Uncertainty Bounds Methodology¶
Quantifying and reporting uncertainty in emission estimates.
Why Uncertainty Matters¶
Emission estimates are approximations. Reporting without uncertainty: - Implies false precision - Violates ISO 14064 requirements - Reduces credibility
Proper uncertainty reporting: - Enables informed decision-making - Supports third-party verification - Meets compliance standards
Sources of Uncertainty¶
1. Emission Factor Uncertainty¶
Factors are estimates, not measurements:
| Source | Typical Uncertainty |
|---|---|
| Provider-specific | ±10-15% |
| Academic research | ±25-35% |
| Model extrapolation | ±40-60% |
| Fallback estimates | ±80-100% |
2. Activity Data Uncertainty¶
Token counts from API responses:
| Source | Uncertainty |
|---|---|
| Verified API response | ±1% |
| Estimated from text | ±5-10% |
| Unknown | ±20% |
OpenClaw uses API-reported tokens: ±1% uncertainty.
3. Grid Carbon Uncertainty¶
Grid intensity varies:
| Source | Uncertainty |
|---|---|
| Real-time API | ±5% |
| Regional average | ±20% |
| World average | ±30% |
OpenClaw default (400 gCO₂/kWh): ±30% uncertainty.
Calculation Method¶
Combined Uncertainty¶
Using root sum of squares (RSS):
Combined = √(EF² + AD² + GC²)
Where: - EF = Emission factor uncertainty - AD = Activity data uncertainty - GC = Grid carbon uncertainty
Example:
Combined = √(35² + 1² + 30²)
= √(1225 + 1 + 900)
= √2126
= 46%
Simplified Approach¶
OpenClaw uses confidence-based mapping for simplicity:
function confidenceToUncertainty(confidence: number) {
if (confidence >= 0.7) return { lower: 0.85, upper: 1.15 }; // ±15%
if (confidence >= 0.5) return { lower: 0.70, upper: 1.30 }; // ±30%
if (confidence >= 0.3) return { lower: 0.50, upper: 1.50 }; // ±50%
return { lower: 0.00, upper: 2.00 }; // ±100%
}
Reporting¶
Point Estimate with Range¶
12.45 kg CO₂eq (8.7 - 16.2 kg, ±30%)
Confidence Interval¶
12.45 kg CO₂eq
95% CI: 8.7 - 16.2 kg
As Percentage¶
12.45 ± 3.7 kg CO₂eq (±30%)
CLI Output¶
openclaw green intensity
Output:
Carbon Intensity Metrics (TCFD)
Per million tokens: 142.50 gCO₂eq
Per API call: 6.7200 gCO₂eq
Uncertainty range: 70% - 130%
Export Formats¶
GHG Protocol¶
{
"scope3Category1": {
"emissions_tCO2eq": 0.01245,
"uncertainty_percent": 30
}
}
ISO 14064¶
{
"emissions": {
"scope3Cat1_tCO2eq": 0.01245,
"uncertainty_tCO2eq": 0.00374
}
}
TCFD¶
{
"uncertainty": {
"lower": 0.70,
"upper": 1.30
}
}
ISO 14064 Requirements¶
ISO 14064-1:2018 requires:
- Identify uncertainty sources
- Quantify where possible
- Report overall uncertainty
- Explain methodology
Required Documentation¶
## Uncertainty Assessment
### Sources
1. Emission factors: No supplier-specific data (±35%)
2. Activity data: API-reported token counts (±1%)
3. Grid carbon: World average used (±30%)
### Methodology
Combined using root sum of squares:
√(35² + 1² + 30²) = 46%
Simplified to ±50% based on confidence mapping.
### Overall Uncertainty
Emissions: 12.45 kg CO₂eq
Uncertainty: ±50% (6.2 - 18.7 kg)
Verification¶
Third-party verifiers assess uncertainty:
| Level | Materiality | Requirements |
|---|---|---|
| Limited | Higher | Plausibility check |
| Reasonable | Lower | Detailed testing |
Most Scope 3 Cat 1 emissions verified at limited assurance.
Reducing Uncertainty¶
Short Term¶
- Use consistent methodology
- Document all assumptions
- Track changes over time
Medium Term¶
- Engage providers for data
- Apply regional grid factors
- Implement real-time monitoring
Long Term¶
- Industry standardization
- Provider disclosure requirements
- Direct measurement capabilities
Best Practices¶
- Always report uncertainty — Never imply false precision
- Use conservative estimates — Upper bound, not best case
- Document methodology — Enable verification
- Update regularly — Improve as data improves
- Be consistent — Same methodology year over year