I often get the question what microclimate sensor I would recommend. To facilitate my answer to that, I decided to summarize my ideas on the matter here. Note that this is a far from exhaustive musing on the different factors to take into account. And most importantly: I would love to hear from YOU what worked and what didn’t, and would gladly update this information here with your insights!

In SoilTemp, we are not using one standard sensor type, we are accepting data from all different kinds of sensors, as long as they measure with 4 hour intervals or less. Most common brands are TOMST, iButton, HOBO and Lascar, yet there is a myriad of other sensors, for example those integrated in weather stations, flux towers, or other long-term monitoring. My personal recommendation is still the TOMST TMS4, a sensor designed by and for ecologists that solves many of the issues one could have.

Some features:

• Around 10 year memory space. This is the game changer for me! The TMS4 measures when it leaves the factory and never stops. It doesn’t overwrite data and simply keeps it all. You can always read out data from only your last measurement campaign or the whole thing. So no data losses as is ever so common with other sensors! Also, they easily measure every 15 minutes, whereas the cheap iButtons and HOBOs wouldn’t survive a whole year with their memory at ~60 minutes, even.
• 4 sensors for the price of 1 (or 2-ish 🙂 ): the TMS4 mimics a little plant, measures temperature just above, at and just below the soil surface, and also measures soil moisture. This has proven véry useful, as the three temperature levels show highly different patterns and drivers in virtually all ecosystems (e.g. related to snow cover).
• Around 10 year battery life, so no issues of sudden breakdown (without option to retrieve data) as with iButtons, and no struggling with replacing batteries as in HOBO
• Standardization: no messing with home-made shields or installation methods, making measurements much more comparable between different studies.
• Free software! iButton software is free as well, but I don’t like that I have to pay for HOBO software.
• Rugged. The TMS4 is fully waterproof (no wrapping in Parafilm as with iButtons or fumbling with dessicant as in Hobo Pendant). The solar shields are the most vulnerable; TOMST provides rings now to secure it to the sensor, but the result is now that instead of disappearing they tend to break under duress). We still have some sensor deaths despite their ruggedness, which mostly relates to them sticking above the soil (e.g. mown down or burned).
• The sensors unique number is written super clearly on the sensor, so messing up sensors and plots has become much less common. Perhaps more importantly even: the output file has a standardized naming that includes the sensor name, so you really have to put in effort to screw this up :).
• Easy to handle data format: all datafiles are (mostly) exactly the same, so less of a mess than I’ve had with HOBOs but especially iButtons to get the data processed.
• They look good. Admit it, it’s much more attractive to have this little mushroom in your plot than it is to have a buried soil temperature sensor.

Some drawbacks, unavoidably:

• Accuracy: as with most cheap sensors, the accuracy is not super high; I usually work with trusting it at 0.5°C. However, especially aboveground and on sunny days, there could be MASSIVE errors on the temperature readings, as the sensor heats up (see Maclean et al. 2021). As with all cheap sensors, extreme caution is thus warranted for interpretation of air temperature measurements. Soil temperatures shouldn’t have that issue. If you want to avoid this, you should work with small thermocouples, which don’t heat up (much) in the sun as they are so small. These are usually substantially more expensive, however. I heard of development of such a sensor within a similar price range, but haven’t seen it commercialized yet.
  The sensors come with optional solar shields, yet these don’t solve the issue (except that they are at least standardized, which is a great jump forward compared with homemade shielding). I especially am not that big of a fan of the lowest shield and prefer installing the sensor sufficiently deep so the middle-temperature sensor is just below the soil surface and as such avoids solar radiation. Here again, however, there is a risk of comparability with other studies, so it is tricky. Note also that having a different installation height of only a centimeter or so can dramatically change the readings on that surface sensor, as so much variation is happening close to the soil surface.
• I’ve gotten repeated messages about read-out issues. Most of these seem to be software-related and TOMST is working very hard at solving those, but you have to stay on top of the software updates and if your sensor doesn’t respond, there is only so much you can do yourself to find the issue. TOMST has been very helpful in solving issues, though, and I have the feeling things have improved (we had a bunch of sensors in Congo that didn’t read out over there, but I managed reading out all of them back home).
• This brings me to failure rates: I have the feeling the TMS4 is doing better than other sensors regarding percentage of erroneous measurements, but there is still a non-trivial amount of sensors that still provides erroneous data (e.g. all readings at -100°C, or sudden outliers of +80°C).
• As for all sensors, you have to buy a reader as well. The TMD-adapter is the price of one sensor (so not better of worse than for the others, but you shouldn’t forget about it and it adds up if you can only buy a few sensors).
• Visibility: the TMS4 sticks out of the soil. This makes it prone to damage by animals (e.g. wild boars are horrible). This can be solved by installing a metal wire frame around the sensor. But it also invites vandals. The latter in its turn can be solved by installing it out of sight of most humans or putting information signs next to it. In extreme cases (e.g., roadsides or agricultural fields), you could install the sensor fully belowground (TOMST even has shorter ‘dwarf’ versions of the sensor for this goal), but be warned that you then loose the important feature of comparability with other studies. They also have a (cheaper also!) smaller version called the ‘Thermologger’ which only measures one temperature. In theory, I think you can also install this one belowground and use it as you would a HOBO temperature logger, but I haven’t tried this. TOMST affirmed me, however, that it is sufficiently waterproof for this.
• The soil moisture measurement is not super accurate. This is a whole big story on its own, but it boils down that there are different levels of uncertainty piling up onto each other (sensor-based, calibration, installation, soil heterogeneity, weather…) which in the end result in the fact that a measurement of 20% could be anything between 10% and 30%. However (!) we do get good result when looking at patterns across hundreds of sensors, so the noise seems mostly random, and we feel we can trust relative measurements (e.g. the slope of the curve when drying out, or the difference in moisture generated by a rain event). So, care is needed, but data still has many good uses. And again, very importantly, this issue is there for virtually all cheap soil moisture sensors.
• Sending sensors across the globe: I’ve had DHL speak up against sending more than two sensors at once due to the Lithium content of the batteries. This made shipping very difficult and/or expensive. As the batteries cannot be removed, there wasn’t much I could do about that. I now however found a small local company who claims that DHL is misguided on the matter and that it’s actually ‘not worse than shipping a bunch of laptops’, so my shipping has resumed without much extra cost. I’ve heard of most collaborators that they simply don’t declare the Lithium when shipping, which solves the issue of course as well.
• TOMST is a small company and the sensor has become very popular. Combine that with the global shortage in electronic equipment and the end result is that delivery could not always be guaranteed. I have the feeling that these issues (e.g. the shortage in equipment) has cleared up now and the company is on top of orders again. Also, I’ve had plenty of delivery issues with iButtons as well, despite the size of the Maxim Integrated company, so these hiccups simply exist.

Some alternatives:

• If you really only need one temperature, you can go cheaper than the TMS4. TOMST has the thermologger for that, as mentioned above, which could be of great help! But things like HOBO or iButton can go even cheaper. For example, I recently heard about iButtons (DS1921G) at a little bit over €20 which is of course unrivaled [UPDATE: looks like they sold out, which makes sense! Now they are much closer to TMS-prices]. However, failure rate is significantly higher there so you should take that into account, and you’d have to boost the measurement interval (we often used 4 hour intervals) to limit the amounts of time you have to go read them out.
• If you want to measure air humidity, there are some good HOBO loggers out there that can do that job for you. They have proven robust to colleagues in the tundra, yet are significantly more expensive than alternatives from iButton and Kestrel. Rumours have been positive and negative about all options, so I can’t make any hard recommendations. In any case, be careful with relative humidity values close to 100%, as these are rarely accurate (and can show things like ‘drift’ in the data).
• If you want to measure at >2 soil depths, you could obviously simply bury your TMS sensor (3 depths) or one of the dwarf versions. However, this limits you to the specific sensor depths of those. I haven’t tried anything else, yet I asked Twitter and got some good responses. I especially like the flexibility of the UGT, which can handle a temperature profile of up to 8 depths. However, I know nothing about the price.
• The ultimate dream is of course remote data transmission using the Internet of Things. This allows you to track sensors on a daily basis, see when there are failures, dramatically reduces data loss due to sensor malfunctions, ánd ensures you are much more on top of microweather events, which in my opinion greatly improves mechanistic understanding of what’s happening. We have been working with TOMST to develop the TMS-NB for our citizen science project www.curieuzeneuzen.be, and these are truly game-changers to me. They are not commercialized yet as it is tricky to get them to work in other countries, and your price adds up (extra hardware, SIM-card, roaming costs, database connectivity…). We have a few secondhand available, though. We have also teamed up with the group of Pieter De Frenne (UGhent) on commercializing their modular (i.e., add whatever microclimate sensor you want) IoT-connected forest microclimate sensor MIRRA. NOT available yet for real usage, though!
• If you want higher accuracy, especially above ground, or higher temporal resolution, you should consider thermocouples. I have little expertise there myself, however. Ilya Maclean (University of Exeter) is developing ‘the ultimate near-surface air temperature sensor’ using thermocouples and is close to commercializing this, but it’s not out in the open yet.