Carbon Accounting for Manufacturing and Industrial Facilities in Australia

A cement plant in NSW burns natural gas in its kilns. That's a combustion emission — Scope 1, fuel-based, relatively easy to calculate from the gas bill. But the actual calcination of limestone into clinker releases about 0.52 tonnes of CO2 for every tonne of clinker produced. That's a process emission. It has nothing to do with energy. You can't find it on a meter reading. And it often accounts for more than half of the plant's total carbon footprint.

This is what makes manufacturing emissions reporting in Australia fundamentally different from property, retail, or professional services. For most sectors, your emissions profile is dominated by electricity (Scope 2) and maybe some natural gas. You can get 80% of the picture from utility bills. For manufacturing, the bills are just the beginning.

We've spent years building data extraction systems for utility bills at Carbonly, and that experience has taught us where automation works well and where it hits a wall. Manufacturing is where it hits a wall — and where we're still figuring out the best approaches for the process emission side. This article is the honest version of what manufacturing facilities actually need to account for under NGER and ASRS, and where the hard parts are.

The Four Emission Source Categories That Manufacturing Facilities Must Report

Under NGER, every emission at a facility falls into one of four categories. Most manufacturing facilities will have sources in all four. That's unusual — an office building only hits one or two.

Fuel combustion (Measurement Determination Part 2) is the one everyone understands. Natural gas boilers, diesel generators, LPG forklifts, furnaces, kilns. You burn something, it produces CO2, CH4, and N2O. The calculation uses fuel quantity multiplied by NGA emission factors. For a Victorian manufacturing plant running gas-fired boilers, you're looking at 51.4 kg CO2-e per GJ of natural gas consumed (NGA Factors 2025). If you consume 10,000 GJ per year, that's 514 tonnes CO2-e just from gas — before you even get to process emissions or electricity.

Industrial process emissions (Part 4) are where manufacturing diverges from every other sector. These are emissions from chemical or physical transformations of materials that aren't fuel combustion. The Measurement Determination has a separate division for each industry type: Division 4.2 for cement and lime, Division 4.3 for chemical industry processes, Division 4.4 for metal production, Division 4.5 for electronics manufacturing. The CER updated its industrial processes guidance in August 2025 specifically because these calculations are where facilities most often get it wrong.

Fugitive emissions (Part 3) cover unintentional releases — refrigerant leaks from HVAC and cold storage being the big one for most manufacturers. Under NGER, you only need to report HFCs with a GWP greater than 1,000. That exempts R-32 (GWP 675 under AR4), but here's the wrinkle: if your refrigerant blend contains any listed HFC with GWP above 1,000, you report all listed HFCs in that blend. Australia holds roughly 55,000 tonnes of refrigerants in systems nationally, equivalent to 100 million tonnes CO2-e if released. And the NGER framework's exemptions for systems under 100 kg likely mean significant underreporting across the sector.

Waste emissions (Part 5) catch many manufacturers off guard. If you operate on-site wastewater treatment — common in food processing, beverage manufacturing, and chemical production — the methane and nitrous oxide from that treatment process is Scope 1. It's your facility, your treatment plant, your emissions. A meat processing plant generating 85% of its fresh water intake as wastewater isn't just dealing with a water management problem. It's sitting on a material emission source.

Process Emissions: The Part That Doesn't Show Up on Any Bill

This is the section where most carbon accounting software — including ours, frankly — has the least automation to offer. Process emissions require industry-specific knowledge, and the data inputs aren't invoices or meter readings. They're production records.

For cement and clinker production, the NGER Measurement Determination offers three methods. Method 1 uses default emission factors — approximately 0.52 tonnes CO2 per tonne of clinker. Method 2 requires you to sample and analyse your clinker composition to derive a facility-specific factor. Method 3 uses continuous emissions monitoring systems (CEMS). Most Australian cement facilities use Method 1 or 2, and the difference matters: a facility producing 500,000 tonnes of clinker per year could see a variance of 10,000–20,000 tonnes CO2-e between methods, which is enough to shift your Safeguard Mechanism baseline position.

For metals production — steel from electric arc furnaces, aluminium smelting, ferroalloy production — Division 4.4 prescribes calculations based on carbon content of inputs (electrodes, reductants, carbonates) minus the carbon in outputs. This isn't something you can estimate from a spend-based approach. You need actual production data: tonnes of raw materials in, tonnes of product out, carbon content percentages. The CER's 2023-24 published data shows primary metal and metal product manufacturing at 27.1 million tonnes CO2-e from 85 facilities. That's an average of 319,000 tonnes per facility — well above the 100,000 tonne Safeguard Mechanism threshold.

Chemical manufacturing under Division 4.3 covers ammonia production (where natural gas is used as feedstock, not fuel — a distinction many facilities get wrong), soda ash, sodium cyanide, and any process using carbonates as raw materials. The emissions from using gas as a feedstock are fundamentally different from burning it for heat. Same molecule, completely different reporting treatment under NGER.

Food processing doesn't have dramatic process emissions like cement or metals. But it has two sources that routinely get missed. On-site wastewater treatment produces methane from anaerobic decomposition of organic matter in effluent ponds. And any fermentation process — brewing, baking, dairy processing — releases CO2 directly. These aren't big numbers individually, but for a food manufacturer running wastewater lagoons across multiple sites, the cumulative total can push you toward or over NGER thresholds. The CER's data shows food product manufacturing at 2.8 million tonnes CO2-e across 421 facilities — an average of just 6,700 tonnes each, but with enormous variance between a small bakery and a major meat processor.

Why the Safeguard Mechanism Changes the Maths for Big Manufacturers

If your manufacturing facility emits more than 100,000 tonnes CO2-e per year (Scope 1), you're covered by the Safeguard Mechanism. That's 219 facilities across Australia as of 2024-25, collectively responsible for 138.7 million tonnes — about 29% of national emissions.

The reform that matters: baselines decline by 4.9% per year to 2030. If your emissions don't decline at least that fast, you exceed your baseline. And exceeding costs real money.

In 2023-24 (the first year of the reformed mechanism), 7.1 million ACCUs were surrendered for compliance — up from 1.2 million the previous year. At a spot price around $37 per ACCU (as of early 2026), that's roughly $263 million in compliance costs across the scheme. The 2024-25 data shows a 48.9% increase in total exceedance, so those costs are climbing.

But here's the part that specifically affects manufacturing: 17 facilities received Trade-Exposed Baseline-Adjusted (TEBA) determinations in 2023-24. If you're a manufacturer competing against imports from countries without carbon pricing — steel, aluminium, cement, chemicals — you can apply for a reduced baseline decline rate as low as 1% instead of 4.9%. That's a massive difference. A facility emitting 200,000 tonnes with a 4.9% decline needs to cut 9,800 tonnes per year. At 1%, it's 2,000 tonnes. The gap — 7,800 tonnes at $37 each — is $288,600 per year in avoided ACCU costs.

If you think you might qualify for TEBA and haven't applied, you're leaving money on the table. The application goes to the Clean Energy Regulator, and you need to demonstrate your cost impact metric based on EBIT. It's not automatic.

Data Collection Across Complex Manufacturing Operations

Here's where we get practical. The theoretical framework of NGER categories is one thing. Actually collecting the data from a manufacturing facility with 15 different emission sources, four buildings, two shifts, and a maintenance team that replaced refrigerant last Tuesday but didn't record how much — that's the real problem.

Energy data is the easiest part, and it's still harder than you'd think for manufacturing. A single facility might have separate meters for production lines, office areas, cold storage, and compressed air. Some manufacturers have sub-metering; many don't. Without it, you're working off total site consumption from the electricity bill and the gas bill. That's fine for NGER (it reports at facility level), but it won't tell you where your emissions are concentrated or where to target reductions. We built Carbonly's utility bill extraction pipeline to handle exactly this type of messy document — but the bill only captures what the meter captures.

Production data for process emission calculations typically lives in manufacturing execution systems (MES), ERP platforms like SAP, or — more often than anyone wants to admit — in spreadsheets maintained by shift supervisors. Getting clinker production volumes, raw material inputs, or wastewater treatment throughput into a carbon accounting system usually means a manual data integration exercise. We're honest about this: there's no AI that reads your SAP production records and automatically calculates cement process emissions. Not yet. Probably not for a while.

Refrigerant tracking is almost always done on paper or in maintenance management systems (CMMS) that weren't designed for emissions reporting. When a technician tops up an air conditioning unit with 15 kg of R-410A, that information needs to make it from a work order into your emissions inventory. R-410A contains R-32 and R-125 — the R-125 component has a GWP of 3,170, well above the 1,000 threshold, so the whole charge is reportable. The calculation method under NGER section 4.102 requires you to know the mass of each HFC component and its GWP. Most facilities track total refrigerant charged but not the component breakdown. That's a reporting gap waiting to bite you during an audit.

Waste and wastewater data comes from environmental monitoring, which is often managed by a completely different team than the one doing carbon reporting. Wastewater BOD concentrations, treatment volumes, anaerobic vs aerobic processing methods — all affect the emission calculation. And the NGER Measurement Determination updated its N2O factors for estuarine effluent in the 2025-26 amendments, so last year's spreadsheet formulas might be using outdated factors.

The Practical Difference Between Methods 1 Through 4

The Measurement Determination gives you a choice of methods for most emission sources, numbered 1 through 4. This isn't just a bureaucratic exercise. Your method choice directly affects your reported number — and by extension, your Safeguard Mechanism position and NGER compliance exposure.

Method 1 uses default emission factors from the NGA Factors workbook. It's the simplest: multiply activity data by the published factor. For fuel combustion, this works well. For industrial processes, the default factors are national averages that may not reflect your actual facility chemistry. A cement plant with a higher proportion of MgCO3 in its raw materials will have different process emissions per tonne of clinker than the default factor assumes.

Method 2 requires facility-specific sampling and analysis. For fuel, this means getting your natural gas composition tested rather than using the default. For process emissions, it means analysing your actual raw material and product compositions. More accurate, but you need a lab and a sampling protocol.

Method 3 and Method 4 involve direct measurement — continuous emissions monitoring or periodic stack testing. These are expensive to set up (CEMS installations run into the hundreds of thousands of dollars) but give you the most defensible numbers. For Safeguard facilities, Method 4 is increasingly common because an accurate number that's lower than the default factor saves you money in ACCUs.

The CER doesn't mandate which method you use (with some exceptions). But here's the catch: once you move to a higher-tier method, the CER expects you to stay there. And if an audit reveals that Method 1 was significantly overstating or understating your emissions, the auditor will question why you didn't move to Method 2. The ANAO performance audit found 72% of NGER reports contained errors — and industrial process calculations were among the most common sources.

Where Your Electricity Emissions Fit In

After all the complexity of process emissions, calculating Scope 2 from your electricity bills almost feels like a relief. But manufacturing facilities are typically the heaviest electricity consumers outside the power sector itself, and the state-based emission factors create real variance.

A manufacturer consuming 15,000 MWh per year in Victoria generates 11,700 tonnes CO2-e at the 0.78 kg CO2-e/kWh state factor. The same consumption in South Australia? 3,300 tonnes. In Tasmania? 3,000 tonnes. That's not a rounding error — it's the difference between being comfortably below NGER facility thresholds and blowing right through them.

For ASRS reporting (which NGER reporters are now pulled into via Group 2), you need to report location-based Scope 2 as your primary method under AASB S2 paragraph 29(a)(v). Market-based is voluntary and supplementary. So if you've invested in a Power Purchase Agreement or GreenPower to reduce your Scope 2, that shows up in the supplementary disclosure but doesn't change your primary reported number.

What a System Actually Needs to Look Like

Most carbon accounting platforms — including many of the enterprise ones — were built for office-based organisations: electricity bills, gas bills, maybe some fleet fuel. Manufacturing needs something different.

You need a system that handles at least four distinct data input streams: utility bills for combustion and electricity, production records for process emissions, maintenance records for refrigerant tracking, and environmental monitoring data for waste and wastewater. And you need an audit trail that connects each reported number back to its source document, because the CER can and does ask for it. Records must be kept for five years from the end of the reporting year.

We built Carbonly's 18-module platform to handle the utility bill side of this really well — our AI Document Processing engine reads eight document formats (PDF, CSV, multi-sheet Excel, Word, PPT, RTF, images, and scanned documents) with 5-tier material matching and confidence scoring, extracting kWh, GJ, billing periods, and meter numbers from hundreds of bill formats without manual data entry. For process emissions, we're still building out the functionality. We won't pretend we've solved the cement clinker calculation through document scanning. That's a different kind of data problem, and it needs a different kind of solution — one that connects to your production systems, not just your inbox.

But several Carbonly modules are built for manufacturing complexity beyond utility bills.

The Anomaly Detection module uses AI-powered pattern recognition with five rule types and a full investigation workflow. When energy consumption at a production line spikes 35% without a corresponding increase in output, that's either an equipment malfunction or a data error — either way, you need to know before it shows up as an unexplained variance in your NGER return. For manufacturing facilities where energy-per-unit-of-output is a key efficiency metric, the anomaly detection catches the problems that spreadsheet formulas silently average away.

The Incident Management module tracks environmental incidents — chemical spills, emission releases, equipment failures, refrigerant venting — with a structured investigation workflow that links each incident to the affected emission records. When a chiller unit vents R-410A during a compressor failure, that's a reportable fugitive emission. Carbonly ensures it flows from the incident report into your NGER inventory with the correct GWP calculation.

The NGER-native compliance engine understands the Measurement Determination's structure — combustion, industrial processes, fugitive emissions, waste — and generates NGER returns directly. For manufacturing facilities that straddle multiple NGER divisions, the platform handles the complexity of different calculation methods applied to different source categories within the same facility.

The Carbon Planning scenario builder lets you model efficiency upgrades before committing capital — evaluate the emissions and cost impact of switching from gas-fired boilers to electric heat pumps, upgrading to variable-speed drives on compressed air systems, or installing rooftop solar on warehouse space. The built-in action library includes LED lighting, solar, and EV fleet conversions, with cost-benefit analysis that shows both the emissions reduction and the payback period.

And the LCA module enables product-level carbon footprinting — critical for manufacturers whose customers are increasingly asking for the carbon intensity of the specific products they're buying. If your client wants to know the CO2-e per tonne of your output, LCA gives you a defensible number that accounts for the energy, process emissions, and materials in your production chain.

The honest picture for most manufacturing facilities is that you'll still need specialist input for the first year of process emission calculations. But the energy, combustion, and fugitive side — which is where most of the ongoing data collection burden sits — that's where Carbonly eliminates the manual work.

Don't wait for a single platform that does everything perfectly. The NGER deadline is 31 October every year, no extensions available. Start with what you can automate now, and build out the process emission side in parallel.

Related Reading: