When 95% of Your Scope 1 Is Diesel: Carbon...

An environmental superintendent at a gold mine once described their diesel accounting process as "counting grains of sand with boxing gloves on." The mine consumed 40 million litres of diesel in a single year. They knew the total. What they couldn't tell you, with any confidence, was how much went to haul trucks versus excavators versus light vehicles versus generators. And under NGER, that split matters.

This is the paradox of mining carbon accounting in Australia. The emissions profile looks deceptively simple. For most metal ore mines, 90% to 95% of Scope 1 emissions come from one source: burning diesel. The emission factor is published. It's 69.9 kg CO2-e per gigajoule, or about 2.7 kg CO2-e per litre. Multiply litres by the factor. Done.

Except it's not done. Not even close. Because the number that matters isn't the factor. It's the litres. And at a large open-cut operation running 24 hours a day, 365 days a year, across a fleet of 200 haul trucks, 30 excavators, 15 drill rigs, a dozen dozers, water carts, graders, and 150 light vehicles, the diesel data is scattered across four or five systems that don't talk to each other. Reconciling them into a single accurate number is where the real work lives.

We built our careers on enterprise data platforms in the resources sector before starting Carbonly. We know what a mining fuel management system looks like from the inside. We know what the bowser logs say versus what the tanker delivery docket says versus what the fuel card statement says. And we know they almost never agree.

Four fuel data systems, four different truths

A typical large mining operation tracks diesel through at least four sources, each with its own version of reality.

Bulk fuel deliveries are the supplier invoices from tanker loads delivered to site storage tanks. These are your "fuel in" records. They arrive as PDF invoices or email attachments, usually monthly, showing total litres delivered per tank per delivery date. This is the most reliable volume data you'll get, because it's what the supplier is billing you for. But it only tells you aggregate consumption. A 60,000-litre delivery to the main diesel tank doesn't tell you whether those litres ended up in haul trucks or generators.

Bowser metering systems (or fuel management systems like Fleetwatch, OPW, or FuelMaster) track dispensing at the pump. Every time a haul truck fills up, the operator swipes a tag, the bowser records the volume, the equipment number, and the timestamp. In theory, this gives you equipment-level consumption data. In practice, tag readers malfunction. Operators use the wrong tag. Night-shift fill-ups get recorded against "unknown vehicle" because the system timed out. And the bowser meter drifts by 1% to 3% per year without regular calibration. On 40 million litres, a 2% drift is 800,000 litres of unallocated fuel.

Fuel card transactions cover light vehicles and road-registered equipment that refuel at commercial service stations, either on-site or off. These generate clean transactional data with vehicle rego, litres, date, and location. But they only capture a fraction of total diesel. On most mine sites, fuel cards account for 5% to 10% of total diesel consumption.

Telematics from equipment OEMs provide fuel burn rates estimated from engine hours and load data. Systems from major equipment manufacturers push data to fleet management dashboards showing litres per hour, litres per tonne hauled, and estimated fuel consumption per shift. This is useful for operational efficiency. But it's modelled, not measured. Telematics estimates typically diverge from actual consumption by 5% to 15%, depending on payload, grade, and operator behaviour.

So you've got supplier invoices saying the site consumed 3.4 million litres last month. The bowser system says 3.25 million was dispensed. Fuel cards add 180,000. And telematics across the fleet models 3.5 million. The gap between the highest and lowest number is 250,000 litres. At 2.7 kg CO2-e per litre, that's 675 tonnes of emissions you either report or don't. And under the Safeguard Mechanism, where every tonne above your declining baseline costs you an ACCU at $35 to $40, that gap has a dollar figure attached to it.

Where the litres disappear

The mismatch between bulk deliveries and recorded consumption isn't a mystery. It has well-understood causes. But most mining environmental teams don't have the time or systems to quantify them properly.

Evaporation from above-ground storage tanks in the Pilbara or Bowen Basin, where ambient temperatures hit 45 degrees Celsius, accounts for measurable losses. Spillage during vehicle refuelling or tank transfer operations creates unrecorded consumption. Tank slosh during transport between satellite depots adds variance. And plain old measurement error, from tank dipstick readings at delivery to bowser meter calibration drift, introduces systematic inaccuracy across the reporting period.

The NGER Measurement Determination doesn't require you to account for these losses individually. Under Method 1 for stationary energy, you report fuel consumed, and the NGA emission factor handles the rest. But here's the problem: if your "fuel consumed" figure is derived from a bowser system that's undercounting by 2% to 3%, your reported emissions are correspondingly low. And when the Clean Energy Regulator audits your facility, they'll compare your reported diesel consumption against supplier delivery records. If those numbers don't reconcile within a reasonable tolerance, you've got an audit finding.

The ANAO performance audit found that 72% of 545 NGER reports contained errors, with 17% classified as significant. Gaps in energy consumption data and missing sources were among the most common problems. For mining companies handling thousands of fuel transactions per month across multiple facilities, these aren't hypothetical risks. They're statistical certainties over a long enough reporting horizon.

We're honest about something here: no software, including ours, can solve the physical reality of fuel losses between a tanker delivery and a bowser reading. What we can do is flag the discrepancy automatically so your team investigates it in real time, rather than discovering a 600,000-litre gap three weeks before the October 31 NGER deadline.

The transport versus non-transport split that nobody gets right on the first try

NGER requires mining companies to classify diesel combustion as either transport or non-transport. This sounds trivial. It isn't.

Transport emissions cover road-registered vehicles. Light vehicles, service trucks, road-registered water carts, buses that transport workers. Non-transport covers everything else: haul trucks (which operate under mine site registration, not road registration), excavators, drill rigs, dozers, graders, pumps, generators, and mobile plant.

The emission factors are different. The reporting categories are different. And the Safeguard Mechanism treats them differently because transport emissions from road-registered vehicles can fall under a separate reporting pathway.

The complication is that mining fleets don't sort themselves neatly. A water cart might be road-registered for the highway between two pit areas but spend 95% of its time on haul roads within the mining lease. A service truck might be road-registered but only ever driven on mine site roads. The classification depends on registration status, not usage pattern. But the fuel management system records consumption by equipment number, not by registration status. So someone needs to maintain a crosswalk table mapping every piece of equipment to the correct NGER category.

On a mine site with 400 pieces of mobile equipment, that crosswalk changes quarterly as plant arrives, leaves, or is re-registered. If nobody updates it, your transport/non-transport split drifts further from reality with each reporting period.

Carbonly's AI Document Processing engine handles this by matching fuel records against your equipment register during data extraction. Each fuel transaction gets tagged to a specific asset and its NGER category at the point of ingestion, not months later when someone's building the annual return. But the register itself still needs to be current, and that's an operational discipline no software can enforce for you.

Remote sites, batch data, and the connectivity problem

Mining operations 500 kilometres from the nearest town don't have reliable broadband. Some don't have reliable mobile coverage. The fuel management system at a remote satellite pit might store data locally and sync when someone drives a USB stick back to the main operations centre. Bowser readings might get written on a whiteboard in the crib room and transcribed into a spreadsheet weekly.

This creates a batch-data problem that's different from what urban businesses face. A property manager in Sydney gets monthly electricity bills by email. A mining environmental team in the eastern Goldfields gets fuel data from three satellite pits in batches, sometimes weekly, sometimes monthly, sometimes whenever the satellite internet cooperates.

The practical consequence is that mining emissions data is never complete in real time. It arrives in chunks, often weeks after the consumption occurred. And each chunk needs to be validated against what's already been recorded to avoid double-counting. A bulk delivery invoice from March might not arrive until May. The bowser system data for the same period was uploaded in April. If both get entered without reconciliation, you've counted those litres twice.

We see this pattern across field site data collection in construction and mining. The people who have the data are on remote sites. The people who need it are in head office. The gap between those two groups is where errors breed.

Carbonly's email ingestion system helps here. Each mining facility gets a dedicated ingest address. Site administrators forward fuel delivery invoices, bowser exports, and fuel card statements to that address as they arrive. The AI engine extracts volumes, dates, fuel types, and equipment identifiers, then flags duplicates and reconciliation gaps automatically. It doesn't solve the connectivity problem. But it compresses the time between "document received" and "data validated" from weeks to minutes.

Fugitive emissions: when diesel isn't the whole story

For coal mines, diesel is only part of the Scope 1 picture. Fugitive methane from coal seams can dwarf combustion emissions at some operations. The Clean Energy Regulator is phasing out Method 1 default factors for Safeguard Mechanism coal facilities, requiring a transition to site-specific Method 2 or Method 3 measurement. From 1 July 2026, all remaining Safeguard Mechanism coal facilities must have moved off Method 1. Currently, 22 of 51 open-cut coal facilities still rely on it.

For metal ore mines, fugitive emissions are usually negligible. But there's a common blind spot: refrigerant losses from air conditioning systems in workshops, processing plants, and site accommodation. R-410A (GWP of 2,088) and R-407C (GWP of 1,774) are standard in mine site HVAC. On a large operation with 200 split systems across workshops, offices, and dongas, annual refrigerant losses of 5% to 10% can add up to 200 to 400 tonnes CO2-e. Not huge relative to diesel. But enough to push a facility closer to its Safeguard baseline, and enough to trigger an audit finding if omitted from your NGER return.

Scope 2 at mine sites is simpler than you'd think (or more complex)

Here's something that confuses people outside the mining sector. Many remote mine sites don't have Scope 2 emissions at all. If your site generates all its electricity from on-site diesel generators, that electricity is Scope 1, not Scope 2. The diesel was already counted in your combustion figures. Scope 2 is only for purchased electricity from the grid.

Grid-connected mine sites have straightforward Scope 2 calculations. Pull the kWh from your electricity bills, apply the state-based NGA emission factor, and you're done. But the state factor matters enormously. A processing plant in Queensland consuming 50 million kWh per year generates 33,500 tonnes CO2-e at the state factor of 0.67. The same consumption in Western Australia's SWIS grid generates 25,000 tonnes at 0.50. And in Tasmania, it would be just 10,000 tonnes at 0.20.

Mining companies with operations spanning multiple states need to apply state-specific factors to each facility. Using the national average of 0.62 kg CO2-e/kWh will produce incorrect results in every single state except by coincidence. Under AASB S2 paragraph 29(a)(v), location-based Scope 2 measurement is mandatory, with market-based reporting available as a voluntary supplement. If you're holding renewable energy certificates or have a power purchase agreement, you'll want to report both methods.

Scope 3: the categories that keep expanding

Mining companies know their Scope 1 inside out (even if the data collection is painful). Scope 2 is manageable. Scope 3 is where it gets murky, and where AASB S2 is forcing new disclosure obligations starting from the second reporting period.

The material Scope 3 categories for mining typically include Category 1 (purchased goods and services), which covers explosives, process chemicals, tyres, grinding media, and other consumables. ANFO alone has an emission factor of approximately 0.17 kg CO2-e per kg. A large open-cut mine might use 20,000 to 40,000 tonnes of ANFO per year. That's 3,400 to 6,800 tonnes CO2-e from explosives alone, before accounting for the broader supply chain footprint.

Category 4 (upstream transportation) covers haulage of inputs to site. Category 9 (downstream transportation) covers ore transport from mine gate to port or processing plant, often by rail or road train. Category 6 (business travel) includes FIFO flights, which for a remote mine site running two-week rosters with 500 to 1,000 workers can represent significant emissions.

We'll be direct about the state of Scope 3 in mining: it's still rough. Spend-based estimates carry 30% to 40% error margins. Supplier-specific data is patchy. Collecting Scope 3 data from suppliers remains one of the hardest problems in carbon accounting, and mining supply chains are long, fragmented, and often dominated by a small number of bulk commodity suppliers who may or may not have their own emissions data.

AASB S2 provides modified liability protection for Scope 3 disclosures during the first three years. That's deliberate. The standard-setters know the data isn't perfect yet. But "protected from private litigation" doesn't mean "free to guess." ASIC can still act on materially misleading disclosures, and the ACCC has been actively enforcing against greenwashing claims that don't hold up to scrutiny.

Safeguard Mechanism: where inaccurate data hits your P&L

Mining facilities are the backbone of the Safeguard Mechanism. The scheme covers facilities emitting more than 100,000 tonnes CO2-e per year, and mining and resources represent the largest share of covered entities.

Baselines decline by approximately 4.9% per year through to 2030. That's not negotiable. A facility that started with a baseline of 500,000 tonnes in FY2024 will be at roughly 403,500 tonnes by FY2028. If actual emissions haven't changed, the exceedance is 96,500 tonnes. At current ACCU prices of $35 to $40 per tonne, that's $3.4 million to $3.9 million. Per facility. Per year.

In FY2024, 147 out of 219 covered facilities exceeded their baselines. That's 67%, up from 18% the year before. Covered facilities collectively surrendered 7.1 million ACCUs. The gap between reported emissions and declining baselines is growing, and the 2026-27 review will set decline rates for the next decade.

What does this have to do with diesel data accuracy? Everything.

If your bowser system undercounts by 2% and you consume 50 million litres per year, that's 1 million litres unreported. At 2.7 kg CO2-e per litre, you've understated Scope 1 by 2,700 tonnes. The CER finds the discrepancy, corrects your numbers, and suddenly you've exceeded your baseline by an additional 2,700 tonnes. At $37 per ACCU, that's $99,900 in credits you owe. Plus the audit remediation costs, the potential enforceable undertaking (see what happened with Beach Energy's NGER undertaking in July 2025), and the reputational risk.

Conversely, if your reconciliation process overcounts, you're surrendering ACCUs for emissions you didn't produce. Same dollar exposure, different direction. Getting the data right isn't a compliance nicety. It's financial management.

What a working system looks like at mining scale

We'll spare you the theoretical framework. Here's what actually needs to happen to produce defensible NGER returns from a mining operation processing 8,000 or more fuel transactions per quarter.

First, establish one source of truth for each fuel type. For bulk diesel, that's supplier delivery invoices reconciled against tank dipstick readings at delivery. Not the bowser system (too many gaps). Not telematics (modelled, not measured). The bowser system is your allocation tool, telling you where the fuel went. The delivery invoices are your total consumption. They should reconcile within 2% to 3%. If they don't, investigate the gap before reporting, not after.

Second, automate the document processing. Diesel delivery invoices from three suppliers, in three different formats, arriving as PDF email attachments on different schedules. Bowser data exported as CSV from the fuel management system. Fuel card statements in yet another format. Manually transcribing these into a spreadsheet is where the errors creep in. The ANAO's 72% error rate in NGER reports didn't happen because people used wrong emission factors. It happened because the underlying activity data was entered incorrectly.

Carbonly processes all of these document types through the same AI extraction engine. PDF invoices, CSV exports, Excel bowser reports, even photographed delivery dockets. Each document gets classified, extracted, validated against expected ranges, and matched to the right facility and NGER emission source category. The 5-tier material matching system handles the fact that one supplier codes diesel as "Automotive Distillate" while another calls it "ULSD" and a third just writes "Diesel." All three get matched to the same NGA emission factor.

Third, reconcile monthly, not annually. The October 31 NGER deadline creates a perverse incentive to defer everything to Q4. Don't. A mining company with five registered facilities should be closing each facility's fuel data within 30 days of month-end. That means reconciling bulk deliveries against bowser totals, investigating discrepancies above threshold, and locking the numbers. By the time October rolls around, you're compiling a return from 12 already-validated monthly datasets, not scrambling to reconstruct a year's worth of fuel records from raw invoices.

Carbonly's Anomaly Detection module flags unusual patterns as they appear. A haul fleet consuming 15% more diesel than the same month last year without a corresponding production increase. A generator running 20% more hours than budget. A satellite pit showing zero fuel consumption for two weeks (which usually means the data didn't upload, not that they stopped mining). These are the signals that catch problems early, when they're cheap to fix.

Fourth, maintain the audit trail. NGER requires records kept for five years. For a mining operation, that means every delivery invoice, bowser export, fuel card statement, and equipment register update needs to be preserved and traceable to the reported number. When the environmental manager who built the current system moves on (and in mining, turnover in environmental roles is high), the next person needs to understand not just what was reported, but why. What assumptions were made about the transport/non-transport split. How the bowser gap was allocated. Which facilities share fuel storage.

This is where Carbonly's Source Tracking and Audit Trail modules matter most. Every emission figure traces back to the source document, the extraction result, the emission factor applied, and any manual adjustments. The methodology lives in the system, not in someone's head.

Don't start with Scope 3. Start with the bowser gap.

The temptation for mining companies approaching AASB S2 obligations is to panic about Scope 3 first. Don't. Your Scope 1 diesel data is your largest emission source, your biggest Safeguard Mechanism exposure, and the number the CER will audit first. If you can't reconcile your bulk fuel deliveries against your bowser dispensing records within 3%, you've got a more urgent problem than supplier engagement for Scope 3 Category 1.

Get the diesel right. Reconcile monthly. Automate the document processing so your environmental team spends time on analysis, not on typing numbers from PDFs. Then build outward to fugitive emissions, refrigerant losses, Scope 2 electricity, and eventually Scope 3 categories.

If your mining operation processes more than 5,000 fuel transactions per quarter and you're still reconciling in spreadsheets, we'd like to show you what automated extraction and reconciliation looks like at that scale. Reach us at hello@carbonly.ai or through the contact form on our site. Per-project pricing, no lock-in contracts.

Related Reading: