Field NotesPlant techNo. 12
How to calibrate a soil moisture sensor for your potting mix
Raw counts and generic curves don’t know your mix. Weighed water does. Here is the gravimetric method, the math behind it, and where the errors creep in.

The short answer Pack your mix into a container of known volume at its in-pot density. Record the sensor’s raw reading at four to six moisture levels from air-dry to near saturation, weighing the mix each time. Then dry a subsample, convert each weighing to volumetric water content (VWC), and fit VWC against the reading.
A capacitive board gives you a count, say 2,200 out of 4,095; a commercial probe gives a percentage from a generic equation. Neither is your mix’s water content until you check it against weighed water.
That check, a gravimetric calibration, takes a few hours per sensor plus two days of drying. Capacitive vs. resistive soil moisture sensors covers how the sensors work.
Why a factory calibration misses your mix
Delta-T’s ML3 manual gives rough permittivities of 81 for water, 4 for soil solids and 1 for air, and treats the solids’ contribution as constant for a given soil, so a new mix needs a new equation. UF/IFAS Extension recommends soil-specific calibration for capacitance sensors because they mostly run below 100 MHz, where readings are more affected than TDR’s by temperature, salinity, bulk density and clay content.
Generic curves assume a soil type. Delta-T’s mineral calibration is for soils under 7% carbon with bulk density above 1.0 g/cm³, its organic one for soils above 7% carbon and below 1.0 g/cm³. Peat, coir and bark mixes sit on the organic side. The cost:
- METER rates its TEROS 10 at ±0.03 m³/m³ on the generic calibration in mineral soil and ±0.05 on its soilless-media calibration, versus ±0.01 to 0.02 when calibrated to the medium.
- In an organic-rich garden soil, Nagahage and colleagues found a commercial probe’s manufacturer equation had an RMSE of 0.08 cm³/cm³ in the field-capacity range, against 0.03 after soil-specific calibration.
- Low-cost boards output a raw signal, not VWC. Schwamback’s team, citing earlier studies, notes that the SEN0193 capacitive board (also used in Nagahage’s study) is sensitive to soil texture and bulk density and needs a local calibration.
The calibration, step by step
You need a rigid, straight-sided container big enough to surround the sensing zone with mix (1 liter here), scales reading 0.1 g and 0.01 g (METER’s specified resolutions), an oven thermometer and a gallon of your mix. Use the board, supply and cable you’ll actually run; METER tells EC-5 users on third-party loggers to calibrate at the excitation voltage they’ll use in the field. Get the container’s volume by weighing the water that fills it to a marked line: 1 g of water is 1 cm³.
- Air-dry the mix in a thin layer and break up clumps. Dry mix is dusty, so do this outdoors or wear a dust mask, and wash your hands afterward. Don’t sieve out bark or perlite; METER notes that for materials like compost, removing large particles changes the material.
- Match your pot density. Divide the mass of air-dry mix that fills one of your pots, firmed as usual, by the pot’s volume. Delta-T notes that compressible composts often need rebuilding at their in-place density.
- Subsample. Weigh two 5 to 10 g subsamples of the air-dry mix immediately (don’t let one big bark chunk dominate either) and set them aside; use their average.
- Pack. Tare the container, pack firmed layers to that density and the fill line, and weigh.
- Read. Insert the sensor straight, in one push, electronics above the mix. Let the packed mix equilibrate for 15 to 20 minutes before you read (Nagahage’s team’s wait), log until readings level off and average them, then reinsert in fresh spots once or twice and average again (METER).
- Wet and repeat. Add about 1 mL of water per 10 mL of volume, which METER says raises VWC by roughly 0.10. Mix evenly, repack to the line, weigh and read. Stop near saturation.
- Dry the subsample to constant weight, let it cool in a covered dish, and weigh it right away.
Sources differ on drying temperature. METER’s default is 105°C (221°F) for 24 hours, but high-organic soils can lose volatile organics at that heat, which the math counts as water, so it recommends 60 to 70°C (140 to 158°F) for at least 48 hours. Delta-T says organic soils and composts are usually air-dried, to avoid burning off volatile fractions. That’s gentler, but METER treats air-dry soil as still holding some water, so air-drying reads every VWC slightly low. If you use a kitchen oven, check with the oven thermometer that it can hold 60 to 70°C (many can’t go that low, and some switch off on their own after several hours). Dry the subsample in a small dish you don’t cook with, keep the kitchen ventilated, and turn the oven off whenever you go out or go to bed; drying can pick up where it left off, so keep going until two weighings a few hours apart match.
Turning weights into VWC
θg = (m_wet − m_dry) / m_dry
θg is gravimetric water content (g/g); m_wet and m_dry are the moist and oven-dry masses of the mix (g).
m_dry = m_air × (s_dry / s_air)
m_air is the air-dry mass you packed; s_air and s_dry are the subsample’s masses before and after drying.
ρb = m_dry / V
ρb is dry bulk density (g/cm³); V is the container volume (cm³).
θv = (m_w / ρw) / V
θv is VWC in cm³/cm³ (the same as m³/m³); m_w = m_wet − m_dry is the water mass (g); ρw is water’s density, 1 g/cm³.
θv = θg × ρb / ρw
For what the numbers mean to a plant, see what is volumetric water content.
Worked example (illustrative numbers)
These numbers are invented to show the arithmetic, not measured. They mimic a capacitive board on a 12-bit ADC (0 to 4,095 counts) that reads lower when wetter.
The container holds 1,000 cm³ and you pack 225.0 g of air-dry mix. A 9.00 g subsample weighs 8.00 g after drying at 65°C, so m_dry = 225.0 × (8.00 / 9.00) = 200.0 g and ρb = 200.0 / 1,000 = 0.200 g/cm³.
| Point | Raw (counts) | Wet mix (g) | Dry mix (g) | Water (g) | θv (m³/m³) |
|---|---|---|---|---|---|
| 1 (air-dry) | 2,905 | 225.0 | 200.0 | 25.0 | 0.025 |
| 2 | 2,480 | 339.0 | 200.0 | 139.0 | 0.139 |
| 3 | 2,230 | 451.0 | 200.0 | 251.0 | 0.251 |
| 4 | 2,005 | 562.0 | 200.0 | 362.0 | 0.362 |
| 5 | 1,845 | 668.0 | 200.0 | 468.0 | 0.468 |
| 6 (wettest) | 1,680 | 779.0 | 200.0 | 579.0 | 0.579 |
Check point 4 the other way: θg = 362.0 / 200.0 = 1.810, and 1.810 × 0.200 = 0.362.
Fit θv against x, the raw reading divided by 1,000. A least-squares straight line gives:
θv = 1.2951 − 0.4524x
R² is 0.967, but RMSE is 0.034 m³/m³ and it predicts −0.019 at point 1. A quadratic fits far better (R² 0.9998, RMSE 0.003, worst miss 0.005):
θv = 2.4372 − 1.4834x + 0.2248x²
At 2,100 counts, x = 2.100, so θv = 2.4372 − (1.4834 × 2.100) + (0.2248 × 4.410) = 2.4372 − 3.11514 + 0.99137 = 0.3134 m³/m³, or about 31% VWC.
Stay inside the calibrated 1,680 to 2,905 counts; past about 3,300, this curve turns back up.
Linear, polynomial or two-point?
METER’s method usually yields 4 to 6 points; both board studies cited here used 11. Give a quadratic at least five or six points and check the residuals as well as R².
METER says the relationship is often linear but sometimes better fit by a polynomial, especially in soils high in organic matter. In a sandy soil, a straight line through all 63 of Schwamback’s sensors still reached R² 0.949; in organic-rich soil, Nagahage’s team fit a cubic (R² 0.98). Plot before you choose.
The two-point shortcut
Mapping a reading in air to 0% and one in water to 100% gives a share of the sensor’s span, not VWC. Nagahage’s team found such a line inaccurate in both media they tested.
Dry and wet points in your packed mix are better. Delta-T’s ML3 manual uses them, but its model is linear by design (in the square root of permittivity), and the manual rules the method out for soils that shrink or turn very hard when dry. Raw counts aren’t linearized: in the example, a line through points 1 and 6 reads 0.07 to 0.08 too wet at points 2 to 4. Add a middle point to test any line.
Sources of error to control
- Air gaps. Air’s permittivity is about 1, so a gap reads as dry. UF/IFAS stresses good contact and careful installation.
- Packing density. If your calibration density is more than about 20% off your pots, METER suggests repeating it at a more realistic density.
- Sensor-to-sensor variation. Four boards of one model differed significantly in Nagahage’s tests, and a universal curve from 63 sensors in Schwamback’s spread widely at the dry and saturated ends. Calibrate and label each board.
- Temperature. Schwamback measured about 0.001 m³/m³ per °C (0.015 over a 20°C swing) for the SEN0193, and Delta-T notes water’s permittivity falls as temperature rises. Calibrate near the pot’s usual temperature.
- Salinity and fertilizer. METER quotes the TEROS 10’s generic accuracy only below 8 dS/m solution EC; Delta-T gives the ML3 up to 3.5% vol salinity error over 50 to 500 mS/m. If you fertilize every watering, calibrate with that solution.
- Supply voltage and wiring. In a one-sensor test, Schwamback’s team got a better fit at 3.3 V (R² 0.871) than at 5.5 V (R² 0.798) and suggests 3.3 V. Treat the board, supply and cable as part of the calibration.
Checking it later, and when to redo it
For a spot check, core the mix beside the sensor with a short tube (METER suggests cutting a 3 to 5 cm length of 1.5 to 2.5 cm diameter conduit or thin-walled rigid tubing), then weigh, dry and compute θv from its volume:
V = π × r² × h
r is the tube’s inner radius and h its length (cm). Coring disturbs roots, so do it at repotting or in a spare pot. If core and sensor disagree by more than the ±0.03 to ±0.05 METER quotes for its generic calibrations, your curve is doing no better than a generic one, so recalibrate.
Redo it when you change mixes, repot or pack differently, swap any of the hardware, or change fertilizer strength a lot. Mixes matter: Nagahage’s sensor did not perform acceptably in a 1:1 blend of their soil with vermiculite.
Keep a sheet with each sensor: the mix, packed density, date, coefficients and the count range they cover.
Sources
- Method A: Soil-specific calibrations for METER soil moisture sensorsMETER Group
- TEROS 10 Tech SpecsMETER Group
- User Manual for the ML3 ThetaProbe Soil Moisture Sensor (ML3-UM-2.1)Delta-T Devices (copy hosted by distributor Instrument Choice)
- Field Devices For Monitoring Soil Water ContentUF/IFAS Extension, University of Florida (EDIS BUL343)
- Automated Low-Cost Soil Moisture Sensors: Trade-Off between Cost and AccuracySensors (MDPI), Schwamback et al., 2023
- Calibration and Validation of a Low-Cost Capacitive Moisture Sensor to Integrate the Automated Soil Moisture Monitoring SystemAgriculture (MDPI), Nagahage, Nagahage and Fujino, 2019
General information, not professional advice for a specific plant or crop. Spot something wrong? Email info@terrathink.io.


