Developing a novel biofiltration treatment system by coupling high-rate infiltration trench technology with a plug-flow porous-media bioreactor.

The sequence of two infiltration steps combined with an intermediate aeration named 'sequential managed aquifer recharge technology (SMART)' proved to be a promising approach to replenish groundwater using treated wastewater effluents or impaired surface waters due to efficient inactivation of pathogens and improved removal of many trace organic chemicals. To minimize the physical footprint of such systems and overcome limitations through site-specific heterogeneity at conventional MAR sites, an engineered approach was taken to further advance the SMART concept. This study investigated the establishment of plug-flow conditions in a pilot scale subsurface bioreactor by providing highly controlled hydraulic conditions. Such a system, with a substantially reduced physical footprint in comparison to conventional MAR systems, could be applied independent of local hydrogeological conditions. The desired redox conditions in the bioreactor are achieved by in-situ oxygen delivery, to maintain the homogenous flow conditions and eliminate typical pumping costs. For the time being, this study investigated hydraulic conditions and the initial performance regarding the removal of chemical constituents during baseline operation of the SMARTplus bioreactor. The fit of the observed and simulated breakthrough curves from the pulse injection tracer test indicated successful establishment of plug-flow conditions throughout the bioreactor. The performance data obtained during baseline operation confirmed similar trace organic chemical biotransformation as previously observed in lab- and field-scale MAR systems during travel times of <13 h.

A model-based analysis of the reactive transport behaviour of 37 trace organic compounds during field-scale bank filtration.

Though bank filtration diminishes the loads of many trace organic compounds (TOrCs) present in the source water, still there is a wide uncertainty on the influence of local environmental conditions on biodegradation processes. This research addresses the fate and transport behaviour of 37 trace organic compounds at a bank filtration site in Germany over a relatively long-time span of six years. Using two-dimensional heat and reactive transport modelling in FEFLOW, TOrCs are classified according to their occurrence in bank filtration wells with a residence time of up to 4 months. We identify 12 persistent compounds, 20 reactive compounds and 5 transformation products formed during aquifer passage. Estimates of first-order biodegradation rate constants are given for six reactive compounds. Minimum biodegradation rate constants (i.e. maximum half-lives) are approximated for eight compounds only present in the surface water. For some compounds, a simple first-order degradation model did not yield satisfactory results and the behaviour appears to be more complex. Processes like sorption, redox- and/or temperature-dependent biodegradation and temperature-dependent desorption are suspected but incorporating these into the model was beyond the scope of this paper that provides an overview for many compounds. Results highlight the ability of the sub-surface to improve the water quality during bank filtration, yet at the same time show the persistence of several compounds in the aquifer.

Exposure-time based modeling of nonlinear reactive transport in porous media subject to physical and geochemical heterogeneity.

Transport of reactive solutes in groundwater is affected by physical and chemical heterogeneity of the porous medium, leading to complex spatio-temporal patterns of concentrations and reaction rates. For certain cases of bioreactive transport, it could be shown that the concentrations of reactive constituents in multi-dimensional domains are approximately aligned with isochrones, that is, lines of identical travel time, provided that the chemical properties of the matrix are uniform. We extend this concept to combined physical and chemical heterogeneity by additionally considering the time that a water parcel has been exposed to reactive materials, the so-called exposure time. We simulate bioreactive transport in a one-dimensional domain as function of time and exposure time, rather than space. Subsequently, we map the concentrations to multi-dimensional heterogeneous domains by means of the mean exposure time at each location in the multi-dimensional domain. Differences in travel and exposure time at a given location are accounted for as time difference. This approximation simplifies reactive-transport simulations significantly under conditions of steady-state flow when reactions are restricted to specific locations. It is not expected to be exact in realistic applications because the underlying assumption, such as neglecting transverse mixing altogether, may not hold. We quantify the error introduced by the approximation for the hypothetical case of a two-dimensional, binary aquifer made of highly-permeable, non-reactive and low-permeable, reactive materials releasing dissolved organic matter acting as electron donor for aerobic respiration and denitrification. The kinetically controlled reactions are catalyzed by two non-competitive bacteria populations, enabling microbial growth. Even though the initial biomass concentrations were uniform, the interplay between transport, non-uniform electron-donor supply, and bio-reactions led to distinct spatial patterns of the two types of biomass at late times. Results obtained by mapping the exposure-time based results to the two-dimensional domain are compared with simulations based on the two-dimensional, spatially explicit advection-dispersion-reaction equation. Once quasi-steady state has been reached, we find a good agreement in terms of the chemical-compound concentrations between the two approaches inside the reactive zones, whereas the exposure-time based model is not able to capture reactions occurring in the zones with zero electron-donor release. We conclude that exposure-time models provide good approximations of nonlinear bio-reactive transport when transverse mixing is not the overall controlling process and all reactions are essentially restricted to distinct reactive zones.

Using travel times to simulate multi-dimensional bioreactive transport in time-periodic flows.

In travel-time models, the spatially explicit description of reactive transport is replaced by associating reactive-species concentrations with the travel time or groundwater age at all locations. These models have been shown adequate for reactive transport in river-bank filtration under steady-state flow conditions. Dynamic hydrological conditions, however, can lead to fluctuations of infiltration velocities, putting the validity of travel-time models into question. In transient flow, the local travel-time distributions change with time. We show that a modified version of travel-time based reactive transport models is valid if only the magnitude of the velocity fluctuates, whereas its spatial orientation remains constant. We simulate nonlinear, one-dimensional, bioreactive transport involving oxygen, nitrate, dissolved organic carbon, aerobic and denitrifying bacteria, considering periodic fluctuations of velocity. These fluctuations make the bioreactive system pulsate: The aerobic zone decreases at times of low velocity and increases at those of high velocity. For the case of diurnal fluctuations, the biomass concentrations cannot follow the hydrological fluctuations and a transition zone containing both aerobic and obligatory denitrifying bacteria is established, whereas a clear separation of the two types of bacteria prevails in the case of seasonal velocity fluctuations. We map the 1-D results to a heterogeneous, two-dimensional domain by means of the mean groundwater age for steady-state flow in both domains. The mapped results are compared to simulation results of spatially explicit, two-dimensional, advective-dispersive-bioreactive transport subject to the same relative fluctuations of velocity as in the one-dimensional model. The agreement between the mapped 1-D and the explicit 2-D results is excellent. We conclude that travel-time models of nonlinear bioreactive transport are adequate in systems of time-periodic flow if the flow direction does not change.

On the validity of travel-time based nonlinear bioreactive transport models in steady-state flow.

Travel-time based models simplify the description of reactive transport by replacing the spatial coordinates with the groundwater travel time, posing a quasi one-dimensional (1-D) problem and potentially rendering the determination of multidimensional parameter fields unnecessary. While the approach is exact for strictly advective transport in steady-state flow if the reactive properties of the porous medium are uniform, its validity is unclear when local-scale mixing affects the reactive behavior. We compare a two-dimensional (2-D), spatially explicit, bioreactive, advective-dispersive transport model, considered as "virtual truth", with three 1-D travel-time based models which differ in the conceptualization of longitudinal dispersion: (i) neglecting dispersive mixing altogether, (ii) introducing a local-scale longitudinal dispersivity constant in time and space, and (iii) using an effective longitudinal dispersivity that increases linearly with distance. The reactive system considers biodegradation of dissolved organic carbon, which is introduced into a hydraulically heterogeneous domain together with oxygen and nitrate. Aerobic and denitrifying bacteria use the energy of the microbial transformations for growth. We analyze six scenarios differing in the variance of log-hydraulic conductivity and in the inflow boundary conditions (constant versus time-varying concentration). The concentrations of the 1-D models are mapped to the 2-D domain by means of the kinematic (for case i), and mean groundwater age (for cases ii & iii), respectively. The comparison between concentrations of the "virtual truth" and the 1-D approaches indicates extremely good agreement when using an effective, linearly increasing longitudinal dispersivity in the majority of the scenarios, while the other two 1-D approaches reproduce at least the concentration tendencies well. At late times, all 1-D models give valid approximations of two-dimensional transport. We conclude that the conceptualization of nonlinear bioreactive transport in complex multidimensional domains by quasi 1-D travel-time models is valid for steady-state flow fields if the reactants are introduced over a wide cross-section, flow is at quasi steady state, and dispersive mixing is adequately parametrized.