Fuzzy Systems Modeling of In Situ Bioremediation of Chlorinated Solvents

The full-scale in situ bioremediation demonstration conducted at DOE’s Savannah River Site revealed a wide range of spatial and temporal variations of concentrations of VOCs, enzymes, and biomass in groundwater and vadose zone monitoring boreholes over the field site. One of the powerful modern approaches to analyze uncertain and imprecise data is based on the use of methods of fuzzy systems modeling. Using fuzzy modeling we analyzed the spatio-temporal TCE and PCE concentrations and methanotroph densities in groundwater to assess the effectiveness of different campaigns of air stripping and bioremediation, and to determine the fuzzy relationship between these compounds. Our analysis revealed some details about the processes involved in remediation, which were not identified in the previous studies of the SRS demonstration. We also identified some future directions for using fuzzy systems modeling, such as the evaluation of the mass balance of the vadose zone groundwater system, and the development of fuzzy-ruled methods for optimization of managing remediation activities, predictions, and risk assessment. Savannah River Bioremediation Experiment. In 1992-93, a large-scale vadose zone-groundwater bioremediation demonstration was conducted at the Savannah River Site by injecting several types of gases (ambient air, methane, and nitrous oxide and triethyl phosphate mixtures) through a horizontal well in the groundwater at a 175 ft depth. Simultaneously, soil gas was extracted through a parallel horizontal well in the vadose zone at a 80 ft depth (Hazen et al., 1997). Table 1 presents the regimes of remediation campaigns, involving in situ air stripping (Campaigns 1 and 2) and bioremediation (Campaigns 3-6). Groundwater samples were taken from 11 monitoring wells. According to the conventional statistical analysis of concentrations measured in all wells, the nitrous oxide and triethyl phosphate injection (Campaign 6) appeared to be the most efficient type of bioremediation (Hazen et al., 1997). The demonstration revealed significant special and temporal variations of VOCs, enzymes, and biomass in boreholes over the field site. An example of time variations of TCE concentration is shown in Figure 1. Such variations in concentrations make some of the results seem ambiguous and difficult to be used in describing spatio-temporal behavior of bioremediation processes over the field site. Table 1. Types of injections and corresponding times Campaign Injection Start date End date Duration (days) 1 No air injection air extraction only 2/26/92 3/18/92 21 2 Air injection 3/19/92 4/20/92 32 3 1% CH4 and air 4/21/92 8/5/92 106 4 4% CH4 and air 8/6/92 10/23/92 78 5a Long Pulsing CH4 and air 10/24/92 12/20/92 57 5b Short Pulsing CH4 and air 12/21/92 1/25/93 35 6 Pulsing 4% CH4, air and continuous nitrous oxide and triethyl phosphate 1/26/93 5/1/93 95 The goal of this research is to evaluate the performance of in situ remediation of chlorinated solvents carried out at the Savannah River Site in 1992-93, using fuzzy systems modeling that is well suited to the analysis of imprecise and uncertain measurements of concentrations in monitoring wells, and to illustrate the directions of further application of this method for remediation. Essential Features of Fuzzy Systems Modeling. Fuzzy systems modeling is an effective method to simulate the performance of a system that is uncertain because of vagueness or “fuzziness,” which are inherent in the 1 Staff Scientist, Earth Sciences Division, E.O. Lawrence Berkeley National Laboratory, Berkeley, California 94720, Tel: 510-486-4852, Fax: 510-486-5686, [email protected] (corresponding author) 2 Center & Department Head, Center for Environmental Biotechnology/Microbial Ecology & Environmental Engineering Department, E. O. Lawrence Berkeley National Laboratory, Tel: (510) 486-6223, Fax: (510) 486-7152, [email protected]