Ag Geek Speak
GK Technology Inc Team Members, Jodi Boe and Sarah Lovas talk about precision agriculture, agriculture mapping, agronomy and drainage.
Ag Geek Speak
20.5 Tiny Bytes: Choosing Quality Imagery
Join Sarah and Jodi as they discuss the role of satellite imagery, its various sources, and how to select quality satellite imagery for use in precision agriculture applications.
Jodi and Sarah cover NDVI (Normalized Difference Vegetation Index) analysis and its relationship with crop yield, guiding smarter management of field crops. They share real-world experiences with tools like ADMS and insights into sourcing reliable imagery from platforms like Sentinel, NAIP, and Landsat. This is a great episode to listen to if you want to learn how to overcome challenges like cloud cover and smoke, blend resolutions for more accurate maps, and unlock the secrets to creating actionable zone maps.
Whether you're a precision ag pro or just starting your journey, this episode delivers practical tips and a fresh perspective on using technology to manage field variability. Tune in and take your crop production to the next level!
Tiny Bytes Episode 5: Raster Monster! https://www.buzzsprout.com/2297340/episodes/15509080
ADMS https://gktechinc.com/adms-product/ and GK Server access information: https://gktechinc.com/consultants/
And now it's time for a tiny bite of knowledge. Hey Sarah, hey Jodi, what are we trying to get out with a zone map?
Sarah:Well, we are trying to understand the consistent variability of crop production in a field. In other words, we want to know consistently, year in and year out, where the areas of a field are that have higher production potential and low production potential. It's a very useful map.
Jodi:I would wholeheartedly agree with that. And how do we actually get there?
Sarah:In order to have a good zone map, we need to put together data layers that actually help us understand where there is greater crop production potential in a field and lower crop production potential in a field over time. Now we can put together a lot of different data layers for doing that, but one of the data layers that we use a lot here at GK Technology is imagery. Imagery can be a really great way to help us understand how the plant is responding differently across a landscape. One of the things that we do at GK Technology is we are actually a raster- based software. ADMS is a raster- based software. What that means? When we make a map, one pixel has one numerical value with it. Now we have a whole podcast about rasters, so we're not going to go into that. You can go back and check the tiny byte files for an episode called Raster Monster. That one turned out really good. I really liked that one, Jodi.
Jodi:I like the tiny byte files. I love that, but really do check it out. And when you think about it, we're trying to figure out where the consistent variability is in the field with these zone images, and what we need to do, that is, we need to use math to figure out where are these values consistent, where are they staying, where are they different, where are they similar to each other, and so that one pixel having one value allows us to do math across layers, which is awesome. A really great source to get those values is imagery, and so today's episode we're going to talk about. You know, where do we get this imagery, how do we use it, how do we pull out those numbers? And then, what are we actually looking for in those images when we're putting together zone maps, specifically?
Sarah:So in order to get into a raster image and we get to that place where we've got one pixel having one value so that we can mathematically consider the production potential across a field, in order to do that we need to take images, us produce vegetative indices or pulling out individual bands of light which can help us understand that. So oftentimes I am using NDVI or normalized vegetative index values. They are scientifically accepted method of understanding just how green the plant canopy is and it is directly correlated to the production potential in a field. And so, basically, what happens with NDVI is it's really a mathematical equation between the near infrared band of light and the red bands of light.
Sarah:Why is that important the near infrared band of light and the red bands of light? Why is that important? Because when our eyes see green, it is because green is reflected, because red is being absorbed. The darker green the plant canopy is, generally speaking, the healthier it is and the more red light that gets absorbed. So we're able to quantify that with that index. Now there's a lot of other vegetative indexes that exist out there Green NDVI, we can even make a vegetative index out of RGB bands of light and furthermore, we can pull out individual bands of light as well, to do different things, but by far and away. Generally speaking, when I am trying to make zones to understand how crop production is different in a field, I'm pulling out the NDVI.
Jodi:When we think about pulling out NDVI and like thinking about images and these values behind them, let's remember that, like, an RGB picture has three bands of light red, green and blue and in order to get to be able to do the math, to make an NDVI value, you need to have infrared band with that as well, and so that's really important to mention now, like, okay, where can we find this kind of imagery that has all these three bands of light the red, green and blue plus also that infrared band? And that's why it's so important to have a good data source for this imagery and, specifically, satellite imagery. It's really important to have a good data source for this imagery and specifically satellite imagery. It's really important to have a good source of imagery. And also, like, when you're putting these images together and let's remember I mean, sarah, when you last put a zone together, how many images did you download to put together that?
Sarah:zone. You know, as we were preparing for this podcast, I thought a little bit more diligently about that and I would definitely say it's very easy for me to download probably 36 to 40 images and I would definitely say, of those images downloaded, I am probably keeping someplace between 20 to 30 images, maybe as low as 15 images to produce one zone. I would never make a zone out of one image.
Jodi:So you're pulling together quite a few images to put together and help paint that picture of consistent variability for that final zone map.
Sarah:Absolutely, and I think that's a really important point, Jodi. But how about now? You answer that same question how many images do you download when you're making zones?
Jodi:out of imagery I think very similar, probably 36 to 50, just depending on how I realize that's a big range, but depending on what state NAIPs I'm using and if I've got really good 2016 data, I'll pull that in too. So it but I use. The point is I use a lot, and what I love is being able to toggle through all of those images quickly and quickly decide whether or not I'm going to keep something, and that's what's beautiful about having access to the GK server is that you can quickly download Sentinel data, naip imagery and even Landsat data back to 1984. And not only that, too, but LIDAR data, which you might not use so much for putting together zone maps, but you might want to use to validate your zones, and so having that makes the process of bringing in those imagery layers and getting a good base to start squashing down and actually making that zone map it is so crucial squashing down and actually making that zone map.
Sarah:It is so crucial. It's very crucial and I love how you bring up our access to the GK Server. You don't have to use the imagery in our GK Server. I am aware of people that download their own satellite imagery or they've got their own imagery sources that they want to be working with. That is fine. Adms software will handle that as well. It's just really handy to use the data that's already there. I know in the past we've calculated up how much data storage we have in our server room and it is a petabyte of storage. It just blows my mind that we've got that kind of storage capacity for imagery. But when you start thinking about the number of states we're having we're north of 40 states that we've got imagery for, and when that library goes back to 1984, that's a lot of imagery, that's a lot of data.
Jodi:So, in short, if you're looking for a way to get imagery quickly and be able to download multiple years of imagery and be able to look, get access to the GK Server. It's going to save you a lot of time.
Sarah:I would agree it does save a lot of time. But let's talk for a second about when you're going through that library of images. I mean, it can be almost a little bit overwhelming. So when you're looking at imagery, what are you looking for? That is an actual good image for zones. And let's let's target it to zones. We can bring in images for other things, but when we're trying to find images that are going to describe the crop production in season, what are you looking?
Jodi:for Jodi, that's a great question and like, really, what I'm trying to get to when I first start looking through the images. What I'm looking for, like what I first try to look for, is like where's the point where I'm hitting maximum canopy coverage in the field, right? So like where am I getting?
Sarah:where am I getting to a point where, like, I'm getting to maximum field canopy coverage, and then also before that point, so when you're looking at that image and knowing that it's that image actually looks red from the satellite imagery right Cause we're actually looking at the light that's absorbed Are you trying to see, like how dark red that image is compared to black? Don't you ever look at it that way? I'm always trying to see like you're wondering where in the heck I'm going with that. Yes, so I think you bring up a good point when you're trying to find that max productivity to production, that when it's when it's really when you have the maximum canopy coverage.
Sarah:When I'm trying to find that point and I am looking at those images and especially those sentinel images, you know, when we think about again what NDVI is, we're trying to figure out where that plant canopy is densely green and the satellite imagery that we're looking at is really how much red light is absorbed. So when we're looking at that imagery, a lot of times we're seeing we don't see an image like what our normal human eyes would see. It looks like a red blob on the screen, to be honest with you, and what I'm looking for is that deep red color with that image, because that's an indication of how dark green it is. The plant is.
Jodi:Oh yeah, and I just look at the NDVI image that's put on the other side. I hardly ever look at the red image, except for when I'm looking at for clouds, right.
Sarah:That's. The other thing that I'm always looking for, though, is clouds. When I'm looking at the raw image itself, I'm looking for two things how deeply red is that image? Because that's an indication of where the plaque canopy is actually green. And then I'm looking for the clouds, and clouds can be just a confounding thing.
Sarah:Now to your point about the other side, where we're actually taking a look at what that raster is going to appear, produced from that satellite image. I want to see how that looks on there. I want to make sure that we're bringing out the variability of that field. If it is completely saturated with a completely full green canopy, that is actually going to be not a very good image. If it's completely green across the entire thing, we're not going to understand where the areas of variability are. So, from that point, I'll take a look at the images earlier in the growing season and a little bit later in the growing season, and I want to quantify those spots and I'll try to put those images together. Clouds we got to hit on the clouds. When you're looking at images and you're looking at clouds, what are you looking at when you're out there?
Jodi:so I again, like I mostly I like looking at so in adms, when you go through the gk server, what you're going to get is you're going to look at the raw image on the left hand side and then, once you extract it, um, you'll see like the, the, the raster preview on the right side, so the, the raw image, has like that what you're looking at is like the deepness of the red or like how green the canopy is, even though you're looking at red.
Jodi:But when I'm going through those I do look for cloud coverage and sometimes it's really obvious, right, like you'll open up an image and it'll just be like one color across the whole screen and it's like, okay, well, that's obviously some cloud cover here.
Jodi:But sometimes it can be a little bit harder to pick that out. And so I typically, if I, if I see any sort of like really dark shadow across the, the raw image, the red side of the image, it's a pretty good indicator that there's a cloud there. And sometimes it's not so easy, like, sometimes it's easy to spot that there's a cloud there, either that that dark black shadow, or like a white wispy cloud that's coming by. But what's really important is maybe you don't catch it on the front side, but as you're going through the imagery, like the NDVI rasters that pop out from the images, if you're going through each of these images and all of a sudden you see an area in the field that is red or like non-productive compared to like the other four images you downloaded for that year, it's a really good sign that you got a cloud in there and it's important to throw that image out and not use it when you're putting the images together for zones.
Sarah:You know it's interesting, when we think about clouds, how smoke in recent years has really made imagery a little bit challenging. It's interesting because it isn't like you can't see the field underneath, but it's like that light is diffuse and so it's really weird. I have sort of been able to make some of those images work, but on a whole it's really frustrating and generally it produces some not good data, I would say.
Jodi:Yeah, I try to avoid them. But again, like, if you want to use the year, you just have to or you don't have to, but just take a look at it, make sure it's representing the field. Variability that you're trying to get at and that's one thing too is, like, when you're picking out imagery, what I'll typically start out with is looking at the NAIP images. So, NAIP images, those are going to be your really high resolution, one meter to 60 centimeter images. This will be like what FSA uses for their imagery.
Jodi:But what I'll do is I'll look through the NAIP imagery first and take a look at okay, what are the patterns that are coming across as I go back in time across this field? And then, when I look through the images in, like through Sentinel, like the satellite imagery at 10 meters, I'll try to make sure that, ok, are these images? Am I seeing the same patterns here? And, of course, like, you're not going to see the same patterns every single year, but it's a good indicator of like, ok, what am I looking for? Is something weird here? It just helps you kind of put a pattern in place when you're going through and picking images out of the server.
Sarah:And I think that's a really great place to start. You and I both know that going through different concepts about how to create zones is an entire podcast and into itself. But that's exactly how I look at my images too. What kind of trends and what kind of patterns am I seeing in here? So that's really good. But the resolution thing is really important.
Sarah:I've talked with a lot of different clients and customers and people about the value of resolution and what that actually means one pixel is covering on a field. So, for example, if we're talking about one meter resolution, you're talking about one pixel, one square from that image covering one meter of that field versus, for instance, third meter resolution means that that one pixel is going to be covering 30 meters in that field. So image resolution makes a big difference. The smaller the increment for the resolution, the more detail you pick up from the field. Now, that can be a good thing, and there's a lot of pressure that gets put on finding those really high resolution images and that can be valuable to a point, especially when it comes to zones.
Jodi:Yeah, cause I don't know about you, Sarah, but I don't want tracks in my zones and I don't want skips in my zones. Those do not represent actual variability in the field. That is just something that's man-made, and so I think we can sometimes get too detailed. So, for instance, if you've got imagery that's collected from like a drone, which can be extremely useful in doing a lot of different precision agricultural things, but what's? It can be a little bit too precise in that it's picking up on wheel tracks or picking up on some different even like, say, if you had a row unit that was a little bit more shallow than the rest of them. It can be picking up on the differences in between seven and a half inch rows, which isn't really useful in putting together zone maps that are trying to quantify and represent consistent variability.
Sarah:Exactly. I think you nailed it, Jodi. That is so true. That data for instance, if you were to see that row unit that you were talking about, where maybe you had one seeding depth, that can be very valuable data to a grower or somebody for certain situations to try to figure out how to fix that issue. But for the purposes of making zones, that's not helping us understand how that variability is actually occurring across the field. And I have worked with drone data before. It was a wheat field in 80, and I was trying to make zones out of it and instead of picking up the actual area of variability within the field, all I could pick up was the tracks during the in-season applications from the sprayer. That's not the actual variability that I was looking for. So there are lots of great places where drone imagery has a place, but for making zones, sometimes it can actually end up being too detailed. But for making zones, sometimes it can actually end up being too detailed.
Jodi:Yeah, so don't get hung up. If you don't have, you know, something like a very precise drone image. That does not mean that you can't make zones, it's going to be okay. There are other resources you can use which provide fantastic imagery that can help you make great zones that really represent your field.
Sarah:Also to the point of resolution. I think it's important for us to remember how big our application equipment is. You know we're dealing with application equipment out there. That's oftentimes, you know, 60 feet for a boom up to 120, 130 feet wide, and sometimes that application equipment gets split into two to three sections and that's as much area as we're going to be able to variable rate. So maybe we're going to be able to get down to 20 feet, maybe 30 feet, and yeah, I know there's some planters out there that are claiming individual rows, but when your prescription export only allows 10,000 polygons, you're not actually going to be getting that kind of detail.
Sarah:That's another podcast as well, but anyway. So when you have that amount of area to variable rate 20 to 30 feet having that image resolution at 10, 10 meters or 30 meters down to five meter resolution, that is detailed enough. And often, even when I'm working with NAEP imagery, which can actually have detail down to 60 centimeters, I really have to smooth that image out. I feel like my zones are better when I am blending that image with images that have a larger, a less detailed image resolution so it kind of has a smoother, a smoother look across the field. Absolutely At the end of the day.
Jodi:Images are a fantastic resource that contain really valuable vegetative data that is predictive of crop yield, and when we put carefully selected images that represent the field together, we can make awesome zones that we can then do variable radon that really represent that consistent variability in the field Well, with that, I hope all of your imagery helps you find your consistent variability Tune in next time for a tiny bite of knowledge from GK Technology, where we have a map and an app for that.