Making of a Model

Jackson: Welcome to the FarmBits podcast, a product of Nebraska Extension digital agriculture, I'm Jackson Stansell
Sam: And I'm Samantha Teten, and we come to you each week to discuss the trends, the realities, and the value of digital agriculture.

Jackson: Through interviews and panels with experts, producers and innovators from all sectors of digital technology- we hope that you step away from each episode with new practical knowledge of digital agriculture technology. Hello and welcome back to the FarmBits podcast for the second episode in our nitrogen management technology series.

Sam: In this episode, we will turn our focus to nitrogen models- one group of tools that has been proposed for improving nitrogen management.
Jackson: Dr. Laila Puntel, assistant professor of soil and water sciences at the University of Nebraska-Lincoln joins us for this episode to discuss the basics of models, why models work when they work and what shortcomings they have.
Sam: This episode sets us up for a follow-up episode featuring a commercially available nitrogen modeling tool, but let's start off getting the expertise in our interview with Dr. Puntel.

Jackson: What are some of the challenges that producers are facing right now in regards to making these nitrogen management decisions and what sort of research are you hoping to contribute to that's helping them make better decisions?

Laila: Sure well, nitrogen is my favorite, all my work my previous work has been in nitrogen since my master's, my phd and then I got into a position at Nebraska that is mostly related to nitrogen management, so it's pretty exciting and really kind of my motivation is around the fact that nitrogen is still very challenging very challenging right. So, and I think the challenges that producers are facing is the uncertainty around that nitrogen right, so basically there's a lot of variability. The nitrogen that maximizes your profit and minimizes the impact environment could be very different year by year as well as within the field. So, kind of picking up what's the best nitrogen rate that you want to select for a field is still being a challenge right, so to begin with that and then along with that kind of what tool is the best to help you to predict that nitrogen right. so correct yeah

Sam: So, we've had another episode that focused on reactive or responsive nitrogen management tools but can you talk about more maybe predictive nitrogen models and what's out there as options?

Laila: Sure, yeah so in terms of tools as you said like active canopy sensors or more the sensor the sensing technology it's more reactive and uses the plant right as kind of our living sensor to tell us how much nitrogen is in the soil how much nitrogen is needed by the plant. Now, some other tools like crop model based tools kind of allow the producer to cover some of the risk related with weather and how the weather interacts with the crop management as well as some of the soil characteristics, so I think there's plusses and minuses but just the complexity of a crop model allowed us to kind of contemplate some other aspects of the system for people who maybe technology's a little bit intimidating or they're not very familiar with these

models.
Sam: Can you just give just a really basic overview of how simple a model can be to maybe how complex it can be and some examples, so that way they know if they've heard of these before.

Laila: Sure, yeah so probably we should have started there yeah what is the crop model mm-hmm. So, if you think about it every day you can simulate right like a virtual reality of how the plant is growing, so every time you're fixing radiation and put it into biomass right so that's how your plant grows, so now imagine if you want to make it a little bit more complex right? You started there now you try to assess like how much nitrogen is in the soil and for that you need the organic matter. You need to know how it decomposes and how much nitrogen is being released, so that's one piece but then the plant on one side, if it grows too much then it's going to try to uptake more nitrogen or less nitrogen right sure, so models account for all that. Every day you kind of have a balance like a bank account sounds equation it's a mass balance equation, but that's when the weather kind of gets into play right because if it rains more if you have cloudy days if you have more radiation then that potential growth is getting bigger and bigger. Right and then like the supply of nitrogen as well, so that's kind of like a simple way of putting that the thing is that there's a lot of other factors that can be included into the model to make it more and more complex some of those are losses right so we know that nitrogen leeches in the ground and so we're starting to add the hydrology concepts right yeah. What texture, how that rate of nitrogen is being lost from the soil profile, how much nitrogen is run off like volatility. If, we get into the complexity of the nitrogen cycling interactions and everything that's the way that you can make the model more complex and hopefully more accurate and hopefully more accurate right but you know gas emissions right like all volatilization things like that. Our process is that we don't have that much data either right it comes back to the data it goes back to the data right.

Sam: Okay, and what are some examples?

Laila: There are some property models that for example we are partnered and collaborating with Granular. They have their own model, but we know that it's a process-based model. What I mean by that is exactly what I just explained as the more simplistic way but pluses and minus. That's what they run into the background right and then we are also working with AdaptN, which that's a model that it's actually the basis dsat, which is a very it's open source model that was adapted in Cornell University to make recommendations of nitrogen and that model then was acquired by Yada a few years ago and then got commercial, so in the background they run a process-based model with all these components that we discussed today including nitrogen losses as well. So, they incorporate all the complexity even cover crops or residue removal things like that, we are also aware of Fluorosat, is another software that we are testing in this project that I mentioned and they use another open source model that is very complex. it accounts for all the mechanisms from losses and and other factors of the management so that is also another alternative. Farmer search we know that also runs a model in the background and probably I can keep naming, and I probably won't forget and in the next years I feel like we're going to see more and more coming up.

Jackson: So, what are some of the strengths and weaknesses that we have there between those two particular, I mean you talked about how the the sensors kind of incorporate the plant. What are some of the strengths of that versus like what are the risks that you're running when you're kind of using only the plant to tell you everything that is going on out there in the field that maybe a model makes up for?

Laila: Sure, I think the the biggest part of that is that the crop model will interpret the weather and what is going to happen into the future as well as what the grower is going to be looking for. So, in the case on a fertigation event or maybe a a round flowering application or things like that, you can already account for those as well as account with what the weather will look like later. I think that's kind of what they complement each other because I feel like canopy sensors or some sensing technology will give you kind of a picture of today, right, how the crop looks like but maybe crop models come into play to see how the growing season will look like and see if that nitrogen that you put on is actually gonna end it up you know yielding more, responding more and give you more profit.

Sam: Yeah and building on that maybe we can step back a bit and talk about the goals of nitrogen management and then how effective these models are in those goals? So, there's different philosophies can you go into that a bit?

Laila: Sure yeah, so I guess your goal I would say is trying to hit the optimal end rate that maximize the yield response and that is because you want to make sure that you are completing the yield response at a point where it yields the high profit to your field and so ideally the goal is to hit that point in every part of the field right. Let's think about an idea where you have sure highly resolution right and we want to target each pixel or each part of our field. Our optimal end rate that would be kind of our main goal now we know that we have all this complexity and the weather, the soil the management only interacting. So, I think that's kind of the challenge that is still there right and just to clarify

Sam: For people listening, the optimal nitrogen rate does not always equal yield maximization.

Laila: That is I mean that's pretty interesting. We tend to think that we want to apply nitrogen to maximize yield and the thing is that maybe that increase in bushels doesn't pay the extra amount of nitrogen that you put on, and that ties back to you know these models and how do they perform well. Sometimes it's really hard to capture the response to nitrogen rather than like hitting the maximum yield right. So, there are two different things.

Jackson: So, you talked about how models basically give you a view at what the growing season is going to look like and for a lot of people out there they probably see the variability year after year and they're like how could you possibly know what the season is going to look like? So, what sort of data has to go into these models in order for them to work appropriately and really give you that season-long picture that you're looking for?

Laila: Sure, models let's start from the beginning right. Models are not perfect and even you know if a statistical model an equation or even a crop model they're not perfect, so they're things that they don't contemplate. A risk like a weather event or a hail event right and a disease or a pest those things usually tend to be out of what the scope of a model at least for now. Let's see what happens in the future right. But, the way that they contemplate all these things happening is because you input, soil growers can either utilize the data that they have been gathering for the last few years either agreed sampling or some soil mapping or sometimes we can rely on what the open sources of of soil data are available to kind of feed that model. But, then we will also contemplate all the management inputs for that grower like planting day right. What kind of variety or hybrid they are using for the crop things like that . And then we can also utilize whatever weather sources they have. They might have their own weather station that we can fit in or there might be other networks available that could be included. So, I think those are the three major ones. All the crop management, soil data and weather to make that prediction.

Sam: Can you dive into the weather data a little bit more about the predictive part of it which I'm sure is the most challenging part but also how you're using the historical weather data and if whether that's many years worth of historical yield data but also what's already happened up to that point in the season, why is all that data really important to make an accurate recommendation?

Laila: Yeah, that's whether data is key right we tend to rely a lot on what's going to happen and those predictive models. So, what happened up to the day that you're going to do the prediction that is key right because that kind of detects, dictates how much nitrogen you have available in the soil, how much water your soil is holding and you know how much growth you have up to that point. So, that's what I always say that you can maximize the power of crop models the later you are into the season right or the more chances you have to put nitrogen on because the more you can utilize whatever happened. Now, it has been shown and actually was part of my research work a few years ago where we did an assemble of weather data, and I think that's very powerful. So, usually we tend to use whatever comfortable forecast we have even if it's seven day three days ten days that that period of time and then kind of assemble that with historical data and then what we found is that that assemble of historical data is actually tend to be very representative on average right on what what could potentially happen the rest of the growing season. There has been some approaches to kind of divide the data into let's say more drier years, more weather years right like tend to do predictions with those two, and I think it will depend on where you are in the country to know if those approaches are better or worse right, but historical weather is very powerful to at least kind of have a probability of what are my chances you know that my yield is going to get into this trend above 200 bushels so below 200 bushels. So, sometimes we look into this predictive models more like to guide us on what's the trend rather than what is the exact value of yield that I'm going to get.

Sam: And, you kind of hit on this a little bit with the regional information you said what works best for your region what about other characteristics as well like the soil data. A lot of these models may have been built for a certain region what are some of those challenges and how can we better inform it for other areas?

Laila: Totally totally and actually you will see that some models were created in other parts of the world and then they were bringing it into U.S., right, but they has been tested and accommodated and validated in multiple environments that all that research and science had been put it in. So, today you can maybe click on a public available model and you might even find hybrids that are grown in United States even though the model might come from another part of the country. So, I would say that the science behind is very robust and they have been validated however as you mentioned we have models that might be approved or very well proved in the northeast region of the United States but then when you come to midwest the limitations are different. The weather conditions are different and one of the things that we learn is that models tend to perform the worst when you have extreme events, and so a flood right or a heat or stress during the growing season, water table drainage systems different factors that are really challenged to capture in a model. So, really at a regional level I think that's part of my excitement in my current position is trying to find out those things that actually can make those models better for our regional conditions.

Jackson: So, I guess kind of on that on that regional scope as well, thinking about weather data a little bit more instead of soil data what sort of resolution do you need to have in your weather data to really have a reliable model and how reliable is it to get that from historical weather data because I imagine we're just starting to get into having really high resolution weather data with the kind of proliferation of weather stations and that sort of thing in our modern era, can you speak to that a little bit?

Laila: Yeah, so when you mean resolution it's like if you need hourly daily both temporal as well as spatial you know if we're talking about doing having a one weather station for every I don't know 10 square miles or something like that versus a weather station for an entire county. You know, because you think about how intermittent weather stations can be out there. Let's divide the conversation into right one is the temporal models in general they run into with daily data, so every day at least you have to have a record of what happened right. So, and that's kind of pretty a standard you can configure your weather stations for like getting that type of data now in terms of the spatial distribution I think one of the things that help especially for growers that are thinking about adapting these tools is have a rain gauge. So, precipitation is one of the ones the variables that could vary so much right and in a spatial grid. And so, I think that kind of validating or accounting for that will make our predictions much better, radiation and temperature those things kind of tend to vary less within the space. So, I think kind of having your own measurement on precipitation is very valuable.

Sam: You mentioned about how the soil and everything can vary or that response nitrogen, response curve can vary a lot even within a field or that economically optimal nitrogen rate can vary a lot. So, what kind of data are you using to potentially help inform how that variability is in the field especially for nitrogen management.

Laila: Yeah so, there's two worlds right one is the one that exists in commercial available tools and the other world is the one that exists in academic world where you can tweak and access to a lot of data right and make a more complex configuration. So, focusing on more commercial versions and what goes into those tools, I think right now we can utilize a lot of graded information from soil mapping with electrical conductivity some elevation maps to account for more specific slope or curvature in the field rather than just relying on two three values from our public available data. We now have those options now that transition is just happening or I see it happening where every year there is a new release within the industry, companies updating or making those inputs more precise because what happened is that if you use you know just one unique value for maybe configuring your slope let's say that affects all your hydrology, right all the water movement how much water you are getting in some parts of the field, and so going back to your point if you want to capture the yield response to nitrogen correctly you have to make the best to kind of represent those different environments right. So, soil data I think is a big one and I'm tending to see that other indirect measurements like electrical conductivity is being used to characterize how the the soil looks like or maybe what's the distribution of texture right. So, I think that's becoming more and more available and some other are using even remote sensing to characterize some of the initial conditions of the model. And so, maybe with an image you can know what happened last year right and how your yields were and how much residue you have, we know all those things will affect how your crop corresponds to nitrogen.

Sam: And, on the podcast before we have interviewed Veris technologies and swat maps yeah how do potentially those types of maps which you know use a lot of technology you just mentioned but how do you make those specific for nitrogen or is it the same type of idea?

Laila: Sure, I can speak on the electrical conductivity in particular. There's two ways that are currently being used, and we are currently testing as well on the research side of things. The first one is that you can implement those to make a better like division of your field where you might have different yield potentials and different yield response. So, that's kind of a more high resolution- I will say discrimination of how different is your field, so that's one part and the model can take that in to simulate different areas of the field. So, that's one key the other one is that if you have other data or the other layers of information like a grid sampling you can actually couple your electrical conductivity with some of those variables and kind of make it as a factor of how they vary across the field, and so those will help you to configure better your soil profile you know in each of the particular field. So, there's multiple ways that this indirect measurements are feeding the models.

Jackson: That's all really really good info. One other thing that kind of affects models, we've talked about beyond just these soil characteristics and weather characteristics, natural environment there are also management variables we've talked about right so how challenging is it to accommodate all these different management variables that go in from timing of applications to products that are being applied and kind of you know whether people are using inhibitors whether biologicals are getting mixed into the system, tillage, cover crops how do those go into a model and how can I make up for them?

Laila: Yeah, I would say the easy the easiest one are about management. I think those are pretty well adapted. I would say like in terms of you know your planting day and what kind of variety or hybrid you're using or previous crop things like that it gets a little bit more complicated. When we get into manure application, cover crops and you mentioned inhibitors, biological stuff like that. So, if you think about it- I think the biologicals are the toughest ones just because models per se do not have any biology of the micro, so that's a new world. It's coming up right at some point we're going to incorporate those where you're going to think about it, what biological product you're applying and you may know what certain bacterias you have there. You might be able to incorporate in your model right and so those living microbes are going to have their own pool into these models, so we were able to simulate them and and there's models that do that but separate from the ones that maybe growers have access to. So, I think that's coming and that's probably the toughest one, but I would say that manure as long as you have some sort of idea or values about how much nitrogen is there how much phosphorus how much organic material there. The models will be certainly able to take those values in. Cover crops, I think there's more advances on that because we have more data too. So, we have more experiments where we know how much nitrogen maybe they will be releasing throughout the growing season and those are values that we need to calibrate. The models right so if a grower knows for example what kind of mixture did they use for their fields like if it was a 50/50 rye and maybe a legume that information will be an input termination date. How did they terminate or how much did they take the residue out they leave their residue in. So, I think they're trying to simplify how to describe a cover crop, so then we can take it in and simulate it.

Sam: What do you think are some of the most common mistakes when it comes to incorporating all this data and using it into form a nitrogen prescription or a nitrogen recommendation?

Laila: I would say first is not being aware of what kind of data do you really need to make a good use of the tool right? We tend to do a lot of assumptions on configuring these tools, but I'm certain that there's a lot of support, and I feel like it's going to be more and more support. So, growers can get you know a crop consultant and advisor that helps them to make sure that the inputs and the configuration is correct. So, I feel like impetus is first the second one is the interpretation. So, I feel like big part of it is understanding what is this showing like what is this figure saying right, and I think in that aspect industry is doing a great job at trying to be a little bit more friendly with the interface and actually showing things that are like more easily adapted to their common practices, right. It tends to bemore intuitive just to go and look into yields right. What is my yield is going to be well if you think about nitrogen management you actually want to know how much nitrogen is still in my soil right and maybe how many bushels I can kind of support with what I have right now, turning that data into something that the grower can take into account that's you know valuable, and I think that it stills needs some help to get a good understanding of what's the output of the models. So, I think those two and maybe I will name a third that is considering that the tool is not perfect. And so if it doesn't seem right it's like because most of the times it's going to be right but maybe you might have a very particular growing season you know and I feel like everything is evolving and so it's going to get better and better at you know trying to hit that optimal end rate. But, acknowledging that is a good step.

Jackson: So, we've talked you just kind of brought up two great points right that you have a lot of data that goes into it. You got to realize that there are you know potential errors there but it also has to really be easy for somebody to pick up and use if you're going to use this on a on a wide scale. So, how do you balance that is there are there any sacrifices that can be made on one side or the other and where are they, I mean where are the trade-offs?

Laila: Totally yeah, that's a great question and it's part of the motivation behind my research and and our team is to find out you know what are the things that you can sacrifice. Do we ask growers to have eight inches you know organic matter value or are we okay with a smart farmer type of shallow organic matter. Can we feed the model anyways. There are so many questions that I think we just still need to work on, and as you mentioned like sensitivity analysis type of thing where you know which ones are causing the the biggest you know weaknesses in the model and what others are really needed. I would say that most of the models still relied on yield goals right and that's a very critical aspect because even the model will track what's going on during the growing season. Some models will still rely at the end of day on what was your yield goal determined it has to have a target, it has to have a target right. So, I like the part that we're still incorporating some of the growers type experience into the decision but I'm actually like very careful about that because it's like we're trying to simplify something that is complex. But, the problem is complex so nitrogen is complex per se, and we have these crop models that attempt to capture all the pieces of the system and now we are trying back to simplify it right. So, before we may have even captured everything with the complex models themselves, and sometimes you know as is the data that is still not quite there. There are some particular fields that may behave very average right there are conditions that are you know pretty simple to simulate but again there might be some aspects or interactions that are very complex.

Jackson: So, what do you think I mean this has kind of come up a few times in terms of how do we actually accumulate enough data to get started with this? Do you have any idea what the lead time might be for somebody who wants to get into modeling, how far out does a farmer need to look or an advisor need to look at his clients and say you know I really think I need to bring a model to my clients, but I don't know if I have the data that's stored up yet. I mean is it three years that they really need to get the right amount of yield data to get everything done on a practical level, how much time do you think you need?

Laila: Do you want the sciency answer or you want the practice? I want a practical answer. I want a practical answer from somebody that does extension and you know does a lot of on-farm research. Laila: Totally, no I feel like ideally you want to have some idea on your yield data right, so yield money if you're going to use the model for doing a variable-rate application so intended to vary the amount of nitrogen across the field, I'll say that accounting for a few years of yield monitor data will get a better you know idea about your yield targets, and how the how the field responds to every year. The minimum data is five probably just because I mean you might hit corn, let's say that you are planning you're making a management plan for corn and you have a corn/soybean rotation so at least you might have two three years of corn. You can incorporate in soybean but we tend to prefer corn to kind of tell us the story about what's happening in there as far as actual yield goes. Or, what was the response to your nitrogen management that you have done in the past. So, I would say that's kind of my advice and on time. But, the other data layers I mean you can get those if you really invested, you can get those in a year. You can and are things that you do it once and pretty much you know you can utilize it multiple times with some of the new tech that we have out there like Smart Firmers like you mentioned earlier and you know some of the EC mapping units you can throw on a planter at this point. There's a way to get it done, again I feel like that time or waiting time to get implemented is gonna get shorter just because I think we might be getting more benefits out of the remote sensing world where you know well maybe with a couple of yield monitor data sets plus remote sensing, we can kind of get it started at least. It's not going to be perfect right but get it started.

Sam: And building off of this amount of time thing but how much time does it actually take to run the model because I know it definitely depends, but this sounds really complex when people are listening to this is some of it automated- can you talk about how you actually get from all this data to a recommendation?

Laila: Totally, well first to clarify there are some softwares available out there where the representative will configure your field for you, so you just need to provide information, so the time requirement there will be just to make sure that you give them good yield data right and you inform them what were your management practices or what you are planning to do with your nitrogen right. So, in terms of sources timing and any kind of future event that you are planning I feel so in that sense that might be kind of easier right because then you can just someone will help you to configure your account. They ran the prescription for you and then you just follow your field and try to interpret you know how your field is behaving. There are some other levels of software where you really have to put more hands on and so in that sense models can can take you an hour or so to configure you know everything. Incorporating the yield data, incorporating soil information the management and everything so it will depend what kind of tool you pay right. So, I guess in that sense we still have a long ways to go right because there's a degree of simplicity to complexity, and it will depend on the growers right, how much time they have to dedicate to it.

Jackson: I've talked to an advisor here in Nebraska who said he was spending about eight to ten hours a field to get everything set up as far as getting all the data in the right place, getting it in the right format I mean it's the upfront cost is big but then once you have everything up front it's easy to rerun over and over again right? So, it's more the startup cost that's the issue more than the continual time input right?

Jackson: Along those lines, where do you think we are in terms of adoption? How many people are actually using the using models out there actually implementing them on their farm, are we in early stages are we even on that kind of uptick where do you think we are?

Laila: I think we are in the uptake okay and maybe I'm basing my my opinion, this is just an opinion on how much services are coming out from the market. If you look at the at the last you know kind of reports across the country, we see like a very steep curve you know kind of the providers on all these cloud-based what we call cloud-based services is increasing a lot. So, that means that a lot of crop advisors are being exposed to all these variety of tools. So, I would say that will kind of get some traction on how many growers are adopting that, but currently I mean with my experience at least in Nebraska what we are seeing is that there's not a lot of growers that are actively using these platforms, so there's a lot of room for for increasing and we are actually working on the extension side of things to try to promote these tools among them and actually help them to get on board and see that it actually is not that complex and that you know as you said it once and then once you understand what you need is easier.

Sam: And where do you see us going in the next five to ten years and whether that be an adoption, but also in how we continue to develop these models?

Laila: Totally, what I see going I think I mentioned it a little bit before, but I think all the automatic input or information coming from remote sensing will be a big player in the future where I see the improvement on some of the configurations being automatic in the background you know without having to get that much input from the grower, so that would be kind of solving our issue between the trade-off. You know, sacrificing some inputs for quality and again that will get a better recruit representation about what's going on in the field. So, I see us going there. To be honest, we still need to test you know all these tools and have a very good understanding on where we are at.

Jackson: So, along that vein would you mind tell us a little bit about the CIG project that we're doing here in Nebraska that's going to help growers understand models a little bit better and what resources might be available to them through this project?

Laila: Sure yeah, the CIG project is supported by the NRCS and USDA we started last year in the fall as a pilot because that was like what the time allowed, so we are kicking off this year with wheat growers and corn growers, so the main idea is that it gives growers the opportunity to test any existing tool for nitrogen management. So, that that contemplates not only crop model based tools but also remote sensing tools like canopy sensors or any other image provider that could be translated into a nitrogen management. So, we see two main outputs for the grower- one is to get exposed to a new tool with no risk right because we will compensate the grower for participating in our study as well as we will cover any costs from either an extra application the extra cost of you know getting into the software or getting an account and the other one is we'll test if that tool actually works for them. So, yeah we are helping a lot of growers that want to test these crop model tools to set up the field and get the information that they need.

Jackson: Sure, and will the information from that just basically be available through the on-farm research network, I mean is it going to be publications how is that information going to kind of be expanded out beyond the growers that are participating to maybe other people in Nebraska and beyond?

Laila: Sure yeah, the information is going to be of course shared through the on-farm research network we have an annual publication where all the results from the network that is going to get published as well as there's going to be a lot of winter meetings where we're going to be recording but also in person meetings to share the results of each of the growers and maybe where growers also have the opportunity to tell how was the experience which is always great which is also good. And then we are going to keep posting like on Twitter and some other media to see the results that are coming up from this and then of course for the growers that are participating, we are collecting a lot of intense data on these sites, so from images source sampling and other analysis so they get to also share that and share it with other neighbors that might be interested also in learning about their fields.

Sam: Yeah so, we wrap all of our interviews up with what one piece of advice do you have for our listeners when it comes to nitrogen management?

Laila: Oh, I would say try them all out try all the tools out and see which one fits you better at least if you know if you have an extension educator or if you have some extension help out there that can you know help you to set it up a test. I feel like that's the the best way of kind of getting a conclusion about what's the tool and if you have questions you know reach out to people that has been working with these tools before, so you can get a better understanding but don't miss that opportunity I think a lot of the nitrogen issues that we have you know not knowing that right amount or right timing these complex tools are coming to solve that issue. It's just we need to kind of make it more friendly in a way and get it closer to the growers, so to the growers they can access to us or anyone working on extension and research to test them out.
Sam: Thank you to Dr. Puntel for joining us today on the podcast and talking more about nitrogen management.
Jackson: Yeah, it's cool to have her on. We interact with her a lot so it's nice to actually sit down and do it a little bit more.
Sam: Yeah, that was a really fun interview and my favorite part was she did such a great job of really describing all the components of a model. It makes us think a lot about how growers are currently deciding how to apply their nitrogen and if whether they're really thinking about all those components and how complex it is, but also balancing with that with making sure it's easy to use and making sure it's practical. So, I thought she did a great job of breaking that down.
Jackson: Yeah and I think i think she did a really good job too of explaining why exactly data is so important to models. I mean at the end of the day model is only as accurate as the data that's going into it, and it is the quality of the data that's being collected is pretty much equivalent to the quality of the model, and I think with weather data it's kind of where that is the most almost every part of a model that we're putting so much reliance on weather in terms of the modeling side of things, so just interesting to think about. Next week we'll have an opportunity to hear a little bit more about a commercial offering for modeling tools that's available through Granular. We'll have Bob Gunzenhauser, who's going to talk to us about about what Granular is doing, and we're pretty excited to bring that episode to you. So, we hope we'll tune in next week.