Crop Water Stress Index.

Every other input on this site measures the plant's environment: the light on it, the air around it, the water and salt at its roots. This one measures the plant's response. A well-watered canopy transpires and runs a few degrees cooler than the air; a thirsty one closes its pores, stops cooling, and heats up. The Crop Water Stress Index turns that temperature gap into a single number, and the plant tells you it is stressed before it ever wilts.

Leaves cool themselves the way skin does: by evaporating water. When a plant is well supplied, its stomata, the pores on the leaf, stay open, water vapor escapes, and that evaporation carries heat away, holding the canopy below air temperature. When water gets hard to reach, the plant defends itself by closing those stomata to stop losing water. The cooling stops, and the leaf warms toward, then above, the surrounding air. Canopy-minus-air temperature is therefore a direct readout of whether the plant is actually managing its water, integrating everything happening at the roots and in the air into one honest signal from the plant itself.

Ask the plant, not the soil.

This is what makes CWSI special. You can have perfect moisture, perfect tension, perfect EC, and still have a plant under water stress because the air is pulling water out of the leaf faster than the roots can supply it, a problem no root-zone sensor can see. CWSI sees it, because it reads the outcome at the leaf. It is also the earliest objective stress signal: the canopy warms hours before a leaf visibly droops, so you can act while it's still cheap to act.

Why the leaf runs cool.

The same transpiration that cools the leaf is the engine that lifts water and dissolved nutrients up from the roots, so a canopy that has stopped cooling has also stopped pulling its feed. That links CWSI directly to the rest of the system: it is the plant-side mirror of VPD (which sets how hard the air pulls), of tension (how hard the roots must pull back), and of root oxygen (whether the roots have the energy to pull at all). When any of those go wrong, the canopy warms.

The index needs context: it's computed, not raw.

Raw canopy temperature alone can mislead, because how cool a healthy leaf should run depends on the air. On a hot, dry day the air pulls hard and even a well-watered plant can't cool as much; on a cool, humid day the baseline is different again. CWSI corrects for this by scaling the measured gap between two references, the gap a fully-watered crop would show and the gap a fully-stressed crop would show under the same conditions, using VPD to set the well-watered baseline. The result is a normalized number from 0 (no stress) to 1 (maximum stress) that means the same thing across weather. This is why CWSI lives downstream of air temperature and humidity: it is derived from canopy temp plus the atmospheric demand, not read straight off a sensor.

How to read it.

The hardware is simple: a non-contact infrared thermometer aimed at the canopy reads leaf temperature without touching it (the inexpensive MLX90614 is the DIY part), and you already have air temperature and humidity from the environment side. A small node that reads the IR sensor and reports canopy temperature, air temperature, and the gap between them is enough to compute the index wherever VPD is known; that node emits canopy and air temperature as plain oat-ods readings, and the index is derived from them downstream. The gatherer lands the temperatures in your own data, tied to the zone.

The trap: reading canopy temperature without the air.

The trap is treating a bare canopy temperature as the answer, panicking on a hot afternoon when the leaf is warm for a perfectly legitimate reason, or aiming the infrared sensor at the pot, the floor, or a sunlit wall instead of clean leaf. CWSI exists precisely to remove that ambiguity: feed it the air conditions, point it at canopy, and a rising index is a real signal you can irrigate by, not a thermometer reacting to the weather.

See the canopy, not one spot: why a thermal image beats a better thermometer.

A spot sensor gives you the average of whatever sits in its cone: leaf, gap, pot, floor, all melted into one number, and there is no way to un-mix them afterward. A thermal image lets you keep only the leaf pixels and throw the rest away. That leads to a conclusion that sounds backwards and is not: a cheap thermal camera accurate to 2 degrees, reading only leaf pixels, beats a precise spot sensor accurate to half a degree that is averaging leaf and floor. The camera removes a systematic error (measuring the wrong things) where the better sensor only shrinks a random one. You are not buying precision; you are buying the ability to know what you are pointed at.

The image carries a second, earlier signal too. A canopy under developing stress does not warm evenly: individual plants and leaves cross the line first, so the spread of temperatures widens before the average moves at all. A spot sensor is blind to that spread; an image hands it to you for free, along with where in the room it is happening, which is the difference between an alert and an alert you can walk to. The hardware side of this, and what resolution the job needs, lives at Imaging in agriculture.

Why a temperature probe in the canopy can't do this.

The tempting shortcut is to hang a cheap temperature probe among the leaves and call its reading leaf temperature. It is not. A probe reads the temperature of the probe: its metal or plastic body absorbs the grow light's radiation nothing like a leaf does, and under strong light it can sit several degrees above the air on its own account. But the deeper problem is the one that kills the idea outright: a leaf transpires and a probe does not. The entire CWSI signal is the leaf running below air temperature because evaporation is carrying heat away. An object that cannot evaporate can never show that. It will read at or above air, always. The probe is not inaccurate at this job; it is structurally incapable of it.

The same probe is excellent where conduction rules and radiation does not matter: buried in the substrate for root-zone temperature, sunk in the reservoir, strapped to a supply line. Keep it there. Leaf temperature is a job for a non-contact infrared eye.

Two reference leaves make the number honest.

The index needs its two baselines: how cool a fully-watered canopy would run and how warm a fully-stressed one would, right now, in this room. Computing those from air readings and theory is where most CWSI attempts go wrong. There is a hardware answer, and it is pleasingly cheap: build two artificial leaves and let them be the baselines. Two thin metal discs, painted matte green so they absorb light the way a leaf does. Keep one wet from a small wick and reservoir: it evaporates flat out, like a leaf with all the water it wants. Coat the other with petroleum jelly so it cannot evaporate at all: a leaf with its pores shut. Bond a contact temperature probe to the back of each, a job that probe is genuinely right for, and the pair reports the wet and dry baselines continuously, as ordinary oat-ods readings.

Then any thermal reading of the canopy converts on the spot: the index is how far the canopy sits between the wet reference and the dry one. And because all three surfaces hang under the same lights, the heating that grow lights add lands on all of them equally and cancels out of the comparison. That solves a scheduling problem that otherwise spoils everything: lights-on thermal readings are normally polluted by the lights themselves, and lights-off is exactly when nobody is working. With the reference pair in place, the lights-on reading a person takes on a normal walk-through is valid. A few dollars of metal, paint, and wick, fixing a physics problem. We have not built this pair yet; the design comes from published research on reference surfaces, and when ours is running, what we learn lands here.