Data and Model Uncertainties associated with Biogeochemical Groundwater Remediation and their impact on Decision Analysis

S. Pandey, V. V. Vesselinov, D. O'Malley, S. Karra, S. K. Hansen

Research output: Contribution to journalMeeting Abstract


Models and data are used to characterize the extent of contamination and remediation, both of which are dependent upon the complex interplay of processes ranging from geochemical reactions, microbial metabolism, and pore-scale mixing to heterogeneous flow and external forcings. Characterization is wrought with important uncertainties related to the model itself (e.g. conceptualization, model implementation, parameter values) and the data used for model calibration (e.g. sparsity, measurement errors). This research consists of two primary components: (1) Developing numerical models that incorporate the complex hydrogeology and biogeochemistry that drive groundwater contamination and remediation; (2) Utilizing novel techniques for data/model-based analyses (such as parameter calibration and uncertainty quantification) to aid in decision support for optimal uncertainty reduction related to characterization and remediation of contaminated sites. The reactive transport models are developed using PFLOTRAN and are capable of simulating a wide range of biogeochemical and hydrologic conditions that affect the migration and remediation of groundwater contaminants under diverse field conditions. Data/model-based analyses are achieved using MADS, which utilizes Bayesian methods and Information Gap theory to address the data/model uncertainties discussed above. We also use these tools to evaluate different models, which vary in complexity, in order to weigh and rank models based on model accuracy (in representation of existing observations), model parsimony (everything else being equal, models with smaller number of model parameters are preferred), and model robustness (related to model predictions of unknown future states). These analyses are carried out on synthetic problems, but are directly related to real-world problems; for example, the modeled processes and data inputs are consistent with the conditions at the Los Alamos National Laboratory contamination sites (RDX and Chromium).
Original languageEnglish
JournalGeophysical Research Abstracts
StatePublished - 1 Dec 2016
Externally publishedYes


  • 1805 Computational hydrology
  • HYDROLOGYDE: 1819 Geographic Information Systems (GIS)
  • HYDROLOGYDE: 1916 Data and information discovery
  • INFORMATICSDE: 1920 Emerging informatics technologies


Dive into the research topics of 'Data and Model Uncertainties associated with Biogeochemical Groundwater Remediation and their impact on Decision Analysis'. Together they form a unique fingerprint.

Cite this