1. Introduction
Lakes are very important ecosystems, providing drinking water resources, shipping, flood control, irrigation, aquaculture, tourism, and other functions [
1]. However, lake eutrophication has become a challenging global environmental issue [
2]. Excessive loading of nutrients (primarily phosphorus and nitrogen) from anthropogenic sources is a major cause of water quality impairment and lake eutrophication [
3,
4]. With the development of wastewater treatment facilities, the influence of point-source pollution in the watershed was diminished [
5], but nutrients transported into water bodies from non-point source pollution (NPSP), including agricultural non-point sources, rural domestic wastewater, and urban and rural storm water, began to comprise a larger proportion [
6,
7,
8]. Agricultural non-point source pollution (ANPS) is of particular concern [
9], contributing up to 50% of total nitrogen (TN) and total phosphorus (TP) in China; it is a major reason for the eutrophication of streams and lakes [
10].
ANPS is challenging to trace and control due to its random and intermittent occurrence and the complicated mechanisms and processes that underlie it [
11]. Quantifying agricultural nitrogen (N) and phosphorus (P) losses and elucidating their major pathways is essential to accurately estimate the environmental risks posed by ANPS pollution and for formulating appropriate control strategies [
12].
Field monitoring methods can provide accurate NPSP load estimates. However, they are time-, money-, and labor-consuming, which usually results in a scarcity of monitored data, making these methods ineffective and limited at large scale. Mechanistic models have been widely used to simulate watershed/field scale hydrological processes [
13] and the N and P circulation in various environmental compartments, including the atmosphere and surface water–vadose zone–groundwater system [
14]. Therefore, modeling helps give a better understanding of the migration and transformation of pollutants in the real world [
15]. Various models were used in the NPSP study, which developed from lump to spatially distributed models or from conceptual to numerical models. The classic conceptual models are the (P)USLE model and the Export Coefficient Model (ECM), which are simple to operate and have a reasonable accuracy [
16]. SPARROW is a model between the numerical and conceptual models, but it is not considered a suitable option for NPSP simulation in small-scale watersheds [
17]. The most widely applied numerical models include WEPP, soil and water assessment tool (SWAT), and HSPF, etc. [
14]. Taking SWAT as an example, it can continuously simulate hydrologic and nutrient processes via a range of sophisticated modules (e.g., irrigation, fertilization, tillage, planting, impoundment, and drainage) [
18]. SWAT has been utilized and verified in many NPSP studies in various watersheds [
19,
20,
21].
In previous studies, the impacts of land use change on NPSP generation and discharge have received attention in SWAT simulations. Li et al. explained the contribution of land use type to the nitrogen export to Erhai Lake through SWAT simulation [
22]. Zhang et al. pointed out that cropland had the highest contribution rate to TN and TP in the Three Gorges Reservoir Area [
23]. However, change of land use is usually not very influential regarding nutrient loads, especially regarding agricultural land use. By contrast, when the cropping system is changed, which implies a different plant structure and fertilizer application and utilization, the N and P losses are altered at field scale [
24]. However, few studies have focused on the evaluation of the N and P balance at field scale, or the estimation of the sub-basin N and P loads exported to the watershed, due to scarce data and the lack of a method to use SWAT simulation results.
Erhai Lake, located in Yunnan province, southwest China, is a typical plateau lake, with a long hydrological residential time. The warm temperature and abundant sunlight are favorable for the growth of algae [
25]. With the fast development of the watershed, and intensive human activities generating excessive pollution loads to the lake, Erhai Lake has undergone eutrophication in recent years [
26]. Erhai Lake Basin is a typical agricultural lake basin; it has been reported that the pollution resulting from agriculture and rural households represents more than 60% of the total pollution loads [
27]. The cropping systems are also undergoing dramatic regulation in Erhai Lake Basin, especially since the local government abandoned the garlic plant, and chemical fertilizer has gradually been replaced by organic fertilizer [
28]. After the implementation of these NPSP control measures, the influence of changes in the cropping systems on the NPSP nutrient loads in the basin urgently needs to be understood, to assess the risks and support the protection of Erhai Lake.
In this study, the SWAT model was applied to simulate NPSP loads in the North of Erhai Lake Basin (NELB). The three major objectives were: (1) propose a method to calculate N and P balance at the field scale and apply it in NELB; (2) propose a method to calculate the N and P export intensity at the watershed scale and apply it in NELB; and (3) examine the effects of different cropping systems on the N and P export using scenario analysis. The results are expected to provide a scientific tool for NPSP control measure evaluation and implementation.
3. Results
3.1. N and P Balance at Field Scale
The main agricultural systems in the NELB were classified as Vegeta., R-B/R-G, R-B/R-R, and C-B/C-R. The N and P balances at the field scale are shown in
Figure 4.
(1) N and P application rates
The average N application were 1150.71, 1431.12, 857.22, and 736.10 kg/ha in Vegeta., R-B/R-G, R-B/R-R, and C-B/C-R, respectively. In the four years considered, there was initially a rising trend, followed by a downtrend. In the same agricultural systems, the average P applications were 224.22, 231.76, 231.76, and 352.88 kg/ha, respectively, all with a gradually rising trend.
(2) N and P losses
The average N loss were 873.89, 1240.62, 766.10, and 603.75 kg/ha in Vegeta., R-B/R-G, R-B/R-R, and C-B/C-R, respectively, with proportionally similar losses in different cropping systems. In the same systems, the average P loss was 15.95, 14.26, 14.81, and 24.50 kg/ha, respectively, with a declining and then a rising trend.
(3) N and P balance
The average annual N balance was 276.83, 190.50, 91.11, and 132.35 kg/ha in Vegeta., R-B/R-G, R-B/R-R, and C-B/C-R, respectively. The N utilization rate in these four cropping systems was 24%, 13%, 11%, and 18%, respectively, with a declining trend until 2015, and then an increase in 2016.
The average annual P balance was 208.27, 217.50, 216.95, and 328.38 kg/ha in Vegeta., R-B/R-G, R-B/R-R, and C-B/C-R, respectively. The P utilization rate in these four systems was 93%, 94%, 94%, and 93%, respectively. The P utilization rate in vegetable and R-B/R-G fields increased in 2014, then declined in 2015, and increased again in 2016, whereas the P utilization rate in R-B/R-R and C-B/C-R fields showed a rising trend before 2015, and then declined in 2016.
3.2. N and P Export Intensity to the Watershed
The N export intensity from individual sub-basins to the watershed outlet from 2013 to 2016 is shown in
Figure 5. The average annual N export intensities to the watershed mouth were 1.005, 0.837, 0.537, and 0.649 kg/ha per year from 2013 to 2016, showing a declining trend. The N export intensity varied in the 35 sub-basins, with the maximum N export intensity in the no. 18 sub-basin (11.224, 7.415, 4.928, and 5.398 kg/ha per year in the four consecutive years), and the second was no. 34 (2.246 to 3.320 kg/ha per year during the study periods). Sub-basins 9, 7, 21, 23, and 26 also had a relatively high N export intensity. The sub-basins with the lowest N export intensity were 1, 3, 15, and 16, with a value of 0.001 kg/ha per year. These areas were usually the headwaters of the watershed, thus the long transport route was beneficial for the purification of N.
The average annual sub-basin P export intensities to the watershed outlet were 0.231, 0.141, 0.146, and 0.124 kg/ha per year from 2013 to 2016, with a declining trend (
Figure 6). The sub-basin with maximum P export intensity was no. 34, with values of 2.892, 1.674, 1.439, and 1.197 kg/ha per year from 2013 to 2016. The sub-basin with the second largest P export intensity was no. 18 (0.939, 0.634, 0.644, and 0.658 kg/ha per year from 2013 to 2016). The intensive P export areas were in the southwest, northeast, and downstream of the river basin, and the area reduced in size over the years.
3.3. N and P Load Change under Different Cropping Systems
The comparison of N and P loads under different cropping systems was analyzed using SWAT, and the results are shown in
Figure 7. Under the BAU scenario, the N loads were 754.25, 687.16, 546.22, and 602.81 t from 2013 to 2016, and the P loads were 32.05, 16.81, 16.43, and 13.81 t in the same four consecutive years. After changing the cropping system, the N loads were 722.23, 670.24, 535.66, and 584.14 t, and the P loads were 30.68, 16.22, 15.12, and 12.61 t in the four consecutive years. Compared with the BAU scenario, the N load in S decreased by 4.25%, 2.46%, 1.93%, and 3.10%, and the P load in S decreased by 4.70%, 3.49%, 9.51%, and 8.81% in the four consecutive years.
From the distribution of N and P loads, the average N loads from the upstream, midstream, and downstream areas were 179.82, 166.31, and 301.42 t, respectively, and the P loads from these three sections of the basin were 6.38, 3.40, and 10.00 t. The N discharge intensities were 0.25, 1.38, and 0.96 kg/ha per year, and the respective P discharge intensities were 0.09, 0.28, and 0.32 kg/ha per year.
When the cropping system was changed, the biggest change in N load appeared in the midstream area; compared to BAU, the N load decreased by 11%. A slight decrease of N loads occurred in the upstream and downstream areas, with a decrease of 0.4% and 0.3%, respectively. After the cropping system change, the biggest decrease in P load occurred upstream, with an annual decrease rate of 23.2%, whereas a 14% decrease in P load occurred midstream, and a 6.5% annual P load increase was noted downstream.
5. Conclusions
To characterize the influence of the changing cropping system on ANSP in the agricultural watershed north of Erhai Lake Basin, a SWAT model was built to simulate the pollutant transport processes. Based on the SWAT results, a method to estimate the N and P balances at the field scale and the N and P export intensities from individual sub-basins to the watershed outlet was proposed. The spatial and temporal changes in the N and P loads in the NELB with a change in the cropping system were discussed.
The N balance was largest in the vegetable fields, and the P balance was largest in the C-B/C-R cropping system. Around 20% of the watershed area contributed approximately 60% of the total N or P load, and the mid- and downstream sections in NELB were the major polluted areas. Changing the cropping system could decrease the sub-basin N and P loads to the outlet by 4–9%. The layout of dryland and paddies should be optimized; the mid- and downstream areas of the NELB should adopt more effective NPSP control measures, rather than just changing the cropping system.
This study proposed a method to estimate the N and P balances at the field scale and the sub-basin N and P export contributions to a watershed outlet, based on simulation using the SWAT model. This is beneficial for evaluating the influence of NPSP in data-scarce areas. However, the uncertainties of the SWAT model, either contained in the model simulation principle or arising from the accuracy of the data, should be addressed by improving the monitoring frequency and developing model functions. The effectiveness of NPSP control should be further improved by optimizing the fertilization rates and irrigation in the agroecological systems.