Carbon intensity of global crude oil trading and market policy implications

Petroleum fuels supply about 32% of global primary energy, and under 1.5 °C climate scenarios future oil supply may plateau or decline by up to 40% from present levels. Certain sectors (e.g., aviation, petrochemicals) have limited near-term alternatives, making accurate life-cycle accounting crucial. Upstream extraction and midstream transport together emit ~1.9 Gt CO₂eq/year across a complex global trade network. Differentiating crude oil by well-to-refinery-entrance carbon intensity (CI) at pathway-level resolution could enable market and policy mechanisms to prioritize lower-carbon crude. However, data gaps, limited traceability, and low-resolution assessments hinder effective implementation. This study addresses the research gap by providing high-resolution, pathway-level CI estimates that incorporate blending and full transportation impacts to support policy and market decisions.

Regulatory initiatives (e.g., California’s Low Carbon Fuel Standard, EU Fuel Quality Directive, ICAO’s CORSIA) aim to differentiate fuels by CI but are constrained by insufficient data and supply chain traceability. Prior academic work, notably using OPGEE, has revealed heterogeneity of field-level upstream CI but typically stops short of mapping to marketable blends and destination-specific pathways. Transportation emissions have often been approximated using generic models (e.g., GREET), limiting resolution and policy relevance. The literature thus focuses more on sector-wide reduction opportunities than on enabling market incentives for low-CI crude trading. This study builds on and extends prior work by combining field-level upstream modeling with blend formation estimation and physics-based transport models to yield pathway- and destination-specific CI.

The study conducts a global, high-resolution life cycle assessment (LCA) of well-to-refinery-entrance CI, covering upstream (extraction) and midstream (transport) emissions. A network model of the crude supply chain is constructed where oil fields, pipeline stations, shipping terminals/ports, and refineries are nodes; pipelines and shipping routes are edges. Data sources include commercial datasets (Wood Mackenzie, GlobalData, Kpler, S&P Global) and public datasets (NASA MODIS for temperature, Shuttle Radar Topography for elevation). Nodes carry geospatial and physical attributes; edges include mode-specific attributes (pipeline diameter, length, elevation change; shipping route distance, vessel class). Heuristic edges approximate intra-field connections where data are missing. Blend formation is estimated per producing country via a multi-objective optimization that maps field volumes to marketed blends. The configuration matrix Θ (fields × blends) is optimized to minimize a weighted cost comprising distance, connectivity, volume matching, and API matching. The approach uses gradient-based optimization with automatic differentiation and momentum, plus an initialization procedure incorporating name similarity, clustering, and a genetic algorithm to reduce local minima issues. Uncertainty is assessed by varying algorithm parameters (e.g., weighting of proximity, connectivity). Upstream CI: Field-level upstream emissions are modeled using OPGEE v3.0c, with energy-based allocation between co-produced oil and condensate. The upstream dataset covers ~98% of global crude and condensate production in 2015 (95% oil, 3% condensate). Midstream CI: Pathways from fields to refineries are determined by shortest-path tracking on the network (weighted by pipeline length and shipping distance). Pipeline emissions are estimated with COPTEM, a first-principles fluid mechanics model using the Energy, Darcy-Weisbach, and Colebrook-White equations, segmenting pipelines into 40 segments and accounting for viscosity-temperature effects and elevation changes. Shipping emissions are estimated bottom-up from AIS-derived activity, vessel parameters, and power-based models for propulsion and auxiliary loads, applying mode- and class-specific emission factors. Barge, trucking, and rail are excluded due to limited data and their minor global share. Results are aggregated at multiple policy-relevant levels: pathway (field→refinery), blend, producer–consumer country pairs, and consumer country totals. Uncertainty bands reflect parameter variations and blend estimation variability.

• Pathway-level well-to-refinery-entrance CI is highly variable: 4.2–214.1 kg-CO₂eq/bbl, with a volume-weighted average ~50.5 kg-CO₂eq/bbl (equivalently, 0.74–39.41 gCO₂/MJ with a volume-weighted average 9.01 gCO₂/MJ or 50.46 kg-CO₂eq/bbl).
• Upstream dominates: global volume-weighted upstream CI ≈ 45.03 kg-CO₂eq/bbl; midstream ≈ 5.37 kg-CO₂eq/bbl (~10% of total). Despite smaller magnitude, midstream CI exhibits large variability across pathways for a given blend, with skewed, multi-modal distributions in export-heavy blends (e.g., Arab Medium, Merey, Basrah Light).
• Blend-level highlights: Canadian Oil Sands Synthetic exhibits among the highest upstream CI (≈144.5 kg-CO₂eq/bbl); Western Canadian Select shows high well-to-refinery CI among top blends. Middle East blends display wide CI range (≈3.4–181.6 kg-CO₂eq/bbl). Countries with fewer/predominant blends (e.g., Saudi Arabia’s Arab Light) show lower CI uncertainty; countries like Canada, Venezuela, Iran show wide inter/intra-blend variability.
• Transport mode/distance effects: Global average per-thousand-mile CI is ~5.56 kg-CO₂eq/bbl for pipelines and ~2.28 for shipping; global volume-weighted averages per barrel are similar for pipeline and shipping (≈2.55 and 2.61 kg-CO₂eq/bbl), but region-specific patterns matter. North American transport CI is elevated (≈8.7–12.1 kg-CO₂eq/bbl) due to extensive pipeline miles and dispersed refineries; Russia’s pipeline-centric exports yield ~1.5–5.1 kg-CO₂eq/bbl midstream CI.
• Shipping routes: Latin America→Asia shipping CI (~10.7 kg-CO₂eq/bbl) exceeds Middle East→Asia (~5.2 kg-CO₂eq/bbl), driven by smaller tankers, longer routes, and lower efficiency. Examples (kg-CO₂eq/bbl): Venezuela→India 15.29; Colombia→China 16.05; Mexico→Japan 14.10; vs Iraq→India 5.18; Iran→China 2.07; Saudi Arabia→Japan 4.20.
• Consumer countries: Among nations with >1 Mbbl/d refining, well-to-refinery CI ranges from ~8.84 to ~86.39 kg-CO₂eq/bbl; producers consuming domestic crude reflect domestic upstream CI and pipeline lengths (e.g., Canada, U.S. higher midstream), while import-dependent regions relying on shipping (notably parts of Asia) show marginally higher refining-attributed CI than regions with pipeline access (e.g., Western Europe).
• Scenario analysis: Prioritizing low-CI pathways under 1.5 °C scenarios to 2050 yields additional cumulative savings of ~1.5–6.1 Gt CO₂eq (average ~4.5 Gt) beyond reductions from supply decline or process controls, comparable to removing ~100 million new gasoline cars (10-year lifetime, 4.6 t CO₂/year).

By resolving emissions at the level of source field to destination refinery pathways and explicitly modeling blend formation and transport, the study addresses the lack of traceability and low-resolution CI estimates that have limited policy efficacy. The results show that a relatively small midstream share can still drive substantial pathway-level variability, creating actionable opportunities for targeted interventions (e.g., optimizing routing, favoring efficient shipping lanes and larger vessels, leveraging pipeline efficiencies). The blend- and country-level aggregations demonstrate how domestic upstream CI and infrastructure shape national footprints, informing both producing and importing countries’ strategies. The quantified heterogeneity underpins market-based instruments that reward low-CI blends and pathways, potentially shifting crude valuation (premiums for lower CI) and enabling regulators (e.g., LCFS, EU FQD, CORSIA) to integrate CI differentiation at refinery gates and within sectoral policies.

This work provides a global, high-resolution, pathway-aware LCA of crude oil well-to-refinery-entrance CI by integrating OPGEE-based upstream modeling, a novel blend estimation algorithm, and physics-based transport emissions models informed by comprehensive datasets. It quantifies wide CI variability across blends, routes, producer–consumer pairs, and consumer countries, with upstream dominating totals but midstream driving significant dispersion. Scenario analysis shows that simply prioritizing low-CI supply chain pathways could deliver additional 1.5–6.1 Gt CO₂eq savings by 2050 under 1.5 °C scenarios, complementing process-oriented decarbonization (e.g., flaring reduction, CCS). The framework supports policy and market mechanisms to differentiate crude by CI and encourages real-time, granular emissions reporting (potentially via digital technologies such as blockchain). Future research should incorporate temporal dynamics in operations and logistics, expand data on gathering/boosting and fugitive emissions, and improve representation of intra-field infrastructure to further enhance accuracy and policy relevance.

• Data gaps: Bottom-up data for natural gas gathering and boosting emissions are unavailable; fugitive emissions are modeled via OPGEE’s component-level estimates but could be improved with more measurements.
• Transport coverage: Barges, trucking, and rail are excluded due to limited data and relatively minor global shares; heuristic edges approximate some missing pipeline connections.
• Static analysis: The LCA is annualized and static; temporal variability in operations, inventories, and shipping patterns is not captured.
• Spatial representation: Fields are represented as point locations rather than spatial extents; intra-field pipelines are not explicitly modeled.
• Blend estimation uncertainty: Sensitivity to optimization weighting and blend property differentiation introduces uncertainty; some countries with many similar blends (e.g., Iran) show higher uncertainty.
• Reporting differences: Results may not align directly with corporate ESG disclosures due to differing system boundaries and assumptions.