Methodology: what is measured, what is modelled, and where the line is
Every number on this site belongs to one of two categories, and we never let them blur. Category one is federal soil survey data, which we reproduce and translate. Category two is our own modelling, which we label as ours everywhere it appears. No county has been refreshed from SSURGO yet. Every county page is on the regional model and says so at the top.
1. Soil data (measured)
Source: the USDA NRCS Soil Survey Geographic Database (SSURGO), queried through the free Soil Data Access service. The underlying survey is decades of field mapping by soil scientists; the browsable front end is Web Soil Survey. It is public domain US government work.
| Drainage class | muaggatt.drclassdcd — dominant condition per soil map unit |
|---|---|
| Wettest class present | muaggatt.drclasswettest |
| Shrink-swell | chorizon.lep — linear extensibility, percent; maximum across horizons above 100 cm, then weighted by component percentage and map unit area |
| Water table | muaggatt.wtdepannmin — shallowest annual depth in cm; SSURGO reports this field only to 200 cm, so 200 means “200 or deeper” |
| Flooding frequency | muaggatt.flodfreqdcd |
| County attribution | legend.areasymbol and legend.areaname, or a spatial intersection — each county page states which |
| Hydrologic group | muaggatt.hydgrpdcd |
| Aggregation | Area-weighted by mapunit.muacres across every map unit in the county |
We aggregate to county level because that is the unit a homeowner can act on without a survey, and because parcel-level soil data invites exactly the false precision we are trying to avoid. If you want your parcel, Web Soil Survey will draw an area of interest for free and we link to it from every county page.
2. The three risk scores (our model, from measured inputs)
The scores are ours. The inputs are not. Each is a transparent function of one SSURGO field, deliberately simple enough to argue with.
Moisture / seepage risk
Area-weighted average of a fixed risk weight per drainage class. The weights follow the USDA definitions of how long each class stays saturated:
| Class | Weight | USDA definition, in plain terms |
|---|---|---|
| excessively drained | 5 | water passes through very fast and the soil holds almost no moisture |
| somewhat excessively drained | 12 | water drains quickly and the profile stays dry most of the year |
| well drained | 22 | water leaves readily; the profile is saturated only briefly after rain |
| moderately well drained | 42 | water leaves slowly enough that the lower profile is wet for part of the year |
| somewhat poorly drained | 66 | the profile stays wet long enough to restrict ordinary use without artificial drainage |
| poorly drained | 86 | water leaves so slowly that the profile is saturated periodically through the season |
| very poorly drained | 96 | water stands at or above the surface for much of the year |
Foundation crack risk
Linear extensibility × 8.2, capped at 100. LEP is the percentage of its dry length by which a soil sample expands when it takes up water — the standard laboratory proxy for shrink-swell behaviour. Below 3% is negligible; above 9% is the range where footing movement occurs without any water reaching the basement at all.
Water table risk
A step function of the shallowest annual water table depth: under 30 cm scores 95, under 60 cm scores 82, under 100 cm scores 66, under 150 cm scores 46, under 190 cm scores 30, and 190 cm or deeper scores 16. A typical US basement floor sits just under 200 cm below grade, and 200 cm is also the deepest value SSURGO reports for this field — which is where the steps come from.
Combined index
0.45 × moisture + 0.32 × cracks + 0.23 × water table. The weighting reflects claim and repair frequency, not severity — a structural pier job costs far more than a sump pump, but it is far rarer. We publish the formula so you can disagree with it.
3. The regional model (used only where SSURGO is not loaded)
Where a county has not been fetched from Soil Data Access, its page runs on a physiographic region model and carries an orange warning saying exactly that. The model is not random. Each county is assigned to one of 49 real physiographic units by state and centroid, and each unit has a documented parent material with the drainage, LEP and water-table ranges that follow from it. Within-region variation is a deterministic function of the county FIPS code, so builds are reproducible.
It is a geologically coherent estimate. It is not a soil survey, and no amount of internal consistency makes it one. It exists so the site is complete and honest rather than half-built and silent.
| Region | Parent material | LEP range | Water table range |
|---|---|---|---|
| Alaska: permafrost and frost-susceptible ground | glacial, alluvial and organic soils, discontinuous permafrost, deep frost penetration | 1.5–5.0% | 30–150 cm |
| Appalachian Plateau | colluvium and residuum from sandstone, siltstone and shale on steep slopes | 3.0–6.5% | 90–200 cm |
| Atlantic coastal plain | unconsolidated sands and clays at low elevation | 1.5–4.5% | 45–150 cm |
| Basin and Range | alluvial fans and lacustrine sediments of former Lake Bonneville | 4.0–9.5% | 150–200 cm |
| Basin and Range desert | alluvial fans and playa sediments, caliche, gypsum, collapsible soils | 3.0–8.0% | 180–200 cm |
| Blackland Prairie and Black Belt | Vertisols weathered from Cretaceous marl and chalk — smectite clay that cracks open in summer | 11.0–18.0% | 60–180 cm |
| Blue Ridge Mountains | deep colluvium and saprolite on steep slopes, boulders in the soil mass | 2.0–5.0% | 120–200 cm |
| Bluegrass karst plateau | phosphatic limestone residuum with well-developed karst | 4.0–8.0% | 120–200 cm |
| California Central Valley | alluvial fan and basin deposits, expansive clays in the basins, hardpan at depth | 6.0–13.0% | 120–200 cm |
| Central Wisconsin Sand Plain | sandy bed of glacial Lake Wisconsin, flat, with a high water table | 1.0–3.0% | 45–140 cm |
| Colorado Plateau | Mancos and Morrison shale residuum, locally gypsiferous | 6.0–12.0% | 150–200 cm |
| Columbia Plateau and Palouse | deep wind-blown loess over basalt | 2.0–5.5% | 150–200 cm |
| Dakota prairie till plain | calcareous till with closed depressions, semi-arid climate | 5.0–9.0% | 60–180 cm |
| Des Moines Lobe and prairie potholes | young calcareous till pitted with thousands of closed depressions | 4.0–7.5% | 20–75 cm |
| Dissected till plain with a claypan | old till with an abrupt clay horizon 30-60 cm below the surface | 4.5–8.5% | 30–90 cm |
| Driftless Area | loess over sedimentary bedrock, unglaciated and steeply dissected | 2.0–4.5% | 120–200 cm |
| Edwards Plateau and Balcones | thin stony residuum over Cretaceous limestone, karst, caves and springs | 2.5–6.0% | 150–200 cm |
| Flint Hills and Osage Cuestas | residuum from Permian limestone and chert on the hills, with expansive clays from Permian and Pennsylvanian shales in the valleys and to the east | 8.0–14.0% | 90–200 cm |
| Florida karst and sand | quartz sand over limestone, active karst, water table often within a metre | 0.8–3.0% | 20–90 cm |
| Front Range: Pierre Shale and Denver Formation | weathered bentonitic shale rich in smectite, with dipping beds | 9.0–16.0% | 180–200 cm |
| Glaciated New England upland | stony basal till over crystalline bedrock, often with a hardpan | 1.5–4.0% | 60–180 cm |
| Glaciofluvial sand plains | outwash sand and gravel of northern Wisconsin, Michigan and Minnesota | 0.8–2.5% | 120–200 cm |
| Glaciolacustrine lake plain | clay laid down on the bed of Pleistocene lakes (the Great Black Swamp, and the Erie and Ontario lake plains) | 5.5–9.5% | 15–60 cm |
| Gulf Coastal Plain | unconsolidated sands and clays, pine uplands and broad flats | 2.5–7.0% | 45–150 cm |
| Hawaiian volcanic soils | basaltic ash and lava residuum, Oxisols and Andisols, locally expansive Vertisols in dry leeward areas | 3.0–10.0% | 90–200 cm |
| High Plains | Ogallala alluvial and aeolian sediments, semi-arid | 3.0–7.0% | 180–200 cm |
| Interior Low Plateau karst | limestone residuum, unglaciated, caves and sinkholes | 3.0–6.5% | 120–200 cm |
| Loess hills | deep wind-blown loess, locally tens of metres thick | 2.5–5.0% | 150–200 cm |
| Mississippi alluvial plain | river alluvium, alternating sands and clays, water table close to surface | 6.0–11.0% | 15–60 cm |
| Missouri Plateau (unglaciated Great Plains) | weathered clay shales and sandstones, Pierre Shale and equivalents | 7.0–13.0% | 150–200 cm |
| Nebraska Sand Hills | stabilised aeolian dune sand, the largest dune field in the Americas | 0.4–1.5% | 90–200 cm |
| Ouachita Mountains | steeply folded sandstone and shale, thin stony residuum | 2.5–6.0% | 150–200 cm |
| Ozark highland | cherty residuum from limestone and dolomite, karst | 2.0–5.5% | 150–200 cm |
| Pacific Coast Ranges | Franciscan mélange and marine sediments, highly expansive clays, unstable slopes | 7.0–14.0% | 120–200 cm |
| Pennyroyal and Western Coal Field | silty residuum and loess over shale, commonly with a fragipan | 4.5–9.0% | 60–180 cm |
| Permian red beds | red clay shales and sandstones of central Oklahoma and north Texas | 7.0–13.0% | 120–200 cm |
| Piedmont | saprolite weathered from crystalline rock, with Triassic basin clays in places | 3.5–8.0% | 120–200 cm |
| Puget Sound and Willamette Lowland | Vashon glacial till and lacustrine silts, with a dense hardpan below | 2.0–5.0% | 45–140 cm |
| Red River Valley (bed of glacial Lake Agassiz) | fine lacustrine clay with a high smectite content | 8.0–14.0% | 20–70 cm |
| Ridge and Valley | limestone residuum in the valleys, sandstone on the ridges, karst | 3.5–7.0% | 120–200 cm |
| Rocky Mountains | stony colluvium and glacial deposits, bedrock close to the surface | 1.5–5.0% | 120–200 cm |
| Sierra Nevada and Cascades | granitic and volcanic residuum, shallow bedrock, coarse stony soils | 1.5–4.5% | 150–200 cm |
| Smoky Hills and central plains | silty loess and residuum from Cretaceous sediments | 4.5–8.5% | 120–200 cm |
| Snake River Plain | aeolian loess over basalt, shallow bedrock | 2.0–5.5% | 150–200 cm |
| South Texas Plains | clayey and loamy sediments of the Rio Grande plain, semi-arid, caliche at depth | 7.0–13.0% | 120–200 cm |
| Southern Atlantic Coastal Plain | marine sands and clays, very low relief, high water table | 2.0–6.0% | 45–140 cm |
| Texas Gulf Coast prairie | Beaumont and Lissie clays — deltaic smectite clay, flat, water table close to the surface | 9.0–16.0% | 30–90 cm |
| Trans-Pecos desert | gravelly desert soils, gypsum and caliche horizons, alluvial fans | 3.0–8.0% | 180–200 cm |
| Wisconsinan till plain | dense basal till (stony clay) under a thin loess cap | 3.5–7.0% | 45–110 cm |
4. Prices (our model, and the weakest thing here)
Price book 2026.08, updated 2026-08-11. Built from published editorial ranges (Angi, HomeAdvisor, This Old House, Bob Vila, Forbes Home) and a per-state labour factor. There is no free machine-readable dataset of US basement waterproofing or foundation repair transactions. If you find one, we will replace this section with it the same week.
The one thing that could fix this is the paid-price database people submit to. Real amounts, by county and solution, no personal data.
5. Radon (expectation, not designation)
EPA assigns every US county to Radon Zone 1, 2 or 3. That dataset is not distributable in a form we can bundle, so the zone shown on our pages is an expectation derived from parent material — uranium-bearing glacial till and black shales run high, coastal sands run low. We label it as expected and link to the official lookup: EPA local radon zones and state contacts. Only a test in your house means anything anyway.
6. Two kinds of county page
Basements are a regional building practice. In the Midwest most houses have one; on the Gulf Coast almost none do. Publishing a basement moisture score for a county where nothing has a basement would be a number with no use, so those counties get a different page built from the same three USDA fields: it leads with soil movement under the slab and footings, and it says at the top that basements are rare there.
The switch is by state, from the prevalence of basements in the housing stock, and it is stated on every page. It changes the framing and the order of the numbers. It does not change a single value.
| Basement-first pages | Midwest, Northeast, Mid-Atlantic, Upper South, Colorado, Utah, Idaho, Montana, Wyoming, Washington, Oregon |
|---|---|
| Foundation-first pages | Texas, Oklahoma, Arkansas, Louisiana, Mississippi, Alabama, Georgia, Florida, the Carolinas, California, Nevada, Arizona, New Mexico, Alaska, Hawaii |
7. The language rule
We write “the soil at this address is classified as poorly drained, which statistically raises the risk of seepage”. We do not write “you have a wet basement”. The first is a fact about a public dataset. The second is a diagnosis we are not qualified to make and have no information to support. Every page on this site is written to that rule, and if you find a sentence that breaks it, that is a bug worth reporting to hello@basementsoil.com.
8. Known limitations
- County averages hide enormous within-county variation. In the Des Moines Lobe and on the Atlantic coastal plain, two parcels a few hundred metres apart can be in different drainage classes.
- SSURGO describes soil, not construction. It knows nothing about your footing drain, your grading or the previous owner’s drainage work.
- Depth to water table is a modelled annual minimum in SSURGO itself, not a monitoring-well reading.
- Where a soil survey area is statewide rather than per county (common in New England), our county attribution is coarser. Those counties are fetched by centroid and the method is stated on the page.
- Radon zone is an expectation, as described above.
- The basement-first / foundation-first switch is by state, not by county. In North Carolina and Tennessee the mountains behave differently from the coastal plain, and a state-level switch cannot capture that.
- Alaska is a different physics — frost heave and permafrost dominate, and SSURGO coverage there is thinner than in the lower 48.
- Manassas Park city, Virginia has no ZIP code of its own, so it has no page. That is 1 of 3,143 counties and county-equivalents.
- Prices are a model. Repeating this because it is the number people screenshot.
Page data updated