Climate Change Impacts on Grey Water Footprint of Agricultural Total Nitrogen in the Yangtze River Basin Based on SSP–InVEST Coupling

Abstract

With climate change, the spatial and temporal patterns of precipitation are altered to a certain degree, which potentially affects the grey water footprint (GWF) of total nitrogen (TN) in agriculture, thereby threatening water security in the Yangtze River Basin (YRB), the largest river in China. The current study constructs an assessment framework for climate change impacts on the GWF of agricultural TN by coupling Shared Socioeconomic Pathways (SSPs) with the InVEST model. The framework consists of four components: (i) data collection and processing, (ii) simulating the two critical indicators (L_TN and W) in the GWF model based on the InVEST model, (iii) calculating the GWF and GWF index (GI) of TN, and (iv) calculating climate change impact index on GWF of agricultural TN (CI) under two SSPs. It is applied to the YRB, and the results show the following: (i) GWFs are 959.7 and 961.4 billion m³ under the SSP1-2.6 and SSP5-8.5 climate scenarios in 2030, respectively, which are both lower than that in 2020 (1067.1 billion m³). (ii) The GI values for TN in 2030 under SSP1-2.6 and SSP5-8.5 remain at “High” grade, with the values of 0.95 and 1.03, respectively. Regionally, the water pollution level of Taihu Lake is the highest, while that of Wujiang River is the lowest. (iii) The CI values of the YRB in 2030 under SSP1-2.6 and SSP5-8.5 scenarios are 0.507 and 0.527, respectively. And the CI values of the five regions in the YRB are greater than 0, indicating that the negative effects of climate change on GWFs increase. (iv) Compared with 2020, L_TN and W in YRB in 2030 under the two SSPs decrease, while the GI of TN in YRB rises from SSP1-2.6 to SSP5-8.5. The assessment framework can provide strategic recommendations for sustainable water resource management in the YRB and other regions globally under climate change.

1. Introduction

Climate change is significantly altering the hydrological cycle and agricultural systems [1,2]. The Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) indicates that the rise in average surface temperature may exceed 1.5 °C within the next two decades [3]. As global temperature rises, the spatial and temporal patterns of precipitation are altered to a certain degree, leading to changes in the distribution of water resources [4]. These changes have profound impacts on agricultural production, and in particular, total nitrogen (TN) emissions in the future will show significant spatial heterogeneity with alterations in precipitation patterns [5]. TN from the excessive use of fertilizers and pesticides in grain production enters water bodies through surface runoff and subsurface leakage, which increases the risk of eutrophication of water bodies [6]. Consequently, water quality and ecosystem health are affected, leading to severe water pollution problems [7].

The grey water footprint (GWF) is an important model for evaluating the pollution in water and quantifying the volume of water necessary to dilute pollutants in waters to safe levels [8]. It is crucial for assessing the impact of the TN from agricultural activities on water resources [9]. Many studies have evaluated the GWF of TN loads. For example, Mekonnen and Hoekstra [10] estimated the global-scale GWF of nitrogen loads from human activities and their impact on freshwater systems. Yi et al. [11] quantified the GWF of nitrogen for different crops at the provincial level in China. Feng et al. [12] developed an integrated model based on GWF to more accurately calculate the excess nitrogen loads in Central Asia and to evaluate their potential pressure on the aquatic environment.

However, despite the wide application of the grey water footprint (GWF), existing research has significant shortcomings in deeply integrating climate change factors into the GWF assessment of agricultural TN. (i) Limited coupling depth: Most studies only consider climate change as a background factor, lacking an in-depth portrayal of the mechanisms of the climate–hydrology–pollution process chain in GWF. For example, studies by Jiang et al. [13] based on SD, Chen et al. [14] based on SWAT, and Arunrat et al. [15] based on CMIP6 mainly focus on overall water footprint (WF) or yield, without deeply analyzing the specific impact paths of climate change on TN migration, transformation, and final GWF. (ii) Lack of a systematic framework: There is currently no comprehensive assessment framework specifically designed to evaluate the impact of climate change on the GWF of agricultural TN, which integrates future climate scenarios with process-based models. (iii) Insufficient integration of future scenarios: in particular, there is a lack of studies that systematically incorporate the differentiated socioeconomic development and corresponding climate scenarios represented by shared socioeconomic pathways (SSPs) into the GWF assessment of agricultural non-point source pollution, especially TN.

Therefore, the core scientific question that this study aims to address is how to systematically quantify and evaluate the spatial and temporal pattern evolution and potential impact of future climate change (combined with different SSPs) on the GWF of TN generated by agricultural activities.

To fill the aforementioned research gap and address this key problem, this study innovatively constructs an assessment framework that integrates SSP climate scenarios with the InVEST model (Integrated Valuation of Ecosystem Services and Tradeoffs). The InVEST model, known for its modularity, strong spatial explicitness, and relatively simplified parameter requirements, is widely used in ecosystem service assessment [16]. Its annual water yield (AWY) module, based on the Budyko water–energy coupling theory, estimates water yield at the pixel scale, which is crucial for calculating the denominator of the GWF index (water resource, W). The nutrient delivery ratio (NDR) module estimates the TN load entering water bodies (i.e., the numerator of the GWF, pollution load: L_TN) by simulating nitrogen transport via surface runoff and subsurface flow. By integrating future climate data (precipitation and evapotranspiration) from SSPs to drive the InVEST model, this framework can simulate the impact of climate change on agricultural TN transport and final GWF under different socioeconomic development pathways.

Therefore, the objectives of this study are (i) to construct an assessment framework for evaluating the impact of climate change on the GWF of agricultural TN, driven by SSP climate scenarios and based on the InVEST model (AWY and NDR modules); (ii) to apply this framework to a typical region (the Yangtze River Basin), quantify and compare the spatiotemporal variation characteristics of the GWF and GWF index of agricultural TN in future years relative to the baseline year under different SSP scenarios (this study selects SSP1-2.6, representing a sustainable development pathway, and SSP5-8.5, representing a fossil-fueled development pathway); and (iii) to construct and calculate a climate change impact index to quantify the effect of climate change on the variation in GWF of agricultural TN. The establishment and application of this framework will provide a scientific tool for gaining an in-depth understanding of the evolution of pressure from agricultural non-point source nitrogen pollution on water resources under climate change. It will also offer quantitative evidence for formulating climate change adaptation strategies for sustainable water resource and agricultural management in the Yangtze River Basin and similar regions globally.

2. Materials and Methods

2.1. Study Area

The third largest basin globally, the Yangtze River Basin (YRB), is located at 90°33′ to 122°25′ E and 24°30′ to 35°45′ N. Its total area is 1.8 million km² and accounts for 18.8% of the land area of China [17]. As shown in Figure 1, in accordance with the sub-watershed boundaries, it is divided into eight regions: Minjiang River (162,602 km²), Wujiang River (87,871.5 km²), Jialing River (161,144 km²), Dongting Lake (261,916 km²), Hanjiang River (153,388 km²), Poyang Lake (162,063 km²), Taihu Lake (36,543.6 km²), and Mainstream (754,976 km²). The YRB flows through mountains, plateaus, basins, hills, and plains, exhibiting a multi-stepped landscape. In addition, the regional and temporal distribution of precipitation in the YRB is extremely uneven because of the subtropical monsoon climate, with an average annual precipitation of 1067 mm. Precipitation is primarily concentrated in May–September and spatially shows a general tendency of low in the northwest and high in the southeast [18].

Due to its unique natural conditions, the YRB is an important grain production base in China, with cultivated land accounting for 27% of China’s cultivated area [19]. However, excessive fertilizer is used in agricultural production, and TN enters water bodies through surface runoff and subsurface leakage, which causes severe water pollution [6]. Excessive nitrogen inputs lead to frequent water eutrophication events (such as algal blooms), and the TN concentrations in some areas significantly exceed the Class III water quality limit (1.0 mg/L) of the environmental quality standard for surface water (GB 3838-2002) [20], which poses a serious threat to drinking water safety, aquatic ecosystem health and water resource functionality. Future precipitation would become more regionally and temporally heterogeneous under the effects of climate change, and the water resources and load export of TN in the watershed would be altered. Therefore, systematic quantification and evaluation of climate change impacts on GWF of agricultural TN is essential.

2.2. Data Collection and Processing

(i)

Data sources

The data adopted in the current study consist of the CMIP6 global climate model (GCM) data under climate change and the driving data for the AWY and NDR modules of the InVEST model (version: 3.14.2). The precipitation and evapotranspiration data from the IPSL-CM6A-LR climate model data in CMIP6 are selected, which contains two scenarios: sustainable development (SSP1-2.6), and traditional fossil-fueled development (SSP5-8.5) [21]. As shown in

Table 1. Input data sources for InVEST AWY and NDR modules.

Data Type	Resolution	Sources
Precipitation	1 km, annual	Baseline year (2020): National Earth System Science Data Center (https://www.geodata.cn/, accessed on 20 May 2025)
	1.3° × 2.5°, daily	Future year (2030): the IPSL-CM6A-LR climate model data
Evapotranspiration	1 km, monthly	Baseline year (2020): National Earth System Science Data Center
	1.3° × 2.5°, daily	Future year (2030): the IPSL-CM6A-LR climate model data
Digital Elevation Model	250 m	Resource and Environmental Science Data Platform (https://www.resdc.cn/, accessed on 20 May 2025)
Land use	30 m	Resource and Environmental Science Data Platform
Root Restricting Layer Depth	1 km	Yan et al. [22]
Plant Available Water Content	1 km	Calculated using the “ASWC” formula proposed by Zhou et al. [23] with soil texture and organic matter data from the Harmonized World Soil Database v1.2
Watersheds	Vector data	National Earth System Science Data Center

, the driving data of the AWY and NDR models include precipitation, evapotranspiration, land use, etc.

(ii)

Downscaling of precipitation data

The Delta method is employed to downscale the precipitation from the IPSL-CM6A-LR climate model. The method adjusts the GCM output data to reflect local climate characteristics more accurately by analyzing the differences between the observed data and the (GCM) output data [24]. The formula for precipitation in the future period is [24]

$${\mathit{P}}_{\mathit{f}}={ \mathit{P}}_{\mathit{o}}·{\displaystyle \frac{{\mathit{P}}_{\mathit{Gf}}}{{\mathit{P}}_{\mathit{Go}}}}$$

(1)

where P_f is the future precipitation reconstructed by the Delta method, P_o is the observed precipitation in the standard period, P_Gf is the future precipitation predicted by the climate model, and P_Go is the precipitation in the standard period simulated by the climate model.

2.3. Grey Water Footprint Model of Total Nitrogen

The GWF of TN quantifies the volume of freshwater required to dilute the agricultural TN load in the study area to the permissible threshold, on the basis of the natural background concentration and the water quality criteria of TN [8]. It is calculated as [8]

$${\mathit{G}}_{\mathit{TN}}={\displaystyle \frac{{\mathit{L}}_{\mathit{TN}}}{({\mathit{t}}_{\mathit{TN}}-{\mathit{b}}_{\mathit{TN}})·{10}^6}}$$

(2)

where G_TN is the GWF of TN (billion m³); t_TN is the permissible threshold in the receiving water body (mg/L) [25]; b_TN is the background value of TN in the receiving water body (mg/L); and L_TN is the TN load entering the water body (kg), which is simulated in Section 2.4.1.

To enable spatial comparison across regions with differing water availability, the grey water footprint index (GI) is calculated as the ratio of G_TN to the local water resources (W, billion m³) (simulated in Section 2.4.2). This index, also referred to as water pollution level in some research [10], is defined as

$$\mathit{GI}={\displaystyle \frac{{\mathit{G}}_{\mathit{TN}}}{\mathit{W}}}$$

(3)

When GI > 1, the water resources of the region cannot dilute the TN load discharged to the permissible threshold, and a larger value indicates more severe water pollution [10]. As GWF calculations in this study rely on model simulations, inherent input uncertainties exist. Considering limitations in the InVEST model’s accuracy for large-scale applications, spatial variations in input data, and potential errors, we estimate a GWF uncertainty range of ±20%. This margin of error aligns with typical uncertainty levels reported in water footprint assessments [26,27] and reflects our regional simulation outcomes. Thus, GI is divided into four grades (Figure 2): low (0 ≤ GI ≤ 0.2), moderate (0.2 < GI < 0.8), high (0.8 ≤ GI ≤ 1.2), and extreme (GI > 1.2) [10].

2.4. Responses of Total Nitrogen Load and Water Resources to Climate Change

The InVEST model, developed by Stanford University and partner institutions, is a modular computational tool for assessing ecosystem services. This study employs two InVEST modules: the NDR model, which simulates the transport of TN from sources to water bodies, producing total nitrogen load (L_TN), and the AWY model, which estimates renewable water resources at the grid scale using the water balance principle, yielding water resources (W).

To ensure the scientific validity and applicability, this study strictly defined defines strict boundary conditions for the InVEST model across three dimensions: (i) Spatial boundary: The simulation range for both the NDR and AWY models is the closed boundary of the YRB. And topological inspection ensured hydrological connectivity throughout the simulated area. (ii) Temporal Boundary: The baseline year is set as 2020, and the future scenario year is 2030. All input climate data (precipitation, evapotranspiration) are processed as annual averages to align with the model’s steady-state simulation requirements. (iii) Process boundary: The NDR model focuses on nitrogen transport dominated by agricultural non-point source pollution. It accounts for pollutant transport through surface and subsurface runoff, but excludes point source emissions, atmospheric wet deposition, and biochemical transformation processes within the river channels. The AWY model calculates water yield based on the rainfall–evapotranspiration balance principle, including vegetation interception and soil infiltration processes, but neglects the influence of snowmelt, lateral exchange of deep groundwater, and inter-basin water transfer.

2.4.1. Simulating Distribution of Total Nitrogen Load Based on the InVEST Model

(i)

Theoretical method

The NDR model simulates the spatial movement of nutrients like nitrogen and phosphorus within watersheds using the mass conservation principle, focusing on ecosystems’ ability to retain pollutants [16]. Its core assumptions are that nutrient transport is primarily driven by surface runoff and subsurface flow, with retention efficiency determined by downslope land use patterns and their spatial arrangement. This study specifically estimates the agricultural total nitrogen (TN) non-point source pollution load within the watershed. The equation of L_TN in the watershed is [16]

$${\mathit{L}}_{\mathit{TN}}=\sum _{\mathit{i}}({\mathit{load}}_{\mathit{surf,i}}·{\mathit{NDR}}_{\mathit{surf,i}}+{\mathit{load}}_{\mathit{subs,i}}·{\mathit{NDR}}_{\mathit{subs},\mathit{i}})$$

(4)

where load_surf,i and load_subs,i are the surface and subsurface TN loads for pixel i in the watershed (kg), respectively, and NDR_surf,i and NDR_subs,i are the surface and subsurface nitrogen delivery ratios, respectively.

Equation (4) integrates nutrient load estimation and delivery ratio calculation, ultimately determining the nutrient export for each grid cell within the watershed. The nutrient load is calculated based on land-use-specific export coefficients and the runoff potential index. The surface nitrogen delivery ratio depends on the Topographic Connectivity Index (IC_i) and vegetation retention efficiency. The IC_i itself is derived from the input Digital Elevation Model (DEM). The subsurface nitrogen delivery ratio is calculated using an exponential decay function, modulated by the subsurface flow path length and retention efficiency. The specific calculation formulas are detailed in the InVEST User’s Guide [16].

(ii)

Model input and calibration for the NDR model

In addition to the environmental data (Table 1), the biophysical table is a key input for the NDR model. This table includes the following parameters: load_n (nitrogen loading for each land use), eff_n (maximum nitrogen retention efficiency), crit_len_n (distance beyond which a land use type retains nitrogen at maximum capacity), and proportion_subsurface_n (proportion of TN that dissolves into the subsurface). Following Sharp et al. [16] and Li et al. [28], the biophysical table used in this study is presented in Appendix A,

Table A1. Biophysical table input into the NDR model.

Description	Lucode	Load_n	Eff_n	Crit_len_n	Proportion_ subsurface_n
Paddy field	11	24.2	0.25	25	0.5
Dryland	12	24.2	0.25	25	0.5
Wooded land	21	3.68	0.7	300	0.1
Shrubland	22	3.68	0.7	300	0.1
Open woodland	23	3.68	0.7	300	0.1
Other woodland	24	3.68	0.7	300	0.1
High-cover grassland	31	8.5	0.4	100	0.15
Medium grassland	32	8.5	0.4	100	0.15
Low-cover grassland	33	8.5	0.4	100	0.15
Rivers and canals	41	0.01	0.05	10	0.3
Lakes	42	0.01	0.05	10	0.3
Reservoirs and ponds	43	0.01	0.05	10	0.3
Permanent glacial snow	44	0.01	0.05	10	0.3
Mudflat	45	0.01	0.05	10	0.3
Beaches	46	0.01	0.05	10	0.3
Townsite	51	14.5	0.05	10	0.2
Rural settlements	52	14.5	0.05	10	0.2
Other Building Land	53	14.5	0.05	10	0.2
Sandy land	61	5	0.05	10	0.1
Gobi	62	5	0.05	10	0.1
Saline soil	63	5	0.05	10	0.1
Marshland	64	5	0.05	10	0.1
Bare land	65	5	0.05	10	0.1
Bare rocky land	66	5	0.05	10	0.1
Others	67	5	0.05	10	0.1

.

Moreover, four parameters need to be input. Threshold Flow Accumulation classifies streams from the DEM, and the Borselli K Parameter calibrates the relationship between hydrological connectivity and nutrient delivery ratio. Subsurface Critical Length refers to the distance traveled beyond which the soil is assumed to retain nitrogen at maximum capacity, and Subsurface Maximum Retention Efficiency represents the highest achievable nitrogen retention efficiency via subsurface flow.

These parameters have physical meaning, but they lack fixed values and must be calibrated within the model. Following Sharp et al. [16] and Li et al. [28], these parameters are adjusted within plausible ranges to align the simulated annual TN output for the YRB in 2020 with the observed value (815.83 kilotons) from monitoring stations of the Hydrology Bureau of Changjiang Water Resources Commission. The optimal parameter combination is identified based on the simulation (800.32 kilotons) closest to observed values: Threshold Flow Accumulation: 1200; Borselli K Parameter: 1.9; Subsurface Critical Length: 650; Subsurface Maximum Retention Efficiency: 0.5.

2.4.2. Simulating Distribution of Water Resources Based on the InVEST Model

(i)

Theoretical method

The AWY module employs the Budyko framework to calculate annual water yield at the grid-cell scale, based on the balance between precipitation (P) and actual evapotranspiration (AET) [16]. The formula for calculating the water yield (W_i) of pixel i is [16]

$${\mathit{W}}_{\mathit{i}}=(1-{\displaystyle \frac{{\mathit{AET}}_{\mathit{i}}}{{\mathit{P}}_{\mathit{i}}}})·{\mathit{P}}_{\mathit{i}}$$

(5)

where AET_i, and P_i are the actual evapotranspiration (mm) and the annual precipitation (mm) for pixel i, respectively.

In Formula (5), AET_i is calculated using the Z parameter and the plant available water capacity, representing the interaction between climate, soil, and vegetation. The specific calculation method is detailed in the Invest User’s Guide [16].

(ii)

Model input and calibration for the AWY model

In addition to environmental data inputs (Table 1), the biophysical table constitutes a critical component of the AWY model’s input parameters. This table contains three key variables: root_depth, indicating maximum plant root depth for each land use class; Kc, representing crop coefficients; and LULC_veg, a binary vegetation indicator (0/1) determining evapotranspiration eligibility. According to Sharp et al. [16] and Li [29], the specific biophysical parameters used in this study are detailed in Appendix A,

Table A2. Biophysical table input into the AWY model.

Description	Lucode	Root_depth	Kc	LULC_veg
Paddy field	11	2100	0.7	0
Dryland	12	2000	0.65	1
Wooded land	21	5200	1	1
Shrubland	22	5200	0.95	1
Open woodland	23	5200	0.93	1
Other woodland	24	5200	0.93	1
High-cover grassland	31	2600	0.85	1
Medium grassland	32	2300	0.65	1
Low-cover grassland	33	2000	0.65	1
Rivers and canals	41	100	1	0
Lakes	42	100	1	0
Reservoirs and ponds	43	100	1	0
Permanent glacial snow	44	100	0.5	0
Mudflat	45	1000	1	0
Beaches	46	1000	1	0
Townsite	51	100	0.3	0
Rural settlements	52	100	0.2	0
Other Building Land	53	100	0.3	0
Sandy land	61	300	0.2	0
Gobi	62	300	0.2	0
Saline soil	63	300	0.2	0
Marshland	64	300	1	0
Bare land	65	300	0.2	0
Bare rocky land	66	300	0.2	0
Others	67	300	0.2	0

.

The Z parameter represents a seasonality factor with a value range of 1–30. This parameter is calibrated using the water yield coefficient, defined as the ratio of water resources to annual precipitation. According to China’s water resources in 2020, the YRB recorded 1274.17 billion m³ of water resources and 1282.0 mm of precipitation, yielding a calculated water yield coefficient of 0.553 [18]. The Z parameter is iteratively adjusted until the AWY model’s calculated water yield coefficient matches the observed value of 0.553. Final calibration established an optimal Z parameter value of 0.7, resulting in a calculated water yield of 1148.91 billion m³ for the YRB, which closely aligns with observed data, confirming the model’s validity.

2.5. Climate Change Impact Index on Grey Water Footprint of Agricultural Total Nitrogen

To quantify the climate change impacts on the GWF of agricultural TN, a climate change impact index (CI) is constructed based on the GI of TN. The GI in the future year and the baseline year of region i are denoted as GI_fi and GI_bi, respectively. As shown in Figure 3, CI is classified into three major categories based on the relationship between GI_fi and GI_bi: CI < 0, CI = 0, and CI > 0.

(i)

When GI_fi < GI_bi, the CI < 0. It indicates that in future climate scenarios, GI_fi decreases, the water pollution level declines, and the negative impact of climate change on GWF of agricultural TN reduces.

(ii)

When GI_fi = GI_bi, the CI = 0. It shows that under the future climate scenario, GI_fi remains unchanged, and the water pollution level remains stable.

(iii)

When GI_fi > GI_bi, the CI > 0. It suggests that under future climate scenarios, GI_fi increases, the water pollution level rises, and the negative impacts of climate change on GWF of agricultural TN increase.

The calculation formula for the CI is

$$CI=\left\{\begin{array}{cc}-\left(b(G{I}_{fi})+0.25\times {\displaystyle \frac{\left|G{I}_{fi}-G{I}_{bi}\right|}{G{I}_{bi}}}\right)& G{I}_{fi}<G{I}_{bi}\\ 0& G{I}_{fi}=G{I}_{bi}\\ b(G{I}_{fi})+0.25\times {\displaystyle \frac{\left|G{I}_{fi}-G{I}_{bi}\right|}{G{I}_{bi}}}& G{I}_{fi}>G{I}_{bi}\end{array}\right.$$

(6)

$$b(G{I}_{fi})=\left\{\begin{array}{cc}0& 0\le G{I}_{fi}\le 0.2\\ 0.25& 0.2<G{I}_{fi}<0.8\\ 0.5& 0.8\le G{I}_{fi}\le 1.2\\ 0.75& G{I}_{fi}>1.2\end{array}\right.$$

(7)

where GI_fi and GI_bi are the GI in the future year and the baseline year of region i, respectively; 0.25 in Equation (6) is the adjustment coefficient reflecting the relative change; and b(GI_fi) is the base value, which is determined by the grade of the GI in future years, as shown in Equation (7).

The range of CI values and their corresponding meanings are presented in

Table 2. Calculation formula and implication description for CI.

Variation Trend	CI Value Range	Implication
GIfi < GIbi	[−1, −0.75)	Negative impact of climate change on GWF of TN reduces, and GIfi is at “Extreme” grade.
	[−0.75, −0.5]	Negative impact of climate change on GWF of TN reduces, and GIfi is at “High” grade.
	(−0.5, −0.25]	Negative impact of climate change on GWF of TN reduces, and GIfi is at “Moderate” grade.
	(−0.25, 0)	Negative impact of climate change on GWF of TN reduces, and GIfi is at “Low” grade.
GIfi = GIbi	0	Water pollution level remains stable.
GIfi > GIbi	(0, 0.25]	Negative impacts of climate change on GWF of TN increase, and GIfi is at “Low” grade.
	(0.25, 0.5)	Negative impacts of climate change on GWF of TN increase, and GIfi is at “Moderate” grade.
	[0.5, 0.75]	Negative impacts of climate change on GWF of TN increase, and GIfi is at “High” grade.
	(0.75, 1]	Negative impacts of climate change on GWF of TN increase, and GIfi is at “Extreme” grade.

.

2.6. Assessment Framework for Climate Change Impacts on Grey Water Footprint of Agricultural Total Nitrogen

This study develops an evaluation framework (Figure 4) to quantify the impact of climate change on the GWF of agricultural TN. The framework consists of four main computational steps: (i) data preparation (integrating climate scenario data (SSP1-2.6 and SSP5-8.5) with geospatial datasets); (ii) key indicator simulation (using the AWY and NDR modules of the InVEST model to calculate W and L_TN, respectively); (iii) pollution quantification (calculating G_TN and GI based on W and L_TN); and (iv) climate impact assessment (comparing GI values between the baseline and future years under both SSPs to derive the CI, which evaluates the overall impact of climate change on agricultural TN pollution levels).

Figure 4 visually presents the framework’s information flow, starting with raw input data, followed by process simulation and footprint calculation, and concluding with impact quantification. The framework creates a comprehensive analytical chain from climate scenario inputs to impact assessment outputs. It guarantees the comparability and repeatability of the results, providing a scientific foundation for water resource management in the YRB.

3. Results

3.1. Spatial and Temporal Distribution for Total Nitrogen Loads

According to the NDR model in the InVEST model, the L_TN of the YRB in the baseline year (2020) and under two climate scenarios (SSP1-2.6 and SSP5-8.5) in 2030 is simulated. The results are shown in Figure 5.

As illustrated in Figure 5, the L_TN value for the YRB in 2020 is 800.32 kilotons. Under the climate scenarios of SSP1-2.6 and SSP5-8.5, the L_TN values for the YRB decrease to 719.77 and 721.04 kilotons by 2030, representing a decline rate of 10.1% and 9.9%, respectively.

Temporally, compared with L_TN values in 2020, under the SSP1-2.6 climate scenario, the higher decline rates in L_TN values among the eight regions of the YRB in 2030 are observed in the Dongting Lake, Hanjiang River, and Wujiang River, with rates of 19.8%, 18.5%, and 17.8%, respectively. The L_TN decline rates in the other five regions range from 0.9% to 10.7%. Under SSP5-8.5 in 2030, the Wujiang River and Dongting Lake experience significant declines in L_TN values at rates of 24.2% and 23.5%, respectively. And the decrease rates of L_TN values in the Mainstream, Taihu Lake, Poyang Lake and Minjiang River range from 1.8% to 9.7%. In contrast, the L_TN values in the Hanjiang River and Jialing River increase, with growth rates of 0.6% and 1.1%, respectively.

Spatially, the ranking of L_TN values among the eight regions of the YRB in 2020 and under the SSP5-8.5 climate scenario in 2030 is as follows: Mainstream > Dongting Lake > Poyang Lake > Hanjiang River > Minjiang River > Jialing River > Wujiang River > Taihu Lake. Under SSP1-2.6 in 2030, the Mainstream (262.59 kilotons) and Dongting Lake (121.22 kilotons) remain dominant, accounting for 53.3% of the YRB. In 2020, despite having the smallest area (2.1% of basin total), Taihu Lake exhibits the highest L_TN per unit area (0.81 t/km²), which is 2.2 times that of the Mainstream (0.37 t/km²). This indicates that pollution intensity is primarily driven by land use patterns rather than area size.

3.2. Spatial and Temporal Distribution for WATER Resources

The W of the YRB in the baseline year (2020) and under two climate scenarios (SSP1-2.6 and SSP5-8.5) in 2030 is simulated based on the AWY model in the InVEST model. And the results are illustrated in Figure 6.

As depicted in Figure 6, the total W value of the YRB in 2020 is 1148.91 billion m³, and the total W values in 2030 under the SSP1-2.6 and SSP5-8.5 climate scenarios decrease to 1005.00 and 934.72 billion m³, with a decrease rate of 12.5% and 18.6%, respectively.

Temporally, compared with the W value in 2020, the Dongting Lake has the highest decline rate of 32.7% under SSP1-2.6 in 2030 among the eight regions of the YRB. The other seven regions show decline rates ranging from 2.2% to 15.3%. Under SSP5-8.5 in 2030, compared with the W value in 2020, the Dongting Lake shows the highest decline rate of 40.4%. The decline rates for the W values in the Jialing River, Poyang Lake, Wujiang River, Taihu Lake, and Mainstream range from 14.3% to 20.4%. Conversely, the W value of the Hanjiang River rises in 2030, with an increase rate of 1.2%.

Spatially, the W values of the eight regions in the YRB in 2020 are ranked as follows: Mainstream > Dongting Lake > Poyang Lake > Minjiang River > Jialing River > Hanjiang River > Wujiang River > Taihu Lake. The W value of the Mainstream is the largest at 391.87 billion m³, accounting for 34.1% of the YRB. Dongting Lake and Poyang Lake follow, with W values accounting for 20.5% and 13.8% of the YRB, respectively. Poyang and Dongting Lakes show the higher W per unit area (0.98 and 0.90 million m³/km² respectively). In contrast, the Jialing River and Mainstream have notably lower densities (0.52 million m³/km² each). Notably, while Taihu Lake’s W per unit area (0.71 million m³/km²) exceeds that of the Mainstream, its exceptionally small area results in the basin’s lowest total W (25.99 billion m³).

Under the SSP1-2.6 and SSP5-8.5 climate scenarios, the W values of the eight regions in the YRB in 2030 are in the order of Mainstream > Dongting Lake > Poyang Lake > Minjiang River > Hanjiang River > Jialing River > Wujiang River > Taihu Lake. The W values of Mainstream are the largest, 362.66 and 335.86 billion m³, accounting for 36.1% and 35.9% of the YRB, respectively, while the W values of the Taihu Lake are the smallest, 23.64 and 21.73 billion m³, accounting for 6.2% and 6.5% of the YRB, respectively.

3.3. Grey Water Footprint Results of Total Nitrogen in the Yangtze River Basin

The GWF results of agricultural TN in the YRB in the baseline year (2020) and under two climate scenarios (SSP1-2.6 and SSP5-8.5) in 2030, calculated according to Equation (2) in Section 2.3, are illustrated in Figure 7.

As shown in Figure 7, the GWF of the YRB in 2020 is 1067.1 billion m³. Under the SSP1-2.6 and SSP5-8.5 climate scenarios in 2030, the GWFs decline to 959.7 and 961.4 billion m³, respectively.

In terms of temporal changes, compared with the GWF in 2020, the GWFs of all eight regions of the YRB decrease under the SSP1-2.6 climate scenario in 2030. Notably, the decrease rate exceeds 17% in the Dongting Lake, Hanjiang River, and Wujiang River, with GWF values dropping from 201.6, 104.6, and 55.9 billion m³ to 161.6, 85.2, and 45.9 billion m³, respectively. And the Jialing River, Mainstream, Minjiang River, Taihu Lake, and Poyang Lake decrease less than 11%, with GWF values declining from 80.8, 373.0, 92.7, 39.4, and 119.2 billion m³ to 72.2, 350.1, 87.9, 38.6, and 118.1 billion m³.

Under the SSP5-8.5 climate scenario in 2030, compared with the GWF in 2020, the GWF values in the Hanjiang River and Jialing River increase, rising from 104.6 and 80.8 billion m³ to 105.2 and 81.7 billion m³, respectively. On the contrary, the other six regions decline, with decreases exceeding 23% in the Wujiang River and Dongting Lake, where GWF values fall from 55.9 and 201.6 billion m³ to 42.3 and 154.1 billion m³, respectively. The GWF values for the Mainstream, Taihu Lake, Poyang Lake, and Minjiang River decrease by less than 10%, dropping from 373.0, 39.4, 119.2, and 92.7 billion m³ to 336.6, 36.5, 113.9, and 91.0 billion m³.

Spatially, the GWF values for the eight regions of the YRB in 2020 and under the SSP5-8.5 climate scenario in 2030 are ranked as follows: Mainstream > Dongting Lake > Poyang Lake > Hanjiang River > Minjiang River > Jialing River > Wujiang River > Taihu Lake. Under the SSP1-2.6 climate scenario, the ranking in 2030 is Mainstream > Dongting Lake > Poyang Lake > Minjiang River > Hanjiang River > Jialing River > Wujiang River > Taihu Lake. Analysis of GWF per unit area further reveals the spatial heterogeneity of pollution pressure: Taihu Lake (1.08 million m³/km²) > Dongting Lake (0.77 million m³/km²) > Hanjiang River (0.68 million m³/km²). Specifically, the GWF per unit area of Taihu Lake is 2.2 times that of the Mainstream (0.49 million m³/km²), highlighting its dual disadvantages of high nitrogen load per unit and limited dilution capacity.

3.4. Grey Water Footprint INDEX Results of Total Nitrogen in the Yangtze River Basin

According to Equation (3) in Section 2.3, Figure 8 shows the GI results of agricultural TN in the YRB in the baseline year (2020) and under two climate scenarios (SSP1-2.6 and SSP5-8.5) in 2030.

As illustrated in Figure 8, the GI for TN in the YRB in 2020 is 0.93, classified as “High” grade. Under the SSP1-2.6 and SSP5-8.5 climate scenarios, the GI values in 2030 remain “High” grade. However, the values diverged significantly: the GI under SSP5-8.5 (1.03) is distinctly higher than under SSP1-2.6 (0.95), with an increase of 8%.

In 2020, among the eight regions of the YRB, the GI of TN in Taihu Lake and the Hanjiang River Basin is classified as “Extreme”, with values of 1.51 and 1.25 respectively. In the same year, the average TN concentration at the main sections of the YRB was 1.96 mg/L, which significantly exceeded the standard limit value of Class III surface water (1.0 mg/L) [30]. Regionally, TN concentrations averaged 1.27 mg/L in the Lake Taihu Basin [31]. Affected by the superposition of agricultural non-point source pollution, the concentration at the confluence section of the Han River Basin rose to 1.73 mg/L [25]. This spatial pattern of water quality is consistent with the simulated spatial distribution of the GI in 2020 in this study.

Under the SSP1-2.6 scenario, the GI values of the eight regions range from 0.74 to 1.63, with only Taihu Lake (1.63) classified as “Extreme” grade. Under the SSP5-8.5 scenario, the number of regions classified as “Extreme” grade increases to three (Taihu Lake 1.68, Hanjiang River 1.24, Jialing River 1.23), and the GI values of all eight regions are higher than the corresponding values under SSP1-2.6, among which the Jialing River shows the largest increase (26%).

Notably, under the SSP5-8.5 scenario (radiation intensity of 8.5 W/m²), the GI values are comprehensively higher than under SSP1-2.6 (radiation intensity of 2.6 W/m²). Specifically, the GI values of five regions in the YRB (Taihu Lake, Dongting Lake, Poyang Lake, Jialing River, and Mainstream) show a monotonic upward trend with increasing radiation forcing. Among these, the GI of Taihu Lake reaches 1.68 under SSP5-8.5, making it the region with the most prominent pollution pressure in the entire basin.

3.5. Climate Change Impact Index Results of Total Nitrogen in the Yangtze River Basin

In accordance with Section 2.5, the CI results of agricultural TN in the YRB in the baseline year (2020) and under two climate scenarios (SSP1-2.6 and SSP5-8.5) in 2030 are illustrated in Figure 9.

The CI values of YRB in 2030 under the SSP1-2.6 and SSP5-8.5 climate scenarios are 0.507 and 0.527, respectively, indicating that the negative impacts of climate change on GWFs increase, and the GI values are both at “High” grade.

Regionally, the CI values of Hanjiang River, Minjiang River, and Wujiang River are less than 0. The CI values of Wujiang River in 2030 under SSP1-2.6 and SSP5-8.5 are −0.507 and −0.521, respectively, which indicates that the negative impacts of climate change on GWFs reduce, and the GI values are both at “Moderate” grade. The Minjiang River’s CI values in 2030 are −0.256 and −0.254 under the SSP1-2.6 and SSP5-8.5, while that of Hanjiang River is −0.274 under SSP1-2.6, indicating that the negative impact of climate change on GWF reduces, and the GI value is at “High” grade. The Hanjiang River’s CI value under SSP5-8.5 in 2030 is −0.003, reflecting a decreasing negative impact of climate change on GWF, with the GI remaining at an “Extreme” grade.

In contrast, the CI values of the other five regions are greater than 0. The CI values of Taihu Lake in 2030 under SSP1-2.6 and SSP5-8.5 are 0.769 and 0.777, respectively, indicating that the negative impacts of climate change on GWFs increase, and the GI values are at “Extreme” grade. The CI values for Dongting Lake and Mainstream in 2030 fall within the (0.5, 0.75] range under the two SSPs, suggesting an increasing impact of climate change on GWFs, with the GI remaining at “High” grade.

Poyang Lake’s CI values are 0.253 and 0.544 under the two SSPs, respectively, indicating a significant increase in the impacts of climate change on GWFs as radiative forcing rises, with the GI grade rising from “Moderate” to “High.” The Jialing River’s CI values are 0.503 and 0.817 under the two SSPs, respectively, reflecting a significant increase in the impacts of climate change on GWFs, with the GI grade rising from “High” to “Extreme.”

In summary, under the two climate scenarios, SSP1-2.6 and SSP5-8.5, the largest decrease in GI due to climate change in the eight regions of the YRB is in the Wujiang River in 2030, while the largest increase in GI due to climate change in the two SSPs is in the Taihu Lake and the Jialing River, respectively.

4. Discussion

4.1. Impacts of Climate Change on Total Nitrogen Load

Compared to 2020, L_TN values across the eight regions of the YRB range from −19.8% to −0.9% under the SSP1-2.6 climate scenario. In contrast, under the SSP5-8.5 scenario, the range widens to −24.2% to 1.1%. This difference is attributed to greater spatial variation in precipitation under SSP5-8.5 compared to SSP1-2.6. Projected precipitation across the YRB in 2030 ranges from 347 mm to 2216 mm (mean: 1071 mm) under SSP1-2.6, and from 280 mm to 2038 mm (mean: 1031 mm) under SSP5-8.5.

Wang et al. [32], based on the SWAT hydrological model, found that the TN loads in the Fuhe River basin, a major tributary of Poyang Lake, decreased significantly from 2021 to 2030 under the SSP1-2.6 and SSP5-8.5 climate scenarios. Furthermore, studies by Chen et al. [33] and Hu [34], based on the InVEST model, indicate that nitrogen loads in the Hanjiang River Basin decline under the SSP1-2.6 future climate scenario. In the current study, the L_TN values of Poyang Lake in 2020 and under SSP1-2.6 and SSP5-8.5 scenarios in 2030 are 89.43, 88.60, and 85.43 kilotons, respectively. And the L_TN of the Hanjiang River decreases from 78.44 kilotons in 2020 to 63.91 kilotons by 2030 under the SSP1-2.6 scenario. In summary, the findings from the aforementioned literature align with the predictions of a declining trend in L_TN in this study for the future.

4.2. Impacts of Climate Change on Water Resources

Compared with 2020, under the SSP1-2.6 scenario, the W values of the eight regions of YRB range from −32.7% to −2.2%. Under the SSP5-8.5 scenario, the variation range of W value in the eight regions of YRB is expanded to −40.4% to −1.2%. This difference primarily stems from lower precipitation and higher evapotranspiration in the YRB under SSP5-8.5 compared to SSP1-2.6, resulting in generally lower W values and more pronounced spatial variations. Specifically, the YRB’s average precipitation under SSP1-2.6 and SSP5-8.5 scenarios in 2030 is projected at 1071 mm and 1031 mm, respectively, while average evapotranspiration is projected at 3979 mm and 4388 mm.

According to Jiang et al. [35], the runoff in the Min-Tuo River Basin shows a downward trend from 2017 to 2050 under the SSP1-2.6 and SSP5-8.5 scenarios. Research by Wu et al. [36] indicated that the runoff in the Jialing River Basin in 2030 under the SSP5-8.5 scenario is lower than that in 2020. In addition, Zhou et al. [37] found that the total annual runoff depth in parts of the YRB would decrease under the SSP5-8.5 scenario. In the current study, the W values of YRB in 2030 under the SSP1-2.6 and SSP5-8.5 scenarios are lower than those in 2020. In particular, the W values of the Minjiang River in 2030 are 94.07 and 96.48 billion m³ under the SSP1-2.6 and SSP5-8.5 scenarios, respectively, both lower than that of 2020 (96.60 billion m³), and the W values of the Jialing River are 73.87 and 66.64 billion m³, respectively, both lower than that of 2020 (83.67 billion m³). Therefore, the W results simulated in the present study are consistent with the findings of the aforementioned literature.

4.3. Correlation Analysis Between W and L_TN

Figure 10 presents the Pearson correlation analysis between W and L_TN of the YRB in the baseline year (2020) and under two climate scenarios (SSP1-2.6 and SSP5-8.5) in 2030.

In 2020, W and L_TN show a 98.6% synchronicity in variation, confirming the “hydrology-driven pollution” mechanism: increased precipitation in wet years enhances surface runoff, mobilizing more agricultural non-point nitrogen into water bodies. For instance, in 2020, Dongting Lake’s W value (235.4 billion m³) and L_TN (151.2 kilotons) both ranked second in the overall basin, illustrating the amplifying effect of precipitation on pollution.

For 2030, under both SSP1-2.6 (r = 0.988) and SSP5-8.5 (r = 0.988) scenarios, W and L_TN maintain a strong positive correlation. Compared to 2020, two typical trends emerge: (i) In the Hanjiang River, under SSP5-8.5, W increases by 1.2%, with L_TN correspondingly rising by 0.6%, demonstrating the enhanced role of precipitation in pollution transport. (ii) In the Dongting Lake, under SSP5-8.5, W decreases by 40.4%, while L_TN only drops by 23.5%, showing that water resource reduction does not proportionally decrease pollution, resulting in a significant increase in the GI (from 0.86 to 1.10).

4.4. Impacts of Climate Change on Grey Water Footprint of Total Nitrogen

This study employs the GI to indicate the pollution level of agricultural TN. A key finding emerges: under both SSP1-2.6 and SSP5-8.5 scenarios, the GWF of TN in the overall YRB in 2030 is lower than that in 2020, but the GI of TN is higher than that in 2020, with a greater increase under SSP5-8.5. This paradox of declining GWF but rising pollution pressure (GI) highlights the crucial role of W in diluting pollutants and how climate change amplifies water pollution pressure by altering the hydrological cycle.

This paradox stems from water resources declining faster than the pollution load. As shown in Section 3.1, agricultural L_TN in the YRB decreases by approximately 10% under both scenarios by 2030 compared to 2020, primarily due to reduced surface runoff from projected lower rainfall. However, W declines more sharply (12.5% under SSP1-2.6 and 18.6% under SSP5-8.5), driven by combined effects of reduced precipitation and increased evapotranspiration. According to the GI formula (GI = GWF/W), a faster decline in the denominator (W) than the numerator (GWF) increases GI. Consequently, future climate-induced water scarcity fully offsets potential water quality improvements from reduced nitrogen loading and actually intensifies pollution. This highlights that in water-stressed regions, even controlled pollutant emissions can heighten water quality degradation risks due to significantly reduced dilution capacity.

Regionally, the Jialing River basin’s GI goes from “High” in 2020 to “Extreme” in 2030 under SSP5-8.5, showing regional differences in sensitivity to climate-induced “hydrological pollution” coupling changes. Furthermore, although both L_TN and W decrease in the Taihu Lake basin, W drops much more, making it the area with the most pollution pressure. These hotspots, most sensitive to the negative effects of climate change-induced water resource reduction, should be prioritized in future water resource management and agricultural non-point source pollution control.

The CI values for the YRB in 2030 are 0.507 (SSP1-2.6) and 0.527 (SSP5-8.5). CI > 0 unequivocally indicates that climate change will negatively affect the GWF of agricultural TN in 2030 compared to 2020, heightening threats to YRB water security. Crucially, the higher CI under SSP5-8.5 confirms that stronger climate forcing (e.g., precipitation decline and evapotranspiration increase) magnifies negative water pollution. Regional CI analysis pinpoints the most affected hotspots: Taihu Lake (CI = 0.769) under SSP1-2.6 and Jialing River (CI = 0.817) under SSP5-8.5. The CI index effectively distills climate change’s complex influence on water pollution into a comparable metric, providing critical guidance for prioritizing regional risk management.

4.5. Limitations of the Assessment Framework

The SSP–InVEST coupling framework constructed in this study systematically assesses the impact of climate change (primarily through precipitation and evapotranspiration) on the agricultural GWF of TN in the YRB. However, the assessment has the following limitations.

First, simplifications in model processes may compromise the representation of complete nitrogen transport pathways. The NDR model primarily simulates nitrogen transport via surface runoff and subsurface flow but does not account for the potential leaching of nitrate into deep aquifers. Additionally, the annual-scale AWY model, based on the Budyko water balance theory, overlooks the contribution of groundwater–surface water exchange to baseflow. Future studies could integrate a groundwater model or incorporate soil profile nitrogen transport modules.

Second, simplified scenario assumptions present another limitation. The core drivers of the framework are precipitation and evapotranspiration data, assuming that land use (LULC) and agricultural management practices (e.g., fertilizer application rates, methods, and crop varieties) in the future year remain fixed from the baseline year. Future work could incorporate dynamic LULC change scenarios.

Furthermore, the study focuses solely on agricultural non-point source nitrogen loads and associated GWF, excluding industrial point sources. Industrial development varies substantially across SSP pathways. In regions where reduced precipitation diminishes water dilution capacity, such as the Taihu Lake Basin under SSP5-8.5, the combined effects of point and non-point sources could heighten water quality risks. Critically, excluding industrial sources may lead to underestimation of GWF pressure in highly industrialized areas under the SSP5-8.5 pathway. Future studies could integrate industrial and domestic wastewater discharge data to quantify the synergistic effects of multiple pollution sources (point + non-point) and their comprehensive impact on GWF under climate change.

5. Conclusions

This study couples the InVEST model with SSPs to construct an assessment framework for analyzing the climate change impacts on the GWF of agricultural TN, applied to the YRB. Based on simulation results for 2030 under SSP1-2.6 and SSP5-8.5 scenarios, the key findings are as follows:

(i)

Both agricultural L_TN and W show downward trends in response to climate change. Compared to the baseline year (2020), agricultural L_TN values in the YRB in 2030 under SSP1-2.6 and SSP5-8.5 decrease by 10.1% and 9.9%, respectively (from 800.32 kilotons to 719.77 and 721.04 kilotons), primarily due to reduced precipitation. During the same period, W decreases more significantly: under SSP1-2.6, it drops by 12.5% (from 1148.91 billion m³ to 1005.00 billion m³), and under SSP5-8.5, it drops by 18.6% (to 934.72 billion m³), driven by combined effects of reduced precipitation and increased evapotranspiration.

(ii)

Although the GWF decreases, water pollution pressure intensifies. Under both scenarios, the GWF of TN drops (SSP1-2.6: from 1067.1 billion m³ to 959.7 billion m³; SSP5-8.5: from 1067.1 billion m³ to 961.4 billion m³). However, GI rises significantly: under SSP1-2.6, GI increases from 0.93 to 0.95, and under SSP5-8.5, it rises to 1.03 at “High” grade. This shows that the reduction in W fully offsets the positive effect of the decreased L_TN.

(iii)

Regional heterogeneity reveals pollution hotspots. Under SSP5-8.5, GI increases by 11%, 28%, 18%, 27%, and 5% in five regions (Taihu Lake, Dongting Lake, Poyang Lake, Jialing River, and the Mainstream, respectively). Taihu Lake exhibits the highest GI (1.68), making it the most severe pollution hotspot in the basin under this scenario.

(iv)

The adverse impact of climate change on GWF is quantified. The CI reaches 0.507 under SSP1-2.6 and 0.527 under SSP5-8.5, indicating growing negative effects of climate change on GWF. The high-emission scenario exerts stronger adverse effects than the sustainable pathway.

The SSP–InVEST coupling framework demonstrates that future management of agricultural nitrogen pollution in the YRB must prioritize the interplay with water scarcity. Although emission reduction measures can reduce TN load, if climate adaptation strategies (such as promoting water-saving irrigation and enhancing water source conservation) are not strengthened, the depletion of water resources will lead to continuous deterioration of water quality. This framework offers strategic insights for sustainable water resource management in global agricultural basins under climate change.