Multiple Regression Models for Estimating Pollutant Concentrations at Agricultural Canals

Abstract

Multiple regression models were developed to estimate total nitrogen (T-N), total phosphorus (T-P), and CODconcentrations for agricultural canals in the Sisatannak District, Vientiane, Lao People's Democratic Republic. The observed pollutant concentrations were normalized by a natural logarithmic transformation and were used as a criterion variable. Using the area of residential zone, paddy fields, and the other zone as predictor variables, p-values were statistically significant, 0.034, 0.002, and 0.016 for T N, T-P, and COD, respectively. Multiple R-squared were relatively low, 0.42, 0.58, and 0.48 for the same indices. Using the population, paddy fields, the other area, and a dummy variable that represents pump area as predictor variables, p-values were largely improved, 0.0004, 0.0000, and 0.0000 for T-N, T-P, and COD, respectively. Multiple R-squared were largely improved, 0.71, 0.87, and 0.80 for the same indices.