Soil Data Sources: Reading the Ground Before You Model It
Weather data is everywhere — forecasts, satellite feeds, public APIs by the dozen. Soil data is where field-scale modeling gets hard. The ground under a field changes over meters, hides most of its properties below the surface, and reveals them only to whoever goes looking with the right instrument. Yet every irrigation run, planting date, and drainage decision leans on assumptions about that hidden layer, and a model fed the wrong soil assumptions will confidently schedule the wrong irrigation.
The good news: decades of public investment have produced soil data sources that are free, documented, and genuinely useful — if you know what each one can and cannot tell you. This entry maps the three families every practitioner draws from, and how they combine.
Three Ways to Know a Field
Think about a field’s soil the way a doctor thinks about a patient:
- The chart — the accumulated written record: public soil surveys, compiled by professionals over decades. Broad, authoritative, and slightly out of date by design.
- The vitals monitor — in-field sensors reporting continuously: what is the moisture right now, at this spot, at this depth?
- The blood panel — laboratory samples: precise, quantitative answers about one place at one moment, at a price that keeps you from ordering them constantly.
A doctor who reads only the chart misses what’s happening today; one who watches only the monitor doesn’t know the patient’s history; the blood panel is definitive but expires the moment conditions change. Soil work is the same: each family covers the others’ blind spots, and the craft is in the combination.
The Chart: Public Soil Surveys
SSURGO and the Web Soil Survey. The Soil Survey Geographic Database (SSURGO) holds the results of the National Cooperative Soil Survey’s fieldwork in the United States — information gathered, in the NRCS’s own words, “by walking over the land and observing the soil” over the course of a century, with many samples analyzed in laboratories along the way. The maps divide the landscape into map units: areas with distinct soil components, properties, and productivity, mapped at scales from 1:12,000 to 1:63,360. Each map unit links to tabular data covering properties a grower actually plans around — available water capacity, soil reaction (pH), electrical conductivity, flooding frequency, and estimated yields among them. The free Web Soil Survey puts all of it behind a browser: NRCS reports soil maps and data available online for more than 95 percent of the nation’s counties.
Two caveats keep SSURGO honest. First, a map unit is not your field: a single unit may contain one to three major soil components plus minor ones, and the mapped boundary won’t follow your fence line. Second, the survey is a record, not a reading — it describes what surveyors found, at survey scale, when they walked the land; it cannot tell you what compaction, drainage work, or this morning’s rain have done since.
SoilGrids. Outside well-surveyed countries — or when you need one consistent global layer — ISRIC’s SoilGrids takes a different approach: machine-learning models trained on soil profile observations from the WoSIS database plus environmental covariates, predicting fourteen soil properties (pH, organic carbon, bulk density, sand, silt, clay, cation exchange capacity, and total nitrogen among them) at 250-meter resolution for six standard depth intervals, worldwide, under a CC-BY 4.0 license. The trade is explicit: SoilGrids is prediction, not observation. Every value is a model estimate with published uncertainty, which is exactly why the project ships uncertainty maps alongside the property maps — read them together, not separately.
The Vitals Monitor: In-Field Sensors
Surveys tell you what kind of soil you have; sensors tell you what it’s doing right now. The reference example is the statewide mesonet — a network of permanent environmental monitoring stations. The Oklahoma Mesonet, a joint project of the University of Oklahoma and Oklahoma State University, measures soil moisture at 5, 25, and 60 centimeters below the sod, sampled every 30 minutes, alongside soil temperature at multiple depths — a public, continuous record of the water actually in the ground, station by station, that shows how moisture moves down the profile after rain and dries back out. Several states run comparable networks; on-farm probes bring the same idea inside your own fence at whatever density you can afford.
The blind spot is the mirror image of the survey’s: a sensor is a point, not a field. A probe reads the few centimeters of soil around it, and translating its raw signal into volumetric water content depends on the soil it sits in — which sends you straight back to the chart to learn what that soil is. Siting matters too: a probe in the one sandy streak of a clay field will cheerfully mislead you all season.
Sensors also rarely stand alone: soil moisture is driven by weather, so any moisture model needs the atmospheric side of the ledger. NASA POWER serves solar and meteorological data derived from satellite observations and models — over 300 parameters, with an agroclimatology community tailored to exactly this use — through a free API. The archive’s Importing from APIs entry covers the mechanics of pulling feeds like it.
The Blood Panel: Laboratory Samples
A soil sample sent to a lab is the ground truth the other two families are calibrated against: measured texture, organic matter, nutrient levels, pH — numbers from your field, not a survey polygon or a model cell. The limits are the ones cost imposes. A sample describes one spot on one day; soil chemistry varies across a field and through a season; and budgets allow a handful of samples per field per year, not a continuous record. That sets the lab’s role in the combination: not your everyday data source, but the periodic reference that anchors the chart and checks the monitor.
Choosing Sources: A Working Combination
| Family | Example | Best at | Blind spot |
|---|---|---|---|
| Public survey | SSURGO / Web Soil Survey, SoilGrids | Soil identity: texture, water capacity, mapped extent — free, documented, everywhere | Today’s conditions; sub-map-unit variation |
| In-field sensor | Mesonet stations, on-farm probes | Timing: continuous moisture and temperature at known depths | One point per probe; needs soil-specific calibration |
| Lab sample | Any accredited soil lab | Accuracy: measured chemistry and texture from your actual field | A snapshot — sparse in space and time |
In practice the combination runs in that order. Start with the chart: pull your field’s map units, texture, and available water capacity from the Web Soil Survey — it’s the free foundation every later step interprets against. Add the clock: a nearby mesonet station or an on-farm probe for live moisture, with a weather feed such as NASA POWER supplying the rain and temperature record that explains what the probe shows. Anchor with the panel: periodic lab samples to confirm what the survey claims and calibrate what the sensors read.
In Action at Dendrology
The archive’s Irrigation Rigging entry describes the two halves of irrigation scheduling — soil-type analysis and real-time soil-moisture monitoring. This entry names where each half’s data actually comes from: the type analysis is the chart (survey texture and water capacity), and the monitoring is the vitals (probes and mesonet feeds), with weather forcing tying the two together.
In the Precision Agriculture Specialist learning path, this entry follows the checkpoint where you map a season’s water and timing decisions — and the path’s later field-dataset checkpoint has you pull a season of daily weather into a CSV. Pair that file with your field’s survey data and you have, in miniature, the same layered picture Dendrology’s modeling work is designed to consume: stable soil identity from surveys, live state from sensors, and weather driving the change between readings.
Beyond the Basics
- gSSURGO — SSURGO repackaged as a gridded (raster) database, the form you want when soil layers need to align with satellite imagery or model grids.
- Soil Data Access — NRCS’s web services for querying SSURGO programmatically, when the point-and-click Web Soil Survey becomes the bottleneck.
- Pedotransfer functions — estimating the hydraulic properties you can’t easily measure (like water retention curves) from the properties surveys record (like texture): the standard bridge between survey data and simulation models.
- Uncertainty layers — SoilGrids publishes per-prediction uncertainty; treating it as part of the data, not a footnote, is what separates careful use from map-reading.
- The view from above — the planned Introduction to Remote Sensing entry picks up the one vantage point this entry leaves out: reading fields from satellites and drones.
Three families, one discipline: know what each source is actually measuring, and the ground stops being a guess — it becomes one more data layer Dendrology can put to work.