Where Is the Plume Right Now?
If you're going to drill monitoring wells, you need to know where to drill. The screening model says “on the centerline.” The heterogeneous model shows the plume bifurcated around a clay lens.
Real ground isn't uniform. An aquifer is made of sand, gravel, silt, and clay deposited over thousands of years by shifting rivers and glaciers. The result is a patchwork of materials with wildly different permeabilities.
Clay lenses are low-permeability layers that deflect groundwater flow the way a boulder deflects a stream. A plume hitting a clay lens splits around it, creating two lobes instead of one smooth blob. Gravel channels do the opposite — they act as preferential flow paths that accelerate the plume in narrow fingers far ahead of the main front.
This is why plumes don't spread as the smooth Gaussian blobs that analytical models predict. At Cape Cod, the USGS tracer test showed a plume that was supposed to be elliptical had actually split into two distinct lobes separated by a 15-meter clay layer. Monitoring wells placed on the centerline — exactly where the analytical model said the peak would be — measured near-zero concentrations. The contamination was 40 meters to either side.
What Each Model Sees
Same source zone. Same aquifer parameters. Same 20-year simulation time. The only difference is how much geological reality each model includes.
The Probability Map Changes Everything
The probability-of-exceedance map (Model C) is the most useful product for monitoring well placement. Instead of asking “where is the plume?” — which implies a single answer — it asks “where is PFOS most likely to exceed the 4 ppt MCL?” That's the question a well placement decision actually needs answered.
In the exceedance map, hot zones (red/gold) indicate cells where 70–100% of Monte Carlo realizations exceed the MCL. Cool zones (green) show 20–40% exceedance probability. Wells placed in the high-probability zones will almost certainly detect contamination. Wells on the centerline — where the deterministic model peaks — may land in the dead zone between the two lobes.
The practical implication: A monitoring network designed from the screening model places wells where the model says the peak is. A network designed from the probability map places wells where contamination is most likely to be found — which may be 30–50 meters off-centerline. The difference is between a network that confirms what you assumed and one that finds what's actually there.
Plumes at 20 years post-release. Model B: MODFLOW 6 GWF+GWT, 200×100 grid. Model C: 100 realizations, exceedance probability relative to 4 ppt PFOS MCL.
What a 2D Model Cannot See
The heatmaps above show plan view — where the plume is on the map. But PFAS contamination has dramatic vertical structure. We downloaded 61 real USGS monitoring well measurements from Joint Base Cape Cod, many from multi-level well clusters that measure concentration at 5–10 different depths at the same location.
The data shows the plume is not uniform with depth. Peak concentrations are typically at 10–15 meters below the water table, dropping to non-detect by 25–35 meters. A 2D model that averages over the full aquifer thickness misses this entirely — and can't tell you at what depth to screen your monitoring wells.
The model works — in 3D. With Cape Cod parameters (K=95 m/d, 10 layers, vertical anisotropy 10:1), MODFLOW 6 predicts the plume front at 2,910m. USGS measured 2,700m — an 8% match without calibration. The plume front varies by layer: 2,120m near the water table, 2,910m at 25–30m depth where recharge has pushed the contaminant, and zero below 15m. This vertical structure is invisible to a 2D model.
Plume Front by Layer
| Depth | Plume Front |
|---|---|
| 50–45m (near water table) | 2,120m |
| 45–40m | 2,600m |
| 40–35m | 2,770m |
| 35–30m | 2,870m |
| 30–25m | 2,910m (maximum) |
| 25–20m | 2,900m |
| 20–15m | 2,840m |
| 15–10m | 2,680m |
| 10–5m | 0m (clean) |
| 5–0m | 0m (clean) |
3D model: MODFLOW 6, 10 layers × 50 rows × 150 cols = 75,000 cells. Cape Cod parameters (Kh=95 m/d, Kv=9.5 m/d, n=0.39, Kd=0.4, recharge=0.0015 m/d). 55-year simulation. Real data: USGS Water Quality Portal, 2019–2020 sampling at Joint Base Cape Cod.