From cows to canvas panels: Measuring emissions in naturally ventilated livestock barns

This lightning talk explains how emissions can be measured in naturally ventilated livestock barns housing cattle and pigs using temporary sensor campaigns and Grafana. The sensors run for 2–3 weeks per campaign, several times per year, so the focus is on collecting consistent data and comparing results across campaigns — not on 24/7 live monitoring.

Sunil Gopalakrishna, a Researcher at the Institute of Agricultural Technology, walks through the full pipeline: sensor nodes, data ingestion, a time series database, and Grafana dashboards. He shares what must be consistent for results to be trustworthy: a shared time reference, a stable naming scheme for sensors and measurements, and clear metadata (campaign, site, animal type, sensor location, and calibration state). Simple quality flags highlight missing periods and obvious spikes so plots are interpreted correctly.

Finally, Sunil shares the campaign dashboards: annotated time series for key events (cleaning, feeding, ventilation changes) and a canvas panel with a barn layout that displays sensor dots on a floorplan and marks when thresholds are crossed. Basic integrity checks (data completeness, “last seen” during active windows, and ingestion errors) help identify data capture problems quickly.

Sunil Gopalakrishna (00:00):

Hi, good evening. My name is Sunil Gopalakrishna. I work as a researcher at Thünen Institute of Agriculture Technology in Braunschweig, Germany. I'm gonna talk about how I'm using Grafana for this particular use case where we have sensors built into barns and hence the title, From Cows to Canvas Panels. We have traditionally closed barns. So barns, housing animals, could be cows, could be pigs, could be hen. And we have changed this build for the last decade or so. We have now a partially ventilated barn. That means we have bay areas where you can see pigs, they come out of these barns, get some fresh air, do their business, get some sun, go back in. Now it becomes very difficult to measure emissions when you have such barns which are, or which have these bay areas. What happens here is you try to measure emissions in this region and due to winds, due to different scenarios of animals being there and not being there, it becomes pretty difficult.

(01:13):

So I work with, in a project that we are trying to measure such barns or emissions from such barns and we developed a Sensor Array Measuring Ball. It consists of four sensors, low cost sensors, measure ammonia, carbon dioxide, particulate matter, and partial, sorry, differential pressure. And when we're doing this, we get this data from these sensors and we wire each of these sensor balls with each other and send them to telecommunication towers or LTE.

(01:53):

And if you look, we try to place them at the interface of these barns and they're all wired to each other. And it's also a possibility that we can scale these balls up. So let's say we can do 50 balls and try and measure the emissions. And once we have these sensor values, we try to send them using wifi. We have tried it, but it's not easily possible because we are in some locations where there's basically no connection. So we use LTE and some of this data looks like this. And this data in turn, as I would say, a hundred kilobyte file because of this packages, we have time-indexed data. So basically, from the left to the right, you see time and particulate matter, carbon dioxide, ammonia values. And this is important for us because we are trying to see how much of ammonia emissions, how much of carbon dioxide is emitted from these such barns, right?

(02:58):

And once we have this data, it could gather up and basically create gigabytes of data, which doesn't happen at this point in time because we are doing campaign measurements, we have measurements happening between seasons. So I decided to go ahead and use an open source solution that was InfluxDB. And in this scenario, you could see I have a few buckets, and from these buckets I could get the data from each of these spheres and also plot them against time. So this is what it looks like. But there were some engineers between us. There were some technicians who realized this doesn't help them so much. So then I thought about it and I decided I would go ahead and use Grafana. And I created a dashboard. And you can see, from top to bottom, I specified what kind of barn it was and how many spheres are we having on these barns and the concentration with respect to time.

(03:55):

And I used the Canvas panel to actually try and create the layout of the barn and the spheres in the barn. And if you look on the top, you can see from each of these spheres we have data in terms of the emissions. And also, I can check and relate the humidity or other factors from the sphere. And I also have other inputs like this video or two videos from different barns from our partners. So they have been installed. And if you look carefully, I don't think it's installed here, our spheres are also being installed across these barns. And that also helps us to correlate the spikes in these emissions.

(04:42):

And we also have annotations on each of these emission values so that we could actually say what period of time we measured, basically talking about the growth period or the fattening period, or if there was an anomaly because we have two or three pigs added or cows added to this barn. And I also created something like the data completeness here. It makes sense if it's live because then you can actually see in an hour how much data was received. And if it goes down, then we already know that the sensors or the server itself has an issue.

(05:17):

So what I would like to conclude here is that it is possible to use an open source tool like Grafana for this such a use case. And you can basically see what's happening here. We pull the data from the spheres, we calculate the velocity separately using the partial differences, and then send it to the Influx. And basically, then you have a dashboard. I use the Canvas panel. I'm a bit, I would say I would like few more additional features in the Canvas panel. I had a great talk with a few of the experts here. And also, I'm very thankful to the community here because I was able to go on the forums and ask questions. I thank you all. Finally, I would like to thank the partners and the team members in my institute. And with that, thank you so much.