Get better air temperature accuracy with this new method

The accuracy of air temperature measurement in microclimate monitoring is crucial because the quality of so many other measurements depend on it. But accurate air temperature is more complicated than it looks, and higher accuracy costs money. Until now.

Most people know if you expose an air temperature sensor to the sun, the resulting radiative heating will introduce large errors. So how can the economical ATMOS 41 all-in-one weather station‘s new, non-radiation-shielded air temperature sensor technology be more accurate than typical radiation-shielded sensors?

We performed a series of tests to see how the ATMOS 41’s air temperature measurement compared to other sensors, and the results were surprising, even to us. Learn the results of our experiments and the new science behind the extraordinary accuracy of the ATMOS 41’s breakthrough air temperature sensor technology.

In this brief 30-minute webinar, find out:

• Why you should care about air temperature accuracy
• Where errors in air temperature measurement originate
• The first principles energy balance equation and why it matters
• Results of experiments comparing shielded sensor accuracy against the ATMOS 41
• The science behind the ATMOS 41 and why its unshielded measurement actually works

Explore why the ATMOS 41 weather station is right for you.

Next steps

• Weather monitoring system comparison: which weather station is right for you?
• Watch another webinar: 7 Weather Station Installation Mistakes to Avoid
• ATMOS 41 FAQs
• Request a ZENTRA Cloud live demo

Questions?

Our scientists have decades of experience helping researchers and growers measure the soil-plant-atmosphere continuum.

• Talk to an expert
• Request a quote

Presenter

Dr. Douglas Cobos is a senior research scientist and the director of environmental research at METER Group, Inc. USA (formerly Decagon Devices). He also holds an adjunct appointment in the Department of Crop and Soil Sciences at Washington State University where he teaches Environmental Biophysics. Dr. Cobos was the lead engineer for the Thermal and Electrical Conductivity Probe (TECP) that was designed by Decagon and flew to Mars aboard NASA’s 2008 Phoenix Scout Lander. His current research is focused on instrumentation development for use in soil and plant research.

Transcript

BRAD NEWBOLD 0:10
Hello everyone and welcome to today’s webinar Stop Hiding Behind a Shield. I’m Brad Newbold with customer research here at METER Group. So today we’ll have 20 minutes for a presentation followed by 10 minutes of Q&A with our presenter, Dr. Doug Cobos whom I’ll introduce in just a moment. But first a couple of housekeeping items. First, we really do want this to be a discussion, so we encourage you to ask any and all questions in the chat pane. And we’ll be keeping track of these for the Q&A session towards the end. Second, if you’re wanting us to go back or repeat something you might have missed no problem. We’re recording the webinar and we’ll send around the recording via email within the next few days. All right, so let’s get started.

BRAD NEWBOLD 0:50
Today we’ll hear from Dr. Doug Cobos, a senior research scientist and director of Environmental Research METER Group who will talk about air temperature accuracy, and present a new method to correct solar heating and temperature errors for radiation exposed sensors. Dr. Cobos has been with METER, formerly Decagon Devices for 14 years, where he currently oversees instrument development for use in soil and plant research. He also holds an adjunct appointment in the Department of Crop and Soil Sciences at Washington State University, where he teaches environmental biophysics. Dr. Cobos was the lead engineer for the thermal and electrical conductivity probe that was designed by Decagon and flew to Mars aboard NASA’s 2008 Phoenix Scout lander. So without further ado, I’ll hand it over to Dr. Cobos to get us started.

DOUG COBOS 1:38
Okay, thanks, Brad. And thanks for that awesome mug shot that I see here in my presentation. Okay, so today we want to spend some time talking about measuring air temperature primarily. And we also want to present a short case study in an experiment we did to try and answer the question you see here of whether or not we can accurately measure air temperature with a sensor that’s exposed to solar radiation. And the reasons behind that question should become apparent as we move forward through the presentation. So why do we care first of all?

DOUG COBOS 2:17
Well, this one’s pretty obvious why we care about air temperature. It is by far the most commonly reported environmental measurement, you’re bombarded with air temperature measurements. Any place you go online, looking at any weather information, or any climate logical information will give you a measurement of air temperature. And the reason that many of us care about that is that it’s a main driver of the human comfort level. Okay, so when I get up in the morning, I check the forecast and see what the predicted air temperature should be that day. And that helps me determine what type of clothes I put on that day. I throw in this slide here that it is a main driver, but it’s not the only driver of human comfort level. And it affects nearly all biophysical processes. And it’s a key component of biological development. So understanding the air temperature is important not only for human comfort, but also for a lot of the science that we engage in. So it’s safe to say that we would like an accurate measurement of air temperature. But I would say that despite, you know, centuries of trying to make this measurement, it is still a lot more difficult to get an accurate measurement of air temperature than most people realize. Okay, so let’s talk about some of those errors and why we might not have an accurate air temperature measurement.

DOUG COBOS 3:48
First reason is pretty clear, if our temperature sensor is not accurate. Okay, if our temperature sensor itself has not been calibrated well, to measure temperature well, then it’s going to be really difficult to measure an air temperature. But calibrating air temperature sensors is relatively easy. Okay, we know how to do that as a science community, we can make a sensor that accurately measures temperature. But the bigger problem with measuring air temperature is getting that temperature sensor to the temperature of the air or the atmosphere. Remember that air temperature is just that it is the temperature of the gas in the atmosphere, just the temperature of the air itself. Nearly all measurements of air temperature are really measurements of the temperature of some sensor that’s immersed in that gas. And if the temperature of that sensor isn’t at exactly the air temperature, then you can get some pretty major errors. And this is what we’re going to focus on today because this is the pieces of the puzzle that causes us to have inaccurate air temperature measurements, the disequilibrium between the temperature of the sensor and the temperature of the air. And you can see that a little bit in the photo on the left where the air temperature there is clearly not 133 degrees Fahrenheit or whatever, that would just be a little bit too hot. This temperature sensor, this thermometer is measuring mostly the temperature of that rock or the temperature of itself as it’s being heated in the sun.

DOUG COBOS 5:31
And so you can see that there are good ways to measure temperature and there are bad ways to measure air temperature. So let’s take a more scientific look at how the different environmental factors affect that air temperature measurement. So let’s see if I can get a okay, there, we got a mouse here, hopefully you guys can see that. Let’s look at this equation on the left hand side. The left hand side of the equation air temperature, T air is what we desire to measure. That equals some measured temperature. This is what we’re actually measuring minus this whole term on the right side, which is our error term. And I want to take a closer look at the components of this. Note that this equation is derived from a simplified energy balance, just a first principles energy balance that takes into account convective heat transfer and radiative heat transfer to and from a thermometer. So let’s look at the different factors here in the numerator. Okay, well, let me make the statement first that this entire error term, for us to get an accurate measurement of air temperature, this term needs to go to zero. Okay, this is our error. So we want to force this thing to zero so that air temperature equals the temperature that we measure. When those are equal, then we’re doing a very good job of measuring an accurate air temperature. So in the numerator, we have an absorbtivity to solar radiation, and a solar radiation term, total radiation, solar radiation that’s incident on our thermometer. Okay, these are in the numerator. So as these increase our error increases. So what does this tell us? Well, we want the air temperature sensor to have a low solar absorptivity. Okay, we want it to absorb not very much radiation, solar radiation. And even more importantly, we don’t want a lot of solar radiation incident on that sensor. So what should we do? Well, we should shade the sensor, of course, and that will drive the numerator towards zero, which drives this error term toward zero. Okay, we have a couple of terms here, this is the specific heat of air, can’t really change that. This is kind of an empirical constant that describes boundary layer heat conductance, and we also lump a little bit of a conductive transfer into this term. So this one’s a bit empirical. This u here is wind speed. Okay, so this wind speed is in the denominator, so we want wind speed to be very high. When wind speed increases, then this whole error term decreases. Well, why is that? Well, the more wind speed you have, the more tightly coupled the temperature of the thermometer is to the actual air temperature. That I think makes intuitive sense as well. The higher the wind speed, the more convective heat transfer you have, and so that thermometer will come closer and closer to the air temperature as the wind speed increases. Okay, this is the characteristic dimension, the d is characteristic dimension, which in the case of most thermometers is just the diameter of that thermometer. Okay, as the characteristic dimension increases, since this is in the denominator, it actually goes into the numerator up here. And so as your characteristic dimension increases, your error will increase. That means, as the size of your thermometer increases, the error will increase. So we want a small temperature sensor. Let’s take a little bit closer look at the main drivers that we have here.

DOUG COBOS 9:17
I grabbed a graph from a Monje and Tanner paper that illustrates these effects pretty nicely. So let’s look at the solar radiation part of this. They plot this in terms of net radiation, this Rn is net radiation in which we could just call that a proxy for solar radiation in this particular case. So what you see here on the y axis, this difference from air temperature, that’s our error, okay, we want this to be zero. This is the error that we’re trying to avoid. On the x axis is thermocouple wire diameters. Let’s ignore that just for a second and focus on the change in error as a result of increasing solar radiation. So you can see that these four lower plots are at a relatively well, a moderate solar load, 400 watts per meter squared net radiation. But as you increase that load to 700 watts per meter squared, you’ll see that this upper plot, okay has a much greater error because of the increase in absorbed radiation. Or we could call that the increase in solar radiation. So it’s pretty clear that as solar radiation goes up, you get much bigger errors in air temperature, or sorry, in the error, yeah, much bigger errors in our air temperature measurement. So let’s look at the effects of wind speed. Okay, here’s wind speed, wind speed is plotted here, this .3, .2, .4, .7, 1.2. So if you look at the lower four, where the net radiation is the same for all of these, you can see that the error clearly decreases as the wind speed increases. So at the higher wind speeds, for instance, 1.2 meters per second, you have a relatively lower error in your air temperature measurement for the reasons that we talked about. And then the other that we want to talk about today is the characteristic dimension or the thermocouple wire diameter as to where this is plotted in the graph. This is your x axis value. So you can see that in all cases, as the wire diameter decreases, so does the error. Okay, so there’s a direct relationship between the diameter of the temperature sensor or the size of the temperature sensor, and the error that you can expect. Note that these data that are shown here are kind of the worst case scenario, these are from thermocouples that are just out in the environment with no shading and no shielding. Okay, but this illustrates some of the things that we can see from the first principle’s energy balance equation that you see on the left.

DOUG COBOS 11:54
Okay, so how would we make a good air temperature measurement? What’s our best configuration? Well, we need to avoid solar radiation, so you need to shade the thermometer. Okay, I hopefully have time for this. I’m gonna go there anyway. This is one of my pet peeves, and I apologize to anybody who’s in the audience that has taken the environmental biophysics course that Colin Campbell and I teach at Washington State University, because they’ve already heard this. One of the things that we rail on when we’re talking about air temperature measurement is you got to make your air temperature measurement in the shade. You can never make a measurement in the sun. It’s just a really bad idea.

DOUG COBOS 12:39
Just an anecdote, we have a grocery store here, Dissmores IGA in downtown Pullman that as many grocery stores do, has a sign out front that advertises daily specials and it also flips through and gives you the time of day. And it gives you the air temperature and in the summertime at certain times of the day sometimes when I’m coming home from work, that air temperature measurement will be off by six or seven degrees C, okay, it will be jacked way up. You’re driving home and it tells you that it’s 109 degrees Fahrenheit, and it’s not 109 degrees Fahrenheit. And it’s a thermometer that sits out directly in the sun, and it has no wind. I am not sure that it’s quite that bad. But anyway, when somebody says oh man, it’s so hot outside, it must be 100 degrees in the shade. Well, my response to that is well you better be measuring it in the shade, and it better be 100 degrees in the shade. If you’re saying that their temperature is 100 degrees, if you’re trying to make that measurement in the sun, then something’s very wrong. Okay, I’ll get off the soapbox and get back to the presentation and say that you better make your measurement in the shade, try and minimize solar radiation. The other ways that you increase the accuracy are to maximize the airflow. So I’ll show you I think on the next slide some radiation shields that do shade the thermometer, but they also maximize the airflow. The best way to do that is to run a fan over it. Okay, pull air across your thermometer to make sure that you have high wind speed and that you the wind speed part of that equation is maximized. You can do that with a fan, but it’s really power hungry. The other way to do that is to use some louvered radiation shields that look a bit like a pagoda but allow wind to flow freely through but still shade that thermometer. Those are not quite as effective as the aspirated shield but certainly don’t require any power. And then finally you want to minimize the size of your temperature sensor. You want to minimize that characteristic dimension. So your best temperature sensor is going to be a tiny sensor in a windfield that has wind passing across it in the shade, and that’s going to give you your best measurement of air temperature.

DOUG COBOS 15:06
And that’s exactly what these aspirated temperature shields do. So on aspirated shield, this is kind of the standard or the reference method for measuring air temperature. You would have a small temperature sensor, this is an apogee shield TS 100 that we used. In this experiment it’s our reference, very small temperature sensor that sits in a aspirated path. Okay, here’s a fan that’s pulling air across the temperature sensors and maximizing the wind speed. And also, there’s obviously shading here to keep that temperature sensor out of the sun. And this is about as good as you can do for an air temperature measurement. So many different manufacturers make these with different flow pads that have been optimized to get maximum wind speed and also minimize the radiation load on the sensor. But the downside of these aspirated shields is that you have to run a fan. And if you’re dealing with a system that doesn’t have mainline power, that means you have to have bigger solar panels. You have to have more batteries to be able to run those fans. And even a little computer fan, it draws quite a bit more power than probably the rest of the entire environmental monitoring system.

DOUG COBOS 16:25
So many systems, many air temperature measurements use passive radiation shields like the ones that you see here with the louver design that allows wind to flow through freely, but relies on natural wind to aspirate or blow a little bit of wind across the thermometer. But these do keep the solar radiation off. Less, much much less power consumption, probably a more common way to make air temperature measurements. But I borrowed some data here from Mark Blonquist and Bruce Bugbee from a book chapter that’s coming out in a ASA publication here pretty shortly that shows the average error in daytime temperature, air temperature measurements from three passive radiation shields and you can see that once you get down below about one meter per second wind speed, you can get some pretty substantial inaccuracies in your air temperature measurements. So there are trade offs between these passive shields and those actively aspirated shields. Now, what I want to talk briefly about today is the air temperature measurement that we make on our new ATMOS 41.

DOUG COBOS 17:43
So ATMOS 41, if you’re not familiar with it is a multiparameter compact weather station that measures precipitation, solar radiation, wind speed, direction, air temperature, vapor pressure, humidity, atmospheric pressure, and some lightning parameters. So it’s basically an all in one weather station that is made to be super easily deployable and makes scientific grade measurements, but without the price tag of some of the high end met stations that are you know, five to $15,000. So why do we care about the air temperature measurement here? Well, obviously we need an accurate air temperature measurement. But look, our temperature sensor, you can see it here is exposed to some solar radiation. Okay, big nono, I’ve just been railing on how in the heck could you possibly think, even imagine making air temperature measurements in the sun or where you have some exposure to solar radiation? Note that this is shaded from high sun angles, but it can get some reflected radiation and some radiation at low sun angles. So our question was, well, we’re going to try and make this air temperature measurement where we know we’re gonna get some exposure to radiation, but we’re using a really small temperature sensor, okay. And the ATMOS 41 measures solar radiation and wind speed. So, if we know the solar radiation, we know the radiation load and we know the wind speed. Can we correct this measurement of air temperature and get a good accurate measurement? So that is the experiment, the case study that I want to talk about today with the question of whether we can use wind speed and solar radiation to correct that air temperature measurement on this ATMOS 41.

DOUG COBOS 19:26
So in our experiment we used that Apogee TS-100 aspirated shield as our air temperature reference to give us our real measurement of air temperature. And then we compare that air temperature measurement from the ATMOS 41, and just a standard passive radiation shield to that reference air temperature measurement from the Apogee TS-100. Collected air temperature data over a variety of environmental conditions, cloudy to sunny, and then what we do is we optimize that energy balance equation so we use that K value and the solar absorptivity value is fitting parameters to force the experimental data to fit the actual air temperature data from the TS-100. And once we optimize those parameters, then we validated our correction in our model with data from multiple units, so let me show you the data we use for our model optimization.

DOUG COBOS 20:27
These were from last September. And what you’ll see here is, excuse me, on the y axis, you have the temperature, air temperature error. So this is the difference between the apogee measurement and the other measurements that we made. And on the y axis is just time. So these are time series data. And if you look at the blue plot, this is the uncorrected data from the ATMOS 41. And you can see during the daytime, we get some overestimation. And it’s fairly substantial up to about a degree, but not quite as bad as we had expected. We’re actually fairly pleased with this. And you’ll note that the orange plot here is the error from an unaspirated temperature sensor. And it’s in a passive shield, one of those passive radiation shields, and you’ll see that the error there is actually pretty comparable, the ATMOS 41 data before they’re corrected. But once we corrected the ATMOS 41, this is the gray plot, you’ll see that it comes into quite a bit better agreement with the actual air temperature. One thing to note is that we had to limit the upper end of the solar radiation that we used to correct the data because we were overcorrecting under very high radiation conditions. And so we had to impose a limit at the top end. And that resulted in a much better data fit. So you’ll notice here that uncorrected ATMOS 41 had a 95% confidence interval of .6 degrees. The non aspirated had an even worse accuracy or confidence interval. And once we corrected the ATMOS 41 data, then that dropped down to about .4 degrees, which is a pretty respectable number, considering that the non aspirated shield is quite a bit higher than that. So this summer, and in fact, these data are just from a couple of weeks ago, we validated our model with with seven ATMOS 41 units instead of a single unit.

DOUG COBOS 22:41
And the black trace you’ll see here is the air temperature from the TS-100. And then all of the other colors, those are the temperature measurements from the ATMOS 41 units after being corrected. And from a plot like this, everything looks super good. Okay, we see quite small errors, but it becomes a lot more apparent when you actually plot the error instead of the magnitude of air temperature. And so what we see here is that we’re nearly always within one degree C of the actual air temperature, all seven units are. And if you look at the bias, we have some with a slight positive bias, some with a slight negative bias, some with no bias. So we’re splitting the difference there. That’s very good. And the 95% confidence interval for each of these units is about .6 or less. And so we’re feeling pretty confident that our measurement sorry, our model and our correction is effective when used across a population of sensors instead of just a single sensor. So our conclusions or the take home point is that we are pretty confident that our first principles energy balance based air temperature correction is effective. It’s a good job of correcting the air temperature and giving a good accurate air temperature measurement even though we have some radiation exposure on our sensor. Our biases are low and we have a pretty tight accuracy spec that we can write there. And one of the things that was a little bit surprising is that the corrected results in air temperature are probably about on par with passive shields and maybe a little bit better than air temperature measurements that you can get from a passive radiation shield and that is pretty encouraging to us.

DOUG COBOS 24:35
One last thing that we wanted to do is give a shout out to our friends at TAHMO. For those of you who aren’t familiar with TAHMO, TAHMO is a Trans Africa Hydrometeorological Observatory. This is a nonprofit group that is trying to really improve the lives of the people in Africa and specifically the African farmer and so If you’re interested in learning more about that, head to the website you see there metergroup.com/tahmo. These guys have been development partners on the ATMOS 41 with us for for several years now and are doing really great things in Africa. And if you wish to support their cause there is an app for your iPhone or Android, where you can get a little exercise, improve your life. And for any little bit of exercise you do, you actually are donating money from METER to TAHMO to help fund the things that they’re doing in Africa. So if you’re interested in that, then feel free to head to the website. Okay. With that, I will take questions. Let’s see if I can figure out how to get the maybe it’s in the chat here. Okay, so if you guys want to write through some questions, please do now. I see them up here. Also, it sounds like there will be a pop up once the presentation is over that will ask you a few poll questions. I’m keenly interested to know how long you would like these virtual seminars to be ran a little bit longer than 20 minutes that we allotted today, but interested to see if people would like these to be longer or shorter. Okay, let me see if I can grab some of these questions.

DOUG COBOS 26:38
Okay, yeah, so this is a good question. It’s asking, what happens to your air temperature measurement if your wind speed or solar radiation sensor fails? Okay, so for instance, what if a bird poops on your pyranometer and you’re not getting good solar radiation data? Or what if one of the ultrasonic transducers goes out? And your ridiculous wind speed measurements? So the answer to that is if the wind speed measurements are compromised, then we ignore that correction factor, and it just reverts back to the uncorrected values which are a little bit worse, but still pretty comparable to a passive radiation shield. If the radiation sensor fails, if it is obscured for some reason, then that’s going to put the correction off a little bit. Okay, so it’s going to underestimate the solar radiation. So you’re going to get a little bit artificially high air temperature measurements, but no worse than leaving the correction off. And we were actually pretty pleased with what was going on with the uncorrected air temperature measurements. Okay. Question here is latent heat of vaporization and the correction. This is from somebody who’s clearly familiar with the full energy balance. And the answer is no, we neglect latent heat of vaporization, because we don’t expect that sensor to be wet. If the sensor were wet, and water were evaporating from it, then we would have to account for that heat exchange, but because the sensor is not going to be wet, we’re not accounting for that.

DOUG COBOS 28:32
That’s a good question. Do you have any experience in using the chimney effect in order to create natural airflow with the passive radiation shield? No, we do not. My experience with the chimney effect has primarily to do with the toilets that our US Forest Service puts in the national forest that they have a big black chimney that pulls air from the stinky stuff and out a chimney. It’s black. And so the air in that rises. And so it keeps that stinky air from flowing back into the area where people might inhabit, but never used that for the air temperature measurement. But it is something that would be intriguing to try and you know, to try and aspirate a air temperature sensor passively under high solar load. So there’s this possibility that that could be a good method.

DOUG COBOS 29:34
Another question here says your energy balance neglects long wave radiation. Is that a problem? That is also a good question. And long wave radiation is a part of the energy balance that we have to take into account. We have a couple of considerations that help us there. One of those is that we have the really small temperature sensor and that temperature sensor is polished stainless steel, which has a very, very low long wave absorptivity. So, that means that it is pretty well decoupled from the long wave energy that’s surrounding it. The other thing that works in our favor is that most of the errors you get from long wave radiation have to do with clear night sky having a very low temperature which can cause you to underestimate air temperature by quite a lot under those conditions. But our sensor remember is under the bottom side of the housing that holds the ultrasonic sensors for the anemometer and does not have any view factor to the sky. So the errors that we see from long wave radiation tend to be very, very small. Okay, another question here asks, can you use solar declination in radiation compensation? Yeah, that would be a really good idea, in fact, because we know that we’re going to be exposed to solar radiation more at low sun angles. If that ATMOS 41 had a GPS built into it, we could do some of that. And because then we’d have time and we’d have location, but it doesn’t have a GPS. The loggers that we make now, the ZENTRA series loggers do have a GPS and so they know time and location. And if we did some two way communication, we can probably do a little bit of correction on that. So that is a good idea. Okay, so we’re out of time, a little bit past time. I was supposed to go till 8:30 but went till 8:33. I apologize for wasting three minutes of you guys’ day. Thanks so much for coming. And please, if you can take just a second to respond to the survey questions at the end, that would be very helpful to us. Thanks.