How the Black Sea ports report is made

The method, the main tables and the sources behind the 3D view, taken from the public report (15 October 2026). Read the full public report (PDF).

How it is made · Who pays · Port damage · If the war ended today · Sources

How this report is made

The questions. The report answers five questions. What does the change in shipping through the Kerch Strait mean for trade, fuel and food supply? When did ship traffic stop? What is the likely global impact? What is the level of damage at Ukrainian ports on the Black Sea and the Danube? How long would it take for shipping to resume in full if the conflict ended today? The answers rest on dated public sources and our own satellite analysis, completed on 15 October 2026.

Events. Every strike and closure is dated and sourced, and each record states whether it is independently reported, confirmed by an operator or government, or a claim by one side of the war. Russian Ministry of Defence statements and Ukrainian claims of ships hit are labelled as claims. Single anonymous reports are labelled as such and never alone make a port "Severe" or "Major".

Ship traffic. Kerch Strait transits and Azov port calls come from IMF PortWatch, which counts ships from their automatic identification signals. Ships sailing with these signals switched off are not counted, so the counts are a floor.

Damage grades. Each Ukrainian port is graded from dated reports, operator statements and our satellite evidence, on the criteria in Section 4. A grade describes the damage to grain and port infrastructure, not to the town. Operating status is recorded separately: a port can be open but empty of ships, or damaged but working.

Satellite analysis. We screened 349 Copernicus Sentinel-1 radar scenes from April–October 2026, cut into 612 port images (two to four orbits per port), and used the 564 port images that passed quality checks; interference removed half of Taman's. We also used Sentinel-2 optical scenes, for nine strikes at six ports. For radar, we compared the backscatter after each strike with a stack of up to six earlier scenes, masked water, and kept only change that persisted and appeared on at least two independent orbits. To tell a strike from normal port activity, we ran the same test on 11 to 30 control windows with no strike in our calendar at the same port (dates three days apart, overlapping). The number of those windows with at least as much change, counted against the strike itself, measures how unusual the strike is. Across eight radar-tested strikes and four measures each (whole port, stable ground, mapped facilities, largest single change), chance alone would put one or two results above every control window. The control windows also overlap, so they count for less than their number suggests. For optical images, we compared the clearest scene before and after each strike, with a burn index and a test for active fire. Radar change at 10–20 metres shows that the surface changed; it cannot name the object or grade the damage. NASA Black Marble night lights measure, on the darkest of the seven nights after a strike, the share of lit area within 10 km of the port that lost at least half its light, compared with one or two windows without strikes. We found no openly available very-high-resolution image of these ports taken after the 2026 strikes (Esri Wayback, Maxar/Vantor Open Data, Copernicus EMS; the newest openly available captures date from 2023 to March 2026). We did not buy commercial imagery; the only published high-resolution images of the strikes we found are Vantor's of Novorossiysk (August) and a satellite image of the EFKO terminal at Taman (provider not named), published by UkrAgroConsult on 8 August, which shows damaged tanks and burn scars; we did not analyse either.

Supply, prices and people. A scenario model runs from October 2026 to September 2027 in quarters.

• Low case: Greater Odesa partly restarts and Novorossiysk returns toward normal volumes from early 2027.
• Central case: current constraints persist, with the Danube, rail, road, Baltic and Caspian routes at their measured capacity.
• High case: further escalation hits the Danube route and holds Russian exports down longer.

Supply comes from port capacities, measured rerouting and official export data. Price effects separate the world benchmark price from the extra cost of landing Black Sea grain (freight, war-risk insurance, longer routes). They are anchored on observed prices around each strike wave and checked against the 2022 and 2023 episodes. Import bills apply these effects to each country's wheat imports and its Black Sea share, at normal import volumes and origin shares (an exposure measure). Extra people in acute food insecurity apply a price-to-IPC response (from WFP's 2022 analysis, halved at the low end to remove the fuel channel) to the latest IPC caseload, weighted by how much of the staple basket is imported wheat. That weighting is our judgement. IPC baselines are old for Yemen (2025) and Bangladesh (2025, partial coverage), and Kenya's covers arid counties only.

Recovery timeline. The phases and their ranges are our estimates, built from the 2022 Black Sea Grain Initiative and the 2023 Ukrainian corridor, mine clearance after the 1991 Gulf war, Haiphong in 1973 and Suez in 1974–75, the 2023 insurance facility, Kernel's 2023–25 terminal repair, and the Red Sea after the 2025 Gaza ceasefires. The time insurers take to remove a war-risk listing is not documented and is our judgement.

Confidence.

• High: the date and fact of the Kerch closure, the collapse of transits, and the halt of foreign ships at Greater Odesa.
• Medium: the damage grades and the size of the grain shortfall.
• Low: the price attribution between the Black Sea and North American and European drought, the extra people in food insecurity, and the recovery months, which depend on terms nobody has yet agreed.

Who pays: extra import bills and hunger, by country

Who pays. The cost falls mainly on the largest wheat importers and on those with little room to absorb it. Among these ten countries, all heavily dependent on the Black Sea, differences in that dependence add only a little; the size of the import bill and the lack of a buffer matter more.

The bill holds each country's imports at their normal volume and origin mix, so it measures exposure to the price change, not spending already incurred. Origin shares are UN Comtrade figures for Egypt and Turkey (2025) and for Lebanon, Tunisia and Yemen (2024); Sudan's is Fastmarkets' share of 2026 import value; Bangladesh's comes from a trade aggregator we could not verify. Kenya's (midpoint of 82% in 2024 and 39% in 2025 in Comtrade), Somalia's and Libya's are our estimates. Yemen's import volume is our estimate (three-quarters of FAO's 5.2 million tonne cereal import requirement), and Somalia's import volume dates from 2020 (UNCTAD). Turkey and Egypt are buying less for now, so their figures are upper bounds. Turkey has a record 23 million tonne crop, has released 5.5 million tonnes of state stocks, and Russian arrivals are down about 90% (USDA). Egypt's July–August wheat arrivals are reported down about three-quarters year on year (Al Manassa, not independently confirmed). Much of Turkey's wheat is milled for re-export, so part of its cost passes to flour importers.

Over October 2026 to September 2027, the extra import bill for these ten countries is about US$1.5 billion in the central case (0.5–3.4 billion). For wheat bought from the Black Sea our central case adds US$40 a tonne to the landed cost now, easing to US$30 by mid-2027 (about US$36 on average over the year). That is above the reported rise of US$16–37 in Egyptian prices to early September (US$16 from S&P Global; the upper end from a single Al Manassa report). S&P's general price for wheat delivered to Egypt, origin not stated, held at about US$313 on 8 October, below the US$320 reported for Russian wheat on 3 September. That rise also includes some freight increase caused by the worldwide fuel shock, which we could not separate out, so this part of the bill is an upper estimate of the Black Sea's share. About four-fifths of the bill is wheat. It would be about US$0.8 billion more if El Niño cuts harvests elsewhere in 2027. Egypt pays most in money, but its fixed-price bread subsidy shields consumers and costs the treasury instead. Yemen, Sudan and Somalia pay most in hunger, because they buy at market prices with no buffer.

In the central case, 1.1–2.4 million more people fall into acute food insecurity (Phase 3 or worse on the Integrated Food Security Phase Classification, IPC) across the six countries with IPC data. About half are in Yemen (0.56–1.2 million) and a quarter in Sudan (0.29–0.63 million). Somalia and Bangladesh add 0.1–0.25 million each, with smaller numbers in Lebanon and Kenya. Yemen's IPC baseline dates from 2025. This is not a global figure. WFP modelling cited by the Council on Foreign Relations (24 September 2026) puts the number pushed into acute food insecurity worldwide above 15 million. We could not find WFP's own publication or its inputs. CFR attributes it to the price rise from the Black Sea closure alone, before drought and other pressures on prices are counted. It is global; ours covers six countries.

The level of damage at Ukraine's ports

The damage is serious but concentrated in particular terminals. Physical damage has shut Kernel's terminal at Chornomorsk, about 0.8 million tonnes a month of capacity. That is a sixth to a fifth of the 4.1–5.0 million tonnes a month that Greater Odesa handled in the first half of 2026 (Kyiv Post; cargo type not stated). Industry estimates in mid-July put the total loss higher, at about a third. Ukraine's main farmers' union gave that figure; an investment analyst put deep-water capacity down from up to 7 to 4–5 million tonnes a month. Part of the gap may be that 4 of 13 large terminals had stopped buying grain, according to industry sources quoted by Reuters; no reason was given, and stopping purchases does not by itself mean a terminal was damaged. Most of the rest is intact but idle, because foreign ships and insurers have stayed away since 20–22 July. The Odesa terminal of ADM, a US grain trader, hit on 31 August, has not reported its condition. Ukraine's port authority counts 160 strikes on ports and 155 attacks on ships from January to mid-August 2026, and 1,002 port facilities damaged since 2022. We grade each port on dated sources, operator statements and our own satellite analysis.

Grades cover physical damage from 2022 to 2026, with most weight on 2026, and are kept separate from whether ships call.

\* Months until the port handles its June 2026 volume (repairs and returning ships). For Mykolaiv, Olvia and Kherson, closed to commercial ships since 2022, months to reopen; see Section 5. Capacity: the Greater Odesa figures split the Kyiv Post's 4.1–5.0 million tonnes a month (first half of 2026; cargo type not stated), split by each port's share of all cargo in early 2025; Izmail and Reni are all-cargo throughput in 2024. Mykolaiv and Olvia: all-cargo throughput in 2021 (Ukrainian Sea Ports Authority), before closure. Mt = million tonnes.

What happened at each port in 2026:

• Chornomorsk: The 10-million-tonne-a-year grain terminal of Kernel, Ukraine's largest grain exporter, hit repeatedly since May; suspended since 13 July
• Kherson: On the front line; near-total destruction reported in 2025
• Odesa: Repeated attacks from 15 July; five on Odesa-region port facilities on 1–3 August (Ministry); port hit 9 August and 5–6 October; ADM grain terminal hit 31 August (operator assessing; no suspension reported). Ships hit at or off the port include the GOLDEN LEO (struck 19 July after leaving Chornomorsk with maize, 10 killed; sank off Odesa on 26 July) and a ship entering port on 3 October. Russia claimed strikes on two Odesa-bound dry-cargo ships at sea on 15 October (not verified).
• Mykolaiv: Nika-Tera terminal destroyed earlier in the war; three foreign-flagged ships hit and two people killed aboard one of them on 17 July; officials report the berths otherwise largely intact. Idle for security, not damage. Reopening needs a secure channel past the Kinburn Spit, mine clearance and dredging.
• Pivdennyi: Rail and production facilities hit 28 January; a warehouse at the TIS terminal reported hit 14 August (one anonymous source; TIS declined comment)
• Izmail: At least 7 dated attacks in 2026 on the port, ships there or its approaches, from April, including 13 August. Russia's Defence Ministry also claimed strikes on 7 September and 15 October (not verified). Satellite: one radar measure, inside the mapped port and industrial outlines after the attacks of 13 and about 17 August, is unusual but could be chance; after the 7 September attack claimed by Russia the radar shows nothing above normal port activity.
• Kiliia: Airstrike on 27 September; port damaged, no casualties
• Orlivka (ferry crossing): Drones hit 1 and 6 September; resumed within a day after 1 September; no resumption confirmed after 6 September
• Reni: Major damage in 2023, since repaired. No 2026 strike confirmed; Russia claimed strikes on naval sites in Reni and Izmail on 13 August (not verified).
• Olvia: No 2026 strike on Olvia found by name; closed to commercial ships since February 2022 because its access channel is not demined. Reopening depends on the same secure channel as Mykolaiv.

Danube. The Danube ports are damaged but working. Danube traffic peaked in 2023, at about 3 million tonnes a month (OSW, 28 July 2026), before this year's low water cut barge loads by up to 60%. In 2024 the three Danube ports (Izmail, Reni and Ust-Dunaisk) handled about 1.4 million tonnes of all cargo a month (17.3 million in the year, GMK Center). In 2025 that halved, to 0.7 million a month, as cargo returned to Greater Odesa. In September 2026 they shipped about 1.1 million tonnes of farm goods, up from 0.7 million in August (Ministry of Agrarian Policy, via Reuters). The minister's wider count for September is 1.3 million. Damage is not what limits them. In July, record-low water cut barge loads by 20–40% (Reuters) and by up to 60% (OSW, 28 July). Satellite water maps show surface water within 10 km of Izmail up by about 2–10 km² on 12 October against early July and early August (much of it flooded vegetation). That is not clearly a recovery of the river; river flow itself was not measured. About 70–80 ships queue at Sulina and Bystre. Ships leaving are also attacked: on 5 October the ROYAD MAMMADOV, carrying grain from Izmail to Ravenna, was struck off Romania and reportedly sank. On 6 and 13 October drones hit four more merchant ships in or near Bulgarian waters and sank two, ALFA WATAN and HS DIGNITY; 21 crew are missing. Bulgaria says it has no clear evidence of who attacked, and the ships' cargoes were not reported. On 15 October two tankers anchored at the entrance to the Sulina canal caught fire, with eight crew missing; Romania has not confirmed the cause. After a ceasefire, Danube volumes would likely fall back as cargo returns to Greater Odesa.

What the satellites show. We analysed 349 Sentinel-1 radar scenes, cut into 612 port images (564 passed quality checks) and before-and-after optical images (Sentinel-2) for nine strikes at six ports between July and October 2026.

• In Ukraine: Chornomorsk, Odesa, Pivdennyi and Izmail (the only Danube port analysed).
• In Russia: Novorossiysk and Taman.
• Not analysed by radar or optical images: earlier 2026 strikes, Mykolaiv, Olvia, Kherson and the other Danube ports (Reni, Kiliia, Ust-Dunaisk, Orlivka). The ADM grain terminal strike at Odesa (31 August), the Kiliia strike of 27 September and the Orlivka strikes of 1 and 6 September were not tested.

A working port changes all the time. For each strike we therefore compared the change in the port with the change in 11 to 30 control windows with no strike in our calendar (dates three days apart, overlapping). Measured over the whole port area, no strike stood out from the control windows. Two stood out on narrower measures. Stable ground is ground that rarely changes when there is no strike; it leaves out berths, stockpiles, rail cars and water. At Chornomorsk in July, change on normally stable ground was larger than in all 15 control windows. At Izmail in August, change inside the mapped port and industrial outlines was larger than in all 21. The results below are the strongest single measures. They support the reported damage but do not prove it. Across eight radar-tested strikes and four measures each (whole port, stable ground, mapped facilities, largest single change), chance alone would put one or two results above every control window. Two such results are therefore weak evidence. The control windows also overlap, so they count for less than their number suggests.

• Izmail, 13 August: inside the mapped port and industrial outlines, 5.9 hectares changed against 2.5 typically, more than in any of 21 control windows measured the same way. About 1 hectare of that is outside the port, mostly at a solar power plant some 4 km north-east of the berths. Counting the port alone (its three districts, the Nibulon terminal and the piers), 4.7 hectares changed against 1.4 typically, and no control window reached it (the largest was 3.6). On all normally stable ground 16 hectares changed (2 of 21 windows as much). About two-thirds of it (10 of 16 hectares) lies in fields on the opposite, Romanian bank, 200–500 metres from the river. Those fields lost greenness after mid-August, which fits harvest or field work, not the port. Across the whole port area the change is within the normal range. Twenty of the 21 comparison windows date from June and July. The river fell through July to its lowest level since 1996 (about 20 July), so only the last few of them cover low water. The only August window began eight days after this strike and four days after another reported attack (about 17 August). The after-strike radar images (14–25 August) also cover that attack, so the change cannot be tied to 13 August alone. A falling river can also change the radar signal along the quays, so this result is weaker than the count suggests.
• Chornomorsk, July: radar change on normally stable ground was larger than in any of 15 control windows. The optical images flag about 3 hectares of burned or darkened ground inside the mapped port and industrial areas (4.2 in all). Some may be ships that moved or cloud-shadow edges, and the before and after images come from different Sentinel-2 tiles. The dark patch near the port in the image of 17 July is the shadow of a cloud, not smoke. The satellite's own cloud mask flags it, and it stays dark in the infrared, where smoke is nearly transparent. No active fire was detected at Chornomorsk. The radar after-images (14–23 July) also cover further strikes on 14, 15 and 22 July; the optical pair (7 and 17 July) covers those of 14 and 15 July. The signal is weak but consistent with the reported damage at the port, including Kernel's grain and oil terminals (loading equipment, silos, tanks and power lines). We did not have the terminals' exact footprints, so we cannot tie the change to them.
• Taman (Russia), strike of 30 July: the next day Sentinel-2 caught active fire over about 5 hectares of the coastal terminal strip, with smoke drifting over the sea. Our burn index is unreliable on this coastal strip, because the before and after images come from different tiles and disagree on where the water is, so we do not report a burned area. The larger fire (3.4 hectares) lies inside the mapped outline of an oilseed-processing plant, and a smaller one (1.7 hectares) about 175 metres west of it. This fits the reported hit on the EFKO vegetable-oil terminal (Section 2), but we cannot confirm the object.
• Novorossiysk (Russia), 12 August: a 1.6-hectare change, seen from four different satellite passes, beside the rail yard of the grain port. It lies about 600 metres from the grain elevator, where NKHP's loading gallery was reported destroyed; we cannot tell whether the change is the gallery. Two of 27 control windows had a cluster as large. Both of these, and the three windows with as much change on stable ground, began 7–19 days after the 10 September strike on Novorossiysk, so they may include damage, clean-up or repairs. None of the 22 windows before the 10 September strike was as large on either measure. A change of similar size appeared at the container terminal, where no grain-terminal strike was reported (the oil terminal and naval base nearby were hit in the same days), so this result is suggestive only. Commercial imagery published by NBC News (Vantor, 12 August) appears to show damaged port infrastructure at Novorossiysk; an industry source quoted by Reuters says NKHP's conveyor gallery was completely destroyed. We did not analyse the imagery.

At Odesa and Pivdennyi the satellites see nothing above normal port activity. Damage was reported at both (confirmed by officials at Odesa; from one anonymous source at Pivdennyi in August), but it is likely too small or too scattered for 10-metre images to separate from cargo and ship movements. Night lights dimmed most, in share, around Pivdennyi. On 20 August, six days after the strike, 57% of a small lit area within 10 km went dark (about 9 of 16 km²), against 27–33% in two earlier windows. At Izmail, which has a similarly small lit area, windows without strikes reached 47–50%. The Pivdennyi drop may be a grid outage in Yuzhne, the town at Pivdennyi port, rather than damage at the port. Mykolaiv's lit area is similarly small. It also dimmed after a Russian claim, not confirmed, that a tugboat was hit there on 1 August: on 6 August, five days later, 51% of its lit area went dark, against 17% in June. Lights fell across Odesa on the same night (38%, against 22% in June). Both are weak evidence. We found no post-strike very-high-resolution image in the open archives we checked (Esri Wayback, Maxar/Vantor Open Data, Copernicus EMS). We did not buy commercial imagery. We found only two published high-resolution images of the strikes. One is Vantor's image of Novorossiysk (August). The other is a satellite image of the EFKO terminal at Taman (provider not named), published by UkrAgroConsult on 8 August; it shows damaged tanks and burn scars. We did not analyse either.

What the satellites cannot see. Openly available 10–20 m images show where the ground surface changed. They cannot see:

• cranes, conveyors, ship-loaders or silo interiors;
• damage under intact roofs;
• anything underwater: mines, wrecks, damaged berths or channels.

A port with no signal may still be damaged. Grading terminal damage needs commercial very-high-resolution imagery or inspection on site. Commercial very-high-resolution images would settle the grades at three Ukrainian sites: Kernel's and ADM's terminals in Greater Odesa, and the Izmail facility clusters that changed after the attacks of 13 and about 17 August. They would also show what was damaged at two Russian sites: NKHP, Novorossiysk's main grain terminal, where the only published images show unspecified port damage, and the Taman terminal. The table below gives the location of the main changes we measured, so they can be checked against commercial imagery.

What it would cost. No assessment of 2026 port damage has been published. The World Bank, EU and UN damage assessment of February 2026 puts Ukraine's transport needs above US$96 billion, but does not report ports separately. The Kyiv School of Economics put direct port damage at US$0.85 billion by January 2024. In May 2025 the deputy minister responsible for ports put the cost of restoring port infrastructure at about EUR 1 billion, with nearly 400 port facilities damaged. Strikes intensified in 2026. Ukraine's port authority counts 160 strikes on ports from January to mid-August alone. The count of port facilities damaged since 2022 rose from 935 on 4 June to 1,002 by mid-July. On that basis we estimate cumulative port damage at US$1.5–5 billion, with low confidence.

If the conflict ended today

Export volumes would recover in months and the first ships would move within weeks; the full pre-2022 network outside the occupied ports would take about three years, set by rebuilding Kherson. Operators and officials report damage concentrated in particular terminals, chiefly Kernel's at Chornomorsk, with no port-wide destruction in Greater Odesa or on the Danube. Our satellite analysis is consistent with this but cannot grade damage. What holds back volumes is insurance and the confidence of shipowners. Mines and the security of the estuary channels hold back the ports closed since 2022 and the full network. Mykolaiv and Olvia, almost a quarter of Ukraine's seaport cargo in 2021 (Ukrainian Sea Ports Authority), could reopen in about a year. Beyond that, the insurers' war-risk list raises costs rather than limiting volumes: it adds notification and a premium but does not bar trade.

These are our estimates for a ceasefire on 15 October 2026, built from comparable episodes:

• The 2022 grain deal. The first ship sailed ten days after signing, and volumes peaked in the third month at 4.2 million tonnes.
• Ukraine's own corridor in 2023. Ukraine exported 7.3 million tonnes of cargo by sea in December 2023, the corridor's fifth month; the corridor carried about 15 million tonnes in its first five months, under fire.
• Mine clearance after past wars. Kuwait and the Gulf in 1991, Haiphong in 1973 and Suez in 1974–75 each took months to years to clear in full.
• The Red Sea in 2025. When the Gaza ceasefire was agreed in January 2025, shipping executives said owners would wait for proof of security before returning to the Suez route. Houthi attacks resumed, and in July 2025 they sank two bulk carriers, MAGIC SEAS and ETERNITY C. A second Gaza ceasefire took effect on 10 October 2025. Even then, Suez Canal transits stayed at about 39–44 a day in every month from October 2025 to early October 2026, against about 74 in 2023 (IMF PortWatch). From February 2026 the Iran war also kept ships away, so the later months overstate the Red Sea's own effect. In the Red Sea, the attacker was also an armed group that kept its ability to strike. A verified Black Sea ceasefire between states would remove more of the risk, so our high case of 12 months is our judgement.

Three facts set the critical path:

1. Mines. Drifting mines have been found in the Black Sea since 2022. Large-scale clearance can only start after a ceasefire. Under the 1936 Montreux Convention, Turkey bars warships of countries at war from the Straits, which has kept donated minehunters out. Ukraine's navy estimated in January 2024 that clearing the trade routes would take 3–5 months and the whole sea 3–5 years. In November 2024 the Navy spokesman put a first phase, clearing berths and ports, at 3–7 months and the main clearance at 3–5 years; he noted that Kherson and Mykolaiv ports remain blocked, partly by mines and Russian control of the Kinburn Spit. These estimates come from one side of the war, and we found none after 2024. Turkey has said the bar holds as long as the war continues; whether and when a ceasefire lifts it is Turkey's call. We assume weeks, not months (our judgement). The Turkish, Romanian and Bulgarian mine-countermeasure group already works in its own waters. 2. Insurance. In 2023, about three months after the corridor opened, a government-backed scheme set war-risk cover at 1.25% of a ship's value, against early market forecasts of up to 5%. Insurers taking the Black Sea off their list of war-risk areas entirely would take longer; we put it at 12–48 months, and this is our judgement. 3. Kerch and Azov. Reopening the strait is a Russian administrative decision that could take days. Azov traffic could return in about a month, once strikes on Azov shipping stop and damaged ships and wrecks are checked (days to four months; our judgement). This applies to sea-going ships. River-sea traffic on the Don and the Volga–Don canal normally stops for winter, from about December to April, so full Azov and Don volumes would return only from spring 2027 (our judgement). Kerch transits already fall in winter: about 11 a day in December 2025, against about 20 in September–November (IMF PortWatch). Most Russian Azov and Black Sea terminal capacity can restart quickly. NKHP, Novorossiysk's main grain terminal, was expected to need up to six months of repairs after the August strike: the company said up to four months, Reuters reported up to six, and a source said staff had been sent home for six months. If repairs have continued, it could be ready at any time between now and mid-February 2027. Taman's damaged grain terminal still could not load in late September, and no repair estimate has been published. A ceasefire would not by itself change the sanctions or insurance terms on Russian trade. Ukraine cannot use the occupied ports of Mariupol and Berdiansk without a territorial settlement.

What a ceasefire today would mean for importers. Our low case, a partial restart from early 2027, is slower than the timeline above, so it gives a cautious upper bound for lost exports after a ceasefire today. Lost Russian and Ukrainian grain exports fall from about 37 to about 16 million tonnes. The extra import bill for the ten countries falls from about US$1.5 billion to about US$0.5 billion. The added people in acute food insecurity in the six countries with IPC data fall from 1.1–2.4 million to 0.43–0.94 million. The low case already assumes lower freight and war-risk costs this quarter. If October–December cargoes are priced at today's costs, the bill is about US$0.7 billion and the added people about 0.54–1.2 million (our model, low confidence).

Where the figures come from

• Kerch Strait: Reuters (10–13 July 2026, citing three grain-industry sources) for the closure; Gorod N (14 July 2026, citing a market participant and a Transport Ministry statement) on the Azov–Don canal; Institute for the Study of War and Critical Threats (11 and 13 July 2026) for claimed ship strikes; IMF PortWatch daily ship transits and port calls, read 15 October 2026; Rostov regional statements (28 August, 9 October 2026); ProZerno, SovEcon, BIO-TON and Kpler for the Azov share of Russian grain exports.
• Russian exports and rerouting: Reuters (13 August, 25 September 2026); Rosselkhozbank (28 September 2026) and the Russian Grain Union for the share of lost volume other routes replace; Delo via UkrAgroConsult (2 October 2026) on unsold grain; IKAR and Interfax (8–9 October 2026) for Baltic shipments; Baltic states' request to the EU (Marine Insight, 12 October 2026); UkrAgroConsult (15 October 2026) for Latvia's position; Ukrainska Pravda (1 October 2026) for Estonia's transit ban; USDA WASDE (9 October 2026), IGC (17 September 2026), SovEcon (30 September 2026), S&P Global (13 October 2026, via Milling MEA) and IKAR (8 September 2026) forecasts; The Moscow Times and SovEcon (17 August 2026), CSIS (11 September 2026) and CFR (24 September 2026) on capacity offline; USDA WASDE (September 2026) and IKAR (14 May 2026) for the baseline forecasts; ProZerno (8 October 2026) for September port shares.
• Fuel: S&P Global, CREA and the Kyiv School of Economics (Kpler data) for Black Sea crude and product exports; IEA Oil Market Report (14 October 2026); Russian government export bans; Crimean authorities' statements; Krasnodar Krai Operational HQ (21 June 2026, via ISW); Kyiv Post (3 September 2026) on Crimean rationing; CSIS (11 September 2026) on fertiliser routes and the Ust-Luga strike; OSW (14 September 2026) on vegetable oil; UkrAgroConsult (8 August 2026) on the EFKO terminal; A-95 Consulting and Enkorr (Reuters, 12 October 2026) on Ukraine's diesel imports; Interfax-Ukraine (A-95 Consulting, July 2026 data) on the shift to the Polish route; AMIS Market Monitor No. 142 (2 October 2026) on fuel and freight; S&P Global (7 September 2026); KSE Russian Oil Tracker (September 2026); World Bank Commodity Markets Outlook (April 2026) and Food and Nutrition Security Update (September 2026), and WTO on fertiliser.
• Ukraine's ports: Ukrainian Sea Ports Authority attack and ship-call counts; Ministry of Agrarian Policy export figures (2 October 2026); Kernel (13 July 2026); Reuters, Kyiv Independent, Ukrainska Pravda and Latifundist strike reports; World Bank, EU and UN Rapid Damage and Needs Assessment (RDNA5, 23 February 2026); Kyiv School of Economics damage estimates (January 2024).
• Satellites: Copernicus Sentinel-1 (radar, 349 scenes, April–October 2026, via Microsoft Planetary Computer) and Sentinel-2 (optical, via Element 84 Earth Search), analysed by us; NASA Black Marble night lights; Esri World Imagery Wayback, Maxar/Vantor Open Data and Copernicus EMS (no post-strike very-high-resolution capture found); Vantor imagery of Novorossiysk as published by NBC News (not analysed by us).
• Markets and importers: FAO Food Price Index (2 October 2026); AMIS Market Monitor (2 October 2026); FAO cereal trade forecast (October 2026); USDA trade tables (October 2026); AMIS Market Monitor; IGC; Platts and Fastmarkets price reports; the Joint War Committee listed areas (16 September 2026); UN Comtrade, Fastmarkets and our estimates for import origins (see Section 3); IPC analyses for Yemen, Sudan, Somalia, Kenya, Lebanon and Bangladesh; WFP Executive Board procurement report (2026); CFR (24 September 2026) for the WFP global estimate.
• Recovery: 2022 Black Sea Grain Initiative and 2023 Ukrainian corridor volumes; Ukrainian Navy mine-clearance estimate (January 2024); analyses of Kuwait 1991, Haiphong 1973 and Suez 1974–75; the 2023 war-risk facility; Turkey's 2024 Straits decision; Kernel's 2023–25 repair; the Red Sea in 2025.

Taken from the public report; figures are our estimates unless a source is named.