How do I build a regional-to-local nested groundwater model in IGW-NET?

Create the regional model with DrawDomain, then add a Zone (ZoneRect or ZonePoly) within it as the local area. Initially set the zone type to PolygonOnly so only the regional model simulates. After running and calibrating the regional model, switch the zone to Active Submodel Domain and check Boundary Conditions from Parent Model in Simulation Settings. The local model will then use the regional solution as its boundary conditions.

How do I calibrate a regional groundwater model to observed water levels?

Use the Calibration tool with IGWServer as the observed data source. Filter to the appropriate aquifer type (e.g., glacial drift only, no bedrock wells) and date range. Sample on the order of 10,000 Static Water Levels as calibration targets. Adjust hydraulic conductivity and recharge multipliers in Domain Attributes to shift the simulated vs observed cloud toward the 45-degree line of perfect match.

How do I design a purge well system for groundwater remediation?

After simulating the contamination plume, use DrawWell to add an extraction well downstream of the source. Assign a negative pumping rate (e.g., -2000 GPM). Resubmit the simulation with Boundary Conditions from Parent Model enabled. The contour lines will bend toward the well (indicating capture), and the simulated plume extent should contract.

Why does the contaminant plume only occupy the upper portion of the aquifer when I add vertical layers?

Because the contamination enters from the surface via recharge, not as a vertically-mixed source. When you subdivide the aquifer into multiple computational sublayers and model the source as Recharge-Quantity & Quality rather than a continuous concentration throughout the column, the plume naturally occupies only the upper portion near the water table — which matches reality for most dissolved-phase contaminants migrating from shallow spill sites.

Mancelona TCE Plume — IGW-NET Case Study

Case summary

This end-to-end case study demonstrates how to use IGW-NET to develop, calibrate, and apply a coupled flow-and-transport simulator for the Wickes Manufacturing TCE plume in Mancelona, Michigan — one of the largest groundwater contamination plumes in the United States and the largest TCE plume in Michigan. TCE is a known carcinogen.

The site's proper treatment demands multi-scale modeling: a large regional domain captures the prevailing flow field and provides boundary conditions to a finer-grid local submodel that resolves plume detail around the source. Once calibrated, the model is used to predict plume migration and design a purge well for active remediation. Finally, vertical computational layers are added to reveal that the plume is not vertically mixed — it occupies primarily the upper third of the aquifer, consistent with a surface-source origin.

The site and the problem

From the 1940s through the 1950s, Wickes Manufacturing operated a plant in Mancelona, Michigan (a small town in the northwestern Lower Peninsula). The plant used chlorinated degreasing solvents — principally trichloroethylene (TCE) — and disposed of waste solvents in shallow pits on site. Over subsequent decades the TCE has migrated northwest through the surficial glacial aquifer, producing a plume that is now one of the largest of its kind in the United States. The plume trends toward a well field that provides drinking water to private residences and nearby municipalities, creating a source-protection problem of direct public-health consequence.

Simulating the plume's transport and planning remediation operations requires detailed flow information at the site, but site-specific flow conditions are influenced by regional flow patterns. The proper approach is nested modeling: a coarser regional model establishes the far-field head distribution and provides boundary conditions to a finer local model that resolves plume-scale detail.

1. Create model domains and overlay the site map

In the IGW-NET modeling environment, after zooming to Mancelona at a regional scale, the DrawDomain → DomainRect tool is used to draw a rectangular regional domain covering the site and surrounding region. Two mouse clicks (one per diagonal corner) are sufficient.

MAGNET modeling environment default screen showing world map with Quick Helper menu for first-time users — Figure 1Default screen setting when first accessing the IGW-NET Modeling Environment. First-time users need to create a free MAGNET4WATER account via the Quick Helper menu.

Navigation tools with Search bar and Zoom level controls, zoomed to Mancelona Michigan — Figure 2Navigation tools zoomed to the site area. After adjusting the zoom level and entering the site name in the Search bar, pressing Enter navigates to the site.

Regional Model domain drawn as red rectangle around Mancelona and surrounding region using DomainRect tool — Figure 3Drawing the Regional Model domain using DomainRect. Single left-mouse click at first vertex, then another at diagonal vertex.

A local submodel zone is then added using ZoneRect, covering Mancelona itself and the area to the northwest. Critically, the zone is initially configured as PolygonOnly in its ZoneAttr Flow Property tab — this flag tells the solver to ignore the zone during the first simulation runs so the regional model can be calibrated independently. After calibration, the zone will be switched to Active.

Local Model area highlighted as green rectangle inside the larger Regional Model domain, centered on Mancelona — Figure 4Local model area added as a Zone feature within the Regional Model domain.

Zone Attributes menu Flow Property tab with Submodel Domain checkbox and Zone Type set to PolygonOnly — Figure 5Zone Attributes menu — assigning the added zone as the submodel domain with PolygonOnly selected. This allows the Regional Model to run first; after calibration, this will be changed to Active.

For visual context, a georeferenced plume map is overlaid using Utilities → Overlay myImage. The image shows the previously-delineated TCE plume extent from traditional hydrogeological field methods — a benchmark against which the model results can be compared later.

Overlay an Image on Maps menu with site map showing well fields and TCE plume extent overlaid on terrain map — Figure 6Site image overlay — the georeferenced plume map showing the previously-delineated TCE plume, Cedar River well field, and Shanty Creek well field.

2. Regional model parameters

The DomainAttr menu is used to assign aquifer attributes. Every scalar is linked to a MAGNET4WATER Data Center raster layer — spatial variability comes from the server, not from uniform manual input:

Top elevation: 10 m Digital Elevation Model
Bottom elevation: Michigan bedrock surface raster (built from well records)
Hydraulic conductivity: Michigan Glacial-Screen to 1st Confining Unit raster
Recharge: Global w/MI Patch long-term mean recharge

Concept reminder: When a property uses the Data Center, the scalar you see (e.g., Cond=22.86 m/day) is a fallback, not the actual value used. The actual values come from the spatially-variable raster — see Fallback vs spatial values for why this matters when someone reviews your model file.

Aquifer Attributes window showing Top Elevation, Bottom Elevation, Hydraulic Conductivity, and Rain Recharge all configured to pull from Data Center rasters — Figure 7Aquifer Attributes — all four domain properties configured from Data Center rasters specific to Michigan.

Conceptual cross-section diagram showing aquifer top at land surface, aquifer bottom at bedrock, unsaturated zone, groundwater system, spatially-variable recharge, and discharge to surface water body — Figure 8Conceptual cross-section of the model inputs — aquifer top (land surface), bottom (bedrock), spatially-variable recharge, and discharge to surface features.

3. Regional simulation

With SIMULATE in steady-state mode, the regional flow field is solved. Head contours and velocity vectors appear over the domain. In highly-gradient areas — near regional discharge features — vector density can make the display crowded; adjusting DomainAttr → Display Settings → Draw Vector Every: 2 grids spaces the vectors for readability.

Regional model output showing head contours and dense groundwater velocity vectors throughout the domain — Figure 9Regional model output. Contour lines are computed hydraulic heads; arrows are velocity vectors (length proportional to magnitude).

Display Settings tab with vector spacing parameters highlighted — Figure 10Display Settings — adjusting vector spacing from 1 to 2 pixels in both X and Y directions for readability.

An X-section (cross-section) is drawn from the regional recharge area to the regional discharge area, displaying vertical hydraulic conductivity variability. In this 2D setup, K_xx = K_zz, so the coloring reflects the Data Center raster variability along the section.

Plan view model with head contours plus cross-section showing colored vertical conductivity variation from recharge to discharge area — Figure 11Plan view display + cross-section view. Colors in the cross-section represent variations in vertical hydraulic conductivity.

4. Evaluate model performance (calibration)

The Calibration tool compares simulated heads against Static Water Levels (SWLs) — measurements of water level taken at the time wells were installed, before any pumping. The data source is set to IGWServer, with filters applied:

Aquifer type: glacial drift only (no bedrock wells — they respond to different flow regimes)
Construction date: after 2000 (to avoid old, less reliable records)

The Well Data Processing Tool is used to randomly sample 10,000 SWLs as calibration targets. The chart customization adds a 1σ confidence band, a moving-window (band-mean) average, and reduces marker size for clarity.

Calibration Chart setup with IGWServer data source, Server Data Filters dialog, Well Data Processing Tool for random sampling of 10000 wells — Figure 12Calibration setup — filtering wells by aquifer type and construction date, then sampling 10,000 SWLs as targets.

Calibration chart with scatter of simulated versus observed heads, default settings on left versus customized settings with confidence band on right — Figure 13Calibration chart — left: default settings; right: customized with confidence intervals and band-mean. Note the model systematically underestimates heads between 250–375 m.

Finding: The default parameters produce a calibration cloud with systematic underestimation of heads in the 250–375 m range. This is a physics signal: the model's K is too high or recharge is too low for that portion of the domain. Both levers can be adjusted via domain-wide multipliers.

5. Adjust K and recharge to improve performance

In DomainAttr, domain-wide multipliers are applied: K × 0.3 and Recharge × 1.1. A K multiplier of 0.3 means every cell's hydraulic conductivity from the Data Center raster is reduced by 70%; the recharge multiplier of 1.1 means every cell's recharge is increased by 10%. The resulting calibration cloud and the band-mean are both centered on the 45-degree line of perfect match.

Calibrated regional model with K multiplier 0.3 and Recharge multiplier 1.1, showing data cloud and band-mean now centered on 45-degree line of perfect match — Figure 14Calibrated regional model results. The cloud of data and band-mean are centered on the 45° line of perfect match.

6. Activate the local submodel

With the regional model calibrated, the local zone is switched from PolygonOnly to Active. Using Utilities → Geometry unlocked, the zone nodes become clickable; clicking any node opens ZoneAttr, where the Flow Property tab is updated: Submodel Domain checked, Zone Type = Active.

Zone Attributes menu with Submodel Domain checked and Active zone type selected — Figure 15Making the submodel active in the Zone Attributes menu.

In the Simulation Settings tab of DomainAttr, the Boundary Condition from Parent Model checkbox is turned on. This tells the solver to use the last regional simulation's head solution as the prescribed-head boundary for the local model.

Concept reminder: Once boundary conditions come from the parent, any polyline or zone boundary conditions you drew locally on the submodel boundary will be overridden. This is by design — see Submodel boundary precedence.

Simulation Settings tab with Boundary Condition from Parent Model checkbox enabled — Figure 16Assigning the boundary condition from the parent (regional) model.

The local model is then simulated, producing a refined head field and a new cross-section showing detailed local flow patterns.

Local submodel simulated results with detailed head contours, velocity vectors, and cross-section showing vertical conductivity variability — Figure 17Local model simulated results — detailed head contours, velocity vectors, and cross-section of vertical conductivity variability.

7. Add the Cedar River as a stream feature

The Cedar River is a salient feature of the site, situated north-northeast of the plume and eventually draining to Lake Bellaire. It is added using DrawLine; click-and-point adds vertices along the river's path. SaveShape finalizes the polyline and opens the Line Attributes menu, where the river is configured as a two-way head-dependent boundary:

Stream stage: equal to aquifer top (DEM)
Stream bottom: stage − 1 m
Leakance: 5 m/day (applied uniformly along all stream reaches)

Cedar River stream polyline drawn on local map, with Line Attributes menu configured as two-way head dependent with stream stage from DEM and leakance 5 m/day — Figure 18Cedar River added as a line feature — two-way head-dependent with DEM-derived stage and 5 m/day leakance.

8. Contamination source and monitoring well

Zooming in to the former Wickes Manufacturing property, a ZonePoly is drawn at the site. In ZoneAttr → Source and Sinks Prescribed, the feature is assigned as a continuous source with concentration 1000 ppm.

Zone polygon at Wickes Manufacturing spill site with Source and Sinks Prescribed tab showing Continuous source concentration 1000 ppm — Figure 19Contamination source at the Wickes spill site — zone polygon with continuous 1000 ppm concentration.

A monitoring well is added downstream with DrawWell, with the Monitoring Well option enabled. This causes IGW-NET to track concentration at that location over time and produce a breakthrough curve.

Monitoring well added downstream of spill site with Well Input Options menu showing Monitoring Well option checked — Figure 20Monitoring well added downstream using the Well tool.

Concept reminder: Transport is auto-detected from concentration sources. The moment the source zone gets a non-zero concentration, transport activates for the simulation — there is no on/off toggle. See Auto-detected transport.

9. Simulate transport and compare with field delineation

After ensuring Boundary Conditions from Parent Model is still checked (it gets unchecked after each simulation), the local model is submitted. The simulation is steady-state in the flow field but transient in concentration — the plume spreads over time using the fixed flow field. Analysis Tools → Analysis → Display Charts opens the monitoring well breakthrough curve, the cross-section, and the mass balance chart.

Simulated plume extent in plan and cross-section views, monitoring well concentration breakthrough curve showing rising TCE, aquifer mass balance chart, and calibration chart all displayed simultaneously — Figure 21Simulated model results — plume extent (plan and cross-section), concentration breakthrough at monitoring well, aquifer water balance, and local calibration chart. The simulated plume direction and extent are broadly consistent with the field-delineated plume from traditional hydrogeological methods.

Model validation: The simulated plume trends northwest and the simulated concentrations at the monitoring well produce a plausible breakthrough curve. The plume extent is broadly consistent with the field-delineated plume on the overlay image — though note the simulated plume is not pulled quite as far west as reality suggests. This residual discrepancy reflects the influence of the Cedar River well field (Shanty Creek Resort), whose pumping wells aren't yet represented. Adding them with DrawWell closes that gap.

10. Design a purge well for remediation

A remediation extraction well is added downstream of the source using DrawWell, with an aggressive pumping rate of −2000 GPM (negative denotes extraction). After resubmitting the simulation with parent-derived boundary conditions, the contour lines bend toward the purge well — the classic signature of capture — and the simulated plume extent contracts dramatically.

Remediation insight: The purge well's capture zone intercepts the forward plume migration. Note the bent contour lines near the well and the much smaller plume extent compared to the un-remediated case. In real remediation design, this kind of simulation informs pumping rate selection, well placement optimization, and treatment system sizing.

11. Add vertical detail — sublayers and a surface source

The 2D simulation so far has assumed the plume is perfectly mixed in the vertical direction. In reality, TCE enters the groundwater system from above via infiltration of recharge through the spill pits. Resolving this requires both vertical computational layers and a surface-source formulation.

Convert the source

In ZoneAttr → Source and Sinks Prescribed for the spill zone:

Uncheck Source Concentrations
Check Recharge−Quantity & Quality
Enter a constant infiltration rate of 10 inch/year
Enter source concentration of 1000 ppm

Add computational layers

In DomainAttr → Simulation Settings:

Check Water Table as Top
Check Number of SubLayers and set to 8

Concept reminder: Sublayers multiply computational cost. With 1 conceptual layer and 8 sublayers, the solver processes 8 layers — see Sublayers vs conceptual layers. The "water table as top" option uses the 2D water table solution to subdivide the saturated thickness efficiently.

When the model is re-simulated, the plume now occupies primarily the upper ~1/3 of the aquifer rather than being vertically mixed. This matches reality for surface-sourced dissolved-phase contaminants migrating laterally in a shallow flow system.

Final insight: Moving from a 2D mixed source to a 3D surface source is not just refinement for its own sake — it reveals a qualitatively different plume morphology. The vertical layering matters for monitoring well design (where should the screen be placed?), for remediation design (which depth interval should be targeted?), and for risk assessment (where is the mass actually located?).

Source Provenance

Original document: IGW-NET Tutorial: Massive Plume Migration in Northern Michigan Community (PDF)
Site location: Mancelona, Michigan, USA — approximately 44.90°N, 85.06°W
Contaminant: Trichloroethylene (TCE) — known carcinogen
Source facility: Wickes Manufacturing, degreasing solvent disposal (1940s–1950s)
Platform: IGW-NET / MAGNET4WATER
Last reviewed: 2026-04-20