Groundwater Model Communication Decoding Models

Welcome to the fourth installment in our six-part series.

In our previous post, How to Build Your Bridge from Reality to Numbers, we embarked on the essential journey of model construction – skillfully transforming the chaotic complexity of the Real World into the structured form of the Model World, and finally, into the computations of the Math World. We explored the critical steps involved in building that conceptual bridge, ensuring your model’s foundation is sound.

Now, you’ve done the hard work: numbers are flowing, graphs are plotting. But here’s the ultimate question in groundwater model communication: what do these numbers really tell us about the Real World? And how do we effectively share those insights? It’s a deceptively simple question, yet its answer determines whether your powerful model truly serves its purpose or ends up as an uninterpretable digital artifact. Remember, the answer you get from your model tells you what’s happening in your Model World, not directly what’s happening in the Real World. This distinction is absolutely critical.

The Peril of “Modeling Spectacles”

Anthony Starfield offers a brilliant analogy: when you transition from the Real World to the Model World, you essentially put on a pair of “modeling spectacles.” These spectacles simplify reality, filter out noise, and help you focus on what’s important for your specific purpose. The danger—what Starfield calls “the worst problem”—emerges when you forget you’re wearing them. You might present your model’s outputs as definitive Real World truths, overlooking all the simplifications and assumptions embedded within your Model World and Math World.

This can lead to serious miscommunication and a rapid erosion of trust. Imagine presenting a detailed groundwater model result to a design team, only for them to ask, “But what if the aquifer is thicker in the north?” And you’re forced to reply, “Our model is calibrated to an assumed aquifer thickness, so we can’t change it.” Confidence shattered. This is the essence of what Dr. Erica Thompson describes as getting trapped in “Model Land”—mistaking the map for the territory. It can even escalate to model risk, where inappropriate use or misinterpretation leads to negative real-world outcomes.

Escaping “Resulting”: Beyond Outcomes to Process

Another subtle yet powerful cognitive trap awaiting model interpreters is what Annie Duke calls “resulting.” This is judging the quality of a decision (or a model’s usefulness) solely by its outcome, rather than by the quality of the process and the information available at the time the decision was made.

Consider, for instance, a groundwater model designed to simulate the probability of a contaminant plume reaching a sensitive area within a specific timeframe, based on its defined Model World parameters. The Math World might output that, given its assumptions, there’s a 30% modeled probability of the plume reaching that area. If, in the Real World, the plume does reach the area, a knee-jerk reaction might be, “The model was wrong! It’s useless!”

However, the crucial point is that this 30% is a probability calculated within the model’s simplified Math World, reflecting the model’s current understanding and specified uncertainties. It is not necessarily the absolute, inherent probability of the complex, often radically uncertain Real World system itself. If the model clearly communicated this model-derived probability and the decision-makers understood the underlying assumptions and limitations, then the model actually served its purpose by informing a good decision process, regardless of the eventual outcome. The model helped them understand the odds as calculated by its internal logic, guiding their decision-making in the face of irreducible real-world uncertainty.

Embracing Uncertainty: The Spectrum of Possibility

To combat resulting, we need to embrace probabilistic thinking. Instead of chasing a single, definitive number (“the plume will arrive on X date”), our models should ideally convey ranges of possibility and degrees of belief based on their Model World constructions (“our model suggests a Y% probability of arrival between A and B dates, given these assumptions”). As Nordstrom reminds us, “scientific knowledge is a body of statements of varying degrees of certainty – some most unsure, some nearly sure, none absolutely certain.” It’s about being “less wrong.”

The “Who” of Communication: Tailoring Your Message

This brings us to a crucial question: Who is this model for?

If you’re the modeler and also the primary decision-maker, your interpretation process might be largely internal and intuitive. You’re already wearing your “modeling spectacles” and understand their specific filters.

However, in many—if not most—professional scenarios, your groundwater model is intended for others: a client, a regulatory agency, a public stakeholder group, or a senior management team. These individuals are often not specialists in hydrogeology or numerical modeling. For them, effective communication isn’t just important; it’s absolutely indispensable. As Nordstrom points out, it’s the modeler’s responsibility to “translate their technical jargon to everyday language that the public understands.”

So, how do we bridge this communication gap between the precise, often abstract, Math World results and the human brain’s preference for intuition and clarity? This is where insights from cognitive science become invaluable.

Communicating Beyond the Numbers: Patterns and Stories

Daniel Kahneman’s work on System 1 (fast, intuitive) and System 2 (slow, logical) thinking reveals that our brains naturally prefer coherent narratives and vivid examples over abstract statistics. While stories can sometimes mislead (leading to base-rate neglect or framing effects), they are incredibly powerful for conveying understanding.

Gerd Gigerenzer argues that humans are much better at grasping probabilities when presented as natural frequencies. For example, instead of saying, “Our model calculates a 0.01% model-derived probability of this slurry wall failing under these conditions,” try, “For every 10,000 similar slurry wall designs in our model, we calculate 1 to fail.” This simple shift in framing transforms an abstract number into a relatable pattern. This framing helps convey the model’s calculated likelihoods in an intuitive way, making the inherent uncertainty clearer to a diverse audience. Gigerenzer and others also argue for the use of visuals that show patterns rather than individual results because people are naturally good at spotting patterns even when the subject matter is new. So if you want others to understand what you are trying to explain with your mathematical model, then you need to communicate in a way that a non-specialist can understand, with “natural frequencies” and visual patterns.

AEM & Anaqsim: Tools for Effective Communication

This is where the strengths of the Analytical Element Method (AEM) and software like Anaqsim become powerful communication tools. AEM’s inherent transparency, derived from its analytical nature, means that the logic of the groundwater model is often much clearer than with opaque numerical grids. You can directly point to how a construction dewatering strategy or a slope drainage system, represented by an AEM element, directly influences the flow pattern. This clarity helps non-specialists grasp the underlying mechanics more easily.

What If? Dynamic Scenario Exploration

Furthermore, the speed and efficiency of Anaqsim allow you to rapidly generate multiple scenarios and visualizations. Instead of one static prediction, you can quickly show:

How an increase in pumping for construction dewatering might affect uplift pressure patterns.
The range of potential pore pressure changes under different hydrological conditions.
The effectiveness of various slope drainage designs.

These dynamic “what-if” explorations, presented through clear visuals and natural frequencies, allow decision-makers to see the patterns of the system’s behavior and the range of potential consequences in the Real World. This empowers them to make more informed decisions, bridging the gap between your Math World insights and their practical needs, and preventing the “delusion of validation” that can arise from misunderstood model outputs.

In our next post, How to spoil a perfectly good model, we’ll confront the inevitable: despite our best efforts, models can still go wrong. We’ll explore common pitfalls and how to recognize them.

Decoding: Bringing Groundwater Models Back to the Real World

Stay in Contact

Legal

Book to join us at our next informal 1 Hour drop in session