Quantification of Uncertainty in Sediment Provenance Model Conference Paper uri icon


  • 2015 ASCE. We quantify and compare model uncertainty derived from sediment provenance or fingerprinting models using mathematical and statistical formulation rooted in the traditional Optimization and Bayesian Markov Chain Monte Carlo Simulation (MCMC) schemes. An ensemble prediction of soil yield percentage estimation from sub-watersheds of the 60 square-mile urbanized Buffalo Bayou Watershed of Houston was accomplished by forcing Markov chain rainfall time series windows generated by USDA's CLIGEN weather generator module of the Water Erosion Prediction Project (WEPP) software. In doing so, we also attempt to present a decision support tool that allows us to tell how much of a land area may be considered in simulation such that the model resolution definitively captures the contribution of soil from different sub-watershed sources. This was done by forcing the Bayesian model with varying lead time rainfall time series. Results for a given watershed contribution area shows that the model uncertainty remains constant after a certain lead time forecast. This allows the user to decide on how much land area to consider and when to stop the simulation on reasonable ground.

name of conference

  • World Environmental and Water Resources Congress 2015

published proceedings


author list (cited authors)

  • Karim, A., Ahmed, I., Boutton, T. W., Strom, K. B., & Fox, J. F.

citation count

  • 1

complete list of authors

  • Karim, A||Ahmed, I||Boutton, TW||Strom, KB||Fox, JF

publication date

  • May 2015