**Sustainable research** practices are transitioning from optional to expected. Horizon Europe's 2024 work programme explicitly requests **environmental impact of research** assessments. ERC reviewers increasingly favor proposals that acknowledge computational **carbon emissions**—not because they require mitigation, but because environmental awareness demonstrates responsible stewardship.
This **Green Algorithms** calculator addresses three proposal needs:
- Horizon Europe Sustainability Sections - Many calls request environmental impact descriptions, particularly for computational projects
- NSF Broader Impacts - Sustainability considerations strengthen broader impacts narratives beyond traditional education/outreach
- Institutional Requirements - Some universities now require carbon accounting for high-performance computing resource requests
For context on how sustainability requirements emerged, see The Sustainability Paradox on short-term grants vs. long-term research.
How This Green Algorithms Calculator Works
The calculator implements the **Green Algorithms** framework, developed by Lannelongue et al. (2021) and validated across computational biology, machine learning, and climate modeling workloads.
Input Parameters:
- Hardware Type - GPU models (V100, A100, RTX 3090) or CPU cores with TDP ratings
- Runtime Duration - Hours of computation for training runs, simulations, or analyses
- Geographic Region - Carbon intensity varies dramatically: 12g CO₂/kWh (Iceland) to 820g CO₂/kWh (India)
- Hardware Efficiency - Power Usage Effectiveness (PUE) adjusts for data center overhead
Output Metrics:
- Total **carbon emissions** in kg CO₂ equivalent
- Equivalent miles driven or tree-years needed for offset
- Cost estimates if carbon pricing applied
These metrics can be incorporated into **grant sustainability** statements. For writing compelling sustainability narratives, avoid the traps described in The Sustainability Smokescreen.
Integrate Sustainability Throughout Your Proposal
While calculators quantify footprint, **Proposia** helps you weave sustainability into research design, methodology, and broader impacts. Our AI identifies opportunities to demonstrate environmental responsibility across all proposal sections.
Try Proposia FreeIncluding Carbon Footprint in Grant Proposals
Strategic placement of **carbon emissions** data strengthens multiple proposal sections:
Research Plan / Methodology: "Training our transformer model on 8 V100 GPUs for 72 hours is estimated to produce 284 kg CO₂ equivalent (US grid average), comparable to a round-trip flight from New York to London. We will optimize hyperparameters on smaller datasets before full training runs to minimize computational waste."
Broader Impacts (NSF): "This project demonstrates **sustainable research** practices by: (1) benchmarking computational efficiency against published baselines, (2) sharing trained models to prevent redundant training runs, and (3) selecting data centers with renewable energy commitments when available."
Ethics Sections (Horizon Europe): "We acknowledge the **environmental impact of research** from computational requirements. Our preliminary estimates suggest 1.2 tons CO₂ equivalent over the project lifetime. We commit to monitoring actual emissions and reporting in project deliverables."
For comprehensive guidance on ethics sections, consult The Ethics Section Nobody Reads (until it kills your grant).
Green Algorithms Best Practices
1. Report Transparently - Acknowledge computational footprint even when large. Honesty demonstrates integrity; concealment suggests ignorance or evasion.
2. Contextualize Emissions - Compare to field norms, not absolute zero. "Our 500 kg CO₂ training run represents a 40% reduction compared to published benchmarks through architectural optimization."
3. Highlight Reusability - Trained models, datasets, and code sharing prevent redundant computation by other researchers. This multiplies sustainability impact.
4. Don't Greenwash - Avoid vague commitments without specifics. "We will explore sustainability options" signals checkbox compliance. "We will benchmark against CodeCarbon and report monthly" signals genuine commitment.
5. Connect to Broader Impacts - Frame computational efficiency as methodological rigor, not just environmental consideration. Optimized code runs faster and costs less—benefits beyond **carbon emissions**.
NSF's Broader Impacts criterion provides natural sustainability integration opportunities. See examples in our Broader Impacts guide.
Geographic Variation in Carbon Intensity
**Carbon emissions** from identical computational workloads vary 70-fold based on electricity grid composition:
| Region | Carbon Intensity (g CO₂/kWh) | Primary Energy Source |
|---|
| Iceland | 12 | Geothermal + Hydro |
| France | 85 | Nuclear |
| USA (average) | 417 | Mixed (natural gas/coal/renewables) |
| China | 681 | Coal-dominant |
| India | 820 | Coal-dominant |
When proposing international collaborations or cloud computing, acknowledging these differences demonstrates sophistication. "We selected Google Cloud's Iowa region (carbon intensity 478 g/kWh) over Virginia (347 g/kWh) due to better GPU availability, accepting a 38% higher carbon footprint."
Complementary Proposal Tools
Ethics & Compliance
Sustainability often connects to ethics sections. Review The Ethics Section for Horizon Europe compliance strategies.
Build Sustainable, Fundable Proposals
**Proposia** doesn't just help you calculate **carbon emissions**—it helps you integrate **sustainable research** principles across your entire proposal. From methodology to broader impacts, our AI identifies opportunities to strengthen environmental responsibility narratives.
Methodology Citation: Lannelongue, L., Grealey, J., & Inouye, M. (2021). Green Algorithms: Quantifying the Carbon Footprint of Computation. Advanced Science, 8(12), 2100707. https://doi.org/10.1002/advs.202100707
Alternative Tools: ML CO2 Impact (machine learning specific), CodeCarbon (Python library for real-time tracking), Green Algorithms Calculator (official web tool).
Data Sources: Carbon intensity data from Electricity Maps and ML CO2 Impact dataset. Hardware TDP specifications from manufacturer documentation.
