Prediction of groundwater quality using transfer functions - an approach to link historical data to future concentrations
Thèse de doctorat : Université de Neuchâtel, 2012.
Groundwater quality has been degraded by anthropogenic activities at many locations. In order to reverse upward concentration trends, environmental programs such as land use changes are frequently implemented. To evaluate the efficiency of such measures, methods linking the evolution of the groundwater quality at public-supply wells to the history of pollutant inputs in their capture zones are... PlusAjouter à la liste personnelle
- Groundwater quality has been degraded by anthropogenic activities at many locations. In order to reverse upward concentration trends, environmental programs such as land use changes are frequently implemented. To evaluate the efficiency of such measures, methods linking the evolution of the groundwater quality at public-supply wells to the history of pollutant inputs in their capture zones are required. The use of numerical flow and transport model is the most comprehensive approach to evaluate and predict pollutant concentration trends. However, the application of this approach is frequently restricted by financial constraints or lack of sufficient data to reliably calibrate such models. Furthermore, the development of a sophisticated numerical model may be disproportionate for an initial estimation of the time-scale of groundwater remediation.
Therefore, this thesis proposes an alternative, simpler approach to evaluate the effect of pollution pressures or remediation programs on the evolution of pollutant concentrations at aquifer outlets. This alternative is based on the transfer function approach which has been commonly applied to the transfer of environmental tracers through aquifers. The method consists of a combination of input concentration trend and travel time distribution represented by a transfer function, using a convolution integral.
While the classical transfer method approach is appealing due to its simplicity, it relies on several assumptions such steady-state hydrodynamic conditions, a homogeneous solute input and a single compartment of solute transfer in the aquifer system. These assumptions limit its use for common scenarios of diffuse pollutions, which often show a spatially distributed input on the catchment and several compartments of transfer within the aquifer, typically the unsaturated and saturated zones. Moreover, steady state conditions are rarely observed.
In an initial part of this thesis, the different methods that can be applied to characterize the transfer functions are reviewed. The analytical models of transfer functions are detailed and new formulations of the exponential model are developed to be applied for input and output zones of limited extent. Furthermore, it is demonstrated that the steady state approximation can be applied when characteristic periods of hydrodynamic fluctuations are smaller than the mean travel time of solute in the system. In a next step, the convolution integral is modified to incorporate spatially varying concentration inputs of diffuse contamination from agriculture. It is demonstrated that the transfer function only needs to be characterized for the section of the catchment in which the contaminant input is modified. Finally, the equation is extended to include distinct transfer compartments of the aquifer system, such as the unsaturated and saturated zone.
The modified transfer function approach is then applied to the Wohlenschwil site, where land use changes have been implemented since 1997. This site is well suited for testing the approach because the land use history is well known and the full cycle of upward concentration trend followed by trend reversal is documented. For this site, analytical and numerical methods are applied to characterize the transfer functions of nitrate. Separate numerical models are established for the unsaturated (HYDRUS) and saturated (FEFLOW) zones, respectively. The analytical method consists in determining the mean transit time of water in the unsaturated and saturated zones of the aquifer, using respectively a water balance approach and the results of artificial tracing, for the parameterization of analytical dispersion models of transfer functions. The nitrate concentration input is reconstructed on the basis of land use history and the recharge rate. Two different recharge scenarios are compared: an average recharge rate and annually varying recharge rate.
Both analytical and numerical methods well reproduce the observed concentration trend. The contribution of each step of land use changes to the water quality improvement is established. A better agreement is found when using annually varying recharge rates. This approach allows reproducing an increasing nitrate concentration trend following a period of low recharge rate.
To summarize, this thesis redefines the framework of application of the transfer function approach. The study demonstrates that the transfer function approach can be applied in most cases of diffuse pollutions affecting groundwater and is particularly suited for remediation programs. The advantages of the transfer function approach are its simplicity and its flexibility, even if it requires a good understanding of the aquifer functioning, in particular the delimitation of the capture zone. The limits of the approach depend on the conditions in which steady-state hydrodynamic conditions can be assumed. Furthermore, the accuracy of the transfer function approach depends on a good quantification of the functions describing inputs and transfers of pollutants in the aquifer system. This study suggests that the choice of one or the other method for the characterization of the transfer functions mainly depends on the availability of data relevant for their implementation. A combination of different methods is also possible, such as the use of analytical models for the unsaturated zone and numerical models in the saturated zone.