An improved source apportionment mixing model combined with a Bayesian approach for nonpoint source pollution load estimation

Abstract

Abstract A nonpoint source (NPS) loads evaluation method based on the Bayesian source apportionment mixing model was established in this research. The model assumed that (1) the pollutant concentration from each source mixed with the others in the monitoring section during transport, (2) transport only considered first-order attenuation, (3) point source pollution had relatively stable emissions, and (4) the measurement error was random, unrelated, and consistent with a normal distribution (mean of 0). All unknown parameters in the model were taken as random variables, and their posterior distributions were derived by Markov chain Monte Carlo procedures based on historical data, literature, and empirical information. The outflow system of the Huaihe River was adopted as a case study to verify the feasibility of the model. Gelman–Rubin, automatic frequency control statistics, and the determination coefficient (R2) verified the reliability. The results showed that the total loads of ammonia nitrogen (NH4+), chemical oxygen demand, total nitrogen, and total phosphorus from NPSs accounted for 16.35–27.58%, 18.78–25.69%, 21.68–29.71%, and 42.11–52.09%, respectively. The parameter sensitivity analysis showed that prior distribution of NPS concentration was the most sensitive one, which should be determined reasonably based on the empirical or historical information.