How Does a Soil Moisture Sensor Work, and How Accurate Are Its Readings?

Your soil moisture sensor says 28 percent, and you want to know whether the bed needs water. Here is the short answer: the sensor does not detect water directly. It measures how strongly the soil around its prongs affects an electric field and converts that into volumetric water content, the share of the soil volume that is filled with water. How accurate that number is depends less on the device than on your soil, on how well the probe touches it, and on whether the calibration matches your soil type.

To use the reading, you need to know what the number physically means and which number is actually good for your soil and plants. Twenty-eight percent can be soaking wet in sand and close to dry in clay.

How does a soil moisture sensor measure water content?

Almost every electronic soil moisture sensor relies on dielectric permittivity, a measure of how strongly a material stores an electric field. Air has a value of 1, dry soil minerals sit around 3 to 5, and liquid water is about 80. Because water stands out so sharply, the water content largely determines the permittivity of the soil as a whole. The electronics measure that value and use a calibration curve to convert it into percent by volume.

A capacitive sensor forms a capacitor from its electrodes and the surrounding soil. More water means more capacitance, which shifts the frequency of an oscillator circuit. Cheap hobby boards and many professional probes work this way.

FDR (frequency domain reflectometry) is a close relative. The probe becomes part of a resonant circuit, and the resonant frequency or the impedance at a fixed frequency reveals the permittivity. Typical measurement frequencies range from roughly 10 to 150 MHz. According to the manufacturer, the Dragino SE01-LB uses FDR and also compensates its moisture value using soil temperature and electrical conductivity.

TDR (time domain reflectometry) sends a very short electrical pulse down the metal rods and times how long it takes to travel to the end and back. The wetter the soil, the slower the pulse. Thanks to its very high frequency components, TDR is less sensitive to salt and serves as the research reference method, at a considerably higher cost. For mineral soils, the travel time is often converted into water content with the empirical Topp equation published in 1980.

Simple two-electrode probes only read electrical resistance, which depends heavily on salt, and their bare electrodes corrode. Tensiometers measure soil water tension in kilopascals, not water content.

What do soil moisture readings in percent mean?

On a dielectric sensor, “28%” usually means 28 percent by volume: one liter of soil contains 0.28 liters of water. The technical term is volumetric water content, often written as 0.28 m³/m³. The rest of the volume consists of solid particles and air-filled pores.

That sets a hard limit. A sensor in real soil can never reach 100 percent. A typical mineral soil has a pore volume of about 40 to 50 percent, so even when every pore is full of water, the reading lands roughly there. A probe that consistently shows much more is likely in standing water or very loose growing media, or miscalibrated.

Two other percentages are easy to confuse with this one. Gravimetric water content relates the water to the mass of dry soil. At a bulk density of 1.4 g/cm³, 20 percent by mass corresponds to about 28 percent by volume. And some hobby sensors display a relative index between “air = 0%” and “glass of water = 100%.” That index has little to do with actual water content and cannot be compared with published tables or other devices, so check the datasheet before comparing two percentages.

What is the ideal soil moisture level for lawns, crops and trees?

There is no single number, because the ideal range depends on soil texture. Two reference points define it. Field capacity is the water content a soil still holds one to three days after heavy rain, once excess water has drained away. The wilting point is the water content below which plants can no longer extract water from the soil. The range between the two is the plant-available water.

The table below lists typical ranges based on the FAO guidelines (Irrigation and Drainage Paper 56).

The table shows why thresholds borrowed from someone else rarely fit: 18 percent is close to field capacity in sand, while clay at 18 percent is already below its wilting point.

Water before the plant has used up a set share of the available water. The FAO gives guideline values of about 40 to 50 percent for turf grass, 30 to 45 percent for shallow-rooted vegetables and about 50 percent for fruit trees such as apple or pear. A worked example for loam with a field capacity of 30 percent and a wilting point of 12 percent: 18 percentage points are available. For a lawn with 40 percent allowable depletion, the watering threshold is about 23 percent; for lettuce at 30 percent, about 25 percent; for an apple tree, about 21 percent. The best way to find your own field capacity is to look at the reading one to three days after a soaking rain.

How do you calibrate a soil moisture sensor for your soil?

Out of the box, a soil moisture sensor is usually calibrated for typical mineral soils, and Dragino states this explicitly for the SE01-LB. In your soil the curve can still be off by several percentage points. A soil-specific calibration compares sensor readings with an independent reference measurement.

The reference is the oven-drying method. Use a sampling ring of known volume, such as 100 cm³, to take an undisturbed sample right next to the probe, and note the sensor reading at the same moment. Weigh the sample, dry it at 105 °C for about 24 hours and weigh it again. The mass lost in grams equals the water in cubic centimeters, and dividing by the ring volume gives the volumetric water content. A kitchen scale with 0.1 g resolution is good enough.

One data point is not enough. Repeat the comparison at no fewer than three clearly different moisture levels: dry, medium and shortly after rain. Fit a straight line through the pairs. If only the intercept is off, an offset will do. If the slope is off too, you need a linear correction.

Where you apply it depends on the device. The Milesight EM500-SMTC accepts a calibration value by downlink for soil moisture, temperature and conductivity that shifts the reported value. That is an offset, not a new response curve, so a slope correction belongs in the decoder or your IoT platform. The Dragino SE01-LB can send raw values in MOD=1 mode, and Dragino explicitly points to a conversion in the platform for other soil types. Correct in one place only: an offset in the device plus a second one in the platform shifts the data twice.

Be careful with the popular two-point calibration in air and water. One instrument maker notes that well-calibrated soil probes often read only 50 to 60 percent in water and slightly negative values in air. That is expected, because they are optimized for real soil. Air and water work as a function check, but they cannot replace a soil curve.

Why do clay and sandy soils give different readings?

There are two separate reasons. The first is physics. Sand has large pores and little internal surface area, so it drains quickly and holds little water. Loam and clay have many fine pores and hold much more. With the same amount of irrigation, loam really does show a higher water content, and that is not a measurement error.

The second reason lies in the measurement itself. Clay minerals bind some water so tightly to their surfaces that it has a lower permittivity than free water. At the same time, clay-rich soils conduct electricity better, which distorts the reading, especially at low measurement frequencies. Depending on the soil and the sensor, the error can go in either direction. Bulk density and organic matter shift the curve as well, which is why potting mixes and growing media often look implausible with a mineral-soil calibration. Stony soil adds low-permittivity material to the measurement volume on top of that.

In practice, compare readings only within the same soil type and set thresholds per location: a sandy raised bed and a loamy orchard need different limits.

Do salt and fertilizer affect soil moisture readings?

Yes, and the lower the measurement frequency, the stronger the effect. Dissolved salts make soil water electrically conductive. A capacitive or FDR probe interprets part of that conduction as additional permittivity and reports more moisture than is actually present. TDR is affected less, although in very saline soils its signal weakens too.

In a garden this shows up mostly after feeding: fertilizer near the probe raises conductivity, and the moisture curve can jump for a few days even though little water was added. Irrigation water carries salts as well; Dragino gives a typical conductivity of under 800 µS/cm for irrigation water.

This is where a sensor that also measures conductivity pays off. Both the Milesight EM500-SMTC and the Dragino SE01-LB record it up to 20,000 µS/cm, so you can see whether a sudden rise in moisture coincides with a jump in conductivity. The SE01-LB also uses conductivity to compensate its moisture value. In strongly saline soils, however, compensation does not remove the influence completely.

Why does a soil moisture sensor also measure soil temperature?

First, temperature is a correction factor. The permittivity of water decreases as temperature rises, and the electronics drift as well. Conductivity reacts even more, rising by about 2 percent per degree Celsius, which is why EC values are normally referenced to 25 °C; the Dragino SE01-LB datasheet states this explicitly. Without a temperature channel, a warm afternoon can look like a small increase in water on the moisture chart.

Frost is a special case. Ice has a permittivity of only about 3, similar to dry rock. When the ground freezes, the moisture reading drops steeply even though the water is still there. The temperature channel tells you right away that you are looking at frost, not drought.

Soil temperature is also useful in its own right: germination, root activity and nutrient uptake depend on it, and in spring it shows when the ground is warm enough for seeding.

How accurate are soil moisture sensors?

Datasheet figures apply to the conditions under which the manufacturer calibrated the sensor. The table compares the two sensors we carry.

These figures are percentage points of water content. At ±3 points, a displayed 25 percent could mean anything from 22 to 28 percent. In a soil that does not match the factory calibration, published research on dielectric sensors reports errors of several percentage points, and a soil-specific calibration typically brings that down to about 1 to 2 points.

Don’t confuse accuracy with resolution: two decimal places are not more accurate, only more finely stepped. Repeatability at a fixed spot is usually much better than absolute accuracy. For watering decisions, what matters most is whether the value is rising or falling and where it sits relative to your own field capacity. A well-installed sensor often does that job even without a custom calibration.

Why is my soil moisture sensor reading wrong?

Before you recalibrate anything, rule out the common causes; most outliers have nothing to do with the calibration curve.

• Air gaps or debris: air around the rods pushes the reading down, and Milesight warns that sticks, bark or roots stuck between the prongs severely affect the data. After heavy rain, freshly loosened soil settles and the value jumps for no obvious reason.
• Partly exposed rods: if the prongs stick out of the soil, the probe measures air as well.
• Frost: a sudden drop at temperatures below 0 °C is frozen water, not dry soil.
• Fertilizer or salt: if moisture rises together with conductivity and it has not rained, salts are the cause.
• Double correction or wrong scaling: an offset in the device plus another in the platform, or a decoder that divides by 10 instead of 100.
• Missing data shown as zero: a lost radio message should never appear as 0 percent on the dashboard.

How and where to install a probe, including depth, distance from roots and choice of location, is covered in a separate article of ours. Here, only one point matters: a new value after moving a probe is no proof of changed moisture, because the probe now sees different soil. Note such changes in the history.

Where can you buy a soil moisture sensor?

Start with the measured quantity. A sensor that reports volumetric water content in percent by volume can be compared with the field capacity and wilting point of your soil; a relative index cannot. Then look at the measuring principle. Capacitive and FDR probes are practical for everyday use but react to salt, so an additional conductivity channel is worth a lot. A temperature channel helps with compensation and exposes frost.

Also consider how you can calibrate: is a device offset enough, or do you get raw values to build your own curve in the platform? Compare the accuracy figures in the datasheets and remember that they apply to the factory calibration. For gardens, fields and greenhouses without power, a battery-powered LoRaWAN sensor that reports its values automatically over a long period, without cables, is the sensible choice.

If you are looking for a suitable soil moisture sensor, you will find our models in the soil moisture sensors category.