Introduction

Watershed environmental issues, especially water quality and soil erosion are becoming an increasing concern in China and other parts of the world (Wang and others 2006, Chen and others 2006). The soil losses and nonpoint source (NPS) pollutants formations in the middle and upper reaches of Yangtze River in central China are highly vulnerable. Destruction of vegetation has led to soil erosion and NPS pollutions in the upper reaches. In the past 30 years, the forest cover has been reduced to half, while the area exposed to severe erosion doubled in size (Yin and Changan 2001, Zhang 2003). Soil erosion on cultivated land not only results in on-site soil degradation and NPS pollutions, but also causes off-site problems related to downstream sedimentation and water pollution (Zhang and others 2003, Martin and others 2006). After Yangtze River floods in 1998, there was wide agreement that agricultural production on sloped land in upper and middle reaches of Yangtze River was one of the most important causes of rapid runoff and soil erosion in the basin and a major contributor to flooding (Eckhardta and Ulbrich 2003). Since 1999, for preventing soil erosion and water pollution, the Chinese government has carried out the so-called “Conversion Cropland to Forest Policy” (CCFP), which encourages the farmers to return hillside crop cultivation to grassland and forest with detailed compensation and subsidy schemes. From the beginning of the 21st century, 14.7 million ha of cropland have been converted to forest (Zhang and others 2003). Despite enormous scales of CCFP, to date, there have been few assessments of the policy impact on regional sediments and NPS pollution (Wanga and others 2006). In this article, we estimate not only the environmental effects of CCFP in preventing soil erosion and NPS pollutants, but also formulate the regression equations between forestry area ratio and environmental index. Then, the regional sediment, organic nitrogen (N), and organic phosphorus (P) can be effectively estimated with forestry ratio.

To assess the effects of CCFP, field experiments or long-term monitoring is very expensive and time consuming. During the repeated monitoring process, there are uncertainties associated with the results and is very difficult without additional resources. Watershed soil erosion and NPS pollution loading are generally a combined result of many influencing factors including climate, land cover, topography, land uses and soil conditions that characterize the watershed (Tripathi and others 2003). When implementing CCFP, it is quite difficult to assess the environmental quality improvements by extensive sampling and monitoring data. Under these conditions, application models become useful and efficient. The Soil Water Assessment Tool (SWAT) is a watershed simulation model system and can be applied to quantify environmental impacts of CCFP in watershed scale (Behera and Panda 2006). During simulation, the climate, land use, soil, topography, and geological variances are all taken into considerations (Arnold and Fohrer 2005). SWAT has been applied by several authors to study impacts of watershed environmental policies in different perspective. Zhang and Jogensen (2005), for example, have applied the model to assess regional point and NPS pollution reductions for best management practices in Denmark. Attwood and others (2000) have used SWAT to evaluate the impacts of natural resource policy in terms of environmental and economic implications at differing spatial scales.

For the advantages of SWAT model system, it can be used for regional environmental benefits assessment for CCFP implementations. The research motivation is to identify the reductions of NPS pollution with SWAT model, which is the foundation for conversion cropland to forest policy decision making. The objectives of this article, therefore, are to: (Ι) estimate NPS pollutant loading in this typical study area with support of SWAT model, (Π) suggest approach to estimate CCFP long-term benefits for regional NPS pollution reduction at watershed level, and (Ш) analyze the linkage and correlation between NPS pollution load change and the various CCFP implementation levels.

Methods

Model Description

The watershed water quality model, SWAT, was developed by the United States Department of Agriculture (USDA) (Arnold and others 1998). The Soil and Water Assessment Tool was selected for this study because of its ability to simulate land management processes and hydrological responses in larger watershed. The Soil and Water Assessment Tool is a semi-distributed watershed model with a GIS interface that outlines the subbasins and stream networks from a digital elevation model (DEM) and can calculate daily water balances from meteorological, soil and land-use data. The surface runoff is predicted at a daily step by a modification of soil conservation service (SCS) curve number method and the peak runoff rate is estimated according to rational formula. The erosion and sediment yields are estimated in each subbasin with modified universal soil loss equation (Ramanarayanan and others 1996). SWAT applies a multilayer storage routing technique to partition drainable soil water content for each layer into components of lateral subsurface flow and percolation into the layer below.

The SWAT model is built to simulate the physical processes of pollutant in watershed as realistically as possible. Most model inputs are physically based. However, it is important to note that SWAT is not a parametric model with a formal optimization to fit any data. In-stream nutrient dynamics have been incorporated into SWAT using kinetic routines from in-stream water quality model. The N processes, soil pools, plant supply, and demand of N can also be simulated by SWAT. The N Circle, with plant biomass, N transported with runoff, lateral flow, and percolation, can have different N formations estimated daily. The Soil and Water Assessment Tool can also model P circle and formations in similar approach to N. While predicting the amount of soluble P removed in runoff, the labile P concentration in top soil, runoff volume, and P soil-partitioning factor are all considered. Sediment transport of P is simulated with a loading function, as is organic N transport (Santhi and others 2006).

The Soil and Water Assessment Tool was applied all over the world with diverse missions. The basic function is to estimate regional NPS pollution load. In recent years, more applications have been focused on regional managements, pollutant source identifications based on simulation results. Under assumed condition, it can assess regional NPS pollution load changes induced by, for example, climate change. Agriculture has been identified as the major contributor of NPS pollution of water resource (Cheng and others 2007). Simulation of agricultural NPS pollutions in relation to crop patterns is one of the strong traits of SWAT.

Study Area

The study area of Bazhong City, Sichuan province is located in central China and in upper stream of the Yangtze River Basin (Fig. 1). There are 65,959 Km2 cultivated land and 70% of them are cultivated slope land. The soil erosion area in Sichuan is 199,800 Km2, which is 40.87% of the total soil erosion area in Yangtze River watershed. There are about 600 million tons of sediments transported from Sichuan into Yangtze River every year. In this area, the natural vegetation has been destroyed by tillage, grazing, and deforestation, which led to soil erosion and ecosystem degradation; therefore, this region is the core area to implement CCFP. Also, 34% of high to medium productivity hillside cropland has been converted to forest or grassland in the last several years.

The climate of Bazhong City is moderate, with the highest temperature at 40.2° C occurring in July and the lowest average monthly temperature is −4.7° C in January. The average annual temperature is 16.0–16.9° C and average yearly precipitation is 1120.7–1203.1 mm. The whole research area is 960,154.45 ha and the mean elevation is 773 m. The terrain is higher in the north and the maximum and minimum altitude is 2464 m and 332 m respectively. The various land uses in this watershed are: agricultural land (dry land and garden land) (48.0%), mixed forest (9.94%), range-grasses (8.12%), rice (paddy) (6.54%), honey mesquite (5.21%), sesbania (3.87%), and range-brush (0.05%). The main land use is agricultural land. The Paddy is defined as a separate land use type because of its small area and lower slope distributions. Most of the soils are purple soil and lateritic red soil.

Model Inputs

The ArcView Geographic Information System interface of AVSWAT model was used to develop input files. The 1:200 000 geographic database with topography, land use, and soils were constructed (Table 1). The watershed climate condition was simulated from 1996 through 2005 using daily historical weather information collected from seven weather stations in Sichuan Province around the study area (Table 2). The SWAT delineates watershed into subwatersheds based on topography. The land use and soil map in this region was overlaid on subbasins. The land use was classified in seven types and codes of each land use were listed in Table 3.

In the research watershed, there are five types of soils. The dominant soil types in Bahe River watershed are purple soil and lateric red soil. The purple soil occupies more than half of the area and lateric red soil dominates another 41.71% in the north part of the watershed (Fig. 2). The main soil properties are listed in Table  4. The typical management practices information, such as cropping schemes, fertilizer application, and tillage operation were gathered from farmer surveys and local administrative agents.

Table 1 Data type scale and data description/properties
Full size table
Table 2 The geographical features of seven weather stations
Full size table
Table 3 The land use type area and percentage in study area
Full size table
Table 4 Soil properties under different soil series of Bahe River watershed
Full size table

Model Calibration

With reference to actual historical monitoring data on stream flow and water quality, AVSWAT model was calibrated (Arnold and others 1995). The calibration simulation period for flow and water quality was started from January, 2004 to December, 2004. The related SWAT model parameters were adjusted (Table 5) to correct the overestimation of average monthly stream flow. After calibration, the curve number (CN2), plant water uptake compensation factor (EPCO), the soil evaporation compensation factor (ESCO), threshold depth of water in the shallow aquifer for “revap” or percolation to the deep aquifer to occur (REVAPMN), amount of shallow aquifer water that moved into the soil profile (GW_REVAP), and groundwater delay coefficient (GW_DELAY) were determined as listed in. As a result, the simulated water flow was acceptable.

Table 5 The SWAT model parameters adjusted during calibration for stream flow
Full size table

The stream monitoring data in Bazhong watershed includes only total N and total P without distinguishing organic nutrients from mineral components. Therefore, SWAT model was calibrated with the reference to the actual monitoring data of total N and total P. The adjusted SWAT model parameters after calibration are listed in Table 6, including the average slope steepness (SLOPE), average slope length (SLSUBBSN), initial soil organic N concentration (SOL_ORGN), initial soil organic P concentration (SOL_ORGP), organic N enrichment ratio (ERORGN), N percolation coefficient (NPERCO), universal soil loss equation soil erodibility factor (USLE_K1), depth of the top layer of the Aledo soil (SOL_Z1), initial residue cover (RSDIN), N in rainfall (RCN), and biological mixing efficiency (BIOMIX).

Table 6 The SWAT model parameters adjusted during calibration for stream sediment and nutrient loads
Full size table

After calibration, the estimated average monthly N density at the outlet of Bazhong watershed was slightly lower than the actual average monthly N yield in the rainy season and slightly higher in the winter (Table 7). The simulated average monthly P density after calibration has similar variance trend as that of total N.

Conversion Cropland to Forest Scenarios Analysis

In order to estimate the reductions in soil erosion and NPS pollution due to implementation of land use conversion from cropland to grassland/forest policies, four scenarios are constructed representing different conditions of implementation of CCFP in the simulated basin: baseline with (I) no conversion, (II) conversion of cropland with slope greater than 25°, (III) conversion of cropland with slope greater than 15°, and (IV) conversion of cropland with slope greater than 7.5°. Under existing conditions, the evaluation of impacts of CCFP was carried out in four scenarios. The first, Scenario A, was original land use condition, 66.26% of land use was agricultural land and 9.94% was forest. In Scenario B, agricultural land on slope bigger than 25° was transferred into forest and other land use did not change; therefore, agricultural land dropped to 383,429.2 ha and forest area climbed to 349,834.3 ha. In Scenario C, agricultural land on slope bigger than 15° was all converted to forest, which resulted in forest area that rose to 562,971.5 ha and agricultural land that decreased to 170,292.1 ha. In the last scenario, D, agricultural land on slope bigger than 7.5° was turned into forest. As the result, forest area occupied 73.43% of regional land use (706,260.7 ha) and agricultural land was cut down to 27,002.9 ha.

To substantiate the four scenarios and simulate and assess their impacts by SWAT, an analytical framework was developed (Fig. 3). Firstly, the land use distribution characteristics was analyzed according to use type and land slope (Table 8). The average agricultural land slope is 16.49°. The study area is a highly mountainous region with limited suitable arable land resources. Consequently, much of the hillside land is also cultivated, including those with slope greater than 25°. There are 252,761.7 ha agricultural land with slope greater than 25°, which is about 26% of the entire study area. The details of the four scenarios are listed in Table  9.

Table 7 The monthly (M) simulated and monitored results after calibration at watershed outlet
Full size table
Table 8 Land use distributions characteristics with maximum, mean slope and their standard deviation (STD)
Full size table
Table 9 Details of the four scenarios
Full size table

The land use distributions of the four scenarios were also mapped in Fig. 4. Based on these four kinds of land use conditions, the regional water flow, soil erosions, organic N, and organic P loading in 10 years were estimated respectively using the calibrated SWAT model.

Results and Discussion

Stream Flow Variation

The volume of surface runoff under different scenarios is simulated for the whole watershed from 1996–2005. Obviously, the overall pattern of stream flow variation is determined by the variation of precipitation (PRECIP). The highest peak runoff value (331.4 m3/s) simulated by the model was in Scenario C and the lowest in Scenario B. The stream flow are very similar between Scenario A and D despite the different land use structure. With sloppy agricultural lands converted into forest from Scenario A–B, the stream flow did not decrease dramatically. However, when the agricultural land located on a slope larger than 15° was converted into forest (Scenario C), the stream flow climbed intensively, ranging from 112.0 m3/s t o 331.4 m3/s, but not the highest in the first three years. With more lands transferred into forestry, the stream flow dropped to an original situation. The distributions of stream flow value in Scenario A and D were plotted graphically with respect to the same line from 78.5 to 302.9 m3/s.

Transported Sediment Variation

Sediment yield is the amount of overland soil loss due to water erosion in the whole study of watershed, which reflects the integrated response of sediment generation processes and stream processes at watershed scale (Fig. 6). shows the change in annual average sediment generation and precipitation of four scenarios from 1996–2005. The sediment yields exhibit similar trends with the PRECIP. Under the original land use condition, Scenario A, the sediment yields of whole watershed ranged from 1535 × 103 to 4456 × 103 t. By implementing the CCFP in scenario B and C, the sediment yields decreased, but not intensively. However, the estimated results of these two conditions seemed to be the same as in Fig.  5. When the agricultural land on slope bigger than 7.5° was changed into forest in scenario D, the sediment yields dropped sharply and ranged from 569 × 103 to 3037 × 103 t. With the exception of four scenario analysis, the watershed sediment yield decreased as the forest area increased.

Fig. 1
figure 1

Location, topography of study area in China

Full size image
Fig. 2
figure 2

The soil type distribution in the Bahe River watershed

Full size image
Fig. 3
figure 3

The research framework for the NPS pollution response for conversion of cropland to forest

Full size image
Fig. 4
figure 4

Land use distribution about Bahe River watershed of four scenarios

Full size image
Fig. 5
figure 5

Comparison between simulated stream flow of four scenarios and precipitation from 1996–2005

Full size image

Organic N Variation

Comparisons of the assessed organic N concentrations in the runoff water under the four scenarios and the corresponding simulated precipitation values in ten years are shown in Fig. 7. The precipitation has a similar trend with estimated results, which proved the NPS pollutants transportations were mainly controlled by the climate conditions. By the analysis of watershed organic N yields, the yield decreased with the forest area climbing, which demonstrated that the implementing of CCFP to prevent NPS pollutions was feasible. In scenario A, the organic N yield was 67.9 × 104 to 394.1 × 104 kg/ha, which dropped to 63.3 × 104 to 251.7 × 104 kg/ha in scenario D, except the simulation error (1284 × 104 kg/ha) in 1996. The estimated value of scenario C in 1996 was also treated as unreasonable. But the other estimations were seasonable for the reference to the local monitored values. By the simulation, the pollutant prevention benefits of different policy implementations level can be concluded clearly for the simulation differences of the four scenarios. Based on those assessments, the CCFP benefits can be estimated in advance and the environmental quality index can be forecasted, which is the guidelines for management.

Organic P Variation

The simulated decreases in sediment and organic N load indicated for efficiency of CCFP implementations and organic P load was also reflected in the similar estimated differences. The simulated organic P yields under the four scenarios over the 10-year period were mapped in Fig.  8. The organic P yields were the organic P transported with sediment out the watershed during the time step. Similar to organic N yield simulations, the estimated values of organic P of scenario A and B in 1996 ware unreasonable and were treated as errors. In Scenario D, when the agricultural land slope was bigger than 7.5° and was transferred into forest, the organic P yields dropped. The yield was between 92.3 × 103 kg/ha and 758.9 × 103 kg/ha. However, at the same time under the original land cover situation, the yields were ranged from 168 × 103 kg/ha and 2438 × 103 kg/ha. Within the other two situations, the variations were not dramatic, but can prove the benefits of CCFP implementations. Compared to sediment variations and organic N estimations, the organic P yields difference of four scenarios over ten years was not dramatic. One important reason is that parts of the organic N come from residue of the ground in forestry.

Fig. 6
figure 6

Comparison between simulated sediments yields of four scenarios and precipitations from 1996–2005

Full size image
Fig. 7
figure 7

Comparison between simulated organic N value of four scenarios and precipitation from 1996–2005

Full size image
Fig. 8
figure 8

Comparison between simulated organic P of four scenarios and precipitation from 1996–2005

Full size image
Fig. 9
figure 9

Order 2 polynomial trendline between estimated results and forest area ratio

Full size image

Statistical Analysis

The modeling approach was applied to estimate the 10-year impacts of implementing the CCFP in Bazhong City in the Yangtze River basin. The ten years’ average results (sediment, stream flow, organic N, and organic P, respectively) of four scenarios were estimated respectively. With the ten years’ average estimations, the order 2 polynomial trendline was mapped to examine the correlation between the NPS pollutants yields and forest area ratio (Fig. 9). The result shows that the regression between stream flow, sediment, and forest area ratio are not significant. However, regression correlation between simulated organic N, P values, and forest area ratio is highly significant with high r 2 of 0.9966 and 0.9980, respectively. With the forest area increase, the organic N and organic P yields dropped intensively, which proves that the CCFP implementation would be an effective way to prevent and reduce regional NPS pollutions. In scenario D, for example, the organic N yield is 204.1 × 104 kg/ha, a reduction of almost 60% compared to the “baseline” land use conditions. Similarly, the organic P yield decreased from the 1162.0 × 103 kg/ha to 434.0 × 103 kg/ha with a 62.7% reduction. The change of average ten years’ stream flow under the four scenarios was not significant. Yet the impacts of land use conversion from cropland to grassland/forest on the sediments formations and transportations are still significant. With the implementation of CCFP in scenario D, reductions in sediment is about 38.0%. Given the area of CCFP would be implemented, these reductions and benefits of the policy are reasonable at the watershed level.

Conclusions

Land use conversion of hillside cropland to grassland/forest is one of the major national policies that have been adopted by the Chinese government to curb ecological destruction, prevent soil erosion, and reduce NPS pollution. This study attempts to assess quantitatively the impacts of this policy on soil erosion and NPS pollution loading.

The need for implementing CCFP and assessing soil and forest conservation practices was climbing extensively for soil erosion and NPS pollutants management. The modeling approach was applied to estimate ten-year impacts of implementing CCFP in Bazhong City in the Yangtze River watershed. With GIS technologies, the agricultural land at four slope grades were converted to forest, respectively, which indicated typical CCFP implementation levels. The ten years’ average simulation about sediment, stream flow, organic N, and organic P in the four scenarios were calculated. By SWAT modeling, the estimations demonstrated CCFP benefit about soil erosion and NPS pollution at regional scale. The results revealed that the organic N and organic P decreased 42.1% and 62.7%, respectively, when agricultural land with slope greater than 7.5° was transferred to forest. With regression principle between pollutant load and forest area ratio, the environmental benefits of CCFP at any level can be calculated. Quantifications of benefits of soil erosion and water quality were necessary for future planning.

The SWAT modeling approach was proved as an effective tool for decision makers to assess benefits of CCFP at watershed level. SWAT system was useful to identify effects of CCFP applied in a new watershed or to quantify long-term environmental consequence of CCFP in a watershed where they have been already carried out. However, we only discussed the land use conversions in this article. The other land management practices, which also cause better regional environmental quality, were not modeled.