factors were set. The transfer area and probability matrix were calculated using a Markov chain and LULC under S1 in 2025 was predicted at 10-year intervals.

(2) ER-SNWDP scenario (S2): Using the land use data for the study area in 2015 and 2020 as the base and end-year, the MCE (Multi-criteria Evaluation) module in IDRISI software was employed to create a suitability atlas for land use. Factors such as DEM elevation, slope, and distance to roads were selected as influencing factors for the suitability atlas. Strict control was imposed on the conversion of water bodies, which cannot be converted into other land classes, while other land classes can be converted into water bodies. In other words, the land use type conversion was limited. With a 5-year interval, the land use under ER-SNWDP in the study area was predicted for the year 2025.

The Markov model was employed to compute the land use transfer matrix and transfer probability for the period 2005–2010, while the CA-Markov model was utilized to simulate LULC in 2015. The confusion matrix provides detailed classification accuracy for forecast map (Table 1). By comparing the forecast with the actual map in 2015, the Kappa value was calculated at 0.8789, which passed the accuracy test¹⁵.

Table 1.

Detailed accuracy assessment of simulated LULC classes.

LULC	Cropland	Forest	Grassland	Water	Built-up land	Bare land	Total	User accuracy
Cropland	308	1	3	1	4	0	317	97.16%
Forest	0	17	2	0	0	0	19	89.47%
Grassland	0	2	11	1	0	0	14	78.57%
Water	0	0	2	38	5	0	45	84.44%
Built-up land	10	1	0	0	92	1	104	88.46%
Bare land	0	0	0	0	0	1	1	100.00%
Total	318	21	18	40	101	2	500
Producer accuracy	96.86%	80.95%	61.11%	95.00%	91.09%	50.00%
	Overall accuracy	93.40%		Kappa coefficient	0.8789

Estimation of carbon stocks in different land covers

Carbon stocks in the terrestrial LULC were estimated using the Integrated Valuation of Environmental Services and Tradeoffs (InVEST) model²⁹. The model necessitates estimating the carbon content in four primary carbon pools for each land-cover type: aboveground biomass (leaves, bark, branches, tree trunks, and other living vegetation above the Earth’s surface), belowground biomass (organic carbon in subsurface living roots), soil organic matter (organic carbon in soil), and dead organic biomass (organic carbon contained in dead vegetation and litter). The calculation formula for the model is as follows:

1

2

where is the carbon stock of each LULC type, is the amount of carbon stored in a given cell, represents the total area covered by each land type, and n signifies the total number of land use types, and , , , and are the carbon density of the aboveground biomass, belowground biomass, soil organic matter, and dead organic biomass of LULC type  , respectively.

Carbon density and its correction

In addition to field investigations and reference to “A dataset of carbon density in Chinese terrestrial ecosystems (2010s)” (www.csdata.org), we consulted the carbon density data of relevant literature in the study area and neighboring areas^30–32 and determined the soil and aboveground carbon density data of Jiangsu Province. A root-to-shoot ratio of 0.2 was applied to calculate vegetation carbon density³³. The dead organic biomass carbon density for each LULC type was set to zero because of its negligible proportion. Subsequently, the data were adjusted to match the actual carbon density of the study area using a carbon density correction formula^34,35 (Table 1), with the calculation formula as follows:

3

4

5

 

Where C_SP is the soil carbon density (kg∙m⁻²) based on the average annual precipitation, C_BP and C_BT represent biological carbon density (kg∙m⁻²) obtained based on annual precipitation and annual average temperature, respectively, and P and T are the average annual precipitation (mm) and temperature (℃), respectively.

The mean average temperature and precipitation of the study area and Jiangsu Province were substituted into the above equations. From 2005 to 2020, the mean average temperature in Jiangsu Province and the study area was 15.8 and 14.6℃ and annual average precipitation was 1066.2 and 848.5 mm, respectively. The revised carbon density data for each land-use type in the study area are listed in Table 2.

Table 2.

Revised carbon density of each LULC type. (t/hm²).

LULC	C_above	C_below	C_soil
Cropland	5.14	0.40	85.29
Forest	17.81	1.38	116.87
Grassland	1.93	0.15	91.53
Water	0.55	0.04	74.46
Built-up land	0.09	0.01	68.49
Bare land	0.09	0.01	10.74

Geo-detector

The continuous variables among the driving factors are categorized into five classes using the natural breakpoint method. The Geo-detector’s factor detection is employed to reveal the primary driving factors of the spatiotemporal evolution of carbon stock in the water-receiving area of the Jiangsu-Shandong section of the ER-SNWDP. The calculation formula is as follows:

6

where i = 1, 2., n; L represents the stratification or partition of variable Y or factor X, N and N_i are the sample sizes of the entire region and stratum i, ² and ² are the variances of the Y values in stratum i and the entire region, respectively. q is the detection power of the detection factor, with a range of [0,1]. A higher value indicates a stronger explanatory power of factor X on variable Y.

Research framework

Based on the features of land use type structure, land use dynamics, and land type transfer in the study area under a 15-year-long time series, the CA-Markov model was applied to explore the future characteristics of LULC. The carbon density-modified carbon module of the InVEST model was used to calculate the carbon stock of the study area, revealing the spatiotemporal distribution and evolution patterns of the carbon stock in the context of the water transfer project (Fig. 3).

Fig. 3.

Research framework.

Results

Analysis of water-receiving area’s LULC dynamics

Figure 4 shows that the main LULC type in the study area was croplands (approximately 65%). Built-up land and water accounted for 18% and 7% of the total area, respectively. In contrast, forests and grasslands comprised a relatively modest share, each constituting approximately 4%. The extent of bare land was the lowest, accounting for less than 1% of the total area. The share of croplands, the dominant land-use type in the study area, underwent relatively little change over the study period. However, the built-up land area fluctuated, with a rapid expansion from 2005 to 2010 of 5,779.39 km², followed by a reduction from 2015 to 2020 of 983.40 km². The water body area remained relatively stable from 2010 to 2015 but began to grow after 2015, with an increase of 14.48%. Changes in the built-up land area and the growth of water bodies may have been closely related to the official operation of the ER-SNWDP in November 2013. The areas of forest and grassland decreased significantly, with forestland decreasing by 945.37 and 20.11 km² and grassland decreasing by 3,705.03 and 73.11 km² from 2005 to 2010 and 2015 to 2020, respectively, with the rates of decrease notably decreasing during the latter period.

Fig. 4.

Changes of area in different land use types in the study area from 2005 to 2020.

As shown in Table 3, from 2005 to 2010, the absolute magnitudes of the dynamic degree across by land-use type in descending order were bare land, grassland, built-up land, water, forest, and cropland. The positive values for water and built-up land indicate that the area increased rapidly during this period. Conversely, forest, grassland, and bare land displayed a negative dynamic degree, indicating swift contractions in the extent of these land-use types. From 2010 to 2015, mutual transformations among different land use types were relatively stable. The dynamic degree for each land-use type was significantly diminished compared to the preceding period, resulting in a comprehensive dynamic degree of 0.21%. From 2015 to 2020, only the water area demonstrated a positive dynamic degree, indicating that the water area showed a continuous growth trend after the ER-SNWDP began operation, whereas built-up land, which represents the need for urban expansion, showed a decreasing trend.

Table 3.

Dynamic degree of land use in different periods.

Year	Land use type						Comprehensive dynamic degree
	Cropland	Forest	Grassland	Water	Built-up land	Bare land
2005–2010	−0.36%	−2.04%	−6.58%	3.75%	3.95%	−10.94%	0.82%
2010–2015	−0.21%	−0.03%	0.01%	0.01%	0.73%	−0.21%	0.14%
2015–2020	−0.11%	−0.05%	−0.19%	2.90%	−0.54%	−4.12%	0.21%
2005–2015	−0.28%	−1.03%	−3.29%	1.88%	2.41%	−5.52%	0.48%

LUCC characteristics under different scenarios

Based on land-use data from 2015 to 2020, the CA-Markov model was employed to predict the spatial distribution of land use for the Jiangsu-Shandong section in 2025 under S1 and S2 (Fig. 5). Owing to the differences in the scenarios, significant variations were observed in the land-use layouts (Table 4). Forest and grassland areas declined rapidly under S1, decreasing in area by 21.34% and 41.1%, respectively. In contrast, under S2, the forest area increased by 1.67% and grassland area by 1.32%. The built-up land area showed an inverse trend to that of forests and grasslands. Under S1, it increased by 6.19%; however, under S2, a decreasing trend was observed, indicating that S2 effectively curbed the disorderly expansion of built-up land. Under both scenarios, the water body area increased; however, under S2, it increased by 15.96%, which was 2.5 times the increase under S1. Under S2, the increased water area amounts to 2251.04 km², which is concentrated primarily in the Shandong Province, nearer to the Bohai Sea, while in S1, this same region corresponds to built-up land.

Fig. 5.

Land use map of the study area in 2025 under the (a) natural variation and (b) ER-SNWDP scenarios (Software: ArcMap 10.8, https://www.esri.com).

Table 4.

Land use in 2015 and in 2025 under different scenarios (km²).

LULC	2015	2025 S1			2025 S2
	Area	Area	Area change	Rate of change	Area	Area change	Rate of change
Cropland	125576.88	127395.61	1818.72	1.45%	123955.03	−1621.85	−1.29%
Forest	8318.37	6542.85	−1775.53	−21.34%	8457.36	138.99	1.67%
Grassland	7548.07	4452.61	−3095.46	−41.01%	7647.99	99.92	1.32%
Water	14099.96	15030.96	931.01	6.60%	16350.99	2251.04	15.96%
Built-up land	36328.33	38578.05	2249.72	6.19%	35566.28	−762.04	−2.10%
Bare land	731.47	651.55	−79.92	−10.93%	607.91	−123.55	−16.89%

Analyzing the direction of the transitions (Fig. 6) reveals that changes in croplands were relatively minor under S1. Under S1, a diminished portion predominantly shifted towards built-up land, with a partial influx into grassland. Decreased forestland was primarily converted into cropland, with smaller portions transitioning to grassland, water bodies, and built-up land. Some water area was converted into cropland but rarely into other types of land; however, cropland, grassland, and built-up land were converted into water area, leading to the increasing trend. Notably, built-up land was largely converted into cropland and water bodies while cropland was converted into built-up land, leading to significant augmentation. Under S2, cropland was mainly transformed into built-up land, forestland into grassland, and built-up land into water bodies, with the greatest increase observed in the water area. It can be seen from the data in Table 3 that the scale of built-up land in the study area was limited and that the water area increased.

Fig. 6.

Transfer matrix of land use types from 2015 to 2025 in the study area under the (a) natural variation and (b) ER-SNWDP scenarios.

Analysis of spatial-temporal evolution characteristics of carbon stock

The InVEST model carbon module was employed to simulate the carbon stock of the study area in 2005, 2010, 2015, and 2020 and to predict the carbon stock in 2025 under S1 and S2. As shown in Table 5, the carbon stocks of the study area in 2005, 2010, 2015, and 2020 were 169677.99 × 10⁴, 168350.76 × 10⁴, 168024.12 × 10⁴, and 168154.83 × 10⁴ t, respectively, showing a fluctuating trend of decreasing followed by increasing. The carbon stock decreased by 1327.24 × 10⁴ t (0.78%) and 326.64 × 10⁴ t (0.19%) from 2005 to 2010 and 2010–2015, respectively, stabilizing gradually. From 2015 to 2020, the carbon stock increased by 130.71 × 10⁴ t (0.08%). The predicted carbon stock for the study area in 2025 was 166795.76 × 10⁴ t under S1, which was 1228.35 × 10⁴ t less than that in 2015. However, the predicted carbon stock for 2025 was 168286.96 × 10⁴ t under S2, an increase of 262.84 × 10⁴ t (0.16%) compared to that in 2015. Thus, under S2, the carbon stock reduction was mitigated and began to increase after 2015. This indicates that the operation of the ER-SNWDP led to effective carbon sequestration in the study area.

Table 5.

Carbon stock of different land use types in the study area from 2005 to 2025. (×10⁴t).

LULC	Year
	2005	2010	2015	2020	2025 S1	2025 S2
Cropland	117358.15	115247.95	114061.48	113444.78	115913.43	112888.35
Forest	12619.37	11333.10	11317.98	11290.61	8902.20	11507.09
Grassland	10534.54	7066.26	7065.75	6997.32	4168.09	7159.28
Water	8908.62	10578.85	10582.02	12116.07	11280.74	12271.42
Built-up land	20080.37	24044.45	24917.60	24243.08	26460.68	24394.91
Bare land	176.95	80.14	79.29	62.97	70.63	65.90
Total	169677.99	168350.76	168024.12	168154.83	166795.76	168286.96

The carbon stocks in the study area showed a relatively consistent spatial distribution pattern and change trend from 2005 to 2015 (Figs. 7 and 8). Specifically, the central region at the junction of Jiangsu and Shandong provinces, as well as the hilly areas in Shandong, had higher carbon stocks, characterized by well-maintained vegetation and strong carbon sequestration capability. Low-value areas of carbon stock were primarily concentrated in the Jiangsu Lake regions, coastal areas, and economic zones near the Bohai Sea in Shandong. This spatial distribution pattern of carbon stocks was closely correlated with the distribution of vegetation and land use types in the study area. High-value areas were predominantly composed of forestland and grassland, whereas low-value areas encompassed water bodies, built-up land, and other land-use types with lower carbon densities. The reduction in carbon stock during 2005–2015 was mainly due to the decline in forestland and grassland as well as the extensive expansion of built-up land.

Fig. 7.

Spatial distribution of carbon stock in the study area from 2005–2025 (Software: ArcMap 10.8, https://www.esri.com).

Fig. 8.

Carbon stock variation under different scenarios from 2015–2025 (Software: ArcMap 10.8, https://www.esri.com).

The spatial distribution of carbon stocks from 2015 to 2025 was largely consistent with that from 2005 to 2015 under S1, demonstrating a decreasing trend. Decreasing carbon stocks were observed in Yantai and Weihai in Shandong and in the central area. This is primarily because under this scenario, the reduction in forestland and grassland was not alleviated; instead, they continued to transform into built-up land, leading to a decrease in carbon stock. Conversely, under S2, the increase in carbon stocks was mainly concentrated in cities such as Binzhou and Dongying in Shandong, as well as in coastal cities such as Lianyungang in Jiangsu. In the central region, the reduction in carbon stock was noticeably alleviated. Under S1, the centers of gravity of increase and decrease shifted 13 km southeast and 40.7 km to the northwest, respectively. Under S2, the centers of gravity of increase and decrease shifted 6.7 km to the northeast and 8.6 km to the southwest, respectively (Fig. 9). The main reason for the shift of the center of gravity of reduction to the west is that the rapid development of urbanization in Jinan, Liaocheng and Jining (Shandong Water Resources Bulletin of China, http://wr.shandong.gov.cn/zwgk_319/fdzdgknr/tjsj/szygb/) has led to a substantial increase in the area of construction land, so the carbon stocks in this region has decreased more. Under the S2 scenario, the center of gravity of increase shifted towards the northeast. Binzhou and Dongying in the northeast region had a water transfer volume of 1.364 and 1.143 billion m³ respectively, in 2020. The higher water transfer volume in these areas may have been the main factor contributing to the rapid growth of carbon stocks in the region.

Fig. 9.

Spatial change of carbon stock (a) Track map of the center of gravity of the reduction of carbon stock from 2005 to 2025 under different scenarios. (b) Track map of the center of gravity of the increase of carbon stock from 2005 to 2025 under different scenarios (Software: ArcMap 10.8, https://www.esri.com).

Effect of water transfer project on regional carbon stock distribution

To explore the impact of the water transfer project on carbon stocks in the study area, carbon stock variations were calculated for 2015–2025 under S1 and S2 (Tables 6 and 7). Under S2, in 2025, the carbon stock was 262.84 × 10⁴ t greater than that in 2015. Conversely, under S1, in 2025, the carbon stock was 1228.35 × 10⁴ t lower than that in 2015. This disparity stems primarily from the different transition probabilities of land-use types under the two scenarios. In S2, the conversion of forestland and grassland to built-up land was significantly reduced, while the transfer area of other land types to forest and grassland was greatly increased. Therefore, carbon stocks show an increasing trend under S2. From 2015 to 2025, the conversion of forestland into cropland, grassland, water bodies, built-up land, and bare land in the study area led to net carbon stock emissions of 588.78 × 10⁴, 21.00 × 10⁴, 27.38 × 10⁴, 267.64 × 10⁴, and 24.79 × 10⁴ t under S1, respectively. The emissions of carbon stock outweighed its sequestration, resulting in an evident decrease in the overall carbon sequestration capacity within the study area under S1. Conversely, under S2, the total sequestration of carbon stock resulting from the conversion of these land types into forestland in the study area was 176.35 × 10⁴, 47.22 × 10⁴, 117.27 × 10⁴, 10.16 × 10⁴, and 28.73 × 10⁴ t for 2015–2025, respectively. In this scenario, the sequestration of carbon stock surpassed emissions, leading to a distinct increase in the overall carbon sequestration capacity within the study area under the influence of the ER-SNWDP.

Table 6.

Carbon stock transition matrix for the study area under the natural variation scenario from 2015 to 2025. (×10⁴t).

2015	Natural variation scenario for 2025
	Cropland	Forest	Grassland	Water	Built-up land	Bare land
Cropland	-	0.58	0.12	−6.89	−860.99	−1.67
Forest	−588.78	-	−21.00	−27.38	−267.64	−24.79
Grassland	−65.33	8.17	-	−46.96	−96.76	−125.63
Water	51.63	1.27	0.91	-	−1.51	−0.03
Built-up land	356.53	5.12	1.42	56.86	-	−1.29
Bare land	126.91	1.90	0.66	17.21	41.84	-

Table 7.

Carbon stock transition matrix for the study area under the ER–SNWDP scenario from 2015 to 2025. (×10⁴t).

2015	ER-SNWDP scenario for 2025
	Cropland	Forest	Grassland	Water	Built-up land	Bare land
Cropland	-	176.35	21.66	−301.11	−1,369.65	−36.28
Forest	−9.30	-	−19.77	−0.36	−12.55	−0.01
Grassland	−1.42	47.22	-	−9.87	−2.44	−0.02
Water	139.28	117.27	62.77	-	−14.43	−25.44
Built-up land	1,644.69	10.16	14.00	104.30	-	−13.16
Bare land	39.99	28.73	29.44	34.47	2.91	-

Analysis of carbon stock driving factors

We selected nine driving factors, including elevation, slope, soil type, NDVI, annual average precipitation, annual average temperature, nighttime light, population density, and land use, to analyze the spatiotemporal evolution of carbon stock in the study area using the factor detection module in the Geo-detector for driving factor analysis. The main driving factor results for carbon stock in the study area from 2005 to 2020 are shown in Tables 8 and 9. According to the analysis of single-factor results, the ranking of the main driving factors remained relatively stable during the years 2005–2020. The q-value of the land use factor remained between 0.5 and 0.6, with an explanatory power exceeding 50% in each year, indicating that land use was the dominant factor influencing carbon stock changes in the study area. Elevation was the second most influential factor, with an explanatory power exceeding 25% in each year, demonstrating that elevation was an important factor affecting carbon stock changes in the study area. In addition, slope and soil type factors maintained an explanatory power of approximately 20% in each year, indicating their significant impact on carbon stock changes in the study area. Furthermore, climate factors showed an explanatory power exceeding 10% in most years, suggesting a certain degree of influence on carbon stock changes in the study area.

Table 8.

Results of the driving factors of carbon stock in the study area from 2005 to 2010.

2005				2010
Factor	q	Interaction Factor	q	Factor	q	Interaction Factor	q
Land Use	0.511	Land Use ∩ Elevation	0.559	Land Use	0.56	Land Use ∩ Nighttime Light	0.612
Elevation	0.262	Land Use ∩ Soil Type	0.557	Elevation	0.279	Land Use ∩ Elevation	0.599
Slope	0.202	Land Use ∩ Annual Temperature	0.548	Slope	0.208	Land Use ∩ Soil Type	0.597
Soil Type	0.188	Land Use ∩ Slope	0.543	Soil Type	0.198	Land Use ∩ NDVI	0.595
Annual Temperature	0.128	Land Use ∩ Annual Precipitation	0.539	Annual Temperature	0.114	Land Use ∩Annual Temperature	0.590

Table 9.

Results of the driving factors of carbon stock in the study area from 2015 to 2020.

2015				2020
Factor	q	Interaction Factor	q	Factor	q	Interaction Factor	q
Land Use	0.571	Land Use ∩ Nighttime Light	0.635	Land Use	0.566	Land Use ∩ Nighttime Light	0.627
Elevation	0.274	Land Use ∩ Soil Type	0.608	Elevation	0.272	Land Use ∩ Population Density	0.609
Slope	0.206	Land Use ∩ Elevation	0.605	Slope	0.199	Land Use ∩ Elevation	0.606
Soil Type	0.198	Land Use ∩ NDVI	0.601	Soil Type	0.18	Land Use ∩ Soil Type	0.599
Annual Temperature	0.136	Land Use ∩Annual Temperature	0.599	Population Density	0.127	Land Use ∩Annual Temperature	0.599

According to the analysis of interaction factor results, there were significant differences in the ranking of main interaction factors between 2005 and 2020. In 2005, “land use ∩ elevation” was the interaction factor with the strongest explanatory power. However, after 2005, the interaction factor with the strongest explanatory power became “land use ∩ nighttime light,” indicating an increasingly apparent interaction between land use and socio-economic factors. Among the main interaction factor results, all significant interaction factors involved the land use factor. Moreover, the explanatory power of land use factors interacting with other factors was shown as a dual-factor enhancement, where the explanatory power of the interaction-driven factors was greater than that of each single factor. This suggests that the interaction between land use factors and other factors had a significant impact on the spatiotemporal evolution of carbon stock in the study area.

Discussion

Impact of water diversion project on LUCC in water-receiving area

Spatiotemporal analysis of land use in the study area from 2005 to 2020 and the CA-Markov model simulation of land use in 2025 under different scenarios revealed that the water diversion project changed the land use cover of the water-receiving area. This supports the results of previous studies. Liu et al.³⁶ studied the Danjiangkou Reservoir area, the core water source of the middle route of the South-to-North Water Diversion Project, and found that from 2000 to 2018, the construction land and water areas increased by 1.31% and 1.39%, respectively, whereas the forest area decreased. After the ER-SNWDP began operation in 2013, the land use in 2015–2020 deviated from the previous change trends, with the water area changing from stable to growing, forest land and grassland area changing from significant decrease to stable, and construction land area restricted to disorderly expansion. This resulted in significant improvements in the ecological condition of the water-receiving area.

LUCC was mainly concentrated in coastal areas such as Shandong near the Bohai Sea, Lianyungang, and Yancheng in Jiangsu Province. Due to their geographical advantages and the need for economic development and urban expansion³⁷, these cities were mainly transformed into construction land before 2015 and then into water areas following ER-SNWDP operation. By simulating LUCC under different future scenarios, with the expansion of the time scale, the difference in the change in cover of the receiving area before and after the water diversion project became more prominent.

Analysis of carbon stock response to land use change

According to the analysis of carbon stock driving factors, it indicated that land use change was the dominant factor influencing carbon stock variation in the study area. The InVEST model was used to obtain the carbon stock distribution under two simulated scenarios from 2005 to 2020 and 2025 and the relationship between land use and carbon stock in the water-receiving area of the ER-SNWDP was analyzed. The results showed that LUCC affected the carbon stocks after the water diversion project (Fig. 8). The changes in carbon stocks from 2005 to 2015, 2015–2025 under S1, and 2015–2025 under S2 were analyzed. The results showed that carbon stocks showed a downward trend from 2005 to 2015 and 2015–2025 under S1. The two main forms of LUCC in these two periods were the expansion of construction land and the sharp reduction of forest land and grassland. Many studies have shown that the rapid growth of construction land is the main reason for the decline in carbon stocks³⁸. Urbanization resulted in the conversion of part of the forest and grassland areas into construction land before the ER-SNWDP began operation; therefore, the carbon sequestration capacity of the ecosystem declined in the study area.

Carbon stocks showed an upward trend from 2015 to 2025 under S2 and the main forms of LUCC change were the expansion of construction land being limited, forest land and grassland areas increasing, and water area increasing significantly. The change from a low carbon density land-use type to a high carbon density land-use type improves the ecological carbon sequestration ability of the study area^39,40. Therefore, the water resources of the water transfer project changed the trend of LUCC, causing a change in the soil and vegetation carbon sinks in the ecosystem, thus leading to a change in the total carbon stocks in the water-receiving area.

Impact of climate change on carbon stock estimation

The analysis of carbon stock driving factors also indicated that climate factors had a certain impact on carbon stock variation in the study area. The primary sources of carbon density data were national and comparable study areas^41,42. However, numerous studies have demonstrated that carbon density is influenced by climate conditions⁴³. Therefore, carbon density was corrected for climate change in the study area. The calibration results are within the range of the carbon density data for Jiangsu and Shandong, indicating the reasonability of the corrected data in this study.

Our study compared the temporal and spatial distributions of carbon stocks before and after carbon density correction, as shown in Figs. 10 and 11. From the perspective of time, after the correction for carbon density, the annual total carbon stocks decreased by nearly 10% and the trend of carbon stocks over the years did not change. The largest difference before and after the correction was in 2010, with a difference of 17,022 × 10⁴ t. The spatial distribution of carbon stocks was consistent before and after the correction and high-value carbon stocks were concentrated in Shandong Province near the Bohai Sea, whereas low-value carbon stocks were distributed in the coastal and lake areas of Jiangsu Province. The specific difference was that the carbon density of each cover was correspondingly reduced after correction. The carbon density of the neighboring province (Jiangsu Province) was used as the initial data for the carbon stock estimation and the meteorological data of Jiangsu Province and the study area were used for correction.

Fig. 10.

Carbon stock for different land use types before and after correction.

Fig. 11.

Spatial changes in carbon stock in the study area before and after correction (Software: ArcMap 10.8, https://www.esri.com).

Studies have shown that the carbon density of biomass and soil organic matter is significantly correlated with annual mean precipitation and weakly correlated with annual mean temperature⁴⁴. Our study helps verify these findings. The average annual precipitation of the study area (848.5 mm) was lower than that of Jiangsu Province (1066.2 mm) and the average annual temperature of the study area (14.6) was lower than that of Jiangsu Province (15.8), which had a significant positive relationship with regional precipitation but a weak relationship with air temperature.

• Impact of water diversion project on the distribution of carbon stock.

The water resource allocation of water diversion project not only affects the spatial distribution of water resources but also influences the spatial distribution of carbon stocks. According to the data on water transfer volume from the ER-SNWDP (Shandong Water Resources Bulletin of China, http://wr.shandong.gov.cn/zwgk_319/fdzdgknr/tjsj/szygb/), Binzhou and Dongying in the northeast of the study area had a significant water transfer volume in 2020, accounting for 35% of the total, with 1.364 and 1.143 billion m³, respectively. From the simulation results of carbon stocks under the water transfer scenario, it can be observed that the areas with a substantial increase in carbon stocks align with the regions of Binzhou and Dongying (Fig. 8). The center of gravity of increase (Fig. 9) also moved to the northeast of Binzhou and Dongying, indicating that water transfer projects may be one of the significant factors influencing the spatial distribution of carbon stocks. However, there are many factors affecting the spatial distribution of carbon stocks, which is difficult to distinguish specifically. For example, the water transfer volume of Jinan is 0.733 billion m³. Due to the acceleration of urbanization in this region in recent years, carbon stocks in this region showed a decreasing trend (Figs. 8 and 9). Currently, there is limited research on the impact of water diversion projects on the distribution of carbon stocks. With the continuous operation of the ER-SNWDP, the impact of water resources on the distribution of carbon stocks may become more pronounced.

Limitations

CA-Markov model lacks consideration of social and economic factors in the modeling process, but it has advantages in long sequence prediction. Through a detailed and accurate evaluation of the simulation results, it is found that the results meet the accuracy requirements. Therefore, the CA-Markov model can effectively simulate the LUCC in the study area. The estimation of carbon stocks by the InVEST model largely depends on the accuracy of carbon density data. Therefore, on the one hand, we conducted field sampling measurements at sampling points, and on the other hand, we corrected the carbon density data based on meteorological data, which to some extent improved the accuracy. Although the InVEST model still has some inaccuracies, such as the simplification of model algorithm and the limitations of the application, the estimation results of carbon stocks can reflect the trend of future changes in carbon stocks.

As the ER-SNWDP was implemented in 2013, the period was relatively short, meaning only limited data after water transfer were available for comparison, which has a certain impact on the evaluation and prediction results. With the continuous operation of the ER-SNWDP, we will further study the spatio-temporal evolution and influencing factors of carbon stock in water-receiving areas in the future.

Conclusions

This study evaluated the dynamic variation in carbon stocks in the context of the water transfer project, taking the Jiangsu-Shandong section of the ER-SNWDP as the research area. The primary conclusions are as follows:

(1) LUCC in the study area from 2005 to 2015 was characterized by the continuous expansion of construction land and continuous reduction of woodland and grassland areas. From 2015 to 2020, following the operation of the ER-SNWDP in 2013, the water area showed a continuous growth trend. LUCC retained its original trend under S1 but under S2, the disorderly expansion of construction land was curbed, the reduction in forest land and grassland was alleviated, and the water area increased.

(2) Carbon stocks in the study area from 2005 to 2020 first decreased and then increased, and the carbon stocks decreased by 1653.87 × 10⁴ t from 2005 to 2015. Due to the implementation of the project the research area had better carbon sequestration capacity, carbon stocks increased from 130.71 × 10⁴ t between 2015 and 2020. Under the natural variation scenario from 2015 to 2025, the carbon stock would decrease by 1228.35 × 10⁴ t. However, under the ER-SNWDP scenario, there would be an increase of 262.84 × 10⁴ t. In addition, the water resource allocation of ER-SNWDP may affect the spatial distribution of carbon stocks. In the northeast region, particularly in the Binzhou and Dongying areas with large water transfer volumes, the increase in carbon stocks was significant, and the center of gravity of increase also tended to tilt these areas.

(3) According to the analysis of driving factors, land use was the dominant factor influencing carbon stock in the study area, with an explanatory power exceeding 50% in all years. Additionally, the results indicated that climate factors also had a certain impact on the variation of carbon stock. According to the results of the interaction factor analysis, the strongest interaction factor after 2005 was “land use ∩ nighttime lights”, indicating that the interaction between socio-economic factors and land use factors gradually amplified the impact on the spatial variation of carbon stocks.

Policy recommendations

The implementation of the ER-SNWDP caused obvious changes in the water quantity in the receiving area. A large amount of water resources has been replenished in water-scarce areas through the ER-SNWDP, which significantly improves the regional water resource carrying capacity and provides a foundation for the sustainable development of the ecosystem and human production and life in water-receiving areas. Therefore, more stringent requirements have been proposed for the management of water-diversion projects and ecosystems.

(1) A scientific mechanism for optimal management of carbon and water resources should be established. Most water resource management methods consider flood control, power generation, and water supply needs⁴⁵, whereas ecological flow is generally considered in terms of ecology⁴⁶. In addition, insufficient consideration has been given to the carbon sink capacity caused by water resources. Therefore, there is an urgent need to study the water resource optimization scheduling method of inter-basin water transfer projects that considers the multi-objective relationship between carbon sink capacity and water quality and quantity to promote the innovation of the optimal allocation mechanism of water and carbon resources.

(2) Land use patterns should be optimized. According to the response process of LUCC to water resources, the relationship between ecological environmental protection and social and economic development was analyzed. The results showed that water resources are the largest rigid constraint. Water is vital for the city, land, people, and production, so should be considered when planning the population and urban and industrial development. Plans should aim to achieve a spatial balance between upstream and downstream, trunk and tributaries, and left and right banks, and firmly follow the path of green, sustainable, and high-quality development.

(3) More attention should be paid to ecological civilization construction while boosting economic development. Water-receiving areas should strengthen monitoring and assessment of the ecosystem to understand their health. Guided by the concept of “green water and green mountains are gold hills and silver mountains”, several ecological parks can be established in areas with low carbon sequestration levels, and green barriers can be gradually established to improve regional carbon neutrality.

(4) Ecological compensation system, or carbon ecological compensation system for inter-basin water transfer should be promoted. The price fluctuation mechanism of water ecological products with instantaneous linkage with ecological environment quality should be established. We should put a “price” on green water and green mountains, so that the “ecological price” such as the freshness of the air, the beauty of the environment, and the landscape index can realize dynamic changes, and vigorously develop the ecological tourism industry, so as to promote the ecological resource property rights such as water right and carbon sink as the object of market trading.

Author contributions

All authors contributed to the Conceptualization and design. Peng, Z.Y.: Conceptualization, Methodology, Writing–review& editing, Visualization, Supervision. Li, M.T.: Data curation, Software. Liu, Y.M.: Data curation, Investigation, Writing–original draft. Yin, J.X..: Data curation, Investigation. Fang, H.Y.: Data curation, Software. Wen, J.W.: Data curation, Software. All listed authors have approved the manuscript before submission, including the names and order of authors.

Funding

This research is funded by National Natural Science Foundation of China (52379027), Open Project Program of Engineering Research Center of High-efficiency and Energy-saving Large Axial Flow Pumping Station, Jiangsu Province, Yangzhou University (ECHEAP024), Open Project Program of Jiangsu Province Engineering Research Center of Knowledge Management and Intelligent Service, Yangzhou University (KMIS202204), Social Science Foundation of Jiangsu Province (24ZHC010), Open Research Fund of State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research (IWHR-SKL-KF202404),Yangzhou science and technology plan project (YZ2023069), Open Project Program of Yangtze River Cultural Research Institute (CJ2316).

Data availability

The data presented in this study are available in the article.

Declarations

Competing interests

The authors declare no competing interests.