Estimating Kelp Farm Yields: Strategies for Accurate Forecasting | Farmer Forum

Welcome and Introduction [00:00:00]

Hey everyone. Good morning, good afternoon, good evening. Welcome to Farmer Forum. Thanks for joining us. Some familiar names in the lineup — love to see that — and some new ones too.

We’re going to go ahead and jump right in because we’ve got another packed hour, and I want to make sure we have lots of time for our speakers to share their knowledge. Happy New Year everyone. Happy January.

If this is the first time you’re tuning in for Farmer Forum, my name is Lindsay Olson. I’m the Director of Training and Support at GreenWave. These conversations are really just an opportunity for us to dig into more technical topics that seaweed farmers around the country and the world may be grappling with as they go through the different stages of their season.

I am particularly excited to talk about this topic today — estimating yields — because it is something that I feel like a lot of folks are struggling with. As the industry takes its next step and we start to see the influx of new buyers and new types of conversations happening between farmers and buyers, there’s more and more of a need to be able to predict how much seaweed is going to be grown on your farm.

Before we get into the details, we’d love to hear where you all are coming from. Please throw in the chat your name, your current role or interest in seaweed, and where you are located. It’s always really fun to read through those introductions.

Session Agenda [00:02:19]

Here’s what we’re going to cover today. I’ll start with an overview of the challenge — why we’re having this conversation and what is pertinent about this topic. Then I’ll share a couple graphs from what GreenWave has seen over the last few years of collecting Kelp Climate Fund (KCF) data. As many of you might know, we run a program called the Kelp Climate Fund where we provide a climate subsidy to folks who are actively growing kelp. In exchange, we ask that they provide monthly biomass samples from their farms during the time the kelp is in the water. Based on that data, we’ve been able to see trends across the industry. We have about 50 participating farms in the U.S. and Canada.

My colleague Dave and I went on a bit of a learning journey on this topic and talked to a lot of farmers — both from traditional terrestrial agriculture and from the seaweed world — and we’ll share some of the takeaways from those conversations.

In the second half of the hour, we’ll hear from our two great speakers: Adelia Myrick from Alaska Ocean Farms in Kodiak, and Diogo Raposo from Seaweed Solutions in Norway.

The Yield Estimation Challenge [00:03:37]

I think this is probably not news to anyone who signed up for a call on estimating seaweed farm yields — you’re probably in the weeds enough for this to be a common understanding. But just to put us all on the same page: the challenge we’re seeing is that there is just such wide variability in pound-per-foot yields — or biomass-per-area yields — within farms, between different lines. You can see really different yields line to line even within the same farm system.

We’re also seeing variability between farms — even farms that are close to one another — and big differences in yields year to year. All of this variability leads to inaccurate forecasts. It makes it hard as a farmer to accurately predict how much kelp you’re going to grow, which can then create inefficiencies in downstream processing. If you think you’re going to grow 10,000 pounds of kelp but you only end up producing 5,000, and you’ve told a buyer you’re going to deliver 10,000 pounds, that buyer counts on that and sets up all of their contingencies that trickle down through the rest of the value chain.

The other part of the challenge is simply that this is a new industry. Compared to terrestrial agriculture that has hundreds of years of data to benchmark against, there really isn’t a standard set of numbers to track your crop’s production against — no average yield, no clear picture of what “good” looks like. We just don’t have that data yet. And because it’s a new industry, farm systems are constantly changing. People are trying out new things every year, innovating and improving, and those added variables make it hard to compare year to year.

KCF Data: Illustrating Variability [00:06:24]

Here are a couple graphics coming from our Kelp Climate Fund data. The first shows variability farm to farm. On the left, these are two different farm sites GreenWave has operated for several years, showing data from two years. These sites are about 30 miles apart in Connecticut and received the exact same batch of seed — yet had almost a 100% difference in yields over two years. The point is that there is huge variability farm site to farm site. Seaweed is going to grow differently in every place it’s growing.

On the right, that beautiful squiggle shows different growth curves of kelp from farms in the Gulf of Maine. Each color represents the path of growth of kelp, tracking the date on the bottom and the pounds-per-foot yield on the y-axis. Even within the same region and the same body of water, you see huge variability farm to farm, even as you see a general shared trend of growth tapering off.

Variability also happens year to year. Comparing dash lines to solid lines for both the red and blue, that’s showing the 2022–23 season compared to the following year — again, almost a doubling of yields between two different years. We saw this across the Gulf of Maine and Southern New England, with roughly a 100% jump in yields on the east coast that season.

Overestimation Bias in KCF Data [00:08:52]

We’ve also noticed in our KCF data that there is a tendency to overestimate yields when you’re relying exclusively on sample data. There’s a real tendency when you go out to your farm to take a sample — a one-foot or one-meter section of line — to pick the section that represents good growth, because that’s where kelp is visibly growing. But when we only sample the best sections and don’t take into account all the lesser-growing sections, it can lead to overestimation.

Across the board, we have found that the estimates the Kelp Climate Fund data is producing are 40% higher than what people actually produce. Our goal in having this conversation is to figure out strategies for how we lower that range and get those estimates more accurate over time.

Lessons from Farmers Across Industries [00:10:06]

We went out and talked to an apple farmer, a crop insurance agent, vertically integrated seaweed companies, and individual seaweed farmers. One big takeaway: estimating yields is hard for all farmers, on land and sea. For me at least, that was comforting — we’re not alone in this struggle. Everyone is dealing with this, and there are things we can do to get better at it. You should feel affirmed that it’s hard to do.

We also learned that an estimate within 15 to 20% of your final production is considered pretty good. If the number you came up with is within that range of what you actually produce, you’re doing well. That said, 15 to 20% is quite a big buffer — if you’re 20% short on an order, that’s significant.

Across agriculture, farmers aim to overproduce to ensure they can fill the orders they have. Knowing that yield estimation is hard, and knowing you’ll need some sort of buffer, you need to build that buffer into your production at the beginning of the season — kind of make that a behind-the-scenes calculation. One apple farmer we talked to sometimes overproduces by 100% — he grows twice as many apples as the order he needs to fill. That provides security in satisfying his target markets, and the leftover apples he’s able to offload in lower-quality markets.

We also learned that historical data from each farm site is key. It doesn’t work to copy and paste data from somebody else’s farm. You have to know the historical performance of your own site — that’s going to give you the most informative data for making these estimates. Terrestrial agriculture relies on hundreds of years of data to benchmark crop production. We don’t have that luxury, but those benchmarks were created by collecting volume per growing area — the same metrics we’ve been encouraging KCF farmers to collect: pounds per foot, kilograms per meter, how much is growing on a given section of line.

Yields can vary within one farm site as well — which Adelia is going to talk about today. One sample from your farm does not represent the entire farm. How can you get more granular in your assessment to really represent how your farm is doing overall?

We also learned that yield estimates are often adjusted — and should be adjusted — throughout the season at key points as you get more observational data. In the apple world, 30 days after the blossoms bloom is the most critical point for assessing crop health and making a much more informed prediction. For seaweed, think about those junctures in the spring when you start to see early growth on your lines — when you’re able to make more informed observations about how the crop is developing. If there’s a storm event or some major disruption, that obviously updates your yield estimate for the year.

Communication among regional farmers improves estimates, too. Farmers who talk to each other tend to have more accurate estimates than those who calculate in isolation.

Key Factors Affecting Yields [00:14:42]

Several factors kept coming up across all the farmers we spoke with:

Seed quality matters enormously. Images of seed spools outplanted in fall 2023 across farms across the U.S. show huge variability in seed quality, and that variability in seed directly impacts variability in crop yields.

Outplanting timing and season length also really matter. How long your crop is in the water determines how much it can grow. If you outplant later and harvest at the same time, you’ve given yourself a shorter growing window. Over the past two years, we have seen that an earlier outplant has produced higher yields overall.

Environmental conditions are the factor that changes every year. Even with perfectly consistent seed quality and planting schedule, yields can vary because we live in the natural world. Water temperature, storms, salinity, runoff events, changes in nutrient availability — those environmental conditions are what drive big shifts in yield estimates year to year.

Historical data from your specific farm is a key anchor for your estimate. Your yield forecast should reference what has actually grown on your farm before.

And then there is an X factor — farmer intuition. Despite all the numbers you’re reading and the benchmarking you’re doing, your gut feeling also matters. Your intuition about how the crop looks and how it seems to compare to other years is important and worth factoring in.

Strategies for Improving Yield Estimates [00:17:21]

So what can seaweed farmers do to improve yield estimates on kelp farms? A few key strategies:

In the meantime, while we’re all working to improve, we need to be thinking about overproducing on farms — to the extent you’re confident in your numbers. Make the buffer calculation a behind-the-scenes endeavor. Your buyer doesn’t necessarily need to know you’re making these calculations. This is for you as a farmer to work out, and by building in that minimum 15 to 20% buffer, you’re shielding yourself from having to have a hard conversation with your buyer while you fine tune your estimates.

Collect data and compare it to previous years. There is no replacement for data coming off your own farm. The more you collect throughout the season, keep it organized, and reference it, the better.

Talk to other farmers. Don’t be shy about asking how other people’s growth is doing — it helps inform everyone. The more we can share about what people are seeing in your region, across regions, the better.

Make yield estimates a dynamic calculation. The most accurate estimate is one that is constantly being updated based on what you’re seeing. Don’t be afraid to revise your estimate as the season goes on and new information comes in.

Adelia Myrick: Farm-Level Sampling in Alaska [00:20:28]

Alaska Ocean Farms, Kodiak, AK — Fifth year of kelp farming

Thanks, Lindsay. I would concur with the description of me as a student. My husband and I are in our fifth year of kelp farming now. Last year we were really working hard to figure out our estimates as the season went on.

The first thing that was really helpful was just — we’re always told don’t cherry-pick the good lines, and we try not to. But I think we have to try even harder. There’s some sort of bias that makes it so easy to pick the lines that look better. I think we may need to flip it so we’re cherry-picking the bad lines, in order to get a more accurate estimate. Historically everything has been high — we’ve been off on the high end — so that was a helpful shift.

Another thing I asked Dave about was that when we use the GreenWave app, we estimate the total biomass of the farm. But what we’re selling to our buyers doesn’t include the holdfasts. We had to make that calculation for how much weight is in the holdfast — Dave said it’s typically between 5 to 10%. Taking that off helped when we were talking to our buyers.

The more frequently you can take samples, the better. Dave was helping me — if I could take samples almost every week, he would help me estimate the growth rate so we could project out from April: what is the growth rate going to be by harvest time in May?

Our farm had sugar kelp outplanted at roughly the same time, but there were just different lines performing differently. We had to spend a lot of time trying to estimate what percentage of lines had what kind of growth on them. The pictures I’m showing were all taken on the same day — same date of outplanting, just very different growth.

One of the most important mindset shifts was in the second photo — you can see my little measuring tape is actually in an area where there isn’t any sugar kelp growing. Actually putting a zero in there helped get the estimates more accurate.

We were continuously, as often as I could, doing measurements — doing more than required. Looking at the growth rate, Dave was able to help me compare it with previous years. We could see that kelp grows at a certain rate in a given timeframe and project out from there.

In the future, it would be interesting to track sunlight and salinity records and see how those impact a year’s growth. But the main takeaways: estimate low, try to get lines that don’t look that great, and do it as often as you can.

Diogo Raposo: Monitoring and Forecasting at Seaweed Solutions [00:25:17]

Seaweed Solutions, Trondheim, Norway — 11 years of seaweed farming

Thank you for having us. I think it’s really valuable to share experiences among farmers — that’s the only way we can actually make this a successful industry.

I’m the Cultivation Coordinator at Seaweed Solutions. I’ve been a seaweed farmer for 11 years, spending some time in Denmark before moving to Norway. I’m based in Trondheim, and our farms are about two and a half hours away. We have two commercial farms located on the islands of Frøya/Ørland — an area that is reportedly the most dense salmon farming area in the world.

Our two farms are each about 20 hectares of effective farming area, divided into 16 squares. The grow lines are 110 meters long, staying between one and three meters depth, with buoyancy points to manage the natural belly that forms over such long lines. We use these farms for R&D work on farming methods, technology, and site selection.

Environmental and Seasonal Monitoring

We do monthly water sampling at cultivation depth, collecting samples for chlorophyll and nutrient analysis in-house. We also have sensors out at the farms collecting temperature, light (above and below water), salinity, and pH. These factors can influence yield, and we’re still building up the data — but for example, if we have a very cloudy year, that may impact yield at the end.

Seasonal Biomass Monitoring Schedule

We do deployments mainly in September and October. We start biomass sampling in March, because that’s when it becomes most representative. In March, we do a quality check looking at length, width, and density of individuals. We divide the month in two — March 1 (first to 15th) and March 2 (16th to end of month).

In early April, we can start adjusting the harvest forecast. We go to the lines and collect the first weight, measuring length and width. We then look at density again in early May. We continue this through the end of June, because that’s when harvest is most relevant — the potential harvest window runs from end of April to end of June. After June, the bryozoan biomass becomes too significant for most markets.

Sampling Protocol Details

We collect three to five replicas per batch — where “batch” can mean different deployment days, different weeks, or different R&D tests. We try to randomly select samples, especially based on depth, since the belly in the lines can create differences. We work along the 100-meter lines and always pick from different locations — from ones far down, from ones further up — to make the selection as random as possible.

For weight, we collect one meter of line and measure the biomass within that meter, including the whole individual — holdfast, stipe, and blades. Very importantly, the sample needs to drip excess surface water. It goes into a basket with holes — like a regular fruit basket — and we let it drip before weighing. This prevents overestimation from surface water.

For length and width, we measure 5 to 10 individuals, three to five replicates. For density, we count individuals within 20 centimeters, three to five replicates, and categorize those individuals into length categories. This categorization matters for future seasons — we want the individuals to be as large as possible. Very tiny individuals that didn’t manage to overcome shading aren’t contributing much, which likely means the seeding density is too high.

Harvest Data and Yield Tracking

The harvest data is the control check on all of this. We harvest the full biomass, divide by the meters, and get a real-world figure to compare against our estimates.

We divide harvest data into three points: harvest on board (biomass that comes onto the boat and is released from the rope, collected in 1,000-liter fish containers), delivery to processing facilities (where boxes are weighed on arrival), and packed product weight after processing. There is loss at each stage — from holdfast and stipe removal, from quality issues, and from the processing method itself. For example, blanching seaweed results in about a 20% weight loss.

We normally consider between 20 and 25% of the total biomass weight coming from holdfast and stipe. If customers don’t want that material, we need to consider it as a loss. But we keep tracking it because hopefully one day there will be a market use for it.

Yield Forecasting Approach

At the beginning of the season, we set a minimum and a maximum yield range based on species, farm site, historical data, deployment and harvest timing, and the processing methods the customers require. If customers mostly want blanched product, we need to produce significantly more than what they’re ordering because of loss along the way.

Season after season, we try to apply R&D knowledge — particularly around depth management, buoyancy, and seeding density — to improve our yield estimates. Over three seasons, we’ve seen a negative correlation between seeding density and yield: reducing density can result in longer biomass and more weight per meter. We’re still figuring out the optimal density.

We still get a range in our estimates, but the framework helps us get closer year over year.

Q&A: Proxies, Benchmarks, and Harvest Efficiency [00:42:22]

Q: Can you use a proxy for yield estimates — like total length or stipe thickness — similar to fish length-to-biomass correlations?

(Diogo): Length can be a proxy, and in the future it may be the area — length plus width — because when we start talking about selective breeding, we may also want wider plants. A combination of length and width could actually make the sampling process much quicker. If we only need to measure five individuals and get a good proxy, that would be really valuable.

(Adelia): We’re not there yet.

Q: Has anyone done work comparing initial seeding density during inoculation to ultimate biomass yield?

(Diogo): Yes, that’s essentially the R&D test I described — the density I mentioned is the seeding density at the time of seeding the twine. We’ve been testing the effects of density on length and yield at harvest, and we’ve seen that negative correlation between density and yield across three seasons.

Q: Do you measure chlorophyll as a way to determine light transmission or nutrient availability?

(Diogo): We measure chlorophyll to understand how nutrients are cycling — it’s more useful to look at chlorophyll and nutrients together, because phytoplankton will take available nutrients first. We also use it to track the settlement of bryozoans. In some tests, we’ve figured out that if we keep taking samples, we’re able to detect bryozoan larvae at some point, giving us about two weeks to harvest before they become significantly present in the blades. However, that detection point tends to fall roughly in the same window regardless, so it doesn’t dramatically change our operations — it’s mostly useful for tracking the nutrient curve at the farms.

An interesting note: when we get a storm around harvest time, chlorophyll goes down, but the nutrients that seaweed can access become more available. We actually see a positive of having a storm — the seaweed grows a bit better because it doesn’t have the phytoplankton competition it would have without the storm.

Q: How do you make harvest more efficient?

(Diogo): I believe in continuity of processes. Ideally you have a line that goes continuously from the start to the end of the day — you keep winching the whole day. That can be achieved with many lines coming together on board. The winching needs to keep happening all day, and all other operations need to work around that. The challenge is getting the right equipment to achieve that in practice.

(Adelia): We’re still hand-cutting, and that is definitely a bottleneck. Moving away from hand-cutting would be a great leap in efficiency.

(Diogo): There’s a farmer in Norway who is successfully harvesting efficiently, and I think a major reason is that he doesn’t use seedstring twine — he uses a direct seeding method, which removes the risk of twine ending up in the product. When you’re using blades to cut and you have twine, there’s a high risk that the rope can get wobbly during the season and cut the seedstring, which ends up in the product. Customers don’t want product with rope — so that’s something we all need to work out together. We’ve also tried water cutting; it brings different challenges, and it needs more development.

Q: Are average yield benchmarks like five pounds per foot for a six-month season measured with or without stipe and holdfast?

(Lindsay): It probably depends. Adelia, you made a comment about needing to fine tune your estimate based on whether or not you include that portion of the holdfast.

(Adelia): I think it’s going to be different for every farm, region, and species. Working with the My Kelp app from GreenWave, which estimates total biomass, my understanding is that when people talk about five pounds per foot for sugar kelp, they are talking about total biomass. So you do want to take off the holdfast percentage when estimating what you’re actually going to be selling.

(Lindsay): This gets to the important point Diogo touched on — having multiple different weights. You have the full biomass weight, the maximum kg per meter growing on the line, and then you have the portion that is actually sellable. And there are different points in the process where you are weighing each of those metrics.

(Diogo): The difference between harvest on board and delivery to processing facilities is the holdfast and stipe. We try to sample some days within the harvest period to get an idea of that percentage. It has been consistently between 20 to 25% for us across past seasons.

Q: What are your average kilo-per-meter (or pound-per-foot) results for each kelp type, and what do you think is possible?

(Diogo): For Alaria esculenta at prime deployment and harvest timing, we can talk about 5 kg per meter. For Saccharina latissima, a bit higher — around 7 kg per meter. Those numbers include holdfast and stipe.

(Adelia): Looking at last year, our sugar kelp averaged 3.4 pounds per foot — that was a hard year, so I think it could be higher. And our Alaria, which was our first time growing that species, was 3.3 pounds per foot.

(Lindsay): Looking back at the KCF data, the average yield per region this past year around the Gulf of Maine was about five pounds per foot, and similarly for Southern New England.

Closing and Next Session [00:57:11]

Thank you so much to Adelia and Diogo for presenting today and for your contributions. It is always great to hear directly from farmers.

We will be back in February on the 13th for a conversation on harvest planning — getting you in the mindset of thinking about the things to come. Please join us back in February for our next Farmer Forum. Best of luck everyone, and have a great rest of your week.