In this episode of the Real Science exchange pubcast Scott leads a discussion on cow monitoring technology and its benefits for cows, farmers and the environment.
Guests: Evine van Riemsdijk from NEDAP Livestock Management and Stefan Borchardt from the Free University of Berlin
In this episode of the Real Science exchange pubcast Scott leads a discussion on cow monitoring technology and its benefits for cows, farmers and the environment.
Ms. van Riemsdijk gives some history of cow monitoring, stating it started for identification of an animal and the feeding station for the purpose of separating milk. (6:58)
Ms. van Riemsdijk said the sensors help see heat behavior as a whole in your barn, they help you find a scout and who has shorter heat periods, when heat starts and helps calculate optimal insemination time, even when you are not in the barn. (15:02)
Mr. Borchardt said that a major drawback of the industry is integration, bringing these different technologies together into the herd management software. As an industry, how can we bring data and programs together to make smarter decisions? (26:29)
Ms. van Riemsdijk said the sensors can be used while breeding cows. It can also be used as an intervention to understand why cows are not cycling correctly. (39:11)
Mr. Borchardt said that when farmers adopt the sensor system, they realize there are some cows already showing a health alarm and most of the time they wouldn’t realize these cows are sick without the sensor. (47:23)
In summary, Mr. Borchardt said that with farms, sensor technology and increased genomic data, we can get to a place where we are managing cows on an individual basis like precision feeding and reproductive management. (52:14)
Ms. van Riemsdijk concluded by saying you can use data points from the sensor in heat behavior in the voluntary waiting period to make more strategic decisions with other experts on the farm. (58:40)
Please subscribe and share with your industry friends to bring more people to join us around the Real Science Exchange virtual pub table.
If you want one of our new Real Science Exchange t-shirts, screenshot your rating, review, or subscription, and email a picture to anh.marketing@balchem.com. Include your size and mailing address, and we’ll get a shirt in the mail to you.
Scott (00:00:09):
Good evening everyone, and welcome to the Real Science Exchange, the pudcast we're leading scientists and industry professionals meet over a few drinks to discuss latest ideas and trends in animal agriculture. Hi, I'm Scott Sorrell, one of your hosts here tonight at Real Science Exchange, and we're here tonight with two expert guests to discuss the importance of cow monitoring and how far that technology has come in recent years for the benefit of the cows, the farmer, and the environment. I first like to introduce Evine van Riemsdijk from NEDAP Livestock Management. I first met Evine in Israel at the IFCN conference where she gave a presentation on Dairy Cow technology. I thought she was a great, would be a great guest for the Real Science Lecture series and indeed she was. Evine, welcome. Glad to have you here at the Real Science Exchange and as is customary what's in your glass tonight?
Evine (00:01:02):
Well, Scott, thank you for having me after having the joy of doing the webinar and following up on this pubcast. I know it's typical to have a good drink like we are in a bar. I have to be honest, I'm drinking a cappuccino today cause I just came back from Oceania traveling New Zealand and Australia for two weeks, and by pretending I don't have jet lag, I might have one if I start on alcohol now. So I keep it on the caffeine with some good dairy in it.
Scott (00:01:30):
Very well. Evine, I see you brought a guest with you here tonight. Would you mind introducing the guest you brought?
Evine (00:01:37):
Yeah, so this is Stefan Borchardt. We've met about a year ago, I think for the first time. I joined NEDAP years ago, but Stefan has been working for a longer time already with NEDAP sensor and their data. And yeah, we had the pleasure of having Stefan and his colleagues visiting us at our NEDAP facility to really do brainstorms with both our data scientists from both teams. Very exciting. And he's now starting with the publications on his research. And yeah, the last paper you just published on the webinar is very, very interesting. So I was hoping you could explain a bit more about this paper during this pubcast.
Scott (00:02:18):
Yeah, yeah. Excellent. Looking forward to that discussion. So welcome Stefan and what's in your glass tonight?
Stefan (00:02:24):
Well, first of all, thanks for having me and a nice introduction. So I have a classic from Berlin, it's called Berliner Pilsner. So it's a typical beer that we drink here and I can prove that we had some, also some fantastic Dutch beers when we visited Evine and NEDAP in the Netherlands.
Scott (00:02:45):
Ah, there are some very good Dutch beers and Belgian beers, which is not all that far away. Anyway, I'll get into that in just a little bit because I am having a Dutch beer. But before I get started with that, going invite back a good friend and valuable resource to the Balchem team, Dr. Ryan Ordway. Ryan, thanks for joining us tonight. And what's in your glass?
Ryan (00:03:08):
Well, Scott, I am in my headquarters office, so I am double fisting it with a coffee and my for those in the US and the southern US, my Bucky's Beaver from the Bucky's famous road stop mug, but it's just water today, so.
Scott (00:03:32):
All right. Very well.
Ryan (00:03:33):
Unfortunately, the CEO doesn't, he doesn't quite agreed to letting me drink in the, in the headquarters office yet, but, we'll, I'm working on it, so
Scott (00:03:42):
We'll get there. All right. Well, in honor of Evine, I'm having an Amstel in a traditional pint, a glass, which I'm customary to drinking out of when I'm in Europe. So I had to get some while I was here at home. So anyway, in the spirit of the pub folks let's raise our glasses to Evine and Stefan. Here's to a great pubcast. Cheers.
Speaker 5 (00:04:04):
Tonight's pubcast stories are brought to you by Reassure Precision Release Choline. Reassure is the most researched, encapsulated choline on the market today, consistently delivering results to your transition cows of higher peak milk, reduced metabolic disorders, and even in utero benefits to her calf leading to growth and health improvements. Visit balchem.com to learn more.
Scott (00:04:26):
So, Evine, you had a terrific webinar back in April during the Real Science lecture series, where you shared some great ways that NEDAP is focusing on cow monitoring to improve the life on the farm for both the farmers and their cattle. I'd like, if you would, can you give us kind of a brief overview of how NEDAP is helping farmers to improve cow health and profitability?
Evine (00:04:48):
Yeah. So need NEDAP’s a company founded in 1929, so almost a hundred years being around but their expertise really focused on dairy since the seventies. So initially it was cows moving from a tight stall barn to a freestyle barn, and there was a need to identify the individual animal to be able to feed it individually to its needs in lactation stage, in its concentrate feeders, and has developed in the past 40 plus years to adding heat detection because the animal's more active during heat, which can be picked up by the sensor. And eventually also more automations like robotic milking, sort gates and also through newer center technologies, also having eating behavior, rumination behavior, other activities indicating health events or health deviations in these animals. Historically always been to help a farmer day by day, identifying a cow, putting a soffer rule if she could eat or be milked knowing when her heat was optum, insemination time, or having that deviation in and behavior to act on a cow, which might be unhealthy and needs a checkup.
Evine (00:06:02):
But now at Farmers having more challenges and data can be key in that strategic decision making, reaching out more and more to other experts in the fields who actually can help farmers with more knowledge and combine all these information dots. For instance, the academics, the veterinarians, nutritionists. So how can they actually build more context around what sensor data is providing? Initially trying to connect all the dots with all these new technologies arising. So Farmer really has a full complete picture and can think longer term in their strategy, the rest of the future.
Scott (00:06:39):
Yeah, interesting. Looking forward to digging into that a little bit more here in a bit. One of the things I'd like to start off with during the webinar, you talked a little bit about the history of cow monitoring and wonder if you could kind of just kind of take us down memory lane real quick on, on how we got from where we started and to where we are today.
Evine (00:06:58):
Yeah, so initially it was a typical standard just for identification but always been a solution in a bigger tool. So a solution to identify an animal, the feeding station and the feeding station would provide feet or a solution to identify a cow in a milking parlor, and then you would know which animal is being milked if you had to separate her milk or you did combine with milk meter data. Yeah. And it is then grown into heat detection with optum insemination times and eventually with help according to the other thing, what is how building these standards globally is, is integrations with farm management softwares, because it's really hard to make decisions on each individual technology or tool you'll have. And that has really helped us scaling.
Scott (00:07:45):
Now I remember back when I first got into the industry, we were introducing computer feeders at the time, and they were, they were all the raids, right? And it was all about identifying individual cows and trying to match where they were in their lactation cycle or, or their size, their how much milk they were giving and then try to feed them specifically. So I'm gonna assume that was one of the first stops on the way of the evolution of this kind of technology.
Evine (00:08:16):
Yeah. And, maybe it's partly also I was founded in Netherlands. It was a very practical challenge the farmers faced when they moved from tight stall where it was like 10, 20, 30 cows in a barn, and they moved to scale up to freestyle barns. And there were some metabolic problems because if you would feed the same ration to all cows in different lactation stages, and that's where these feeding boxes or computerized feeding stations were introduced. And then there was a need to still identify that animal at that moment, at the feeling station to have a rule how much she could eat at what times. And so that was the first practical need to have identification of a cow in the freestyle barn. Yeah. So she could still be the cow wonder group, but individually managed.
Scott (00:09:03):
Yeah. I'm surprised to see during some recent travels in Europe that those are still quite popular over there and not so much here in the states anymore. I'm kind of curious, the technologies that you guys have, are they, are they intended for just large dairies, or what are they more, most appropriate for?
Evine (00:09:21):
So what we commonly see, it starts being introduced dairy farms from 55-60 cows. So any size where, for instance automated milk systems like milking robots are introduced that the start where a cow has a identification plus heat and health monitoring. It's commonly followed that these technologies are more suitable for larger farms. But actually the biggest return of the investment comes in smaller farms, especially on the heat detection part if you miss one cow. And if you miss one cow out of a 60, or you miss one cow out of 200, the impact is much bigger or missing one cow out of a 60.
Scott (00:10:01):
Yeah, makes sense. You know, I was just kind of thinking today, been hearing a lot about Chat GBT artificial intelligence and kind of curious, what role does AI have in your systems today? And then maybe Stefan, this is a question for you. What do you see AI playing? What role will they play in the future with this kind of technology? Yeah,
Stefan (00:10:23):
So I'm a veterinarian by background, so I think our yeah, our experience with large data is limited very quickly. And that I think was one of our smartest move a couple years ago that we hired a data scientist to make sense of all the data. All the volume that we get from the sensors. And I think that's, I think one of the limitations that we also see in our curriculum that most of the vets, they have no clue how to deal with these kind of data. And what is interesting, I think for us is, especially this communication with data scientists. I remember the first time I had to explain him what is a cow in heat? And, and I think now he has a clear understanding, okay, what's the physiological background behind that? And I said, I think also he was able to teach us some very basic elements of coding. And I think artificial intelligence is becoming more and more an issue for us because with these large data volumes, we can apply some of these AI techniques to make sense about the physiology. But I think for us usually the biology of the cow is on the first spot. And then I think we've, we want to yeah, make sense of these data and these kind of techniques can help us.
Scott (00:11:56):
So you mentioned biology. So, the technology we currently have today is measuring behavior and then translating that into what we believe has gone from a bot biological perspective. Do you see somewhere down the road where we'll actually be measuring biology or will we continue down the path of measuring behaviors?
Stefan (00:12:16):
Well, I think right now, if we think about these accelerometer systems, I think we measure behavioral response and then we translate that into something that is actionable for the pharma, like potential disease or heat. But I think there are several other technologies available if we think about inline progesterone measurement or probably inline measurements of some substances that we don't know yet. So I think the future is bright and I think what we, what we see today, even only with these accelerometer systems, I think that most of the potential that they have is unused. So most of the farms, they simply have them as an alert generator for a cow in heat or a diseased cows. But I think as we have shown in this study that being mentioned, but also in a couple other studies that there's more into it things like estrogen intensity or resumption of estros expression after carving. So I think that's very interesting area right now. And I mean there are also a couple scientists in the US hardly working on it. So
Ryan (00:13:36):
The sky's a limit. I mean, as we talk about precision as a nutritionist you know, as being focused on nutrition, it, you know, we're, when we talk about overall profitability on the farmer right now, we're probably spending a lot of money maybe, or losing a lot of money, either underfeeding certain cows on amino acids, on choline, on minerals, things like that, maybe overfeeding on some other ones. And I think, you know, as we get more of this data and can really aggregate it together and get closer to understanding that biology, it's really gonna unlock the potential for precision feeding, which is better for everybody, more profitable for the producer, and certainly healthier and more productive for the cow. So very exciting.
Scott (00:14:26):
Yeah. Evine you had part of your, your presentation title was revealing her secrets, and I found it interesting that you talked about the fact that a cow is a prey animal and so that they're very good at hiding those secrets. Right. And because they don't want to be eaten. And I'm curious, so what kind of, what buckets would your technology focus on in terms of what areas do you wanna reveal those secrets? And I heard Stefan talk a little bit about reproduction. Where else are you guys looking to monitor and reveal those secrets?
Evine (00:15:02):
Yeah, I would say that heat shield epxressed, that's a pretty strong hormonal response. So that's what you will still see when you're in the barn, given that you're in the barn. So that's where sensors really help you to find a scout, have maybe shorter heat periods and find them more maybe at times when you're not in the barn or at night. Plus the sensors help with explaining when did the heat starts and when it reach a certain level to confirm it's a heat and it helps calculating the optimal insemination time. So it's not really cow hiding, but just helping you see our whole heat behavior while you're not always there. The secrets we're revealing is, like you say that, that cow being a prey animal, so instinct, even if she would feel weak, is try right to stay with the group.
Evine (00:15:50):
Act like the others in the group. Try not to stand out, interestingly, because she now carries a device which tracks her 24/7 and that device actually learns about each individual animal's rhythm. So we see even in the same farm, on the same management and feeding times, every cow kind of has her own rhythm. And that might has to do with her age with a hierarchy when she's eating, when she's resting, and then also when she's ruminating. So it doesn't matter what she does in the day, as long as she's very consistent. Because cows are actually very nice to predict cause they're so consistent. Revealing her secrets means the sensor will also pick up when her eating time, the eating time deviates, her rumination, time deviates, maybe her inactivity goes up or she shows other activity. So even though we're still not diagnosing what she has, we are just creating the indication something is off because she's having more inactive hours, she's eating less, she's ruminating less. And depending then on the checkup by the farmer it can be different source of health incidences. Sometimes it means she stops eating very rapidly. That's what we see with those displaced omasums. Sometimes only rumination goes off. If it's something maybe with her leg she might rest more. So we can still not diagnose what she has, but at least we create an alarm to someone at the prime check on her cause something is off in our daily rhythm.
Scott (00:17:22):
Do you see the point where we will maybe through AI have the ability to diagnose the problem specifically other than just saying, Hey, there's a problem here.
Evine (00:17:33):
Well maybe Stefan can even comment better on it. Cause I know he knows a lot about transition disease and he's looked into different behaviors around transition disease. What commonly see is they do deviates and the patterns look pretty similar, even with different transition disease. Stefan might be able to comment more on it.
Stefan (00:17:52):
I think we can see clearly different patterns of behavior depending on the disease. So as Evine mentioned, a cow with a displaced upper has a very strong variation. But if you think about other diseases like metritis, ketosis, mastitis, I think it really much depends on the severity of the disease. So if we think about the mastitis, for example, it might be just a very mild case. But if you think about something like toxic mastitis, this is probably a very severe behavioral difference. But what was interesting when we, when we look at the pattern, so most of the time we can identify these cows earlier than the hertz men can do it. And we see, especially for diseases around calving even some of the cows they show deviations in the dry period already. And I think that's really interesting if we would be able to predict which cow would will be sick and which cow will be healthy. I'm not really sure if we right now, if we are able to yeah, tell the farmer what is the disease. But I think the sensor is just a tool that can help you. But at the end I think somebody has to make a decision and has to has to make an examination or an a proper intervention. But, who knows what will be in 10 years.
Scott (00:19:34):
All right. I'm kind of thinking, you know, we've got technologies that can measure milk output, they can measure conductivity, which is maybe, you know, is related to mastitis. Maybe there's know some hormone detection. And so maybe as time goes on we'll be integrating some of those data into the behavioral data and be able to predict that much more. Sorry, yeah, go ahead. Yeah,
Ryan (00:19:56):
I was just gonna ask a question. So being that, you know, back on a comment about herd animals, is there, are you doing any predictions based on overall herd behavior? Because as you mentioned, Evine with, you know, an animal that gets sick and or not quite feeling well and hides her, hides her secrets so to speak. I mean, the rest of the herd I think will normally react to that as well. So are there any predictions, predictive data now or maybe that you're trying to gather in the future to look at other animals and see, okay, maybe they're not the ones that are sick but their herd bait is and it's impacting them out of, I mean, use the term concern, but as they are, you know, work is a herd. Is that is that possible today or in the future or,
Evine (00:20:48):
Yeah, so how, so how it's being used now is it would still depend and it's gonna be a bit tricky on data entry. How do you find where the cow is? She's wearing a sensor, but now it's about integration, fire management software. Is she assigned to the right pen in that farm management software? But on larger farms where you would say you have your low production group, your high production group, you dry, correct. As long as the cow, the data is entered correctly, she is in that pen. You can compare her to her to her pen mates and you still need a larger group to get a conclusion. So if it's a dry up and the three cow, yes she'll deviate. But how it's now commonly used actually on these larger pens, if she's in a high production group and, and she’ll deviates the dashboards will already tell you how much she deviates from the group average.
Evine (00:21:41):
So now you already know if she's the only one or if the whole group is at the same level. That can also actually combine the data points from different cows in the same group and create a group alert. And that will might tell you something about the feed being off or not being fed at the right times. Because if they go down in eating time as a group, then you know, there's something else you have to change. It might not be the health, it might actually be feed management. The other nice tool is on these dashboards you can actually show how many animals are eating at the same time and it'll actually help you optimize your feeding strategy. It'll tell you if there's always feed available. Sometimes on farms we see gaps. There's no feet in front of them between 4:00 and 7:00 AM because you see 0% cows or close 0% cows eating between 4:00 and 7:00. And then you see spiking after seven towards at 80-90% in which we, in the graph concedes feed delivered. So then that group data will definitely help you explain, hey, maybe we should push up the four feet more often. We should feed more often, we should feed fresh feed. They all stop eating and there's free feet in front of them. There might be something wrong with the feed. And that's where you guys are experts on. Right. That's definitely where I see role of nutritionist to work with these group patterns.
Stefan (00:23:00):
Yeah, so I think I can add to this, I think one of the really nice features is this group level monitoring. So before we always use time cam time lapse cameras to evaluate feeding management if we troubleshooting hertz. But with these sensors, I think you can, you can really dig into the feeding timing, but also probably the amount. And I think it's really helpful. And one anecdote that I always tell now is one of the farms that we work with, they applied a NEDAP device to the tractor that they used for pushing the feed. And with this tech, you can actually tell when they move the tractor and when the feet was pushed. So I think it's sometimes the farmers invent something that we didn't even think about. And I think that's really, really interesting for us to use the data, not only for the individual call, but also on the group level.
Scott (00:24:06):
That's funny. It's pretty good. Perhaps we should put the monitors on the farmers as well. Yeah.
Ryan (00:24:11):
Yeah, that's pretty pretty brilliant. One, one question I had and this is maybe jumping into a slightly different topic, but I'd actually, Evine knows at a conference in the Middle East and met some folks that, or it's actually a Dutch company that is a lighting company. And I was pretty. I got to speak to them quite a bit over the course of the week and it was really fascinating to me. I mean, of course that research has been out for a while in, in dairy with lighting and things like that. But is there any integration into some of your sensor data where you can actually look at that photo period effect with some of the new lighting LED lighting that we're installing in barns and being able to capture that data along with some of the other biological data that we're trying to move towards?
Evine (00:25:10):
Yeah, at the moment there are no practical applications, but we know it could be technically feasible, especially with more and more companies moving towards the clouds, we're moving to the cloud too, and it allows actually to create better connectivity between smart devices on the farm. So we foresee that fully smart barn where actually all the devices could talk to each other, and that could be the future. We are looking into, we have a cow positioning system, and at the moment it tells the farmer where a cow is in the barn, and it's very practical for robotic milking systems. There's always these few cows which don't show up. And if you do three, four row barn and you have to find her, that's quite a long walk through the barn to find a cow to fetch. So now it's practically being used very real time knowing where the cow is, finding her and getting her to the robot, but to work with different universities, actually oncologists who look into that position data of the cow. Does she have preferred spots in the barn? Have what, what's her daily behavior throughout the barn? Is there a social structure in that barn? So we can foresee those types of technology in the future? Being able to talk more to each other, that's still a long long way of development to go.
Stefan (00:26:29):
Yeah, I think one of the major drawbacks in the industry is the integration. So we are involved in the project where we have sensor data from an activity monitoring system. We have a sensor in several pens that's measuring light, temperature, humidity, CO2 and ammonia. And then we have a camera that's evaluating the condition of the cow, but also we are looking into lameness identification. But I was not aware of the problem to bring these different technologies together into the herd management software because I think the, for the farmer, it's a real struggle if he has five different screens and then he has to come up with a conclusion from these five different sensors. So I think they want to have a simple solution within their dairy com, PC data, whatever. And I think that's, right now, to me, one of the major issues is how we can bring these data together to make smarter decisions. And I think your example with the light is on spot. So I think really, really interesting. Yeah,
Scott (00:27:45):
I think that's where AI can really come into play, right? I mean, it's not just bringing the data together, it's then interpreting this mass amount of data and what does it mean bringing meaning to it. So anyway, go ahead Ryan. You had a question?
Ryan (00:27:58):
Oh, no, I was just gonna say, I don't know, that was one, one thing of a, I think we were talking about at the we were together at Euro Tier is, you know, the challenge I always had and I'm retired from feeding cows. But you know, when I was out on farms every day, it was you know, it was always a challenge. And this was back before we had sensors, but
Ryan (00:28:21):
Being able to make sense of all the data and the, and the challenge I always had is, you know, if you had a you know, didn't have the a farm management team, you know, you would have the, the veterinarian who was gathering data, you would have the reproductive company gathering data, the nutritionist, and then of course the dairy management team themselves. And a lot of times you would never come together until you actually, you know, if you had a management team, you would come together once a month perhaps and talk about it, but by then you're not doing anything predictive. You're doing, looking in the past and then trying to react to it, which usually that whatever that was, is, was over that event. So I think I completely agree. I think it's exciting to be able to have something that can integrate it in real time and then provide data that is actually sort of already sorted through, so to speak, to give you something that is predictive and understanding to everybody, because otherwise you have so much data out there.
Ryan (00:29:28):
I mean, it's just like our smartphones, right? You have so many different apps and so much data that most of it goes unused because you just don't know how to aggregate it into one seamless event. So I think that we can get that figured out. And I think, Scott, your, your comment on the AI will help us to do that. But it will take a multi-professional approach in terms of having all the different stakeholders in the science on the farm working with the dairy to put it together. So you, you have all the pertinent points coming together.
Evine (00:30:05):
Well, actually it's got, it's got a nice bridge. To what I've learned in the past two weeks in New Zealand, Australia, we were learning there from farmers and veterinarians how they use sensor data in their seasonal gathering. Because over there they have to, they would like to have them in calf within six weeks. And there is a limit to the time when they will still try with a bull eventually after AIing. And interestingly, the vets play a major role in doing data analysis for these farms. It's a service they provide. So it's very interesting to learn from these veterinarians how they all work with a dashboard on farm, but some of them are very advanced in hiring data scientists in their clinic. They have to automate report tooling. They could include different sensors. It doesn't matter what sensor it was to vet could still provide support with that context. So we were amazed actually how far they already were with working with sensor data and pulling in so many more data points and still having a very structured way of supporting these farmers in a reporting tool or consultancy tool for, well, it's still, like you say, you're still looking in the past, but at least they can now have a very proactive strategy for next breeding season. So we might actually be able to learn from them with our seasonality to what we do here year round to make an intervention.
Stefan (00:31:32):
So when we think about sensor data, we usually, we use three different topics. So the one and probably the most simple is descriptive. So for example, okay, how many, how many percent of your cows were actually in heat within the first cycle after the world waiting period? And then it might be predictive. So a certain digital phenotype might be associated with the risk of disease. And, then I think the third one is prescriptive. So if we see pattern A, what should we do with these cows? And I think for reproductive management, I think we are going into all of these three different levels. And I think it's, I think now an exciting area that I think different groups in Cornell or in Florida, they developed, so some prescriptions, what you should do with cows, for example, that have been an asterisk within the volunteer rating pad, which is some indication that this cow might be reproductive wise problem. So, I think we are getting there, but you are absolutely right. I think it needs to be very simple for the farmer. So we have to find a solution for them based on the different data streams that we have on the farm.
Scott (00:32:57):
Yeah, absolutely. Stefan, while you have the floor Evine shared some of your research that you've done recently during her presentation. Would you mind kind of giving us an overview of what was the hypothesis or the thesis going into the research and what were some of your key findings?
Stefan (00:33:15):
Sure. So we took data from a large firm that I was working with. So before I joined the clinic of animal reproduction, I worked as a heardsman and a veterinarian. And I helped them develop some of the, yeah, protocols for reproduction, but also for transition cows. And they have been pretty successful with the NEDAP system. So they have a drug rate of around 31%. And, we were wondering if we can use the sensor data within the volunteer rating period. So within the first 60 days to identify cows that have a problem to get pregnant. And basically what we did is we differentiated cows into having no estrus event within the voluntary ready pit, one estrus event or two or more estrus event. And, then we looked at different reproductive outcomes. And actually it turned out that the cows that were on us with, they had worse fertility in terms of reduced pregnancy per AI, but also the reduced chance of getting pregnant within 250 days.
Stefan (00:34:26):
And then we took the data from the transition period and all the health records and we looked into risk factors. And I think that's no surprise. So most of the diseases happening around calving, they were associated with a greater risk for being estrus. And I think that's a nice way how you can use health data from the farm, but also sensor data to identify cows. That will be yeah, that will create some, some trouble in your repro program. And I think what I said before, I think now it's the time to think about a specific intervention for these cows. And we did a similar study on different farms here in Germany as well, using an another sensor system. And, actually, it was the same results. So I think using sensor data within the volunteer rating period can help us identify cows. That might be that might be in trouble. Ryan (00:35:37):
Is there any data being collected with the, you had mentioned stuff on the equipment where there's a producer putting it on the feeding equipment for pushing up. Is there anything with the harvesting equipment? because that was always a challenge that I that, you know, I always dealt with as a nutritionist is always seemed like the harvest crews did their own thing and didn't really integrate with the rest of the stakeholders on the farm. And that always created a challenge because then you would be stuck with, okay, we have a reproductive challenge, we have a nutrition challenge, and it always would go back to the feed, is there anything we can collect with this data to integrate it? That if we do end up with a challenge, we can then go back and say, okay, was there something that happened right during that harvest collection, maybe during the forage collection that was could give us an idea for obviously not being able to solve that problem because it's already an issue, but see if there's any type of trend data perhaps even with the, you know, a person who was, who was harvesting or maybe it was the time or environment or, or something different that we can start bringing in.
Stefan (00:36:54):
Yeah, so when I was working there as a hertzman, we had the opportunity to test a proof of concept NIR sensor in the mixing wagon, which is I think right now yeah, something pretty popular. And we used the sea looking and it was the first time that we compared, for example, NIR data in the TMR on specific feed stuffs. And we matched them with the data from the lab and it was for the dry, it was pretty good. And we see same technology here also for the corn chopper. So they implement these NIR sensors for example. And I think this is one example that where we can think of actually having data from the forage harvest going into feeding the cows. But, but other than that I'm not aware.
Stefan (00:37:54):
I think there are a couple interesting applications. For example, the University of Wisconsin, they developed an app for the kernel processing score. You can do it with your, with your phone. So I think there are several things out there that might be interesting. But, as a veterinarian usually I focus more on the cows, but it's definitely always a discussion that we have when we run into problems in transition cows. So usually go back to the basics including feeding management as I explained before. So, I think a time lapse camera for me is usually worth more than blood sample. And I think that's probably an area where we see most of the issues happening on the farm. It's not complicated stuff, usually it's the basic things that can go wrong.
Scott (00:38:53):
Evine. I'd like to circle back with you relative to using this technology for breeding animals. Do you use this to replace some of the synchronizing programs or to use it in as a supplement to that? Can you talk to that just a little bit?
Evine (00:39:11):
Yeah, so the examples where we see this is it can be a tool actually. I think even on that farm where Stefan has done his research, we start actually with visual observation. The combination, of course, it's the sensor technologies. So the sensor picks up cows in heat. They confirm them, breed them, but they put to a limit on it. So there is your voluntary waiting periods, you're not breeding them. You can use sensor data to already understand which cows are cycling. So when voluntary waiting period stops and you can breed them, you will breed them without hormones. But it can also be used as an intervention actually to understand these cows are not cycling correctly and Stefan knows more about what could be done, but also what is in that farm that day 80 put a cap by saying, okay, we do allow them for natural heat, but at some point we have to use those hormones for the final cows. So after day 80, if they haven't shown any natural heats, that's when we put them on the same program. So it depends on where you are and the cost of these programs. In countries where the cost is very high, then the centers are preferred. If the cost of these programs are very low and it fits in the management as it is right now, I think they'll perform very well too. So it can be supplementing each other sometimes can be replacing because of cost.
Stefan (00:40:37):
And I think the adoption rate of these a hundred percent time there protocols as you have is pretty low here in Europe. So we don't have a lot of farms that use, for example, a double lossing protocol for the first service. And that might be also one of the, one of the differences is I think usually when we present data from these protocols and, and we know they work pretty well, we always get skepticism and criticism from the veterinarian, but also from the farmer. And, and at the end also the consumers. So consumers are pretty critical here in Europe when it comes to systematic use of home protocols.
Scott (00:41:25):
Makes sense.
Evine (00:41:27):
An example where what we see sensors replacing is like the till choking, castile choking is also a way to understand cows which are in heat, however, it still doesn't tell you what's the optimum time for insemination. And the sensor does provide you the additional information when our heat started and then calculating what time is optimal for insemination. So that's where we do see sensors replacing the very labor intensive tilt shocking protocols.
Stefan (00:41:56):
But I think even for the herds that implement a double offspring protocol for first service, I think many of them also implement AI activity monitoring technology for subsequent services. Because the earliest time to diagnose an open cow is when she comes into heat 21 days after breeding. And there are a couple farms that use for example, the fertility protocol for the first service, but then they use a mix of activity monitoring system and a recent strategy,
Scott (00:42:35):
Kinda changing directions just a little bit, what we learned about the impact of cow parity on, on behavior have, have you guys looked at that and, and any key findings?
Stefan (00:42:49):
So, Evine, I think you did a study on the behavior around partition between different parities?
Evine (00:43:00):
Yeah, so for instance, what we know is especially first parity cow, primiparous cows, they have whole different eating times the miltiparous cows. So on conclusion from that study was should we enable these primiparous cows to have more feed allowance? Not that you will restrict feed another cow, but with hierarchy in the barn, thus a first cat heifer actually does get enough feed, that's what she needs to have. Or if there's too much hierarchy in this group. So there's one suggestion, if they require different feeding times in their behaviors could it be like a management strategy to house them differently? However, it still has to fit in your management, right? That's one thing.
Stefan(00:43:48):
But I think also for the, for the algorithm to detect cows that maybe are sick, I think what we try to look at is at least split the cows into first parody and second parody or greater because maybe also we can fine tune some of the algorithms in identifying sick cow or maybe also a cow in heat because as been mentioned, they have different behavior patterns.
Evine (00:44:20):
Yeah. And the other thing we see is the eating but you can see all their cows very steady in a group. They'll make sure they eat and when they eat, they'll eat good time. And you can see when we create a barcode, what is animal doing by the minute in a day. Could see younger cows having more restless behavior and older cows in way more solid behavior. Cause well, they know the hierarchy in the group, they know where the good feed is, they know where the good cubicles are. So again, there's something where these animals might be telling us more about than what we are actually using at the moment.
Scott (00:44:57):
And can we gain some direction relative to stocking density or available bunk space then using this technology?
Evine (00:45:06):
Yeah, potentially it's still hard to say just out of eating time, anything on, on stock density. Still we don't know the context of the sensor company that's sometimes the challenge there. We might see that in peak times 80% of the cows are eating, but maybe there's only for 80% of the cows feed bunk space. We see sometimes an another example in that webinar, a screenshot a for a robotic milker who has a robotic feed pusher too, at this farm, constantly 25% of cows are eating. But that's optimal management in a smaller farm, which is very easy to achieve. I'm a bit hesitant to say based on sensor data, you can say there was a question of webinar two, can you reduce your bunk space or adapt your stocking density? Still we don't know what's happening in that group. We don't know if it's a restless group or a very steady group. So again, we need more data points. And someone on farm who builds context with the sensor as a tool.
Ryan (00:46:12):
Evine, Is there any so to speaking of parodies, are there any breed difference data that you're collecting?
Evine (00:46:21):
No, not that specifically. No. We have background of the farms have a certain breed, but we haven't really specified in the different data we get back.
Scott (00:46:33):
So all of our discussions so far has been relative to the lactating cow. Is there a use for this technology with dry cows?
Evine (00:46:43):
I think especially there is, and I think it's underestimated what the value of sensor is for dry cows. And very commonly, unfortunately, we see that usually the housing of a dry cow group is just outside our data range. And at the moment, law firms, since that's okay, we rather see the cow year round. And again, Stefan knows more about transition disease, but I think how a dry cow behaves can already be predictor of how easier transition will go. So we really say there needs to be more information being generated about these dry cows and it can help her during transition, but sta might be able to comment even more on it.
Stefan (00:47:23):
Yeah, so I think when the farmers adopt the sensor system, I think most of the time what they are really curious about is that they are already some cows in the close up pen showing an health alarm. And I mean, to be honest, most of the time they wouldn't realize that these cows are actually sick without the sensors. So I think that's quite often that they run into these kind of cows that have some indication of sickness behavior actually before partition. So I think that's something that we've noticed. And then for these controlled energy diets, I think some of the farms, they use the group level feeding, but also rumination behavior also as an indicator for sorting. It's more anecdotal evidence, but nothing proof with scientific studies so far. But, actually I think that's quite interesting. So one of the farmers told me that when they have poor quality of chopped straw, they can see a different pattern in feeding and rumination behavior.
Scott (00:48:44):
You know, while we're on the subject of life stages, what about the calves? Are any opportunities to use these kind of behavioral sensors on calves?
Evine (00:48:55):
Yeah, so unfortunately the way these sensors are designed for the cows, they don't work in very young calves. And that's really because cows are very stable in their behavior. And you have a context here. You have a cabin dates, you have a lactation curve, which is predictable with calves. And I have a pretty long history already in calves. That would be my wish if it would be easier to measure than manage them with sensor data. But they develop so quickly and their behavior changes so quickly. So from a very as again, nature instinct makes it a calf tries to hide the first couple days, doesn't move as much, then it starts exploring context might be different. It might be an individual hatch, it might be a paired housing, it might be a group, it might be an outer feeder.
Evine (00:49:44):
So for an algorithm it's really hard to learn, especially when a calf changes every day and becomes expressing different behaviors every day. Like what's the standard? And however, I do really like what these current outer feeders already provide. I've seen the latest apps these outer feeders provide. It's, but again, it's a lot of information. So I think they're already good sensors, forecasts, especially on these outer feeders. What would be now ideal is at some point where we can connect the dots as well. Thus events in early calf food based on maybe outer feeder data or other records being entered in fire management software gonna be connected when you first having calfs behavior can be really complete completed cycle
Scott (00:50:31):
Very well. Kind of curious as we kind of get toward the end of this, if you look into your crystal ball, where's the technology gonna take us? Let's say in 2050? What does it look like? We've already talked a little bit about this and integrating different technologies, but so what's a life in a farmer a day in a farmer's life looked like? Does he just show up to his desk in the morning and the computer spits out, okay, go do this and that, that becomes a farmer's life? Or give me an idea of what you guys see the future looking like?
Evine (00:51:10):
I think we have to realize we're still working with biological creatures who are in an environment. They're always subject to an environment. So I don't think with AI would be always be able to predict our day will go or what a day would be like simply by the output in the morning. I usually like the quotes from especially actually Dutch farmers who moved to Canada, US, to skill and they moved from a small family farm to a big operation, now they're managing their staff and people and they commonly say, oh, I don't have to manage the cows. The cows do what they have to do. It's about managing the people. So I think at crystal ball, I think staff on farm, there will still be staff on farm, but they will do more specialized work instead having to do all the routines in a day. Yeah. So it might have even a bigger effect on the people on the farm than on the animal on the farm, but unless it's crystal ball right, we don't know how smart we get. Yeah,
Scott (00:52:10):
Yeah. Well I, Stefan, you have any thoughts on that? Yeah, go ahead.
Stefan (00:52:14):
Yeah, I think even if we see the farm's getting bigger and bigger, I think with the sense technology, but also with other traits like genomic data, we come back to an area I think where we precisely can manage cows on an individual base. I mean, Ryan, you mentioned the precision feeding. We see some parlors actually offering individual concentrate based on specific patterns. And I think we see this also for reproductive management that we might fine tune reproductive management for individual cows. SoI think in my crystal ball, I would think that we are probably better able to manage individual cows based on their needs whether it's feeding but also reproduction and maybe also the health monitoring side of it
Scott (00:53:17):
Very well. As we kind of move toward last call here, is there any big topics that we've yet to cover?
Ryan (00:53:24):
You know, Stefan, what you had just said. You know, it's interesting what was once old is new again. You know, back 1970s, 1980s, every parlor had a grain feeder. You know, it was a very non-specific, you know, each cow just get a dump of grain and more of it was just to get them to come into the parlor and things. But it seems like we're moving back now. We have the data, now we have the information, we're sort of, you know, we didn't necessarily have it wrong originally. We just didn't know what we were, we didn't know what we didn't know. And it's interesting as we move back towards that precision feeding that some of the technologies that we used to have are coming back, but this time actually with purpose and and understanding of actually what we're, what we're trying to accomplish. So it's, I find it fascinating when you look at the, look at the historical side of things that it's you know, looking in the past, we can learn about the future.
Evine (00:54:29):
Well actually, Ryan, to complete the observation. Yeah. We see big trend in the necessity of proper animal identification again. So at the beginning was the original question, can we identify an animal in a group with accurate identification? And it's very important actually becoming more important again and again, like you say, individual feeding identification, the in the parlor sword gates. because if you don't have proper identification, you're still not being able to manage the animal individually. So it's an observation we have too. We go back with the technology we've been proven for so many years and that's key at the moment.
Stefan (00:55:08):
Yeah, and I think I just mentioned the genomic site. So what we are really also paying attention to is the connection of digital phenotype. Like, for example, estrous expression and genomic data. So we just started a couple projects where, for example, we were able to show that cows that have a higher genomic merit for their pregnancy rate, they actually have a higher chance to resume estruus cyclicity. They show stronger estrous expression. And I think the sensor data might also help us to identify new traits for genomic selection, for example, which is I think very, very interesting.
Scott (00:55:51):
Hmm. Yeah, great conversation. And they have indeed already called last call then. And with that, I'm gonna ask you guys to give us one or two key takeaways that the audience ought to take with them today from the conversation. And Ryan is our special co-host today. We're gonna start with you.
Speaker 5 (00:56:08):
Tonight's last call question is brought to you by NiaShure Precision Release Niacin. Niacin is a proven vasodilator for heat stress reduction and a powerful anti lipolytic agent for lowering high blood niacin in transition cows protected with Balchem’s Proprietary Encapsulation Technology. You can be sure it is being delivered where and when your cows need it. Learn more at balchem.com/niashure.
Ryan (00:56:35):
Oh. I think the biggest takeaway I have is I think don't underestimate technology. I think you know, those of us in the, have worked in the dairy industry long enough know that we tend to shy away from data until the last possible moment. We look, you know, at the, I'll say average consumer iPhones, things like that. And we think, oh, there's no application for that in our business. And I think we're quickly seeing that we may not be adopting things fast enough. So, you know, everybody needs to keep their eyes and ears and more, more specifically mines open to ideas and, and things that can work.
Scott (00:57:23):
Yeah. Well said. Ryan. Stefan, what comments do you have for us?
Stephen (00:57:28):
Well, I think one of the major challenges right now from my perspective is integration of different data streams into the herd management software, and create an easy solution that's applicable to the farmer and give him some actionable insights. And I think from a reproductive perspective, I think it's really interesting to see how we can, yeah, individually manage cows based on their yeah, digital phenotype. And I think I would like to see that farms or veterinarians, yeah, actually fully or use the, the full potential of all the data. And not only that it's showing a cow and heat and, and new breeder. I think there's more into it and I, I think we are just scratching on the surface right now and, and starting to understand all the potential that's in there.
Scott (00:58:32):
Hmm. Yeah. Thank you for that. Evine, why don't you give us a couple closing thoughts as we close this out.
Evine (00:58:40):
Yeah. I guess say I'm very impressed with the research from Stefan's group and especially how he's expressing or showing how current data points like heat behavior in a voluntary waiting periods so much more value than we've been using it before. And I think it's key towards more strategic decision making with other experts on the farm. For instance we didn't really go into it in this, this pubcast, but where I can really use those data points and voluntary waiting period, for instance, also to choose, do I want to breeder to sex semen or to be semen, so many more strategies that can be applied, but already start looking at the data in the voluntary waiting period instead of waiting until you want to breed her. So we're very excited to learn what other publications you guys will come with.
Scott (00:59:31):
Yeah. Thank you. So thank you Evine, Stefan for your many contributions to the industry and the knowledge on this very interesting conversation. Ryan, thank you for joining me again and helping me out. And to our loyal listeners, thank you as always for coming along for this episode and sticking with us and look forward to more topics with you. We hope you learned something today. We hope you had some fun, and we hope to see you next time here at the Real Science Exchange, where it's always happy hour and you're always among friends.
Speaker 5 (00:59:59):
We'd love to hear your comments or ideas for topics and guests. So please reach out via email to anh.marketing@balchem.com with any suggestions and we'll work hard to add them to the schedule. Don't forget to leave a five star rating on your way out. You can request your Real Science Exchange t-shirt and just a few easy steps, just like or subscribe to the Real Science Exchange. And send us a screenshot along with your address and t-shirt size to anh.marketing@balchem.com. Balchem’s Real Science Lecture Series of Webinars continues with ruminant focused topics on the first Tuesday of every month. Monogastric focus topics on the second Tuesday of each month, and quarterly topics for the companion animal segment. Visit balchem.com/realscience to see the latest schedule and to register for upcoming webinars.