Groundwater may be out of human sight, but it is a vital part of the hydrologic system. One way in which scientists seek to understand the hidden processes in the subsurface is to use computer-based models that apply the laws of physics to simulate groundwater flow. A review article published in Reviews of Geophysics explores the benefit of diverse types of observations that can be used to calibrate such models. Here the authors explain some of the field-based and theoretical developments in this field.
What is the purpose of groundwater flow models?
In most cases, the primary purpose of numerical groundwater flow models is providing robust predictions of the future behavior of groundwater. Such predictions are vital for a multitude of groundwater management questions, including the provision of safe drinking water or limiting the distribution of contaminants such as micro-pollutants, waterborne pathogens and antimicrobial resistance genes (Berendonk et al. ).
The protection of groundwater dependent ecosystems can also be facilitated through groundwater flow models. Groundwater flow models are thus key tools for management, policy and research.
What are some of the different types of groundwater flow models?
Groundwater flow models vary greatly in their complexity. A simple model might, for example, simulate the flow of groundwater towards a drinking water well, while a complex model can simulate coupled groundwater and surface water flow across vastly different spatial and temporal scales.
With the latest generation of integrated flow simulators, it is now possible to simulate fully coupled groundwater and surface water systems under consideration of flow, heat and mass transport. Integrated flow models are key for sustainable basin-wide and cross-boundary management of water resources and policy making.
Why is it necessary to calibrate groundwater flow models?
Natural groundwater-surface water systems are very complex and feature heterogeneous physical properties. Our ability to include this complexity in models is limited.
Even where field-based measurements of subsurface properties are available, the complex spatial distributions of these properties are never known. Groundwater flow models, whether they are simple or complex, therefore, always need to be calibrated against different hydrological observations in order to produce robust predictions.
Which hydrological observations have traditionally been used for groundwater flow model calibration?
Observations of groundwater levels and surface water discharge are usually used to calibrate flow models – and for good reason: they can be obtained quickly in large numbers, for a small cost, and at high precision. While they do contain valuable information, from research on groundwater flow modelling done during the past three decades we now know that using only these two classical observation types does not allow the unknown hydraulic properties of groundwater-surface water systems to be sufficiently constrained and, consequently, produces model predictions with considerable uncertainty.
Why is the calibration of models through observations challenging in the case of groundwater flow?
Two aspects of groundwater flow systems create challenges for groundwater flow model calibration. On the one hand, groundwater flow systems are characterized by spatial and temporal heterogeneity in both structural properties as well as the forcings that drive groundwater flow. The hydraulic conductivity of the subsurface, for example, can vary by many orders of magnitude on the scale of just a few meters, while forcings and states such as groundwater recharge, groundwater levels, and groundwater discharge can vary significantly within short periods of time in response to rainfall or snowmelt events, for example.
On the other hand, many of the physical equations which govern groundwater and surface water flow are non-linear. Generally, the more processes a groundwater flow model simulates, the more complex the equations which need to be solved become. As a consequence, more parameters have to be calibrated. Depending on the observation types used for flow model calibration, additional processes need to be considered in the flow model: for example, if observations of tracer concentrations are to be used for calibration, the flow model needs to include the capability of simulating mass transport. This, in turn, increases the number of parameters which need to be calibrated.
While it is often beneficial to include additional observation types, too many additional processes and parameters may result in increased uncertainty because of the increased number of calibration parameters. There thus exists a trade-off between the increased complexity which a flow model needs to address and the benefit of additional observation types. Finding the sweet spot of model complexity is challenging and depends on the specific purpose of each model.
What developments in measurement techniques have enabled more accurate observations for flow model calibration?
Developments in hydrological tracer techniques are of particular note. Stable water isotope analysis is now a standard tool employed in many hydrogeological investigations (Jasechko ). High-resolution measurement technologies such as atom trap traces or low level counting now allow concentrations of the ultra-rare radioactive isotopes of Krypton (81Kr and 85Kr) or of Argon (37Ar and 39Ar) in groundwater to be differentiated from atmospheric background concentrations, filling important gaps in the groundwater residence time analysis toolbox (Loosli and Purtschert ; Schilling et al. ).
Moreover, portable measurement technologies such as portable mass spectrometers now allow simultaneous measurements of inert and reactive gases (He, Ar, Ne, Kr, N2, H2, CO2, O2, CH4) dissolved in groundwater, on-site, in near real-time and for a fraction of the cost of laboratory-based technologies (Brennwald et al. ). While these on-site technologies are lower in precision compared to the high-resolution laboratory-based technologies, the ability to measure dissolved gas time series on-site and in a spatially distributed manner provides insights into the dynamics of groundwater-surface systems beyond any existing technology.
Significant advances have likewise been achieved in airborne technologies such as remote sensing and using unmanned aerial vehicles (Brunner et al. ; Tang et al. ): These have enabled, for example, high-resolution digital terrain models and spatially distributed observations of water temperatures, evapotranspiration, soil moisture or groundwater storage variations to be obtained even in very remote regions of the world.
Meanwhile, geophysical techniques such as ground penetrating radar or electrical resistivity tomography allow structures and hydraulic properties of the subsurface to be inferred in a non-intrusive way, dramatically reducing the necessity for large numbers of expensive boreholes to be drilled.
Recent advances in measurement technologies have therefore increased both direct information on the underlying hydraulic properties of groundwater flow systems as well as information on hydrologic system states, both of which are key information for groundwater model construction. Our review assesses the benefit of using diverse observations for groundwater flow model calibration, focusing on non-classical observation types.
What are some of the unresolved questions where additional research, data or modeling is needed?
Important current challenges around groundwater flow modeling revolve around model complexity. While including diverse observation types is generally beneficial for flow model calibration, research is still needed into how to choose the right degree of model complexity, which optimally balances the gain of information through the inclusion of additional observation types versus the increase of unknowns which the simulation of additional processes introduces.
Another challenge related to model complexity is coping with the high computational demand of fully integrated groundwater-surface water flow models. While integrated flow models have the potential to result in the most insights into the behavior of real-world flow systems, they are computationally so demanding that their application to large spatial and temporal scales or to operational forecasting and controlling is still limited. Research is needed to make fully integrated flow simulators more accessible and more efficient. One interesting current research direction is running flow simulators on now widely-available computational cloud infrastructure (e.g., Kurtz et al. ).
While our review focuses on ways to improve the robustness of models, it is the real-world applications that drive this work. Improved predictions of groundwater flow will serve as a basis for more robust and more sustainable water resources management, including optimizing the abstraction of groundwater for drinking water, maximizing yield while minimizing negative economic and ecological impacts and limiting the propagation of contaminants through hydrologic systems.
—Oliver S. Schilling (email: firstname.lastname@example.org) and Peter Cook, Flinders University, Australia; and Philip Brunner, University of Neuchâtel, Switzerland
Schilling, O. S.,Cook, P., and Brunner, P. (2019), How diverse observations improve groundwater models, Eos, 100, https://doi.org/10.1029/2019EO126933. Published on 05 July 2019.
Text © 2019. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.