Weather Data: Why Accuracy is More Complicated Than You Think

Weather Data: Why Accuracy is More Complicated Than You Think

Don’t unwittingly compromise your weather data by underestimating all the factors that influence accuracy. Dr. Colin Campbell discusses what these factors are and how to plan for them.

You need data you can trust

Think weather data accuracy is about sensor specifications? Think again. There are a host of other factors that influence accuracy, and if you don’t understand what they are, your data can steer you in the wrong direction and put your projects at risk.

What you need to know

In this 30-minute webinar, Dr. Colin Campbell explains how you can unwittingly compromise your data by underestimating these important factors. Learn:

  • How microclimates influence accuracy
  • How many measurement sites you need to deal with variability
  • How installation affects accuracy and important best practices to keep in mind
  • Why you need to measure more than just weather parameters to understand what’s happening at your site (critical ancillary measurements)
  • Why the scientific theory behind how a station makes its measurements matters
  • Why models using internet data are not good enough
  • How a station that requires significant maintenance can derail accuracy
  • How using affordable research-grade stations to fill in data gaps between premium-quality setups can be a cost effective way to increase your accuracy
  • Why your data visualization and management system matters in terms of accuracy
  • Case studies that show why you need to think about the big picture

Dr. Colin Campbell has been a research scientist at METER for 20 years following his Ph.D. at Texas A&M University in Soil Physics. He is currently serving as Vice President of METER Environment. He is also adjunct faculty with the Dept. of Crop and Soil Sciences at Washington State University where he co-teaches Environmental Biophysics, a class he took over from his father, Gaylon, nearly 20 years ago. Dr. Campbell’s early research focused on field-scale measurements of CO2 and water vapor flux but has shifted toward moisture and heat flow instrumentation for the soil-plant-atmosphere continuum.


See All Webinars

Weather Data: Virtual, In-Field, or Regional Network—Does It Matter?

Find out how different weather data sources compare and how those data affect the accuracy of common environmental models used by growers.


Weather monitoring 101– Which weather station is right for you?

Dr. Doug Cobos explores how to choose which system is right for you and the sweet spot for price vs. maintenance vs. accuracy in your unique application.


7 weather station installation mistakes to avoid

Dr. Doug Cobos discusses insider tips and best practices that will improve your accuracy and save you time, effort, and future headaches.


Case studies, webinars, and articles you’ll love

Receive the latest content on a regular basis.


Hello, everyone, and welcome to Weather Data: Why Accuracy Is More Complicated Than You Think. Today’s presentation will be about 30 minutes, followed by about 10 minutes of Q&A with our presenter, Dr. Colin Campbell, whom I’ll introduce in just a moment. But before we start, we’ve got a couple of housekeeping items. First, we want this webinar to be interactive, so we encourage you to submit any and all questions in the Questions pane. And we’ll be keeping track of these for the Q&A session toward the end. Second, if you want us to go back or repeat something you missed, don’t worry. We will be sending around a recording of the webinar via email within the next three to five business days. All right, with all of that out of the way, let’s get started. Today we’ll hear from Dr. Colin Campbell, who will discuss common barriers to weather data accuracy. Dr. Campbell has been a research scientist at METER for over 20 years following his PhD at Texas A&M University in soil physics. He is currently serving as Vice President of METER Environment. And he’s also adjunct faculty with the department Department of Crop and Soil Sciences at Washington State University, where he co teaches environmental biophysics, a class he took over from his father Gaylon over 20 years ago. Dr. Campbell’s early research focused on field scale measurements of CO2 and water vapor flux, but has shifted toward moisture and heat flow instrumentation for the soil plant atmosphere continuum. So without further ado, I’ll hand it over to Colin to get us started.

Yeah, thank you, Brad. It’s a pleasure to be with you all today and talk about something that I’ve been thinking about for several years now as we developed our ATMOS line of instrumentation. The title we gave it was “Why accuracy is more complicated than you think.” And I may be the poster child in some ways for the loss of accuracy in experimentation over the years. And so maybe this is more like a little fireside chat that I tell you all the things I did wrong, and hopefully they’ll benefit you going into the future. I’m gonna get started just but with one of those stories. I titled this slide “Bird takes down PhD project.” As Brad mentioned, I got my PhD at Texas A&M University. And what I did was study the biomass accumulation in rice. And I did this with a eddy flux tower that you can see there and the right hand slide or the right hand picture on the slide. Back then this system was pretty complicated. It was right at the end of the 1990s, and it was just before they’d released the all-in-one e ddy covariance systems that are now pretty ubiquitous around the world. We had to develop the system all by ourselves. And it took a lot of work and a lot of complicated effort. This picture is actually taken from our meteorological tower that’s out at this research site. And on that tower, we took all the kind of non eddy flux measurements that we were making, including something pretty simple, solar radiation, or at least I thought it was simple. We used to go out to this project every week. We’d walk along the levee that’s there kind of in light color in the background near the solar panels. As I’d walk along that levee, I’d look over at our instrument tower, and there’d be this pretty little Redwing Blackbird sitting there, singing its song, and I kind of enjoy the intrusion of nature into this project that we were working on. Lo and behold, as I started analyze the second year’s worth of data, that Redwing Blackbird ended up having a pretty important impact on the project. It was sitting around, as I noticed, but didn’t pay particular attention, our solar radiation sensor, our pyranometer. And what I found when I finished the research was that the very presence of that bird and some of the bird droppings it left impacted our solar radiation measurements such that when we got around to trying to produce some of our graphs that needed to be put in my dissertation, we found that the solar radiation data that we collected weren’t any good in the second year of the project. Now, luckily, we spent a lot of time making sure we had redundant measurements out there and I had some measurements of photosynthetically active radiation, which we did ended up using a conversion to get back into total solar radiation. So we could produce the graphs I’m showing on the right, but at the end of the day, we were pretty lucky because our goal in one of the chapters of my dissertation was to put together biomass generated from carbon exchange rate compared to total intercepted solar radiation. And without that backup, I would have been pretty lost. And it turns out that on the projects we participate in, we’ve got to be pretty careful about the measurements we’re making. Now, I borrowed this slide from one of my colleagues here, Doug Cobos, who talked about weather stations a couple of years back. And he nicely put in a concept here that we work, kind of a working concept we use here. On the right, I have a little equation, it’s not that exciting, value is equal to performance divided by price. And we kind of have that as a working concept as we develop instrumentation here, meaning that we’re trying to optimize this ratio. We want to increase the performance, and as best we can reduce the price. And if we think about weather stations in this way, we have weather stations on the market that run really the gamut from the very bottom of the performance price ratio all the way to the very top. In fact, I was talking to someone in our marketing group the other day, and I said, Hey, how many times are weather stations searched per month on Google, and they came back with 27,000. And as far as instrumentation or instrumentation suites that I’ve seen, and I’ve seen a few of these numbers, that is way, way higher than is typical. So we know there’s a lot of interest out there. And when we think about what makes a station right for a certain application, there’s a lot of things we think about, things like robustness, accuracy, installation, and maintenance, the measurements we actually want to use, the data acquisition, whether or not it’s capable for running all year long, and what are its power requirements. Now, you might not think, Hey, this is all rolled up into this question of accuracy. But I would argue that it really is. And as I was thinking about preparing for the seminar, I had some pretty big vision for all the things we’d be able to cover today. And as I looked at it, all the things that I want to do, I may not get to everything that was on the invitation email, but I really want to talk through some of these things on this list, because they do impact accuracy, maybe in a way you haven’t thought about. So when we think about various weather stations, we go all the way from a hobby weather station, something that that you might just put outside your house, just so that when you get up in the morning, and you kind of wander around the house, you might see hey, what’s going on right here, how’s the wind speed, what’s the direction today, maybe you’re going out for a run or a bike ride and you’re interested in that, you might want to know it, did it rain, should I go and turn off my irrigation system, something like that—ranging all the way up to Aviation Weather, and WMO weather. I actually fly in a personal airplane a lot with my brother who’s a pilot, and it turns out that we depend a lot on these kinds of instrumentation to make sure of exactly the conditions we’re going into at the airports were landing in. I hadn’t thought about it a ton when I was just using commercial aircraft. But when you start running around in private aircraft, you want to make sure that these weather statements are correct, are absolutely true. Sometimes end up kind of in the middle of all this, these are what I kind of think of as the Mesonet area. We work with a couple of these, at least one in the AgWeatherNet here in Washington State. There’s just a picture there on the left hand side of all their stations that they use across state. We also work with Montana Mesonet. There’s actually just one of their sites and some of the data coming out there. So it’s kind of a mid level requirement for performance there. And we’ll talk about that in a little while. But when I think of this overall, I think of trying to balance. And one of my biggest goals in life is really finding a perfect balance in things, and in weather station, it ends up being that ratio, that performance divided by price. And so I had a drawing made just to you know, this is what goes on in my mind is when I think about weather station, is there a Goldilocks kind of version of the weather station, what’s just right for you?

So I want to talk now through a lot of these accuracy pitfalls. I’ll tell some stories. I’ll give you some examples. But mostly what I’m trying to do here is just get you some take home, some things that you may be able to jot down as we talk that you haven’t thought of potentially that are going to be important as you think, what is my next weather station or what is my next set array of weather stations? One thing that that I don’t really like to come up but it’s really important is this idea of maintenance, labor and accuracy. Those are tied together and really important. But when I deploy instrumentation, as I’ll mention in a moment, I wish we could just set it and forget it. So we’ll talk about the impacts there. One of the things I’ve been looking at over time recently is just microclimates and the variability in weather in the situation of these microclimates. And so I’m going to talk a little bit about some measurements we made there. Then I’m gonna go through and talk about just unexpected impacts on sensor readings. And finally finish up with this idea of connectivity and visualization. I’m going to try to make a case for you to think about that, as it connects to accuracy. Does connectivity and visualization have anything to do with accuracy? Well, maybe not absolutely directly, but it does, I think, really impact some of those things. And I’ll try to make a case for that at the end. So this idea of leave it and forget it. So METER Group has built instrumentation for many, many years. And one of our goals that we’ve focused on, especially in the last few years, is really making those things last a long time in the field. But I want to reiterate that longevity doesn’t necessarily imply that the calibrations of these instruments are going to be stable over time. Now, sometimes it does. I work a lot in water content sensors, and we’ve tested those over the years and the stability of the calibration is quite good. I wouldn’t worry about that. So if you’re suddenly wondering, oh, no, is he suggesting I go dig up all my water content sensors and send them in and calibration? No, not suggesting that. Likewise, temperature on an ATMOS 41 or other, I mean this is where I’ve tested other weather stations potentially. These don’t seem to drift. I’ll give you some data that suggests that’s true. Now, others should be on your concern list. These I’m listing here things like barometric pressure, solar radiation and relative humidity. They, in the manual, for example, that METER hands out, there is kind of an expected one to two recalibration cycle that you should be on. If you’re anything like me, you kind of do this scenario. All right, title, “What not to do.” It doesn’t rain a lot here in eastern Washington state where METER Group is located. And even when it does, it’s pretty light. And we were trying to think, you know, we want to test this ATMOS 41 when we were developing it in some pretty harsh and rigorous locations. And so one of the places we picked was near College Station, Texas, where Texas A&M is located. We have friends down there that could help us with finding a site to put these stations up. And so we deployed all these weather stations down there. Now, the good news was that we enjoyed lots of rainfall there. Rain doesn’t fall in Texas like it falls in Washington. And I discovered that pretty quickly. Growing up, I wondered why you’d ever have an umbrella in eastern Washington. We just didn’t need it. When in the first rainstorm in Texas, I suddenly discovered having an umbrella in a backpack is pretty critical, if you want to be anything less than absolutely drenched, should a rainstorm move through rather quickly. That was really good news for ATMOS 41s because when we got these heavy rainstorms, we could test the rain, the drop counting rain gauge that we have on there, which I’ll talk about a little later in the presentation. So we benchmarked all that performance. And then essentially we forgot about these ATMOS 41s that were down there because the project had finished and we didn’t actually get them back. So it provided some interesting data recently, when we started deploying now the ATMOS 41W, our new wireless weather station that we wanted to figure out how it performed in the field compared to the ATMOS 41. And so I gathered some of this data for this presentation to show you what things change over time and what don’t. So here’s air temperature on the vertical axis, on the left hand vertical axis and it’s just days, time, on the bottom axis. So basically here just in the month of August, in our same month now, earlier in the month, we’re comparing on the left hand side of that dotted red line temperature of several units before we made a recalibration change, and then to the right of that red dotted line, what happened afterwards. So we can have some confidence that the temperature really doesn’t change much with time. We don’t have to worry about recalibration then. But when we start to look at other sensors, we see a little bit of drift. On the left hand graph, we have relative humidity on the vertical axis and again time on the horizontal, with that red dotted line being the difference between the pre and post cal. There is not a ton of error there. If we graphed it in absolute error, we’d see a little bit, especially with that orange yellowey line. After a recal, things look quite good. Those match up quite well. On the right hand side, we can see just a little bit more drift. This is now an atmospheric pressure sensor. One of those has drifted quite a bit compared to the others. Now we do here blow up the graph. So we’re only talking about a couple of tenths of kilopascals there, but you can see after we put in recaled versions of those, how consistent they are. Now the next one we’re going to look at is solar radiation, pre and post calibration. Now we have that graph with time on the right hand side now, solar radiation and watts per meter squared on the vertical axis, and again, the same time period on the x axis. So what we see here is quite a bit more variation, as we’ve left those sensors out for a few years down there. And after we replaced those sensors, how consistent the values are. On the left hand side and that graph, we’re just comparing to ATMOS 41s pre and post calibration, see a much more consistent calibration on the one to one line after calibration. So the take home message here is really, Look, do your recalibrations. If the manual suggests that you need to recal every year or every two years, don’t skip it because it’s really nice to leave your instrumentation the field. Go out and take care of it.

So another thing that that can hurt our accuracy is simply not measuring the location we’re interested in. Because weather stations tend to be fairly expensive, we often either want to use just a local weather station—I don’t know how many times I’ve thought, oh, I’ll just use the weather at the airport, here locally, to tell me something about data from a field that’s a few kilometers away. This can be a challenge if we are dealing with microclimates. And just an example on the bottom, it’s a picture, just a airplane picture of METER Group, and actually has that kind of horizontal line. That goes down to a little valley that’s near us, probably have than 100 meters lower. But we have a series of ATMOS 41s on METER Group’s roof there on the right hand side, that dot on the top roof. And down in the bottom, we have some ATMOS 41s as well testing that kind of more native environment down there. What we can see after we analyze the data was a six degree Celsius difference with about 220 meters of horizontal distance. And you wouldn’t necessarily think that’s happening. But if you are interested in pest management, disease modeling, plant growth models, even human comfort, this actually becomes a big deal because you are going to be using a temperature that’s not associated with the location you’re at. So this concept— and I’d like to unpack this more in another virtual seminar. We’ve got more data on this, and it’s certainly something I’ve been thinking quite a bit about. I’d like to unpack this a little bit more to say how many weather stations do we need per distance to really quantify an area adequately.

Next I’d like to step on and talk about solar radiation. Now here, I’ve graphed a sports field that I work on. If you’ve seen any of my other seminars, you might have seen this before. On the sports field, we’re not interested in any specific weather data, the temperature, the solar radiation, the wind speed, we’re actually interested in combining those things together to give an estimate of evapotranspiration. And so on the y axis there, the vertical axis, we have reference evapotranspiration in millimeters. And we have actually over about a month period, we have the ET, going from about an average of five millimeters, just a little bit lower to in the upper four millimeter range. Why am I showing you this while we’re talking about weather stations? Well, solar radiation is the driver of the energy balance. It’s the only real energy input in the system while we have sent several energy outputs in latent heat flux or evapo transpiration, sensible heat flux, soil heat flux, all of those are kind of sinks of this input. Accurate evapotranspiration requires that we measure things like solar radiation to get a value that we can use. Now, are there equations out there that don’t use solar radiation and still give ET? There are, but their accuracy is often questioned.

Now, surprisingly, when you look at all the weather stations out there, actually very few of them come with solar radiation sensors. So my suggestion as you look to either build out or purchase a weather station is, check before you buy, make sure the instrumentation that you’re going to use has a solar radiation sensor if you’re going to need things like evapotranspiration. And then be sure to check the accuracy that you’re gonna get out of that system. Because it turns out that soil radiation sensors are pretty hard to build. We actually integrate one from from our sister company Apogee. And we love them because their accuracy is really solid. The other thing you need to do is verify their performance in the field, where your sensors matter. So I threw on this picture just to, again, smile at some of the things we do at METER Group. Many years ago, we put together a promotional catalog and our intrepid marketing photographer went out in the field to take some pictures of the beautiful EM50, our old data logger and rain gauge temp RH sensor and our solar radiation sensor. I’ve circled that in red. Now, you probably already realized, but it took us a little time to notice that the way we’d set it up, which was not in a working system, had the solar radiation sensor below the data logger and below a significant amount of the post that you see there. This is not great. Don’t set up your your solar radiation sensor below the level of your data logger because it’s very likely going to get shaded and cause you similar pain to what I experienced with my Redwing Blackbird. Our suggestion would be to put it up on top of the post, one of the top of the stack, and make sure that if you’re in the northern hemisphere, for example, put it facing south so that the sun can hit that all day. So just watch where you’re putting these things.

I want to jump on to errors in air temperature measurement. This is one of the most interesting to talk about. And it’s also one of the most frequently done badly. All measurements of air temperature are really the measurement of the temperature of the sensor and not the air. When there’s a big difference between sensor temperature and air temperature, that results in big problems, major errors. And I like this picture on the right hand side just to show that if you put one of these, kind of a home temperature sensor, sitting out on the concrete or something like that, on the rock, we can get some pretty high temperatures that are not representative of the actual air temperature. Now just because I enjoy talking about environmental biophysics, it’s one of my favorite things, I needed to throw up just one equation here for the discussion. Environmental effects on air temperature are determined by this equation where the air temperature here—need my pointer just a second, I know how to do this—this as the air temperature right here. It’s equal to the measured temperature minus a couple of different things. The upper part of this equation here, that’s the short work wave absorptivity of our temperature sensor times the shortwave solar radiation coming in. And that is divided by this sum down here, which is specifically related to wind speed. So basically all this equation says is, if we’re wanting to measure air temperature, our sensor measurement T-measured is going to be adjusted by whatever solar radiation is coming in, divided by the impact of the wind speed. And that’s all contained on the right hand side talking about all the different details there. Why are we interested in it? Well, this is how we make a good air temperature measurement. It really takes into account those two things. First of all, we need to shade the thermometer to make sure that we’re not getting solar heating. And then ideally, we aspirate the temperature sensor by moving a lot of air over it. But when you do this it actually is pretty power hungry and we can’t run low cost or low power systems doing this. So we set about an effort to try to figure out, what if we approached this a little differently? If we measure both solar radiation and wind speed, is it possible for us to then correct the air temperature? And that’s exactly what we did. So here’s the air temperature measurement of the ATMOS 41. So the temperature is, well let me point out, this is the air temperature measurement right here. It’s nice and small. It’s just in a small stainless steel needle. At very low sun angles, at a zenith angle of like 85 degrees or something, the sun can hit that sensor just briefly, but in general it is shaded. And because it’s a small air temperature measurement, because it’s mostly shaded, we can actually measure solar radiation and wind speed. And using that information, we can correct the air temperature measurement. And it actually has. It performs extremely well, in reality. So here’s some data that we collected. This is air temperature error on the y axis and in time. So it’s been a lot of years since we did this, back in 2015. Can see why that that Texas setup was out there so long uncalibrated. This was during the same time period. The orange and the blue lines are uncorrected data of the ATMOS 41 and a louvered radiation shield sensor and they look pretty similar, although the ATMOS 41 is actually just a little bit better, just in its native condition. The gray line actually represents what we’ve done with a model to deal with air temperature and wind speed. And as we show here at the bottom of the screen, we’ve reduced the average error and the 95% confidence interval. So we get really good air temperature measurements out of that. If we actually go in the field and look at what that looks like in specific ATMOS 41s compared to a reference aspirated temperature measurement, we see extremely good consistency over all these six ATMOS 41s compared to an aspirated version. So in the field, they look great. Should have mentioned on the y axis here we have air temperature and x axis we have just time.

Okay, so let’s talk about precipitation. I’m going to talk about two more things precipitation, and then just finish up briefly with wind speed. And then we’ll talk about connectivity. So when we talk about precipitation, I write this subtitle of “So simple – not.” When I was at a conference one day, and we just finished making the ATMOS 41, and I was showing a really good friend of mine all the cool features of that, I got to the rain measurement. And I just said, Hey, rain is rain, right, we just measure it with a drop counter, and that’s it, we move on. And he was standing there with a colleague from a prestigious university, who I didn’t know, but apparently was a world expert on rain measurement and rain collection. And I spent the next hour having a very useful and deep lecture on why rain measurement should not be forgotten. And I haven’t forgotten that lecture. And so I’m here to say, precipitation isn’t simple, it’s not easy. There are a lot of things you should consider when you make those measurements. Here I’m just talking specifically about the measurement itself. And there are a lot of options out here. I’ve kind of, I got with a colleague of mine who does a great job drawing for me. And she put together these drawings, and I really appreciate it because it gives you a good idea of what’s going on in each one of these. So we have there top left is a disdrometer that hears the tapping of raindrops on a plate and it takes that and produces an amount of rain out of that. It uses sensing to figure out the size of the drops. You have the tipping bucket rain gauge. That’s probably the one that’s most common in the industry. You’re probably all familiar with that. It just simply fills a certain volume, at that point the bucket tips, dumps the rain out, and it starts filling the other bucket. The load cell, top right, it’s not as common a measurement out there. In fact, I was looking for a picture on the internet and didn’t really find one. These measurements just take a volume of water in a large container, and they’re able to take up snow and liquid water both because they use a solution inside that will actually melt the snow when it comes in. So they get their total precipitation gauge. They can be pretty accurate with their load cell but one of the challenges, we’ll mention it again in a second, that they can’t be super accurate like our next one, the bottom left, the drop counter. We can count every drop coming through. That’s what the ATMOS 41 does right now. And using that we can get something like .0017 millimeters of rain resolution, which is pretty incredible. The middle one down there is radar. That comes up on some all-in-ones, similar idea to the disdrometer, where we are judging precipitation size. And then finally the tipping spoon. Same idea as the tipping bucket, but instead of having something that swaps back and forth, the drops are just coming into the spoon, and it tips quickly, comes back up, and it continues to fill the bowl of the spoon. So, really on rain, we ask ourselves, What problem will I accept? Each of these things that I’ve talked about has accuracy issues. As I mentioned, for some, I’ll just quickly go through these. Disdrometers and radar struggle with drop size, and consistency, so often making errors in their in their total precipitation amounts. Tipping buckets and spoons struggle with tip volume. You can imagine if that’s tipping back and forth, especially if it’s going quite quickly, we can have some residual in the bucket, we can also get some dust and dirt down inside that fills the bucket of the spoon, and that’ll change the tip volume, so they need maintenance as well. Drop counters struggle at high rain rates, so they do quite well at lower rain rates, but if it gets too high, it struggles to see the difference between one drop and the next. And load cells, as I mentioned, have sensitivity issues. So picking the right thing depends on the error you can live with. And I wanted to mention here, we’re just, as I mentioned, releasing the ATMOS 41W, our new completely wireless system that contains everything all in one unit, instead of needing a data logger or anything outside there. We’ve kind of done something I think it’s kind of fun, is taking the drop counter and added a tipping spoon there to improve our accuracy. So you kind of look at both those things as challenges and say, Hey, why don’t we put these together to make sure we get all we need to in one device. So I put together a performance and price graph here, much like I showed you at the beginning of the presentation. But here we’re doing it just for the precipitation gauge. And you can see the tipping spoon is probably the cheapest on the list, going all the way up top to the load cell, which is the most expensive. And then we kind of have a mixed bag in terms of performance and price. The radar price is quite high, and the performance is pretty good, but I don’t think nearly to the level of the load cell. The disdrometer really fails. Although it’s a little more expensive in some incarnations than some of these other things, like the tipping devices and the drop counter, tt doesn’t really differentiate itself above them in terms of its ability to perform. So just something to think about that there are some challenges in our precipitation measurements that you got to take care of.

Okay, our last measurement, and then we’re just going to wrap up with talking about connectivity. The question I always have had is, Are all anemometers created equally? Sonic anemometers have an advantage and we see more and more of those on the market today. We use one in the ATMOS 41 and ATMOS 41W. A problem with some of these other measurements and the good thing about sonic anemometers is it is accurate at low wind speeds. There’s no starting threshold, there’s no stopping threshold, which can afflict some of these other measurements. There’s no bearings to wear out over time. And they are pretty low maintenance. There’s no dirt getting on the system to cause problems. And so in my opinion, there’s a lot of things to love about the new sonic anemometers that are coming out on the market. I want to just touch on one thing before I get to our summary slide. Connectivity and visualization are really important. And in a lot of ways, the devils are in the details. I want to tell you a little experience I had recently where we wanted to set up a site to just look at the performance of all in one weather stations out there to compare to a tier one standard site. This tier one site had all the nice bells and whistles that we’d expect of a kind of towards a World Meteorological Organization type site. And so we purchased a bunch of these all in ones and put them out there. There was one in particular that stated, Hey, this is a wireless system, go ahead and put it out there and set it all up. You’re good to go. When we got out in the field, that was not the case. In fact, it seems like we were missing maybe a $900 piece of connecting equipment that wasn’t in any information we found out there to get it on the cloud. And in fact, we’re still fighting that battle right now. It left me asking the question, What good is a weather station if you can’t get the data to where you need it? I mean, accuracy is all well and good. You can have the most accurate weather station anywhere. But if you don’t have the data in time to make decisions with it, it’s really not accurate at all. So I wanted to just talk briefly about the ATMOS 41W that METER Group is releasing right now. If the data stays with the station, as I said, does it matter how accurate the system is. So when you’re looking at a system, please keep in mind, connectivity must be included in the decision making that you’re doing. Visualization and decision support software is critical. So you want to test drive before you buy. And the ATMOS 41 Weather Station, the W, is something we put together to try to make sure all of this just came stock. If you throw this out on a fence post somewhere, within three minutes, you’re going to be able to observe your data out on the cloud, and be able to start making decisions with weather. Not sure you really should be in three minutes, but you’ll be able to if you need to. So the question I have for you to think about as we finish up here is, what do you want to— or how do you want to experience your weather data? For me, sometimes I want to experience it in an Excel spreadsheet. People make fun of me around here, I’m kind of old school as you might have imagined. A lot of people are using R and Python to do this work. I use Excel, when it’s really time to get down and gritty with the data. But a lot of times I simply use our ZENTRA Cloud offering, where I can easily get the data on the cloud. You can see in the middle picture there, just the map of one of the places I’m researching that happens to be the BYU campus, where we’ve got a lot of these systems out there to help control irrigation. On the right hand side, it’s the easily obtainable, evapotranspiration graph for that field. And we can quickly make determinations of how much water we need to be applying etc. So think about that. How do you want to experience your weather data? And make sure the system you’re investing in has that to meet your needs.

So in summary, accuracy doesn’t simply come from a spec sheet. I and probably you go on the internet a lot, poke on a spec sheet and start reading, how accurate is this thing? But think about this more broadly. Number one, you may not be getting the whole truth there. Sometimes that’s true. But that’s not really what I’m poking at. When I’m poking at is there a lot of other things that we need to consider. Obtaining great results out in the field means optimizing instrument value for your application. What is your Goldilocks? What is just right? Also agreeing up front that the labor for the station upkeep is going to be something you invest in. Know which sensors require recalibration and how often. And then do it. And then budget for enough stations to cover the climactic variation. This is important, I hope to get back to this topic in another virtual seminar. Finally, the thing I spent the most time on, ensure that the most important measurements are done right. That can have all kinds of meaning. Solar radiation — is that sensor even there on the system you’re considering, and is positioned correctly to get good radiation? And do you have a Redwing Blackbird that likes to come sit on it? That’s also a concern. Temperature, is it corrected for solar radiation and wind? Are you going to make good measurements of temperature over time? Precipitation, what is your choice for how that’s going to be measured? Is that up to standards that you’re expecting? And finally, wind speed, how much maintenance you’re going to do out there? How important is the wind speed to your conclusions? And finally, just that that quick little plug for data delivery. I know it was a short thing at the end of this presentation, when you’ve listened a long time. But really don’t forget that. I love sitting in my office and getting data and being able to analyze the systems that I’m overseeing, without ever having to go there. And in this day and age, I’m not particularly close to my research areas. So with that, I’m going to finish up and turn the time back over to Brad. There’s a lot of my personal information there. You’re welcome to connect with me if you’d like on LinkedIn. And I’ll take some questions.

All right. Thanks, Colin. So we will use about the next 10 minutes or so, we’ll do 10 minutes, to take some questions from the audience here. Thank you to everybody who has submitted your questions already. We’ve got a bunch that have come in. And there’s still time to submit your questions if you’d like. We’ll try to get to as many as we can before we finish. If we do not get to your question live, we do have them recorded and Colin or somebody else from our METER Environment team will be able to get back to you via email to answer your question directly. So Colin, just one thing that there have been several questions here regarding weather monitoring and agricultural applications. You mentioned in a segment you talked about microclimate and I was just wondering if you could kind of touch a little bit more in more depth about how those either growers or those doing agricultural research can use weather monitoring for their various agricultural projects and research applications.

Sure, and I probably can’t here cover the breadth and depth of those things. But what one of the things I’ve been interested in for the last several years is irrigation, irrigation intelligence in agricultural settings. And one of the key areas that’s attached to that is trying to understand better how evapotranspiration, for example, changes in certain areas. And so working quite closely with a grower in southern Idaho, for example, you might have seen them in other virtual seminars that I’ve done. And we asked that question, you know, what is the impact of a microclimate in one location compared to another? What’s the impact of where we put our weather station in comparison to the field? the original weather station that this grower was getting information from was kind of located at the top end of the valley he’s working in. But more importantly, it was located kind of in a section of ground nobody wanted to farm, which ended up being a pretty important issue, because right there, in that particular microclimate it’s much drier, because it was sitting out with no irrigation water. It was sitting in a pretty rocky area that ended up getting quite a bit of solar heating. And what we saw as I started to analyze the data that we were getting an overestimate of their reference ET. And in particular, for the crop modeling, we were getting a daily overestimate of the high temperatures, which was ending up driving the predictive models for how quickly the crop was growing to go well beyond what was actually happening in the field. When we deployed some of these ATMOS 41s in the field themselves, we not surprisingly saw much lower temperature, comparatively. And therefore our predictive ET was lower. Now, I mean, that’s just one instance, though, when we think about, for example, fruit growers in the center of the state of Washington, and we know that the down on the bottom versus up on the side hills and things, we get a vastly different temperature, especially at night, when we have concerns with freezing during the early spring. And but at death. These are all important things to consider, and things, I can’t go into more depth here, but something to think about that there are microclimates out there, and things that we need to address. And we have the opportunity to address them now that we have better instrumentation available.

There was one question and this kind of goes along with that. Do you have any insight on the state of the science as it is right now, with regards to forecasting microclimates?

Yeah, so we’ve been working a little bit with this, just actually working with other instrumentation to do some of this forecasting some of this gridded data gathering where they take weather stations around and try to produce a gridded product that will tell you hey, here’s the information for your specific location. In terms of forecasting, we have a group here at METER that’s actually focused that’s focused on predictive frost monitoring. So they’re looking like 6 to 10 hours into the future if they can predict what might occur in terms of potential freezing situations. And that’s actually getting quite good. But that is derived off one, let’s say ATMOS 41 at one location where we can predict what it is into the future. And then we can follow what it actually was and see how good our models are doing. And as I look over those data, I’m pretty excited because we’re getting nice consistent information in terms of those predictions. Now, these gridded products that are that are working to kind of predict this spatial microclimate information for whether, those haven’t been as good so far. We were doing some comparisons, as I mentioned, this tier one versus tier two site that we are connected with. And we’re doing this all in one comparison at, we’ve compared one of these data products that actually trying to predict the temperature there. And as you might expect, and I’ve only done the temperature so far, we’re just kicking this off, we’re seeing an overestimate of two to three degrees Celsius. And interestingly, we’re seeing an underestimate at night of two to three degrees as well. So the overall temperature envelope and this is just sitting on the top of a hill here in the eastern Washington area, about two to three degrees Celsius, both sides.

All right. Along with this, one last question with with regards to ag applications. If somebody is concerned about pest prediction, would it be more useful to have a weather station within or below the canopy of their crop?

Yeah, that’s a good question. Um, and that really all depends on where the models were made. And that’s something we’re kind of running into. So last year, I was quite interested in coddling moth prediction and started doing quite a bit of looking into the literature out there and the current models. And what I found was that there was an opportunity to do a better job at temperature measurement and therefore modeling in these orchards. But at the time these models were made, it really is important to know where they actually took their temperatures from, right. So if we took our temperatures from right where the the moths are going through their lifecycle, that would be great. And we have the ability to do that now. But if the people generating the models were taking their temperatures from weather stations above the canopy, or even outside the orchard, for example, then that’s not going to match up super well because what we’re predicting goes well beyond the things that had been done in that modeling. So what I would say is, that’s an issue. I also want to suggest that I talked to some growers and said, Hey, let’s, let’s figure this out, we can do a better job of spraying, they were all excited, but then they put it out, you know, we’re so busy that we can’t get in the field, really, maybe even within a week of, you know, tops, let’s say, a week of when we see that there is an issue that we got to go spray for. And I said, Oh, wow, that’s gonna be, you know, what good is all this accuracy if you can’t get out there on time? So I don’t know, many things to think about there.

All right, we do have a couple of questions in here dealing with how many weather stations to install. So one of these questions in particular is asking if I’ve got one research site or one field, do I put in several of these weather stations in each of these different segments? Or can I just pop one right in the middle of the field and that will be good enough?

Yeah, thanks, Brad. That’s actually a really great question. And the answer that is, from our experience, especially with something as you mentioned, like ET, a single weather station out there for multiple plots on 20 acres of land is more than enough in our experience, because right now that weather station, just has generated a reference ET value, and so we’re not going to be able to get accurate enough to give you information on the plot level. If we were interested in that, maybe I’d stick some soil moisture sensors out there, some water potential, water content sensors to get specific water loss in those sub plots, and try to generate our information about actual evapotranspiration for there from that, but a single weather station should be more than sufficient.

All right. We’re hitting our time. So I think we’ll do one more question here. And this one, there have been others that have been similar to this one. But they’re asking, saying, I’ve used a lot of temperature sensors in the typical non aspirated radiation shield. How much of an impact does inaccurate air temperature have on calculations like ET, you know, what you were just talking about there?

We’ve done some analysis on some of this. And I would say that it does really depend on the situation. If we’re talking about evapotranspiration, it does have some impact, as I mentioned, that solar radiation was going to have a much greater impact than temperature. And we’re going to see just, you know, I mean, we’re talking about tenths of millimeters on a day, you know, maybe a tenth, or two tenths of a millimeter if we’re really inaccurate in our temperature. I don’t have a good number for that. But that’s the general idea. We need to, the idea is we need a much better measurement of solar radiation before we worry too much about temperature. But in terms of plant growth modeling or pasture disease modeling, that’s going to be a much more important area where we need to get the temperature right. And Brad, by the way, one of the things that came, one comment that came through that I liked was that both it, how quickly in terms of gusts and how easily in terms of kind of the die down or the wind, the momentum of the cup anemometer. Those are also issues that sonic anemometry overcomes. So thank you for your comment. Yeah, there are more good things about sonic anemometry than I even mentioned.

All right. I think that’s going to wrap it up for us today. Thank you, everybody for joining us. And we hope that you enjoyed this discussion as much as we did here. Thank you again for all of your great questions. Like we said, there were several that we did not get to, and we do have them recorded and Colin or somebody from our METER Environment team will be able to get back to you to answer your question via email. Please consider answering the short survey that will appear after the webinar is finished, just to let us know what types of webinars you’d like to see in the future. And for more information on what you’ve seen today, please visit us at Finally, look for the recording of today’s presentation in your email. And stay tuned for future METER webinars. In our next month’s webinar, it will be our Office Hours Live Q&A session with Doug Cobos and Chris Chambers who are research scientists here at METER Environment. And they will tackle your questions about measuring weather parameters, so related to today’s webinar here. And we will be sending a link in our follow up email. So that’s all. Thanks again. Stay safe and have a great day.

icon-angle icon-bars icon-times