HR: 14:00h
AN: B13A-03    [Abstracts]
TI: Scaling of Microbial Competence for Sediment Remediation
AU: * Li, M
EM: mengyl@umich.edu
AF: Environmental and Water Resources Engineering, Department of Civil and Environmental Engineering, The University of Michigan at Ann Arbor, 181 EWRE Building 1351 Beal Ave, Ann Arbor, MI 48109-2125 United States
AU: Adriaens, P
EM: adriaens@umich.edu
AF: Environmental and Water Resources Engineering, Department of Civil and Environmental Engineering, The University of Michigan at Ann Arbor, 181 EWRE Building 1351 Beal Ave, Ann Arbor, MI 48109-2125 United States
AB: Reliable characterization of the spatial distribution of sediment site attributes, such as contaminant concentrations or microbial activity depends on how well sampled values represent all values throughout the entire study site. In addition to the reliability of samples described statistically, the physical scale of the samples may further introduce uncertainties. Whereas geostatistical tools have been developed to interpolate the attribute values in space, these do not explicitly take into account the uncertainties associated with the various scales (field, lab, mesocosm) at which the data have been collected. Hence, a model to evaluate uncertainties arising from the various sampling scales, is required to properly sample and interpret data from large sites such as contaminated sediments. Here, we describe a statistical model to optimize the reliability of sampled data on a multi-scale basis. The model not only serves as a tool to evaluate relationships over different scales by their covariances, but also make further use of these covariances as basis for a precision-optimized estimator. Unlike conventional geostatistic tools which are based on the point-to-point spatial structures, the multi-scale model introduces a new framework for spatial analysis in which regional values at different scales are anchored by the correlations of each other. The model is developed using least-squares optimization for estimations of different scales, by which the estimation variance can be evaluated for the estimation of all scales combined. The estimation by the new model is expected to have less unfavored smoothing effect than the conventional kriging approaches in the neighborhood of sampled locations. Preliminary results from a comparison to indicator kriging of a spatial dioxin dataset from the Passaic River indicate that both estimation models agree with regions with lower values, while the multi-scale model preserves the features in the regions where hot-spot measurements exist. Information on the smallest scale appropriate for the dataset is also honored by using the multi-scale model, while kriging approaches gives artificial extrapolation at the near-distance variation despite the sampling scheme of the data set. Uncertainties introduced by sampling equipments can subsequently be analyzed by the multi-scale model after the evaluation of estimation uncertainties completes, to meet the practical end of the model developed. Two evaluation tools for spatial estimation models, cross-validation and jackknifing, are performed on both the multi-scale model and the conventional kriging approach in order to assess the competence of the developed model. Both evaluation tools are similar in concept in that subsamples are removed from the original data set to be estimated by the rest of data points, while focusing differently on either the overall estimation performance or the regional estimation capability. The comparison, using the Passaic River dataset, will be used to inform the applicability and objectivity for both models.
DE: 0400 Biogeosciences
DE: 1869 Stochastic processes
SC: Biogeosciences [B]
MN: 2005 Joint Assembly