Interpreting Results

A simulation produces multiple 3D maps simultaneously: concentrations, metabolites, isotopic signatures, and residence times. Without a systematic reading method, this abundance of data can quickly become overwhelming. This page provides a structured approach to interpreting outputs, recognizing inconsistencies, and identifying zones or questions where the model provides decisive information.

Reading a Plume by Order of Questions

Rather than examining all outputs at once, it is more efficient to proceed through a sequence of questions. Each question points toward a specific visualization mode and prevents data fragmentation.

Question	Target Visualization Mode	What to Observe
Where is the plume located?	Concentration isosurface (at regulatory threshold)	Footprint, dominant direction, leading edge
Has the plume reached deep levels?	Vertical cross-section along flow axis	Stratification, aquifer/aquitard interfaces
Are there multiple sources?	Horizontal slice at fixed depth	Multiple lobes, discontinuities
Is degradation active?	Parent/metabolite ratio map	Zones with low ratios (high degradation)
Which zones are isotopically active?	Residual δ - Initial δ map	Enriched zones (fractionation via degradation)
How does the plume evolve over time?	Temporal animation	Front progression, stabilization, or recession
What is the mass flux leaving the site?	Surface integration at downgradient boundary	Cumulative mass flux, instantaneous flow rate

Comparison with Measurements

A simulation that fails to reproduce field measurements requires recalibration. Conversely, a simulation that matches them all perfectly is often over-fitted, which can mask weaknesses in predictions outside of sampling points. The recommended approach distinguishes between three comparison zones.

Zone	Validation Criteria	Warning Signal
Calibration Points	Deviation smaller than measurement uncertainty	Higher deviation: inconsistent model
Validation Points (not used for calibration)	Correct order of magnitude	Deviation > factor of 10: model failure
Unsampled Zones	Consistency with known context	Prediction incompatible with geology

The distinction between calibration points and validation points is essential. If all available data points are used to adjust the model, none remain to test its robustness elsewhere. Reserving one or two outlying boreholes as an independent test is a robust practice that requires discipline.

For a well-instrumented site, a pragmatic rule is to use approximately 75% of points for calibration and keep the remainder for validation. The deviation between prediction and measurement at validation points is a much better indicator of model quality than the deviation at calibration points, which can be artificially optimized.

Recognizing Internal Inconsistencies

Some inconsistencies appear directly in the results without requiring external comparison. Detecting them early prevents the publication of an invalid model.

Concentration exceeding the source. If a downgradient point shows a concentration higher than the source, there is a configuration error (underestimated source, undeclared inflow, unstable time step).
Leading edge moving faster than effective velocity. A plume cannot advance faster than v/R. If it does, either velocity is overestimated or the retardation factor R was not applied.
Dominant metabolite without a parent in any zone. A metabolite can only exist where its parent has been present. Complete isolation suggests an initialization error or a kinetic coupling failure.
Isotopic signature inconsistent with concentration. In a high-degradation zone, one expects both a drop in concentration and an enrichment in isotopic signature. If both move in the same direction (dropping concentration without enrichment), it suggests dilution rather than degradation.
Mass balance non-conservation. The engine displays the total mass balance. A significant drift is often the sign of a time step that is too large.

Sensitivity and Uncertainty

A result without an indication of uncertainty is potentially misleading. While the engine does not explicitly calculate global uncertainty, several qualitative indicators help frame its reliability.

Indicator	Interpretation
Variability of known input parameters	Wider literature ranges imply higher uncertainty
Sampling point density	Well-instrumented zones yield more reliable predictions
Measurement reproduction deviation	Poor reproduction = high uncertainty everywhere
Temporal projection horizon	Uncertainty grows the further out we project
Multi-hypothesis comparison	Variation amplitude between plausible scenarios

It is tempting to associate a single simulation with a precise figure and include it in a report as a reference value. This is almost always misleading. A range with low and high bounds, paired with the scenarios that produced them, provides a more honest communication of what the model can actually say.

Distinguishing What the Model Says vs. Does Not Say

A simulation model answers certain questions with reasonable reliability while remaining silent or hazardous on others. Categorizing questions correctly is one of the most valuable skills for a user of this tool.

Type of Question	Typical Reliability
Order of magnitude of plume footprint	High (if hydrogeological conditions are well-defined)
Dominant flow direction	High (if piezometry is available)
Identification of active vs. passive zones	Moderate (requires local geochemical data)
Projected concentrations at a specific point	Moderate to Low (depending on time horizon)
Precise date of threshold exceedance	Low (high sensitivity to parameters)
Point-source origin of existing contamination	Low (if multiple plausible sources exist)

Producing a Synthetic Interpretation

Moving from a raw model to a distributable interpretation requires summarizing what was observed, what was learned, and what remains uncertain. A three-paragraph framework helps structure this synthesis.

What the model shows. Plume shape, active zones, concentration projections. Factual description without interpretation.
What it means. Reading results in operational terms (resource is threatened, natural attenuation is sufficient, remediation is necessary).
What remains open. Poorly constrained parameters, zones without data, hazardous projections beyond a certain horizon.

This three-step structure aligns perfectly with the **FAIH** philosophy (Factual, Analytical, Interpreted, Honest) used in the IsoFind Reports module. Drafting your interpretation in this framework during the simulation phase greatly simplifies the final report writing.

Learn More

Multi-Hypothesis: Framing uncertainty through multiple scenarios.
Export and Sharing: Extracting visualizations for reports.
FAIH Philosophy: A writing framework consistent with this approach.