Responses Of Agricultural Nonpoint Source Pollution To Import-export Balances Of Nitrogen And Phosphorus

Agricultural non-point source pollution, deriving from agricultural production and domestic sewage, poses serious problems including water eutrophication, soil pollution and land degradation. The objectives were to study: ○1 Limitation of the data. Agricultural non-point source pollution is difficult to monitor and regulate because of the extensive, random and diffusive nature, which make it difficult to reflect the characteristics of pollutant transport in heterogeneous landscape by designated monitoring. ○2 Imperfection of the researches. Previous studies have explored the influence of individual watershed factor on nutrient loss and migration process. Correspondingly, there were less studies on the impact of complex landscape factors. ○3 Complexity of impact factors. It is difficult to discuss the factors on nonpoint source pollution due to the complexity of impact factors. This study focused on transport characteristics, dynamic changes and their relationships with non-point source pollution. ○4 Uncertainty of the dimension. In essence, non-point source pollution is resulted from nutrient budget in space and time. There is a need to balance nutrient budget in each landscape unit before discharging into streams.For dealing with diffuse pollution emerged in the Danjiangkou Reservoir, water source area of the Middle Route Project under the South-to-North Water Transfer Scheme, the study presented in this paper was conducted in the typical Hujiashan catchment and undertaken based on reconnaissance field surveys and routine monitoring. Information technologies such as remote sensing and geographic information system were combined with the Bayesian maximum entropy approach, spatial regression model, AnnAGNPS model and nutrient budget model for estimating water quality at unmonitored locations, exploring the relationships between catchment characteristics and stream water quality at different scales, and studying the influence of nutrient budget in the landscape units on non-point source pollution. This study can provide useful information for the process and mechanism of nutrient cycling and migration. The main results and conclusions of the present study are as follows:(1) Based on water quality data of total nitrogen and total phosphorus derived from routine monitoring data, the Bayesian maximum entropy approach using river distances instead of using Euclidean distances was employed to develop a practical approach for estimating the spatio-temporal characteristics of nutrient concentration along unmonitored river segments. The Bayesian maximum entropy approach using river distances was determined to classify nutrient pollution into two levels, of which the first level represents total nitrogen concentration of 1 mg/L and total phosphorus concentration of 0.2 mg/L, whereas the second level represents total nitrogen concentration of 2 mg/L and total phosphorus concentration of 0.4 mg/L. The ranges of total nitrogen and phosphorus were 1.04~6.81 mg/L and 0.01~7.77 mg/L, respectively. Proportion of impaired river segments affected by total nitrogen during the wet season was 89%, which is higher than 85% during the dry season, whereas proportion of impaired river segments affected by total phosphorus displayed the opposite result. The middle and lower catchment suffer from more nutrient pollution.(2) Spatial autocorrelation analysis and spatial regression model were used to identify the degree of spatial autocorrelation and exploring the relationships between pollutant transport and catchment characteristics at different scales. Results show that spatial lag regression models at the 100~150 m buffer riparian scale explained 88% and 89% of the variations for total nitrogen and phosphorus, and well predict the influence of catchment characteristics on nutrient pollutant transport. A negative correlation is found between forest land-use and nutrient concentration, whereas residential land and cultivated land displayed the opposite trend. Spatial regression models overcome the influence of the data collinearity and correlation. In contrast with ordinary least squares regression, spatial regression models can better explain the variations in nutrient concentration and their relationships with catchment characteristics, indicating the better predictive ability.(3) In consideration of landscape patterns at the catchment and 100 m buffer scale, we explored the relationships between landscape characteristics and spatiotemporal variations of stream water quality, by combining Spearman’s rank correlation analysis, stepwise regression analysis and redundancy analysis. The results showed that the ranges of standardized coefficients of variation for ammonia nitrogen and total phosphorus were 69.8-207.6% and 52.0-146.1%, indicating the significant spatiotemporal variations. Cropland and residential land were the primary sources of stream water pollution, which explained 58.6% of variations of NH3-N at 100 m buffer scale. Landscape metrics including contagion, patch densities of forest and residential land, largest patch index of forest and residential land, and aggregation intensities of forest and cropland had significant effects on stream water quality(p<0.05). Landscape metrics in the whole catchment accounted for 71.1-81.6% of total nitrogen and 74.5-83.8% of total phosphorus. Furthermore, seasonal variations of landscape pattern significantly influence stream water quality. In the dry season, landscape metrics can better explain the variations of total nitrogen and total phosphorus, while landscape metrics explained the variation of ammonia nitrogen better in the wet season.(4) Wulongchi catchment has a relatively serious non-point source pollution, which was thus chosen as a case study area for identifying critical pollution source area at the landscape unit scale. Sample analysis and nutrient budget model were combined to determine nutrient concentration in vegetation and soil and calculate nutrient budget in each landscape unit. Results show that the ranges of total nitrogen and phosphorus surplus were 1~946 kg and 1~386 kg. The highest nutrient surplus is distributed in residential land followed by cultivated land, whereas little nutrient surplus in forest and water bodied. Nutrient surplus is distributed in each landscape unit of cultivated land in the first half of the year, of which cole landscape units have the highest nutrient surplus. In the second half of the year, there is little nutrient surplus in sweet potato and peanut landscape units, whereas corn landscape units show loss state.