171 documented strikes on 34 Russian refineries — up to 2,469 km from the front
Ukraine's General Staff claims 42.7% of Russia's refining capacity is out of action; independent estimates vary by definition, from roughly 20–36% effectively or strike-related offline to 58% when other outages are included. This dataset of 34 struck refineries maps strike depth against damage severity and tracks cumulative capacity loss against those estimates.
Refineries struck since March 2024
Documented strike events across tracked sites
Combined capacity of struck refineries
Strike depth vs. damage severity — weak correlation
Key findings
Since March 2024, Ukrainian long-range drones have struck at least 34 refineries and gas-processing plants across seven Russian macro-regions — from Novoshakhtinsk, 181 km from the front line, to Omsk, 2,469 km deep in Siberia.
The campaign's geography tells a story that distance alone doesn't explain. Reach has expanded steadily, but outcome depends more on strike persistence and facility vulnerability than on proximity to the front.
Distance is a weak predictor of damage. The correlation between strike depth and severity is only r = −0.03 — deep strikes on Omsk (2,469 km), Tyumen (2,050 km), and Permnefteorgsintez (1,526 km) still achieved "Severe" to full "Shutdown" outcomes on some of Russia's largest facilities.
Repeated strikes compound damage. Facilities hit 6+ times — Volgograd, Ryazan, Slavyansk-ECO — accumulated some of the highest cumulative capacity losses in the dataset, regardless of distance.
This dataset's cumulative modeled exposure across 23 refineries exceeds half of national nameplate capacity — an upper-bound model that does not subtract later repairs or restarts.
Central Russia and the Volga basin — the two regions with the most struck refineries — are also the two where Reuters and Euromaidan Press have reported the most acute local shortages.
of Russia's designed refining capacity, per Ukraine's General Staff — the highest official estimate to date.
"More than 20% offline" — independent analysts put functional disruption at closer to a third.
— International Energy Agency, cited in Al Jazeera, Jul 2026
Every struck refinery, mapped
Each colored marker is one struck refinery, sized by pre-strike capacity. Use the map toggle to show five operating refineries where no public strike was found by the research cutoff. These map-only facilities are excluded from every strike count and capacity-loss calculation.
Strike depth vs. damage severity
Each point is one refinery. The x-axis is the great-circle distance from the facility to the nearest point on the active contact line — a strike-depth proxy, not the actual drone flight path. Bubble size scales with pre-strike refining capacity. Hover any point for details.
Pearson correlation, distance vs. severity rank
Weak and not statistically meaningful given the sample size — proximity to the front line does not reliably predict how badly a facility is damaged.
Deepest confirmed strike — Omsk refinery, Siberia
Russia's single largest refinery (442,000 bpd) was struck for the first time in July 2026, more than 2,400 km from the nearest contact-line sector.
Targeted refineries, ranked by capacity
Pre-strike nameplate capacity for the 32 facilities with published figures (Oryol and Astrakhan GPP lack reliable numbers), colored by reported damage severity. Nine of the ten highest-capacity refineries on this list have been hit more than once.
Cumulative modeled exposure crossed the "critical" threshold by mid-2026
The dark line is cumulative modeled exposure across this dataset's 23 refineries with usable capacity-loss estimates, assuming losses persist at their last-reported fraction. It is not current capacity offline. Markers are dated external estimates.
Cumulative modeled exposure, as of Sep 2026
Runs above most third-party estimates because it assumes no repairs — a deliberate upper-bound read, not a claim of higher accuracy than independent reporting.
Repair and restart milestones
A source-linked record of publicly documented repair and restart events. Coverage is incomplete: no recovery record means unknown, not still offline.
| Source |
|---|
Regional fuel markets: cumulative exposure
Each panel tracks cumulative offline capacity as a share of this dataset's struck refineries within that macro-region — a proxy for how exposed each regional fuel market is to the tracked strikes, not the region's full refining base. The dotted line marks 50% of tracked regional capacity offline, roughly where outlets began reporting local rationing.
How this dataset was built
Scope
The dataset covers 34 Russian refineries and gas-processing plants reported struck by Ukrainian long-range drones between March 2024 and September 2026, compiled from reporting by the Caspian Policy Center, Ukraine's Center for Countering Disinformation, Reuters, the Moscow Times, Euromaidan Press, and other outlets cited per-row in the data table below.
Distance proxy
"Strike depth" is the great-circle (haversine) distance from each refinery's coordinates to the nearest of eight reference points along the active Ukraine–Russia contact line as of mid-2026 (Sumy border sector through the Kherson/Dnipro sector). This is a straight-line proxy — actual drone routing is typically longer and rarely public.
Severity & estimated loss
Severity is a four-level scale coded from reporting language: Light-Moderate (rank 1.5), Moderate (2), Severe (3), and Shutdown (4). Estimated barrels-per-day lost multiplies each refinery's nameplate capacity by a loss fraction inferred from the same reporting — a full shutdown scores 1.00; partial unit damage scores lower.
Evidence status
Each refinery record identifies its source type, confirmation status, confidence, and impact basis. Ukrainian official claims and Russian official statements are attributed as such. Reuters industry-source reports and other open-source reporting are separate categories. These labels describe provenance; they do not mean this project independently verified events on the ground.
National cumulative modeled exposure
Cumulative modeled exposure sums estimated bpd losses across the 23 refineries with usable estimates, ordered by each event's reported date and divided by ~6.5M bpd national nameplate capacity. It holds every recorded loss fraction indefinitely and therefore is an upper-bound exposure model, not an estimate of capacity currently offline. External markers retain their sources' definitions.
Recovery events
Recovery events are dated, source-linked milestones, not a complete operating-status history. Measurements preserve the source's unit and evidence basis; unit nameplate capacity, reported throughput, and observed throughput changes are not treated as interchangeable. Later strikes do not erase earlier restart records.
Why this runs higher than other estimates
This tracker is monotonic and does not subtract repairs. No independent national outage percentage newer than mid-July was publicly available by 13 August. The best-defined snapshots put strike-related capacity offline above 36% at end-June, S&P Global CERA's all-cause downtime near 58% in the week ending 10 July, and Kpler's effectively offline capacity at roughly 20–27% in mid-July. These are not interchangeable.
Limitations
- Cumulative modeled exposure does not subtract repairs or restarts and must not be read as capacity currently offline; incomplete reporting may also understate historical peak disruption.
- Two facilities (Oryol, Astrakhan GPP) have no reliable public bpd figure and are excluded from all capacity-weighted charts, though included in the strike-depth scatter and the full data table.
- Regional panels use each region's tracked-refinery capacity as the denominator, not the region's total refining base — a proxy for exposure among struck facilities, not a market-share figure.
- Severity coding is qualitative, derived from open-source reporting language rather than verified engineering assessments.
Update policy
The ledger is reviewed event-by-event when credible new reporting appears. Facilities must be inside Russia and physically process crude oil or hydrocarbon feedstock; depots, terminals, pumping stations, petrochemical-only sites, and facilities in occupied Ukrainian territory are excluded. Astrakhan GPP is a disclosed category exception. New records retain a direct source and evidence label; corrections are documented in the changelog. The last verified date appears in the hero, footer, dataset metadata, and versioned release.