The overall objective of this study is to quantify the contributions of nitrogen and phosphorus loads in surface water and identify the critical source areas in a medium-sized scale watershed in the southeast of China. The method of integrating grid-based Geographic Information Systems with empirical models is applied. The curve numbers method is used to estimate surface runoff. The Universal Soil Loss Equation is employed to predict the soil loss in the Jiulong River watershed. Empirical methods for calculating nitrogen and phosphorus discharge are further utilized to calculate loads based on the spatial analysis of a grid system. Spatial distribution of nitrogen and phosphorus loads in terms of sub-watersheds, land-use type and counties/districts in the Jiulong River watershed is represented. Various contributions and sources of nitrogen and phosphorus loads in surface water for each county or district in the Jiulong River watershed are calculated and evaluated. Study results show that excessive fertilizer use and livestock breeding contribute significant nitrogen and phosphorus loads for many counties. The contribution of livestock breeding to these nutrient loads for the Zhangzhou and Xinluo districts were 40 to 47% and 51 to 62%, respectively. Critical source areas and major contributors of non-point source nutrients are identified. The quantitative data of nutrient loads from agricultural non-point sources in the Jiulong River watershed estimated from the study could provide the scientific basis for management and control of nutrients and pollution.

Introduction

With successful control of industrial point source pollution, non-point source (NPS) pollution, especially resulting from agricultural activities, has been identified as a significant source of water-quality pollution in China. In agricultural catchments, such as the Jiulong River watershed in Fujian province, the major NPS pollutants are nitrogen, phosphorus and sediment. Excessive amounts of soluble inorganic phosphorous (P) and nitrogen (N) cause eutrophication in water bodies, which may lead to water-quality issues in water supplies and decrease the diversity of aquatic ecosystems.

One approach to control NPS pollution efficiently is to identify the critical source areas by quantitative evaluation of NPS pollutants. However, processes of NPS pollution are not only very complex and uncertain, but vary spatially and temporally across a catchment in response to a combination of land conditions and topographic factors (Pullar and Springer, 2000). The geographic information system (GIS) has proved to be a viable tool to generate, manipulate, and organize spatial data such as land use, soil, water at the watershed scale. Environmental models have been highly developed since the 1980s and have provided possible solutions with the capability of modelling NPS processes exactly. Many physically based models such as AnnAGNPS and SWAT can simulate the complex processes of NPS, quantify the N and P loads and visualize the spatial distribution of NPS pollutants by coupling with GIS technology (Francos et al., 2001; Baginska et al., 2003; Tripathi et al., 2003). Empirical models such as USLE were typically used for medium-sized watersheds by integrating with grid-based GIS (Sivertun and Prange, 2003; Shi et al., 2004).

Our paper demonstrates that grid-based GIS can be integrated with empirical models by a) evaluating and quantifying the major contributors of N and P loads in a medium-sized agricultural watershed in southeast of China; and b) identifying and describing spatially the critical source areas of NPS pollution in the Jiulong River watershed.

Material and methods

Study watershed

Jiulong River watershed, located in southeast of China (116°46′55∼118°02′ 17 E, 24°23′53∼25°53′ 38N), is the second largest watershed in Fujian province and covers about 14,000 km2. It is situated in a subtropical zone with a monsoon climate. Annual average temperature is 19 to 21C°. Annual precipitation averages 1400 to about 1800 mm, of which 70% occurs between April and September, inclusively. The major agricultural crops are bananas, oranges, litchi, logan and rice. The area has experienced extensive soil erosion and losses of nutrients from intensive agricultural activities and livestock breeding. Agricultural NPS pollution has become serious in recent years.

Models and methods description

There are three major processes of NPS pollution, runoff, soil losses and pollutants discharge. The Soil Conservation Service Curve Number method (SCS-CN) was used to estimate the runoff volume from each grid. The Universal Soil Loss Equation (USLE) was provided to predict soil losses for each grid for annual yield, and then empirical equations to calculate nutrient losses during NPS processes were further applied to simulate N and P loads for each grid in the Jiulong River watershed.

The SCS-CN method, used to estimate runoff volume for each grid in the Jiulong River watershed, is calculated as follows (SCS, 1985):

formula
where Q is runoff volume (in mm), P is rainfall (in mm), and S (in mm) represents an upper limit of the amount of water that can be abstracted by the watershed through surface storage, infiltration, and other hydrologic abstractions. For convenience, S is expressed in terms of a curve number (CN) which is a dimensionless watershed parameter ranging from 0 to 100. A CN of 100 represents a limiting condition of a perfectly impermeable watershed with zero retention and thus all the rainfall becomes runoff. A CN of zero conceptually represents the other extreme, with the watershed abstracting all rainfall with no runoff regardless of the rainfall amount (Gumbo et al., 2002). The watershed parameter CN can be determined from empirical information. The SCS has developed tables of initial CN values, a function of the watershed soil type and land cover/use condition. These are listed in SCS-SA User Manual (Schulze et al., 1992).

The USLE (Wischmeier and Smith, 1978) is applied in a GIS environment to determine the average annual soil loss for each grid in the Jiulong River watershed. The USLE predicts soil loss for a given site as a product of six major factors, whose values at a particular location (e.g., a given grid cell) can be expressed numerically. The soil erosion is calculated as:

formula
where A = annual soil loss in tons ha− 1 yr− 1; R = rainfall erosivity factor (Joules mm m− 2 h− 1); K = soil erodibility factor (tons Joule− 1 mm−l); L = slope length factor; S = slope steepness factor; C = crop and management factor; P = conservation supporting practices factor.

According to research from Jiao (1991), nutrient losses in forms of particulate N (PN) and particulate P (PP) discharge from agricultural non-point source pollution are calculated as follows:

formula
where LSkt in kg ha−1 is the nutrient losses in particulate form; a is a constant; CSkt in kg mg− 1 is the N and P concentration in the top soil layer; and Xkt in tons ha− 1 yr− 1 is the average annual soil loss.

According to Jiao (1991), nutrients in forms of dissolved N and dissolved P discharge from agricultural non-point source pollution are calculated as follows:

formula
where LDkt in kg ha− 1 is nutrient loads in dissolved form; CDkt in mg l− 1, is nutrient concentration in surface runoff; Qkt in mm is the runoff volume.

GIS layers or coverages were spatially organized with the same resolution (the cell size was 100 × 100 m) and coordinate system. Land use and vegetable map were processed and generated from Land sat-TM data by unsupervised classification with ERDAS software and ARC/INFO software. A soil map was obtained from soil surveys at 1:200000 scale. The vector maps were further transformed into raster files with grid cell size of 100 × 100 m for the use in GIS-based USLE processing, whose factors such as C and P should be presented spatially based on land use and crops. The K factor should be described based on the soil map. Digital elevation model (DEM) data of Jiulong River watershed was processed based on 46 DEMs with a scale of 1:50000 by using ARC/INFO GRID. A DEM of the Jiulong River watershed was re-sampled with the appropriate cell size of 100 × 100 m. The L and S factors in the USLE model were generated, the boundary of the Jiulong River watershed was defined and fourteen sub-watersheds were delineated using the DEM and ARC/INFO software (Huang, 2004). Most sub-watersheds' outlet flows were determined by surface water gauges in the Jiulong River watershed (Figure 1).

Model application and evaluation

The values for variables in empirical models were mainly generated from GIS database and monitoring data in the field during storms. The factor Xkt was obtained from applying USLE in GIS environment in the Jiulong River watershed (Huang et al., 2004); CSkt was obtained from soil survey information in the Jiulong River watershed; CDkt was obtained from data monitoring in five typical sub-watersheds of the Jiulong River watershed; Qkt was obtained from SCS-CN method mentioned above. The CNs were adjusted depending on the antecedent moisture condition (AMC) before each storm.

N and P loads from agricultural non-point sources for different sub-watersheds

Spatial variability in different sub-watersheds was obvious in the Jiulong River watershed (Figure 2). The spatial variability in N and P loads related greatly to the extent of agricultural activities and soil losses. The sub-watersheds that exit in Zhangzhou plain, including JL10, JL11, and JL14, use excessive fertilizers for intensive agricultural activities, and showed high levels of dissolved N and P loads.

On the contrary, the sub-watersheds exiting in the mountainous area, especially the Lonyan areas, such as JL1, JL2, JL3, JL4, JL5, which exhibit high elevations, high slope, sufficient rainfall and extensive agricultural activities, show low levels of dissolved N and P loads and have relatively high levels of particulate N and P. Additionally, sub-watersheds, such as JL9 and JL13, the orchards cultivated extensively in hilly areas, suffered from intensive soil erosion and both dissolved and particulate N and P were high.

N and P loads for different land-use type

N and P loads from agricultural non-point source pollution were assessed by different land-use type in the Jiulong River watershed. The value of total N loads for each land use type varies from 1.03 to 40.67 kg ha− 1 y− 1. Among each land-use type, banana and orchard have the higher value in total N loads. This is followed by young orchard, rice land, vegetable land, cropland, and village in turn. Dissolved N occupies a large amount of total N for banana and orchard. The total P loads for each land-use type varies from 0.35 to 4.29 kg ha− 1 y− 1. Among each land use type, banana and young orchard have the higher value in total P loads with rice land, village, shrub, orchard, and woods, following. As far as bananas, young orchard and rice land are concerned, dissolved P occupies a large portion of total P. For bare land, shrub, village and woods, particulate P constitutes a large portion of total P.

N and P loads and sources from agricultural non-point source pollution for different counties

N and P loads and theirs sources from agricultural non-point source pollution were further analyzed and evaluated by different counties and districts in the Jiulong River watershed.

Huang (2004) found four sources led to agricultural non-point source pollution: soil losses, excessive fertilizer use, rural sanitary waste, and livestock breeding. As a whole, for total N, sources from excessive fertilizer use, livestock breeding, rural sanitary waste and soil losses occupied 53.4, 21.0, 13.3, and 12.4%, respectively. For total P, sources from excessive fertilizer use, livestock breeding, soil losses and rural sanitary waste contributed 40.8, 31.4, 14.7 and 13.1%, respectively. Obviously, excessive fertilizer use and livestock breeding are the major contributors for total N and total P. Sources and contributors of N and P loads from NPS vary in different counties in the Jiulong River watershed (Table 1).

Livestock breeding contributes a big proportion in N and P loads for Zhangzhou and Xinluo Districts, which calls for treating effectively the water quality discharged by livestock industry or diminishing the scale of operations.

In runoff from these counties belonging to Zhangzhou, especially the Zhangzhou plain, including Longhai, dissolved N and P loads are higher than among the counties in the Jiulong River watershed. The cause of such phenomenon are mainly because there are many bananas and orchards planted in such areas and fertilizer is overused and beyond the requirement of plant growth to a great extent. The N and P in fertilizer that cannot be taken in by crops, and thus remains in the top soil, is leached during storms.

On the other hand, orchards developed in hilly areas, like Changtai, Pinghe, and Nanjing counties, often utilize excessive fertilizer. The extent of agricultural activities and soil losses is intensive, which leads to both dissolved and particulate N and P having high levels.

Conclusions

This study presented an approach to quantitatively evaluate N and P loads from agricultural NPS in a medium-sized watershed by integrating grid-based GIS with empirical models. The study results show that: a) the grid-based GIS with empirical models were suitable for analyzing relatively large watersheds for critical places, (the critical areas for controlling N and P discharge from agricultural NPS were identified in terms of sub-watersheds, land-use type and counties/districts); and b) the major contributors of agricultural NPS were identified and evaluated in terms of counties/districts.

As a whole, excessive fertilizer use and livestock breeding are the major contributors of total N and total P from agricultural non-point source pollution. The quantitative study results are expected to be helpful for local governments or stakeholders making scientific decisions on the water quality control in the Jiulong River watershed.

Acknowledgements

Authors wish to acknowledge the government of Fujian province for providing financial assistance to conduct this study (No. 2002H009). Monitoring stations of EPA in Zhangzhou city are also acknowledged by authors for providing support to analyze the samples.

References

Baginska, B., Home, W. M. and Cornish, P. S.
2003
.
Modelling nutrient transport in Currency Creek, NSW with AnnAGNPS and PEST
.
Environm. Model. Software
,
18
:
801
808
.
Francos, A., Bidoglio, G., Galbiati, L., Bouraoui, F., Elorza, F. J., Rekolainen, S., Manni, K. and Granlund, K.
2001
.
Hydrological and water quality modeling in a medium-size coastal zone
.
Physics Chem. Earth
,
26
(
1
):
47
52
.
Gumbo, B., Munyamba, N., Sithole, G. and Savenije, H. H. G.
2002
.
Coupling of digital elevation model and rainfall-runoff model in storm drainage network design
.
Physics Chem. Earth
,
27
:
755
764
.
Huang, J. L.
2004
.
Study on Agricultural Non-point Source Pollution of Jiulong River Watershed Based on GIS and Environmental Models
,
Ph.D. thesis
Xiamen University
.
(in Chinese)
Huang, J. L., Hong, H. S. and Zhang, L. P.
2004
.
Study of predicting soil erosion in Jiulong River Watershed based on GIS and USLE
.
J. Soil Wat. Conserv.
,
18
(
5
):
75
79
.
(in Chinese)[CSA]
Jiao, L.
1991
.
The application of USLE and nutrient loss equations on investigation of non-point source pollution in Xihu Lake
.
Environm. Pollu. Control
,
13
(
6
):
5
8
.
(in Chinese)[CSA]
Pullar, D. and Spinger, D.
2000
.
Toward integrating GIS and Catchments Models
.
Environm. Model. Software
,
15
:
451
459
.
Schulze, R. E., Schmidt, E. J. and Smithers, J. C.
1992
.
SCS-SA User Manual PC Based SCS Design Flood Estimates for Small Catchments in Southern Africa
,
Report no. 40
Pietermaritzburg, , South Africa
:
Department of Agricultural Engineering University of Natal
.
SCS (Soil Conversation Service)
.
1985
. “
Hydrology, section 4
”. In
SCS National Engineering Handbook
,
Washington, D.C
:
U.S. Department of Agriculture
.
Shi, Z. H., Cai, C. F., Ding, S. W., Wang, T. W. and Chow, T. L.
2004
.
Soil conservation planning at the small watershed level using RUSLE with GIS: a case study in the Three Gorge Area of China
.
Catena
,
55
:
33
48
.
Sivertun, A. and Prange, L.
2003
.
Non-point source critical area analysis in the Gisselo Watershed using GIS
.
Environm. Model. Software
,
18
:
887
898
.
Tripathi, M. P., Panda, R. K. and Raghuwanshi, N. S.
2003
.
Identification and prioritization of sub-watersheds for soil conservation management using SWAT model
.
Biosyst. Engin.
,
85
(
3
):
365
379
.
Wischmeier, W. H. and Smith, D. D.
1978
.
Predicting rainfall-erosion losses—a guide to conservation planning
,
Agricultural Handbook 537
Washington, D.C
:
United States Department of Agriculture, ARS
.