Heifer Grazing GHG Analysis: Results + Questions for the RuFaS Team

[mikehsu90]

Hi @JoeWaddell and @KFosterReed,
Carbon Yield has been working on a project to model GHG (greenhouse gas) intensity differences between confined and pasture-grazed replacement heifers on a Wisconsin freestall dairy using RuFaS, and we wanted to share what we've found and get your input on a few structural questions before we finalize our documentation.

Our goal is to help document what's possible with the existing RuFaS infrastructure — including clear caveats — so that co-ops and practitioners can use it responsibly.

What We Did

Baseline Scenario

• Heifer Housing: Confined, freestall
• Heifer Diet: 7-ingredient purchase TMR
• Manure (Growing Pen): AlleyScraper + 120-day slurry

Heifer Pasture Scenario:

• Heifer Housing: Pasture
• Heifer Diet: 97.3% cool-season grass (ID 92) + 2.7% mineral (ID 301)
• Manure (Growing Pen): ManualScraper + 1-day storage, crust cover

Adult cows are confined in both scenarios.
Repo: https://github.com/carbon-yield/rufas-grazing-heifers
Branch: heifer-pasture-analysis

Results

Baseline Scenario: 1.504 kg CO₂e/kg FPCM
Heifer Pasture Senario: 1.470 kg CO₂e/kg FPCM
% Change: -2.2%

The reduction is driven primarily by manure management (avoiding 120-day anaerobic slurry storage), not feed change. Enteric methane increases slightly (+0.9%) due to higher fiber in the pasture diet — which appears physically correct.

Questions

Q1: Steady-state / heifer replacement rate

Jim Eckberg flagged this in a conversation with us, and we wanted to raise it directly with you. RuFaS initializes by simulating every animal to reach a steady-state herd. Our concern is that when we change the heifer enterprise (housing + diet), RuFaS may be treating the number of growing heifers as effectively fixed rather than allowing it to respond to changes in replacement rate, longevity, or first-lactation productivity.

• Is the count of growing heifers in the simulation actually recalculated when heifer inputs change, or is it held static by the input configuration?
• If it's static, could changing heifer management practices (which may improve longevity and reduce culling) meaningfully distort the comparison?
• Is there a recommended way to model this kind of enterprise-level change in RuFaS?

Q2: Manure methane from pasture direct-deposit

In our heifer pastuer scenario, we model heifer manure as 1-day storage with crust cover + ManualScraper (our best proxy for pasture direct-deposit). We see a large drop in manure CH₄, which is the expected physical result from eliminating anaerobic slurry storage.

• Is ManualScraper + 1-day storage the correct RuFaS approach for representing pasture direct-deposit, or is there a more accurate configuration?
• How does RuFaS calculate the volatile solids available for enteric vs. manure CH₄ when animals are on pasture? We want to make sure we're not double-counting or misattributing emissions.

Q3: Grazing module / Hector Menendez's prior work

Jim mentioned that Hector Menendez at SDSU worked on a grazing module attempt ("Grass is Greener on the Other Side" AMP grant) and that you may be aware of this work, Kristin.

• Is any of that work incorporated into the current RuFaS codebase, or is it documented somewhere we could reference?
• Is there a plan or roadmap for a formal grazing module? If so, we'd love to understand the timeline and whether our scenario work could inform that effort.
• If we were to draft a proposal for a grazing module contribution, what would be the right way to approach that conversation?

Q4: Pasture diet representation

We used cool-season grass (feed ID 92) at 97.3% + mineral supplement (ID 301) at 2.7% as the heifer pasture ration. This is our best approximation for a Wisconsin grazed grass diet.

• Does this feel like a reasonable representation, or would you suggest a different feed ID or composition for a Midwest cool-season grazing system?
• Are there any known limitations in how RuFaS handles 100%-forage rations for growing heifers specifically (e.g., energy balance, growth rate modeling)?

Q5: Soil carbon

RuFaS does not currently model soil organic carbon (SOC) sequestration under permanent pasture. This is the most significant gap in our analysis — managed grassland can sequester 0.2–1.0 tonnes CO₂/ha/year, which would likely further favor heifer grazing.

• Is there any SOC modeling planned for RuFaS, or is this considered out of scope?

Thanks very much for your time. This is genuinely exciting work and we're glad to be building on RuFaS's foundation. Happy to jump on a call if any of this is easier to discuss verbally.

Mike Hsu
Carbon Yield
hsu@carbon-yield.com

[KFosterReed]

Hi @mikehsu90! Thank you for the excellent questions! See some responses below that have input from several of the RuFaS maintainers. I will also send you an email to follow up more about the grazing/pasture stuff :)

Q1: Steady-state / heifer replacement rate

• The input value for heifer numbers is only an initial value for day 1 of the simulation; from there the numbers change daily based on calves born, calves kept (function of stillbirths, male calf rate, and keep female calf rate), repro failures, and duration in the system (age at first breeding and repro performance, resulting in age at first calving).
• RuFaS doesn’t currently model heifer mortality or culling explicitly aside from repro failures, so that comparison between groups would be lost. Any differences in growth or health would need to be captured in the repro and growth parameters to show up as a difference between groups.
• We are working now to update how RuFaS handles this type of herd dynamics at baseline and to understand what type of scenario comparisons would be useful to facilitate. We would love to chat to understand more about what you have in mind!
• In your workflow, are you allowing for a “burn-in” period so the herd can reach a steady state? We also recommend running at least 4 iterations for each scenario and taking the average. For smaller herds especially, this is important for comparisons.

Q2: Manure methane from pasture direct-deposit

• Manual scraping with a 1-day slurry storage was the best RuFaS work-around we had for a daily spread management system in our previous versions. In the recently released v1.0 there is a daily spread option where you can add directly to the pens that are spreading manure.

Q3: Grazing module / Hector Menendez's prior work

• Sadly, there has been no implementation of the grazing module yet. We did have an initiative and a few students who worked on it over the years but nothing ever made it out of the model concept phase.

• Representation of grazing is something we’d love to have in the RuFaS model but we don’t have a timeline for it yet as it is not in the near term (~6-12 month) development objectives in the RuFaS maintainer group.

• There are several ways you could approach developing a grazing module and have it pushed to the RuFaS repo. I’d love to learn more about what you are thinking so this would be a great item for discussion with the RuFaS leadership team (The Program Management and Leadership committee, aka the PML).

Q4: Pasture diet representation

• I think the cool-season grass the right choice although you might consider including some legumes as well since pastures will often have some legumes seeded with the grass so I think anything from a 50:50 split of the pasture grass and the pasture legume (feed ID 105) to a 80:20 split would be reasonable. Here is a nice article from the University of Minnesota about mixed sward grazing (https://blog-crop-news.extension.umn.edu/2022/03/benefits-of-legume-grass-mixtures.html).

• At this point, there is no feedback from the diet to the growth rate of animals as we haven’t developed a way to reduce the requirements (i.e. growth rate) in a similar way that we reduce milk production. See Heifer ration formulation 'reduction' method #577 to track progress on how this may be implemented in the future. In the meantime, if a ration formulation fails for a pen that isn’t comprised of lactating cows, we report the failure to formulate, and then one of two things happens.

• If there is a previous ration available for a pen, we continue to use that ration, assuming it should be sufficient, if not perfect.

• If this was the first formulation period, the simulation will stop. This is simply because we don’t have an example of a successfully formulated ration for the given herd/simulation.

Q5: Soil carbon

• Currently the RuFaS Soil and Crop Module has an adaptation of Daycent’s soil carbon model built in to the soil nutrient cycling methods. We are working towards a more robust evaluation of this part of the model but it is still in development. There are many ways I think of that we could extend the current model to better represent a field in pasture (e.g. a grazing like tillage/soil disturbance model, addition of manure carbon, improved grass senescence and response to grazing) but a step closer would be to simulate a cool season grass field with daily or frequent manure application. The Soil and Crop Module Scientific Documentation has more information about how soil carbon is represented in RuFaS. There is definite room for improvement but it is a start! We’d love to hear more about how this does or does not meet the needs for your use case.

[mikehsu90]

Hi @KFosterReed , thank you and the team for such a thorough response. This is exactly the context we needed. Know we are chatting next month, but I had a few updates now that I have had the chance to look into this more deeply. Replying by question:

Q1: Steady-state / heifer replacement rate

Helpful clarification on herd dynamics - this addresses a gap in our understanding of how to interpret cross-scenario comparisons. We are running 4 seeds per scenario (random seeds 41–44) and taking the average, consistent with your recommendation. We have not implemented an explicit burn-in period beyond what the simulation initializes from - would welcome any guidance on what a typical burn-in looks like in practice (number of days, how to identify when the herd has stabilized).

Jim Eckberg was the one that had flagged the herd dynamics to us - I believe he's been in conversations with you and your team about this specifically.

Q2: Manure methane from pasture direct-deposit

We've already updated to DailySpread in our current scenario (v2, April 2026). The change resolved an artifact we encountered with the 1-day slurry workaround: manure N2O was spiking by ~3,300% relative to baseline, which turned out to be triggered by routing manure through slurry storage even for 1 day. Under DailySpread, manure N2O is essentially unchanged from the confined baseline (9.5 kg/yr vs. 9.5 kg/yr), and manure CH4 is reduced by 13.7% - which is the physically expected direction. Results are much cleaner. Thank you for confirming this is the correct approach.

Q3: Grazing module

Understood that this is not in the near-term roadmap and that prior efforts didn't reach implementation. We appreciate the transparency.

We are interested in discussing a potential contribution path through the PML. Our project (Scaling Dairy Heifer Grazing) has a specific use case: quantifying heifer grazing as a GHG-reduction practice for dairy carbon programs and co-op Scope 3 accounting. A grazing module that captures direct deposition, pasture nutrient cycling, and seasonal forage dynamics would directly serve that use case - and we imagine it would benefit the broader RuFaS user community as well.

From our analysis so far, the minimum viable representation for program-level GHG accounting would include: (1) pasture-specific manure N and C cycling, (2) some representation of seasonal forage quality variation, and (3) ideally, a mechanism for diet quality to feed back into heifer growth rate. We'd welcome a conversation with the PML to understand what a contribution process would look like and whether our use case aligns with the team's longer-term vision.

Q4: Pasture diet representation

We implemented the 80:20 grass-legume split you suggested (ID 92: 77.81%, ID 105: 19.45%, ID 301: 2.74% mineral). This aligns well with the University of Minnesota guidance on mixed sward grazing - thank you for the reference. The legume addition also reduced feed LUC slightly relative to our pure-grass scenario, which is directionally sensible (legumes fix nitrogen, reducing off-farm input requirements).

We understand and have documented the growth-rate limitation: if the pasture ration fails to meet heifer energy requirements, RuFaS falls back to the prior ration rather than reducing growth targets. We're treating the current results as directional — a conservative floor — and will monitor issue #577 for when the heifer ration reduction method is implemented.

Q5: Soil carbon

This is the most significant remaining gap for our use case. Our program partners need to be able to quantify soil carbon sequestration under perennial pasture to make a credible case for heifer grazing as a practice intervention — without it, the ~4% GHG intensity reduction we're currently showing is likely a substantial undercount.

The near-term step you described - simulating a cool-season grass field with daily or frequent manure application - is exactly what we'd like to configure. We've already read through the Soil and Crop Module SciDoc and have a few specific questions before we start building:

• What crop species name should we use for cool-season perennial grass in the crop schedule? Is there an existing species defined in the model (e.g., analogous to corn_silage) or would this require defining a new one?
• Is there an example field configuration we could use as a starting point for a perennial grass system? The existing examples are all corn/alfalfa rotations.
• For the manure schedule, should we configure daily application to approximate DailySpread outputs, or is there a recommended approach for tying the Soil and Crop Module to the manure module's direct-deposition output?

We're happy to contribute the resulting configuration files back to the repo as an example pasture scenario once we have something working. Looking forward to our chat next month - and thank you again to the whole team for the depth of these responses.