An inversed Bayesian modeling approach for estimating nitrogen export coefficients and uncertainty assessment in an agricultural watershed in eastern China

Quantification information for nonpoint source pollution is required to develop a Total Maximum Daily Load (TMDL) at the watershed scale. All export coefficient models and some complex mechanistic models rely on the pollutant export coefficients to quantify and identify nonpoint source pollution. Typically, pollutant export coefficients are estimated by monitoring field plots or small catchments with a single land-use or by monitoring and statistically analyzing mixed land-use watersheds. However, these approaches underestimate export coefficients by neglecting in-stream pollutant retention and they also neglect spatio-temporal variability for export coefficients within a watershed. In addition, these methods do not address the uncertainty associated with estimations. This study combined the export coefficient model, a stream pollutant load model and a Bayesian statistical method to inversely estimate pollutant export coefficients for multiple land-use types from commonly available stream monitoring data. The efficacy of this inversed Bayesian modeling approach was confirmed by determining total nitrogen (TN) export coefficients for farmland, residential land and woodland for six catchments of the ChangLe River watershed in eastern China covering 72 field observation dates in 2004–2009. In-stream retention processes were considered in estimating TN export coefficients. The temporal (across 72 field observation dates) and spatial (across 6 catchments) variability of TN export coefficients for each land-use type within the watershed are discussed, as well as the uncertainties associated with TN export coefficient estimation. This inversed Bayesian modeling approach overcomes the shortcomings involved in current widely used approaches for estimating pollutant export coefficients; thus it can more efficiently support development of TMDL programs, particularly in circumstance where limited data are available.