Episode 6: Helping Growers Bridge the Technology Gap

Episode 6: Helping growers bridge the technology gap

Dalyn McCauley, Oregon State University researcher and G.A. Harris Fellow, discusses her research on crop-damaging weather events and how scientists can help growers implement changes that will improve their quality and yield.

Next steps:

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


The views and opinions expressed in the podcast and on this posting are those of the individual speakers or authors and do not necessarily reflect or represent the views and opinions held by METER.


see all podcasts

Episode 5: The fascinating backstory: ATMOS 41/ClimaVUE 50 weather station

Stories, challenges, bloopers, and successes behind the development of the ATMOS 41/ClimaVUE™50 weather station: a METER/Campbell Scientific collaboration


Episode 4: How to be field ready

Veteran scientist and author Sara Vero talks about her upcoming book Fieldwork Ready and how to make sure you're ready for any surprise your field experiment throws at you.


Episode 3: Making measurements on mars

Dr. Doug Cobos discusses challenges his team faced measuring water content and thermal properties on Mars, how NOT to burn down a NASA building, and much, much more….


An abstract illustration of multiple green dots of different shades above faint grey dots with a faint lines connected a few of the dots

Grow your expertise with the latest podcasts

Receive the newest content on a regular basis.


Hello everybody and welcome to we measure the world, a podcast produced by scientists, for scientists.

I think the as that kind of just approaching the conversation, we had maps ready for him like this is what the data can show you these like very, you know, tangible maps that show you this side of the vineyard is a little hotter, you know, for for whatever reasons, but just showing him the maps and like, this is what the data can provide. And then yeah, we talked a little bit about kind of futuristic potentials of like, Okay, what if you were driving your tractor around, and there was an IR sensor on one of the booms, and it was just capturing the vineyard, like having to integrate some kind of sensors into a practice that you’re already doing rather than saying, Oh, we want you to fly a drone twice a week or something.

That’s a small taste of what we have in store for you today. We measure the world explores interesting Environmental Research trends, solutions to research issues, and tools to better understand the entire soil plant atmosphere continuum. stay current on applied environmental research, measurement methods, and more. Thanks for joining us. Today’s guest is Dayln McCaulay, Dayln was a 2018 ga Harris fellowship winner interested in developing a site specific decision support tool for on farm management of crop damaging weather events. Today, she’ll be discussing her master’s research project, which was focused on collecting environmental data to develop predictive models for detecting early onset of downy mildew and vineyards for disease and risk management. Dayln, thanks so much for being here. So just to start things off, can you give us a brief introduction into what inspired you to start doing this type of research, you know, what kinds of problems do vineyard growers face that kind of prompted this research that you’re doing? Yeah,

yeah, so a grower contacted us he was concerned with powdery mildew disease on his vineyard, and just was looking for some kind of way to better inform his management of it. powdery mildew was a pretty big issue in the Pacific Northwest vineyards industry. It’s, you know, highly dependent on weather really likes moist, temperate kind of cloudy areas, which is exactly the Pacific Northwest. So it’s a big concern of a lot of growers. That’s, like I said, almost always present. And it can be detrimental in the form of, you know, yield losses, quality concerns with the grapes. And then if it gets bad enough, sometimes you have to, you know, slash and burn an entire vineyard just to eradicate it. So it’s a really costly disease. Of course, fungicide applications are really costly trying to mitigate it. So yeah, just looking for a better way to manage it. Because it’s everywhere.

Yeah, definitely. So what kinds of instrumentation Are you using in these vineyards?

So we use the instruments that we got form Meter to make those four weather stations, and we distributed them across the vineyard, when basically four different locations that we suspected to have pretty different weather conditions for different microclimates that we suspected. And so yeah, we measured use the Atmos 41, just to get general weather parameters, we use the soil moisture, sensors and matrix potential sensors. And then also that leaf wetness sensors. And so yeah, we distribute them across the vineyard. And then we just track general weather parameters. And then also calculated the powdery mildew risk index, which is a temperature based index developed by UC Davis IPM pest management program, it gives you a disease pressure of powdery mildew disease pressure, basically from zero to 100. So 100 would indicate a severe, powdery mildew pressure. So we also tracked that in the four different locations just to see, yeah, how it varied and how the farmer can use that information to better inform his fungicide application management.
Awesome. Yeah, that’s so cool. So right now, I mean, it seems that everybody’s talking about machine learning. And I know you weren’t able to get the machine learning part of your plan yet. But how were you going to apply it or, or how are you planning to apply in the future?

So our, idea was to integrate basically three different types of data, one, the weather data and the four locations. And then secondly, we were hoping to quantify the actual disease incidence, the spatial and just like a severity of the disease incidents with four different locations, integrating that data, as well as some remote sensing data. So we did surveys of the vineyard, using a hyperspectral, uh, hyperspectral reflectance sensor, as well as a infrared thermometer to get canopy temperature. And so we’re hoping to use that to combine the spatial aspect with the temporal aspect of the weather data. So we’d have like timestamps of spatial maps that we could then link to Historical weather, weather records, and then again link that to the disease presence. So we’re hoping to stack all of these data types on top of each other and run them through a machine learning algorithm, such as an artificial neural network or something, just to develop some kind of these heat maps of disease pressure, that we could link to weather and canopy temperature, something that the farmer could say, Okay, let’s just use an infrared thermometer and take a temperature of this leaf and put it into this model. And ideally get out with some kind of disease pressure value or a disease potential for that, that section of the vineyard for that time of year, given the weather, you know, combining all these things. So we haven’t got there yet. But we do have the majority of those data sets kind of ready to stack into some kind of model.

Nice. Yeah, you know, we’ve been associated over the years with a lot of these different projects. And that’s a familiar theme that machine learning is more complicated than people expect. But the coolest part is that you are trying to put it all together. So on a another note, I love the spatial and temporal aspects of this project. What did you learn as you collected all the data

for my thesis, and then the publication that we’re putting out, we mainly just focused on that spatial point measurements from the weather stations, and then that temporal data, so we were able to see that they were definitely, we had the four weather stations, but we saw that there were probably three distinct microclimates that had different susceptibility to the disease. And then, given that, we also saw that using the powdery mildew risk index, we saw that there are definitely, again, three different areas that could benefit from unique spray schedules. So for example, one of the stations that was kind of higher on the on this ridge in the canyon had much higher severity throughout the whole season than another station that was kind of lower and hotter. So in that case, the powdery mildew risk index indicated that that that higher disease pressure area should have had fungicide spray weekly, whereas the other one should have had fungicide spray, maybe every three weeks. So that was our main result was just showing that. Yes, there are different microclimates across a single operation, and they could benefit from more targeted unique disease management strategies.

Definitely, definitely so so as we think about the the impact on the grower, the ways research can really push change that will have a positive impact on the environment, did you get a feeling for what the the environmental impacts might be of, for instance, a one week versus a three week schedule?

Right? So I didn’t go into a cost analysis or anything, just because when you’re using just the risk and probabilities, it kind of creates these pretty complicated, you know, how could I say that this disease schedule would save you this amount of money, because you don’t actually know if it’s going to work. But I think it came down to that, you know, across the the four stations that the farmer could have saved to fungicide applications throughout the season, which doesn’t seem like a lot. But you know, the Think of it takes them probably two full days to do the whole operation. And definitely a lot of fungicides, which are one of the most expensive inputs for vineyards. So two whole applications over several years, that kind of marginal, like savings would actually be beneficial.

Yeah, that’s so interesting. I mean, there’s time and money saved. But then there’s also an intangible benefit, which is the lower impact on the environment. But you, you mentioned the higher location and the lower location. And I was gonna bet just from what I heard you say that down on the bottom would be kind of a wetter environment. And that was going to be where you’d see more powdery mildew. Did the results surprise you?

Yeah, that result did surprise me. I think the lowest elevation happened to be the hottest location as well. And part of that disease risk index is that you know, any temperatures over 95 degrees is harmful for the fungus. So that surprises in that I think that was the main driver as to why the lowest location was less susceptible, especially over July and August when it was the hottest. And then yeah, it was definitely surprising to see that the highest one had the highest severity, especially being that from historical experience. The farmer always recognized that powdery mildew developed in the Chardonnay first which was kind of like this middle elevation, whereas the Grenache was the highest elevation and that showed to have the most of your incidents. And that either tells us that, you know, maybe the farmers is not looking in that location. Or maybe there’s something wrong with that disease risk index and that it’s not taking into account the conditions in the Chardonnay that were actually more susceptible to the disease such as can it be bigger or that can’t be big. In that the the inside of the canopy was more moist and had less radiation, which is harmful for the fungus. But yeah, that was definitely interesting.

So you did this analysis where the models were showing you one thing, and you said you did hyperspectral imaging, as well as IR measurement. Were there other measurements that were kind of validating what you were seeing in terms of that powdery mildew index?

Yeah, so the hyperspectral, we did initially to see if we could develop some kind of disease index, spectral index, that kind of the caveat that I should mention is that this vineyard, you know, historically, always gets powdery mildew every single year. And this year, the year that we did the study, you know, fortunately for the grower, a little bit unfortunate for research, they didn’t get powdery mildew. So we weren’t able to link that kind of final, you know, you know, real valid ground truth measurement. But with the thermal canopy measurements, we did see some things spatially that made sense. We we mapped there in that lower vineyard, the serraj, we started that one corner was always a lot hotter throughout the entire season, then, basically, the corner on the edge of the vineyard was a lot hotter than the corner in the middle. That corner happened to be right next to a gravel road and didn’t have any vegetation, it just kind of Rocky from everywhere next to it. And so we just assumed that that was you know, there was higher evaporative demand coming from that gravel road that was increasing the heat and basically stressing that corner of that vineyard a little bit more. So it’s kind of interesting thing that the farmer didn’t recognize it wouldn’t know about that we were able to catch with the surveys and the measurements.

Yes. So that leads into another question was the grower aware of the microclimates in the vineyard, or maybe he didn’t have a sense of that, as he walked around the vineyard,

I think he knew when he went up the hill, it was windier, and a little cooler, and that when he went to this other vineyard, it was hotter. But he didn’t, For my knowledge, he didn’t use that to change name his practices, except for one with the Syrah. He always knew that those were ready to harvest a couple of weeks earlier, just because it was a lot hotter down there. And they had accumulated, you know, more growing degree days. So he, he identified that that one was normally hotter, and that he should harvest a little bit sooner. But other than that he wasn’t aware of these kind of finer scale differences in a microclimate such as across a single row, one side to the other.

Yeah, I asked that that question. Because we’ve had similar experiences, one grower had dry spots in his field that were affecting his yield, there was like a 25%, drop in yield at those locations. And when we showed him how measuring water potential can prevent plant stress, the next year, he installed water potential sensors everywhere on his farm. And he’s still very actively using them. Were there similar impacts as you showed this vineyard grower the data?

Yeah, similarly, so we use the four Atmos and we actually gifted him one of them to keep that has been your permanent leak. So he was interested in mainly the the powdery mildew risk index, kind of just having that number that he could base his race schedule off of he wasn’t as interested in, in the canopy surveys, just because I think, you know, yeah, he could have someone come into his vineyard and do that for him every summer. But it’s not as tangible. It’s not as available for him to actually use on a daily basis. So he was really interested in just having that on site tracking of some kind of disease index that could help inform, you know, even if he doesn’t use it, just some kind of double check for his own management decision making like, okay, it’s high severity. Now, I’m going to, I’m going to do an extra herbicide application here, or I’m not going to do one today, because it’s low. So he wanted to definitely keep that part of the information. And so, I think he’s still, I think we took it down for the winter for him. But he had that for this whole season as well. tracking that index.

Yeah, that is really interesting, the connection between instrumentation and decision making. And you made a really important point about the usability of the data. You know, he was thinking, as long as the data show up on my computer, I’ll look maybe changed my management practices. But he thought hyperspectral imaging and IR temperature may be a bridge too far. Can you talk about those conversations? I mean, you must have been thinking, I want my research to be really impactful. How could I take the next step here? But what is the next step in trying to get these valuable practices in?

I think the, as to kind of just approaching the conversation, we had maps ready for him like this is what the data can show you these like very, you know, tangible maps that show you this side of the vineyard is a little hotter, you know, for for whatever reasons, but just showing him the maps and like this is what the data can provide. And then yeah, we talked a little bit about kind of futuristic potentials of like, Okay, what if you were driving your tractor around and there’s an IR sensor on one of the booms and it was just capturing the vineyard, like having to integrate some kind of sensors into a practice that you’re already doing, rather than saying, Oh, we want you to fly a drone twice a week or something, we talked about the potential of using data like that, again, it’s still a little bit too far fetched as of right now, but as far as doing research and developing these kind of tools for farmers, you know, that’s what we need to start thinking about is how do we make it easy for them to use this data really, in things like that. Yeah, monthly drone images, maybe of hyperspectral, that you could plug into a platform that just gave you a map or Yeah, some kind of periodic survey that you could do of your vineyard. But it’s hard to bridge that gap with the more complicated data products.

Yeah. So when you talk about some of the barriers to entry for the technology, is it cost? Is it time? Is it a combination of those things? I mean, I think of vineyard production as a fairly lucrative endeavor. But I don’t know, did you come away with any feelings for what those barriers are?

I think, really, with some of them, it’s the perceived value, he says, okay, so that start side of my vineyard is a little hotter than the rest. But what does that mean for what I get at the end of the day, and we didn’t, I wish we would have, but we didn’t, we weren’t able to track yeild or anything to kind of show him a real economic benefit of incorporating some of this information. That’s why I think disease is an important or a particularly like, effective one. Because, you know, a disease outbreak is entirely detrimental to a vineyard. So they’re kind of, they’re looking at it as a high risk variable. So it’s like, I think they’re more likely to put in the effort and invest in those kind of sensors, then, then maybe, you know, a corner of his vineyard is getting a little it has a little bit less yield. I don’t know what the actual impacts of that heat map would have been on his financials. But yeah, I think it might be the perceived value of the data. So getting that across them somehow being good first step.

Yeah, I really appreciate that perspective. Because when we think about research, and how it impacts others, you know, we’ve got to show them return on investment, to get people to change, they’ve got to feel a need, you know, some pain even. And it seems like at least in the year that you were doing it no powdery mildew equals no pain. But if the plants are wilting, or if the sweetness of the berries is better here than there, that might get them asking questions. Now, in this study, did you measure soil moisture, either water potential or water content?

Yeah, we did. Soil moisture and water potential in the same location. That was mainly to track irrigation events just to see the spike in the signal to identify irrigation frequencies. And those were relatively equal across the entire vineyard, I think one of them was a bit of a measurement bias or experimental setup bias, and that most of them are right under a drip emitter, and then want the one on the very steep slope, which was the highest elevation, it was always a little bit drier. And I think the water was just rushing right by it. So it wasn’t like pooling at the sensor itself. And so that one was generally had lower soil moisture than the rest of them.

Interesting. Yeah. So were there any major bumps in the road in terms of deploying instrumentation or trying to get all the data put together there? What were some of the challenges that you faced? And how do you overcome them?

I’d say the major bump the road is not having the powdery mildew. And so there, we just switched gears to really looking at disease pressure and potential disease, just going to form like, changed our research questions to look at that. But other as far as instrumentation, with vineyards, something I didn’t realize, but I come from an engineering backgrounds. This is my first agriculture like application and being out in the field. The solar we had solar panels on the weather stations and through like middle of July, all of a sudden, all the weather stations go dead. And it’s because the canopies started growing over them. And solar, solar power just reduced and they couldn’t stay. So we had to cut back the vines a little bit. So it was kind of an interesting bump. Other ones I, I ended up making my own, like Arduino based data loggers to query the sensors. And so just difficulties with that kind of open source platform. In the beginning, there was kind of difficulties with the real time clock and sinking the data and just more like technical issues with getting the station’s running. But all most of those were were solved in the first couple of weeks.

So you didn’t have a ton of farming background coming in and you’re an engineer. When you went to the farm. What were your perspectives when you got there?

Yeah, I guess being an engineer just want to like record everything. I was like wanting to do the yield. I wanted to do you know leaf assessment samples, I wanted to do weekly surveys instead of bi weekly surveys, I just wanted to collect all of the data. And I just quickly realized that that is just so not feasible a lot of the times are you can’t capture all the variability when you’re in the environment. Engineering, you’re always kind of in controlled environments, when you’re learning in engineering school. And so just having that open environment that was so variable, it was a little overwhelming to be like, Okay, what are the most influential variables for the problem that I’m addressing? And how do I account for variability, it’s just natural variability and try to minimize that for, for control for that in my studies. Yeah, so just kind of narrowed down four locations and the sensors I had are sensors I had, and I could kind of go a little crazy with the surveys and try to do as many as I could and get as many plants that I could in them. But then even realizing that at the end, all the data from the surveys took so long to compile and get into a format that was actually usable, you know, you can take pictures and hyperspectral sensor measurements, that is just such high dimensional data that it’s hard to actually incorporate into some kind of model. So just realizing the the realities of collecting data in the field. And using that in a useful way. It’s definitely kind of eye opening.

I totally agree. figuring out ways to aggregate, which is the state of the art right now can be pretty overwhelming. So next question, how would you balance the cost of the measurement system with the value of the data at these different sites? If you had your own vineyard? How would you distribute some of these systems?

Yeah, yeah. So I think so we put out four we put two in the same vineyard block, but two different sides of it. And we found that those were basically identical weather microclimates. So yeah, I think, if I had my own vineyard, I would want to place a bunch out just to assess the variability and then narrow it down to the the four or five most different unique climates. And yeah, definitely putting at least temperature, relative humidity sensors, you know, distributed across the vineyard just to see one, like a lot of the disease indices are just temperature based. So that’s a good record to have. And then growing degree days, frost, of course, I think just a single temperature measurement distributed could really benefit management and targeted decision making. It’s always hard though just to pick which where where to put them, you know, where is the most variable, and it’s kind of the hard thing, but the cost is, initially, you kind of need that in that first survey to, to realize that, okay, station one and station three are basically the same, we don’t need two stations in this vineyard. And so it’s kind of a high upfront cost, maybe to just have that high spatial resolution, you know, information and then narrow it down to the four or five most influential places. And this vineyard particularly I think, three would have done, one have done fine, one in the in the lowest part one in the highest part, and then one somewhere in the middle, I think that would have been sufficient to capture the most variable places.

Yeah, maybe the idea would be initial surveys to figure out where the variability lies, maybe satellite imagery? That’s something we’ve been poking at, and trying to look at variability and trying to spot historical locations that may be more interesting.

I think that another thing to think about is just the operational, you know, at what resolution can can the production apply? Like, if you ever having super high spatial data in a single vineyard? Well, they’re not going to just spray every plant differently, you know, he’s probably going to spray an entire row or if not an entire block similarly, so just looking at the resolution of the actual production and operation activities. It’s important thing to keep in mind.

Yeah, it seems like there’s a real balance to what we can do and what is truly feasible. And if we’re trying to design this for mass agriculture, you know, what does that look like? And so what are you working on now?

Yeah, so I’m working. I’m working at Oregon State’s my faculty research assistant at the North Willamette Research and Extension Center, so it was an agricultural extension center, working with the nursery production program, some kind of like the sensor technician, I’m working on sensor based sensor controlled irrigation for container nursery industry in Oregon. So yeah, right now I’m working on lots of sensor development for integrating sensors into control, irrigation control program. So I developed these mini lives emitters that we’re going to use to, to weigh containers and then control irrigation based off of depletion and container mass as just a way to make irrigation a little bit more data driven for container nurseries. And then we’re still working with these low cost distributed weather stations as well. I’m going to be working in hazelnuts, putting them out in a hazelnut orchard and developing some Crop water stress index using canopy temperature and local weather parameters as well.

Nice. Yeah, we’ll definitely be interested to hear how that goes. And it looks like our time’s up for today. Thanks so much Dalyn for taking time to share your research with us. And we definitely wish you the best of luck in all of your future endeavors. And if listeners have any questions about this topic or want to hear more, feel free to contact us at metergroup.com or reach out to us on Twitter @meter_env. You can also view the full transcript from today in the podcast description. That’s all for now. Stay safe, and we’ll catch you next time on We Measure the World.

icon-angle icon-bars icon-times