Advancing Digital Agriculture for Improving Agronomic Decisions

Ignacio Ciampitti, Professor, Farming Systems, Deparment of Agronomy, Kansas State University Author

05/22/2024 Added

6 Plays

Description

This presentation will focus on discussing the bottlenecks for translating science into actionable management and topics related to digital agriculture, and use of new technologies such as satellite imagery and new data visualization to assist producers on developing relevant agronomic decisions.

Searchable Transcript

Search:

[00:00:00.750]The following presentation is part
[00:00:02.670]of the Agronomy and Horticulture seminar series
[00:00:05.790]at the University of Nebraska-Lincoln.
[00:00:08.310]Welcome, everyone.
[00:00:10.710]Thanks for being today at this special seminar
[00:00:15.090]with our guest speaker, Dr. Ignacio Ciampitti.
[00:00:18.810]He's a professor in farming systems
[00:00:21.510]from Kansas State University,
[00:00:23.100]and he's also the director
[00:00:25.110]of the Digital Agriculture and Advanced Analytic Institute
[00:00:29.070]at K-State.
[00:00:30.270]Ignacio has a, here his master's and bachelor's
[00:00:34.140]in University of Buenos Aires,
[00:00:36.120]and then his PhD at Purdue University.
[00:00:38.524]His lab focuses on using field, statistics, remote sensing,
[00:00:43.770]and modeling research to try
[00:00:45.630]to select best management practices
[00:00:48.060]to improve farming systems.
[00:00:49.380]He has an extensive records of publications
[00:00:52.170]that includes more than 150 peer-reviewed articles
[00:00:55.290]in the last five years.
[00:00:57.750]He's also providing leadership in different societies.
[00:01:02.880]He's member of the board of,
[00:01:04.860]of members of the Crop Science side of America.
[00:01:07.500]He serves also as an associate editor
[00:01:09.810]in different journals like Crop Science or in the board of,
[00:01:13.945]editor of board of other journals
[00:01:15.360]like European Journal of Agronomy.
[00:01:16.860]I would like to mention also
[00:01:18.000]that I was his first PhD student,
[00:01:21.090]but after me, like 20 grad students came after,
[00:01:24.360]and they graduated with him.
[00:01:25.560]He has a great team of,
[00:01:27.390]I think almost 14 grad students today.
[00:01:30.660]And in the last 10 years,
[00:01:31.650]I think he also providing mentoring
[00:01:34.740]for more than 120 visiting scholars in his team
[00:01:37.290]that is, helped him to do all this research.
[00:01:39.960]And I think at some point,
[00:01:41.160]part of my research when I was in his program,
[00:01:43.500]so I think, we invited him in this, you know,
[00:01:46.500]this thinking about the digital act topic
[00:01:50.310]within the seminars this semester.
[00:01:52.500]So we have some topics related on that topic.
[00:01:55.860]And we thought that Ignacio was a great additional.
[00:01:58.320]Thank you, Ignacio, for accepting the invitation right away
[00:02:01.500]and to be present in the seminar.
[00:02:03.150]We know that we have a complex agenda,
[00:02:05.850]but it was, we're happy to have you here.
[00:02:07.470]So with that,
[00:02:08.580]we're going to start and for those in the audience
[00:02:11.910]and the people online who will have questions at the end,
[00:02:14.760]so just keep that in mind.
[00:02:17.820]Thank you.
[00:02:18.690]Okay, thanks, Ashamo, for the invitation.
[00:02:23.040]It's a pleasure to be here.
[00:02:24.500]So I usually like to move,
[00:02:25.740]so I will try not to move too much.
[00:02:28.920]Trying to keep on this space.
[00:02:30.210]So, and as Sharma was saying I think that one of the things
[00:02:33.660]that we tend, I mean,
[00:02:34.800]and I will enjoy the most beyond doing a lot
[00:02:37.800]of bureaucratic paperwork and administration on my work,
[00:02:44.340]is also specifically mentoring grad students.
[00:02:46.680]And then that is reflected in all the work
[00:02:49.620]that I will be showing a little bit today.
[00:02:51.870]So most of the work that we will be showing is just part
[00:02:55.140]of grad students' projects.
[00:02:56.370]So I mean, one of the first questions
[00:02:59.070]when we look at this topic and we're still kind of exploring
[00:03:03.420]what does it mean digital ag,
[00:03:04.860]and trying to understand how this idea
[00:03:07.680]of involving data decisions can help farmers, you know?
[00:03:11.820]And then the way that I'm thinking in many situations
[00:03:14.550]is the best way to show some of the complexities
[00:03:17.880]to talk about a few topics.
[00:03:19.410]So, and I divided a little bit on trying
[00:03:21.990]to show some concepts on this idea of moving
[00:03:25.683]from traditional agronomy to something
[00:03:27.690]that is more quantitative,
[00:03:30.870]discussing a little bit about this,
[00:03:34.469]attaching uncertainty to estimations.
[00:03:36.330]So, I mean,
[00:03:37.639]thinking about how we are calculating these uncertainties.
[00:03:40.440]I mean, I always call this as being more honest
[00:03:44.370]on our research when we discuss with farmers too.
[00:03:48.030]And then I will show a little bit about a few examples
[00:03:50.280]on these, how it's connecting digital ag,
[00:03:53.490]and then a little bit of what the world
[00:03:55.350]that we're doing today on looking at more this data science,
[00:03:59.010]developing new open databases and really moving from paper
[00:04:03.000]to developing interacting tools.
[00:04:06.420]I mean the research papers are always a good accomplishment,
[00:04:10.230]but they finish on research papers
[00:04:13.020]unless they are being transformed into something
[00:04:15.000]that people can use, so.
[00:04:17.520]I will just present very briefly some
[00:04:19.470]of these case studies.
[00:04:21.570]I promise I will be on time or close. (chuckles)
[00:04:26.809]And let me start with something
[00:04:27.642]that I usually like to start in many situations
[00:04:29.310]so we can get this more up to personal level.
[00:04:31.590]So I usually put this slide in many situations,
[00:04:34.620]even if I do a research presentation
[00:04:36.390]because it's what we do in the lab.
[00:04:38.820]I mean, how we work with people.
[00:04:42.240]We work at personal level,
[00:04:43.560]we work with people in multiple ways.
[00:04:46.500]I truly believe on how we are mentoring students
[00:04:49.290]and open door policies all the time.
[00:04:52.500]We increase communication.
[00:04:54.270]We think that we communicate and sometimes we think
[00:04:56.850]that we don't communicate too much.
[00:04:58.650]So we find ways of communicating
[00:05:00.720]and trying to get students to be leaders, okay?
[00:05:04.800]Empowering students and then the other thing
[00:05:07.320]that we have discussions even with some
[00:05:08.970]of the professors this morning is this idea
[00:05:11.280]of how we move our research quickly
[00:05:14.010]into kind of a presenting,
[00:05:16.350]I mean solutions for farmers and for people every day.
[00:05:19.620]So, and I can tell you that many of the research ideas
[00:05:23.370]that I'm presenting today, many of them on those are not,
[00:05:26.670]I will not take the credit, many times that are coming
[00:05:28.980]or by discussing with the students
[00:05:30.570]or by discussing with farmers
[00:05:32.160]or by discussing with policymakers.
[00:05:35.640]And I would say those are the best ideas usually
[00:05:38.400]'cause they're coming by by kind of a team collaboration.
[00:05:40.830]So, and as Ashamo was saying, and I say this all the time,
[00:05:44.700]so last 11 years on our lab, we just,
[00:05:49.350]I think that we finished 11 PhD students.
[00:05:52.620]We're in track to finish two more this year.
[00:05:55.110]So we're gonna be 13 and we have seven masters,
[00:05:58.740]and we are on track to finish,
[00:06:00.592]and I have three masters this year.
[00:06:01.425]So, I mean close to more than 20 students
[00:06:05.310]in the last 11 years, so.
[00:06:08.545]And it's because of this concept.
[00:06:09.378]I mean we believe that many of our students,
[00:06:12.210]I feel that five or six,
[00:06:14.250]I mean many of them they finish in academia,
[00:06:17.310]which is very nice to see people working in universities
[00:06:22.410]or universities and that is probably the best
[00:06:25.740]when someone is asking what is your impact about,
[00:06:28.950]I usually say I don't have a really good number
[00:06:32.100]on how many farmers are we attracting,
[00:06:34.530]are they using what we do,
[00:06:35.790]but I can tell you something that I have good numbers
[00:06:39.060]and it's about, I know how many faculty or students,
[00:06:42.810]I mean on my team, we have a professor in Georgia,
[00:06:45.540]we have professors in Missouri, Nebraska, Wealth,
[00:06:49.920]Ohio, Spain, Australia, and to a K-State now.
[00:06:57.960]So, and that is an impact.
[00:07:00.060]I mean, that is a true impact.
[00:07:01.440]We can, I can clearly see that.
[00:07:03.090]So, and I say I can answer it that way very easily.
[00:07:08.370]And I always say my work is well done when I finish
[00:07:11.400]and I see that my students, they know much more.
[00:07:14.970]And it happens all the time, what I know,
[00:07:17.190]and it happens many times, I mean,
[00:07:18.600]you see your students just going well above and beyond, so.
[00:07:22.800]Moving away from this,
[00:07:23.820]a little bit of that personal connection,
[00:07:26.340]let's discuss a little bit about this,
[00:07:27.840]some of these concepts so we can start framing the idea
[00:07:30.630]of this digital ag, is simple and not that simple.
[00:07:37.320]When you think about digital ag
[00:07:38.670]or when you think about this,
[00:07:39.900]what happens in agronomy in the last a hundred years,
[00:07:42.900]it happens as an example that,
[00:07:46.920]many times we are working on traditional agronomy
[00:07:50.246]who are doing a lot of multidisciplinary work.
[00:07:52.470]We're working with people in ag econ,
[00:07:55.290]putting numbers to our research.
[00:07:56.730]We're working with people in entomology or plant disease
[00:08:00.510]or microbiology, but in many situations,
[00:08:03.167]we are not really all the time,
[00:08:04.860]sometimes connecting with people in geography
[00:08:07.200]or people in computer science, so.
[00:08:10.140]And this is one of the things
[00:08:11.130]that we start thinking about how we can move
[00:08:13.290]to connecting different disciplines
[00:08:16.560]and start helping farmers to get to the decisions.
[00:08:20.520]I mean using data and more data, okay?
[00:08:24.390]And this, when we kind of discussed this concept
[00:08:26.850]of moving from this traditional agronomy
[00:08:29.550]and you see that I emphasize all the time,
[00:08:31.650]I wanted to make sure that agronomy is on the center.
[00:08:35.510]We have many situations.
[00:08:36.343]I mean, it could be in computer science or it could be
[00:08:38.730]from GIS or it could be from other disciplines
[00:08:41.370]that they're leading projects.
[00:08:43.620]But at the end of the day,
[00:08:44.490]they don't understand what they're doing.
[00:08:47.199]And mainly because sometimes,
[00:08:48.032]they don't have the foundational agronomy
[00:08:50.160]on crop of physiology or soil science
[00:08:53.490]and then or grasslands.
[00:08:55.710]And then unless you understand a little bit
[00:08:57.530]of the mechanistic, you can connect and correlate NDVI
[00:09:01.590]with the number of cows in a field.
[00:09:04.920]I mean, and you will find a correlation.
[00:09:06.600]I'm telling you that, there is a paper on that.
[00:09:10.999]And you can correlate NDVI with whatever you want.
[00:09:13.500]The question is what are the mechanistic,
[00:09:15.480]what is the process behind that?
[00:09:17.465]I mean, when you look at these indices coming
[00:09:20.310]from satellite,
[00:09:21.143]they're connecting to intercepting radiation.
[00:09:23.460]So they have a connection with how much radiation
[00:09:26.100]is being captured by the crop.
[00:09:27.780]So it's a process behind,
[00:09:28.980]it's not just a kind of a black box idea, so.
[00:09:33.900]And then we start thinking about, okay,
[00:09:35.700]one example when you think about moving
[00:09:38.010]into more data intense analysis, we do studies,
[00:09:42.510]we collect data,
[00:09:43.770]which is the traditional approach if you see here,
[00:09:46.410]and sometimes based on this data,
[00:09:48.090]which make recommendations.
[00:09:50.400]Okay, so,
[00:09:51.870]and then what about if we move a little bit
[00:09:54.240]from there using the same data that we collected,
[00:09:56.820]but now we are doing a little more dynamics.
[00:09:58.830]So we are collecting data from fields
[00:10:01.830]and now we start looking at forecast,
[00:10:04.590]water content before planting,
[00:10:06.300]while looking at water forecast
[00:10:07.650]from the next coming weeks.
[00:10:09.690]And we start building this probabilistic analysis,
[00:10:12.480]which basically is what is the preference of the farmers,
[00:10:16.080]in this case, I put plant density for corn as an example.
[00:10:19.980]And then we start building models that they look complex,
[00:10:23.490]but they might look complex,
[00:10:26.250]but at the end of the day, they're not that complex.
[00:10:28.890]They basically use information coming from yield trials,
[00:10:33.060]from USDA to establish what is the optimal potential
[00:10:35.910]from that section.
[00:10:37.170]And then you start collecting data
[00:10:38.580]on whether water holding capacity from the field.
[00:10:41.790]And overall what we start doing
[00:10:43.530]is developing probabilistic models.
[00:10:45.480]What is the probability that in this field,
[00:10:47.460]on this location, if I plant in this density,
[00:10:50.250]I have a response, okay?
[00:10:54.043]And we have an example.
[00:10:54.876]So this is one of the studies that we published
[00:10:56.430]that we did a probabilistic model analysis
[00:10:58.470]on optimal plant density for corn,
[00:11:01.050]of course consider the precipitation
[00:11:02.970]and the temperature during the season.
[00:11:05.070]I mean, I just highlighted two years
[00:11:07.508]that they were very different,
[00:11:08.341]2012, quite drought and rowdy.
[00:11:10.560]And then if you look at 2018,
[00:11:13.140]very different in terms of wet.
[00:11:16.050]And what we start plotting and start getting
[00:11:18.630]is this response functions.
[00:11:21.030]I mean, and then we start basically plotting
[00:11:23.370]what is the probability of optimal plant density,
[00:11:25.980]in this case, in a very dry year
[00:11:27.453]that we are using less number of plants.
[00:11:31.795]And this was basically the estimation from those counties.
[00:11:34.080]And when you move to (indistinct) years,
[00:11:36.120]then you start seeing how that probability
[00:11:38.730]for response is changed.
[00:11:41.550]And it's nothing new,
[00:11:42.383]it's just using the data we're collecting from the field
[00:11:45.180]and it's integrating the data
[00:11:46.590]and putting in probabilistic models, okay?
[00:11:48.930]And getting to the farmers the idea
[00:11:50.520]that they are the ones taking the decision,
[00:11:52.740]but every decision that I'm taking
[00:11:54.780]has a probability attached, okay?
[00:12:00.180]This is one of my favorite topics
[00:12:01.560]because it's probably, I mean, I don't know.
[00:12:05.280]I mean, it's probably something that I like, we like to,
[00:12:07.920]I mean, we have been doing this studies on action in corn
[00:12:10.500]since my PhD have helped from many people on taking this.
[00:12:16.740]And I'm fascinated that we have been doing this
[00:12:18.870]for probably the last a hundred years and maybe we're,
[00:12:22.410]I always say that we probably are not even close
[00:12:24.330]to solve this, sorry.
[00:12:29.550]And why we're not that close
[00:12:30.810]is because I mean this response,
[00:12:33.000]like if you think about this case study yield response
[00:12:35.850]in maize, I mean nitrogen rates
[00:12:38.610]and there are a couple of factors.
[00:12:39.630]Now, regardless of what model you want to use,
[00:12:42.300]we can discuss many models, quality plateau,
[00:12:45.090]linear power law.
[00:12:48.210]In this case, there are two concepts or two problems here.
[00:12:51.150]One is this one which is how much nitrogen is coming
[00:12:54.120]from the soil because that is the intercept,
[00:12:57.360]is the yield for that crop when we are not applying
[00:13:00.240]in action.
[00:13:01.200]And the challenge is also this one,
[00:13:03.390]what is our optimal end rate, okay?
[00:13:06.180]And the challenge is what is our optimal end rate?
[00:13:08.370]Guess what?
[00:13:09.881]Depends on these two, okay?
[00:13:11.880]Because if we don't know this, then it's hard to know this,
[00:13:15.600]okay?
[00:13:16.680]And there's a correlation.
[00:13:17.670]So, and we started doing some studies
[00:13:20.370]and we have based many studies.
[00:13:22.350]I mean, we collected a database from thousands of studies
[00:13:26.700]across entire US and we plot many functions.
[00:13:30.780]I mean, and we can discuss again another day
[00:13:33.000]about functions.
[00:13:35.070]But the honest approach is that when we look at,
[00:13:38.896]this was for example, one study,
[00:13:41.790]we are very close to Topeka in Kansas,
[00:13:44.550]20 years of data is a corns in rotation.
[00:13:48.150]So 20 years of data, right?
[00:13:49.890]Corn in rotation and you say, okay,
[00:13:53.550]I have 20 years of data on what is the optimal end rate
[00:13:57.330]for corn.
[00:13:58.530]I already solved the problem.
[00:14:02.310]So the optimal end rate is 146, okay?
[00:14:06.810]So when you go to a farmer
[00:14:08.070]and the farmer asking me a question, I say, don't worry,
[00:14:10.710]your optimal end rate is 146,
[00:14:13.050]however already 20 years of data, I can tell you that.
[00:14:16.770]What I didn't mention to that farmer is that, well,
[00:14:19.380]it depends, which is the best that we use as an agronomist.
[00:14:24.567]Now, it depends is coming all the time.
[00:14:26.010]It means,
[00:14:27.720]it depends means that depending on your weather conditions
[00:14:30.630]and depending on early season, depending on thermalization,
[00:14:35.899]depending on what is the weather around flowering,
[00:14:38.280]if you have heat, drought, well,
[00:14:40.470]sometimes you might get, your optimal is a hundred pounds
[00:14:44.010]because it saturates very fast,
[00:14:45.420]and sometimes you might need 200 pounds, okay?
[00:14:50.430]And this a challenge, you know,
[00:14:51.480]because at the end of the day,
[00:14:52.830]we haven't done too much studies on trying to estimate
[00:14:56.130]and calculate uncertainties.
[00:14:58.980]So what is the uncertainty attached to the estimate?
[00:15:01.890]If I tell you your optimal end rate is 146,
[00:15:05.130]what is the uncertainty attached to that?
[00:15:08.040]I mean, in this cases, for example,
[00:15:09.990]we know that uncertainty attached
[00:15:11.580]to developing an optimal end rate was 50 pounds, okay?
[00:15:16.830]Could be less 50, plus 50.
[00:15:18.870]And we know that in the cases that we have more data,
[00:15:21.330]in many situations we are usually between 30 to 50 pounds.
[00:15:25.170]So when a farmer is, and when I'm telling a farmer
[00:15:28.110]that your optimal end rate is 120,
[00:15:30.360]I know that the optimal end rate could be 90
[00:15:32.700]or it could be 150, okay?
[00:15:37.560]And one of the issues
[00:15:38.996]is that when we start measuring uncertainty,
[00:15:40.650]I'm not talking about the value
[00:15:42.027]and the estimate uncertainty.
[00:15:44.070]The problem with uncertainty is that depends on the weather.
[00:15:48.990]So the management contributes only 1%
[00:15:52.650]to explain that uncertainty.
[00:15:56.100]81% of the uncertainty is being explained
[00:15:58.410]by the weather conditions
[00:15:59.610]and many times the weather conditions around flowering.
[00:16:02.340]So that month around flowering time for corn, precipitation,
[00:16:06.840]solar radiation, temperature, is basically what defines
[00:16:10.860]if that crop is gonna be a high yielding crop,
[00:16:14.310]a medium yielding crop,
[00:16:15.450]or is gonna finish with losing 50% of the yields.
[00:16:19.140]Because if I have two or three weeks without rain, without,
[00:16:21.960]I mean high temperature, hot stress,
[00:16:24.030]I mean then most likely I'm gonna lose
[00:16:26.670]and apply more nitrogen can compromise
[00:16:28.680]and can present as an stress situation, okay?
[00:16:32.790]But we start bringing more and more is to talk to farmers
[00:16:35.943]about this concept of uncertainty, okay?
[00:16:38.701]And trying to tell them that understand
[00:16:40.590]that every value has an uncertainty attached
[00:16:44.700]and it's really hard to predict and to mention
[00:16:47.130]and to talk about a true value, okay?
[00:16:49.920]Is this fantasy that at some point we're gonna know
[00:16:52.230]what is the optimal end rate in all the fields,
[00:16:54.330]but it's more a fantasy, it's really difficult to predict.
[00:16:59.190]I mean we have this, the study looking at all the fields
[00:17:03.270]and if you look at the predictive versus observe,
[00:17:05.430]so you have usually to make a prediction
[00:17:07.320]of even with the data,
[00:17:09.390]I mean we have a prediction error that is 34%, okay?
[00:17:14.987]So, and this is something that is important to mention
[00:17:16.560]because I always tell even to my students,
[00:17:19.950]how complex does need to be?
[00:17:23.160]Can we develop something that is simple,
[00:17:25.140]but is in the right direction?
[00:17:26.940]And sometimes when you start looking at the error
[00:17:29.250]by working on biological systems, complex systems,
[00:17:33.330]you start thinking more about this question
[00:17:35.130]about how complex needs to be a really good decision tool
[00:17:39.420]if we already have that error.
[00:17:44.040]So changing but moving on similar directions.
[00:17:47.790]I mean, many people working on digital ag is talking
[00:17:50.760]about satellite and then this, a couple of years ago,
[00:17:54.396]they start talking more about satellite data,
[00:17:55.860]satellite imagery, and of course there are multiple sources
[00:18:00.270]and of course there is multiple people working
[00:18:02.160]on satellite, yield forecast, an example,
[00:18:05.550]every day you might wake up and have another paper
[00:18:07.830]on yield forecast and now they're telling you
[00:18:09.870]that you can predict corn yields
[00:18:11.370]with maybe one bushels better and a new paper on that.
[00:18:16.410]So okay, we start thinking on many times when we work
[00:18:20.901]with satellite in a different direction,
[00:18:21.900]we start thinking about what are the different sources
[00:18:24.600]that we have available?
[00:18:26.400]How come farmers use this information?
[00:18:28.980]Can farmers use this information to take any decision?
[00:18:33.390]And of course, depending on what satellite you are using now
[00:18:35.790]because we're looking at,
[00:18:36.780]here we're looking at the same field.
[00:18:39.330]You start seeing that if I look using this type
[00:18:41.670]of satellite source, I don't even, can look at the pivot,
[00:18:44.700]then I'm looking closer to the people,
[00:18:46.710]then I'm really looking at,
[00:18:48.060]and then I'm here starting to say, oh, I can see problems.
[00:18:52.770]So having high special resolution
[00:18:54.570]is what is getting us closer
[00:18:56.070]to really start seeing these issues that we can face
[00:18:58.620]in the field and also start asking us questions
[00:19:02.040]about how do you use this information?
[00:19:05.250]What decisions can I use?
[00:19:06.990]I mean,
[00:19:08.538]what decisions can I take by just looking at this data?
[00:19:10.770]Because satellite has been around for many decades.
[00:19:15.210]The same was even yield monitor.
[00:19:18.750]But why today satellite seems to be something fancy
[00:19:22.590]and as many people will say sexy or,
[00:19:26.812]and I think that happens to the yield monitor,
[00:19:28.230]the same, is just simplicity.
[00:19:31.500]So, yield monitor is being used today
[00:19:33.600]because someone developed something that is so simple
[00:19:36.900]that now you can upload the yield monitor
[00:19:39.000]to whatever center, planter company you're using
[00:19:42.540]and you can create prescriptions, okay?
[00:19:45.960]When I started my position 11 years ago and I talked to,
[00:19:50.177]I mean these guys using precision ag,
[00:19:53.580]maybe 5% of the farmers knew
[00:19:55.980]how to use the yield monitor data.
[00:19:58.500]They even didn't know how to extract the car
[00:20:01.290]from the combine and what to do with that.
[00:20:04.890]And even today we're working on a project
[00:20:06.690]that we're helping farmers to say, okay,
[00:20:08.430]you collected 10 years of yield monitor data.
[00:20:10.530]What do you do with that?
[00:20:13.380]What is the value of the data, right?
[00:20:15.030]The data per se doesn't have any value
[00:20:17.520]unless there is an interpretation or something behind.
[00:20:20.790]There is no value of the data alone.
[00:20:24.600]And we start working on this direction,
[00:20:26.070]we start looking at satellite.
[00:20:27.360]And then there was one question that we start asking.
[00:20:30.300]I mean we have a PhD student working on this project
[00:20:34.140]and most of the projects that you'll see
[00:20:36.660]coming from our site,
[00:20:37.830]they usually have a very strong component
[00:20:39.780]from a field or crop, ecophysiology.
[00:20:44.315]So we start working with a company
[00:20:46.410]and we start talking to farmers.
[00:20:47.610]And of course when you talk to a farmer
[00:20:49.920]and there was this big idea of drones and we say, okay,
[00:20:55.260]we'll do this, we'll fly the farmer field
[00:20:57.960]and we create a map.
[00:20:59.550]Then we, the same day we collected a land satellite imagery
[00:21:03.240]and then we connected a sentinel imagery.
[00:21:05.640]And I didn't put any labels and I said to the farmer,
[00:21:08.010]which one is the one that you really think
[00:21:10.110]that you can take a decision?
[00:21:12.930]And guess which one the farmer point?
[00:21:17.190]Which one?
[00:21:19.320]Yes.
[00:21:20.520]He looked at us and he's like,
[00:21:22.110]I don't even know if I have a planter or a combine.
[00:21:25.410]I mean, what do I do with this pixel versus that one?
[00:21:29.520]There is not even a machine.
[00:21:31.200]I mean, we're not even there yet to say how do I manage
[00:21:34.500]that much probability?
[00:21:35.490]And maybe sometimes it's not good to know.
[00:21:38.640]Sometimes it is better not to know.
[00:21:41.812]This is a situation that I don't need
[00:21:43.128]to know all the variability, okay?
[00:21:45.090]But it's good to know that hey, there is variability,
[00:21:47.400]ah, good, ah, and now I can identify this,
[00:21:50.353]a 60 acre field, okay?
[00:21:53.579]60 acre field.
[00:21:54.990]The farmer was doing same rate, same nitrogen.
[00:21:58.200]And then we start looking at this field and I say,
[00:21:59.850]no, what are you doing?
[00:22:03.140]And here is 180 (indistinct),
[00:22:05.217]160 (indistinct) corn,
[00:22:06.150]80 (indistinct) corn.
[00:22:09.180]Here is the next growing in season on soybeans.
[00:22:12.090]Here is a soybean that is usually 30 (indistinct),
[00:22:15.138]here is a soybean that is 50 (indistinct).
[00:22:17.580]So there are many situations that you don't need
[00:22:19.500]to have the fancy equipment or the fancy variable rate
[00:22:22.320]or the fancy, you just need to know where is the variability
[00:22:24.927]and start playing with it, okay?
[00:22:28.196]And those are the things that we start working with farmers
[00:22:30.300]on this within field variability and interpretation, so.
[00:22:33.510]I remember many years ago that we put a simple publication
[00:22:36.420]because many farmers who are coming with this idea
[00:22:38.910]of saying, I don't even know what to do with satellite.
[00:22:41.760]And it looks complex, it looks nice.
[00:22:43.800]I mean, I can get really nice fancy maps,
[00:22:47.310]but what do I do with that, okay?
[00:22:49.830]And today you start seeing more people,
[00:22:51.690]more people like creating tools, no?
[00:22:55.020]In fact, there was a change in the last five years
[00:22:57.210]that the satellite companies now,
[00:23:00.630]they start moving into creating tools.
[00:23:03.510]So we are working, we have been working with planet
[00:23:05.610]for many years and then the last five years you see
[00:23:08.400]that evolution, no?
[00:23:10.157]They went from, okay, I'm a satellite company,
[00:23:11.730]I just built the satellite, launched to space.
[00:23:14.790]I know how to do that.
[00:23:16.140]I know how to collect the data.
[00:23:17.340]But now they're like, oh wait,
[00:23:19.050]I can do something else with this.
[00:23:21.858]And they start developing now biomass proxy.
[00:23:24.930]So now they can say, oh, I can use this information
[00:23:27.960]and tell to grassland farmers how much basically biomass
[00:23:33.390]or estimates of biomass production in the field, okay?
[00:23:37.950]And I always say, okay, the only way that I rely on that,
[00:23:41.430]those decision tools,
[00:23:42.630]if I know that they are based on true data,
[00:23:45.960]field data, okay?
[00:23:48.510]And we're still working,
[00:23:49.530]we're working with them on a project
[00:23:50.880]that's looking at phenology, so.
[00:23:53.010]I mean we start working with a crop service company.
[00:23:58.230]I mean this company is scouting fields.
[00:24:01.140]Again, we have a few fields in Nebraska,
[00:24:04.041]but we haven't got there yet.
[00:24:04.874]So, but the idea is that we are collecting,
[00:24:08.280]going to farmer fields, collecting data,
[00:24:11.380]visiting these fields many times during the year
[00:24:14.088]and really collecting data on what is a crop,
[00:24:16.110]what is a stage.
[00:24:17.940]And when they told me that it has this data, I say,
[00:24:20.640]okay, can I use it?
[00:24:23.599]It's like, they were like,
[00:24:24.540]what are you planning to do with this?
[00:24:26.450]I mean, we have been collecting this data
[00:24:27.690]for the last 10 years
[00:24:28.680]and we don't even know what to do with this.
[00:24:31.230]And then we're always wondering,
[00:24:32.910]should we keep collecting it?
[00:24:34.170]And I'm like, yes.
[00:24:36.150]I say, if we have this data,
[00:24:37.520]is the first time that you can start getting data
[00:24:39.510]at farmer level.
[00:24:41.280]And the other way,
[00:24:42.420]the only way that you can start building models
[00:24:44.580]that make sense is by looking at data
[00:24:47.989]that is coming from the ground.
[00:24:50.101]And we start working with planet
[00:24:52.110]and we start getting information developing their,
[00:24:55.513]some access to satellite data.
[00:24:57.600]But in this case, we start accessing data
[00:25:00.060]that is daily data.
[00:25:02.970]So it's called plant fusion.
[00:25:04.380]So what they do is they kind
[00:25:06.780]of use different satellite source
[00:25:08.700]and they basically build mosaic
[00:25:11.910]and we get access now to daily temporal data,
[00:25:16.860]do any daily, a question for another day.
[00:25:19.950]But we answer that question in the paper,
[00:25:22.340]we don't really need daily.
[00:25:25.473]But one of the things that was beautiful about this paper
[00:25:27.000]is like for the first time we start thinking about, well,
[00:25:30.000]yeah, crop phenology is something
[00:25:32.156]that many times we don't have data on crop phenology.
[00:25:35.250]And if we can get more access to data on crop phenology,
[00:25:37.530]then we gonna start looking at crop estimations,
[00:25:41.250]crop mapping, even thinking about water irrigation.
[00:25:45.210]When is the moment that the farmers need to stop irrigation?
[00:25:48.000]I mean, very important aspect.
[00:25:51.090]And this is what we get most of the times, right?
[00:25:53.130]I mean, many of you are familiar with this table.
[00:25:57.580]If you look at this table,
[00:25:58.609]this is the best phenology that we get.
[00:26:01.610]This is the best crop phenology
[00:26:02.443]that we get all the time during the season.
[00:26:04.530]So if you look at, this was, I mean,
[00:26:06.773]one year I just took Kansas and then you say northwest,
[00:26:09.000]west, central, by district,
[00:26:11.190]and it's telling you what is the percentage of the fields,
[00:26:14.850]okay?
[00:26:16.386]And for those that I know how this area is being collected,
[00:26:18.570]they know also how precise it is.
[00:26:22.994]I'll leave it there.
[00:26:24.510]But with satellite, we can do this if we start getting data.
[00:26:28.560]I mean if we start getting even daily data,
[00:26:30.270]we can start mapping
[00:26:31.260]because any crop follows a pattern
[00:26:35.130]that has basically a peak.
[00:26:37.597]And this is like basically bell-shaped,
[00:26:41.610]I mean type of curve.
[00:26:43.619]I mean maximizing around maximum peak of leaf area
[00:26:47.318]around flowering time and then basically declines.
[00:26:50.070]And with this type of approach,
[00:26:51.600]you can start basically thinking, okay,
[00:26:53.190]I can start training the models
[00:26:55.560]and then informing the moments of different phenology.
[00:27:00.226]And that is exactly what it is.
[00:27:01.059]So we have two studies, I mean, that are published
[00:27:03.840]and now we are working on expanding these two other crops.
[00:27:07.770]And the main goal is basically to understand that in order
[00:27:10.470]to build these tools, you need to have data from the ground,
[00:27:15.240]okay?
[00:27:16.701]Unless you don't have data from the ground,
[00:27:18.240]building those tools,
[00:27:19.170]you are basically connecting something versus something,
[00:27:23.040]but you don't really understand the process, okay?
[00:27:26.943]And this is what we're working on many farmers right now is,
[00:27:29.760]I mean it's moving to now on a kind of another pilot project
[00:27:33.900]that is, we're working on soybeans
[00:27:37.371]and really trying to understand, can we transfer the model?
[00:27:39.930]So we usually develop a model for one place in the state,
[00:27:44.398]and then this model, for example,
[00:27:45.231]we transfer to another place in the state and it works okay.
[00:27:50.140]And if it works well, that means that most of the models
[00:27:53.124]or the decision tools that you are developing,
[00:27:54.570]they tend to be a little more universal, okay?
[00:27:57.570]Of course every time, you need to get more data
[00:27:59.400]because it might change based on the stress conditions.
[00:28:03.240]I mean like specifically more on the sorghum,
[00:28:06.120]on the side of sorghum,
[00:28:08.220]there is some latency and that is affected, I mean.
[00:28:11.433]So if you have latency or if you have the crop not evolving
[00:28:13.500]as the speed that should be.
[00:28:15.967]So these tools are less precise.
[00:28:20.673]Okay, I think that hopefully I'm okay with the time.
[00:28:25.200]Another example, no,
[00:28:26.310]because we talk about publications,
[00:28:28.080]we talk about mentorship.
[00:28:29.340]So we start also thinking about a couple of years.
[00:28:32.865]And this, most of the credit goes to one of my friends
[00:28:36.120]at this point, Peter Kyveryga from Iowa Soybean,
[00:28:40.171]he has been working probably for the last several years
[00:28:42.630]on thinking ideas on how he goes from data to visualization,
[00:28:47.610]okay?
[00:28:48.600]How many times we just have so many papers,
[00:28:52.020]is trying to go through all the numbers on publications
[00:28:55.770]and you say, okay, what is the impact, right?
[00:28:58.522]Because it's a paper published in a journal.
[00:29:01.952]Well, yeah, it's good because it's there,
[00:29:06.600]it's good because maybe the student published the paper,
[00:29:09.495]great.
[00:29:10.328]But the true question is the legacy.
[00:29:12.333]Where is the legacy?
[00:29:13.166]And in many situations it's coming from transferring,
[00:29:16.410]moving these papers into something that people can use.
[00:29:20.645]And this was an example, we did a quick study.
[00:29:23.880]We have soybean seed rate responses
[00:29:26.040]to multiple farmer fields.
[00:29:28.650]And what we start building was a simple tool,
[00:29:31.440]not perfect, and it's called a simulator, okay?
[00:29:36.154]Simulator means that when you are getting into a plane,
[00:29:41.366]those pilots are flying the plane.
[00:29:42.720]They usually need to do many hours of simulation
[00:29:44.970]before they fly the plane.
[00:29:46.950]So here's the same,
[00:29:47.970]those farmers that they wanna do whatever rate,
[00:29:50.970]they're gonna collect many hours of a simulator
[00:29:54.600]and they're gonna try to understand what is the availability
[00:29:56.610]that I have in my field.
[00:29:59.190]And if I have availability in my field, I will input,
[00:30:01.650]I will put all three years of yield monitor for soybeans.
[00:30:05.310]And what we are doing is a process
[00:30:06.840]that is all automatically done
[00:30:09.210]that basically we're identifying the variability, right?
[00:30:12.570]Putting a grid in all the variability
[00:30:16.530]and then building the calculations
[00:30:18.450]of the response based on the responses
[00:30:20.850]that we have in the literature.
[00:30:22.590]And then what we do is a recommendation.
[00:30:24.660]A recommendation of what we'll seeing in rate.
[00:30:27.311]We developed a map, we have basically a profit analysis
[00:30:30.780]and we have basically a kind of a break even analysis
[00:30:33.330]based on prices, okay?
[00:30:36.870]I mean many of the things that we're doing,
[00:30:38.670]not only here but international.
[00:30:40.260]So we have a, I mean I'm also the director
[00:30:42.870]of the Digital Geospatial Consortium,
[00:30:45.240]so we are working in Africa, in Asia, in many countries.
[00:30:50.250]I mean building, using remote sensing tool to try
[00:30:52.860]to understand if intensification is a process or not,
[00:30:57.360]and trying to understand if there is an option
[00:30:59.550]for intensification and helping farmers
[00:31:01.470]to quantify practices and changes on those practices.
[00:31:05.778]So for Senegal, for example, we developed a tool
[00:31:10.746]that is very simple,
[00:31:11.579]but the idea was to start connecting data sets.
[00:31:14.640]I mean, for example,
[00:31:15.510]when policymakers who are taking decisions,
[00:31:18.210]sometimes they are taking decisions based
[00:31:19.830]on productivity only.
[00:31:21.660]And what we say is, okay, we have productivity data,
[00:31:26.683]there is weather data, there is social data,
[00:31:29.520]there is human nutrition data and there is economic data.
[00:31:33.000]But each of these data sets are in multiple places
[00:31:36.810]and they really never interact.
[00:31:39.273]They really never come.
[00:31:40.365]I mean, why?
[00:31:41.198]Because each of these data sets,
[00:31:43.260]they are being collected by different specialist scales.
[00:31:48.519]And then many times they're not even, no one has the time
[00:31:50.490]to go through all the data, decorating the data,
[00:31:53.610]looking at unifying the skills.
[00:31:56.400]And I say at some point if we wanna start doing
[00:31:59.730]or thinking about sustainable (indistinct),
[00:32:02.343]we need to start connecting data sets.
[00:32:05.049]And that's exactly what we did.
[00:32:06.961]We just start working on connecting data
[00:32:09.339]that now you can start basically, and this,
[00:32:12.225]all the data we have available,
[00:32:13.440]you can do simple graph of looking
[00:32:15.570]at how precipitation is impacting by price,
[00:32:19.230]economic production.
[00:32:20.460]And you could even compare this district
[00:32:22.620]versus this district and trying
[00:32:24.450]to get some insights why I have more mortality
[00:32:29.530]in one district versus another district.
[00:32:30.540]And if the mortality is connected to precipitation or not.
[00:32:35.070]In Senegal specifically,
[00:32:36.510]there's a lot of connection between the success of pregnancy
[00:32:41.100]of women, even with chest precipitation
[00:32:43.500]and the production of the crop
[00:32:45.720]because it impacts the nutrition of the woman, okay?
[00:32:48.960]So, and it's funny to see that we have people working
[00:32:51.930]in human nutrition, in social science, in remote sensing,
[00:32:54.960]but they never talk to the ag side.
[00:32:57.930]And you have that people, are they basically,
[00:32:59.640]they were using many times precipitation as the factor.
[00:33:02.340]And I say, why do you use precipitation?
[00:33:04.110]Why do you use productivity?
[00:33:05.160]What is productivity?
[00:33:06.960]I'm like, productivity.
[00:33:09.450]So by talking to people from multiple disciplines,
[00:33:12.780]you start making those connections.
[00:33:14.190]So, and this,
[00:33:16.035]one of the things that we like about this project is helping
[00:33:18.150]to connect multiple disciplines, so.
[00:33:22.980]This is something that we have been working
[00:33:24.420]for the last couple of years.
[00:33:27.180]Quality, you know, the quality of the crop
[00:33:28.920]is becoming something that is, I would say,
[00:33:32.865]if you think about, in this case,
[00:33:33.750]the example is for soybeans.
[00:33:35.580]But if you look at oil and protein concentration.
[00:33:39.030]So we have funding from (indistinct),
[00:33:42.348]and we have funding from north central soybeans.
[00:33:44.263]And then what we decided to do is look at some point we need
[00:33:47.517]to have a really good idea of what is the oil quality,
[00:33:49.890]how much variability we have on soy quality
[00:33:52.020]and farmer fields?
[00:33:53.550]And we set into this idea of saying, okay,
[00:33:55.410]we'll try to see if we can collect in three years close
[00:33:58.560]to 500 pharma fields across the entire US, okay?
[00:34:02.700]I mean we have one here in Nebraska is helping us
[00:34:05.130]with the data collection.
[00:34:06.690]Probably she didn't know that we're doing another year.
[00:34:08.460]So (indistinct) can explain them yet later.
[00:34:11.940]Just a joke on the side.
[00:34:13.950]But I wanna show a little bit
[00:34:15.963]about what are we doing with this data?
[00:34:17.880]What is the purpose of using this data, okay?
[00:34:20.790]And one of the main goals is to start thinking
[00:34:22.650]about quality.
[00:34:23.850]In the future and you will start hearing more and more,
[00:34:27.210]we already have a few farmers
[00:34:29.215]that they can segregate by quality.
[00:34:30.180]We have a few farmers in North Dakota
[00:34:33.120]and Michigan already selling high protein.
[00:34:36.300]We have a few companies that are already making contracts
[00:34:38.910]for high protein soybeans.
[00:34:40.800]But this year you start hearing more
[00:34:42.450]and more now interest on the oil side.
[00:34:46.350]And then we start seeing questions about,
[00:34:48.450]can you tell me what,
[00:34:50.467]how much oil variability we have on farmer fields?
[00:34:53.520]And we say, yeah, we might be able to do that.
[00:34:55.200]We just developed some nice integration
[00:34:58.980]between field data from this is more than,
[00:35:03.258]I mean we have many, Kansas and Iowa,
[00:35:04.950]the first, we satellite data.
[00:35:07.860]So we have basically using, working with many sources
[00:35:10.830]and then we just put a pipeline on,
[00:35:14.271]can we make predictions on protein and soybean oil?
[00:35:18.274]And here's an example.
[00:35:19.107]So it's a field,
[00:35:20.171]a prediction on basically a protein prediction
[00:35:23.070]and then field two protein prediction, oil prediction,
[00:35:26.310]I mean, and there is variability.
[00:35:28.388]If you look at the fields,
[00:35:29.497]sometimes we have in some sections of the field, 41%,
[00:35:31.980]in some sections we have 36.
[00:35:33.930]So you have five points
[00:35:35.160]or six points of protein variability in some fields.
[00:35:39.030]And if you look at the oil side, the same,
[00:35:40.860]sometimes you have situations in the field that we have 24%
[00:35:43.650]and we have sections that we have 19%.
[00:35:46.920]But there is a lot of variability, okay?
[00:35:49.752]Is the model perfect?
[00:35:51.150]Again, complexity, the models are not perfect,
[00:35:53.580]but they're not bad.
[00:35:55.126]I mean, we have prediction errors
[00:35:56.352]that there are only two points or less than two point
[00:35:58.050]on protein and less than one point in oil, okay?
[00:36:02.974]And I'm also always telling to my student,
[00:36:04.380]I mean always about this, I say,
[00:36:06.330]I mean if I can predict the value,
[00:36:09.720]I will not be just in a university, a professor,
[00:36:13.110]I will be in another place taking a very nice long vacation.
[00:36:16.800]What I really want to do is just
[00:36:18.570]to understand the segregation.
[00:36:20.940]Can I help farmers to segregate?
[00:36:22.980]Can I help farmers to understand if there are sections
[00:36:25.675]of the field that are high and low?
[00:36:27.420]That's what I want.
[00:36:28.920]I mean, I'm not trying to go and predict the value,
[00:36:32.010]but what we can do with this data
[00:36:34.072]and then with all the data we are collecting,
[00:36:36.600]we have collected more than 200 fields across the entire US,
[00:36:40.260]we can start building these maps.
[00:36:44.188]So now we can start in building maps.
[00:36:45.300]They are telling us what is the variability of protein
[00:36:48.080]and oil and the entire country, okay?
[00:36:52.620]And what is useful, yeah, it might be useful
[00:36:54.660]because if you think about why many international countries
[00:36:57.900]are buying soybeans,
[00:36:58.860]it's because of the exports and the quality, okay?
[00:37:02.340]And if we can start building maps and knowing,
[00:37:04.770]well, what is the high protein level?
[00:37:06.720]So I have high protein levels in this section
[00:37:08.940]of the country, in this section of the country.
[00:37:11.130]And there are few pockets,
[00:37:12.390]like here in Kansas are living in Nebraska,
[00:37:15.090]there are a few pockets of high protein levels, okay?
[00:37:20.844]It is a high map, high level, no,
[00:37:22.020]but it's just telling us what is the variability.
[00:37:25.650]And with this,
[00:37:27.176]you can start mainly make all the predictions.
[00:37:28.980]So I mean, Marina was helping us on this project,
[00:37:32.040]so she collected data from Nebraska
[00:37:34.530]and how we can use this data from Nebraska, okay,
[00:37:37.110]we can develop a tool, okay?
[00:37:40.680]So we can start developing these visualization tools.
[00:37:44.100]And what you do with these visualization tools
[00:37:46.438]is that you start using this information to inform decision.
[00:37:49.680]These are in dry basis, so the numbers are a little lower.
[00:37:52.980]So, but our main goal
[00:37:55.080]is that basically we can start putting more and more data
[00:37:57.600]into farmers county fields.
[00:37:59.790]And then you start comparing what is your county mean,
[00:38:02.220]for example, for Lancaster compared to the national mean
[00:38:05.550]and below or above,
[00:38:07.110]what is the oil percentage and below or above, okay?
[00:38:11.681]And trying to get some new data on predictions, okay?
[00:38:14.310]Of course as we get more data,
[00:38:16.350]these tools are gonna become much better.
[00:38:18.660]So if we get less data, we don't have, we have,
[00:38:22.202]in many of these situations,
[00:38:23.035]we're not collecting data from all the counties.
[00:38:25.770]So the error in those counties
[00:38:27.390]that we don't have data is bigger, okay?
[00:38:29.820]In the places that we have more data
[00:38:31.200]is gonna be more precise, okay?
[00:38:33.390]And our next step is looking not just,
[00:38:36.180]I mean at the county level,
[00:38:37.290]but what we are doing now is looking
[00:38:39.930]at within field variability, okay?
[00:38:42.120]Trying to get into a farmer field.
[00:38:44.370]That's what we're working on.
[00:38:47.820]Something just to leave behind, food for thought,
[00:38:51.750]I will say like this, we are also working,
[00:38:55.020]and I know that there are some efforts here
[00:38:57.280]in the department.
[00:38:58.125]So we're also working on developing open databases, okay?
[00:39:01.230]When I'm working internationally or even working
[00:39:03.480]in many places, one of the main challenges
[00:39:05.370]is that people wants to run models
[00:39:07.740]or people wants to get solutions or they say,
[00:39:09.780]can you connect satellite with this?
[00:39:11.100]And I'm like, with what data?
[00:39:13.920]In many places, there is not data available.
[00:39:16.650]And I'll rephrase it.
[00:39:18.870]In many places, there is a lot of data available,
[00:39:21.000]but not accessible, okay?
[00:39:25.053]And we start working on doing this more and more and more.
[00:39:27.693]And looking at start opening data, it's a lot of work.
[00:39:31.170]So, it's a lot of work.
[00:39:33.600]So I would say,
[00:39:34.500]don't try to do this with a first year student please,
[00:39:38.220]because that person's gonna quit.
[00:39:42.931]It is a lot of work and it requires a lot
[00:39:43.800]of really consistency
[00:39:46.200]and requires a lot of basically search.
[00:39:49.080]But it's important because this is what basically
[00:39:51.090]is gonna be the future of the discipline, 20,
[00:39:55.773]30 years or 50 years in the future,
[00:39:57.240]people are gonna say,
[00:39:58.890]I'm not sure what the guys 50 years ago did,
[00:40:02.970]because I'm not even sure where the data is.
[00:40:06.150]Oh, let's do again,
[00:40:07.230]it's really important to understand an action rating corn.
[00:40:11.160]We'll go again with another 50 studies,
[00:40:14.490]another round of affordability studies, okay?
[00:40:18.240]So it's awesome.
[00:40:20.100]But it's good if you have data
[00:40:22.140]because then your research questions are gonna be more
[00:40:24.450]directed towards the data that you have available, okay?
[00:40:29.160]So these are an example of a systematic review.
[00:40:33.455]We are now finishing, we finish another systematic review
[00:40:36.120]for Southeast Asia on carbon.
[00:40:38.670]We are finishing another kind of a database
[00:40:40.860]for Senegal on fertility.
[00:40:43.290]So we're kind of opening up these data sets
[00:40:46.590]and one of the critical aspect when you work international
[00:40:49.620]is to make sure you have a partner
[00:40:52.110]or many partners on the ground because you can do searches
[00:40:56.820]and you can work a lot online.
[00:40:58.080]But I mean, most of the data
[00:41:00.180]that is international is not published
[00:41:03.810]and most of the data is not in our English
[00:41:06.270]or it could be in any other languages
[00:41:08.220]and it could be in many reports.
[00:41:10.800]So you need to have someone
[00:41:12.120]that knows how to find those and bring them back, okay?
[00:41:17.646]So we're trying to bring all this data back to life.
[00:41:19.470]That's the main message.
[00:41:21.900]And in this open also if someone wants
[00:41:24.000]to any work on dilution in notioning in many crops
[00:41:26.910]is you can just get here and use the data, okay?
[00:41:30.870]I mean we just are showing the examples
[00:41:32.520]of the different dilution curves.
[00:41:35.280]I mean as, I mean concentration nitrogen versus biomass
[00:41:38.460]for maize, for sugarcane, for rice, for wheat, okay?
[00:41:42.810]And all these data sets are open.
[00:41:45.840]What is the importance of this
[00:41:47.726]is that we start building these type of approaches.
[00:41:49.800]Now we can start building what are the relationships
[00:41:53.130]of natural nutrition index versus, I mean,
[00:41:55.530]relative yield in different crops
[00:41:58.816]and start comparing between crops.
[00:41:59.700]And then guess what, from here we can jump
[00:42:01.890]into remote sensing and start building decision tools, okay?
[00:42:07.020]That's the main goal, is, unless you have some good data,
[00:42:09.540]unless you have some foundational,
[00:42:11.160]you cannot just run models
[00:42:13.350]and do data science or digital tools. (laughs)
[00:42:17.280]The foundational science is as relevant
[00:42:19.710]or even more important.
[00:42:21.120]And then we wanna make sure
[00:42:22.740]that the people doing this are agronomist too.
[00:42:27.510]A couple of things, I always say phenotyping,
[00:42:29.670]we have good discussions of phenotyping.
[00:42:31.987]We're always thinking about, okay, how do we do agronomics
[00:42:35.551]or something that is helping farmers do phenotyping?
[00:42:38.736]So I mean, you have seen this in the past,
[00:42:41.790]you can do this with cell phones today,
[00:42:43.290]you can just take a picture count, boom, right?
[00:42:48.870]But 10 years ago, and you shouldn't probably remember
[00:42:52.834]because we have been working for more than 10 years on this,
[00:42:54.480]we start looking at sorghum and we,
[00:42:58.710]and I say at some point we're gonna figure it out,
[00:43:01.802]and we start working on multiple ways, volumetric,
[00:43:03.480]allometric models, developing cylinder equations.
[00:43:06.840]When you work on sorghum, it doesn't work,
[00:43:09.390]because it's not a cylinder,
[00:43:11.670]it's not a volume like this, it is just a mess.
[00:43:16.200]It's nice looking, but it's messy.
[00:43:19.350]But, and we look at different options.
[00:43:21.732]So we finally have a student who start working with us.
[00:43:23.760]This is Pedro.
[00:43:24.930]So we have in our team, people from computer science
[00:43:27.900]and engineers working in the team.
[00:43:30.990]So finally a couple years after,
[00:43:32.640]I start hiring people that they were no agronomist
[00:43:35.040]and I'm trying to transform them
[00:43:38.830]into agro engineers, something, you can just, I don't know,
[00:43:44.732]pick the name, but beautiful about working with these guys.
[00:43:47.010]They said they don't understand so much,
[00:43:48.330]but then when you start talking to them,
[00:43:49.770]they're coming and saying, oh, okay,
[00:43:51.840]that's what you wanna do then.
[00:43:55.588]And this is something that we just put together.
[00:43:56.421]I mean, I don't know if I can click,
[00:43:58.624]but it's, yeah, I can do it.
[00:44:01.050]But it's simple.
[00:44:01.920]It's like a farmer can do real estimations
[00:44:04.980]taking pictures with the sorghum.
[00:44:07.500]They put the name, they put all the information just,
[00:44:11.100]I mean, it is just playing around.
[00:44:12.630]So you just put number of acres, put how many plants,
[00:44:16.830]you put the hybrid if you want.
[00:44:18.420]You put all this information and then the main goal
[00:44:21.675]is that you'll be taking pictures of these panicles.
[00:44:24.090]And the main goal is that you will get grain counts, okay?
[00:44:27.570]Getting grain counts.
[00:44:28.500]And then you can decide if you do this after flowering,
[00:44:31.590]if your growing season, it's gonna be good, really good.
[00:44:36.072]Very optimistic.
[00:44:37.380]You took the pictures of the field,
[00:44:38.730]you see that we took a picture in the field, not in the lab.
[00:44:42.060]You select which one you wanna do, use,
[00:44:43.800]because the software is telling you,
[00:44:45.450]I'm giving you the option,
[00:44:46.470]you select which one you want to use.
[00:44:49.350]So this is a part that we need high speed internet. (laughs)
[00:44:53.430]Why?
[00:44:54.923]If we do this in the field,
[00:44:55.756]it takes time to retrieve the data, no,
[00:44:56.589]but it gives you these grain numbers
[00:44:59.610]where you just keep going
[00:45:00.720]and then basically you just get what is the yield.
[00:45:03.390]And then based on your growing season,
[00:45:05.100]you can change your grain number, grain weight.
[00:45:08.820]You can get a report and then you can say to someone,
[00:45:12.199]you are working for a crop insurance company and you say,
[00:45:13.350]I was there.
[00:45:14.906]I can tell you that, yeah, I put together a report
[00:45:16.470]and then you get a report out and you just send the report,
[00:45:21.210]okay?
[00:45:22.140]Just a simple summary report.
[00:45:23.790]So these are the things that you can start building, no?
[00:45:26.280]I mean, but to release, you need data from the field first.
[00:45:30.060]You need to, we have tons of data.
[00:45:32.130]We have like more than thousands of images and counting,
[00:45:37.851]okay?
[00:45:38.684]To that.
[00:45:39.517]So let me close, hopefully I'm on time,
[00:45:42.330]but I wanna showcase the team.
[00:45:44.852]I mean, I'm leaving behind many people,
[00:45:45.960]but I wanna make sure you see that,
[00:45:48.399]who are the ones doing the work, agronomist,
[00:45:51.270]and then sometimes not many, too many agronomists,
[00:45:54.030]we have mixing here, engineers, data scientist,
[00:46:00.494]people from different backgrounds.
[00:46:01.327]Because I feel that that is what really enriched the concept
[00:46:03.750]of developing these new tools.
[00:46:07.320]I mean, I always discuss about this,
[00:46:09.480]building these decision support tools are complex,
[00:46:12.900]but we need to keep it simple.
[00:46:15.009]We need to find ways that if you can make a one improvement,
[00:46:18.030]if you can just improve something
[00:46:20.130]or helping the farmer to take one decision,
[00:46:22.170]even if it has some error,
[00:46:24.660]it's even better than what today is using in many situations
[00:46:27.690]because it's based on sometimes, I mean,
[00:46:29.580]whatever is coming from their way, okay?
[00:46:35.716]And this, the institute that we are developing,
[00:46:38.280]so one of the things about the institute,
[00:46:40.320]just to close is I will always,
[00:46:44.370]I was discussing with Martha this morning,
[00:46:47.829]I mean you need to have a vision
[00:46:50.253]and then you'll have the leadership following.
[00:46:53.370]And that happens, you know, we have a vision
[00:46:55.110]and idea of what we wanna do on digital ag.
[00:46:57.300]And then we put together people from the college of art,
[00:47:00.879]from engineering, from arts and science.
[00:47:04.228]And when we start putting people together,
[00:47:05.713]we realize that we have more in common.
[00:47:09.165]And then the only thing that we needed to do
[00:47:09.998]is to spend some time so we can find a way
[00:47:12.300]to communicate in the same language.
[00:47:16.623]But it's very interesting when you start going
[00:47:18.867]in that direction because it helps you to understand
[00:47:21.000]how to solve complex problems, okay?
[00:47:23.100]And knowing that many of these problems,
[00:47:25.530]they need more help from other disciplines.
[00:47:28.890]But I always say, I mean, if we can train more people
[00:47:31.860]that can do interdisciplinary work
[00:47:36.171]and I don't wanna have an agronomist
[00:47:37.770]that is an statistician,
[00:47:39.060]but I wanna have an agronomist that understands statistics.
[00:47:42.720]I wanna have an agronomist that can speak
[00:47:44.580]with someone in geography or GIS, it's not a GIS scientist,
[00:47:50.040]but he knows okay, how to speak with the other person, okay?
[00:47:54.150]And get to solutions.
[00:47:55.941]I mean, and this is what we truly believe
[00:47:57.300]on this concept of complexity, you know,
[00:48:00.062]and getting to solutions, so.
[00:48:02.010]Yeah, we'll close it here.
[00:48:03.090]So I mean, just a matter of time.
[00:48:05.250]Thank you.
[00:48:10.890]Thank you, Ignacio, for the presentation.
[00:48:15.180]So we open the floor for questions.
[00:48:18.270]As a digital presentation,
[00:48:20.105]we have a great digital audience we are talking
[00:48:21.990]on the online version, so happy for that.
[00:48:25.680]That's good.
[00:48:27.093]Yeah.
[00:48:27.926]So we open the floor for any question.
[00:48:28.759]Yeah.
[00:48:35.010]Somebody has to be first.
[00:48:36.690]Yes. (laughs)
[00:48:37.800]Early on in your presentation,
[00:48:39.240]you talked about the importance of weather.
[00:48:41.640]Yeah.
[00:48:43.020]The number 81 stuck in my mind
[00:48:45.270]and we realized that's probably a seed average
[00:48:50.970]and that includes some irrigation.
[00:48:53.730]Yeah, mostly.
[00:48:55.680]Now, part of my question is what portion
[00:48:59.130]of that weather is light quality?
[00:49:03.510]A big portion because radiation is a factor.
[00:49:06.582]So we have three main factors there.
[00:49:08.220]Water, temperature, and radiation.
[00:49:10.920]Those are the three main components.
[00:49:13.080]So radiation is still a big factor there.
[00:49:15.570]Solar radiation.
[00:49:16.620]So if we have cloudy days
[00:49:18.180]or if you have days where not really sunshine
[00:49:21.000]around flowering time,
[00:49:22.590]that was something that was impacting the growth rate.
[00:49:25.800]And then if it's impacting the growth rate of the crop,
[00:49:28.560]it's gonna impact the amount of nitrogen
[00:49:30.722]that that plant needs.
[00:49:31.980]Would you comment about temperature?
[00:49:34.650]Yes.
[00:49:36.038]Is it too high?
[00:49:36.871]That's the issue with-
[00:49:37.704]Yes.
[00:49:38.537]Heat is stress.
[00:49:39.370]In this case, in many of these situations,
[00:49:40.440]I mean, mainly in Kansas too.
[00:49:42.960]I mean high stress, heat,
[00:49:45.150]and the problem that we are seeing even when I was at Purdue
[00:49:50.548]on my time in Indiana is that we have this small episodes,
[00:49:55.020]like sometimes they were five days, no, I'm very erratic.
[00:49:59.070]They were not long,
[00:50:00.420]but we have basically high night temperature
[00:50:04.694]that usually that was a big issue.
[00:50:05.580]It's not the, usually sometimes high night temperature.
[00:50:10.890]And also the other problem is that in those days,
[00:50:13.440]the minimum temperature is going up.
[00:50:16.800]The minimum is usually going up too.
[00:50:19.380]One last question pertains to your protein monitoring.
[00:50:22.320]Yeah.
[00:50:23.669]Or estimation.
[00:50:24.502]Yeah.
[00:50:25.335]Have you used these combines
[00:50:27.461]that have protein monitors built into them?
[00:50:29.460]Yeah, as a, before, yeah, before we do this project,
[00:50:32.700]I was the one helping John Deere
[00:50:35.640]to calibrate the protein sensors.
[00:50:37.975]I mean, we were the first one doing calibrations
[00:50:39.420]for protein sensor for wheat.
[00:50:42.300]And we finished, we did some studies for soybeans too.
[00:50:48.167]So what I envision in the future is that, I mean,
[00:50:51.840]if you wanna see the big picture,
[00:50:53.070]is that farmers might have an access to these protein maps.
[00:50:56.787]And these protein sensors, for example, they're already out.
[00:50:59.640]So, and then farmers are having access
[00:51:02.880]to these protein maps.
[00:51:03.990]Now you can have basically put that protein map
[00:51:07.020]in your combine and start making segregations.
[00:51:09.540]So you can segregate areas of the field
[00:51:12.124]that you have high protein or low protein.
[00:51:14.010]And also the other important aspect on the protein sensor
[00:51:17.280]is too that you are getting information that is very key,
[00:51:21.720]is connected to the nitrogen removal
[00:51:24.510]'cause you are also getting data
[00:51:25.500]on how much nitrogen are you removing in those fields.
[00:51:28.680]What about oil sensors?
[00:51:30.810]That sensor has both.
[00:51:33.180]It's a small NIR, so it's a small NIR,
[00:51:36.720]imagine that the ones that use in the lab.
[00:51:38.700]So it's the same type of concept.
[00:51:40.110]It's an NIR that basically makes prediction
[00:51:42.420]of protein anno together.
[00:51:44.580]Yeah, yeah.
[00:51:46.005]Good.
[00:51:46.838]Excellent.
[00:51:47.671]Yeah.
[00:51:49.174]Follow up on question.
[00:51:50.203]Yeah.
[00:51:51.036]About the 81%.
[00:51:53.322]Yeah, on the weather.
[00:51:54.165]Uncertainty, 20% was related to soil and 1% is the none.
[00:52:00.821]Now the context of that variability or uncertainty is,
[00:52:05.160]in my opinion,
[00:52:06.030]it was related to your trying to get a better decision
[00:52:10.140]on making a fertilization decision.
[00:52:12.210]Exactly.
[00:52:13.748]I assume that's sometime before flowering.
[00:52:16.020]So you're gonna have to make a priority estimate
[00:52:21.515]of what the weather's gonna be.
[00:52:22.470]Yeah.
[00:52:23.340]And given that uncertainty,
[00:52:26.370]how favorable is the uncertainty based on weather prediction
[00:52:31.020]for the two weeks just before you,
[00:52:34.290]just after you have to apply your fertilization event?
[00:52:37.860]Let's say you're doing an in-season fertilization.
[00:52:39.660]Yeah, yeah, yeah.
[00:52:41.050]To try to optimize response.
[00:52:42.480]So maybe you wanna comment on that one, so.
[00:52:45.479]Yeah, and I always, may I, I always tell this, I try to,
[00:52:48.528]I mean, when I talk to farmers,
[00:52:49.361]I always give this a kind of an analogy.
[00:52:51.420]So I will use the analogy, hopefully helps, when I say,
[00:52:54.420]when farmers go to the field early season, right?
[00:52:57.150]I mean, I have many farmers, are they in Kansas last year?
[00:53:00.311]That they went to the beginning of the season
[00:53:01.800]and they apply 150 units of nudge.
[00:53:04.920]So this like going to a casino, I don't know how many
[00:53:06.960]of you really like to go to a casino.
[00:53:09.480]I mean, I don't, I don't like it.
[00:53:12.450]I mean, but is going to a casino and then you have $150
[00:53:16.350]and you just say red or black
[00:53:19.500]and you just put $150 to the red and the guy,
[00:53:22.560]the (indistinct) is just throwing the ball
[00:53:24.450]and the ball start going and you just hope that it goes red.
[00:53:28.890]Because if you go red, you make a really good decision.
[00:53:32.430]But, exactly, okay?
[00:53:36.210]So for me, this paper or this research
[00:53:39.720]is helping the farmers to think about can we get closer?
[00:53:43.860]Can we go back to the in season?
[00:53:45.900]Because when we go to the in season,
[00:53:48.210]we're getting closer to the true value.
[00:53:50.160]As you're getting closer to flower
[00:53:51.510]and you're getting closer to what is the real value
[00:53:54.869]that you are gonna need of nitrogen.
[00:53:56.100]And I always say, I use the same analogy and I say, okay,
[00:53:58.960]go to a casino, but just now don't put it on the red,
[00:54:02.600]on the black.
[00:54:04.127]Wait until the guy drops the ball and the ball is dropping
[00:54:08.130]and it's getting close to a new chest out for one moment.
[00:54:11.760]You know exactly probably that it is gonna go red
[00:54:14.130]because you are seeing that it's slowing down.
[00:54:16.710]And that is exactly what we're doing with the in season.
[00:54:18.630]Basically, we're trying to get closer
[00:54:21.240]because we know that if we get closer,
[00:54:22.800]we might know what is gonna be the weather in two weeks,
[00:54:28.989]if I don't have any rain, should I recommend a farmer
[00:54:30.750]to apply in action or that guy would have a problem
[00:54:33.390]that basically is not gonna make a crop?
[00:54:36.450]And if I recommend, I make that recommendation,
[00:54:38.880]I make a recommendation probably knowing
[00:54:40.890]that the guy is not harvesting.
[00:54:42.960]Last year we had many farmers, in central Kansas
[00:54:46.474]that already didn't harvest, that they chop the corn
[00:54:50.100]and they still put 150 units up front.
[00:54:54.090]And for me, they see some technology is something that, yes,
[00:54:57.480]I understand all the complexity
[00:54:58.860]and I understand the implications
[00:55:02.490]of a farmer entering back to the field
[00:55:05.130]when you really want to go to vacation.
[00:55:07.710]I really understand all that,
[00:55:09.240]but I'm like, we need to move in that direction.
[00:55:11.730]Maybe we need to move in the direction
[00:55:13.080]that we are helping farmers to get some incentives.
[00:55:17.610]So I think that that's the direction that we're moving.
[00:55:20.820]Hopefully we'll move some funding to help farmers
[00:55:25.350]to get an incentive to become better towards the land,
[00:55:28.560]to become better, I mean,
[00:55:29.640]to really make in season as a practice
[00:55:32.280]that should be done forever.
[00:55:36.990]Ignacio, thank you for a great presentation.
[00:55:40.080]So today here in INR,
[00:55:42.990]we have data driven nutrient management and water quality.
[00:55:48.030]So what I didn't hear from you is,
[00:55:51.180]and we have a lot of nitrate issues here in Nebraska.
[00:55:54.360]Could you comment on how these tools, you know,
[00:55:57.690]you talked about quality protein oil,
[00:56:01.591]but what about below ground?
[00:56:02.460]And how does this innovation in thinking
[00:56:07.020]about data driven agronomy help us to think
[00:56:10.530]about the trade off?
[00:56:11.970]Yeah, I mean maybe the easy example for me
[00:56:14.520]is as I always tell to my students, I mean,
[00:56:18.337]and in fact I always,
[00:56:19.170]I told this to a company a couple of weeks ago.
[00:56:20.970]I say, why do you wanna measure roots
[00:56:22.800]if you know what's going on with the plant about ground?
[00:56:26.610]And the guy looked at me and saying, but I'm,
[00:56:28.779]do you know that I came here to give you money to research?
[00:56:31.680]But I say, okay, I understand and I appreciate that,
[00:56:34.620]but I say, why do you wanna research on roots
[00:56:37.020]if you don't yet know how to manage the upper ground?
[00:56:41.460]And then the same question goes to this is like,
[00:56:44.910]by knowing the, a supply in action,
[00:56:48.300]by knowing how much you're removing from the field,
[00:56:51.120]you can start building maps on and budgets.
[00:56:54.120]So if you wanna measure nitrates,
[00:56:56.010]do it in a way that is meaningful,
[00:56:57.690]that you just go to points that are hotspots,
[00:57:00.450]that you already know
[00:57:01.283]that you might have excessive nitrogen, okay?
[00:57:05.310]And do it in a meaningful way
[00:57:06.450]because the questions are, can we really develop a network
[00:57:09.810]and measuring a million fields
[00:57:12.090]of nitrate across the entire state?
[00:57:13.710]Or we need to be smart on thinking
[00:57:15.340]about how we can start developing proxies?
[00:57:18.150]No, we need to always have a science.
[00:57:20.550]I'm not saying that the science of how to measure
[00:57:22.890]and extract and having information collected,
[00:57:25.620]but then from there I will challenge everyone to think
[00:57:27.750]about how do you go from measuring one field
[00:57:30.570]that you probably use a million or $10 million
[00:57:33.030]to that project to now say, develop a tool,
[00:57:36.690]develop something that is practical, that might be connected
[00:57:39.660]to something that farmers can take a decision.
[00:57:42.720]And remember that when we do, and Sherman,
[00:57:46.380]you guys are doing this survey,
[00:57:47.640]which is great on digital lab, by the way,
[00:57:50.628]congratulations on that.
[00:57:52.260]Many times farmers don't adopt technology
[00:57:54.240]because it's too complex, it's too expensive,
[00:57:58.830]it's too hard to use.
[00:58:02.040]So if we keep going in the same direction
[00:58:04.020]of building complex predictive,
[00:58:05.940]high predictive models with 25 variables,
[00:58:11.550]I think that we know the answer.
[00:58:14.880]First, we need to keep asking the government
[00:58:18.000]to keep adding more funding into NSF and USDA
[00:58:21.030]because if you read the news,
[00:58:23.070]they're applying to cut on NSF, so.
[00:58:27.000]So start thinking about those scenarios
[00:58:29.130]and say how we can move and I like the term data driven,
[00:58:34.650]but I will try to separate ourselves from saying
[00:58:37.890]that we're data driven.
[00:58:39.480]Because for me, data driven could be anyone.
[00:58:42.270]But in agronomy we are trying to create solutions
[00:58:44.610]that they are data informed solutions.
[00:58:47.550]You can be data driven and take awful decisions.
[00:58:52.620]So we need to be careful about the term of using data driven
[00:58:55.620]because you can, I mean, I have seen many times, no,
[00:58:59.250]I mean, I would probably not get anyone,
[00:59:01.170]but computer scientists that they run big complex models,
[00:59:04.590]machine learning, deep learning and all this.
[00:59:07.359]When you see a paper that has these terms of saying,
[00:59:10.890]I will always laugh with my students,
[00:59:12.600]like a deep learning vision heretical approach.
[00:59:15.840]I mean, and you say I don't even understand the title,
[00:59:17.790]so I don't even read the paper.
[00:59:19.950]The pain is when you have these people leading topics
[00:59:22.380]that they are ag-based topics
[00:59:24.540]and they create complex models with 25, 30 variables.
[00:59:28.170]And you know that you are not even able
[00:59:30.000]to replicate this in a farmer field
[00:59:32.550]because you can measure probably one correct, okay?
[00:59:37.065]And that is data driven, but it's not data informed
[00:59:40.260]because you didn't move from data to interpretation
[00:59:43.410]because you cannot downscale that to a farmer field
[00:59:46.860]and to take a decision.
[00:59:49.260]So, we have bigger problems yet to solve
[00:59:52.440]when we move in too much data
[00:59:55.260]that we need to solve these problems.
[00:59:56.400]So making sure that we don't create models
[00:59:58.530]that they are too complex or we don't create,
[01:00:01.260]as I was talking with Carolina today on this idea
[01:00:03.780]that agronomists always want to have R squares of 0.9,
[01:00:09.300]right?
[01:00:10.133]And I'm like, wait, if I have a 0.5, I'm jumping,
[01:00:13.560]and if I see that the train is positive, done,
[01:00:15.720]if I can help the farmer
[01:00:16.770]to take a decision much better way in a simple model,
[01:00:19.860]I will not add more variables.
[01:00:22.980]Someone in the future after me,
[01:00:25.530]you are gonna come and gonna do something much better.
[01:00:27.930]Congratulations.
[01:00:29.010]That was a step of progress.
[01:00:31.140]You shouldn't be trying to solve all the problems
[01:00:32.930]in one step,
[01:00:34.770]but you should always at least try to think
[01:00:37.110]that you are going into the right direction.
[01:00:39.990]The right direction in many situations
[01:00:41.430]is building data informed,
[01:00:43.800]making sure that the farmers are part of these tools
[01:00:46.290]because we're testing these tools with farmers.
[01:00:48.720]We're applying to send these applications
[01:00:50.310]to farmers now to give us feedback.
[01:00:52.500]If they don't like it, I need to change it.
[01:00:55.320]Simple.
[01:00:56.153]Because if they're not using it
[01:00:57.420]and they don't like it, it's too complex.
[01:00:59.550]They don't know what to do next.
[01:01:02.130]So then I be that tool for me.
[01:01:05.310]So I'm willing to think about those ideas, no?
[01:01:08.190]Sometimes they are a little more complex
[01:01:09.810]involving stakeholders, making sure you get feedback,
[01:01:12.180]but we need to do it, okay?
[01:01:15.738]And nitrate quality, huge problem, right?
[01:01:18.120]Huge problem.
[01:01:20.041]And I think that the questions is,
[01:01:21.030]how do we start building solutions from multiple angles?
[01:01:27.164]And believe it or not, some of those solutions might be
[01:01:29.310]to start the less complex.
[01:01:32.220]If you can gimme an action budget map of how much you apply,
[01:01:36.780]how much you remove, for me,
[01:01:38.670]that would be my first starting point
[01:01:40.320]because that will be something that I can work
[01:01:41.880]with the farmer, the farmer can do it.
[01:01:43.710]And then from there I can start identifying hotspots
[01:01:46.440]on potential places where I should say, okay,
[01:01:48.900]we should go and measure this.
[01:01:50.250]We should go and look at data on nitrate
[01:01:52.080]and understand if when we are leaving nitrogen behind,
[01:01:55.260]if it's staying there or if it's moving.
[01:01:59.790]We're working on complex systems.
[01:02:01.320]Nothing stays in one place,
[01:02:03.900]but we need to start from some point, okay?
[01:02:07.365]That's a little bit the response there.
[01:02:12.291]We have time for one more question.
[01:02:16.440]Professor Ignacio,
[01:02:17.310]thank you for this wonderful presentation.
[01:02:20.130]I'd like to know more about that mobile app.
[01:02:23.430]Yeah.
[01:02:24.816]For the yield estimation of sorghum.
[01:02:25.649]Yeah.
[01:02:26.482]Can you enlighten more on what parameters you consider?
[01:02:30.510]Because the panicle, the whole panicle,
[01:02:34.470]we cannot see just taking a picture
[01:02:36.600]because on the other side there are more number
[01:02:40.099]of primary panicles, right?
[01:02:41.042]Yeah.
[01:02:41.875]So how do you consider for that variation?
[01:02:44.220]Because it might differ from one plant to another.
[01:02:47.400]Yeah.
[01:02:48.780]And what the model does is it's a very simple,
[01:02:50.280]like a machine vision.
[01:02:51.660]So using a yellow B model of deep learning.
[01:02:54.720]So, and then what it does is you take a picture
[01:02:57.720]and if you see four or five panicles,
[01:02:59.460]it will bring all the panicles, okay?
[01:03:02.280]What you need to decide as your, as a farmer, right?
[01:03:04.950]Because, or you need to decide which one is a panicle
[01:03:07.934]that you want to use for the yield estimation.
[01:03:09.990]So you might see all panicles around the back
[01:03:12.545]that they are halfway or small,
[01:03:15.300]but you are the one taking the decision.
[01:03:16.800]We don't wanna decide for you, we decided to do that,
[01:03:19.680]because we don't wanna just make an assumption
[01:03:23.280]based on what the farmer wanted to look, okay?
[01:03:26.280]And we have done these ideas in the lab now.
[01:03:28.980]We were taking the,
[01:03:30.180]the way that we start all these process 10 years ago
[01:03:33.300]was like taking thousands of thousands
[01:03:35.220]of panicles from Chinese, from India,
[01:03:38.340]from different type of compact,
[01:03:41.130]I mean the ones that (indistinct), I mean,
[01:03:42.900]and taking pictures, counting (indistinct),
[01:03:44.670]and creating a database.
[01:03:47.610]And then now we're getting closer
[01:03:49.530]to say we have a better basically flow off,
[01:03:53.190]because we start with idea of a cylinder
[01:03:56.490]didn't work really well,
[01:03:58.110]and then we start working on edge identification
[01:04:01.140]and then we just were cropping the image
[01:04:03.060]and it worked better.
[01:04:03.893]And now we are doing like more density map.
[01:04:07.170]So it's a density function.
[01:04:08.280]So I mean, I'm looking at kernels, okay?
[01:04:11.700]And again, if you look at,
[01:04:13.334]if I show you the equation behind,
[01:04:14.970]it's an equation that is very simple.
[01:04:16.380]It's just relating grain number versus the volume.
[01:04:21.300]And it's an equation that if I tell you
[01:04:22.890]it has an higher square or 0.7, 0.75, it's perfect.
[01:04:27.030]No, but do we need perfection?
[01:04:30.090]Who's gonna count those kernels?
[01:04:31.620]Good luck.
[01:04:32.760]I would count many already.
[01:04:34.020]So, and no one counts.
[01:04:36.060]I can tell you because I talked to many, so many farmers,
[01:04:38.820]when you are telling me that you have 3000,
[01:04:41.250]3000 or 3020 is the same for me.
[01:04:44.790]So even as, I mean even having errors of 15%
[01:04:48.570]of (indistinct), or I mean,
[01:04:50.220]or relative (indistinct) square is nothing on this count.
[01:04:52.830]So yeah.
[01:04:54.870]And as I always say, just to finish,
[01:04:56.370]and I mentioned that to many of the ones I visit today,
[01:05:00.540]there's nothing to the estimation of sorghum today.
[01:05:03.690]Nothing, okay?
[01:05:05.850]We are now working on creating something that's perfect,
[01:05:09.060]but something that is something.
[01:05:12.810]Well, thank you, Ignacio, for that closing.
[01:05:17.861](audience applauding)

The screen size you are trying to search captions on is too small!

You can always jump over to MediaHub and check it out there.

Comments

0 Comments

Advancing Digital Agriculture for Improving Agronomic Decisions

Description

Searchable Transcript

Comments icon comment

Related Channels

Comments