3.2.1. Spatial Variation in Nutrients

As an important component in water pollution, SWAT simulates the transformation and transportation processes of N and P nutrients. Figure 4 shows the simulated spatial distribution of nutrient (TN and TP) leaching loads from each subbasin and the concentration in each river reach. The nutrients in Figure 4a,c are illustrated in terms of load per unit area (kg/ha) to make the value comparable among different subbasins with distinct

total areas. The spatial distribution of non-point-source pollution strongly depends on the heterogeneities of soil properties, crop types, vegetation types, farming activities, etc. [46]. The TN load is mainly concentrated in the mountainous cultivated field of the upstream areas of the Bai River, such as subbasins 3 (22 kg/ha·y), 5–8, whereas TN load to the Tang River is lower, ranging from 2.9 kg/ha·y (subbasin 44) to 11.8 kg/ha·month (subbasin 39). The intensity of TP load is about half that of TN load. Again, subbasin 3 leached the most TP into the river, at a rate of 10 kg/ha per year. In addition to the upper Bai River regions, TP load is also high in the southeast TRB, including the upper Tang River regions and the other tributary of the Tangbai River, the Gun River in Hubei Province. In the middle flat land, only small amounts of N and P loads are observed. This may be attributed to (1) the low altitude and (2) the occupation of shajiang black soil (Figure 2b), which belongs to a clayey soil (40–54% of clay among four layers), thereby retaining nutrients much longer and eroding less [47].

The mean monthly TN and TP concentrations at the watershed outlet are 1.9 mg/L and 1.1 mg/L, respectively. Regarding the variation along river reaches (Figure 4b,d), none have mean TN or TP concentration that met the Class III standard (i.e., ≤1.0 mg/L for TN and ≤0.2 mg/L for TP). Only 441 km (30.5%) of the river total length had TN concentrations belonging to Classes IV–V. The Bai River is more polluted with N than the Tang River. Except for reaches flowing through the northern mountainous forests, all reaches of the Bai River have TN concentrations inferior to Class V (Figure 4b). The Tang River, in contrast to the Bai River, has TN concentration in the mainstem within Class V and is only inferior to Class V in three tributaries. The magnitude of TP concentration on average is lower than that of TN, but its quality class is significantly worse. Only 18 km of river reaches are of Classes I–V. The remaining 98.8% of the rivers have TP concentrations that are inferior to Class V (Figure 4d).

Table 3 lists the nutrient inputs through fertilization and leaching for different land covers. The average N application for the whole TRB is estimated to be 114.3 kg/ha·y. The highest N fertilization occurs in hay land (HAY), where auto-fertilization is applied based on nitrogen stress and heat units (Section 2.2.2). The elemental N and P fertilizers applied to generic dryland (AGRL) are large because of chemical fertilizer application, as shown in Table 2. The three land covers with the highest TN leaching per unit area are barren land (BARR), dryland AGRL, and pasture (PAST), corresponding to 14, 10, and 9.7 kg/ha·y, respectively. The highest TP leaching is observed in agricultural fields with values of 8.4 and 7.1 kg/ha·y for RICE and AGRL, respectively. The nutrient loads from fields are dominated by organic N (ORGN in Equation 2, 92% and 86% of TN for AGRL and RICE, not shown) and sediment mineral P (SEDP in Equation 3, 68% and 77% of TP, not shown). SWAT simulates three forms of organic N/P: active organic N/P, stable organic N/P associated with humic substances, and fresh organic N/P associated with plant residues [48]. Such predominant organic N and sediment P may partly due to that (1) SWAT tends to overestimate organic N but underestimate dissolved N [49,50], and (2) the accuracy of organic N and sediment P relies on how sediment is simulated [49,51]. The SWAT model has been found to tend to overestimate organic N but underestimate dissolved N [50]. The high nutrient leaching from fields is due to the loosening surface, ponding fresh water on the soil surface, harvesting activities, and rich content of nutrients in the fields; thus, sediment and nutrients can be easily washed away or infiltrated into streams during heavy rainfall. Organic N accounts for 100% of TN losses in pasture (PAST), barren land (BARR), and forest (FRST), which is caused by rain erosion. Forests leach much less nutrients than other land covers, with 2.7 kg/ha·y for TN and 0.9 kg/ha·y for TP, owing to their complex vegetation coverage and interception effects of higher leaf area. This difference in nutrient loss is consistent with previous studies [52,53].

The annual mean loads of N and P leaching from HRUs to stream for the whole TRB are 18.6 and 12.4 kiloton, respectively (see Table 3), whereas the corresponding loads flowing through the watershed outlet (reach 56) are slightly lower at 16.2 (13% less) and 11.6 (6.5% less) kiloton (see Figure 5). When area factions of different land covers are **(km2)**

**Area (%)**

considered, AGRL land is the predominant source of nutrient pollution in the Tangbai River, constituting 79.7% of TN and 85.2% of TP. Such serious non-point-source pollution should be controlled by strengthening dryland management and emphasizing conservation tillage to reduce sediment and nutrient losses. URLD BERM 1638.0 6.8 121.2 ‐ 5.6 1.5 917.0 252.5 RICE RICE 1083.2 4.5 76.2 49.3 6.3 8.4 680.8 913.5 WATR ‐ 657.6 2.7 ‐ ‐ ‐ ‐ ‐ ‐ PAST Panicum 567.3 2.3 ‐ ‐ 9.7 3.1 550.1 173.8

**Pfert (kg/ha∙y)**

**Table 3.** The type of plant, area and area ratio, elemental nitrogen, and phosphorus fertilizers ap‐ plied through calendar fertilization and auto‐fertilization in HRU (Nfert, Pfert, see Section 2.2.2), and total N and total P transport from a certain HRU into reach (TN, TP) in load and load per unit

> **TN (kg/ha∙y)**

**TP (kg/ha∙y)**

**TN (ton∙y)**

**TP (ton∙y)**

**Table 3.** The type of plant, area and area ratio, elemental nitrogen, and phosphorus fertilizers applied through calendar fertilization and auto-fertilization in HRU (Nfert, Pfert, see Section 2.2.2), and total N and total P transport from a certain HRU into reach (TN, TP) in load and load per unit area per year for individual land-use/land-cover (LULC) classes. HAY Hay 392.0 1.6 321.6 12.5 3.0 1.5 118.3 60.4 URBN BERM 221.3 0.9 226.2 ‐ 6.6 0.3 147.1 7.5 ORCD ORCD 142.7 0.6 2.9 ‐ 4.1 1.1 57.9 15.0 UIDU BERM 83.2 0.3 129.2 ‐ 6.0 1.5 49.6 12.5


age to reduce sediment and nutrient losses.

*Water* **2022**, *14*, x FOR PEER REVIEW 13 of 21

area per year for individual land‐use/land‐cover (LULC) classes.

AGRL WWHT/CORN 14,913.9 61.6 153.9 69.6 10.0 7.1 14,850.9 10,566.6 FRST FRST 4462.2 18.4 ‐ ‐ 2.7 0.9 1223.9 387.2

**Nfert (kg/ha∙y)**

**Figure 5.** Seasonal mean TN and TP loads transported with water out of the individual river reaches averaged over 2000–2019. Reaches numbered before 53 belong to the Bai River, and those numbered 17–54 belong to the Tang River. Reach 57 is the downstream tributary called the Gun River, and 56 is the watershed outlet of TRB. **Figure 5.** Seasonal mean TN and TP loads transported with water out of the individual river reaches averaged over 2000–2019. Reaches numbered before 53 belong to the Bai River, and those numbered 17–54 belong to the Tang River. Reach 57 is the downstream tributary called the Gun River, and 56 is the watershed outlet of TRB.

## 3.2.2. Seasonal Variation in Nutrients

3.2.2. Seasonal Variation in Nutrients We divided the year into three seasons according to the monthly variation in precip‐ itation: the non‐flood season (from October to March), spring flood season (April and May), and main flood season (June to September). The respective mean seasonal total pre‐ cipitation from 2000‒2019 are 174.9 mm (21% of the mean annual), 150.6 mm (18%), and 497.4 mm (61%). Figure 5 shows the TN and TP loads at each representative reach for the three seasons. The nutrient loads are overwhelmingly dominant in the main flood season, We divided the year into three seasons according to the monthly variation in precipitation: the non-flood season (from October to March), spring flood season (April and May), and main flood season (June to September). The respective mean seasonal total precipitation from 2000–2019 are 174.9 mm (21% of the mean annual), 150.6 mm (18%), and 497.4 mm (61%). Figure 5 shows the TN and TP loads at each representative reach for the three seasons. The nutrient loads are overwhelmingly dominant in the main flood season, constituting on average 78% of the annual TN load and 84% of the TP load. Using the watershed outlet (reach 56) as an example, the TN loads in spring flood season and main flood season are approximately 3.5 and 9 times that in the non-flood season, respectively; the TP loads are 3.3 and 16 times that in the non-flood season, respectively. The simulated low nutrient pollution in non-flood seasons is because the pollutants gradually accumulate

on the land surface when there is little or no rain, regardless of the point-source pollution emissions. The spring flood season is the first period of concentrated rainfall after the dry winter; thus, it allows the pollutants accumulated in the soil and floating in the air to be washed away into the stream. the dry winter; thus, it allows the pollutants accumulated in the soil and floating in the air to be washed away into the stream. Comparing the loads of different reaches, a clear increasing trend of nutrients is ob‐ served from upstream to downstream, suggesting relatively lower rate of pollutant deg‐

constituting on average 78% of the annual TN load and 84% of the TP load. Using the watershed outlet (reach 56) as an example, the TN loads in spring flood season and main flood season are approximately 3.5 and 9 times that in the non‐flood season, respectively; the TP loads are 3.3 and 16 times that in the non‐flood season, respectively. The simulated low nutrient pollution in non‐flood seasons is because the pollutants gradually accumu‐ late on the land surface when there is little or no rain, regardless of the point‐source pol‐ lution emissions. The spring flood season is the first period of concentrated rainfall after

Comparing the loads of different reaches, a clear increasing trend of nutrients is observed from upstream to downstream, suggesting relatively lower rate of pollutant degradation than accumulation. For example, except for the two reaches 28 and 35, which are tributaries of the Bai River (Figure 4b), TN and TP loads increase continuously along reaches 3 to 53. The changes among reaches 40, 45, and 53 are limited because the leaching losses from the respective subbasins are low (Figure 4a,c). The TN load in the Bai River (reach 53) is around 1.5 times that of the Tang River (reach 54), accounting for around 53.4% and 35.3% of total nutrient loads at the outlet reach 56. The TP loads in the Tang River and the lower reaches of the Tangbai River contribute more to that in reach 56, as a result of increasing phosphorus loss from paddy fields (RICE, Table 3) which are distributed mainly in the southern part of TRB (Figure 2a). radation than accumulation. For example, except for the two reaches 28 and 35, which are tributaries of the Bai River (Figure 4b), TN and TP loads increase continuously along reaches 3 to 53. The changes among reaches 40, 45, and 53 are limited because the leaching losses from the respective subbasins are low (Figure 4a,c). The TN load in the Bai River (reach 53) is around 1.5 times that of the Tang River (reach 54), accounting for around 53.4% and 35.3% of total nutrient loads at the outlet reach 56. The TP loads in the Tang River and the lower reaches of the Tangbai River contribute more to that in reach 56, as a result of increasing phosphorus loss from paddy fields (RICE, Table 3) which are distrib‐ uted mainly in the southern part of TRB (Figure 2a).

#### 3.2.3. Nutrients in the Trans-Provincial Key Sections 3.2.3. Nutrients in the Trans‐Provincial Key Sections

*Water* **2022**, *14*, x FOR PEER REVIEW 14 of 21

The trans-provincial analysis focuses on three controlled sections, including the outlet of TRB (Zhangwan section on reach 56) and the junctions of Henan and Hubei province (Diwan section and Bukou section on reaches 45 and 46). Figure 6 illustrates the monthly distribution of nutrient concentrations in these sections. Both the multi-year mean and median results are shown to eliminate the disturbance from large outliers. The trans‐provincial analysis focuses on three controlled sections, including the out‐ let of TRB (Zhangwan section on reach 56) and the junctions of Henan and Hubei province (Diwan section and Bukou section on reaches 45 and 46). Figure 6 illustrates the monthly distribution of nutrient concentrations in these sections. Both the multi‐year mean and median results are shown to eliminate the disturbance from large outliers.

**Figure 6.** Mean and median values of monthly TN (**a**,**c**) and TP (**b**,**d**) concentrations in reaches 45 (Diwan section) on the Bai River, 46 (Bukou section) on the Tang River, and 56 (Zhangwan section) of the watershed outlet. The dashed horizontal lines are the respective concentrations of water qual‐ ity Class III. The three curves in (**c**,**d**) are SWAT‐simulated mean monthly streamflow for the indi‐ vidual reaches. Comp(I–III) in the legend is the respective compliance rate of Classes I–III water in the period 2000–2019. **Figure 6.** Mean and median values of monthly TN (**a**,**c**) and TP (**b**,**d**) concentrations in reaches 45 (Diwan section) on the Bai River, 46 (Bukou section) on the Tang River, and 56 (Zhangwan section) of the watershed outlet. The dashed horizontal lines are the respective concentrations of water quality Class III. The three curves in (**c**,**d**) are SWAT-simulated mean monthly streamflow for the individual reaches. Comp(I–III) in the legend is the respective compliance rate of Classes I–III water in the period 2000–2019.

Nutrient pollution is concentrated during the flood season (April to September): it increases sharply in April, decreases in June, peaks in July, and then gradually declines with decreasing streamflow. In accordance with Figure 5, the nutrient concentration is high in the spring flood season (April and May) despite the small proportion of annual rainfall and streamflow. The N and P leaching losses under heavy rainfall during the spring flood season far exceed those under weak rainfall during the non-flood season, which is consistent with previous findings [54]. This in turn causes the nutrient concentration to exceed the water quality standard. For the Bai River, the mean TN concentration in the spring flood season can be even higher than that in the main flood season. Interestingly, we find the mean TN (Figure 6a) to be much larger than the median value (Figure 6c), while the difference is insignificant in the remaining months. This indicates that the extremely high TN concentration on the Bai River can occur during spring flood seasons. In contrast, the TP concentration in the spring flood season is only about half of the peak value (in July). This is because paddy fields (RICE), which are the main source of P pollutants (Table 3), lay fallow in the non-flood season (Table 2); therefore, phosphorus accumulation is limited in winter, and the amount of P flushed out from paddy fields is also not high during the spring flood season.

Considering the water quality standard, the Class III water quality compliance rate for monthly TN concentration over 2000–2019 is around 40%, and for TP, it is only 30%. The higher compliance rate of TN than TP is consistent with the spatial distribution results in Figure 4. On average, only months in the non-flood season (October to March) may reach Class III or Class V water. The TN pollution in the Bai River is more serious than that in the Tang River; but the opposite is true for TN pollution.

#### *3.3. Effect and Quantification of Eco-Compensation*

Stretching across Henan province (Nanyang city) and Hubei province (Xiangyang city), TRB has a high proportion of traditional agriculture. Nanyang has 86.8% of the total agricultural fields of TRB, while Xiangyang has only 13.2%. The boundary between wheat–corn fields and rice fields almost coincides with the provincial boundary between Henan and Hubei provinces. Nanyang city has been accustomed to growing wheat and corn for thousands of years for geographical reasons. The major soil types in Nanyang are shajiang black soil (SJHT) and yellow-cinnamon soil (HHT, Figure 2b). Both are typical low-yield soils due to their low organic matter content and poor soil structure [55,56]. The wide extent of dryland farming in Henan province, along with intensive human activities, has accelerated soil erosion, leading to serious nutrient losses.

To realize the simultaneous growth of the economy and restoration of water ecosystems in Nanyang city, increasing crop yield, for example, through implementing more scientific crop management (e.g., agricultural industrial agglomeration) and a higher portion of organic fertilizer [57], is necessary but not sufficient. It is an inevitable trend to abandon part of drylands, and transform and upgrade the local industrial structure. This section predicts the effect of converting dryland to forest on water quality in streams across the provincial boundary. Note that we assume that dryland fields (AGRL) in Hubei province and all paddy fields (RICE) remain unchanged because (1) this study investigates trans-boundary eco-compensation, the logic of which is the downstream Hubei province may benefit from the ecological contribution of Nanyang in the upstream and, in turn, compensate for this contribution; (2) the soil type in the paddy fields is mostly paddy soil (SDT), which is one of the three high-yield soils in China; thus the ecological benefits of retiring paddy fields are likely to be much fewer than the economic benefits of retaining farming.

Figure 7 shows the changes in pollutants and reference eco-compensation values at trans-provincial sections under different 53 Grain for Green (GFG) settings (see Section 2.3). Overall, a clearly more noticeable nutrient load decrease is observed compared to the TN/TP compliance rate increase with the increase in GFG area. For instance, up to 60% area achieves only limited improvement (within 10%) for the TP compliance rate while the TP load is reduced by more than 50% (Figure 7b,d). Under the same horizontal axis, there is a disparity in results because the HRUs for GFG are randomly selected, either in the upper regions of the Tang River or Bai River. However, for the whole basin, the different selections in HRUs bring only limited difference in the nutrient pollution (i.e., thin ribbons for the basin outlet shown in Figure 7e). If all dryland fields in TRB in Henan province could be converted to forest (i.e., 100% GFG area), the water compliance

rate in the Diwan and Bukou sections would increase by 32%–39%. The months with the worst water quality are mainly concentrated in Spring (March to May, Figure 8e,f). However, the improvement in TN at the outlet section is stronger than that in TP, with maximum increases in the compliance rate in TN of 29.8% and in TP of only 13.2%. Under the 100% GFG, the further accumulation of TN and TP loads in the downstream Hubei province is responsible for 38% and 62% of the N and P transported out of the basin, respectively. This implies that TP pollution is more extensive in Hubei and therefore cannot be controlled by eco-protection measures in Henan province only. province could be converted to forest (i.e., 100% GFG area), the water compliance rate in the Diwan and Bukou sections would increase by 32%–39%. The months with the worst water quality are mainly concentrated in Spring (March to May, Figure 8e,f). However, the improvement in TN at the outlet section is stronger than that in TP, with maximum increases in the compliance rate in TN of 29.8% and in TP of only 13.2%. Under the 100% GFG, the further accumulation of TN and TP loads in the downstream Hubei province is responsible for 38% and 62% of the N and P transported out of the basin, respectively. This implies that TP pollution is more extensive in Hubei and therefore cannot be con‐ trolled by eco‐protection measures in Henan province only.

*Water* **2022**, *14*, x FOR PEER REVIEW 16 of 21

the TP load is reduced by more than 50% (Figure 7b,d). Under the same horizontal axis, there is a disparity in results because the HRUs for GFG are randomly selected, either in the upper regions of the Tang River or Bai River. However, for the whole basin, the dif‐ ferent selections in HRUs bring only limited difference in the nutrient pollution (i.e., thin ribbons for the basin outlet shown in Figure 7e). If all dryland fields in TRB in Henan

**Figure 7.** Effects of converting drylands (AGRL) in Henan to forest in terms of TN and TP leaching into reaches from upstream subbasins and compliance rate of water quality (Class III water) in the individual controlled sections (**a**–**e**), and the value of eco‐compensations in CNY based on oppor‐ tunity cost (upper line, **f**) and GFG compensation standard (lower line, **f**). The ribbons are the range and means of the respective results for GFG based on certain area ratios, and the scatter diamonds and stars are results based on slopes. **Figure 7.** Effects of converting drylands (AGRL) in Henan to forest in terms of TN and TP leaching into reaches from upstream subbasins and compliance rate of water quality (Class III water) in the individual controlled sections (**a**–**e**), and the value of eco-compensations in CNY based on opportunity cost (upper line, **f**) and GFG compensation standard (lower line, **f**). The ribbons are the range and means of the respective results for GFG based on certain area ratios, and the scatter diamonds and stars are results based on slopes. *Water* **2022**, *14*, x FOR PEER REVIEW 17 of 21

**Figure 8.** Comparison of mean monthly TN (**a**,**c**,**e**) and TP (**b**,**d**,**f**) concentrations in pollutant control sections between three different sets of GFG scenarios: converting all non‐flat drylands (AGRL with slope > 2.5%) in Henan province to FRST (**a**,**b**), one scenario of randomly selecting 20% AGRL to FRST based on HRUs (**c**,**d**), and converting all AGRL land to FRST (**e**,**f**). **Figure 8.** Comparison of mean monthly TN (**a**,**c**,**e**) and TP (**b**,**d**,**f**) concentrations in pollutant control sections between three different sets of GFG scenarios: converting all non-flat drylands (AGRL with slope > 2.5%) in Henan province to FRST (**a**,**b**), one scenario of randomly selecting 20% AGRL to FRST based on HRUs (**c**,**d**), and converting all AGRL land to FRST (**e**,**f**).

As sloping lands are more susceptible to soil erosion, the scatter dots in Figure 7 de‐ pict the two slope‐specific GFG scenarios. It is evident that GFG of sloping drylands (slope > 15°, area = 1.5%) and non‐flat drylands (slope > 2°, area = 21.8%) are more effective in

21.8% non‐flat dryland GFG is 3.5 times the mean reduction under the 20% slope‐inde‐ pendent GFG; however, water quality compliance rates in this section are not much dif‐ ferent from each other. This inconsistency is because although GFG of sloping/non‐flat drylands reduces the pollutant load considerably, mainly in the main flood season (June to September), TN or TP concentrations during this period are still worse than Class III

To summarize, if the water quality standard is the only benchmark, then improving the nutrient concentration at the provincial boundary sections requires control of both the amount of nutrients being washed out and the timing of their washing out. Thus, enhanc‐ ing the monitoring and management of water quality, along with the ecological operation of water projects, are the key engineering measures that allow water bodies to self‐purify

The payments of eco‐compensation are calculated based on the GFG area and crop yields (Figure 7f). In the baseline scenario without any GFG, the mean annual yields of maize and wheat in TRB within Henan province are 3839.1 and 3740.7 kiloton, respec‐ tively. The sum of this is close to, but slightly higher than, the statistic annual grain pro‐ duction of 7105.9 kiloton in Nanyang city in 2019 [58]. The total average annual value of both crops simulated by SWAT is CNY 18.4 billion. The compensation amount consider‐ ing agricultural opportunity costs increases almost linearly with the increase in GFG area, from CNY 1.81 billion (for 10% GFG) to CNY 18.4 billion (for 100% GFG) (upper line in Figure 7f) per year. When compensation is based on merely the GFG area, the value also increases proportionally, but grows more slowly (lower line in Figure 7f). Under the 100% GFG scenario, the lower value is only one‐third of the upper value. This indicates that even under the same ecological restoration strategy, the amount of eco‐compensation available to farmers may vary considerably due to the different quantitative criteria

(Figures 8a,b and 6a,b).

more effectively.

adopted.

As sloping lands are more susceptible to soil erosion, the scatter dots in Figure 7 depict the two slope-specific GFG scenarios. It is evident that GFG of sloping drylands (slope > 15◦ , area = 1.5%) and non-flat drylands (slope > 2◦ , area = 21.8%) are more effective in controlling nutrient loss than slope-independent GFG scenarios. Using the TN load at the Diwan section (on the Bai River, Figure 7a) as an example, the TN reduction under the 21.8% non-flat dryland GFG is 3.5 times the mean reduction under the 20% slopeindependent GFG; however, water quality compliance rates in this section are not much different from each other. This inconsistency is because although GFG of sloping/non-flat drylands reduces the pollutant load considerably, mainly in the main flood season (June to September), TN or TP concentrations during this period are still worse than Class III (Figure 8a,b and Figure 6a,b).

To summarize, if the water quality standard is the only benchmark, then improving the nutrient concentration at the provincial boundary sections requires control of both the amount of nutrients being washed out and the timing of their washing out. Thus, enhancing the monitoring and management of water quality, along with the ecological operation of water projects, are the key engineering measures that allow water bodies to self-purify more effectively.

The payments of eco-compensation are calculated based on the GFG area and crop yields (Figure 7f). In the baseline scenario without any GFG, the mean annual yields of maize and wheat in TRB within Henan province are 3839.1 and 3740.7 kiloton, respectively. The sum of this is close to, but slightly higher than, the statistic annual grain production of 7105.9 kiloton in Nanyang city in 2019 [58]. The total average annual value of both crops simulated by SWAT is CNY 18.4 billion. The compensation amount considering agricultural opportunity costs increases almost linearly with the increase in GFG area, from CNY 1.81 billion (for 10% GFG) to CNY 18.4 billion (for 100% GFG) (upper line in Figure 7f) per year. When compensation is based on merely the GFG area, the value also increases proportionally, but grows more slowly (lower line in Figure 7f). Under the 100% GFG scenario, the lower value is only one-third of the upper value. This indicates that even under the same ecological restoration strategy, the amount of eco-compensation available to farmers may vary considerably due to the different quantitative criteria adopted.

## *3.4. Implications for Eco-Compensation*

In China, the main factors accounting for eco-compensation include the following three with reference to guidelines in other countries. (1) The opportunity-cost factor: additional consideration must be given to opportunity costs, which are the non-ecological benefits being sacrificed by environmental protectors to protect the environment [59]. (2) The polluter-pays factor: the actors causing the pollution should pay to correct the wrong, thereby limiting their pollution activities and effectively reducing the environmental freeriding behaviors [60]. (3) The beneficiary-pays factor: beneficiaries of ecosystem services should reimburse the upstream providers of water-related environmental services either in full or according to a share of the total [61,62]. Generally, the finance of eco-compensation in China comes from both the government and directly responsible stakeholders [41]. The calculations performed in this study, especially Figure 7f, can provide the bases for estimating opportunity cost. Although the money involved may seem large, it does not reflect the actual amount of funds needed for reforestation but provides reference boundaries of the investment that needs to be paid to the upstream by downstream water users. Local ecological requirements and financial affordability should be considered when accounting for eco-compensation and making decisions.

Funding for watershed protection can be implemented in two ways. The first is paying for ecological projects, such as water project construction, including sewage treatment and ecological restoration plants, and their subsequent maintenance. The second is to directly subsidize the contributors of the affected areas, such as paying out the protection funds to the residents, enterprises, and local governments involved on an annual or quarterly basis to compensate their losses due to changes in crop production and lifestyles. The former project payment can promote regional sustainable development and maintain long-term operation, but fixed, static investments may lack flexibility [63]. In contrast, the latter subsidy, although highly flexible and easily gains trust from local residents, can easily turn into consumption expenditure [64], thereby deviating from the policy objective of agricultural transformation. In this study area, the eco-compensation fund needs to be allocated in a combination of project support and subsidies. That is, to achieve a good water ecosystem status in the downstream TRB, it is necessary not only to carry out integrated management at the watershed level but also to provide cash payments to the upstream government and residents, especially to farmers who return farmland to forest.

The standards of eco-compensation can be determined through negotiation based on project cost analysis and water value assessment [65] or on flexible gambling agreements [66]. The value assessment can be calculated by incorporating both direct cost and opportunity cost, such as the money inputs for pollutant dilution and the lost development opportunities due to ecological protection activities. The gambling agreement is more often utilized for eco-compensation regarding trans-regional water protection, where negotiation is made between upstream and downstream in terms of water quality standards and pollutant thresholds in the controlled river sections. If the water quality at the junctions meets the standards, the co-protection fund should be allocated upstream as compensation for protecting the watershed; conversely, the fund will be given downstream as compensation for purifying water.

## **4. Conclusions**

The Tangbai River Basin (TRB) is a typical trans-provincial watershed in central China that experiences severe non-point-source agricultural pollution. Based on the SWAT model simulations in TRB, the spatial and temporal distributions of nitrogen and phosphorus pollution are described in terms of nutrient load and concentration. Their responses to Grain for Green (GFG) ecological restoration measures are then investigated. With respect to trans-regional eco-compensation, we evaluated the effects of location, area, and the slopedependency of GFG on the water quality along the provincial boundary sections, using TN and TP as indicators. It appears that GFG measures may be more effective at reducing nutrient loads than increasing the water quality compliance rate. The monetary values of the corresponding eco-compensation are quantified based on crop production changes and the GFG area. We found that the compensation amount using opportunity cost can be as high as three times the amount typically paid based on area. This study could provide suggestions for eco-compensation and offer new ideas for controlling non-point-source pollution of trans-boundary rivers.

**Author Contributions:** W.W., H.Z. and Y.L. planned the study and prepared the initial data. W.W. performed computations and data analysis, and wrote the initial draft. H.Z. supervised the writing of the paper. Y.F. coordinated a part of computation. H.Z. and H.F. provided financial support. The revised version was written by W.W., with specific contribution from H.Z. and J.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the National Natural Science Foundation of China, (No. 52179009, 51909035 and U2040206).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The public datasets used for SWAT model building have been properly cited in the main text, while the data of streamflow and nutrient concentration adopted for model validation can be found in https://doi.org/10.6084/m9.figshare.20402019.

**Conflicts of Interest:** The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
