Brennan: Precision technologies on extensive beef production systems

Brennan: Precision technologies on extensive beef production systems

Show Video

introduce uh Dr Jam Brenan or Jamie most the time right um I first met Jamie we were trying to figure this out a while ago we were both working on our phds I believe and at an SRM meeting and we're both talking about data and uh one of the things I thought was really interesting about Jamie's beginning into his grad school and going on as he took a real approach I think in data analytics and coding and how that applies to rangeland science as well as beef cattle production on rangeland um I know he spent stints all over the Western Us New Mexico Colorado then HED in South Dakota um where he did his pH New York at South Dakota State and is now you guys call it is it West River is that yeah West River yeah West River there in Rapid Cities where he's based out he's been doing some really really interesting work in beef cattle rangeland systems and intensive data collection technology and data analytics so but that uh it's all you yeah thanks well thank you Sam uh for the introduction thank you guys for having me today to come talk to you it's a real honor to be here first time in B heard a lot about over the years um so as Sam said my name is Jamie Brennan I'm an assistant professor in the department of animal science at souo State University and extension livestock raising specialist um just a little bit of background real quick just add on what Sam just said uh so I was actually born in Missouri and grew up in Missouri um and then I worked for about 10 years I did my undergradu worked at Colorado State University and then spent about 10 to 12 years working in Western rangeland systems um so New Mexico it's SP about five years at the horona experimental range down there worked in Nevada Eastern Colorado um and a little bit of work um kind over here on the Eastern side of the state as well and so I did my PhD and I completed it range science at SDSU in 2019 and so given that I had about 10 12 years of background collecting research data I really decided to focus I I had reading plot frames down I had that down half so I figured I probably read about 15,000 plot frames over that time um and so I really decided to focus a lot of my specialization data science and so anticipating with a lot of this Precision technology remote sensing tools that are coming online um we really need people that have the skills to be able to handle these larger data sets and handle them efficiently so we could actually turn those in a meaningful time frame into meaningful metrics that we could use to better improve management for livestock production on the landscape and so that's kind of where I ended up now in my current position um and as we kind of start moving towards uh Precision livestock technology for beat production systems and so we talk about Precision livestock technology there's a lot of examples that exist out there on the landscape right now a lot that are coming online currently these are things like activity collars that you put on animals that are able to detect different behaviors that these animals are doing um it's pretty amazing if you go into a modern Dairy facility and what kind of Technology they have available within those things like robotic milkers automatic feeders or Precision feed mixing machines that are out there uh smart ear tags weight detection cameras so cameras that can actually reconstruct that 3D image of animals get a mount of body weight body condition score off those um and even kind of geromes that kind of basically construct a 3D image of those animals that are out there from most Dairy system so there's a lot of tremendous advances that have happened primarily though been focus on Dairy and within feed loock systems out there so there's a lot less stuff that's been done I on extensive Range Systems uh for this topic and I kind of view he production on extensive range land system is it's kind of the final frontier for prision agriculture um there it's you know very even slow to adopt and a lot of that has to do of challenges and things like having power to devices right so if you don't have infrastructure for power you need to have solar or battery power to be able to drive sensors you have communication difficulties so if you have no cell service you need to be able to get that information signal out for that to be able to turn into some kind of management decision within some kind of time frame that actually Mak sense for the producer that's using the and so that's kind of where we've headed in the past couple years um what we're kind of focused on is this Cottonwood Precision ranching initiative that we began over the past five years out of our Cottonwood research facility and some of the Technologies we're looking at there we'll get into um you can see here is our SDSU wood Field Station it's located in Western South Dakota on pretty typical Northern mixedgrass Prairie communities out there it is one of the oldest continuously operated experimental stations in the nation and so we have I think it was established in 1907 we actually have weather data going back on a daily basis in 1907 we have a long-term grazing study that we've implemented out there since 1942 under three different stocking rates so we've overlaid a lot of these Technologies on that long-term stocking rate study that we have out there uh and really kind of our hope is we've got a lot of history and research associated with Stasia now let's take it towards 21st century and how do we start incorporating some of these Technologies into that production system and so some of the things that we're looking at specifically um on our research station include things like virtual fence technology uh manipulate livestock Distribution on the landscape uh We've also been utilizing smart scale so these are scales that we put in pasture that are put in front of your existing water tank and they essentially get a front end weight on your animals every time they go up and drink and then read an RFID tag on them uh we got Precision feeders so feeders that could uh measure intake of individual animals and they go to feed as well as you can program these to individual animals that talk about a little bit later uh where you can say you get three pounds of supplement you get five pounds of supplements tailoring these as opposed to the herd level to the IND visual level out there uh things like Green Feed we're measuring anic emissions of animals on pasture uh we kind of built a whole Armada of these now we've got 20 of these that are deployed throughout the state on our experimental station within feed lot studies on our uh campus facilities as well as on beef advice of producers across uh you know South Dakota as a whole and then things like remote sensing and grided climate data sets so this is kind of where I like to live a lot of times is that there's just a tremendous amount of free and open- Source data that exists out there in the world a lot of times it's free and open source data it might take a lot of programming skills to be able to extract that information and turn that into meaningful metrics for producers but I think it's a huge resource and I would say most of in range science we probably had much more advances in remote sensing side of things but this is where I really enjoy some of this work because one it's out there and it's available and two it's free and most other data collection is expensive right and so how do we better leverage these existing Technologies data sets that exist out there incorporate those into our decision policy and so overall when we talk about Precision livestock technology the promise of precision livestock technology is that we're going to increase efficiency so increase operation efficiency maybe increase beat pattle efficiency than the landscape we're going to reduce our environmental footprint associated with that maybe there less inputs that are going into our production system or through things like virtual fencing they were better able to manage or bracing or forage resources out there on the landscape and overall improve sustainability so whether these things are actually realized I don't think we're there yet but I think this is what is sold a lot of times when we talk about Precision mindology so here's what the promise is for these things what does that actually look like on a practic now um a lot of this talk I kind of decided to go in this direction of how do we actually at least in my opinion maximize the impact of these Precision livestock Technologies and so we're looking at things like the economics associated with the what's the practicality of deploying these Technologies on pasture systems but really I see the integration of multiple different Technologies together is really being kind of the future Precision Liv technology it's not just this one widget or this one ear tag it's this one ear tag that's maybe integrated with remote sensing data and then integrated with another sensor to be able to build a whole realistic picture of that animal production system that we're on out there in addition to integrating these multiple different Technologies together how do we incorporate these data streams to feed into AI or machine learning models to be able to develop predictions and then that last part there is how do we then take those things and feed them into animal nutrition models exist out there again the better build this whole production system as opposed to just studying one or single piece of technology and so that's what I'm going to talk about for most of the rest of this talk um some of the work that we've been doing on integrating these different Technologies together and maybe how we've been kind of thinking about this um as opposed to maybe more traditional you know economics or does it work and so this is a little bit of a confusing diagram but I really like this one so we had a Gra uate student named Anna dagel that had put this and she had adapted it for paper by Dr teski in 2019 of this conceptual framework of basically Precision technology and kind of when she was presenting this I kind of laughed she was uh mentored by myself and my colleague Dr Menendez I kind of thought it's funny that over here on this side you have data management data preparation and machine learning and so how do we take all of these thousands or millions or ions data points and funnel those into pipelines into machine learning models and then my colleague on the other side of that equation lives over here on kind of System Dynamics and mathematical U nutrition modeling so how do we feed all this information and there we can incorporate these into system syncing models animal nutrition models and this is unfortunately our poor grad student that's caught in the middle with these two worlds I'm trying to instruct them on this um which you done a very good job on but I think this really conceptualized a lot of the direction that we've been going pulling all these millions of data points processing them efficiently funing them to statistical models that then get funneled into nutrition models as well and so for that I'll give you a little bit of an example of one of the studies that our our grad student did Anna dagel and so she had a study that was looking at Precision supplementation effects heer development and reproduction on extensive range land systems and so kind of the premise for this work is that low input range based systems uh developing heers on Range can potentially reduce cost uh as opposed to developing on things like corn residue stocks or developing within dry out systems or some other potential benefits that have been shown where maybe you have improved grazing management or grazing behavor when it comes to the springtime that these animals been out grazing for that whole time period and so what you're able to do under these systems is better maximize an existing resource which is your dormant season forage and so this dormant season forage tends to be of lower quality according to Noah you could even get down to two or 3% free protein on some of these forages that I learned this morning from him uh so often times if you're going to supplement animals on these low input range based systems you of to require supplementation right we need to be able to supplement these animals with protein or potenti energy uh to increase and reach each reach your average daily gain or performance gance that you're shooting and so there's a lot of challenges associated with this if you're supplementing animals at the herd level these include things like over consumption or underc consumption um and supplementing at that herd versus that individual level so just because I'm putting out five pounds of supplementation per head per day if I'm feeding these animals in a bunk it might be that you're slightly more dominant but you're going to knock neighbor over and eat three pounds of their feet and so you're going to get seven pounds you're going to get three pounds you're going to get one and a half you're going to get six so overall at a herd level you're getting five pounds per head per day but at the individual level you're off that Target right we're not quite sure how because we're not measuring and so kind of the objectives of this study is one we wanted to be able to deploy some of these Precision feeders and smart scales and be able to train he ERS to use these things in pasture um it's really key for a lot of these different Precision Technologies is that you have some kind of training phase associated with them you can't just put out a Precision supplementation feeder and expect all of them to go to it use there has to be a time frame essentially when they're locked into a dry lot period where they learn to adapt and learn to associate like this is Handy machine that I need to go to be able to feed off uh we want to look at differences in performance and supplement intake for conventional versus supplement heers and then we want to develop a Precision systems model that integrates got a real-time daily weight gain data to prescrib supplement delivery to Target nutrient requirements for individual efforts right and so I'll show you what that looks like in a minut how this feeding trial was set up is basically we had 50 heers or 60 heers 30 in each treatment groups 31 prision group and 31 to what we call traditional bunk feeding system and they were fed uh 5 lbs per head per day of try to steal drink pellets and our goal was to try and Target the weight of these heers 60% mature cow weight at time of breeding so approximately 840 pounds pretty common management recommendation for our part of the country and you can see here um here's a traditional bunk fed group sitting around waiting to get fed and then this is our super smart feed uh producer unit they all call their units all about the exact same thing and mix up what they are but this is what our red wagon essentially as we call it and how this thing works is your animal goes up it'll put its head in there it'll read that RFID tag on slowly drop uh your distiller grain pellets until they reach their allotment for the day and then it'll cut them off for the rest of the day so they're not able to access and so we supplemented effers from November until uh May um in this experiment and then we deployed in those pastures as well these smart scale units that you can see there so uh within each pasture we had a smart scale unit to be able to track and collect that front end weight data for those animals um and then we did some calibrations across a bunch of different classes of animals to see how that partial body weight full body weight um we had an R squ actually about 097 for those we have a pretty good sense that there's a good correlation between those two so looking at heer performance you can see our average daily gain in our control there is about 1.93 lb per H per day um and then our our Precision group we had about 1.76 PBS per head per day um for performance you see here these are actually all the individual animal weights that we got over the course of that season um and really they start diverging kind of somewhere around the first couple days maybe two months of that production system where those trajectories of those two groups of animals start emerging as so if anos work um some of the bigger Tak homes associated with that is that uh our we have a big difference in terms of control of intake and so our control group that's supposed to be pounds but that's 870 pounds per head that we fed in terms of supplement to that control group versus that Precision group was 683 and so it's a pretty big difference in the amount of supplement that we fed between these two different animals and that's with both of them having essentially a lot of five pounds per head per day and so some of the things we found from that Precision feeder is that not all animals ate that full five pounds per day nor did they eat every day and so you have some animals that are what we call regular users that are going to go up they're going to get their full lot on the daily basis once they're cut off they're cut off we have some that are wor you know maybe they'll get three or four or five pounds a day not quite their whole lot before they go out and start grazing with the rest of them and then there's a subset of about five or 10% they might skip three or four or five days in a row they go to the sphere right and so they're a lot less interested in utilizing these Technologies and we're seeing that across a lot of our different technologies that there's always going to be a subset of animals that just are not interested utilizing these things putting their head in there and that has some consequences Downstream I think of what we think about in terms of selecting animals Precision technology maybe you're selecting animals that are more docile more easily trained on those that's kind of some of the bigger findings we saw there one of the big differences too within that Precision supplementation is that eliminated over consumption so you're not actually able to feed more than that five pounds per day I think we had about uh less than 1% where they did actually exceed that uh just through a software glitch in that program on average Precision group has significantly lower average daily gain than our control group and then we looked at pregnancy rates know we did have a significant difference in pregnancy rates it's likely just sample size in this was Focus for this but the ' 83% pregnancy rate in that control bre versus 67% in that so although that's not a significant difference it would be hard to sell that to the producer has a significant difference right that's a huge difference in terms of reproduction between these two systems and that's likely due to one these animals being out there within this feeder and maybe not getting that Reg supplementation even if I'm getting not quite five pounds per day at the bunk I'm probably still getting something every single day when we put that feed out there verus that feeder or maybe they're just selectively getting feed whenever they feel like it maybe they're nutritional plane on a lower basis and so what we did after um her trial ended is that we basically used the data that we derived from this first experiment comparing traditional versus bunk feed to develop a range supplementation model and so the kind of the the overall premise of this is let's say we have a heer and our current body weight is 694 pounds if we look at our current average daily gain from the smart scales we can basically project out that at time of breeding she's going to weigh 793 pounds as she's on this current trajectory and that's under 840 lbs Target weight right so we want them to weigh 60% of mature body weight a Time breeding and so there's our Target so we need to adjust how that animal is gaining over the remaining time period that we have there to be able to get them to that Target weate so that's essentially what our desired average daily is right so maybe our current is 1.4 pounds per head per day or 1.4

pounds per day and we need them to be at 1.7 pounds per day so under this scenario we probably want to increase that individual supplementation right because we want them to gain fast what we have out there on the landscape and that's basically what this model is um that we put together with our student there so how this model works is eventually you have kind of our first inputs here which go from our smart scale so these are our login credentials associated with our seock uh we have a start and stop date of when we want a query data so let's say we want to go the 1 of November until um what our current date is today that's when we want to stop for that data and then what's our trial and start stop dates we start on the 1 of November we're going to carry this out until June 1st and then what's our Target body weight and so what we did is from that adg model we wrote what's called an API which is automatic programming interface and essentially what that does is it's a piece of script or code that'll query that data direct from the cloud pull that data into our statistical programming software and basically calculate these metrics in real time so within a daily basis so we could estimate current body weight of those effers at the individual level what's our predicted weight what's our current average daily gain and then what's our desired average daily gain for the rest of that trial and then from there we feed this into an animal nutrition model we can say here's your current body weight what's your forg TDN what your supplement TDN and from that information basically kicks out of what's your anticipated FL whole body weight of gain so what's our predicted from the NM equations what's our predicted average daily gain at this level of supplementation and then that final component there is basically of the supplement optimization model and basically it'll run through and say okay if you need to increase supplementation it's going to increase it at 01 increments per kilogram until your desired rate of gain matches your full body rate of gain is this confusing or I see a couple thumbs up uh and basically all it is just iterating around in this cycle and trying to determine okay I want this heer who weighs 694 pounds I need to feed her 6.7 pounds per day be able to get her average daily game where I want it to be and so from this model um we basically for this first step is tested this on the heers that we had from that data set and so we applied this model on two we intervals uh to these heers and that's basically what this uh these these modeled results are for these different peers of what we should have been supplementing over the course of that time period um versus what we actually did and really it breaks down to kind of three different classes of of supplementation patterns that we have here so in that first one a you can see that red is kind of that fixed what we fed they're allowed five pounds per head per day or three kilograms per head per day that blue is basically what we fit based of our nutrition model what we did and so in this case we're probably over supplementing them a where we're giving them you know 2.3 kilograms per head per day but maybe that animal only needed 1.5 kilograms per head per day to get to the 840 so we might used excess supplement on those animals under that scenario that they don't need to be able to get that that Desir weight gain under be kind of that opposite scenario where maybe we're feeding them at 2.3 pounds

per day and they're not gaining fast enough and we need to be ramping up that supplement on that individual and then C we have a little bit of jitteriness there but it's pretty close to that line or we're supplementing pretty close to the level that they actually should be out there so I ran through all those kind of two week scenarios whether we're like supplementing with the right amount over supplementing or under supplementing we actually had about 69% of the time we predicted that we're over supplementing these animals and under supplementing them 31% of those feeding occurrences that we have and then so that's where we wrapped up with Anna's work and then we have a new student named hadly dots that actually uh conducted this experiment last winter and so we're working through that data right now but we actually implemented this model on these heers uh for this past winter we were dynamically adapting individual heer supplementation every two weeks based off on this model and so some of the early conclusions from that looks like yeah we did have actually an impact on those heers for that control group they were over by a pretty significant amount of that 840 Target with a pretty wide dispersive the final body weights versus that Precision supplement group was much narrow variation within that group U that saw as a whole and um there of other things in there that we're kind of looking towards the future of how do we better capture things like uh know bringing in weather climate API data to be able to P in that so if we need to increase energy for these animals because you're going to have netive 20 have weather how we start incorporating maybe some of these nutrition on overall it went pretty good last year um looking forward to student brighten that up presenting some of that work SRM this year um next kind of to illustrate these two um virtual fencing and how we're thinking also about integ some of these Technologies together is everybody familiar with virtual fencing here couple hands maybe not everybody um on there so virtual fencing 101 basically how this works is that you have a collar that's on these animals and this collar has GPS enabled on it and essentially it works like you would think of a dog shot collar able to train dogs virtual and so with this base station we have here we're able to communicate Thea software to these collars basically draw pasture boundaries anywhere we want out system right so I could draw a pasture boundary right here in the middle of this and as this animals slowly approaching that boundary it's going to get a beep if it continues it's going to get a electrical Cube uh on there to turn that animal around uh in that system and over time you could essentially train these animals to associate that beep with that shock sound and then better manage your animals on the landscape so we've been utilizing virtual fence now for about four years out of the Cottonwood station looking at things like what's the impact of like virtual rotation versus continuous grze and how does that impact animal behavior animal performance what are the economic cost associated with virtual fencing versus traditional fencing and so uh for student Logan kind of ped some of that work for us um some of the things found is is really over the past four years we found very little influence of virtual fence on animal behavior doesn't seem like it's really significantly increasing grazing resting walking time associated with animals on the landscape as well as no real differences in animal performance so that's kind of one of the benefits I think of this technology I think anytime you roll out something new whe that's a new grazing management strategy or a new necklace on an animal that's administering zaps to them you can anticipate a world where that might negatively impact performance and as a producer I would be concerned about what is the impact of all these things fundamentally on performance of animals because ultimately that's where I make my money um so we haven't found that really it impacts animal performance all that well and the other thing is that we found is that technology actually if it's working properly it can be pretty effective at controlling animals on the landsc now I couldn't get a bigger asteris than that on my PowerPoint this is a huge asterus right here of if the technology is working properly that that's not always the case that's not a case for a lot of Precision Technologies that they're always working prop so over the years we've had a number of issues in terms of things like collar retention on these animals we up to like 50% of collar have fallen off of them um battery issues stuff like that kind of ran through Dam of issues run into them uh but that being said if they're on if they're working properly you can't actually effectively control animal landscape and so there's promise there but there is quite a ways to go towards uh maybe integrated more fully on Ranch systems or maybe more turny I guess those and so for part of Logan's work that he did um he looked at uh two different Technologies one is these smart scales we deployed the smart scales out on pasture and then we had those virtual fence collars that we essentially treated for his study as um GPS collection devices and so these things were set at five minute intervals um and we actually got pretty good regular data off this virtual fence to for the first two years of this project and so one of the questions Logan was interested in um my student was looking at you know what is this energetic cost associated with raising right so if we have animals and feed lot we know basically you're brought your lunch every day and you got to go to B eat your feed we have an animal that's out there on the landscape right we have a heterogeneous landscape we have different topographies that are out there we have diversity of forages we have water sources so this animal has to go out graze on the landscape climb Hills climb downhills find its forage and then maybe it's got to go back to water after it's done all that for a couple hours and the point being is that the energetic cost associated with this is not zero right it takes some energy to go out there and graze and get your lunch out there on the landscape and so some of the work Logan did is that basically took this equation uh that was developed by Dr it's looking at net energy for maintenance activity and so we've got net energy for maintenance requirements and then on the top there what's that net energy for maintenance activity how much energy are these animals expending to go out through in the landscape and Grace and collect that forage out there and how can we potentially estimate that using some of these different Precision Technologies have available to us and so you can see from that equation kind of the big components there is you these coefficients of the derived period from studies and you have resting times of big components you have kilometer flat travel so that's travel across flat lands like you would see on this floor here if I was walking across it and then we have kilometers ascending travel so if I'm walking up that mountain that I see over there that's going to be an increased energetic cost to be able to climb that mountain be able to work walk on flat Landscapes and then that last part is that full body weight for those animals that we have there and so from this we basically d dve these metrics uh for each individual animal um based off one the GPS collar so from that GPS collar it's fairly easy to actually parse that data out into resting and moving behaviors uh associated with that so we could say from that GPS collar we could estimate hours per day of resting for that individual animal and then we could participate part set out to either flat or ascending travel based off of overlaying this on a digital elevation P so reach sequential GPS point we can see did it increase or decrease in one meter of elevation and from there you can pull it into ascending or flat travel and then that last part from the smart scales is that we're able to estimate daily full body weight on these individual animals on the landscape so basically how do we pull all these information and Technology together feed them into this animal nutrition equation try and build a better picture of that energy for activity associated and that's basically what uh this is here is your net energy for maintenance activity cost per day um these are individual heers across 2021 and 2022 steers I apologize um what we found is there's really no difference in terms of stocking rate rotation versus continuous grazing strategies within those systems but um kind of the point here is can we take all these data points technologies that are delivering these and actually turn these into an estimate actually meaningful in terms of energy expenditure on pasture for these animals and then lastly another one we're talking about that I've been working on uh for a little bit is how do we develop a Precision grazing management model on these and kind of thinking along the same lines of we' got collars out there we' got smart scales we've got remote sensing data how do we start incorporating these pieces of information together to be able to tell it fig and so for stocking rates uh you know stocking rates are hugely important for the field of range management uh it's basically an estimate of number of animals per unit of time for unit of land out there and they're really important to be able to make sure we're meeting our forage utilization targets so we want to take half leave half or if we want to over underg Grace for whatever management reason we need to have an appropriate stocking rate set to see if we're actually hitting our management targets uh minimize overg grazing associated with these and like I said meet those management objectives now the problem with stocking rates right is that one we tend to set them and then not revisit them years so even though our cow has gone from a th400 pounds right doesn't matter wash at the end but really the problem for for stocking rates is they require knowledge kind of three things right they require knowledge of forage production they require knowledge of animal body weight they require knowledge of dry matter intake to be able to accurately calculate stocking rates problem and all of these things are Dynamic Through Time right so if we have stalker steers that are out there your steers are gaining weight throughout that time period and maybe you could estimate what a midseason weight is and use that for stocking rate calculator but again your weights are Dynamic Through Time your forage production is often Dynamic Through Time right as you go through the growth season you have rapid growth in the spring then you start sessing and dropping off um after there and that quality is also changing As you move through time and then dry matter intake you know we tend to apply maybe two and a half% of body weight to that call that dry matter intake good but that can also potentially be dynamic Through Time whether that's heat Associated or other factors be driving changes day and so kind of the objective of this and kind of going back to the figure of of how do we better uh Drive Impact from some of these Precision Technologies is is really let's develop a machine learning model so we can make forage quality and quantity predictions on the landscape right so that's going to be our forage prediction model over there how do we incorporate smart scales to estimate daily animal performance within this system how do we then feed metrics that were driving off of those into a dry matter intake model so this is a model that ENT essentially is deriv froment equations that's toate for its dry matter intake and then can we apply this to determine you know when we should rotate animals essentially on the landscape based off these factors and so for this forage prediction model um basically wrote apis kind of similar to the smart scale where we're pulling in uh data now from Google Earth engine so we're looking at things like Sentinel imagery that we're pulling off of these as well as prism climate data and then creating metrics things like Growing Degree Days hum precipitation how far does that deviate from normal um kind of stuff like that that were then feeding into a beiji random forest model uh to be able to predict uh forage quantity and forage quality over time I've actually improved these predictions a bit since I made this figure but generally speaking we do a pretty good job of estimating biomass as well as estimating ADF systems so from these we can then extract and estimate daily Forge biomass predictions and daily ADF predictions now the other part of this is that smart scale model and so from these smart scales that are deployed in pasture basically able to get daily weights on these animals clean up kind of bad erroneous data points that are within those systems and then from this it fit spline models for each individual animal and from that spline model we could then estimate daily weight and average daily gain off of those models for animals that are out there one of the reasons we don't spine models is partially because it's able to capture nonlinear Trends have up here as well and so we tend to assume right if you matter your adg in the traditional sense you get a weit at the beginning of your range trial and a weit at your end of your range trial and you've got a straight line that just connects those to your average daily this maybe takes a little bit more of that Nuance in there using that daily data that they have available and then we feed these into our dryer intake box and so this takes the inputs of ADF which is coming off of our forage prediction model on a daily basis we have body weight at time T body weight time t plus one so that essentially be today and yesterday for those and then days on feed within their system would be one we calculating feed those through Nome equations I'm not a ruminant nutritionist I can get these you really need to see that was verified by ruminant nutritionists that are these are equations that we're using that were essentially calculating dry met intake for an individual level on an individual Day based off these metrics in our model there and so what this looks like on a practical standpoint you can see at the top there is basically our estimate off of our remote sensing model what's our daily ADF uh over the course of the grazing season you can see there what's our kind of uh spline estimated weight for these steers that are out there on pasture throughout the whole season and then that bottom one is essentially is what is that daily dry matter intake estimated for this individual animal and so we can start building some of these things and so this is for one pasture what's our estimated DMI for each individual animal accumulative DMI as move throughout That season there right and so generally speaking when we do this I was kind of surprised but yeah almost every time they're between two and 3% es based off these metrics that we're feeding in there so I think you know DMI on Range lands is kind of this fully grail I guess uh I will say of being able to calculate it but I think we're well within the realm of reality we're kind of putting some of these bottles together and so from this cumulative forage consumption we could then go to this next part is essentially a grazing utilization so how do we take inputs things like number of hectares and percent forage allocation and try and determine when should we rotate animals based on quality model parameters other so what does this look like well here if I took all those individual steers that are in that pasture of that cul dry matter intake and I basically add them all together as for a whole herd and how much this herd is intaking is what that red line is there if you look at that green line that's essentially Forge that's been allocated so our total production let's say we're going to allow 25% Harvest efficiency associated with that uh for animals and basic basically when those two lines cross then we've consumed as much forage as we allocated to them on the landscape and that's when we should be rotating out there and so this is kind of what i' say is maybe our business's usual method where we could say all right here's our pasture we want to break this into three rotations we're just going to use our long-term average forage estimate of you know 1,700 pounds breaker that's out there or maybe I pulled that from the web soil survey to get an estimate of my forage production and then what's our midseason weights for our animals and we're not really accounting for its Dynamics but we're just going to do about 25 days per rotation when we calculated as kind of our business as usual versus a dry year which is the 2021 and you can see there in rotation one we' get about 23 days worth of brazing so pretty close um rotation two we'd have about 15 days and rotation three about 12 days estimated from our model when they and so what's happening here is one is you're going through your animals are increasing in weight so we're incorporating some of those Dynamics into this our forage is sessed and dried out and maybe you're starting to lose biomass that's out there on the landscape and so do some of those Dynamics out there um really trying to better capture these dynamics that are happening through a brezing season on the landscape conversely if we looked at a wet year like 2023 we had a lot of production and I think looked like we barely even got on top of the graphs out there our first rotation under this is about 63 days we have about 3,000 pounds per acre which is phenomenal over Western sou um and so again you know this largely a modeling exercise at this point I think until we test of experimentally but it is showing that we are capturing some of these dynamics that are on maybe we need to be better paying attention to these especially when we start incorporating some of these Precision Technologies and so whether it's the heer supplementation or whether that's through stocking rates um you know there's there's a lot of work how can we integrate some of these virtual fencing but how do we improve grassland management and more closely hit our Target with these Precision Technologies right so if I'm under a normal stocking rate let's say I said it I've said it for the past 40 years and you're going to have a wet year or maybe you're under stocked and you're going to have a dry year where you're overstocked and all these other years because you know we never have a normal year doesn't EX landscape and so if you average all of those over that whole time you're probably going to hit that Center Point right you average all of those points but within any given year you're missing that Target that's same the heers within any given heer you're probably missing that supplementation individual so how do we move that Target using Precision Technologies better hit that or at least get closer to it um as we start incorporating some of these Dynamics and so a lot of this there's still a lot of work um evaluate how close we actually are to these targets I'm not a modeler that's that's kind of Hector's World sometimes where you can take that and run with it but I like to be I'm more of an experimental guy to see how close we actually are to the mark on some of these as well as I want to evaluate some of our future work is potential for compounding ER so these different modules from different technologies that are getting plugged in there is a potential that you're going to add error from here add error from here add error from here and how is that getting compounded as we're moving throughout these whole models out there it's kind of some of our future work now um I think we got about 15 minutes here maybe five I don't know depending on how many questions ask but as you'd imagine there's a lot of practical considerations you think there's a lot of practical considerations for cion technology anybody here used a lot of precision technology all right we're going to have a support group meeting a later actually not it's all sunshine and Roses right all works 100% of the time nonstop fail right yeah that's yeah that's not the case there's just a tremendous amount of limitations to implementing these things on extensive production systems um and I i' like to think we've learned them all but I think every year we learned some kind of new limitation or a new thing that we didn't think of that failed um that we really should have anticipated or maybe the manufacturer should have anticipated uh potential point of failure throughout there uh one animals need to use the devices and use them correct ANS don't always have the most respect for research that we do out here and so if you're a smart scale if you have a wire and maybe this wire was slightly exposed and they didn't touch it for the first 74 days of your trial and then one day effort decided oh I could reach up gra that wire RI all of your dat all your machine is all uh basically not functional uh this is one of my favorites from the early days so this is an earlier version of the vent virtual fence towar um that had these nylon straps and these are essentially what administered that electrical queue within that this thing gets flipped around it could be sending zaps to the cows come home L they're not going to come home on the landscape uh in addition the sound the speaker is now nestled up against their neck so they're probably not hear the be associated with that either um C failure was a big one like I said I think the first year with these we had about 75% retention the second year I think it was about 46% calor retention so about 54% of them fell off uh over a three Monon raising period which is not an insignificant amount that is out there it makes system much less effective um third year we altered them and that work about 95% eff so it is possible to keep them on better but again trials that we were learning as we're going through uh for all this equipment climate is a big factor you know we're in the Northern Great Plains we're routinely in the summer we'll have days where we'll hit 105 degrees winter time we usually get down to30 40 degrees out there that's a huge range of environmental conditions that you putting equipment in that it's got to be a little work under those scenarios right so it's very hard on equipment batteries hard power to be able to operate under those different environmental systems you have communication protocols it's always a challenge on Western Range lines that's one of the biggest hampers for a lot of these Technologies is that it's one thing to be able to collect that data if you're actually going to use that data to inform management you need to be able to get that information out and be able to turn it into a management recommendation within a time frame that's meaningful right if I found out that uh I before prediction model and let's say okay here I am in December and I have 2,300 pounds of forage produced this past year can I do anything with that information maybe a little bit but it doesn't really help me at the time I've already raised managed system now you're telling me after the fact here's what my for is so timely delivery of information really requires good communication protocols and that's a challenge when you don't have cell service you don't have power on these extensive Range Systems um and so communication is always a challenge battery life uh is a big one um you know some of these virtual fence collar we found especially if you're in much more intensive rotations or they're interacting with those boundaries more frequently your GPS is on a lot more you're giv more stimuli you potentially could drain a battery in a couple to maybe multiple months of your more extensive production so battery life is always going to be challenge data access platforms um you know a lot of this let's say if you want to utilize different data different Technologies in a dream world it would all be standardized in the same format we acquire it all the same that's not the case if you got five different technology platforms you have five different potentially data delivery systems to get your information that's always going to be a challenge and then labor we can't discount This and like we said in the beginning the promise of precision technology is that we're going to save labor make it more efficient these things take a tremendous amount of time and effort that labor cost is not zero right these aren't tury where I just put these collars out and I don't have to learn the system they're just going to manage their animals there there's a lot of effort that goes into be able to learn the system the animal behavior put these devices together um you know for a green feed units we've got two technicians that are employed full-time just to keep Green Feed units running across the SA and that's with the expensive data package plan maintenance plan with the company and so the labor is not zero on these things it's really important to consider if you start getting into this what is that cost associated with that and then programming skills um is a big one and etc etc we're still learning every day we're still learning tricks and tips for these Technologies programming skills are really kind of maybe how Sam and I got together to begin with um and so this is a quote that I don't know if I ever said it often times contributed to me but we have unprecedented amounts of data that we're collecting on Range Lanes generating maybe a billion data points a summer now and that has a huge learning curve to be able to efficiently process this data and information right you can't just open this up in Excel and expect to start making Headway a lot of these larger data sets and so if you're going from maybe 600 animal weights um total data points that you were doing at the beginning in the end so now we've got you know 47,000 animal weights on an individual level that all need to be cleaned and processed and handled efficiently that's a lot of a learning curve that happens there and so it's something that I've really been promoting uh it's the value of Open Source Code and not just open source code but the value of sharing that code between researchers and research entities and so we've got a couple papers out there so if you're utilizing like sea loock equipment for example we can publish that API dat code so you can go in there and automatically pull that information off the website and calculate your metrics we use it most frequently to be able to just determine if the equipment's working properly so we essentially get daily reports for all of our systems and let's say I have a green feed unit and I was getting you know 50 good data points per day then all of a sudden that drops to zero and then it drops to zero the next day now I know I've got some kind of problem I don't know what it is maybe the maybe it doesn't have any feed in it maybe the system shut down and failed you know maybe they don't have access to it the the ranch manager move pastures and didn't tell me and now my system's out there about animals but again developing these protocols be able to pull that information turn it into meaningful management insights and then troubleshoot things happen so you don't find out a month or two into your trial that your equipment was time which is pretty terrible feeling um I don't know if anybody commiserate but uh yeah other things uh developing open source platforms and then how much time and effort that saves and So within our cow CAF unit on our campus we have an incenting system uh that basically measures Dr matter intake and water intake for for animals and feed lot um just wrote a quick code for a grad student and they said that probably saved them about three months worth of data cleaning and they ran that on all the past three years of incen data that they had there and they estimate it probably would have taken graduate super labor over a year to be able to go through and process and handle that and so these are the kinds of things or if you could develop these codes if you could share them with people that are also working in this field um I'm a big proponent of this and you ever say hey I'm thinking of using these Technologies reach out to me I'll sure code happy to do that um if it saves people labor because then you can ask more interesting questions if you're not just spending all of your time data R other things uh understanding the biological system is important these Technologies are never going to replace good fundamental knowledge of range management and animal science it's just not going to replace things and so even though something might look good in an algorithm and let's take our heord uh supplementation model if my model my model doesn't know that I can't go from two pounds per day to 10 pounds per day that potentially having some kind of consequence that animal health right and so we need to as animal scientists be able to think about what are some of these limitations that even though an algorithm spits out a number should we go with them maybe we should Step Up supplementation every 10% for three days until we reach that high level right so it's really important especially too when you start working with some of these companies that maybe they're have a really good talented engineering staff that maybe don't fully understand the production system that they're trying to apply that engineering stuff to so they need to know that hey you might tip an animal over if you increase the supplementation that fast on there so it's really important to understand your biological system and then a virtual fence there's a whole bunch of considerations things like GPS here um you know you might think I'm outside the boundary I'm in the management zone or I where I'm supposed to be and maybe I'm just getting a 10 or 15 meter signal that or I can't virtually fence out access to water right it's Common Sense thing but it has happen where somebody's put in a virtual fence boundary because they can just draw it on the screen anywhere and then say hey how are they going to get to this water that's outside of there that's not going to be effective at training those animals or maybe my manager doesn't know where this virtual fence boundary is and she drops B A over here and so they all run and just break right through get hey and now your system so really takes kind of a different level of thinking or retraining your brain how you're going to imp getting a little long on time uh I will say you know extensions role in Precision ranching there is just a huge amount of companies that are getting into the space technologies that are coming out Flames that I am less comfortable with than maybe others are within these spaces and buzzwords like Health metrics real time data driven analytics right AI powered AI power could just mean a linear model that's running behind the scenes we don't know it's black box but there's a huge amount of companies that are getting involved in space um and I think technology a lot of times is outpacing maybe the research is being able to evaluated and I worry that more than a few of these might become snake oil that's being sold out so it give be a challenge one to keep up with current technology but actually evaluate what are our stakeholders needs and so really icr role not just from research from an extension standpoint as one cinate research and education to stakeholders but really be that mouthpiece to be able to provide unbiased information associated with these technology because we've had a number of different virtual offense Field Days um and just really talk to them about realistic picture of if you're going to implement virtual offense what's the labor cost what's the time to learn the system how effective is it actually on the landscape you what's colar retention one of the things we kind of did at our most recent ones we actually collored animals for producers to show them that process that one you need good cattle handling uh facilities to be able to handle put a collar on an animal without hurting yourself and two so it's kind of a Counterpoint so if you see a video of maybe a nice 4 cow that's getting a collar placed on it and it just nonchalantly saunters out of the shoot versus maybe some of our more rangy cows that are really fighting and bucking and and get a picture of that actually so trying to paint that realistic unbiased information picture for producers that might be considered that I like to thank you know I'm one person talking out here the representative much larger team sponsors that are out there just like to thank all the funders and acknowledgements there and if you have any time have to answer question for your heer study did you notice any differ any differences inat two groups correct weights birth weights weights births sorry birth weights I'd have

2025-01-17 15:15

Show Video

Other news

Google's Quantum AI Is Actually Mind-blowing 2025-02-09 16:23
Arm Disappoints With Outlook, Vanguard Slashes Fees | Bloomberg Technology 2025-02-08 20:01
Primitive Technology: Flywheel blower smelt/Monsoon begins 2025-02-07 20:07