Units
1. AI in Environmental Engineering: Introduction, Boundaries, Responsibility and Ethics 2. Water and Wastewater Treatment Process Support 3. Air Quality and Emission Analysis 4. Waste Management and Circular Economy 5. EIA and Environmental Reporting 6. Environmental Data and Monitoring Analysis 7. Climate, Sustainability and Carbon Footprint 8. Geographic and Remote Sensing Data Interpretation 9. Regulatory Research and Verification 10. Environmental Data Analysis, Validation and Security with Python
Unit 7 / 10

Climate, Sustainability and Carbon Footprint

Gains:

  • Ability to configure carbon footprint calculation with scope 1/2/3, CO2e and emission factor
  • Ability to produce inventory draft, reduction scenario and sustainability report section with AI
  • Ability to verify the emission factors and GWP values used by AI with official sources

You're in the environmental department of a food production facility, and management is requesting the first corporate carbon footprint (CAI) report at the request of export customers. He has 12 months of electricity bills, natural gas consumption, fuel records of the company vehicle fleet and a list of suppliers. At the meeting, someone says, "Let's ask ChatGPT, it will be finished in an hour." TRUE; An LLM can produce an outline of the inventory, the accounting framework, and the narrative sections of the report in minutes. However, if the emission factors and GWP values ​​used by the model are incorrect or outdated, the number produced will not pass inspection, no matter how professional it looks. In this unit, we cover structuring the carbon footprint, accelerating it with AI, and the most critical step — verifying the factors with the official source.

Scopes: 1, 2 and 3

According to the GHG Protocol, corporate emissions are divided into three scopes. Placing an emissions in the correct scope is essential to avoid double counting and ensure consistency of the report.

Scope

Description

Typical environmental engineering examples

Scope 1

Direct emissions — sources owned/controlled by the company

Boiler/process natural gas combustion, company vehicles, fugitive

Scope 2

Indirect emission of purchased energy

Electricity purchased from the grid, purchased steam/heating-cooling

Scope 3

Other indirect emissions in the value chain

Purchased raw materials, waste disposal, employee transportation, logistics, product use

Scope 3 is generally the largest but least vague category; Based on supplier data and average factors. It is imperative to clearly write the report boundary (organizational boundary) and which Scope 3 categories are included.

CO2e and GWP

The climate impact of different greenhouse gases is not the same. To convert them all into a common currency, CO2 equivalent (CO2e) is used:

CO2e = Σ ( gas_mass_i × GWP_i )GWP = Global Warming Potential (100-year, IPCC AR reference)Example 100-year GWP values (source version must be confirmed): CO2 = 1 CH4 ≈ 28-30 N2O ≈ 265-273 Some HFCs = thousands

Attention: GWP values vary depending on the version of the IPCC assessment report (AR4, AR5, AR6). For example, the GWP of methane is given as 25 in AR4, 28 in AR5, and approximately 27-30 in AR6. LLM often does not specify or confuses which version it is using. You should write in your report which IPCC version you base it on and confirm the value from that report. False GWP biases the entire CH4/N2O contribution proportionally.

Sample carbon footprint calculation

Consider scope 2 electricity emissions. The formula is simple; what is critical is the correct emission factor.

Scope 2 emissions = Electricity consumption (kWh) × Grid emission factor (kg CO2e/kWh)Example: Annual consumption = 480,000 kWh National grid factor = 0.42 kg CO2e/kWh (← Obtain from OFFICIAL SOURCE) Scope 2 emissions = 480,000 × 0.42 = 201,600 kg CO2e = 201.6 tonnes CO2e / year

Similar logic for Scope 1 natural gas, factor source is different:

# Scope 1: natural gas combustion natural gas_m3 = 55_000 # annual consumption, m3ef_natural gas = 1.9 # kg CO2e / m3 (confirmed from NATIONAL/IPCC source) scope1_gas = natural gas_m3 * ef_naturalgasprint(scope1_gas / 1000, "ton CO2e") # -> 104.5 tons CO2e # Scope 2: electricityelectricity_kwh = 480_000ef_grid = 0.42 # kg CO2e / kWh (NATIONAL grid factor)scope2_sieve = electricity_kwh * ef_gridprint(scope2_sieve / 1000, "ton CO2e") # -> 201.6 tonnes CO2etotal = (scope1_gas + scope2_sieve) / 1000print("Total (Scope 1+2):", round(total, 1), "tons CO2e") # -> 306.1

Tip: Don't embed emission factors as a "magic number" in the code. Comment the source and date next to each factor (e.g. # ef_grid: 2024 national grid average, [source/date]). When the factor is updated, the report remains traceable and the auditor asks "where did this number come from?" When you ask, your answer is ready.

Inventory and scenario drafting with AI

LLM is strong at categorizing inventory, setting up accounting, and listing mitigation scenarios.

WEAK PROMPT:"Calculate the carbon footprint of our factory."STRONG PROMPT:"Set up a Scope 1 and 2 carbon inventory DRAFT for a food production facility. Inputs: electricity 480,000 kWh/year, natural gas 55,000 m3/year, diesel fleet 18,000 L/year. Your task:1) Place each item in the correct scope and write the calculation formula.2) Emission DO NOT fill in your factors NUMERIC; replace them with a placeholder '[EF — to be verified from national/IPCC source]' and indicate which official source it will be taken from. 3) Wherever you use GWP, clearly state the IPCC version (AR5/AR6). 4) Label any step where there is uncertainty or assumption with the 'ASSUMPTION' tag. Do not give a made-up factor or made-up source name.”

This prompt prevents the model from doing its most dangerous hallucination in the first place—generating a plausible-looking but made-up emission factor or a nonexistent source (fictitious report/DOI). You build the model skeleton and fill in the factors from the official source.

Mitigation scenario

Logic of a mitigation scenario: calculate baseline with current factor, apply intervention, show difference in tonnes CO2e.

Scenario: Providing 30% of the grid electricity from self-production with Rooftop Solar Power Plant Base Scope 2 = 201.6 tons CO2e Drawn from the grid = 70% → 480,000 × 0.70 × 0.42 = 141.1 tons CO2e Reduction = 201.6 − 141.1 = 60.5 tons CO2e / year

Verify emission factor and GWP source

The crucial principle of this unit: AI's framework is valuable, its factors are dubious. Verification workflow:

  1. For the grid electricity factor, use the most current value from the national authority (e.g. the country's official grid emission factor publication).
  2. Use IPCC guidance or national inventory report as basis for fuel combustion factors.
  3. Use the table of the IPCC assessment report (AR5 or AR6) for GWP values ​​and write the version in the report.
  4. Record the publication year and version of each factor; Factors may be updated every year.

mini case

A logistics company noticed that the CAI report it received from a consultant used a factor of 0.25 kg CO2e/kWh for Scope 2. The report was prepared with the help of an LLM and the model gave this value as an "average European grid factor". However, the official current grid factor of the country where the company is located was 0.44 kg CO2e/kWh. With the right factor, Scope 2 emissions have nearly doubled, making the company's declared reduction target unrealistic. The fix was caught in internal audit before it was released to the customer; because the engineer had the discipline to question the source of every factor. The picture changed completely when the “average” value of the LLM was replaced by the official country-specific value.

Common mistakes

  • Placing the emissions in the wrong scope or double counting by counting the same item in two scopes.
  • Using the emission factor produced by LLM without verifying its source.
  • Not specifying the IPCC GWP version (AR5 vs AR6) or mixing versions.
  • Using an overall "world average" for the grid factor rather than the official country-specific value.
  • Hiding the uncertainty of Scope 3 and presenting it as a single definitive number.
  • Unit error: mixing kg with tonnes, m3 with kWh; bypassing energy content cycles.
  • Forgetting that factors are updated on a yearly basis and carrying the previous year's value to the new report.

In summary

  • Corporate carbon footprint Scope is structured as 1/2/3; boundaries and scope should be clearly defined.
  • CO2e converts different gases into a common unit with GWP; The GWP value depends on the IPCC version and must be specified.
  • The basic calculation is as simple as “consumption × emission factor”; The whole difficulty is choosing the right factor.
  • LLM is a powerful drafting assistant for inventory skeleton, account structure and scenario construction.
  • LLM can hallucinate emission factor and source name; Numerical factors should always be filled in from the official source.
  • The AI ​​output is a draft; Factors, GWPs and totals cannot be reported without verification by official source and internal audit.

Application task

Draft a Scope 1 and Scope 2 carbon inventory for a small facility (either self-selected or hypothetical). First, write a strong prompt and generate the scope classification and accounting skeleton from the LLM with placeholder factors; Then fill in each emission factor and the GWP value you use by finding it from a real official source (national grid factor publication, IPCC AR5/AR6 table) and write the source and year. Calculate total tonnes CO2e in Python or manually, check unit consistency. Finally, define a reduction scenario (e.g. energy efficiency or self-generation) and calculate the annual reduction in tonnes CO2e. At the end of his report, "What information from LLM have I corrected/corroborated?" Add a short note titled.

Previous Next