ABSTRACT
Phosphorus (P) is transported into the water resources mainly due to soil erosion. Using the structural equation modeling, this study investigated the relative contribution of different P-producing sources in the Pasikhan River watershed, south of the Caspian Sea, Northern Iran. For this purpose, 79 surface soil samples and 14 suspended sediments were taken from the potential P sources and the river. These sources included undisturbed and degraded rangelands, forests, rice fields, tea gardens, and gullies. The built-in structural equations in PLS software were used for modeling purposes. The overall model fitting index (goodness of fit (GOF) = 0.591) showed the model's strong forecasting capability. Degraded rangelands, gullies, rice fields, and tea gardens significantly contribute to P in the river sediments. In particular, the highest contribution was related to degraded rangelands and gullies (intensity of the effect = 0.63 and 0.47). Finally, the results showed that gullies' contribution was 28.26% to the P production in river sediments, while the other sources had a relatively equal contribution. In conclusion, this study elucidates that undisturbed rangelands and forests exhibit negligible contributions to phosphorus levels in river sediments. Overall, the results confirmed that structural equation modeling is a robust and efficient approach to identifying P sources.
HIGHLIGHTS
Mass erosions play a more important role in sediment production than surface erosion.
Structural equation modeling proved to be a promising approach in estimating the origin of phosphorus sources.
Gullies and disturbed rangelands are the major contributors to the production of the river phosphorus.
INTRODUCTION
Population growth, industry developments, agricultural activities, and urbanization have led to severe soil erosion and subsequent water pollution in rivers, lakes, and reservoirs (Varol et al. 2013; Zeinalzadeh & Rezaei 2017; Wen et al. 2019; Ebrahimi et al. 2022a,b). Among the most critical consequences of soil erosion, one can point to soil fertility reduction and the unfavorable transfer of nutrients and organic matter (e.g., N, P, K, and sediments) to the surface waters (Troeh Frederick et al. 2003). In most cases, erosion is considered only in terms of the quantitative soil mass loss in the eroded area (Xingchang et al. 2004; Singh et al. 2008), which is not all that matters. Phosphorus (P) and nitrogen (N) are the two primary nutrients that behave differently in soils during water erosion. Due to the high surface area and adsorption capacity of clay particles, P firmly attaches to the soil particles and moves with them during the erosion process, much more than the soluble form (Hatch et al. 1999). Therefore, the primary mechanism of P loss is in particulate form (with suspended sediment), the values of which fluctuate widely (Hatch et al. 1999; Asadi 2016). The P attached to the surface of the clay particles gradually separates from it and runs into the water in a soluble form (Shoja et al. 2017; Zhou et al. 2020). During surface runoff in cultivated lands, 80% of P losses are in the form of sediment-bound particles (Sharpley et al. 1992). While in runoff that flows through grasslands and meadows, uncultivated soils, and forests, less sediment is transferred, and most downstream P is soluble in water (Sharpley et al. 2003).
Lagoon ecosystems are particularly important among water resource systems and cover about 6% of the planet. They are rich natural ecosystems with various ecological and environmental functions. Widely acknowledged for their extraordinary biodiversity, these ecosystems are serving various ecological and environmental functions (Newton et al. 2018). By serving as habitats for animal and plant species, lagoons contribute significantly to the conservation and preservation of biodiversity. Providing crucial ecosystem services, including nutrient cycling and sediment formation, underscores their importance to human well-being and the overall health of the environment (Rodrigues-Filho et al. 2023). Lagoons offer valuable natural resources, recreational opportunities, food sources for human consumption, and minerals for industrial applications. Additionally, lagoons play a role in protecting coastlines from erosion and storm surges, acting as natural barriers that dissipate wave energy. To ensure the health and sustainability of lagoon ecosystems, it is crucial to implement conservation measures, such as nature-based solutions and adoption of ecosystem-level management plans. These efforts can preserve habitat heterogeneity and ensure the provision of essential services for human welfare and the well-being of the ecosystem (Lloret et al. 2008; Massarelli et al. 2023; Rodrigues-Filho et al. 2023). Due to its remarkable ability to reproduce, Anzali Lagoon, located in northern Iran on the southern shore of the Caspian Sea, is an outstanding international lagoon with a prominent role in ecology, economy, and environment. It is vital to the survival of many plants and animals and is also an essential dwelling for birds, reptiles, amphibians, fishes, and invertebrates. Anzali Lagoon was officially listed in the Ramsar Convention in 1975 (Ramsar Convention Bureau 1975). Birdlife International, a global partnership of non-governmental organizations, has also identified this lagoon as an exceptional bird habitat (Evans 1994).
The practical function of the lagoons in removing nutrients is one of their unique features. In this regard, studies on the Anzali Lagoon have also proved its influential role as an intermediate filter of nutrients between the rivers entering the lagoon and the Caspian Sea. In recent years, unfortunately, due to the excessive entry of nutrients from urban, agricultural, and industrial effluents, the main functions of this lagoon have been seriously endangered, demanding immediate attention and management. This lagoon is included in the list of Montreux (JICA 2012) due to its current status.
One of the major issues associated with the lagoons is eutrophication phenomenon. Eutrophication is a serious threat to lakes and surface waters, mainly due to the pollution induced by human activities. Erosion in upstream lands, the entry of various elements, mainly nitrate and phosphate, into the river, and finally, accumulation in water reserves such as lagoons is the leading cause of eutrophication. It is known that phosphorus is considered a significant limiting factor in the eutrophication of water bodies (Smith et al. 2017; Ni et al. 2019; Moyle & Boyle 2021).
The presence of an excessive amount of nitrogen and phosphorus, slow water flow, moderate temperature, and salinity level in water contribute to eutrophication. Moreover, human interference including agriculture and disposing industrial waste can accelerate the process. Eutrophication results in the excessive growth of plants and algae, which can disrupt the balance of the aquatic ecosystem and decrease water clarity. The elevated levels of nutrients can trigger harmful algal blooms, which produce toxins that are detrimental to aquatic life and can pose risks to human health. Eutrophication can also lead to low-oxygen or anoxic conditions, causing fish mortality and the decline of other aerobic organisms. Its interference with drinking water treatment processes can also lead to health issues (Yang et al. 2008; Chislock et al. 2013). The economic costs associated with eutrophication can be significant, imposing financial burdens on affected communities and industries (Carpenter 2005; Chislock et al. 2013; Costa et al. 2018).
To address the main cause of eutrophication, Asadi (2016) examined the sediment, P, and organic matter of the Pasikhan River; one of the major inlets to the Anzali Lagoon. This researcher stated that 245.3 tons of P enter the Anzali Lagoon annually through the Pasikhan River. He also stated that in wet seasons only 20% of the total P is soluble. However, in dry seasons this contribution reaches 50%, which indicates the significant share of clay particle bonded P.
It is imperative to identify the prominent locations of P origin to take controlling and protective actions to maintain and rehabilitate lagoons, especially the Anzali Lagoon. Although many methods have been exploited to identify the origin of sediments, none targeted nutrients such as P, to the authors' knowledge. This research used the structural equation modeling (SEM) method to trace P for the first time. Therefore, this study aimed to investigate the method of SEM based on partial least squares in determining the sources of P production in the watershed.
MATERIALS AND METHODS
Study region
Land use . | % . | Land use . | % . |
---|---|---|---|
Forest | 61.01 | Rural | 1.45 |
Disturbed rangelands | 24.2 | River | 1.22 |
Undisturbed rangelands | 5.36 | Other | 0.54 |
Tea garden | 3.69 | Gully | 0.30 |
Rice field | 2.23 |
Land use . | % . | Land use . | % . |
---|---|---|---|
Forest | 61.01 | Rural | 1.45 |
Disturbed rangelands | 24.2 | River | 1.22 |
Undisturbed rangelands | 5.36 | Other | 0.54 |
Tea garden | 3.69 | Gully | 0.30 |
Rice field | 2.23 |
To generate the vegetation profile map, the Landsat satellite image captured in August 2,010 over Guilan province was acquired from the Glovis site. Comprising seven distinct bands, this image was processed using ERDAS software to merge the bands and create a comprehensive satellite image of the entire province, which subsequently served as a baseline reference for subsequent analyses. Subsequently, the Landsat satellite image specific to the basins was generated through the combination of bands 1, 2, 3, and 4 using ERDAS software. This resultant image, representing a single-band configuration, was employed in the creation of a plant profile map. The assessment of vegetation coverage within each basin involved a quantitative analysis wherein the number of pixels corresponding to vegetation in a particular area was divided by the total number of pixels and multiplied by one hundred to determine the percentage of vegetation coverage.
The hydrometric method was used to measure the soil particle size distribution (particles smaller than 2 mm) (Gee & Dani 2002). Organic carbon content was measured through the Walkey and Black method (Walkley & Black 1934), to measure phosphatase enzyme in samples, a method exploited by Tabatabai et al. (1994) was used. Available P was measured by the Olsen method at a wavelength of 880 nm by a spectrophotometer (Olsen 1954). Calcium and magnesium were measured by the method of Mehlich (1953). All measurements were carried out in three replicates.
Trophic state
Modeling
Cronbach's alpha and composite reliability
In this section, Cronbach's alpha and combined reliability are evaluated and presented in Table 2. In order to calculate the reliability, the combined reliability criterion is used, which has advantages over the traditional approaches such as Cronbach's alpha. The optimum limit for composite reliability is considered a value greater than or equal to 0.7, according to Nunnally (1978), as presented in Table 2. It can be inferred from the reported values for Cronbach's alpha and composite reliability, that the model has a sufficient level of reliability.
Sources . | Composite reliability . | Cronbach's alpha . |
---|---|---|
Optimum limit | ≤0.70 | ≤0.60 |
Undisturbed rangelands | 0.92 | 0.89 |
Degraded rangelands | 0.93 | 0.90 |
Forest | 0.87 | 0.80 |
Tea garden | 0.91 | 0.86 |
Rice field | 0.91 | 0.89 |
Gully | 0.89 | 0.85 |
Sources . | Composite reliability . | Cronbach's alpha . |
---|---|---|
Optimum limit | ≤0.70 | ≤0.60 |
Undisturbed rangelands | 0.92 | 0.89 |
Degraded rangelands | 0.93 | 0.90 |
Forest | 0.87 | 0.80 |
Tea garden | 0.91 | 0.86 |
Rice field | 0.91 | 0.89 |
Gully | 0.89 | 0.85 |
The Sobel test
VAF index
If the indirect path is significant, both a and b and their product can be inferred to be significant, so the VAF can be evaluated. The variable has no mediation effect if its VAF value is less than 0.2, it has a partial mediating effect if the VAF lies between 0.2 and 0.8, and if it is more than 0.8, it has a full mediation effect.
Evaluation of parameters weight factor
Weight factors resulting from model are presented in Table 3. In this table, clay is considered a mediation parameter. Therefore, the weight factor is not calculated for the clay. The weights are calculated by evaluating the correlation of indices of a structure with their parent structure, and the appropriate value is equal to or greater than 0.4 (Hulland 1999). Kline (2015) also stated that the weight factor is a value between zero and one. If the weight factor for a parameter is less than 0.3, it has a weak correlation with the structure and is dismissed. The weight factor between 0.3 and 0.6 is acceptable and is considered desirable if it is greater than 0.6. Results represented in Table 3 indicate that P and phosphatase have the highest weight factor in each source of P production.
Sources . | Parameter . | Weight factor . | Sources . | Parameter . | Weight factor . |
---|---|---|---|---|---|
Undisturbed rangelands | Clay | Tea garden | Clay | ||
OC | 0.73 | OC | 0.77 | ||
P | 0.88 | P | 0.86 | ||
Ca + Mg | 0.81 | Ca + Mg | 0.89 | ||
Phosphatase | 0.85 | Phosphatase | 0.90 | ||
Degraded rangelands | Clay | Rice field | Clay | ||
OC | 0.76 | OC | 0.82 | ||
P | 0.85 | P | 0.81 | ||
Ca + Mg | 0.85 | Ca + Mg | 0.67 | ||
Phosphatase | 0.87 | Phosphatase | 0.75 | ||
Forest | Clay | Gully | Clay | ||
OC | 0.83 | OC | 0.84 | ||
P | 0.80 | P | 0.87 | ||
Ca + Mg | 0.75 | Ca + Mg | 0.85 | ||
Phosphatase | 0.91 | Phosphatase | 0.88 |
Sources . | Parameter . | Weight factor . | Sources . | Parameter . | Weight factor . |
---|---|---|---|---|---|
Undisturbed rangelands | Clay | Tea garden | Clay | ||
OC | 0.73 | OC | 0.77 | ||
P | 0.88 | P | 0.86 | ||
Ca + Mg | 0.81 | Ca + Mg | 0.89 | ||
Phosphatase | 0.85 | Phosphatase | 0.90 | ||
Degraded rangelands | Clay | Rice field | Clay | ||
OC | 0.76 | OC | 0.82 | ||
P | 0.85 | P | 0.81 | ||
Ca + Mg | 0.85 | Ca + Mg | 0.67 | ||
Phosphatase | 0.87 | Phosphatase | 0.75 | ||
Forest | Clay | Gully | Clay | ||
OC | 0.83 | OC | 0.84 | ||
P | 0.80 | P | 0.87 | ||
Ca + Mg | 0.75 | Ca + Mg | 0.85 | ||
Phosphatase | 0.91 | Phosphatase | 0.88 |
Two validity matrices of latent variable correlation and the Fornell–Larcker criterion (1981) are used to examine the discriminant validity of the measurement model (Table 4). Fornell & Larcker (1981) stated that discriminant validity is acceptable when the amount of average variance extracted (AVE) for each structure is greater than the shared variance between that structure and other structures (i.e., the square of correlation coefficients of structures) in the model. Accordingly, the acceptable divergent validity of a measurement model implies that a structure in the model interacts more with its characteristics than with other structures. The latent variable correlations section in the output file in Smart PLS software is used in this study. For the matrix's main diagonal, the square of the AVE is used.
Sources . | Undisturbed rangelands . | Degraded rangelands . | Forest . | Tea garden . | Rice field . | Gully . |
---|---|---|---|---|---|---|
Fornell–Larcker matrix | ||||||
Undisturbed rangelands | 0.81 | |||||
Degraded rangelands | 0.74 | 0.87 | ||||
Forest | 0.69 | 0.63 | 0.84 | |||
Tea garden | 0.68 | 0.76 | 0.65 | 0.82 | ||
Rice field | 0.55 | 0.44 | 0.66 | 0.60 | 0.90 | |
Gully | 0.53 | 0.43 | 0.47 | 0.73 | 0.62 | 0.85 |
Latent matrix | ||||||
Undisturbed rangelands | 1.00 | |||||
Degraded rangelands | 0.74 | 1.00 | ||||
Forest | 0.69 | 0.63 | 1.00 | |||
Tea garden | 0.68 | 0.76 | 0.65 | 1.00 | ||
Rice field | 0.55 | 0.44 | 0.66 | 0.60 | 1.00 | |
Gully | 0.53 | 0.43 | 0.47 | 0.73 | 0.62 | 1.00 |
Sources . | Undisturbed rangelands . | Degraded rangelands . | Forest . | Tea garden . | Rice field . | Gully . |
---|---|---|---|---|---|---|
Fornell–Larcker matrix | ||||||
Undisturbed rangelands | 0.81 | |||||
Degraded rangelands | 0.74 | 0.87 | ||||
Forest | 0.69 | 0.63 | 0.84 | |||
Tea garden | 0.68 | 0.76 | 0.65 | 0.82 | ||
Rice field | 0.55 | 0.44 | 0.66 | 0.60 | 0.90 | |
Gully | 0.53 | 0.43 | 0.47 | 0.73 | 0.62 | 0.85 |
Latent matrix | ||||||
Undisturbed rangelands | 1.00 | |||||
Degraded rangelands | 0.74 | 1.00 | ||||
Forest | 0.69 | 0.63 | 1.00 | |||
Tea garden | 0.68 | 0.76 | 0.65 | 1.00 | ||
Rice field | 0.55 | 0.44 | 0.66 | 0.60 | 1.00 | |
Gully | 0.53 | 0.43 | 0.47 | 0.73 | 0.62 | 1.00 |
Moreover, based on the latent matrix table (Table 4), it is understood that the correlation of each structure with its constituents is more than that of other structures. As a result, the discriminant validity of the model is confirmed. As inferred from the results, the appropriateness of the convergent validity criterion is confirmed.
Convergent validity and discriminant validity
Table 5 presents the output results of the model for AVE. Convergent validity is another criterion for fitting measurement models in the SEM method. Fornell & Larcker (1981) proposed using the AVE as a measure of convergent validity. The desirable range of AVE is equal to and higher than 0.5. As can be seen, the results of all the studied parameters indicate the appropriateness of the convergent validity criterion – AVE.
Sources . | AVE . |
---|---|
Optimum limit | ≥0.5 |
Undisturbed rangelands | 0.65 |
Degraded rangelands | 0.75 |
Forest | 0.71 |
Tea garden | 0.68 |
Rice field | 0.81 |
Gully | 0.72 |
Sources . | AVE . |
---|---|
Optimum limit | ≥0.5 |
Undisturbed rangelands | 0.65 |
Degraded rangelands | 0.75 |
Forest | 0.71 |
Tea garden | 0.68 |
Rice field | 0.81 |
Gully | 0.72 |
Evaluation of the structural model
After measuring the validity and reliability of the measurement model, the structural model is examined through the relationships between the independent variables. In addition, in this study, the most widely used criteria have been used to evaluate the structural model. These criteria include the coefficient of significance (T-values), the coefficient of determination (R2), and the goodness of prediction coefficient (Q2).
Coefficient of significance (T-values)
Coefficient of determination (R2)
One of the main advantages of the PLS method is the ability to reduce errors in measurement models or increase the variance between structures and indices. The essential point here is that the value of R2 is calculated only for the dependent structures of the model, whereas in the case of independent structures, it is considered zero. The R2 value is in the range of 0–1 and indicates three levels of the fit of the structural model: weak (0–0.19), medium (0.19–0.33), and strong (0.33–0.67) (Hair et al. 2021; Salimi & Nazarian 2022). Table 6 represents the R2 value for each variable. It can be observed that the amount of R2 for all parameters is at a moderate to strong level, except for the clay-undisturbed rangelands relationship.
Endogenous variables . | R2 . | Sources . | Q2 . |
---|---|---|---|
Clay-undisturbed rangelands | 0.17 | Undisturbed rangelands | 0.839 |
Clay-degraded rangelands | 0.45 | Degraded rangelands | 0.521 |
Clay-forest | 0.36 | Forest | 0.458 |
Clay-tea garden | 0.54 | Tea garden | 0.744 |
Clay-rice field | 0.63 | Rice field | 0.388 |
Clay-gully | 0.71 | Gully | 0.354 |
P sediment | 0.65 |
Endogenous variables . | R2 . | Sources . | Q2 . |
---|---|---|---|
Clay-undisturbed rangelands | 0.17 | Undisturbed rangelands | 0.839 |
Clay-degraded rangelands | 0.45 | Degraded rangelands | 0.521 |
Clay-forest | 0.36 | Forest | 0.458 |
Clay-tea garden | 0.54 | Tea garden | 0.744 |
Clay-rice field | 0.63 | Rice field | 0.388 |
Clay-gully | 0.71 | Gully | 0.354 |
P sediment | 0.65 |
Q2 = (1 − SSE/SSO).SSE is the prediction error when using the model prediction. SSO is the Sum of Squares of Observations for each variable.
Q2 index
This criterion, introduced by Geisser (1974), determines the model's predictive power. They believed that models with an acceptable structural fit should be able to predict the characteristics of the endogenous structures of the model. Accordingly, if in a model, the relationships between structures are properly defined, the structures will be able to have a sufficient impact on each other's characteristics, and thus the hypotheses are correctly confirmed. Three ranges of 0–0.02, 0.02–0.15, and 0.15–0.35 for the Q2 value of an endogenous structure correspond to its weak, medium, and strong forecasting capability, respectively (Chin 1998; Salimi & Nazarian 2022). Suppose the value of Q2 for an endogenous structure is zero or less. In that case, it indicates that the relationship between the other structures of the model and that specific endogenous structure is not well explained. As a result, the model needs to be modified. This criterion shows the model's predictive power based on three intensities: weak, medium, and strong. Table 6 presents the results of the Q2 index. According to Table 6, the model has outstanding predictive power because the Q2 index of structures has a magnitude greater than 0.35.
PLS software was used to perform SEM. Due to the insufficient quality of the PLS visualized output, Microsoft PowerPoint was used to reproduce the graphs according to the values and paths.
RESULTS AND DISCUSSION
Trophic state index
Modeling
In the above formula, is an indicator for each structure's shared average used to evaluate the fit of the measurement part of the model. This criterion is used to evaluate the quality of measurement models and shows the variability of the indicators (questions) explained by the related structure. is the average value of R2 of the model's endogenous structures, which is used to examine the fit of the structural part of the model (Hair et al. 2016). Thus, the process through which the GOF for the research model was calculated is as follows. It should be noted that only the shared values of the first-order hidden variables should be included in the calculation of communalities. The values of related to all the model's dependent variables should be considered to calculate , both first and second-order. Table 7 represents the results of communalities and for mediation variables. Given that the GOF is 0.591, the model's overall fit is considered ‘very strong’ according to Wetzels et al. (2009) and Salimi & Nazarian (2022).
Sources . | Communalities . | . |
---|---|---|
Optimum limit | ≥0.50 | ≥0.33 |
Clay-undisturbed rangelands | 0.65 | 0.17 |
Clay-degraded rangelands | 0.75 | 0.45 |
Clay-forest | 0.71 | 0.36 |
Clay-tea garden | 0.68 | 0.54 |
Clay-rice field | 0.81 | 0.63 |
Clay-gully | 0.72 | 0.71 |
P sediment | 0.57 | 0.65 |
Sources . | Communalities . | . |
---|---|---|
Optimum limit | ≥0.50 | ≥0.33 |
Clay-undisturbed rangelands | 0.65 | 0.17 |
Clay-degraded rangelands | 0.75 | 0.45 |
Clay-forest | 0.71 | 0.36 |
Clay-tea garden | 0.68 | 0.54 |
Clay-rice field | 0.81 | 0.63 |
Clay-gully | 0.72 | 0.71 |
P sediment | 0.57 | 0.65 |
.
The greater effect intensity of degraded rangelands compared to forests and agricultural lands can be attributed to several factors. Excessive livestock grazing and poor vegetation of the studied rangelands are among the issues observed during the field studies. Excessive grazing in rangelands reduces vegetation cover and compacts the soil surface, exposing the soil to raindrops and resulting in more runoff and surface soil loss. In addition, livestock weight pressure on the soil surface leads to the destruction of granular soil structure (Bayat et al. 2017) and a subsequent release of P. It is also observed that the effect of paddy origin in P production (0.36) is higher than in tea gardens (0.26). The amount of runoff P in agricultural lands is affected by using P-enriched fertilizer. Also, due to tillage operations in agricultural lands, soil porosity is high, runoff penetrates the soil, and further loss of P is prevented. Also, the residual straw in the agricultural lands after the harvest is effective in the amount of element loss. It reduces the release of elements compared to degraded rangelands and mud. Examining P-producing sources based on the mediation index (clay) also confirms that the two sources of rangelands and forest have no role in the P of sediments.
At high slopes, the surface flow velocity and erosion intensity increase which can be attributed to the decrease in permeability and increase in the runoff volume (Ekwu & Harrilal 2010). In sloping rangelands, changing the land use to dry farming due to reduced vegetation and soil resistance leads to the intensification of water erosion (Peng & Wang 2012). Rangeland ecosystems are affected by climate change, human activities, and management strategies. The results of previous studies indicate that excessive and continuous grazing and consumption of vegetation by livestock reduces soil surface cover and carbon in the soil and increases flow velocity, compaction level, and density, leading to the increased rate of permanent soil erosion and degradation (Marcos et al. 2003; Reeisi et al. 2005).
CONCLUSIONS
This study investigated the contribution of different production sources of phosphorus in the river sediments. The SEM method in PLS software was used to determine the relative contribution of each resource. This study showed that the two sources of undisturbed rangelands and forests do not play an essential role in supplying P out of the Mobarakabad sub-watershed. Phosphorus transfer and its outflow from the watershed are mainly related to four sources: gully, degraded rangelands, rice fields, and tea gardens. Field surveys and aerial photographs have also confirmed that erosion is high in the gullies, causing soil particles to flow out of the watershed. Severe surface erosion is also occurring in the degraded rangelands, which causes the release of particles and P. Therefore, the relative contribution of locations where the erosion rate was high was significant in supplying P to the river sediments. The results showed that the agricultural sector (rice fields and tea gardens) also significantly impacted the outflow of P from the watershed. In these lands, a large amount of phosphate fertilizers is used annually without soil and plant analysis, often much more than the plant needs. Excessive phosphate fertilizers leave the land and enter rivers and water sources due to soil erosion. The modeling results showed that tea gardens are more critical than rice fields in P outflow from the watershed. The tea gardens in this watershed are located mainly on hills and areas with very steep slopes. Due to the high slope of tea lands and heavy rainfall in the region, erosion is high in the lands under tea cultivation, and nutrients such as P are easily transported downstream along with soil particles by runoff. Rice fields are located in the flat section in the studied area, and P is removed from them through surface runoff.
In general, P-producing resources can be divided into two groups for management. The first group consists of degraded rangelands and gullies, where P is mainly a part of their constituent minerals, and erosion and sediment transport are very high. The second group is agricultural lands, where the P is supplied through human activities (fertilization). The type of management strategy that should be taken in each group is different. In the first group, there is a need for physical measures such as tree planting, rangeland management, livestock grazing control, and protection of the gully wall. However, in the second group, it is possible to prevent the excessive use of fertilizer by raising farmers' knowledge and awareness about fertilizers. In general, the results of this study showed that SEM is a very promising approach through which sensitive and critical information about the watershed can be obtained. It helps improve the situation of the watershed more efficiently in terms of time and cost.
FUNDING
This study was funded by the Iran National Science Foundation (ISNF, Grant No. 96003440).
ETHICAL APPROVAL
The authors have considered the subject of plagiarism, and this article is without problem.
AUTHORS’ CONTRIBUTION
E.E. and H.A. conceived of the presented idea. E.E., H.A., and M.R. developed the theoretical framework. E.E., H.A., and H.B. developed the theory and performed the computations. H.A., E.E., and H.B. verified the analytical methods. E.E. and M.R. carried out the experiments. All authors discussed the results and contributed to the final manuscript.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.