REFLECTIONS ON USING SOIL MOISTURE SENSORS 2025

No, the photo is not a bad photoshop. That is just how it came out.

During the 2025 cropping season, soil moisture sensors were installed as part of projects looking at digital tools for use in precision agriculture. One of the units was a set of three water mark tensiometer type sensors connected to a LoRa transceiver. The other unit was a single TDR type sensor, also connected to LoRa transceiver.

Locations

Four sites in two fields were selected (eight sites total). A map of soil conductivity, measured with an EM38 sensor, was used to select two points of relatively high and two points of relavtively low conductivity. Once in the field, the final site was selected to be a couple of meters off the wheel tracks of the spray unit.

The fields were selected based on that there was yield mapping in them and that they would be within the coverage of the LoRa network. One of the fields that was originally selected was at the edge of the the expected LoRa coverage, but given it was at a low point in the landscape and surrounded by trees, there was a need to test connectivity. The tested showed that there was no coverage. It was revealed by the farmer that they often lose satellite connectivity in this field, because of the trees, so no LoRa coverage was probably not that big of a surprise. The second field was thus move to a location near to our office with strong LoRa coverage, but unfortunately without EM38 mapping. These sites were instead selected to be down a single sow row but at points that should vary in soil: top of the hill, bottom of the ditch, in the really sandy patch of the field.

Installation

The units were installed in pairs, with a single TDR and three watermark sensors at each of the eight sites. The TDR sensor was installed to 15 cm. The watermark sensors were installed at 15, 30 and 45 cm. The TDR sensor was installed by digging a hole to 20cm and inserting the probes at the right depth. Soil was backfilled around the sensor with care, then the rest of the hole filled. To insert the watermark sensors, holes of a depth that sat the middle of the sensor at the desired depth were drilled with a “spiralborr” (wood auger?), the sensor inserted, and the hole back-filled will finer dirt and a lot of water. The backfilled soil was packed in with a folding ruler (tumstock). The wood auger needed an extender to reach the deepest depth. It was also fairly worn out after drilling some 40 holes. It was a fairly cheap version, but it was being used in abrasive soil and hit a lot of stones (thus why the total number of holes is more than 8 sites x 3 sensors + 1 for the pole the transceivers were mounted on). Using a battery powered drill and the wood auger made it really easy to drill the holes.

Each site took approximately one hour to set up: dig/ drill holes, insert sensor, backfill, start transceivers, all while trying to not walk on the crop.

Starting the transceivers was done by opening them and inserting the battery. It was possible to have a paper stopper between the battery and connector which could just be pulled out and the box sealed, but the choice was made to open the boxes fully so that the LED that signalled that the unit was working could be seen. This took a bit of extra time, to open and close the boxes fully, but only a minute or two.

During the season

The situation was monitored from the desk. Some connections were lost, either temporaily or more or less permanently. This was likely a result of the poles supporting the transceivers lying down.

The growers didn’t mention any issues with needing to work around the sensors, except for ensure the spray boom was lifted a little during spraying so that it didn’t crash into the transceiver.

Removal

Finding the units was easy using just the coordinates, to 5 decimals in WGS84 degress-decimal, and Google maps on the mobile. The hope was that the units could just be lain on the ground for the harvester to work over. After observing the harvest from the cabin, it was decided that the best option to remove the sensors without impacting the yield or yield mapping was for the sensor cables to be cut, the transceivers removed, and the sensors dug up once the field was harvested. Finding the sensors after the harvester had passed over was simiarly easy using the coordinates, and in most cases the cables were visible directly. In the case that the cables had been driven over by the harvester, it was somewhat more difficult to find the cables as they were stuck to the soil surface and covered in charf. It took 20 minutes of scratching around to find them. But this did confirm that cutting the cables and removing the units was the right move.

Digging up the sensors took a bit of effort and time – the type that would make a user swear to never use this system again. That said, it was only the sensors to 45 cm that took a lot of effort. Those at 15 cm were up in a matter of two shoves-full.

Improvements

The ultimate improvement would be to have sensors that don’t need to be removed so often. Next would be have a system were there sensor could just be pushed in and pulled out directly.

Start up without having to open the box: so, a switch plus the LED indicator visible on the outside? Probably possible but not cheap.

Better poles that won’t lie down but still won’t damage a sprayer.

RTK precision for recording sensor location and finding them again.

Ensure the cables will not be driven over: in the case of the farm 1, harvest was done with guidence, so it would have been possible to use the field’s A-B lines to ensure the sensors were a long way off them.