Prediction of Uncertainty in Watershed Scale Sediment Provenance Model Using Tracers

Abstract:

The study, conducted on the urbanized Buffalo Bayou Watershed in Harris County, Texas, draws on two methods to quantify and compare the uncertainties in sediment fingerprint or provenance model. There are two sources of uncertainty. One is from the physically based watershed sediment yield model due to spatial variation in the watershed terrain slopes. The other is from the use of long-range episodic rainfall time series data. The work finds the Bayesian Markov Chain Monte Carlo (MCMC) method as more mathematically robust compared with the traditional optimization based Monte Carlo Simulation (MCS) method for prediction of sediment yield fraction estimate uncertainty. Thereby, the work attempts to close the existing gap in research in this area of uncertainty comparison. The link between Markov Chain rainfall time series and the Bayesian MCMC simulation model is established using an erosion process parameter that promotes uncertainty in the statistical soil yield fraction estimates. The research hypothesis was tested using physical and statistical computational models in soil erosion and geostatistics with the use of rain gauge data and soil biogeochemical properties.