Characteristics of Nutrient Transport in Runoff from Different Land-Use Types on Maozhou Island in the Li River Basin

Abstract

Non-point source pollution poses a severe threat to the water quality of the Li River. This study conducted field monitoring of pollution loads from different land-use types on Maozhou Island in the Li River during the 2023 rainy season. Runoff water quality from vegetable plots, orchards, and bamboo forests consistently exceeded standards, with vegetable plots being the primary source of pollution. Their total phosphorus (TP) concentration exceeded standards by nearly 25 times, contributing the highest annual load. The transport of pollutants (TP, total nitrogen(TN), chemical oxygen demand(COD_Cr)) was closely correlated with suspended solids (SS), with the finest particles (<5 μm) identified as the primary carrier exhibiting the strongest pollutant enrichment capacity (e.g., in vegetable fields, the correlation coefficient r between < 5 μm particles and TP was >0.85, p < 0.01). Rainfall patterns significantly influenced pollutant concentrations; TN and TP levels increased with preceding dry days, while phosphorus output from vegetable plots decreased with rising average rainfall temperature. Compared to bamboo forests, vegetable plots and orchards exhibited lower soil adsorption capacity. This study recommends a connectivity-based strategy prioritizing the interception of heavily enriched fine particulate matter (<5 μm) through runoff control and enhanced wetland retention functions. These findings underscore the importance of controlling fine particulate matter for reducing non-point source pollution and maintaining ecological health in the Lijiang River basin.

1. Introduction

Surface runoff pollution refers to the phenomenon whereby pollutants from the atmosphere, land surface, and soil are dispersed into surface water and groundwater through the leaching and flushing effects of rainfall runoff, leading to water quality deterioration. Primary sources of pollution include the excessive use of fertilizers and pesticides in agriculture, soil erosion, livestock manure, urban surface runoff, and deforestation. The range of pollutants associated with surface runoff is extensive, encompassing pesticides, fertilizers, and herbicides in agricultural runoff [1,2]; toxic chemicals and petroleum hydrocarbons in urban runoff; sediments and soil particles from construction sites and farmland; bacteria, nitrogen, and phosphorus from livestock farming; and salts and acidic substances from abandoned tailings.

Runoff-induced non-point source pollution constitutes the primary form of such contamination. This pollution exhibits characteristic diffuse properties, including extensive geographical distribution, significant temporal and spatial fluctuations in pollutant loads, and substantial challenges in research, prevention, control, and remediation [3]. Its severity is also strongly influenced by regional rainfall patterns and natural environmental conditions, collectively contributing to its highly random and sporadic nature. Based on pollution source types, runoff non-point source pollution can be categorized into several forms, including urban runoff pollution, agricultural production pollution, rural livestock manure pollution, forest surface runoff pollution, and runoff pollution from mining areas and construction sites.

Through bibliometric analysis, scholars such as Lei Xiao found that agricultural runoff pollution has received greater attention in global research than urban runoff pollution. This disparity may stem from differing land-use structures: in urban areas, stormwater runoff typically exhibits high permeability, making pollution control relatively manageable [4]. Conversely, tracing pollution sources proves considerably more challenging in rural environments such as agricultural fields and livestock farming zones, thereby complicating the prevention and management of agricultural non-point source pollution. Pollutant loads in water bodies such as Beijing’s Miyun Reservoir, Tianjin’s Yuqiao Reservoir, and Anhui’s Chaohu Lake primarily originate from non-point sources, with agricultural non-point source pollution being the principal contributor [5]. Data from the National Pollutant Source Census conducted by China’s Ministry of Ecology and Environment indicates that key pollutants from agricultural sources include total nitrogen (TN), total phosphorus (TP), chemical oxygen demand (COD), and ammonia nitrogen [6]. Non-point source pollution in China’s agricultural sector is severe, presenting multiple significant challenges for prevention and control efforts.

Runoff-induced non-point source pollution poses major threats to ecological environments and socio-economic development. Its detrimental effects include: accelerating water eutrophication and degrading water quality, thereby jeopardizing drinking water safety [7,8]; reducing aquatic biodiversity and weakening ecosystem resilience [9,10]; degrading soil quality and diminishing agricultural product quality [11]; diminishing the scenic value of water bodies, damaging rural ecological environments, and hindering the development of water tourism [12].

As a typical example of a river with mid-stream islands (hereafter referred to as islets) in China, the Li River features widespread distribution of such islets. The natural environment on these islands differs markedly from the surrounding terrestrial landscapes, boasting near-natural ecosystems rich in woodlands, shoals, wetlands, and diverse vegetation species. These ecological resources, complemented by the surrounding river waters, karst peaks, and bamboo groves, create the Li River’s distinctive scenery. Not only do they provide habitats for wildlife, but the islands’ water-enclosed nature fosters a comfortable summer microclimate, attracting increasing numbers of visitors seeking secluded beauty. However, due to erosion from natural disasters and economic development, such as tourism, statistics indicate that among the 252 islands along the Li River stretch from Guilin city to Yangshuo, 178 have suffered human interference. Ten of these larger islands, each exceeding 20 hectares, have long been home to clustered villages, agricultural cultivation, and tourism development. Untreated domestic sewage flows freely into the river, rubbish is dumped indiscriminately, poultry roam freely, runoff channels silt up, gravel banks lie exposed over large areas, and plant communities have degraded. This high degree of ecological fragmentation severely impacts the Li River’s landscape, making island conservation an urgent priority.

This study focuses on Maozhou Island, a representative inhabited alluvial island in the Li River, to address non-point source pollution from soil erosion. Following preliminary surveys, specific land patches, including vegetable plots, orchards, and bamboo groves, were selected as sub-catchments for detailed investigation. Through a year-long monitoring programme, rainfall runoff, soil, and water samples were collected and analyzed. The research aims to: (1) characterize runoff pollution by examining event-mean concentrations and spatial distribution patterns across land uses; (2) identify influencing factors such as rainfall characteristics, soil properties, and vegetation cover; and (3) estimate pollutant loads using event-based and annual methods, supported by GIS-based spatial analysis. This comprehensive approach helps clarify the spatial dynamics of runoff pollution, providing a basis for effective mitigation strategies in similar riverine environments.

2. Materials and Methods

2.1. Geographical Profile of the Study Area

Maozhou Island is located south of the ancient town of Dayu in Lingchuan County, Guilin City. Its geographical location is shown in Figure 1. This alluvial island, formed by the erosion and sedimentation of the Li River, is surrounded by a tributary of the Lijiang River. As a terrace landform unit, the island exhibits flat topography with slopes ranging from 0 to 10 degrees. Its stratigraphy shows rhythmic deposition, characterized by an upper layer of silty clay and a lower layer of gravelly pebbles. The distinct binary structure, combined with predominantly natural soil banks featuring weak erosion resistance, makes the island susceptible to bank collapse during flood events due to scouring, erosion, and surface runoff.

2.2. Sample Collection and Indicator Analysis

During the monitoring period from October 2022 to September 2023, a total of 32 effective rainfall events were recorded. To ensure comparability, the 12 most representative rainfall events from each of the three land-use types—vegetable fields, orchards, and bamboo forests—were selected for analysis (totaling 36 runoff events). Sampling points were selected based on the following principles: (1) Representativeness: covering the three primary land-use types within the study area and their typical microtopography; (2) Comparability: establishing duplicate sampling points within each land-use type; (3) Accessibility and safety.

Twelve sampling points, relatively highly impacted by human activity, were established for sample collection. Water sampling points included: Waterfront Platform (W1), Ferry Terminal (W2), Maoli Bridge (W3), Constructed Wetland (W4), Western Island Water Intake (W5), Fisherman’s Wharf (W6), Area beneath Bridge at Compound 51 (W7), Vegetable Garden Well (W8), Well 49 (W9), Vegetable Plot (W10), Bamboo Grove (W11), and Orchard (W12). Additionally, soil samples were collected from the Vegetable Plot (S1), Bamboo Grove (S2), and Orchard (S3). Precise coordinates for all sampling points were accurately recorded using a Global Positioning System (GPS).

Following the cessation of runoff, dissolved oxygen (DO) in the water collected in the sampling bucket was immediately measured on-site using a multi-parameter water quality analyzer (Hach HQ40d: Hach Company, Loveland, CO, USA). Water depth was determined using a gauge, and a stratified sampler was employed to collect uniformly mixed water samples. After rainfall events, the rainwater accumulated in the collection vessel was thoroughly stirred before sampling. Surface water samples from the Li River were collected 0.5 m below the water surface using an acrylic water sampler, while well water samples were obtained using a pressure sampler. For both river and well water, on-site measurements of DO, pH, and temperature were performed with the Hach HQ40d. Three replicate samples were collected at each designated sampling location.

All water samples were promptly sealed, transported to the laboratory under controlled conditions, and refrigerated at 4 °C for preservation until analysis, which was conducted as soon as possible.

Surface soil samples (0–5 cm depth) were collected using a portable stainless steel soil spade. At each sampling point, three parallel sub-samples were taken and combined in the field to form a composite sample. All soil samples were transported to the laboratory and immediately placed in a clean, cool, and well-ventilated area for natural drying.

All water samples were collected between mid-October 2022 and late September 2023. All analytical procedures adhered to the standard methods outlined in Standard Methods for the Examination of Water and Wastewater (22nd Edition, 2012) [13]. Specific methodologies were as follows: total phosphorus (TP) was determined using the ammonium molybdate spectrophotometric method; total nitrogen (TN) was determined using the alkaline potassium persulphate digestion-ultraviolet spectrophotometric method; ammonia nitrogen (NH₃-N) was measured using the Nessler’s reagent spectrophotometric method; nitrate nitrogen (NO₃⁻-N) was determined using the ultraviolet spectrophotometric method; nitrite nitrogen (NO₂⁻-N) was measured using the spectrophotometric method; chemical oxygen demand (COD_Cr) was determined by the potassium dichromate method; manganese dioxide index (COD_Mn) was determined by the potassium permanganate method; suspended solids (SS) were determined by gravimetric analysis.

Soil characterization employed the following analytical methods: Total nitrogen (TN) was determined by alkaline potassium persulfate digestion coupled with ultraviolet spectrophotometry; available nitrogen (AN) was measured by alkaline hydrolysis diffusion; nitrate nitrogen (NO₃⁻-N) was quantified by ultraviolet spectrophotometry; ammonia nitrogen (NH₃-N) was analyzed using the Nessler’s reagent colorimetric method; TP was determined by acid-soluble molybdenum antimony antispectrophotometry; available phosphorus (AP) by sodium bicarbonate extraction-molybdenum antimony antispectrophotometry; organic matter (OM) by potassium dichromate oxidation-colorimetric method. Additionally, soil particle size distribution was measured using a Malvern 3000 laser particle size analyser(Malvern Panalytical Ltd., Malvern, Worcestershire, UK).

Quality Assurance and Control

To ensure data quality, stringent quality assurance and control procedures were implemented throughout the analysis process. All water quality and soil parameter determinations were conducted using triplicate samples, with relative standard deviations maintained below 5%. Instrument calibration and result validation were performed using nationally certified reference materials. For key parameters (TP, TN, COD_Cr), each batch of samples included analysis of blanks and spiked recovery samples, yielding recovery rates between 85% and 115%. All samples were processed within four hours of collection and stored at 4 °C in the dark until analysis.

2.3. Data Analysis

2.3.1. Pollutant Concentration Equation

To characterize the extent of runoff pollution in the monitored area during a rainfall event, the event-mean concentration (EMC) can be employed. The EMC represents the flow-weighted average concentration of pollutants over the entire runoff process of a rainfall event. It is defined as the ratio of the total pollutant mass to the total runoff volume during the rainfall-induced runoff process. This metric serves as a valuable tool for analyzing and assessing the pollution level of surface runoff [14]. The calculation formula is given by Equation (1):

$$\mathrm{EMC}={\displaystyle \frac{M}{V}}={\displaystyle \frac{{\displaystyle {\int }_0^t{C}_t\cdot Q_t\cdot dt}}{{\displaystyle {\int }_0^t{Q}_t\cdot dt}}}$$

(1)

where M represents the total pollutant mass (kg), V denotes the total runoff volume (m³), C_t is the pollutant concentration at time t (mg/L), and Q_t refers to the runoff flow rate at time t (m³/s).

The EMC values for pollutants in rainfall runoff from the Vegetable Plot, Orchard, and Bamboo Forest were calculated using the average concentration (SMC) method.

$$\mathrm{SMC}={\displaystyle {\sum }_{i=1}^n\frac{W_i\times EM{C}_i}{{\displaystyle {\sum }_{i=1}^n{W}_i}}}$$

(2)

In the formula, Wᵢ represents the runoff volume (m³) for rainfall event i, and EMC_i denotes the event mean concentration (mg/L) for the same rainfall event.

2.3.2. Correlation Analysis

The Pearson correlation coefficient is a statistical measure used to quantify the linear relationship between two continuous variables. Its values range from −1 to 1, where 1 indicates perfect positive linear correlation, −1 denotes perfect negative linear correlation, and 0 signifies no linear correlation. The strength of correlation is categorized based on the absolute value of the correlation coefficient |r|: 0.0–0.2 (extremely weak), 0.2–0.4 (weak), 0.4–0.6 (moderate), 0.6–0.8 (strong), 0.8–1.0 (extremely strong). This classification is widely adopted in environmental statistics [15]. The Pearson correlation coefficient is calculated by dividing the covariance of variables by the product of their standard deviations. This method applies to continuous data assumed to follow a normal distribution.

2.3.3. Analysis of Soil Adsorption Data

The equilibrium adsorption capacities of soil for phosphorus and ammoniacal nitrogen are calculated using Formula (3):

$$q_e={\displaystyle \frac{(c_0-c_e)v}{m}}$$

(3)

where q_e is the equilibrium adsorption capacity (mg/g); c₀ represents the initial concentration (mg/L); c_e denotes the equilibrium concentration at adsorption/desorption (mg/L); v refers to the volume of the solution added to the sample (L); and m is the dry mass of the soil sample (g).

The adsorption isotherms for phosphorus and ammonia nitrogen in soil are plotted with the equilibrium concentration c_e of dissolved inorganic phosphorus or ammonia nitrogen in solution as the x-coordinate, and the amount q_e of soluble inorganic phosphorus or ammonia nitrogen adsorbed per unit mass of sample as the y-coordinate.

This study employed the Freundlich and Langmuir adsorption isotherm models to fit the adsorption isotherm data. Langmuir and Freundlich represent two classical models describing soil adsorption behaviour. The Langmuir model, based on the assumptions of monolayer adsorption and uniform adsorption sites, is suitable for describing systems dominated by chemical adsorption; whereas the Freundlich model is an empirical formula applicable to describing multilayer physical adsorption on non-uniform surfaces.

The Freundlich adsorption isotherm model is expressed as:

$$q_e=K_F{c}_e^{1/n}$$

(4)

where K_F is the Freundlich adsorption coefficient (L/g), and 1/n is an exponential factor related to the adsorption system characteristics (typically with 0 < n < 2). All other symbols retain their previously defined meanings.

The Langmuir adsorption isotherm model is expressed as:

$$q_e={\displaystyle \frac{q_m{K}_L{c}_e}{1+K_L{c}_e}}$$

(5)

where K_L represents the Langmuir adsorption coefficient (L/g), and q_m denotes the maximum monolayer adsorption capacity per unit mass of sample (mg/g). All other symbols maintain their previously defined meanings.

Non-linear least squares were employed to directly fit the original equations of the Langmuir and Freundlich models (Equations (4) and (5)) using OriginPro 2022 software, thereby obtaining more accurate model parameters (K_F, n, K_L, q_m). This approach avoids errors potentially introduced by linearisation transformations.

In this study, different models were employed to fit the adsorption data for phosphorus and ammonium nitrogen, respectively. For phosphorus adsorption, the data exhibited good agreement with both the Langmuir and the Freundlich models. For ammonium nitrogen adsorption, experimental data demonstrated a favourable linear relationship within the studied concentration range; the Henry linear adsorption model was adopted for description. The model expression is:

$$q_e=K_H{c}_e+q_0$$

(6)

where K_H denotes the Henry adsorption coefficient (L/g), q₀ represents the intercept (mg/g), and its negative value indicates desorption occurring at extremely low solution concentrations. EC₀ (mg/L) is the zero adsorption equilibrium concentration, i.e., the c_e value when q_e = 0, calculated as: ${EC}_0={\displaystyle \frac{-q_0}{K_H}}$.

2.3.4. Calculation of Pollution Load

The pollution load from field rainfall runoff can be calculated using the following formula:

$$L_i={\displaystyle {\int }_0^{T_i}{C}_i(t)Q_i(t)dt}$$

(7)

where Lᵢ represents the pollution load (g) of rainfall event i; C_i(t) denotes the instantaneous pollutant concentration (mg/L) in surface runoff at time t during rainfall event i; Q_i(t) refers to the corresponding runoff discharge rate (m³/s) at time t; and T_i is the total duration (s) of rainfall event i.

The Soil and Water Assessment Tool (SWAT) is a watershed-scale model that integrates topography, geology, soil properties, land use, weather conditions, and management practices to simulate hydrological processes and the migration and transformation of associated pollutants. It is particularly suitable for estimating pollution loads from agricultural runoff [16].

3. Results and Discussion

3.1. Characteristics of Runoff Surface Source Pollution

3.1.1. Characteristics of the Distribution of Field Rainfall Runoff Pollutant Concentrations in Different Land Patches

Runoff pollution levels were assessed against the Class III water standard specified in China’s Surface Water Environmental Quality Standard (GB 3838-2002), applicable to centralized surface water sources for domestic drinking water. Relevant limits are: TN ≤ 1.0 mg/L, TP ≤ 0.2 mg/L, COD_Cr ≤ 20 mg/L [17].

Figure 2 presents box plots of the event mean concentrations (EMC) of pollutants in rainfall runoff from vegetable plots, orchards, and bamboo forests. The variability in EMC values across these land types follows the descending order: Vegetable Plot > Orchard > Bamboo Forest. Under identical rainfall conditions, this pattern reflects the influence of anthropogenic cultivation intensity. A study examining the impact of runoff nitrogen under different land-uses [18] demonstrated that tillage disrupts soil structure and enhances pollutant transport. The box plots show that both the mean and median EMC values for each land type are skewed toward the lower quartile, indicating a higher frequency of EMC occurrences near the minimum observed values. This distribution pattern aligns with the findings of Zhou et al. [19], who reported similar clustering of EMC values in the lower half of the box plot in their investigation of land use optimization effects on phosphorus loss in a small watershed within the Three Gorges Reservoir Area.

3.1.2. Correlation Analysis of Runoff Pollutants in Different Land Patches

Pearson correlation analysis of runoff pollutants across the Vegetable Plot, Orchard, and Bamboo Forest (Figure 3) reveals strong positive correlations among TN, TP, COD_Cr, and SS, indicating a potential common source for these pollutants-a finding consistent with previous studies [20,21]. Additionally, SS exhibits a strong positive correlation with DO and a strong negative correlation with pH. This pattern can be explained by the weakly acidic soil background of the area, where increased erosion and particulate transport elevate SS concentrations, thereby further reducing pH levels. Furthermore, dissolved nitrogen species, including DN, NH₃-N, NO₃-N, and NO₂-N, show strong positive correlations with pH. This relationship stems from the fact that DN, as the sum of NH₃-N, NO₃-N, and NO₂-N, inherently correlates with its constituent forms. Moreover, NO₂⁻-N, as an intermediate in the nitrogen cycle, can be oxidized to NO₃⁻-N or reduced to NH₃-N, leading to close interrelationships among these nitrogen species.

In summary, the strong correlations observed between SS and other water quality parameters suggest that implementing measures to reduce suspended solids in runoff could simultaneously mitigate multiple pollutants. This insight provides a strategic direction for integrated rainfall runoff pollution management.

3.2. Factors Influencing Runoff Pollution Output from Different Land Patches

3.2.1. Impact Analysis of Rainfall Characteristics on Runoff Pollution

Pearson correlation analysis (Figure 4) revealed that runoff pollutants (TN, COD_Cr, and SS) were generally positively correlated with antecedent dry days and negatively correlated with the cumulative rainfall in the preceding seven days, indicating that longer dry periods and lower antecedent rainfall lead to higher pollutant concentrations. This aligns with existing literature, where rainfall parameters such as inter-event duration, intensity, and total amount significantly influence runoff pollution loads [22,23,24,25,26]. In the less disturbed Bamboo Forest, TN and SS showed a positive correlation with rainfall intensity, suggesting that stronger scouring effects enhance pollutant mobilization—a pattern consistent with studies reporting amplified nitrogen and phosphorus losses under higher intensity or longer duration rainfall [27,28,29].

Furthermore, a negative correlation was observed between certain pollutants and mean rainfall temperature, particularly in the sparsely vegetated Vegetable Plot, where higher temperatures increase evaporation and reduce runoff volume. This inverse relationship is supported by studies showing significant negative correlations between temperature and TN in bioretention systems [30], as well as enhanced microbial activity in warmer periods that promotes nutrient removal in constructed wetlands [31,32].

3.2.2. Analysis of the Influence of Soil Adsorption Characteristics on Runoff Pollution

Based on soil phosphorus adsorption experiments, the adsorption isotherms for phosphorus in vegetable garden, orchard, and bamboo forest soils are shown in Figure 5a. The results reveal notable differences in phosphorus adsorption characteristics among the three land-use types, with the strongest adsorption observed in Bamboo Forest soil, followed by Orchard and Vegetable Plot soils. These variations are consistent with studies reporting distinct phosphorus adsorption behaviours under different land uses [33,34]. The nonlinear adsorption trends observed align with findings by Zhai et al., who reported significant disparities in non-point source pollution contributions across land types [35]. In this study, soil adsorption affinity was so strong that saturation was not reached even at the highest initial phosphorus concentration applied. The lower adsorption capacity in the Vegetable Plot and Orchard is likely due to long-term agricultural fertilization, which has led to phosphorus accumulation and the near-saturation of available adsorption sites. This explanation is supported by the high level of available phosphorus detected in the Vegetable Plot (124.25 mg/kg), indicating a strong potential for phosphorus release during rainfall-runoff events, consistent with studies showing a positive correlation between soil available phosphorus and phosphorus mobilization [36]. Such elevated release and subsequent transport of phosphorus pose a considerable contamination risk to adjacent water bodies.

As shown in Figure 5b, the initial phosphorus concentration range for adsorption isotherm experiments was set at 0–100 mg/L. This upper limit substantially exceeded the highest total phosphorus concentration observed in runoff during this study (approximately 25 mg/L), aiming to fully investigate the soil’s maximum adsorption potential and ensure the experimental curve could encompass extreme pollution scenarios potentially encountered in the field. Soil adsorption affinity proved exceptionally strong, failing to reach saturation even at the highest applied initial phosphorus concentration.

The results of quantitative model fitting for adsorption experimental data are presented in

Table 1. Fitting parameters of the Freundlich and Langmuir isotherm models for phosphorus adsorption on soils from different land-use types.

Land-Use Type	Freundlich Model			Langmuir Model
	KF (L/g)	n	R2	KL (L/g)	qm (mg/g)	R2
Vegetable Plot	0.084	0.731	0.951	—	—	—
Orchard	0.194	1.060	0.952	0.111	2.060	0.960
Bamboo Forest	0.371	1.439	0.953	1.133	0.730	0.976

(phosphorus) and

Table 2. Fitting parameters of the Henry linear isotherm model for ammonium nitrogen adsorption on soils from different land-use types.

Land-Use Type	Henry Model
	KH (L/g)	q0 (mg/g)	R2	EC0 (mg/L)
Vegetable Plot	0.065	−0.022	0.997	0.332
Orchard	0.068	−0.036	0.998	0.533
Bamboo Forest	0.066	−0.069	0.997	1.040

(ammonium nitrogen). Regarding phosphorus adsorption (Table 1), the trends observed in the adsorption isotherms of soil for phosphorus indicate a non-linear relationship (Figure 5a), exhibiting characteristics closer to both the Freundlich and Langmuir models. The Freundlich model demonstrated an excellent fit for all three soils (R² > 0.95). The Freundlich adsorption coefficient (K_F = 0.371 L/g) was highest for bamboo forest soil, indicating its strongest adsorption capacity, consistent with its highest position in Figure 5a. The Freundlich exponent n values exceeded 0.7 for all soils, with bamboo forest soil exhibiting n > 1, indicating a favourable adsorption process where adsorption intensity slightly increased with concentration. The Langmuir model also fitted well for orchard and bamboo forest soils, with the bamboo forest soil exhibiting the highest Langmuir constant K_L (1.133 L/g), further confirming its high affinity for phosphorus. However, the phosphorus adsorption data for vegetable garden soil failed to fit the Langmuir model effectively. This may be attributed to long-term fertilization causing highly heterogeneous adsorption sites that tend towards saturation, thereby violating the Langmuir model’s fundamental assumption of homogeneous monolayer adsorption.

For ammonium nitrogen adsorption (Table 2), the isotherms exhibited good linear characteristics within the studied concentration range (Figure 5b), and the Henry linear model was employed for description. The Henry adsorption coefficients (K_H) for all three soils were remarkably similar (0.065–0.068 L/g), indicating comparable fundamental adsorption capacities for ammonium nitrogen. However, the intercepts q₀ were all negative, with the bamboo forest soil exhibiting the most pronounced negative value (−0.069 mg/g). explaining why the curves intersect the x-axis at low concentrations in Figure 5b. This indicates that when ammonium nitrogen concentrations in solution fall below a threshold (i.e., the zero adsorption equilibrium concentration, EC₀), the inherent exchangeable ammonium nitrogen in the soil is released into solution (net desorption). Bamboo forests exhibited the highest EC₀ value (1.040 mg/L), signifying their relatively weak capacity to retain ammonium nitrogen, requiring higher environmental concentrations to initiate net adsorption. This contrasts markedly with phosphorus adsorption results, highlighting fundamental differences in retention mechanisms for nitrogen and phosphorus—two critical nutrient elements—across distinct land-use types.

This observation is consistent with findings by Li Zhuo [37], who reported that cultivated land exhibits a higher ammonia nitrogen adsorption capacity than woodland and shrubland in the Liao River Basin. Within the experimental concentration range, the ammonia nitrogen adsorption isotherms generally followed a near-linear trend. At initial concentrations below 60 mg/L, the Vegetable Plot showed slightly stronger adsorption compared to the Orchard, whereas the trend was reversed at concentrations ≥ 60 mg/L. Additionally, desorption occurred at lower concentrations, resulting in a crossover-type linear isotherm that intersects the concentration axis rather than passing through the origin.

3.2.3. Analysis of the Effect of Sediment on Runoff Pollution

The analysis of

Table 3. Sediment content of runoff water samples under different treatment methods.

Date	Treatment	Hybrid	Precipitation 30 min	Precipitate Removal Rate
9 June 2023	Vegetable plot	640 mg/L	203 mg/L	68.3 percent
	Orchard	245 mg/L	38 mg/L	84.5 percent
	Bamboo forest	864 mg/L	350 mg/L	59.5 percent
23 August 2023	Vegetable plot	2605 mg/L	1978 mg/L	24.1 percent
	Orchard	593 mg/L	162 mg/L	72.7 percent
	Bamboo forest	206 mg/L	137 mg/L	33.4 percent

and Figure 6 reveals that the sedimentation removal efficiency of SS significantly influences the concentrations of TP, TN, and COD_Cr in runoff. Among the three land-use types, the Orchard exhibited the highest SS sedimentation removal rate, attributable to its greater proportion of coarse and fine sand particles (55.92%), which aligns with the principle that larger particles settle more readily, a finding consistent with previous studies linking particle size to settling velocity [38]. Moreover, variations in rainfall intensity influenced particle size distribution; for instance, the higher-intensity event on 9 June 2023 (15.0 mm/h), likely mobilized larger particles and enhanced SS removal [39,40]. Overall, SS reduction corresponded with decreased pollutant levels, with the Bamboo Forest showing the strongest influence of SS on runoff, particularly for TP and COD_Cr, which achieved removal rates of 82.83% and 77.49%, respectively, after centrifugation. Sedimentation and centrifugation treatments further illustrated differences in pollutant–particle associations across land uses. After 30 min of sedimentation, elevated residual concentrations of TP and COD_Cr in the Vegetable Plot, COD_Cr in the Orchard, and TN and COD_Cr in the Bamboo Forest suggested these pollutants are more closely bound to fine particles that resist natural settling. Centrifugation, which removes finer particles, led to increased removal of TP and TN in the Vegetable Plot and the highest elimination of TP and COD_Cr in the Bamboo Forest. These results underscore that pollutants in rainfall runoff are closely tied to sediment transport, where soil erosion not only elevates suspended solids but also promotes the transport of adsorbed nutrients such as nitrogen and phosphorus into water bodies, supporting related findings in riverine studies [41].

A comparative analysis of pollutant concentrations before and after sedimentation indicated a link between pollutant affinity and particle size. To further investigate this, grain size analysis was performed on runoff samples from different land uses (Figure 7a). Results revealed that fine particles below 63 μm dominated the particle size distribution across all land types, accounting for 58.34%, 47.10%, and 39.58% of the mass fraction in the Vegetable Plot, Orchard, and Bamboo Forest, respectively. This prevalence of fine particles is attributed to land management practices, frequent cultivation in the Vegetable Plot, for instance, fragments soil aggregates and enriches the fine fraction in runoff, consistent with findings on agricultural soil structure alteration [42]. Correlation analyses between pollutants (TP, TN, and COD_Cr) and particle size fractions further elucidated carrier–pollutant relationships (Figure 7b–d) and in the Vegetable Plot, TP, TN, and COD_Cr exhibited strong and significant correlations with the <5 μm fraction, indicating that particles below 63 μm are key pollutant carriers, with adsorption capacity increasing as particle size decreases, a pattern consistent with previous studies [43,44]. A similar trend was observed for TP in the Bamboo Forest. In contrast, no significant correlations were detected between pollutants and any particle size fraction in the Orchard. Interestingly, in the Bamboo Forest, TN and COD_Cr correlated more strongly with larger particles (150–200 μm for COD_Cr), suggesting land-use-specific differences in particle–pollutant affinity.

3.2.4. Analysis of the Effect of Vegetation Cover on Runoff Pollution

In addition to the intrinsic soil adsorption processes, field observations further demonstrate the role of surface conditions in mitigating runoff pollution. As shown in

Table 4. Concentration of runoff pollutants in the Orchard under different vegetation cover levels.

Dates	Vegetation Cover	Quantity of Rainfall (mm)	Rainfall and Temperatures (°C)	Runoff Volume (m3)	SS (mg/L)	TP (mg/L)	TN (mg/L)	CODCr (mg/L)
23 April 2023	55.0 percent	17.4	14.4	0.088	4815	7.68	4.80	53.8
11 May 2023	75.0 percent	25.6	17.8	0.169	2043	2.35	3.86	46.8
9 June 2023	85.0 percent	41.2	25.1	0.013	245	1.15	2.47	43
25 June 2023	95.0 percent	97.6	23.2	0.023	65	1.78	1.57	26.6

, increased vegetation cover in the Orchard significantly reduced SS, with concentrations on 11 May and 25 June 2023, being 57.58% and 73.37% lower than those on 23 April and 9 June, respectively, indicating effective control of soil erosion. TP decreased by 69.40% on 11 May compared to 23 April, but increased by 54.70% on 25 June relative to 9 June, a reversal potentially influenced by temperature-dependent release and transformation processes. These results overall demonstrate that vegetation cover contributes to reducing multiple pollutants in runoff, though its efficacy can be modulated by rainfall characteristics and environmental factors such as temperature. This aligns with previous studies reporting that higher vegetation cover enhances the retention of sediment and nutrients in agricultural landscapes [45,46,47].

3.3. Runoff Surface Source Pollution Load Study

In order to assess the applicability of rainfall-runoff models and identify key pollution sources, we analyzed monitoring data across different land-use types on Maozhou Island. The pollution load models for each site, simulated using Minitab software 2024, are summarized in

Table 5. Mathematical model fitting results for pollutant load in runoff from different land parcel plots.

		TP	DP	TN	DN	NO−3-N	NH3-N	CODCr	SS
Vegetable plot	ɑ	16	−5.21	220	−3.18	−4.9	−1.99	9.3	29.1
	β1	−281	−94	−1914	25	144	−19	−56	759
	β2	28	94	1903	−24	−143	19	56	−759
	β3	282	93	1930	−25	−145	19	56	−761
	β4	−5.3	2.07	−80	1.76	2.42	1	−1.5	−7.5
	β5	1.03	−0.481	19.3	−0.333	−0.17	−0.167	0.28	4.37
	R2	0.195	0.322	0.309	0.198	0.173	0.123	0.021	0.532
Orchard	ɑ	23.5	−2.66	163	−0.03	−2.43	1.24	11.9	34.8
	β1	−127	−54	−1292	14	192	−6	−11	1049
	β2	126	54	1284	−14	−193	6	11	−1050
	β3	129	54	1304	−14	−192	6	12	−1051
	β4	−8.7	0.23	−59.1	0.53	0.98	−0.19	−2.4	−9.2
	β5	1.72	0.065	14.2	0.03	0.33	0.057	0.31	5.31
	R2	0.366	0.131	0.434	0.239	0.304	0.104	0.126	0.656
Bamboo forest	ɑ	−0.93	−5.49	−2.72	−0.739	0.86	0.81	−9.69	4.33
	β1	243	379	362	−39	−128	−99	610	341
	β2	−242	−377	−361	39.7	128.9	99.7	−608	−340
	β3	−242	−378	−362	39.8	130.2	100.5	−611	−341
	β4	−0.15	0.76	1.38	0.388	−0.76	−0.646	4.71	0.3
	β5	0.58	0.721	0.611	0.026	−0.035	0.175	0.083	0.707
	R2	0.855	0.766	0.916	0.929	0.937	0.891	0.944	0.769

. The simulation results reveal a stark contrast in model performance. In the anthropogenically disturbed Vegetable Plots and Orchards, models based on individual rainfall parameters showed poor goodness-of-fit (0.021–0.056), rendering them unsuitable. In contrast, the minimally disturbed Bamboo Forest exhibited high goodness-of-fit values (0.766–0.944), affirming the model’s reliability. This clear divergence underscores that the level of anthropogenic interference is a critical factor in the model’s applicability.

The R² values presented in Table 5 indicate a strong correlation between pollutant concentrations in the Bamboo Forest and rainfall parameters. The regression equations are statistically significant, confirming the validity of the model. Therefore, the field runoff pollution load for each pollutant in the Bamboo Forest can be estimated using Equation (8):

$$\left\{\begin{array}{l}TP=-0.93\times h^{243}\times T^{-242}\times I^{-242}\times I_{\max }^{-0.15}\times D^{0.58}\\ DP=-5.49\times h^{379}\times T^{-377}\times I^{-378}\times I_{\max }^{0.76}\times D^{0.721}\\ TN=-2.72\times h^{362}\times T^{-361}\times I^{-362}\times I_{\max }^{1.38}\times D^{0.611}\\ DN=-0.739\times h^{-39.1}\times T^{39.7}\times I^{39.8}\times I_{\max }^{0.388}\times D^{0.026}\\ N{O}_3-N=0.86\times h^{-128.5}\times T^{128.9}\times I^{130.2}\times I_{\max }^{-0.76}\times D^{-0.035}\\ N{H}_3-N=0.81\times h^{-99.2}\times T^{99.7}\times I^{100.5}\times I_{\max }^{-0.646}\times D^{0.175}\\ CODcr=-9.69\times h^{610}\times T^{-608}\times I^{-611}\times I_{\max }^{-4.71}\times D^{0.083}\\ SS=4.33\times h^{341}\times T^{-340}\times I^{-341}\times I_{\max }^{0.3}\times D^{0.707}\end{array}\right.$$

(8)

Based on spatial analysis using the Inverse Distance Weighting (IDW) method in ArcGIS 10.8.2 (Figure 8), the annual runoff pollution load per unit area exhibits a clear spatial structure influenced by land use. As shown in Figure 8a,b, TP and SS loads are highly concentrated in the central agricultural zones (Vegetable Plots and Orchards) due to fertilization and tillage, while the surrounding Bamboo Forest shows significantly lower levels. This pattern is consistent with findings by Hu Xiaoli [48], confirming that agricultural soil disturbance amplifies the export of particulate pollutants. These results underscore the critical influence of land management on sediment and phosphorus transport in runoff.

Beyond particulate pollutants, dissolved and nutrient-type contaminants, including DP, TN, DN, NH₃-N, NO₂-N, and COD_Cr (Figure 8c,d) [49], show more complex distributions, with elevated levels in both agricultural areas and the Bamboo Forest. The high TN and COD_Cr in the Bamboo Forest likely originate from natural processes such as leaf litter accumulation and decomposition. In contrast, dissolved nutrients (DP, DN, NH₃-N, and NO₂-N) may result from the interception and retention of agricultural runoff. Notably, NO₃⁻-N was predominantly retained in the peripheral Bamboo Forest, suggesting this ecotone serves as a functional buffer, intercepting and transforming contaminants from upland areas. This highlights the dual role of bamboo ecosystems as both a source of organic pollutants and a sink for dissolved nitrogen, reinforcing the importance of conserving riparian buffer zones to mitigate non-point pollution.

4. Conclusions

This study unequivocally confirms that runoff pollution from Maozhou Island’s primary land-use types (vegetable fields, orchards, bamboo forests) consistently exceeds water quality standards, with vegetable fields being the predominant contributor to annual pollution loads. Key distinctions lie in phosphorus (highest in vegetable fields) and unique EMC patterns for different nitrogen forms, which are significantly influenced by rainfall characteristics, soil adsorption capacity (bamboo forests strongly adsorb phosphorus, while showing overall weak adsorption for ammonia nitrogen), and suspended solid (SS) concentrations. Crucially, SS exhibits a strong positive correlation with pollutants (TP, TN, COD_Cr), with the finest fraction (<5 μm) carrying the highest pollutant association. These findings underscore the necessity for targeted emission reduction strategies. Guided by the concept of connectivity, it is recommended to focus on intercepting fine particulates (<5 μm) that concentrate pollutants through runoff regulation and enhanced wetland retention functions, thereby effectively controlling non-point source pollution migration from the island. Future management measures should concentrate on reducing soil disturbance in agricultural activity zones such as vegetable plots, whilst utilizing natural or semi-natural ecosystems like bamboo forests as ecological buffer zones to intercept and transform pollutants originating upstream.