 As Gary said, we're located in Southeast Nebraska, actually Richardson County, which is the most southeast county in the state, farm in a variety of topography from hills to bottoms and soil types from sand to heavy clay. The goal of our project was to accurately measure the remaining nitrogen needs of corn in season and produce a variable rate prescription, nitrogen prescription, that could be flown on with a plane to address those needs. There are multiple challenges in determining how much nitrogen to supply for each acre of corn. Overall growing conditions can vary from year to year, and I should mention we're dryland, so if it rains or doesn't rain, it's a very important factor in what kind of yields we can raise. In addition to the yield potential within the field, you can see things from the topography, soil types, fertility and rainfall, they're all going to play a factor in addition to what the weather is overall. The weather can also affect the nitrogen availability that's in the soil, how much of the nitrogen can be gained from organic matter mineralization, but we also have to be concerned about the amount of losses that can occur in wetter conditions and also due to our coarser textured soils. The red and orange dots on this map represent wells with high nitrate levels. For the environment and our own health, we need to be managing our nitrogen with the goal to not only to only apply what the crop will use. We felt by waiting until the initial application of pre-plant nitrogen was nearly depleted, the growing conditions that have occurred up to that point can be taken into account to help calculate the remaining nitrogen that will be needed. By waiting until that initial application is nearly depleted, the growing conditions can actually be experienced in season and factored in to calculate remaining nitrogen needed. These next six slides give a brief overview of our project. In dry land corn production there's two application choices, ground or air. One consistent advantage of the air over ground is no crop damage due to stand loss from the wheel tracks. Also if the field is too wet for ground applicators, a plane can get the job done. This project wouldn't have succeeded without a team effort. I've provided the experience for field operations. Nathan and Laura Thompson provided all the technical knowledge and skills to set up the plots, program and fly the drone, analyze the data and also write the prescriptions for the nitrogen application that will be made later. We used, I guess I should have mentioned, in this slide I'm applying various rates of nitrogen in the different strips that we set up and that will be explained shortly. We used a GPS auto steer to be able to plant on the anhydrous strips. This allowed us to make sure each seed had equal access to the nitrogen in each of the nitrogen rate strips. The entire plots were flown weekly to measure the nitrogen status in the crop. When the drone imagery started to show that the corn was running out of nitrogen, GPS reference prescription maps were generated and sent to the aerial applicator. Besides the cart scale weights of grain harvested out of the center passes of each of the strips, the combine also recorded yield data from the entire plot. Nate will now provide some in-depth look at the drone imagery that went into our project. The goal of this project was to use a multi-spectral sensor to, in effect, measure the amount of nitrogen in the corn plant. In the past, nitrogen, I don't know if some of the other people are with chlorophyll meters or a spad sensor or something like that, but in the past they were usually hand-held or mounted on some equipment that you would then drive through the field. Well, now with the advent of advanced technology, they've become a lot smaller and now you can mount them on things like drones or airplanes or things like that. So you avoid a lot of the problems like Dean was saying with driving through the field and creating crop damage or not having good conditions for things like that. And it's cheaper to fly the drone than it is to get a big piece of equipment through the field and it's easier, not invasive, more frequent. So what we're looking at is, like I said, we're directly measuring the nitrogen content of the corn plant and that's done sort of indirectly. We look at five different wave bands, but what we're really concerned with or the important one is the red band. The red band directly relates to chlorophyll content, which then you can relate to nitrogen content. We take the information we get from the red band and we make some indices. You've heard of them. NDVI is a really popular one. It's just a way of relating different indices into just an easy to understand ratio. The one we used, and I'll explain it a little later, but we used NDRE. And the reason we used NDRE is because the red edge area, and you can kind of see with this graph here. The red edge band is more responsive to slight changes in the nitrogen content of the corn plant. And what happens is, as the corn plant grows and it gets greener, the NDVI index begins to saturate. So this is an example of a comparison of two indices. On the top is NDVI, and the bottom is NDRE. This is the same flight at the exact same time, and you can tell that the NDVI has saturated where you can't see any differentiation. This is a nitrogen plot of different nitrogen rates, but you can see on the NDRE the differences start to show up. You can see the different nitrogen rates within the plot there. Yeah, red is less nitrogen, less chlorophyll, less red band reflectance. So in order to understand if a corn plant really is deficient in nitrogen, you need to have a reference point of a corn plant that is not deficient. So when we set up these studies or we do these prescriptions, we always have an area of the field that's a small area, but it's more or less over applied nitrogen. So we know that that corn there is fully sufficient in nitrogen, that way we have something to compare to. When we compare those, the fully sufficient corn to our target area, we create something called a sufficiency index. So this is an example. This is a field where you can tell that it's visually apparent, just with our naked eye, that corn there is deficient in nitrogen, whereas say the corn there on the left is fully sufficient, has all the nitrogen it needs. So what we do is we'll measure this all, we'll fly the drone over it, we'll get an index number for the sufficient plant, we'll get an index number for the deficient plant, and then we just make a ratio. We divide the insufficient plant by the sufficient, and then we come up with our sufficiency index. So that's kind of the key to creating our prescriptions and relating what we know is to be sufficient to our target area, which may or may not be. So when we first started this project, we had some choices to make as far as drones. There's lots there on the market. We went with a DJI Inspire II. It's actually built and intended for filmmaking. It's meant to be a videography drone, more or less. They make all kinds of fancy video cameras for it. We chose it because of its combination of price and features. At a time of about 25 minutes on average in perfect conditions, I'm going to fly up to 58 miles per hour, which it really isn't important for us. Important detail here is that it has some payload capacity at about a pound and a half, which is more than enough for our sensor. It can fly in winds up to 22 miles an hour, which means it can compensate, it can keep itself in the air, which is important because in late June, early July, in Nebraska, it can be windy sometimes, and we don't want to be kept out of the field just because it might be beautiful sunny skies, but if the wind's blowing, we want to still be able to fly. It's got its own built-in GPS that can navigate itself. It's got two batteries, and it costs about $3,000. The sensor, there's also many, many different choices out there on the market. The one we went with is made by a company called Micasense. The red-end sensor, as far as they're lined up, it's an older model now. These things are continuously changing and improving, but it's a five-band, multi-spectral sensor. It senses in the blue, green, red, red edge, and near and for red. It's pretty light, five ounces or so, small size, eight centimeters per pixel ground sample distance. What that means is that at 400-foot altitude, which is the FAA limit, each pixel on the picture that it takes represents eight centimeters on the ground, so about two and a half, three inches. It makes one capture per second, and it also includes, and this is important as well, includes a down-willing light sensor and its own external GPS, so it doesn't share any information with the drone, which has its own GPS. It has its own connected GPS system and a down-willing light sensor. The importance of the sensor, you want to be able to compare imagery across time, especially for a research project like this. We want to be able to take a picture on Monday and go back the next Monday and take more pictures and be able to directly compare those two, even though the first day it might have been overcast and the second day it may have been bright sunny, because again, this is measuring reflectance, and so reflectance is going to depend on how much light you have. This is correct for our light conditions, more or less, so this sensor comes with a down-willing light sensor that measures those ambient light conditions. Every picture it takes, it records also all the, just the amount of light that's coming in just from the sun and the sky or whatever, and before every flight, we have to calibrate it. I should have brought it as an example, but there's a picture here. We have a reflectance panel. It has a little square about this size, and it's painted a special color that's been accurately measured in a lab, and we know exactly what that reflectance is. So before we start a flight, we'll take a picture of that with the drone. We'll pick the drone up, and it's kind of cumbersome if you're doing it by yourself. You pick the drone up and you hold it over here, and then on your smartphone you can control the camera and take a picture, and then it's taking information from the light sensor and everything, and so then when we go in the process that we go in and we tell the computer what the values are of that panel, and it corrects all our imagery based on that panel and the information from the sensor. So we fly passes back and forth across the field, and it would be impossible to do that by hand, so we use software to control the drone and plan the missions and draw it. It's really simple. It's on an iPad or your phone or any smartphone or iPad can do it. This program, maybe you've heard of it, is called Drone Deploy, and it's pretty simple. It's pretty amazing really how just you plan out your mission and you tell it how high you want it to go, what your overlap is, which is another thing. We shoot for 75% overlap. It just helps, it stitches better, makes a complete, better, more complete map. So you can set all that information up and you hit a button and it just takes off and does its thing, flies the mission, and then comes back, all on its own. You don't have to do anything. So there are some legal considerations when you're getting into flying drones. It's all for commercial purposes like what we'd be doing. It's all under part 107 rules of FAA. If it's greater than .55 pounds, it has to be registered. Just like a plane. We have a registration number and we have to have it on the drone, just like a tail number on an airplane. Commercial use, you need a remote pilot certificate. It's kind of like a pilot's license. It's not quite as in-depth as like a manned aircraft but requires some study and some time. There's some restrictions. We have to stay below 400 foot above ground level. It has to be within line of sight at all times and you can't fly it at night and the line of sight is important. It's unaided line of sight, so you can't use a binoculars. These drones are able, I think, the max transmission range for the controllers, like four miles. Well, there's no way you can see this four miles away up in the air, so you have to be able to see it at all times with no binoculars or unaided sight. It's always important to check VFR maps, make sure you're not in any restricted air spaces or there's a website you can use to do that. Like I said, there's lots of mission planning apps and programs we use drone deploy. That's free and it's easy to use. Set for 400 foot, 75% overlap. Dream conditions, ideally less than 20 miles an hour. That's what the drone rated at. It probably could do better, but I wouldn't ever try it. Light conditions are important. It does have that light sensor is there to compensate for changing light conditions, but what it can't do is it can't handle partly cloudy conditions where you have big puffy clouds and shadows moving across the field. It can't quite compensate for that. You want uniform cloud conditions. You want either completely sunny or just some high clouds that are kind of opaque or just complete overcasts, but the big puffy clouds don't really work very well. Time of day, a couple hours of solar noon. It's not so much of an issue in the summer probably because there's quite a bit of time where the sun's fairly close to being directly overhead, but that's simply just a limit shadows, the effect of shadows. Even in a cornfield, you're going to get a shadow effect from row to row. You really notice that if there's any trees along the field, shadows will cast over the field and that will throw your imagery off. So you really want to be close to solar noon, limit the shadows, and then you've got the most directs straight on light. And this was a challenge this year in our county. Richardson County was awfully dry. If you waited too long into the day, the leaves would start to roll because the plant was conserving moisture. Well, you can't sense the leaves. You can't measure the chlorophyll content when the leaf's all rolled up. So you've got to be cognizant of that. So the final challenge is with flying, well, not the final one, but another one. There's quite a few. Data storage. So if you fly at four and foot, which is the most efficient altitude, 70% overlap, which is probably a little less than what we normally do. An acre field is going to have 13 and a half gigabytes of data. It's almost 6,000 images. So we were doing this like once a week for each one of these fields. And it added up. We have terabytes of data at home. So this is the output. This is what you get. Each row here going across is one capture. So if you notice, each row going across is one capture. So there's four captures here. And you can see, I don't know if you can see back where you are here on the screen, but you can tell the drone is moving this way. And so this is in time. This is the next one, next one, next one. And so this is the output you get. And they're all black and white because what they're doing is they're just measuring the reflectance for each given band. So the blue hasn't actually blue on the image that you get. It's just each pixel is a value representing the reflectance. When we take it into the computer, the computer then assigns blue to blue and green to green and red to red. And then you get a color image like what you would see if you were flying with the drone. But every image is geotagged. And then we stitch it together with image processing software. Software we used last year was Pix4D, which is specially designed and made for this. The year before, we used software that was made by Micasense. And this coming year, it'll probably be a different software because it's constantly changing it. And in fact, Pix4D changed right in the middle of the growing season last year. And we kind of scrambled and find an alternative. It's pretty computer intensive. It's about six minutes per acre. So on average. So there are some fields that take a couple hours. The computer sits there and checks through it. After we get stitched image, it'll stitch each band individually. And then we can use those to calculate our indices, NDRE or NDVI or whatever. So once we have our indices and our stitched images, we can then overlay those in GIS software. We use software called QGIS. It's free on the internet. It's kind of similar to ArcMap if anyone's ever used that. We can overlay all those imagery, you know, Calculator Index. And then we combine it with Calculator SI. We combine it with the base rate of N, well, organic matter, previous crop credits. So if you had soybeans before, we give ourselves nitrogen credit for that. And then we consider a yield goal. And that was all go into the equation that generated prescription. So first location, 2017, we had one. This was the plot layout. We had three base rate, four base rates, three that we were testing. We had two drone management rates of 75 pounds and 100 pounds base rate as in hydrosimonia. And then we had the traditional farmer management that was 160 pounds base. And then we also had two high-end reference blocks, 225 pounds. That went to the side of acreage. It was included on the acreage. This plot was about 90 acres. Each reference block, I think, was 300 foot long of one pass. Wide. 300 feet wide. Yeah, it was. OK. It's a small little, I don't know, a couple acres probably for each high reference block. 30 acres each for each treatment. We flew it about 11 times. We flew it quite a bit because what we're looking for is we were looking for that moment where the lower base rate strips started to show deficiency when compared to the references. We really weren't sure when exactly that was going to happen. So we flew it quite often. June 24th was apparently the day that, and you can tell, this is, these are the same flight, same data. This is an RGB view, so true color view. And this is the NDRU view. And you can see where the 75 pound rates really started to show up. Really a lot, yellow or red, meaning lower nitrogen content. So what was happening here is these base rates were starting to run out of nitrogen. The corn was showing deficiency. That's something we could see at the NDRU and we couldn't see in the true color. That was, it's a combination of the red, green and blue bands. So it's a true color. It's, yeah, here. This is just an actual plain old photo from just a digital camera from the same, you know, probably within a couple days. I don't know if it was the same day compared to NDRU. So you can see, I mean, what your eyes see, it all looks the same. But the index is showing the deficiency. So we took the imagery and then all those factors I mentioned before, like organic matter and yield goal and end credits and everything. This is the prescription we came up with. It was a variable rate prescription for an airplane. It was, yeah, 200 foot long blocks because when the airplane flies down the path, that's how far they, they can change the rate every one second. They go about 200 foot in one second. So that limited our block size. So the, the variable rate prescription ranged from 60 to 120 pounds of urea, of N as urea. The farmer rate, the farmer decided he was going to put on extra 40 pounds on his rate, on his, on his managed strips. So 24th is the imagery we used to apply. And then the 14th was a couple weeks later after application. I think it was like two days, three, the 28th, so a couple days later we applied. A couple weeks later we flew it again to see what happened. And you can see that everything was more uniform. It appears that the lower rate strips caught up, so to speak. Everything's more uniform and homogenous. So there is apparently a response to the fertilizer reapplied with the plane. We did, I don't remember what it was, I don't think we recorded it. We recorded that in 18. We tried, we tried to time the application to go right before that rain. So it's not just sitting there and losing it off. We wanted to get watered in. The urea that did go on was, was had a stabilizer on it. So we did have a little more time to hopefully get a rain. Yeah, it was coded or, yeah, yeah. Okay, so this is a, this is a statistical analysis of the NDRE response. So you can see these are grass representing, oops, representing the reflectance on the 24th of June. These letters here represent that there's a statistical, statistical difference between the three treatments. This is a graph of the values on the 14th of July. This gray line here is the reference box, if you're wondering. And then again on 14th of September. So you can see they, overall, they decline. They're probably the greenest right before Tassel late June. And then as the crop matures and dries down, they obviously decline. But they stayed pretty uniform for the rest of the year. Here's a table showing their base rates, average in season rate from the prescription of an airplane, and then the total N rates for all three treatments. This is the yield. The yield was the same across all three treatments. So then since we used less nitrogen on the drone treatments, there was obviously some efficiency gain and pounds of N per bushels of grain produced. And these are really good, pretty good efficiency numbers. And then a partial product or a partial profit calculation. And they're all fairly similar in profitability. It is more expensive to apply urea with a variable rate airplane than it is to put it on as a hydrous. So that's why even though we saved about 25 pounds on average, the profitability calculations came out pretty much even. Dean's going to go through our two 2018 here fields. In 2018, we continued the project on two new fields. Bay-Oyesal, a neighboring farmer, had agreed to be part of the grant project and provided us with a field for testing. We reduced our drone management pre-plant rates from two to one. We chose to go with a hundred pound base rate for the drone managed fields, since it carried us further into the growing season and allowed us to take into account more of the growth factors that occurred up until that point. This north field site was on gently sloping upland. We did frequent flights through June as we had the year before to watch for signs for the first signs of any nitrogen shortage before any permanent yield reduction could occur. Because of dryers soil moisture profile and lower than normal rainfall in 18, the nitrogen level differences between the strips weren't as obvious as in 17. On June 27th, after flying the field with the drone, we'd made prescription maps and sent them to the aerial applicator. Because of the uniformity of the drone imagery for the field, we were able to have the plane apply a flat rate of 25 pounds of nitrogen per acre. Also note the rainfall that we received on the 30th of June. Here's a comparison of the drone imagery just before fertilizing and then a week and a half later. The improvement is subtle compared to the 2017 imagery. The drone likely made the drought, likely made the moisture a bigger yield limiting factor than the fertilizer. And to summarize this field, we started with the 100 pounds for the drone based and then applied 25 pounds for a total 125. The farmer managed strips had 160 pounds applied replant. The yields came out very close to the same, insignificantly different statistically. The thing that was statistically different was the improvement in efficiency of nitrogen on the drone strips. To get the NUE number, you basically just divide the pounds of nitrogen used by the eight bushels of corn you harvest. The profit per acre was slightly lower for the drone strips, although it wasn't statistically different. This is due to the fact that this in 18, the farmer strips didn't get any additional nitrogen as they had in 17. This is the layout for the south site. It was a bottom field but experienced less rainfall than the north site. Using the same two replant N rates. Well, excuse me, I should say the drone rate was still the 100 pounds. The farmer rate in this case was 180 pounds base rate. We use the same flight times for this field also. As we noted earlier, there's only minor differences seen in the drone versus farmer strips in 2018. Again, because of the uniformity of the drone imagery for the field, we were able to have the plane apply a flat rate. In this case, the calculations indicated that we could get by with 53 pounds of nitrogen on the drone strips. And also note that the rainfall following this flight was in this location was considerably less. Here's the imagery before and after fertilizing, showing only minor differences. Something to note, we continued flying the field on a slightly less frequent basis, but later in the summer, a little over a month later, on August 9th, here's an imagery that we captured. I think this really starts to show that under drought conditions, the different soil types and water holding capacities of this field had significant impact on the plant health of the crop. The accuracy of the drone imagery is very impressive, especially when you compare it to the combine yield monitor data that we gathered in October harvesting the crop. To summarize this field, the drone imagery strips received 153 pounds of nitrogen, the farmer strips 180. The yields came out basically the same, within a fraction of the bustle of each other. Again, we saw slight improvement in the efficiency of nitrogen in the drone strips versus the farmer strips. And this resulted in a insignificant but slightly lower profit because the farmer strips weren't side dressed due to the drought. The cost for materials and application are listed below. This is a picture from one of two field days that we held in 17 and 18. A later slide that will come up will list the web link to a YouTube video of the entire field day. At that point, I think I'll let Nate summarize what we've seen over this two-year project. So in summary, we saved nitrogen for all three sites, averaged 30 pounds an acre, 25 to 35 pounds. No yield loss in any site or year. In all years, the drone plus the sensor approach produced greater nitrogen efficiency. So we used less pounds of end to produce a single bushel of grain. The profitability is fairly similar. Like Dean said, in 2018, it was affected by the fact that the farmer chose not to apply side dress. If we weren't in a research situation where we were testing a method, we probably would have made the decision to not to apply and the drone managed strips either. But we really wanted to test the response to the fertilizer based on what the drone was saying. We really wanted to see what was going to happen. So we did it anyways. If we were really in a more practical real life situation, we probably wouldn't have applied. And then the drone imagery in that case probably would have been used to confirm the farmer's intuition that it was too dry, that we hadn't really lost any nitrogen. So there was no need to apply any more end. Yeah, low rainfall in those two in both years, actually. One of the big challenges was timing that nitrogen application to coincide with rainfall. Because if you fly it on and it doesn't rain for a week, well, you've lost a lot of your end. And that's going to affect your efficiency. So we would like to really see how this works in a supposed wet or normal year. Some future ideas we're thinking about is this would probably work fairly well in a center pivot environment where you have the ability to furtigate. So you don't have to worry about timing your nitrogen correctly with rainfall. And even going beyond a drone, but even just putting the sensor on the pivot itself is something you can do. So if we're non-arrogated, we're looking at obviously lowering the base rate, the starting rate, so we can better time our applications for when the plant needs it or when we think we're more likely to get rain. But also the other option is you could use a higher base rate closer to what you think you may need and then monitor with the drone to see if you have any loss events that require maybe some spot applications or something like that. So this is something I think we're going to be trying next year is do it on more of a contour area, having different high rate blocks throughout the field will fly and just observe maybe not necessarily use it to apply nitrogen, but just observe how things change in the compare yield and try to find out where the optimum nitrogen rates are going to be. And possibly in addition to high rates, also ask some small low rate blocks so you get an early warning that you're approaching the end of the nitrogen available. Right. Using those low areas is kind of like a litmus test for timing your nitrogen application. Any questions? He was asking if we took into account the cost of the labor and time and materials of flying the drone into our profitability calculations and we did not. We just don't have a good handle on exactly how much that would cost. I mean, we were flying it so much more often than what you would in, I guess, a real life application. Estimates are probably $2 to $5 an acre, is probably what it would cost. But the problem with a lot of our time on it is similar whether you're flying a 40 acre field or you're flying a 100 acre field. The cost of it is really kind of the same, so it's kind of hard to put a per acre cost on it. In perfect conditions, you can do 80 acres with this setup. Now, I mean, there's other drones out there that will, like a fixed wing will probably, he was asking how much time we'd get out or how many acres we'd get out of the flight. So it's about 80 acres in perfect conditions. Other setups can do longer. So, OK, you're asking if you can use prior years data to kind of predict when, yeah. Yeah, well, the thing about this is what we're kind of doing here is we're, in our nitrogen equation, we're taking into account the weather. So that crop is going to show deficiency at different times, mostly based on the weather. Right, and that's it. Yeah, I mean, we thought about that, but we're really not sure how we would approach how to exactly go about that. It's been thought of. Yeah, yeah. And actually, with this here, kind of one of the goals of this is kind of what you said, you know, there's been a thought where you do something called a nitrogen ramp, where you apply different base rapes, and then you compare it to yield, and then you chart it, and then you can find where that graph kind of tops out and determine your optimum or most efficient end rate. So I don't know if that's kind of what you're, yeah. Yeah, it's a good, it's a cool idea. I mean, that would be more requires. Yeah. Yeah, in 2018, we had the plane fly flat rates because the imagery didn't show enough difference to justify doing a variable rate. In 2017, they did fly a variable rate, and it ranged from what? 60 to 120. 120 pounds, so it varied a lot. The imagery suggested that we could efficiently place nitrogen at those different rates along the strips. It was almost a quarter mile long strips in the 2017 project.