Evaluation of the Synergistic Control Efficiency of Multi-Dimensional Best Management Practices Based on the HYPE Model for Nitrogen and Phosphorus Pollution in Rural Small Watersheds

Abstract

Non-point source pollution (NPS) from agriculture is a primary driver of water eutrophication, necessitating effective control for regional water ecological security and sustainable agricultural development. This study focuses on the Chenzhuang village watershed, a typical green agricultural demonstration area in Jiangsu Province, using the HYPE model to analyze hydrological processes and Total Nitrogen (TN) and Total Phosphorus (TP) migration patterns. The model achieved robust performance, with Nash–Sutcliffe Efficiency (NSE) values exceeding 0.7 for daily runoff and 0.35 for monthly TN and TP simulations, ensuring reliable predictions. A multi-scenario simulation framework evaluated the synergistic control effectiveness of Best Management Practices (BMPs), including agricultural production management, nutrient management, and landscape configuration, on TN and TP pollution. The results showed that crop rotation reduced annual average TN and TP concentrations by 11.8% and 13.6%, respectively, by shortening the fallow period. Substituting 50% of chemical fertilizers with organic fertilizers decreased TN by 50.5% (from 1.92 mg/L to 0.95 mg/L) and TP by 68.2% (from 0.22 mg/L to 0.07 mg/L). Converting 3% of farmland to forest enhanced pollutant interception, reducing TN by 4.14% and TP by 2.78%. The integrated BMP scenario (S13), combining these measures, achieved TN and TP concentrations of 0.63 mg/L and 0.046 mg/L, respectively, meeting Class II surface water standards since 2020. Economic analysis revealed an annual net income increase of approximately 15,000 CNY for a 50-acre plot. This was achieved through cost savings, increased crop value, and policy compensation. These findings validate a “source reduction–process interception” approach, providing a scalable management solution for NPS control in small rural watersheds while balancing environmental and economic benefits.

1. Introduction

Agricultural Non-Point Source Pollution (AGNPS) refers to pollutants entering surface and groundwater bodies in a widespread, dispersed, and trace manner. The process can be primarily summarized as follows: rainfall generates runoff, which leads to soil erosion when it flushes over the land, subsequently causing pollutants in the soil to enter water bodies [1]. According to global monitoring data, approximately 80% of freshwater systems and 50% of terrestrial areas are affected by NPS, with 75% of these regions experiencing combined nitrogen and phosphorus pollution issues [2,3,4]. As China’s agriculture develops and the use of fertilizers and pesticides increases, the proportion of NPS has been rising annually [5], making agricultural NPS the leading driver of degradation in global freshwater systems [6]. At the same time, agricultural NPS leads to the eutrophication of water bodies through sediment transport and nutrient discharge, significantly reducing aquatic biodiversity and thereby impacting the health of entire ecosystems. Therefore, NPS is one of the most significant obstacles to green agricultural development, and its prevention and control pose a challenging issue for agricultural transformation and economic development in developing countries [7]. In conclusion, accurately simulating and analyzing nitrogen and phosphorus loads in small rural watersheds, quantitatively parsing their spatial distribution characteristics, and selecting appropriate management measures to control pollution are of critical importance for watershed water pollution control and sustainable development [8,9].

Currently, both domestic and international research on NPS modeling primarily employs physical models to simulate the generation, migration, and transformation processes of pollutants [10,11]. In response to this need, researchers worldwide have been actively exploring various hydrological and water quality models, extensively applying them to simulate water quality indicators such as TN, TP, and Chemical Oxygen Demand (COD). For instance, Venishetty et al. used the SWAT model to assess water quality pollution at two different geographical scales [12]. Malagó et al. applied the SWAT model to simulate water and nutrient fluxes in the Danube River basin, evaluating the effects of different environmental policy implementations on water quality improvement, thus providing scientific guidance for watershed environmental management in the region [13]. Sun et al. systematically analyzed the distribution characteristics of point and NPS in the MinJiang River basin (MRB) using the SWAT model and optimized relevant management measures by integrating the Response Surface Method (RSM) with the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) [14]. Although the SWAT model is widely used for simulating hydrological processes and pollutant transport, its relatively simplified representation of pollutant transport mechanisms (particularly chemical processes such as phosphorus adsorption–desorption), combined with its daily time-step resolution and design focus on major large watersheds, may limit its accuracy in simulating the rapid hydrological responses and pollutant transport dynamics characteristic of small watersheds [15]. Additionally, the HYPE model, with its more detailed process representation and flexible parameter system, provides an effective tool for the accurate prediction and management of agricultural NPS. The HYPE model is a semi-distributed watershed model that achieves medium to high precision hydrological process simulation through sub-basin division, making it particularly suitable for simulating pollutant migration and transformation. Its distinctive feature lies in the integration of hydrological–water quality–ecological multi-process coupling, a rich nutrient cycling module, and support for multi-source data assimilation, thereby excelling in watershed management decision support. For example, Yin et al. used the HYPE model with multi-site and multi-objective parameter calibration methods to simulate the hydrological and water quality processes in the Hongru River agricultural watershed, analyzing the temporal variations in TN and TP concentrations and loads with crop rotation [16]. Capell, R. et al. used the HYPE model to simulate nutrient changes in the Baltic Sea, systematically assessing the potential impacts of local measures on the entire large basin [17]. This indicates that the HYPE model performs well in simulating and predicting changes in water quality indicators. However, existing applications are mostly concentrated in large basins and marine areas. In contrast, research on the application of the HYPE model at the rural small watershed scale, particularly in typical agricultural regions of China, remains limited, and its applicability has not been fully validated. To address this gap, the present study will establish a tailored HYPE model parameterization framework for agricultural small watersheds, leveraging high-resolution hydrological and water quality monitoring data from a typical rural catchment in Southern China.

BMPs are systematic approaches developed in the United States to address agricultural NPS, encompassing management methodologies, technical measures, and operational protocols [18]. Extensive research has demonstrated BMPs’ effectiveness in controlling sediment transport and nutrient loss in agricultural NPS [19,20,21,22]. These practices are recognized as alternative management strategies that mitigate NPS pollutants by regulating farmland runoff, sediment, and nutrient losses [23]. BMPs are typically categorized into two types: structural and non-structural measures. The latter has gained wider adoption due to its cost-effectiveness and environmental compatibility [24]. Non-structural measures employ a source-control approach that primarily regulates pollutant generation and diffusion at their origin through optimized agricultural production techniques, improved tillage methods, and refined irrigation management systems [25], operating through three hierarchical dimensions: agricultural production management, nutrient management, and landscape configuration management [26]. Agricultural production management and nutrient management primarily mitigate NPS by controlling pollutant dispersion, as demonstrated by Jiang et al., who identified optimized agricultural practices (particularly fertilizer application) as critical for phosphorus reduction and water quality improvement [27], Meanwhile, M. Lee et al. employed the SWAT model to show significant decreases in TN and TP through fertilization reduction in small agricultural watersheds in South Korea [28], and Himanshu et al. utilized SWAT modeling to evaluate different agricultural production management approaches for controlling sediment and nutrient losses in India’s Marol watershed, subsequently proposing the most suitable BMPs for the region [29]. Landscape configuration management enhances water quality by optimizing land-use types and patterns, both controlling non-point source (NPS) pollution at the source and intercepting pollutants during transport. Its practical feasibility has been widely validated and applied in NPS pollution control. For instance, Wang et al. [30] established a scientific basis for water quality improvement in the Jialing River’s Chongqing section, and Lai et al. used the SWAT model to demonstrate significant impacts (p < 0.05) of landscape patterns on nutrient export [31]. Furthermore, Sith et al. integrated SWAT with MODFLOW to show that land-use modifications and BMPs substantially reduced pollution loads in Okinawa’s Todoroki River basin [32]. The “source–sink” landscape theory is the core basis for optimizing land use. This theory classifies landscapes into two categories: “source landscapes” (such as farmlands and residential areas) are prone to generating and discharging pollutants through runoff; “convergence landscapes” (such as forest land and wetlands) retain pollutants through adsorption, filtration and biotransformation. By increasing the spatial proportion of the “convergence landscape” and optimizing its spatial layout, the process interception of non-point source pollution can be achieved. This mechanism has been widely applied in the practice of river basin pollution control [33]. In summary, current studies employ NPS pollution models to quantitatively evaluate pollution reduction under various BMP scenarios, thereby identifying optimal management strategies tailored to specific watersheds and providing scientific recommendations for NPS pollution control. Research has demonstrated that an integrated modeling approach combining NPS models with BMP scenario analysis has emerged as a critical research paradigm in watershed pollution management [34]. However, there remains a gap in applying this model to highly intensive agricultural small watersheds in China, particularly in quantitatively assessing the environmental benefits of transitioning to green agricultural production and studying the synergistic mechanisms with other management optimization measures.

Therefore, this study focuses on the typical rural small watershed of Chenzhuang village in Jiangsu Province, which serves as an important water source conservation area for Lake Tai. In recent years, it has actively promoted and implemented green agricultural production methods, making pollution reduction and prevention significant in both practical value and scientific research. Chenzhuang village was selected in 2014 as an experimental base for rural transformation planning and ecological innovation under the Science and Technology Service Network Initiative (STS) of the Chinese Academy of Sciences, with a focus on promoting green natural farming technology centered on source reduction. This approach avoids the use of pesticides and plastic mulch, instead cultivating crops and producing agricultural products through natural methods such as fostering indigenous microorganisms, extracting and reinjecting plant nutrient solutions, and organically enriching the soil to grow crops and produce agricultural products [35].The ecological transformation experiment in Chenzhuang not only preserved the natural landscape of the village but also significantly improved local water environmental conditions, providing an ideal setting for exploring synergistic pathways between agricultural development and water resource protection. The study will systematically explore the effects of green natural farming techniques on improving water quality in the watershed, with a focus on analyzing the role of agricultural production management and nutrient management in controlling NPS in the watershed. Additionally, the “source–sink” landscape theory will be integrated to quantitatively analyze the synergistic effects of green natural farming techniques and landscape optimization measures, as well as the contribution of landscape configuration to pollution control. In summary, the research specifically focuses on (1) optimizing the agricultural hydrology module to better represent field-scale hydrological processes and (2) refining nitrogen and phosphorus transport–transformation parameters. This systematic parameterization approach aims to provide both methodological support for the precise prediction and management of agricultural NPS and a technical framework for small watershed water quality modeling systems. Ultimately, from the three dimensions of production management, nutrient regulation, and landscape configuration, the optimal non-engineering management plan suitable for the study area is selected, providing scientific evidence and technical support for the efficient governance of agricultural NPS in the watershed and the coordinated advancement of agricultural green development and water ecological protection.

2. Materials and Methods

2.1. Data Sources

The data used in this study include spatial sequence data (digital elevation model (DEM), land-use types, soil types), time series data (temperature, precipitation, runoff, water quality), and statistical data (data related to agricultural activities).

Table 1. Main data sources and information for the study area.

Data Type	Monitoring Stations	Resolution	Time	Data Source
Digital Elevation Model (DEM)	\	30 m	\	Chinese Academy of Sciences Resource and Environmental Sciences Data Center
Land Use	\	30 m	\	National Center for Basic Geographic Information Global Surface Cover Data
Soil Types	\	1000 m	\	Nanjing Institute of Soil Science, Chinese Academy of Sciences
Temperature	\	0.1° × 0.1°	2017–2021	China Meteorological Administration Climate Center
Rainfall	\	0.1° × 0.1°	2017–2021	China Meteorological Administration Climate Center
Traffic	Chenzhuang	\	December 2017–September 2021	Project Testing
Water Quality (TP, TN)	Chenzhuang	\	December 2017–September 2021	Project Testing
Livestock	Chenzhuang	\	December 2017–September 2021	Field Research
Agriculture-related data	Chenzhuang	\	December 2017–September 2021	On-site Research

presents the data sources and resolution, while

Table 2. Planting management information for major crops in the Chenzhuang village watershed (based on field research).

Major Crops	Year	Month	Operation
Rice	1st	April	Planting
	1st	August	Irrigation
	1st	September–October	Harvest
Corn	1st	April	Planting
	1st	June	Top-dressing
	1st	July	Artificial pollination
	1st	August–September	Harvest
Wheat	1st	November	Planting
	1st	December	Fertilization
	2nd	March	Top-dressing
	2nd	May	Irrigation
	2nd	June	Harvest
Beans	1st	June	Planting
	1st	August	Fertilization
	1st	September–October	Harvest
Tuber crops	1st	May	Transplanting seedlings
	1st	July	Fertilization
	1st	October–November	Harvest

lists the crop planting systems and fertilization schemes, including planting schedules, fertilizer application, and tillage practices.

2.2. Research Methodology

2.2.1. Overview of the Study Area

Chenzhuang (Figure 1), located at the southern end of the Maoshan mountain range on the border between Zhenjiang City and Changzhou City, is the origin and watershed of the western Taihu Lake water system and the eastern branch of the Qinhuai River water system. It is situated in the upstream water conservation area of the Taihu Lake basin, making its geographical location unique. The region has intense agricultural activities, with the application of chemical fertilizers dramatically increasing in recent years, leading to a TN exceedance rate of up to 83% in the basin, highlighting significant non-point source nitrogen and phosphorus pollution from agriculture, which poses a serious threat to the eutrophication of downstream waters [36].

However, systematically analyzing and quantifying the effectiveness of green natural farming in watershed pollution control remains a pressing scientific challenge. This study aims to establish an integrated approach, based on Chenzhuang’s practical experience, that synergizes efficient agricultural NPS management with sustainable agricultural development. This study takes Chenzhuang village as a representative case of ecological transition village, focusing on a 50-acre experimental plot within the watershed to systematically investigate the internal mechanisms through which green agricultural practices improve water quality and enhance ecological benefits. The research aims to provide empirical references for pollution control in similar regions while offering crucial theoretical support and practical paradigms for exploring synergistic pathways between agricultural sustainability and ecological conservation.

Water quality sampling points: a total of 10 monitoring points (P1–P10) were established across the watershed), with locations strategically selected to represent the following:

(1)

Key agricultural runoff pathways (P1, P2, P4);

(2)

Proximity to residential areas (P9);

(3)

The watershed outlet connecting to Lita Reservoir (P10).

This design captures spatial heterogeneity in pollution transport and aligns with hydrological connectivity principles [36]. Among these, the monitoring data of five points (P1, P2, P4, P9, and P10) are relatively complete and were used for model validation.

2.2.2. Model Construction

The fundamental prerequisite for hydrological runoff process simulation lies in establishing appropriate sub-watershed delineation, where both the partitioning methodology and its accuracy directly govern the extraction of sub-watershed characteristics, consequently determining the reliability and precision of watershed hydrological simulations [37]. This study employed the digital elevation model (DEM) of the Maoshan Chenzhuang rural watershed, utilizing ArcGIS software with its “ArcHydro Tools” extension module to derive the watershed drainage network and boundaries while maintaining minimal DEM modifications. With Lita Reservoir designated as the watershed outlet, the study area was systematically partitioned into 12 distinct sub-watersheds.

Based on ArcGIS 10.8, 12 sub-watersheds were delineated and 7 soil–land-use combinations (SLC) were determined, as shown in

Table 3. SLC in Chenzhuang village area.

Portfolio	Land Use	Soil Types
1	Cultivated Land	High-activity leaching soil
2	Forest land	High-activity leaching soil
3	Water areas	High-activity leaching soil
4	Residential Land	High-activity leaching soil
5	Cultivated Land	Saturated cohesive soil
6	Forest land	Saturated cohesive soil
7	Water areas	Saturated cohesive soil

. Key parameter calibration was conducted using Parameter EsTimation (PEST) automatic optimization combined with manual adjustments, with NSE as the optimization objective to calibrate critical hydrological and water quality parameters.

Table 4. Key parameters and properties of the HYPE model.

Parameters	Meaning	Type	Range of Values
lp	Factors for calculating potential evapotranspiration and soil water limit	General Parameters	0.8~1
rivvel	River flow velocity, m/s	General Parameters	0.01~10
epotdist	Coefficient of potential evapotranspiration varying with depth	General Parameters	1~10
wcep	Effective porosity, %	Soil Types	0.01~1
cevpph	The phase of the potential evapotranspiration sine function, days	General Parameters	1~20
fertdays	Fertilizer application days	General Parameters	1~365
freuc	Adsorption isothermal formula parameters	Soil Types	1~250

lists the main parameters and their properties of the HYPE model.

2.2.3. Model Accuracy and Evaluation Metrics

This paper employs commonly used statistical indicators and data visualization to evaluate the simulation accuracy of hydrology water environment models. The evaluation indicators include NSE, R-squared (R²), and percent bias (Pbias). The calculation formulas and meanings of these indicators are summarized in

Table 5. Explanation of model performance indicators.

Metric	Description	Value Range and Evaluation Criteria:
NSE	Reflects the model’s ability to simulate parameter dynamics and quantifies the relative magnitude of residual variance compared with observed data variance	Negative infinity to 1; close to 1 indicates good simulation performance, close to 0 indicates overall credibility but significant process errors, <0 indicates an unreliable model
R2	Indicates the degree of fit between simulated and observed values	0 to 1; closer to 1 indicates better simulation performance
Pbias	Measures the cumulative deviation between simulated and observed values, assessing the model’s overall water balance performance	Optimal value is 0; positive values indicate underestimation, negative values indicate overestimation

.

Moriasi et al. developed a model performance grading evaluation method for hydrological models [38]; therefore, this paper adopted this evaluation method to assess the established HYPE hydrological water environment model. The model’s performance can be rated as “good” if the daily step runoff simulation achieves an NSE > 0.50 and the monthly step nitrogen and phosphorus simulations achieve an NSE > 0.35.

The specific calculation formulas for the evaluation metrics NSE, R², and Pbias are as follows:

$$NSE=1-{\displaystyle \frac{{\displaystyle \sum _{i=1}^n{(S_i-O_i)}^2}}{{\displaystyle \sum _{i=1}^n{(O_i-\overline{O})}^2}}}$$

(1)

$$R^2={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n(O_i-\overline{O})(S_i-\overline{S})}}{\sqrt{{\displaystyle {\sum }_{i=1}^n{(O_i-\overline{O})}^2}}\sqrt{{\displaystyle {\sum }_{i=1}^n{(S_i-\overline{S})}^2}}}}$$

(2)

$$Pbias={\displaystyle \sum _{i=1}^n\frac{S_i-O_i}{O_i}}\times 100\%$$

(3)

In the formula, O_i represents the observed flow rate, m³/s; S_i represents the simulated flow rate, m³/s; $\overline{O}$ represents the mean of all observed values, m³/s; $\overline{S}$ represents the mean of all simulated values, m³/s; and n is the number of days for observed and simulated values.

2.3. Watershed Best Management Practice Scenario Settings

To scientifically evaluate the effectiveness of different management strategies in controlling NPS (particularly nitrogen and phosphorus pollution) in the Chenzhuang rural small watershed, this study constructed multiple scenario simulation schemes based on the core classification framework of BMPs (agricultural production management, nutrient management, landscape configuration management), combined with the pollution control theory of “source reduction–process interception.” The general approach to scenario design was as follows: using the baseline (current situation) as a reference, we independently modified agricultural production practices, nutrient input structures, and landscape patterns (especially the “source–sink” ratio), and then we finally combined optimal measures to assess synergistic effects. All scenarios were closely aligned with the actual conditions of the Chenzhuang village watershed, policy context, and literature support. The specific scenario designs are as follows.

2.3.1. Agricultural Production Management Scenario

Agricultural production management reduces pollution sources through adjustments in agricultural planting methods and farming activities. Research indicates that before the introduction of green natural farming techniques in Chenzhuang village, farmers generally adopted a traditional biennial cropping pattern, which resulted in prolonged bare soil periods, especially during the rainy season, exacerbating soil erosion and the risk of nutrient loss through surface runoff. After establishing an ecological innovation experimental base in Chenzhuang village, crop rotation techniques were promoted. Crop rotation is widely recognized as an effective non-engineering BMP that can improve soil structure and reduce sediment and nutrient loss, thereby enhancing water quality [39]. Therefore, this study set up scenario S2: Implementation of Crop Rotation, based on the baseline scenario (S1: Traditional Farming). This scenario follows the standardized plan from the Chenzhuang village Natural Farming Promotion Center, which involves planting leguminous crops after the main crops are harvested, forming a double-cropping rotation system, to quantify the reduction effect of crop rotation on nitrogen and phosphorus loss compared with traditional farming.

2.3.2. Nutrient Management Scenario

Agricultural fertilization constitutes the primary source of nitrogen and phosphorus pollutants in watersheds. Optimizing nutrient input structures, particularly by reducing reliance on chemical fertilizers and increasing organic fertilizer substitution ratios, represents the most fundamental “source reduction” strategy for controlling agricultural NPS. To quantitatively assess the improvement effects of different nutrient management models on the water quality (TN, TP) of the Chenzhuang village sub-watershed and to quantify the environmental benefits of nutrient measures (organic fertilizer substitution) promoted by Chenzhuang’s green natural farming techniques, this study established gradient nutrient management scenarios (S3–S6). Based on the “Jiangsu Rural Statistical Yearbook,” government website information and field surveys, the level of chemical fertilizer application in Chenzhuang village before the implementation of natural farming (i.e., baseline scenario S1) was determined. To investigate the sensitivity of water environment responses to different agricultural fertilizer reduction levels and provide refined references for policy-making, this study established gradient scenarios with 15% reduction intervals to analyze the specific change processes.

2.3.3. Landscape Pattern Scenario

Landscape configuration management achieves the spatial redistribution of non-point source pollutants by rationally arranging the quantity and spatial structure of landscape components within a region, thereby controlling the spread of NPS. Based on the “source–sink” landscape theory, “source” landscapes (such as farmland) are prone to generate and export pollutants while “sink” landscapes (such as forested areas) can effectively intercept and transform pollutants. Therefore, increasing the proportion of “sink” landscapes is a key BMP for process interception. In the case of Chenzhuang village, the proportion of residential land is low and difficult to reduce, and the water area is also challenging to expand. Thus, the only feasible way to increase “sink” landscapes (forested areas) is through “returning farmland to forest.”

To ensure regional food security and implement farmland protection policies, according to the “Implementation Opinions on Strengthening Farmland Protection, Enhancing Farmland Quality, and Improving Occupation–Compensation Balance” by Jiangsu Province, the farmland area in the study region needs to be maintained at a necessary level. Referring to the minimum per capita farmland area model [40] and relevant regional studies while considering the actual distribution of farmland in the Chenzhuang village sub-watershed (especially farmland adjacent to water bodies) and the spatial needs for ecological restoration, the ecologically feasible upper limit for converting farmland to forest within the study area was determined to be approximately 3% of the total area under the constraint of farmland protection. To evaluate the dose–effect relationship of converting farmland to forest, three conversion gradients were set: 1% (S7), 2% (S8), and 3% (S9). The reduced farmland area is entirely converted to forest land, while other land types remain unchanged.

In summary, the optimal management practice scenarios for the Chenzhuang village area were set as shown in

Table 6. BMP scenario simulation settings.

Type	Scenario Analysis	Conditions Setup
Base Scenario	S1	Traditional farming methods, fertilizer application, and landscape patterns
Agricultural Production Management Scenario	S2	Crop rotation
Nutrient Management Scenario	S3~S6	Reduce chemical fertilizer use by 15% to 50% and replace it with organic fertilizer
Landscape Pattern Configuration	S7~S9	Convert 1%~3%of farmland to forest
Comprehensive Scenario	S10~S13	Green Natural Farming (S2 + S6) to all combination (S2 + S6 + S9)

.

3. Results

3.1. Hydrological Simulation Results

The HYPE model can well-reproduce the runoff process of the Chenzhuang small watershed and its response to rainfall.

Table 7. Evaluation indicators for the runoff simulation calibration period and validation period of Chenzhuang rural small watershed.

	NSE	R2	Pbias
Rate Setting Period	0.813	0.846	0.68%
Verification Period	0.721	0.739	11.16%

presents the evaluation metrics for the daily runoff simulation model performance during the calibration and validation periods. Table 7 presents the performance evaluation indicators of the model during the periodic period and the validation period. The results show that the NSE was much higher than the “good” performance standard for day-scale runoff simulation proposed by Moriasi et al. [38] (NSE > 0.50), indicating that the overall performance of the model was good and it could effectively simulate the dynamics of runoff in the basin. The changing trends of the simulated values and the measured values were highly consistent, indicating that the HYPE model could reflect the relationship between the local climate and hydrological processes in the Chenzhuang small watershed. The overall simulation effect of runoff was good, meeting the requirements of hydrological input accuracy for subsequent water quality simulation.

3.2. Water Quality Simulation Results

As shown in Figure 2, the HYPE model effectively captured the trends in TN and TP concentrations, with the simulated values aligning closely with the observed values at various peaks and troughs.

Table 8. Evaluation indicators for the nitrogen and phosphorus simulation rates during the calibration period and validation period of Chenzhuang rural small watershed.

Simulated Indicators		NSE	R2	Pbias
TN	Rate Setting Period	0.558	0.817	−0.67%
	Verification Period	0.543	0.715	18.87%
TP	Rate Setting Period	0.531	0.772	23.67%
	Verification Period	0.477	0.769	28.06%

further quantifies the model’s performance: the NSE values for TN and TP during the calibration period were 0.558 and 0.531, respectively, and 0.543 and 0.477 during the validation period, all meeting the monthly-scale water quality simulation acceptable standard (NSE > 0.35) recommended by Moriasi et al. [38]. Additionally, the model was sensitive to rainfall events and agricultural activities, successfully capturing significant increases in TN and TP concentrations during periods of concentrated rainfall and fertilization activities (Figure 2), with the timing of simulated peaks closely matching observed peaks. Considering the dynamic trend fit shown in Figure 2 and the “acceptable” NSE indicators in Table 8, it can be concluded that the HYPE model is generally capable of simulating the key characteristics of nitrogen and phosphorus loads in the Chenzhuang small watershed, particularly in response to trend changes and peak events. This provides a feasible model tool for subsequent evaluations of the effectiveness of different management practices (BMPs) on water environment improvement.

Table 9. Key sensitivity parameter calibration results of the HYPE model.

Parameters	Category	Physical Significance	Initial Value	Optimized Value	Relative Comprehensive Sensitivity
rivvel	\	River flow velocity, m/s	0.4	0.6	0.0074
fertdays	\	Fertilization time	100	120	0.0026
freuc	High-activity leaching soil	Adsorption isothermal formula parameters	180	195.213	0.07512
	Saturated cohesive soil		180	19.237	0.05846
epotdist	\	Coefficient of potential evapotranspiration varying with depth	5	3	0.00053
wcep	High-activity leaching soil	Effective porosity, %	0.2	0.046	0.0021
	Saturated cohesive soil		0.3	0.159	0.0085
cevp	Cultivated Land	Potential evapotranspiration rate, mm/(d · °C)	0.25	0.34	0.0048
	Forest land		0.12	0.25	0.0069
	Residential Land		0.02	0.015	0.00034

shows the results of some key sensitive parameters calibration results.

3.3. Analysis of Scenario Simulation Results

The previously mentioned hydrological and water quality simulation results indicate that the constructed HYPE model effectively characterized the hydrological processes and the transport and transformation patterns of nitrogen and phosphorus in the Chenzhuang village small watershed. It also clarifies that agricultural fertilization activities are the key drivers of nitrogen and phosphorus pollution in the watershed. To explore feasible ways to effectively reduce agricultural non-point source nitrogen and phosphorus pollution and improve watershed water quality, this study, based on the “source control–process interception” management concept, set up multi-dimensional BMP scenarios. The calibrated and validated HYPE model was used to quantitatively assess the reduction effects of different scenarios on TN and TP concentrations at the watershed outlet from 2018 to 2021. The simulation results were analyzed in detail from five aspects: agricultural production management, nutrient management, landscape pattern optimization, validation of green natural farming techniques, and comprehensive measures.

3.3.1. Effects of Agricultural Production Management

The crop rotation measure (S2) significantly reduced nitrogen and phosphorus outputs in the watershed. As shown in Figure 3, compared with the baseline scenario (S1), the implementation of crop rotation resulted in a notable decrease in TN and TP concentrations at the watershed outlet. By comparing the scenario simulations of traditional farming practices (S1) and crop rotation (S2), it was found that crop rotation significantly reduced the nitrogen and phosphorus concentrations in surface runoff during the rainy season by shortening the period of bare soil and increasing ground cover. The simulation results demonstrated that the crop rotation measures significantly reduced the TN load. During the 2018–2021 period, the multi-year average concentration decreased from 1.92 mg/L under the baseline scenario to 1.69 mg/L, representing an 11.8% reduction. Regarding TP reduction, crop rotation decreased the annual average TP concentration from 0.22 mg/L in the baseline scenario to 0.19 mg/L, with an annual average reduction rate of 13.6%. During peak periods such as June 2018, TP concentration decreased from 0.43 mg/L to 0.40 mg/L, a reduction of 7.5%. The results indicate that crop rotation is more effective in reducing nitrogen during peak flood stages. Additionally, the synergistic control effect of crop rotation on nitrogen and phosphorus pollution exhibited seasonal variation: the reduction rates for TN and TP during the rainy season (June–September) were significantly higher than those in the dry season (January–March). This difference may have been related to the forms and transport pathways of nitrogen and phosphorus: TN is primarily in dissolved form and migrates quickly with surface runoff, while TP is mainly in particulate form, with its loss more influenced by soil erosion. It is recommended to integrate landscape optimization measures such as cropland-to-forest conversion with existing practices to enhance particulate phosphorus interception, thereby achieving more comprehensive pollution control.

3.3.2. Effects of Nutrient Management

Different nutrient management strategies have significantly varied effects on reducing nitrogen and phosphorus pollution loads. Figure 4 visually compares the dynamic changes in TN and TP concentrations under various nutrient scenarios. Among these, the S6 scenario, representing the core nutrient measure of Chenzhuang’s green natural farming (50% organic fertilizer substitution), was the most prominent. As seen in Figure 4a, compared with the baseline scenario (S1), the annual average TN concentration in the S6 scenario decreased from 1.92 mg/L to 0.95 mg/L, achieving a reduction rate of 50.5%. The annual average TP concentration decreased from 0.22 mg/L to 0.07 mg/L, with a reduction rate of 68.2%. This highlights the particularly significant effect of organic fertilizer substitution on phosphorus (TP) reduction. During the active agricultural periods each year, the TN and TP concentrations in the S6 scenario showed substantial decreases compared with other nutrient management scenarios. This indicates that appropriate nutrient management measures can effectively reduce nitrogen and phosphorus losses at the watershed scale, thereby lowering the risk of eutrophication in downstream reservoirs. Therefore, the adjustment of nutrient management practices in the Chenzhuang area has significantly improved NPS in the watershed.

3.3.3. Scenario Simulation of Optimized Landscape Patterns

Based on the “source–sink” landscape theory, this study systematically evaluated the regulatory effects of reforestation on nitrogen and phosphorus outputs in the watershed by setting landscape configuration scenarios with different forest cover rates (S7~S9). The simulation results showed that, compared with the baseline scenario (S1), all three landscape optimization scenarios effectively reduced TN and TP concentrations at the watershed outlet, with the reduction effect increasing as the forest proportion rose. As illustrated in Figure 5, among the three scenarios, the reduction rates of TN and TP exhibited a clear upward trend with increasing forest proportion, and the reduction effect was enhanced non-linearly with more forest cover. Compared with S1, the S9 scenario achieved a multi-year average reduction rate of 4.14% for TN and 2.78% for TP.

Additionally, the regulatory capacity of reforestation on TN was significantly higher than that on TP, indicating that forest ecosystems have a greater advantage in intercepting dissolved pollutants. Increasing forest area not only effectively reduces nitrogen and phosphorus losses but also improves watershed water quality to some extent. The analysis of Figure 5 reveals that the effect of optimizing landscape patterns was more pronounced during the rainy season, while its impact was relatively limited during the dry season due to restricted pollutant transport pathways. This was mainly influenced by natural conditions (reduced rainfall-driven transport) and human activities (less nutrient input during the agricultural off-season).

3.3.4. Verification of the Effects of Green Natural Farming Techniques

Since the introduction of green natural farming techniques in the Chenzhuang village area in 2016, the villagers’ production and lifestyle have initially shifted towards more sustainable practices, contributing to the gradual improvement of the local ecological environment. The actual application effects of green natural farming techniques (S10 scenario) in Chenzhuang village closely align with the model scenario simulation results. Figure 6 compares the distribution of TN and TP concentrations at five monitoring points (P1, P2, P4, P9, and P10) before and after the implementation of natural farming using box plots. The shaded areas in the legend represent the period before the implementation, while the hollow areas indicate the results after implementation. As shown in Figure 6, the median, mean, and range of TN and TP concentrations at each point significantly decreased after green natural farming technique implementation. For example, at the P10 station, which is the watershed outlet, the average TN concentration dropped from 2.6 mg/L to 0.88 mg/L, improving water quality from below Class V to Class III. The average TP concentration significantly decreased from 0.19 mg/L to 0.06 mg/L, reaching Class II water quality standards.

Green natural farming techniques have significantly reduced N and P concentrations in the area through the synergistic effects of source reduction (reducing fertilizer input) and process interception (increasing ground cover through crop rotation).

3.3.5. Comprehensive Scenario Simulation

Based on the synergistic effects of agricultural production management (S2), nutrient management (S6), and landscape configuration (S7/S8/S9), this study constructed integrated scenarios (S11–S13) combining multiple measures to quantitatively evaluate the regulatory efficiency of the “source reduction–process interception” integrated governance model on nitrogen and phosphorus pollution in the watershed. The simulation results showed that the integrated scenarios significantly outperformed single measures in reducing TN and TP and exhibited a clear gradient enhancement effect and spatiotemporal synergy.

Figure 7 presents the governance outcomes of the integrated scenarios from 2018 to 2021. As the reforestation ratio increased from 1% (S11) to 3% (S13), the annual average concentrations of TN and TP decreased stepwise. Specifically, the average concentration of TN decreased from 0.70 mg/L in S11 to 0.63 mg/L in S13, a reduction of 10%; the average concentration of TP decreased from 0.052 mg/L to 0.046 mg/L, a reduction of 11.5%. Further analysis revealed that starting from 2020, the average concentration of TN was reduced to 0.49 mg/L, meeting the Class II surface water standard; TP remained consistently stable at the Class II surface water standard. Since 2020, TN and TP concentrations have consistently met Class II surface water standards, demonstrating the spatiotemporal synergistic advantages of the “source interception–process interception” approach.

To provide a comprehensive comparison of the effectiveness of different BMP scenarios,

Table 10. Summary of TN and TP reduction rates and water quality improvements under BMP scenarios.

Scenario	Management Measures	Concentration (mg/L)		Reduction
		TN	TP	TN	TP
S1	Traditional farming	1.92	0.22	/	/
S2	Crop rotation	1.69	0.19	11.8%	13.6%
S6	50% organic fertilizer substitution	0.95	0.07	50.5%	68.2%
S9	3% farmland to forest conversion	1.84	0.214	4.14%	2.78%
S10	S2 + S6	0.88	0.06	66.2%	72.7%
S13	S10 + S9	0.63	0.046	67.2%	79.1%

summarizes the reduction rates of TN and TP, as well as water quality improvements under all scenarios (S1–S13). The results demonstrate that integrated scenario S13 achieved the highest reduction rates, with TN and TP concentrations reduced by 67.2% and 79.1%, respectively, compared with the baseline scenario S1, and consistently met Class II surface water standards since 2020.

4. Discussion

4.1. Application of the HYPE Model in Rural Small Watersheds

This study successfully constructed, calibrated, and validated the HYPE model based on observed hydrological and water quality data from the Chenzhuang village small watershed. The model demonstrated good performance in daily runoff simulations, effectively capturing the hydrological dynamics and their response to rainfall in the watershed. This indicates that the HYPE model has good applicability for runoff simulations in rural small watersheds, with better performance during the flood season compared with non-flood periods. The HYPE model was constructed with a daily time step mainly based on the following considerations: (1) The HYPE model can already achieve a good balance between the simulation accuracy and computational efficiency of hydrological processes on a daily scale, and it is particularly suitable for long-term simulation in small and medium-sized river basins; (2) the water quality monitoring data in the study area is on a monthly scale, and the daily step size can meet the requirements for water quality trend analysis. It is noteworthy that during the calibration and validation periods, the HYPE model exhibited some degree of overestimation or underestimation of peak runoff, affecting the model’s hydrological accuracy. For instance, during runoff peaks, the model might simulate values lower than the actual peak runoff. Similar issues have been observed in the application of the HYPE model in other watersheds abroad. For example, Jiang et al. found in their study of a watershed in central Germany that peak runoff was often underestimated under abnormal rainfall conditions [41]. Possible reasons include (1) the HYPE model uses a daily scale as the minimum time unit for simulation, while short-duration, high-intensity rainfall typically causes runoff peaks with smaller time units, leading to errors in peak runoff simulation; (2) the study area, Maoshan Chenzhuang rural watershed, is relatively small, resulting in a quicker runoff response to rainfall. Combined with the first reason, this could lead to a greater impact of rainfall on the simulation results in a study area. Although the daily step size may weaken the peak runoff response of short-term heavy rainfall events, through parameter sensitivity analysis and calibration (such as river flow velocity (“rivvel”) and soil effective porosity (“wcep”)), the model can still effectively capture key hydrological events (such as peak nitrogen and phosphorus loads during the rainy season). Additionally, incorporating a sediment transport module could address peak runoff underestimation and enhance model accuracy. This is particularly relevant for particulate phosphorus (PP) transport dynamics, which depend on sediment movement during high-flow events. However, this study did not include a sediment transport module due to the primary focus on nitrogen and phosphorus transport processes and the limited availability of sediment monitoring data in the Chenzhuang watershed. Future research could explore coupling sediment transport modules from other models (e.g., SWAT) to improve peak runoff simulation and better represent particulate phosphorus dynamics. And for future research, we also hope to enhance fieldwork in this area and collect data with higher precision.

In this study, the HYPE model achieved high accuracy in simulating nitrogen and phosphorus transport processes in rural small watersheds, effectively characterizing the relationship between fertilization intensity and surface runoff pollutant response. The model’s accuracy in simulating hydrology and water quality is higher during the rainy season, where it can better capture changes in nitrogen and phosphorus concentrations. Particularly during the early stages of agricultural activities, such as manure application after sowing, the model accurately reflects the increase in N and P concentrations. Additionally, studies by Rodríguez-Blanco et al. [42] have shown that approximately 67% of phosphorus outputs in Galicia’s mixed land-use catchments are associated with precipitation events, further illustrating that storm runoff events are major contributors to TP load, a finding consistent with this study. During the rainy season, heavy rainfall significantly erodes the land, leading to increased nitrogen and phosphorus loss, which the model can effectively simulate. However, there are some errors in runoff simulation during low-flow periods in the dry season, where the model may not accurately capture small changes in runoff. Furthermore, the HYPE model inadequately represents the transport and transformation processes of particulate phosphorus (such as TP adsorbed onto sediment), potentially leading to TP simulation biases. Future improvements could involve coupling sediment modules from other models. The watershed’s TP concentration exhibits clear temporal variation, generally showing higher values during the farming season and lower values at other times. This indicates a strong response relationship between TP and fertilizer application. Therefore, optimizing fertilization timing and reducing tillage can effectively suppress phosphorus loss due to soil disturbance, significantly reducing agricultural phosphorus output, consistent with findings by Santos et al. [43]. TN concentrations are higher from December to February, likely due to the Lunar New Year period when some migrant workers return home, increasing the population in the Maoshan Chenzhuang area and thus raising TN concentrations due to human discharge. The population increase has a weaker impact on TP, so during the Lunar New Year, when agricultural activities are inactive, TP concentrations remain low. This aligns with the findings of Chen Wenjun et al. [36]. During the calibration and validation periods, TN’s Pbias was −0.67% and +18.87%, respectively, while TP’s Pbias was as high as +23.67% and +28.06%, indicating a systematic underestimation in the model’s simulation of TN and TP. This suggests challenges in the model’s simulation of TP, particularly regarding the transport and transformation processes of PP adsorbed onto sediment, which may be inadequately represented. This relates to the complex forms of TP in the environment and its transport process being strongly influenced by sediment transport. Additionally, groundwater processes were explicitly integrated into the HYPE model structure to account for subsurface hydrological contributions. Groundwater discharge to downstream sub-basins was regulated by the state variable “grwtolake”, which partitions groundwater flow between sub-basins, while the recession coefficient “rcgrw” controlled baseflow generation as a function of watershed retention capacity. These parameters were calibrated during model optimization (Section 2.2.2), ensuring the realistic simulation of baseflow dynamics during low-rainfall periods. However, uncertainties in groundwater parameterization (e.g., spatial heterogeneity of aquifer properties) may partially contribute to the observed dry-season runoff biases. Future studies could enhance groundwater representation by incorporating high-resolution aquifer data or coupling with modular subsurface flow models to improve PP transport simulation, as PP adsorption is influenced by groundwater–surface water interactions. Overall, the HYPE hydrological and water quality model constructed in this study demonstrates good applicability and reliability in the Chenzhuang village small watershed. The successful application in Chenzhuang village indicates that the HYPE model can effectively simulate hydrological and water quality processes in small, intensively farmed watersheds. Given the similarities in land use, climate, and agricultural practices, the HYPE model can be adapted with appropriate parameter adjustments to provide a reliable tool for water resource management and pollution control in the Taihu Lake basin or other similar regions. Additionally, research on pollutant transport and management in small watersheds can elucidate detailed hydrological processes and delayed pollutant transport mechanisms, demonstrating significant scientific value [44]. These findings provide a robust scientific foundation for watershed management practices.

4.2. Evaluation of Pollution Control Efficiency Under Multi-Dimensional Scenarios

To scientifically assess the reduction potential of different management measures on nitrogen and phosphorus pollution, this study found that agricultural production management, nutrient management, and landscape configuration measures all play significant roles in pollution reduction and control in rural small watersheds. The integrated scenario simulations (S11–S13) clearly revealed a significant synergistic effect between “source control (S2 + S6)” and “process interception (S7–S9)” measures, which is an effective approach to addressing water environment pollution in rural small watersheds. Based on simulations of agricultural production management and nutrient management scenarios, the reduction effects of green natural farming techniques on N and P were found to be nearly equivalent to the combined effectiveness of S2 and S6. This finding is consistent with conclusions from studies conducted in other watersheds under various scales and conditions [45,46]. Taking the Chenzhuang village area as an example, with its high intensity of agricultural activities and concentrated rainfall during the rainy season, there is a necessity for the synergy of agricultural production management, nutrient management, and landscape optimization. Therefore, the key is to determine the optimal configuration ratio of each measure based on the actual conditions of the study area, selecting the most suitable non-engineering management plan. This study shows that the comprehensive application of agricultural production management, nutrient management, and landscape configuration can significantly reduce nitrogen and phosphorus pollution loads in rural small watersheds. This efficient “source–process synergistic interception” governance model is also applicable to other regions with high farmland proportions and concentrated rainfall. Policymakers and watershed managers can draw on this model, adapting and implementing similar comprehensive BMPs in accordance with local conditions to achieve water quality improvement goals and sustainable agricultural development. The methods and results of this study provide a reference for pollution control in small watersheds with intensive agriculture in developing countries.

4.3. Economic Feasibility Analysis of Best Management Practices

While the BMPs proposed in this study have significant environmental benefits, their economic feasibility is also an important factor that requires in-depth exploration. Using cost–benefit analysis, this study quantified the implementation costs (such as land loss and initial investment) and economic benefits (such as cost savings, policy compensation, and additional income) of each measure based on scenario simulations and field survey data from a 50-acre experimental field in the Chenzhuang small watershed, and it calculated the net benefits. This experimental area, as a key experimental plot within the Chenzhuang small watershed, represents the typical characteristics of intensive agricultural regions, and its results can serve as a reference for pollution control in the entire watershed and similar areas. The economic feasibility of each measure is analyzed as follows.

• Green Natural Farming Techniques (S10):

Costs: Initial transformation investment is required (such as purchasing organic fertilizers and technical training).

Benefits: Green natural farming techniques significantly reduce the costs of chemical fertilizers and pesticides, and they achieve long-term benefits by improving crop quality and yield (20–30% increase), as well as sales prices. Although crop rotation introduces some hidden labor costs, farming remains the primary income source for local residents. This study indicates that, after accounting for these hidden costs, the net benefit still increases by approximately 350 CNY per acre. For the 50-acre experimental field, the estimated annual net benefit is about 17,500 CNY. The acceptance rate among farmers is 82%, indicating intrinsic economic motivation. This demonstrates significant long-term economic benefits and provides an economic foundation for promotion.

2.

Reforestation (S7–S9):

Costs: Field surveys in the Chenzhuang area indicate that the annual farmland output value is about 1500–1800 CNY/acre. In the S9 scenario (3% reforestation, 1.5 acres), the annual loss is approximately 2250–2700 CNY.

3.

Compensation and Benefits:

Policy Compensation: According to the “Jiangsu Provincial Forest Ecological Benefit Compensation Fund Management Measures,” the compensation standard for reforestation is 150 CNY/acre/year (80 CNY ecological compensation + 70 CNY management fee). The S9 scenario compensation is 225 CNY/year. Jiangsu Province has been continuously promoting the ecological compensation mechanism since 2014, providing stable policy support for reforestation.

Understory Economic Benefits: Promoted by the Chinese Academy of Sciences team, the potential net benefit of understory economy (such as medicinal herbs) is estimated at about 300 CNY/acre/year. The S9 scenario yields 450 CNY/year.

The economic accounting of the plots in this area is shown in

Table 11. Economic impact statistics of implementing scenario S13 on 50 mu experimental fields in the Chenzhuang small watershed.

Project	Formula	Amount (CNY/Year)	Description
Total loss	Loss of arable land × annual average output value per arable land	2250~2700	Loss of agricultural land output
Policy compensation	compensation standards for loss of cultivated land	225	/
Understory economy benefits	Loss of arable land × annual value of crops grown under forest canopy	450	Net income from understory crops
Technology cost reduction	Remaining arable land × natural farming method saves costs	16,975	Applied to the remaining farmland
Net income	Profit−loss	14,950~15,400	Overall net revenue

. Additionally, during policy implementation, the provincial compensation (150 CNY/acre) and understory economic benefits (300 CNY/acre) will be bundled and distributed to ensure farmers’ basic income. Based on the above analysis, using ecological public forest compensation instead of land acquisition standards aligns with the sustainability characteristics of reforestation, and the investment payback period is relatively short. The synergy between reforestation policy and natural farming innovation can transform ecological constraints into economic gains.

According to the «Statistical Bulletin on the National Economic and Social Development of Jurong City», the per capita disposable income of all residents in the city throughout the year is 51,864 CNY. For a typical 10-acre household in Jurong—this represents a 3000 CNY (5.8%) income gain. While the absolute net benefit of S13 (15,000 CNY/year for 50 acres) appears modest at the field scale, its per-acre impact represents a meaningful income uplift for smallholders—especially when contextualized against regional incomes and integrated with policy support. The results demonstrate that the BMPs proposed in this study, particularly the integrated scenario (S13), not only achieve significant environmental benefits (water quality compliance) but also exhibit strong economic feasibility through a triple-path approach encompassing policy compensation, technological innovation (cost-saving and income-boosting natural farming techniques), and agroforestry economy, thereby realizing an “environment–economy” win–win scenario.

4.4. Localization, Implementation, and Transferability of Best Management Practices

To ensure the practical application and scalability of the BMPs evaluated in this study, this section addresses the localization, implementation challenges, and transferability of these measures to other rural small watersheds, particularly focusing on the barriers to adoption, applicability to other regions, and specific policy recommendations.

• Barriers to Adoption of Green Natural Farming Techniques

The adoption rate of green natural farming techniques in Chenzhuang village reached 82%, as determined through conducted field surveys. The primary barriers include the following. (1) Uncertainty of economic costs and short-term returns: some farmers are concerned about the high initial investment in green agriculture (such as organic fertilizers and crop rotation cultivation) and the unstable short-term returns. Although the cost–benefit analysis shows long-term economic feasibility (such as the conclusion that the annual net income of 50 mu of experimental fields increases by approximately 15,000 yuan under scenario S13), some farmers prefer the immediate benefits of traditional agriculture. (2) Insufficient technical knowledge and training: Green agricultural technologies (such as crop rotation, cover crop planting and precise fertilization) require certain professional knowledge and operational skills. Some farmers, especially elderly ones, lack relevant technical training or have a low acceptance of new technologies, which limits their willingness to adopt them. In addition, in the promotion of best management measures, there are certain obstacles to the implementation of the measure of returning farmland to forest. Some farmers in Chenzhuang village have a relatively small land scale, making it difficult to effectively implement some BMPs that require a large space (such as landscape optimization measures or process interception facilities).

To overcome these barriers, targeted interventions such as subsidized technical training programs, financial incentives for adopting organic inputs, and the development of robust local markets for green agricultural products could significantly enhance adoption rates. Additionally, establishing demonstration plots showcasing successful outcomes has proven effective in building farmer confidence and reducing perceived risks, as clearly demonstrated by the high adoption rate (82%) achieved following the success of the 50 mu experimental plot in Chenzhuang village.

2.

Transferability to Other Watersheds

The above issues indicate that promoting BMPs requires a full consideration of farmers’ economic affordability, technical capacity, and actual needs to enhance adoption. To ensure methodological adaptability when scaling this study, we propose the following: (1). Measures designed based on local conditions: Different watersheds have significant differences in climate, topography, soil types, rainfall patterns and agricultural production methods. Therefore, before applying BMPs, a detailed investigation of the basin characteristics should be carried out, and targeted measures should be designed and adjusted. For instance, due to concentrated rainfall and high-intensity agricultural activities, Chenzhuang village has emphasized the synergy of “source control” and “process interception”. In addition, when resources are limited, measures should be used or combined reasonably. For instance, sloping farmland should be given priority for returning to farmland, and flat land should be given priority for crop rotation. (2). Enhance farmers’ participation and training: Farmers are the key subjects in the implementation of BMPs. Their understanding and operational capabilities of green agricultural technologies should be improved through technical training, demonstration projects and on-site guidance. Therefore, we suggest that before implementing different BMPs, each research team should conduct on-site experiments in the villages of the study area, which is more conducive to rural villagers accepting and participating in the relevant renovations. (3). Improve economic incentives and policy support: When promoting BMPs in other river basins, it is recommended that the government and relevant departments establish a complete subsidy mechanism and ecological compensation policy to reduce the initial economic burden on farmers implementing BMPs. In addition, efforts should be made to promote the construction of green agricultural product markets, enhance farmers’ expectations of economic returns through price premiums and brand promotion, and encourage their long-term participation.

The Chenzhuang village case study shows that the integrated BMP framework is effective in small watersheds with high agricultural intensity and seasonal rainfall. To ensure scalability and sustainability, these measures should be integrated into a cohesive framework supported by local governments and research institutions. This study demonstrates that combining financial incentives, ecological compensation, and technical support can overcome adoption barriers and achieve both environmental and economic goals.

5. Conclusions

This study, based on the HYPE hydrological and water environment model, focused on a typical green rural small watershed in Jiangsu Province. It constructed a distributed model system incorporating hydrological processes and nitrogen and phosphorus transport and transformation modules, systematically evaluating the synergistic control efficiency of various BMPs on agricultural non-point source nitrogen and phosphorus pollution. The main conclusions are as follows:

• The HYPE model, constructed and improved according to the actual conditions of the study area, demonstrated good applicability at the rural small watershed scale, effectively simulating watershed hydrodynamics (daily runoff NSE > 0.7) and nitrogen and phosphorus transport processes. This provides a reliable tool for assessing the environmental effects of BMPs. The model performs better during rainy seasons and high-flow periods but requires further improvements in depicting low-flow conditions during dry seasons, peak runoff events, and particulate phosphorus transport processes.
• Agricultural production management, nutrient management, and landscape configuration all contribute to varying degrees of pollution reduction and prevention in rural small watersheds. Crop rotation, by shortening the bare soil period and increasing ground cover, significantly reduces nitrogen and phosphorus loss intensity during the rainy season, with annual average TN and TP concentration reductions of 11.8% and 13.6%, respectively. Nutrient management can substantially reduce TN and TP outputs, with a more pronounced reduction of phosphorus in the study area. Landscape configuration management effectively intercepts pollutants by increasing the proportion of “sink” landscapes, with better interception effects for dissolved nitrogen compared with particulate phosphorus. The green natural farming techniques implemented in the study area, through in-depth analysis, achieved a synergistic effect of source reduction and process interception, significantly improving the watershed outlet water quality from below Class V to a stable Class II surface water standard.
• The integrated management strategy (S13), combining agricultural production management, nutrient management, and landscape configuration, shows significantly better pollution reduction effects than individual measures, with TN and TP annual average concentrations reduced to 0.63 mg/L and 0.046 mg/L, respectively, consistently meeting Class II surface water standards since 2020. This comprehensive strategy maximizes pollution load reduction efficiency and achieves water quality improvement and sustainable agricultural development through a complementary mechanism of natural farming to reduce source emissions and erosion risk.
• Cost–benefit analysis indicates that the integrated scenario (S13) achieves an annual net income increase of approximately 15,000 CNY within the 50-acre experimental field. This is achieved through cost savings and income generation from natural farming, ecological compensation, and understory economy, demonstrating good economic feasibility. It is recommended to establish a three-dimensional policy system of green agriculture subsidies, ecological compensation funds, and technical training to promote the large-scale application of BMPs.

In summary, this study not only validates the effectiveness of BMPs in controlling rural NPS but also provides scientific evidence and references for pollution control in the Chenzhuang village area and other similar regions. It aids in formulating more reasonable agricultural management and landscape planning strategies, achieving effective control of NPS and continuous water quality improvement, and provides scientific decision support for optimizing water environment management in rural watersheds and building “beautiful countryside”.