Environmental Tort Sourcing: Vetting Hydrogeologists for Groundwater Claims

Environmental expert witness sourcing for groundwater contamination fails in a predictable place. The retained hydrogeologist can describe a plume in plain language but cannot defend the numerical model that puts contaminant mass under the plaintiff's property at a specific time. That model is the load-bearing exhibit in most groundwater tort and coverage disputes, and it is where opposing counsel concentrates the Daubert attack.

Two mechanisms create the exposure. First, groundwater fate-and-transport modeling is a distinct competency from field hydrogeology or generalist geology, and a strong resume in one does not imply the other. Second, under Kumho Tire Co. v. Carmichael, the reliability gatekeeping of Federal Rule of Evidence 702 reaches technical and experience-based experts, not only classic laboratory science. A hydrogeologist offering a transport model is squarely inside that scope. Vetting therefore has to test the modeling chain itself, from conceptual site model through calibrated code output, not just count credentials.

Treat credentials as a filter for baseline qualification, not proof of reliability. The relevant layers:

• State Professional Geologist (PG) licensure. Most licensure exams derive from the ASBOG framework administered through state boards. Confirm the license is active in a state, and ideally in the state governing the site, because some venues expect a locally licensed opinion on subsurface conditions.
• Hydrogeology subspecialty, not generic geology. A PG can be a mineralogist. Confirm the specific discipline is groundwater flow and contaminant transport. The National Ground Water Association's Certified Ground Water Professional (CGWP) and American Institute of Hydrology certifications signal subspecialty depth beyond the base license.
• Professional Engineer (PE) boundary. When the model output drives remedial design, capture-zone pumping, or containment engineering, that work may cross into engineering practice. Confirm either PE credentialing or a clean division of labor with a licensed engineer, so the opinion is not attacked as unlicensed engineering.
• Contaminant-class match. The physics differ by contaminant. Light nonaqueous phase liquids (petroleum, LNAPL) float and biodegrade; dense nonaqueous phase liquids (chlorinated solvents, DNAPL) sink and pool on aquitards; PFAS and dissolved metals behave differently again. Confirm the expert's track record is in your contaminant class, because sorption, degradation, and density behavior are not transferable assumptions.

Credentials that do not map to groundwater transport of your specific contaminant are a resume, not a qualification.

Before any code runs, a competent hydrogeologist builds a conceptual site model (CSM): the narrative of source, pathway, and receptor expressed in physical terms. A numerical model inherits every error in the CSM and then reports the error to three decimals. ASTM E1689 provides the standard guide for developing conceptual site models for contaminated sites, and a defensible engagement will reference that discipline.

Vet the CSM against four transport mechanisms the expert must account for explicitly:

• Advection. Bulk movement of dissolved contaminant with flowing groundwater, set by seepage velocity.
• Dispersion. Spreading of the plume around the average flow path, controlled by dispersivity values that are notoriously easy to inflate.
• Sorption and retardation. Partitioning onto aquifer solids, which slows the contaminant relative to water. The retardation factor depends on the sorption coefficient and aquifer bulk density and porosity.
• Biodegradation or decay. First-order loss of mass over time, often the difference between a plume that reaches a receptor and one that does not.

If the expert cannot state which of these mechanisms dominate at the site and why, the downstream model is precise and wrong. That is the single most common exclusion trigger.

A transport model is only as sound as the site characterization feeding it. Audit the data density, not just the conclusion.

• Gradient control. Flow direction and magnitude require at least three properly surveyed monitoring wells to triangulate a hydraulic gradient. A gradient inferred from one or two wells, or from topography alone, is a red flag. Vertical gradients require nested or clustered wells screened at different depths.
• Hydraulic conductivity from real tests. Conductivity is the master variable for velocity. Confirm it was measured through slug tests or aquifer pumping tests analyzed by accepted methods (for example Theis, Cooper-Jacob, or Bouwer-Rice), not assigned from a textbook table for the soil type.
• Seasonal and temporal coverage. Gradients reverse and water tables move. A single synoptic round can misstate flow direction. Look for multiple gauging events across seasons.
• Sorption inputs. Retardation depends on site-specific fraction of organic carbon and partitioning, ideally from soil analysis rather than generic literature defaults.
• Laboratory defensibility. Concentration data driving calibration should carry documented chain of custody and laboratory QA/QC. Analytical gaps here become impeachment on the model's own inputs.

This is the technical core. The governing physics is the advection-dispersion-reaction equation, and seepage velocity follows Darcy's law scaled by effective porosity. Ask which code was used and why, then test the calibration.

• Code selection. Common, publicly documented tools include USGS MODFLOW for the flow field, MT3DMS or MT3D-USGS for solute transport, MODPATH for particle tracking and capture zones, and SEAWAT where density-driven flow matters. EPA screening tools such as BIOSCREEN and BIOCHLOR, and analytical solutions of the Domenico type, are legitimate for early screening but are not a substitute for a calibrated numerical model in a contested case. Confirm the tool matches the question.
• Calibration and history matching. The model must reproduce observed heads and measured concentrations, not just run. Ask for calibration residuals and the scaled root-mean-square error. A commonly used industry target keeps the scaled error low (often cited near ten percent), but the defensible point is that calibration quality is quantified and disclosed, not asserted.
• Sensitivity analysis. A credible modeler shows how outputs move when conductivity, dispersivity, and decay rate vary. Absence of sensitivity work signals a result tuned to a conclusion.
• Mass balance and numerical hygiene. The model should close its mass balance, and grid discretization should control numerical dispersion so the reported spreading is physical, not an artifact of coarse cells.
• Reproducibility. Insist the native input and output decks exist and can be rerun by an opposing modeler to the same result. A black-box output no one can reproduce is fragile under cross-examination.

The General Electric Co. v. Joiner analytical-gap principle applies directly: the court can exclude an opinion where the leap from data to conclusion is too great. Uncalibrated or unreproducible transport modeling is exactly that gap.

In multi-defendant environmental torts, the decisive question is often not whether contamination exists but whose it is and when it was released. This is a separate competency, and generalist hydrogeologists frequently overreach here.

• Chemical fingerprinting. Distinguishing commingled plumes by contaminant ratios and characteristic compounds.
• Degradation sequencing. Chlorinated solvents degrade along known daughter-product chains (for example PCE to TCE to DCE to vinyl chloride). The ratio of parent to daughter products informs relative plume age and can support or defeat an allocation theory.
• Compound-specific isotope analysis (CSIA). Isotopic signatures can differentiate sources and confirm degradation, strengthening or undercutting a single-source narrative.
• Groundwater age dating. Techniques such as tritium-helium and chlorofluorocarbon dating bracket recharge age, which constrains release timing.

Vet the expert's confidence claims closely. Back-calculating a precise release date from plume geometry carries large uncertainty, and an expert who states an exact year without an uncertainty band is exposing the case to a reliability challenge.

These are the observable signals that a modeling opinion will not survive gatekeeping. Any one warrants escalation before retention:

• A transport model presented with no calibration to site heads or concentrations.
• Dispersivity or decay values set to defaults with no site-specific basis, tuned to reach the desired plume extent.
• Flow direction inferred from fewer than three wells or from topography alone.
• A non-reproducible model where native files are unavailable or the result cannot be independently rerun.
• Opinions offered only in litigation, with methods that have no regulatory or peer-reviewed acceptance.
• Precision overstatement: exact arrival times or release dates with no stated uncertainty.
• Predictions extrapolated well beyond the calibrated period or domain.
• Discipline mismatch between the expert's contaminant experience and the site's contaminant class.

Build reproducibility and disclosure into the engagement so the record is defensible before the first deposition.

• Native model deliverables. Require versioned input and output decks, not just a PDF of results, so the model can be rerun and audited.
• Documented assumptions and sensitivity runs. Every key parameter, its source, and the sensitivity of the conclusion to it should be written into the report, consistent with expert-disclosure obligations.
• Reliance materials list. A complete inventory of data relied upon, closing the gap opposing counsel probes.
• Independence and scope. Confirm no result-driven scoping in the engagement letter, and separate any remedial-design engineering into appropriately licensed hands.

None of this guarantees admissibility, and no vetting protocol should be represented as doing so. What it does is convert a subjective hiring decision into an auditable technical record that a court, an adjuster, or an allocation panel can follow. That is the procurement objective.