Premium
Models to Determine First‐Order Rate Coefficients from Single‐Well Push‐Pull Tests
Author(s) -
Schroth Martin H.,
Istok Jonathan D.
Publication year - 2005
Publication title -
groundwater
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.84
H-Index - 94
eISSN - 1745-6584
pISSN - 0017-467X
DOI - 10.1111/j.1745-6584.2005.00107.x
Subject(s) - mixing (physics) , spark plug , plug flow , mathematics , volumetric flow rate , flow (mathematics) , complete mixing , plug flow reactor model , aquifer , set (abstract data type) , soil science , mechanics , statistics , thermodynamics , environmental science , continuous stirred tank reactor , computer science , chemistry , physics , geology , diffusion , geotechnical engineering , geometry , groundwater , programming language , quantum mechanics
Push‐pull tests (PPTs) have been successfully employed to quantify various microbially mediated processes in the subsurface. Current models for determining first‐order rate coefficients ( k ) from PPTs assume complete and instantaneous mixing of injected test solution in the portion of the aquifer investigated by the test, i.e., the system is treated like a well‐mixed reactor. Here we present two alternative models to estimate k that are based on different mixing assumptions, i.e., plug‐flow and variably mixed reactor models. Rate coefficients estimated by the models were compared using a sensitivity analysis and numerical simulations of PPTs. Results indicated that all models yielded reasonably accurate k estimates (errors < 13%), while best accuracy (errors < 1%) was obtained using the variably mixed reactor model. The well‐mixed reactor model generally overestimated true (simulation input) k values, whereas true k values were consistently underestimated by the plug‐flow reactor model. However, estimates of k obtained with the latter models bracketed true k values in all cases. As the variably mixed reactor model is more difficult to apply, we suggest using the well‐mixed and plug‐flow reactor models to obtain intervals for k estimates that will encompass true k values with high certainty. In an example application, we used all models to reanalyze a published PPT data set to obtain k estimates for nitrate consumption in a petroleum‐contaminated aquifer. Similar results were obtained for all three models (relative differences < 10% between k estimates), indicating that all three models are robust tools for estimating k values from PPT experimental data.