%A GONG Dian-lin, HONG Xi, ZENG Guan-jun, WANG Yi, ZUO Shuang-miao, LIU Xin-liang, WU Jin-shui %T Prediction of Water Quality in Rivers in Agricultural Regions Typical of Subtropics in China Using Multivariate Linear Regression Model %0 Journal Article %D 2017 %J Journal of Ecology and Rural Environment %R 10.11934/j.issn.1673-4831.2017.06.004 %P 509-518 %V 33 %N 6 %U {http://www.ere.ac.cn/CN/abstract/article_11221.shtml} %8 2017-06-25 %X

Knowledge about relationships of nitrogen (N) and phosphorus (P) concentrations in river water with landscape pattern is the premise of scientific management of catchment water environment. Shiyan River Catchment, in Changsha County, Hunan Province was selected as the object and Pearson correlation analysis, variation partitioning analysis, and multivariable linear regression analysis were applied to the exploration of effects of land use on N and P concentrations in river water at the catchment outlets and development of a model for predicting quality of the river water. Results show that 1) the water in the outlets has long been ruled into the category of Grade V minus in quality, as specified in the National Standard for Surface Water Quality (GB 3838-2002), with the concentration of ammonium-N (NH4+-N), nitrate-N (NO3--N), total-N (TN), dissolved-P (DP), and total-P (TP) varying in the range of 5.67-22.46, 0.76-2.85, 13.41-45.55, 0.86-5.00, and 1.99-9.94 mg·L-1, respectively; 2) N and P concentrations in the river water at the outlets of the catchment were significantly related to land use, landscape, population and livestock density (P<0.05), significantly and positively to proportion of forest land in area, and livestock density, and significantly and negatively to proportions of farmland and residential settlement in area, and population density. Besides they were also significantly and positively related to landscape maximum plaque index (LPI), and negatively to landscape shape index (LSI); 3) N and P concentrations in the river water could be predicted through analysis of land use patterns (farmland, residential area, tea garden, etc.) and landscape pattern indices[patch density (PD), LPI, LSI, contagion index(CONTAG), and landscape segmentation index (DIVISION)], using the multiple linear regression model (calibrated R2:0.132-0.320). The model worked particularly well for predicting NO3--N concentration in the river water. All the findings may serve as an important scientific basis for protection of the water environment and rational programming, utilization and management of the landscapes in agricultural catchments in the subtropical hilly regions of China.