Chapter 17 — T-PROGS 3D Geology

17.1 Why T-PROGS Exists — Between Single-K and Full Stratigraphy

Every groundwater model has to decide how to represent the aquifer's conductivity. The choices span a spectrum from very simple to very detailed; T-PROGS occupies a specific, useful middle ground that most modeling benefits from when the data supports it.

17.1.1 The spectrum of K representations

T-PROGS sits between scattered-point interpolation and full stratigraphic modeling — more realistic than smoothed K surfaces but far cheaper than hand-constructing every geological unit. It uses the borehole data you have to inform a statistical model of stratigraphy, which then generates K fields consistent with the observations.

17.1.2 What "realistic" means here

The central claim of T-PROGS: the K fields it produces look like geology. Concretely:

• Discrete lithologic units. You see distinct sand layers, clay lenses, till beds — not smooth color gradients. This matches what geologists see in cross-sections and well logs.
• Sharp contacts between units. A sand layer meets a clay bed along a distinct contact surface, not a gradual transition. K changes by orders of magnitude across that contact.
• Realistic geometry. Layers have appropriate thickness, lateral continuity, and proportions matching what borehole observations suggest. Sand lenses are lens-shaped, not spherical blobs.
• Stratigraphic ordering. Younger deposits sit on older, with transitions following geological reasonable patterns (a sand unit doesn't typically sit directly on bedrock without an intervening till in glacial terrain, for example).
• Borehole honoring. At every borehole location, the simulated lithology matches the observed record exactly — the data is not smoothed or averaged.

These are qualitatively different fields from what continuous interpolation produces. A sandy aquifer with a clay lens will show, in T-PROGS output, an actual clay lens with sharp edges; in kriged output, the same aquifer shows a K minimum that smoothly decreases and increases through the clay area. The hydraulic behavior differs: the T-PROGS clay lens diverts flow; the kriged minimum only partially impedes it.

17.1.3 The four-class typology — AQ, MAQ, PCM, CM

T-PROGS produces a categorical field — every cell belongs to a discrete material class. IGW-NET uses a standard four-class typology grounded in hydrogeologic function rather than in lithologic description alone:

The typology is hydrogeologic, not lithologic. Drillers describe lithology in many ways — "coarse gravelly sand", "silty clay with stones", "weathered shale" — but what matters for groundwater modeling is how the material transmits water. AQ/MAQ/PCM/CM groups the thousands of possible lithologic descriptions into four classes organized by hydraulic function. "Gravelly sand" and "coarse outwash" are both AQ because they both transmit water freely; "silty clay" and "glacial till with high clay fraction" are both CM because they both effectively block flow.

Throughout this chapter, material categories are referred to both by their class labels (AQ, MAQ, PCM, CM) and by the material numbers (1, 2, 3, 4) they appear as in the Borehole Simulation Options dialog. Material 1 = AQ, material 2 = MAQ, material 3 = PCM, material 4 = CM is the standard mapping.

17.2 Transition Probability Geostatistics — The Method

This section gives an intuition for what T-PROGS is actually doing under the hood. Full geostatistical theory is beyond this manual's scope, but understanding the core idea helps you use T-PROGS well and recognize when its assumptions apply.

17.2.1 The core idea — transitions between materials

Rather than modeling K as a continuous random variable, T-PROGS models the probability of transitioning between material categories as you move through space. Given you're currently in "sand" at some location, what's the probability the next point (in some direction, at some distance) is also "sand" versus "clay" versus "till"?

This probability structure — the transition probability — is estimated from the borehole data:

• Proportions — how much of the observed lithology is each material (e.g., "30% sand, 20% silt, 40% till, 10% dense till")
• Mean lens thickness — in each direction (vertical, horizontal), typical thickness of each material (e.g., "sand lenses average 3 m thick vertically, 50 m laterally; till beds average 8 m thick vertically, 500 m laterally")
• Juxtaposition tendencies — which materials tend to occur next to each other (sand often adjacent to silt; clay rarely directly adjacent to dense till without intervening layers)

From these statistics, T-PROGS simulates realizations — specific 3D arrangements of materials that are statistically consistent with the observed structure. Each realization honors the boreholes exactly (observed material at observed depth, at the observed location) and fills in between with materials distributed per the transition probability model.

17.2.2 Why categorical treatment matters

The categorical framing is the key distinguishing feature of T-PROGS. Kriging and similar continuous-interpolation methods assume K is a smooth function of space with some uncertainty. Transition probability methods assume the material type is a discrete choice at each point, with spatial structure governing how those choices cluster into layers and lenses.

This matters because:

• Real lithology is discrete — you don't have "62% sand" at a point; you have sand or you have clay. The categorical model matches the data-generating process.
• K contrasts are preserved — a 5000× K contrast between sand and clay shows up as a 5000× contrast in the simulated field. Continuous interpolation would smooth this contrast and lose most of the hydraulic effect.
• Flow behavior is different — preferential flow along high-K channels, flow focusing at stratigraphic windows, and stagnant zones in low-K pockets all require the sharp contrasts that categorical simulation preserves.

17.2.3 What T-PROGS doesn't do

Important limits to understand:

• T-PROGS simulates structure, not properties within each material. Each material has one Kxx value (and optionally Kxx/Kyy, Kxx/Kzz ratios); within the material, K is constant. For finer-grained heterogeneity within a single lithologic category, you'd overlay random fields or use a more refined material classification.
• T-PROGS is statistical, not deterministic. Different runs produce different realizations — all consistent with the borehole data and the transition probability statistics, but differing in where specific lenses are between wells. For critical applications, multiple realizations can be run and compared (stochastic simulation, Ch. 19).
• T-PROGS needs data. Without boreholes, there's nothing to estimate transition probabilities from. For regions without borehole data, T-PROGS isn't the right tool.
• T-PROGS doesn't model specific geological processes. Glacial outwash, fluvial channels, delta progradation, lacustrine sequences — these have specific depositional signatures that a process-based geological model would capture. T-PROGS captures their statistical end-product but not the processes.

17.3 Setup — The Borehole Simulation Workflow

In IGW-NET, T-PROGS is accessed through the Zone Flow Properties → Conductivity → Borehole Simulation option. This section walks through the five-step workflow.

17.3.1 The workflow

Create a DM Zone covering the entire model domain

Zones → Zone = DM (or equivalent zone-type selection). The DM (Domain Material) Zone is the zone that T-PROGS will populate with heterogeneous K. For most models this zone covers the whole aquifer; you can have sub-zones inside it for specific features (contaminant source areas, excluded regions) that override with different K.

Open Zone Attributes → Flow Properties → Conductivity

In the Flow Properties tab of the Zone Attributes dialog, check Conductivity to enable K configuration for this zone. Under Conductivity, the sub-option radio buttons let you choose: Constant, Scattered Points (IDW/Natural-Neighbor/Kriging), Random Fields, or Borehole Simulation. Select Borehole Simulation.

Open Borehole Simulation Options

Click the '...' options button next to Borehole Simulation. The Borehole Simulation Options dialog opens — this is the main T-PROGS configuration panel.

Import the .tp file

Check Import and upload your .tp file (BarronLakeTP.tp in the Barron Lake case). The .tp file is the T-PROGS output — the 3D categorical material field. For automated-coverage regions, the Data Center delivers this file pre-computed; for other regions, you produce it from an upstream T-PROGS workflow.

Configure the Material table

The Material table has one row per material type with columns for Kxx, Kxx/Kyy, Kxx/Kzz, porosity, Sy, and Ss. For each material, fill in the K value — the central parameter — plus anisotropy ratios and storage parameters. These convert the categorical material field (integer material IDs 1, 2, 3, ...) into the actual K values and properties used by the simulation.

17.3.2 Material K values — the central parameter

The most important input is the Kxx (horizontal conductivity) per material. Using the four-class typology introduced in §17.1.3, typical K ranges per class:

These are typical starting values; calibration (Ch. 18) often refines them based on observed heads and fluxes. The important thing is to get the orders of magnitude right — AQ should be ~10,000× more permeable than CM, not 10× — because the hydraulic behavior depends on the K contrast between productive units and confining layers. The Barron Lake values in the rightmost column fall squarely within the typical ranges; they represent specific tunings for that site's glacial stratigraphy (§17.4).

17.3.3 Anisotropy ratios

Kxx/Kyy and Kxx/Kzz ratios capture directional variation in K within each material:

• Kxx/Kyy (horizontal anisotropy) — typically 1 (isotropic) for most glacial deposits. Can be elevated for alluvial fan deposits with preferred flow direction, but that's unusual in most groundwater contexts.
• Kxx/Kzz (vertical anisotropy) — typically 10 for glacial deposits (horizontal K 10× vertical K, reflecting layered deposition). Can be higher (50, 100) for highly-laminated sediments; can be 1 for massive rock or isotropic materials.

The IGW-NET defaults (Kxx/Kyy = 1, Kxx/Kzz = 10) work well for most glacial-terrain applications. Site-specific calibration may adjust them.

17.4 The Barron Lake 4-Material Example

Barron Lake's T-PROGS configuration is an instructive example of how glacial stratigraphy maps onto the 4-material T-PROGS framework. This section unpacks the specific choices.

17.4.1 The 4 materials

The Barron Lake model uses the standard 4-class T-PROGS typology (§17.1.3), with K values tuned to the observed glacial stratigraphy in southwest Michigan:

The K values span 5 orders of magnitude (from 53 to 0.001 ft/day = 53,000× contrast) — capturing the full range of permeabilities in the glacial deposits. This is the hydraulically-realistic range; a single bulk K would average all four into something like 1 ft/day, missing both the productive aquifer behavior and the confining-layer effects. Note that the class values for Barron Lake (53, 5, 0.01, 0.001 ft/day) fall squarely within the typical class ranges given in §17.1.3 — they are project-specific tunings within the standard typology, not free-parameter choices. The four K values together characterize the entire 3D heterogeneous K field that T-PROGS produces from the borehole data.

17.4.2 How the stratigraphy plays out in 3D

17.4.3 Why this structure matters for the Barron Lake model

The stratigraphic heterogeneity is not cosmetic — it affects the model's answers:

• Lake-aquifer coupling strength depends on the K of the material directly beneath the lake. If the lake sits on outwash (Material 1), exchange is strong; if it sits on till (Material 3 or 4), exchange is weak. T-PROGS tells you which.
• Well yield patterns depend on which stratigraphic units the well intersects. Wells screened in outwash yield well; wells in silty sand yield modestly; wells in till yield poorly. The 11 private wells in the Barron Lake study have different productivity based on their screened units.
• Contaminant transport preferential paths follow high-K units. A spill at the lake edge migrates rapidly through outwash lenses and slowly through surrounding till — pattern invisible in single-K models.
• Regional flow organization — the mainstem regional flow follows high-K units between lower-K confining layers, producing the layered flow patterns observable in detailed head measurements.

17.5 Regional Zonation — When One Set of 4 Parameters Isn't Enough

The Barron Lake single-zone configuration is the baseline T-PROGS setup: four K values for the whole domain, characterizing the complete 3D heterogeneous K field. For larger or geologically more complex models, a single set of four values may not capture spatial variation in the underlying geology. This section covers when and how to escalate from one-zone to multi-zone T-PROGS.

17.5.1 Why regional zonation is needed

The four-class typology (§17.1.3) groups materials by hydraulic function — AQ is coarse-grained productive units, CM is fine-grained aquitards. But what counts as an AQ in one region may be hydraulically different from AQ in another region. Two examples:

• Glacial outwash AQ vs. fluvial fan AQ. A glacial outwash aquifer (clean sorted sand, modest K around 50 ft/day) and a fluvial fan aquifer (cobbles and coarse gravel with high K around 500 ft/day) are both "AQ" — both are the most productive unit in their region — but their K differs by an order of magnitude. Using a single AQ parameter for both misrepresents one or the other.
• Glaciated vs. unglaciated terrain. In glaciated terrain, CM is typically dense till (K around 0.001 ft/day). In unglaciated sedimentary terrain, CM might be shale or claystone (K closer to 0.0001 ft/day, ten times lower). A model crossing the glacial boundary would need different CM values on each side.

When the domain spans meaningfully different geologic settings — physiographic provinces, glaciated boundaries, depositional environments, bedrock geology changes — one set of four parameters can't capture the full variation. The solution: divide the domain into spatial regimes or zones, each with its own set of four K values.

17.5.2 How regional zonation works

Each spatial regime is defined as a zone (using IGW-NET's standard zone infrastructure, same as for submodels in Ch. 13). Within each zone, T-PROGS produces its own 3D categorical material distribution, and the four class K values (AQ, MAQ, PCM, CM) are specified per zone. The total parameter count scales linearly with the number of zones: N zones × 4 class K values = 4N parameters.

Each additional zone adds four parameters and requires the calibration data to constrain them. The zone count shouldn't exceed what the observational data can constrain — more zones without supporting data produces an underdetermined calibration where multiple K-field configurations fit the observations equally well (non-uniqueness).

17.5.3 How to decide where zone boundaries go

Zone boundaries should follow real geologic contrasts, not arbitrary convenience. Useful guides:

• Physiographic maps. National or regional physiographic province maps (USGS for the US, provincial surveys for Canada) show where geologic settings change. These boundaries are usually appropriate T-PROGS zone boundaries.
• Surficial geology maps. Where the surficial geology changes — glacial to interglacial, moraine to outwash, alluvium to bedrock — the near-surface aquifer materials likely change, and zonation is warranted.
• Known aquifer-system boundaries. Published regional groundwater studies often delineate distinct aquifer systems with different properties. Use those boundaries.
• Observed head patterns. If your initial single-zone model calibration shows systematic regional biases in residuals (all wells in the north biased low, all wells in the south biased high), the residual pattern itself suggests where zonation might help.

17.5.4 Escalation path — when to add zones

The practical workflow: start simple, add zones only when evidence demands.

17.5.5 Connection to calibration (Ch. 18)

Regional zonation directly affects calibration strategy. With 4 parameters per zone, the calibration is well-posed as long as you have enough observations per zone. Typical rule of thumb: at least 5–10 meaningful observations per 4 calibrated parameters, distributed so each parameter is independently constrained. Chapter 18 covers calibration mechanics, UCODE parameter estimation, and residual interpretation in detail — zonation is one of the first design decisions that shapes the calibration problem.

17.6 Regional Data Coverage — Water-Well Lithology Integration

T-PROGS needs borehole data. What makes it practical for day-to-day modeling is automated access to water-well lithology databases for a meaningful portion of North America.

17.6.1 The covered regions

DataNET integrates water-well lithology databases for:

• Approximately 10 U.S. states — state-specific water-well record databases (Michigan's Wellogic is the prototype; other covered states have similar state-run databases)
• All Canadian provinces — each province maintains a water-well database; all are integrated

Within the covered regions, the end-to-end T-PROGS workflow is automated through the Data Center:

For regions outside the covered coverage, the methodology is identical but the data-acquisition path requires user involvement: users supply borehole data from whatever local source is available (USGS NWIS for US federal data, national geological survey databases for other countries, site-specific drilling records for local projects) in the T-PROGS input format.

17.6.2 Wellogic and its equivalents

Michigan's Wellogic is the most widely-referenced example of a state water-well database integrated with MAGNET4WATER. It contains lithologic logs from tens of thousands of wells drilled across Michigan over decades of well construction, documented by drillers as a regulatory requirement. The depth of coverage — one well every few hundred meters in developed areas — provides excellent T-PROGS input for most Michigan groundwater modeling.

Other covered states have equivalent state-run databases with comparable coverage (the exact name and administrative context varies by state). The DataNET integration abstracts these differences: regardless of which state you're in, the workflow is the same — request borehole data, receive the .tp file.

Canadian provincial databases operate similarly, maintained by each province's water-resources or environment ministry, with coverage determined by provincial well-drilling records.

17.7 When to Use T-PROGS

T-PROGS adds configuration complexity and interpretation overhead. This final section helps you decide whether that cost is worth paying for your specific modeling question.

17.7.1 Strong signals for T-PROGS

Use T-PROGS when any of these apply:

• Stratigraphic control is known to matter. You have prior knowledge — from consulting reports, regional hydrogeologic assessments, or past calibration experience — that the aquifer is meaningfully layered and that heterogeneity drives behavior.
• Transport modeling (Ch. 12). Contaminant transport is especially sensitive to preferential flow paths. T-PROGS reveals the high-K lenses that route plumes; smoothed K fields miss them.
• Wellhead protection in complex geology. Delineating capture zones in layered aquifers with confining units requires realistic vertical structure.
• Coupled-lake modeling (Ch. 15) in glacial terrain. The K directly beneath the lake controls the coupling strength; T-PROGS gives you a defensible value.
• Client or regulatory expectations for realistic-looking 3D geology. Cross-sections and 3D visualizations are much more convincing with T-PROGS heterogeneity than with smoothed fields.
• Automated data coverage. Your domain is in the 10 covered U.S. states or any Canadian province — the setup cost is minimal, so the threshold for using T-PROGS drops.

17.7.2 Signals to skip T-PROGS

Skip T-PROGS when:

• Bulk K is sufficient for the question. Regional flow pattern; long-term water balance; first-pass scoping studies.
• No borehole data available. Outside covered regions without user-supplied data, T-PROGS has nothing to work with. Use scattered-point interpolation of whatever wells you do have, or a random-fields approach.
• Steady-state models where head pattern is the primary output. Calibrating a bulk K often matches observed heads as well as calibrated T-PROGS K values would, at a fraction of the setup cost.
• Very large regional models where computational cost of 3D heterogeneous K is prohibitive.
• Homogeneous aquifers. Some aquifers really are nearly uniform — thick sandstone formations, massive carbonate platforms. T-PROGS on these is overkill and the result is indistinguishable from bulk K.

17.7.3 The escalation path