OFR21: Imagery Informed Fertigation with Jackson Stansell
In this video, graduate student Jackson Stansell shares about his work developing an automated platform that will allow users to make more informed nitrogen management decisions. The tool uses high-resolution imagery to assess corn nitrogen needs and determine if additional nitrogen fertilizer is needed. Producers are able to utilize the tool on a weekly basis to make more informed decisions about nitrogen fertigation.
icon search Searchable Transcript
Toggle between list and paragraph view.
[00:00:11.720]All right, well, I'm gonna go ahead and kick things off.
[00:00:14.450]My name is Jackson Stansell.
[00:00:16.010]I am a Master's student
[00:00:17.310]in Agricultural and Biological Systems Engineering
[00:00:20.180]at the University of Nebraska
[00:00:21.680]and I've been working on sensor-based fertigation
[00:00:23.960]over the past couple of years during my master's program,
[00:00:27.310]alongside Tyler Smith, Dr. Brian Krenke,
[00:00:30.490]Dr. Daran Rudnick and Dr. Joe Luck.
[00:00:33.100]Those are kind of the core members of our team.
[00:00:36.060]And the first thing I really wanna address
[00:00:38.070]as far as sensor-based fertigation management goes
[00:00:40.430]is what our motivation is for working
[00:00:43.300]on sensor-based fertigation.
[00:00:46.080]So, you know, as people who are here in Nebraska
[00:00:48.990]and just generally kind of in the great Plains region,
[00:00:51.250]we know that efficient nitrogen use is important
[00:00:53.360]for achieving acceptable water quality.
[00:00:55.940]This is something that is definitely a focus
[00:00:58.080]here in Nebraska,
[00:00:59.130]especially with our significant groundwater stores
[00:01:02.540]that we have available.
[00:01:04.560]We know also know that optimal nitrogen use
[00:01:06.720]results in higher profits.
[00:01:09.130]If we're actually matching what that optimal nitrogen rate
[00:01:11.890]should be in a field, then we're also going to achieve
[00:01:15.680]the appropriate yield and actually
[00:01:17.160]the economically optimal nitrogen rate,
[00:01:20.130]which is really what we're trying to aim for
[00:01:22.110]with using sensors to direct our fertigation management,
[00:01:26.450]coupled with kind of this efficient nitrogen use
[00:01:29.360]and kind of our groundwater quality issues.
[00:01:32.120]We know that there's a potential for regulation
[00:01:34.110]coming down from kind of local regulators,
[00:01:37.070]particular entities and we need methods and technology
[00:01:40.470]to mitigate the impact of these necessary regulations
[00:01:43.820]on productivity of corn
[00:01:46.120]to be able to retain profits, retain yields, where they are
[00:01:49.490]and maybe even improve those with enhanced management.
[00:01:53.160]So what we've seen in the literature
[00:01:55.790]is that fertigation inherently offers
[00:01:57.690]many nitrogen use efficiency benefits.
[00:02:00.090]It offers the opportunity to split nitrogen applications
[00:02:02.900]throughout the course of the growing season
[00:02:04.920]and also allows for very timely applications of nitrogen.
[00:02:09.760]Particularly once you get into the reproductive
[00:02:12.400]growth stages of corn,
[00:02:14.650]you can see some benefit to fertigation applications
[00:02:17.430]during those stages.
[00:02:19.230]We also know that prior research has demonstrated
[00:02:21.510]that using sensors to guide fertigation decisions
[00:02:24.120]may maximize nitrogen use efficiency benefits
[00:02:27.350]and that's exactly what we are seeking to build on
[00:02:29.990]with this research.
[00:02:32.200]So to get into the method that we have developed
[00:02:34.880]over the past few years,
[00:02:36.710]I'd like to go ahead and begin with what our first step is.
[00:02:39.240]And our first step is pre-seasoned data collection
[00:02:41.870]and site setup.
[00:02:43.200]So what this consists of is collecting yield history data
[00:02:46.370]for a particular field, getting all the different
[00:02:49.360]soil spatial variability data.
[00:02:51.350]So soil EC, elevation slope, as many layers as we can.
[00:02:56.670]And then also collecting some information
[00:02:58.420]about the irrigation system itself,
[00:03:00.880]it's operating parameters, as well as the fertigation pump
[00:03:04.450]that we're going to be using at that particular site.
[00:03:07.310]And so collectively that data,
[00:03:09.630]especially that soil spatial variability and yield data
[00:03:12.660]comes together and we're able to kind of characterize
[00:03:15.680]the spatial variability within that field
[00:03:19.240]and we make what's called an indicator block,
[00:03:21.740]establishing application with the prescription
[00:03:24.980]that is made according to all of that soil
[00:03:27.510]spatial variability data.
[00:03:29.290]And so essentially what this indicator block
[00:03:30.827]and establishing application is,
[00:03:32.860]is we're going out and putting rate blocks
[00:03:34.610]at very specific points in the field
[00:03:37.100]so that we can sample those with our imagery
[00:03:39.220]throughout the course of the growing season.
[00:03:41.360]And we do that by V6.
[00:03:42.950]So this application can be made
[00:03:45.250]with any sort of variable rate-capable ground equipment,
[00:03:47.900]whether it's a dual product planter,
[00:03:49.570]whether it's an anhydrous bar,
[00:03:51.370]or if you even wanna use the center pivot,
[00:03:53.550]we've shown this year that you can actually use
[00:03:55.610]a center pivot to put these in.
[00:03:58.250]But once we make that indicator block
[00:04:01.600]we get into what I call our weekly cycle
[00:04:04.400]over the course of the growing season.
[00:04:06.280]So this starts at V6 and each week we go out
[00:04:08.900]and we collect imagery,
[00:04:10.460]whether that be from satellite, drone or manned aircraft.
[00:04:15.410]We then process that imagery
[00:04:17.010]and this is particularly necessary for drone imagery,
[00:04:20.800]where you have a lot of small images taken
[00:04:23.820]as you cover a field.
[00:04:25.140]They have to be stitched together to make one cohesive image
[00:04:28.340]of the entire field.
[00:04:29.857]And so once we have that one cohesive image
[00:04:32.250]of the entire field that's multi-spectral
[00:04:35.400]and one allows us to produce the NDRE vegetation index,
[00:04:39.410]which is defined in the on-farm research book.
[00:04:42.820]We actually get into our analytics process
[00:04:45.460]and so we have an algorithm on the back end
[00:04:47.290]that is able to essentially assign sufficiency indices
[00:04:51.650]to each individual sector within a field.
[00:04:54.140]So this is looking at our research field setup
[00:04:56.700]where we have a lot of these field sub-regions.
[00:04:59.170]These are 15 degree sectors out in the field
[00:05:02.150]and so you can kind of see the management zones
[00:05:05.180]and for each one of those individual management zones
[00:05:07.890]and sector intersections,
[00:05:10.482]we can assign a nitrogen sufficiency index to that location.
[00:05:15.130]Then using that analytics layer,
[00:05:17.090]we can actually go in and make our fertigation decisions.
[00:05:20.060]So these are go or no-go fertigation decisions.
[00:05:23.040]Either this sector needs fertigation or does not.
[00:05:26.360]And then using those decisions,
[00:05:27.870]we're able to generate a fertigation prescription.
[00:05:30.720]And this fertigation prescription basically assigns
[00:05:33.190]30 pounds of nitrogen breaker to be applied
[00:05:35.669]to all those sectors that indicate
[00:05:37.610]that they need fertigation and zero to all the ones
[00:05:40.590]that say that they don't need fertigation.
[00:05:43.360]So, as you can see here,
[00:05:45.150]the low sufficiencies from our analytics layer
[00:05:48.350]translate to those sectors that are receiving fertigation.
[00:05:53.610]So once that fertigation prescription is made,
[00:05:56.210]we go out and actually make that fertigation application.
[00:05:59.700]So this process is completed weekly,
[00:06:01.580]delivering those go or no-go decisions
[00:06:04.030]until R3 is observed at the time of image capture.
[00:06:08.040]And I'm saying time of flight here
[00:06:09.360]because generally we're using drone imagery
[00:06:11.720]to make these decisions.
[00:06:14.840]So overall, the objective of this method
[00:06:16.570]is to match the corn nitrogen uptake curve
[00:06:19.870]using on-off nitrogen application control with fertigation.
[00:06:24.540]So here, each one of these stairsteps
[00:06:26.920]would indicate a fertigation application
[00:06:28.870]triggered by a sensor and as you can see,
[00:06:31.070]we're trying to just match that uptake curve
[00:06:34.180]for corn as best as possible, which should result
[00:06:36.880]in the optimal amount of nitrogen being applied.
[00:06:41.320]So far, I think we're seeing pretty good results here.
[00:06:44.470]So this is actually a graph
[00:06:46.060]showing different nitrogen uptake curves
[00:06:48.990]and the one I want you to focus on is this black line.
[00:06:51.740]This is basically translating if we went out and applied
[00:06:56.270]the UNL recommended N,
[00:06:58.620]if we were to say that that translated
[00:07:00.860]directly to crop nitrogen uptake
[00:07:02.650]over the course of the season,
[00:07:04.260]that's what that black line would be.
[00:07:06.130]And then our treatment is here in the blue
[00:07:08.410]and you kind of see those stairsteps,
[00:07:10.210]following that upswing in the curve
[00:07:12.740]and really matching, overall,
[00:07:14.880]the total rate that we need out there
[00:07:16.420]but with varying timing and amounts
[00:07:19.070]to match that uptake curve for the corn.
[00:07:24.560]So we've looked at this method that we've developed
[00:07:26.618]at 10 different sites over the past two years,
[00:07:29.360]five in 2019 and five again, here in 2020
[00:07:34.130]and most of our sites have been
[00:07:35.430]in the central part of the state,
[00:07:37.120]with a couple up in the Northeastern part of the state.
[00:07:39.920]And then also one at the Eastern Nebraska
[00:07:42.150]Research and Extension Center in the (audio cuts out).
[00:07:45.970]Generally the way that our treatments are laid out
[00:07:48.050]in these studies, we have 12 different sectors.
[00:07:52.420]Each one of these sectors are 15 degrees.
[00:07:56.160]We have four total reps out in the field
[00:07:58.180]and generally we have a grower management treatment
[00:08:00.770]and we have two sensor-based management treatments.
[00:08:05.840]Generally we'll call them risk averse and risk tolerant
[00:08:08.100]but I'm going to get into exactly
[00:08:10.300]what those treatments are here next.
[00:08:13.740]So we have three different approaches
[00:08:16.530]for sensor-based fertigation management
[00:08:19.452]that we have investigated as treatments
[00:08:21.980]over the past two years.
[00:08:24.320]We have our Risk-Tolerant Last 60 treatment,
[00:08:26.960]which is implemented for only the last 60 pounds
[00:08:29.490]of applied nitrogen out of a grower's total nitrogen goal.
[00:08:32.730]So if a grower had a total nitrogen goal of 180 pounds,
[00:08:36.610]then we would actually kick our method in
[00:08:38.670]at 120 pounds of applied nitrogen
[00:08:40.880]and use the sensors to manage that last 60 pounds, okay?
[00:08:44.810]The reason why we call this risk tolerance
[00:08:46.810]is because we're requiring more indicated deficiency,
[00:08:49.960]more indicated nitrogen deficiency
[00:08:52.240]to trigger a fertigation event, okay?
[00:08:54.810]And goal of doing that is to prioritize nitrogen savings
[00:08:58.670]over saving yield, okay?
[00:09:00.060]So we're really trying to maximize
[00:09:02.100]how much nitrogen we can save with this particular method.
[00:09:06.750]We also have the Risk-Averse Last 60 treatment,
[00:09:08.970]which similarly to the Risk-Tolerant Last 60
[00:09:12.240]is implemented for only the last 60 pounds
[00:09:14.250]of applied nitrogen
[00:09:15.910]but we require less indicated deficiency
[00:09:18.580]to trigger a fertigation event.
[00:09:20.270]So this means that we're going to apply nitrogen
[00:09:22.170]more frequently throughout the growing season.
[00:09:24.990]And the goal here
[00:09:25.830]is ultimately to prioritize our yield savings.
[00:09:28.350]We still wanna save nitrogen
[00:09:29.750]but we also wanna make sure that we're not losing
[00:09:31.820]that yield potential.
[00:09:33.930]Lastly, we have our Risk-Averse Full Season treatment,
[00:09:37.260]which is only different
[00:09:38.220]from this Risk-Averse Last 60 treatment
[00:09:40.620]in the length of time for which it is implemented.
[00:09:43.760]Okay, so we implement this treatment
[00:09:45.900]all the way from V6, or post indicator
[00:09:48.860]block establishment, onward.
[00:09:50.910]Okay, so this is V6 to R3,
[00:09:52.870]the entirety of the fertigation management
[00:09:54.900]for the season is being managed using sensors, okay?
[00:09:58.867]And the reason why we're wanting to do this
[00:10:00.540]for the full season is that we wanna optimize
[00:10:03.080]both nitrogen and yield savings at the same time.
[00:10:06.950]So in 2019, we looked at the Risk-Tolerant Last 60
[00:10:10.530]and Risk-Averse Last 60 treatments.
[00:10:13.280]And then in 2020, we looked at the Risk-Averse Last 60
[00:10:15.817]and the Risk-Averse Full Season.
[00:10:20.120]So to get into some aggregate results,
[00:10:21.770]I'm not gonna get too much
[00:10:23.310]into individual site results here
[00:10:25.360]but I'm happy to answer any questions on those.
[00:10:27.890]What we have graphed here is the difference
[00:10:30.470]between our sensor-based fertigation management treatments
[00:10:34.570]and our grower management treatments for efficiency,
[00:10:38.790]which is shown here on the x-axis
[00:10:41.060]and profitability, which is shown here on the y-axis.
[00:10:44.460]If we were both more profitable and more efficient
[00:10:47.036]than the grower in any particular implementation,
[00:10:50.010]we would be located in this upper right-hand quadrant here.
[00:10:53.240]If we were less profitable but more efficient,
[00:10:55.233]then we would be in this bottom right-hand quadrant here
[00:10:58.570]and so on and so forth, okay?
[00:11:00.747]And so what we can see from this graph
[00:11:03.060]is that the vast majority of our implementations
[00:11:06.510]with sensor-based fertigation management
[00:11:08.860]have been more efficient than typical grower management.
[00:11:12.410]What we also see is there's a lot less consistency here
[00:11:15.790]in terms of profitability, right?
[00:11:19.220]So overall, 94% of our implementations
[00:11:21.720]have been more efficient than typical grower management
[00:11:24.760]and actually, the only particular implementation
[00:11:27.620]that was not more efficient is this one over here,
[00:11:29.680]in the bottom-left hand corner.
[00:11:31.820]And we were actually 16 days late,
[00:11:34.040]just due to kind of a labor backup.
[00:11:37.310]This was during late June of 2019,
[00:11:39.430]so we've since corrected that issue
[00:11:41.140]using some software automation.
[00:11:44.300]But that is really the only site
[00:11:46.210]for which we have not been more efficient
[00:11:48.010]than typical grower management.
[00:11:49.840]On the profitability side,
[00:11:51.790]53% of our implementations have been more profitable
[00:11:54.550]while 47 have been less profitable
[00:11:57.240]than typical grower management.
[00:12:00.620]So on average, across all of our sites,
[00:12:02.960]taking those site-by-site differences
[00:12:05.090]with typical grower management,
[00:12:07.240]what we've seen is that our profitability
[00:12:10.880]for Risk-Averse Last 60 treatment
[00:12:12.970]is better than typical grower management by $1.14 per acre.
[00:12:18.290]Our risk tolerance is worse
[00:12:20.050]than typical growth management by only 81 cents per acre.
[00:12:23.330]And then with our Risk-Averse Full Season,
[00:12:25.180]which has only one year of data so far,
[00:12:27.630]we had a loss on average of $12.22 per acre.
[00:12:32.610]I'm gonna get back to that number in a minute
[00:12:34.010]because that is extremely concerning
[00:12:35.560]if you ever see this on a graph
[00:12:38.810]but we do see that we've increased efficiency
[00:12:41.040]with every single treatment on average versus the grower,
[00:12:43.490]which corresponds to what I've already been saying.
[00:12:47.130]Now, getting back to this Risk-Averse Full Season
[00:12:49.210]number here, what I wanna show you
[00:12:51.810]is that we had one particular site
[00:12:53.350]and this is Howard County.
[00:12:54.620]It's the last site in the book
[00:12:56.140]if you're looking at the section
[00:12:57.230]of sensor-based fertigation management studies
[00:13:00.220]that had these two points here
[00:13:01.890]that are significantly less profitable
[00:13:03.473]than the typical grower management on that field.
[00:13:07.570]And at this particular site,
[00:13:09.910]what I'm showing you here is our 2019 yield map.
[00:13:12.260]This is when we did not have any treatments on this field.
[00:13:15.140]And then I'm showing our 2020 yield map,
[00:13:16.940]which is when we did actually run this study
[00:13:18.850]on the south side of this particular field.
[00:13:22.260]What you can see is that there's a lot of consistency
[00:13:24.500]in terms of the yield patterns on this field,
[00:13:26.670]especially here in the southeastern corner
[00:13:30.210]of this field, where we have a large hill,
[00:13:32.450]it's very sandy, a lot of slope.
[00:13:35.628]And so it's just really
[00:13:37.380]a low yielding area in this field, okay?
[00:13:40.490]If we take this particular rep out,
[00:13:42.360]which I met with the grower
[00:13:44.090]on this particular study last week to review results
[00:13:46.720]And he said, "You know, you're really
[00:13:48.140]not gonna do anything on that particular hill,
[00:13:50.040]as far as yield goes."
[00:13:51.750]And so he was comfortable with at least looking at it
[00:13:53.890]without this rep in there.
[00:13:56.140]If you take this single rep out, not even this entire site,
[00:13:59.760]both of those points moved to being more profitable
[00:14:03.060]than typical grower management.
[00:14:05.810]And that actually flips our entire averages
[00:14:08.230]across all our sites for both of our risk averse treatments,
[00:14:13.320]both the Last 60 and the full season
[00:14:15.640]to being more profitable than typical grower management
[00:14:18.661]with the Risk-Averse Full Season
[00:14:20.500]coming in at $1.72 an acre more profitable.
[00:14:24.050]And you also see a fairly significant,
[00:14:27.030]I guess it's about a pound of grain per pound of nitrogen
[00:14:29.340]increase in terms of our actual efficiency
[00:14:32.940]from the sensor-based fertigation management treatments.
[00:14:36.590]And overall that would move us to 65% of our sites
[00:14:40.340]being more profitable than typical grower management.
[00:14:44.870]So the bottom line here,
[00:14:46.800]I know I've already been preaching on this a little bit.
[00:14:49.180]Sensor-based fertigation management
[00:14:50.640]does lead to increased nitrogen use efficiency.
[00:14:53.050]I'm very confident in saying that here today.
[00:14:55.870]Secondly, sensor-based fertigation
[00:14:57.994]can lead to increased profit
[00:14:59.700]but outcomes are pretty inconsistent.
[00:15:02.270]So I'm just gonna call it 53 to 65% of the time,
[00:15:06.180]whether you wanna leave that rep in there or not,
[00:15:08.830]we're between 53 and 65% of the time
[00:15:11.470]being more profitable than typical grower management
[00:15:13.540]and that's something that we need to improve.
[00:15:16.560]Full-season sensor-based management
[00:15:18.220]shows a lot of potential but with only one year of data,
[00:15:21.000]there's a lot more tuning and a lot more data
[00:15:23.140]that we need to collect
[00:15:24.270]in order to determine whether or not
[00:15:25.660]you can actually implement sensor management
[00:15:27.830]over the course of an entire growing season for fertigation.
[00:15:32.210]We have some ongoing work.
[00:15:33.390]We're looking to adjust our cutoff timing,
[00:15:35.670]so extending our fertigation applications
[00:15:37.810]past R2 into our R3 and R4.
[00:15:41.300]We're also looking at doing some rate adjustments
[00:15:43.090]by growth stage so when we have that rapid uptake,
[00:15:45.520]should we increase it to 40 pounds, 50 pounds, an acre
[00:15:48.790]in those sectors that need fertigation?
[00:15:50.980]And how should we adjust this for soil type?
[00:15:52.800]We've been doing a lot of studies on sandy fields
[00:15:56.870]but we kinda wanna look at how this would react
[00:15:59.630]on heavier textures and so we're looking
[00:16:02.320]at different soil types and then seeing
[00:16:04.030]if we need to modify the method in any way.
[00:16:07.270]Indicator block implementation alternatives.
[00:16:09.460]As I mentioned, you can put those in with the pivot.
[00:16:11.880]That's something we've shown this year,
[00:16:13.250]so we're gonna keep looking at ways
[00:16:15.660]to make that more effective.
[00:16:17.920]And we're also automating this process through software,
[00:16:21.650]which is my project right now
[00:16:23.770]and it's pretty exciting to see it all come together
[00:16:26.700]and hopefully that'll decrease the amount of time
[00:16:28.700]it takes to actually do this process,
[00:16:31.420]implement it on a week-by-week basis
[00:16:32.970]throughout the growing season.
[00:16:35.400]I just wanna thank our funders for this research real quick.
[00:16:40.230]And then my contact info is here.
[00:16:42.850]If anybody has any questions, we only have 30 seconds,
[00:16:45.170]but I'll do my best to answer them quickly.
Log in to post comments