France's Missing July: Reading the EU Grain Cut

On July 30 USDA's EU attaché cut MY2026/27 grain production to 266 MMT, with corn down to 52.5 MMT and wheat to 133.8 MMT, naming heatwave losses in France, Hungary, Poland and Germany. We added France as a monitored grain region this week, which means we can check that claim against our own numbers rather than repeat it. The check passes, and more emphatically than we expected: July 2026 was the hottest, the driest and the most water-stressed July the French grain region has recorded in the sixteen years we hold — a clean sweep, on all three measures at once. Two things that are not in that finding also matter: the wildfires dominating French coverage are mostly in the wrong part of the country to touch cereals, and the Black Sea half of the same price story is about ports, not crops.

The finding A clean sweep, on a short record

The French grain region ran a mean daily maximum of 28.77°C through July against a 22.60°C normal, took 10.1mm of rain against a 60.1mm normal — 17%, and posted a crop-water-stress z-score of +1.34. Each of those is the most extreme value in our sixteen-year series (2011–2026), and it is not close on any of them: the next-driest July managed 26% of normal (2022), the next-hottest was +3.13°C (2018), and the highest prior stress reading was +1.14 (2020). The final dekad of July recorded 0.0mm.

That corroborates the attaché's direction independently — we are not reading their report back to you, we are computing the same physical conditions from reanalysis data and arriving at the same place. It is worth being precise about which crop that damages, because the two are on different calendars.

Which crop Wheat was cooked in June; corn is being cooked now

French soft wheat fills in May–June and harvests in July — the damage window for échaudage, or heat shock during grain fill, is already closed. Over that window 2026 ran +3.85°C, the hottest May–June in our record, with a peak daily maximum of 37.1°C. But rainfall over the same two months was 83% of normal — dry, not extraordinary, and only the fourth-driest fill window in sixteen years. So the wheat story is a heat story, not a drought story, which lines up with the attaché's specific claim of a sub-7.0 MT/ha national yield against 7.4 last year, and with a harvest running roughly three weeks early.

Corn is the opposite, and it is the crop still exposed. French corn silks and fills through July and August — precisely the window that just recorded 17% of normal rainfall and a +5.03°C heat anomaly. The attaché's line that the official French assessment rates more than 30% of the corn crop poor-to-very-poor as of mid-July sits directly on top of our most extreme reading in the series. Of the two cuts, the corn number is the one our data speaks to most directly, and the one with time left on the clock to get worse.

The fires Real, historic, and mostly in the wrong latitude

France is having an exceptional fire season — roughly 115,000 hectares burned, reported as the country's largest single wildfire since 1949, with area burned running several times the twenty-year average. It is a genuine disaster and it is understandably leading coverage. It is also, for the most part, not happening where the grain is.

The major fires have concentrated in the Mediterranean south and the southwest — Aude and Hérault, the Pyrénées-Orientales, Gard, the Landes, Gironde — between roughly 42.7°N and 44.8°N. Our French grain region spans 46°N to 50°N: the Beauce, Centre-Val de Loire, Poitou, Île-de-France and the northern basin, which is where the soft wheat, barley and corn actually are. That is a gap of one to three degrees of latitude, and the landscapes that burned — garrigue, maritime pine, vineyard country — are not cereal ground.

We should flag the exception rather than smooth it over: the Fontainebleau fire, southeast of Paris at roughly 48.4°N, falls squarely inside our region box. It is woodland rather than cropland, and one forest fire is not a production event — but it is inside the boundary, and a claim of "no fires in the grain belt" would be wrong. The honest formulation is that the fires and the crop losses share a cause rather than one causing the other. The same heat and the same missing rain produced both. Treating burned hectares as a proxy for lost grain would double-count a single drought.

The Black Sea A logistics shock wearing a weather costume

The same attaché report names Russia's Kerch Strait blockade and damage at Ukraine's Chornomorsk port, alongside the EU crop cut, as joint drivers of a mid-July surge in European cereal prices. Those disruptions are real and well documented: the Kerch Strait — through which roughly a quarter of Russian grain exports normally move — was closed to shipping in mid-July, Chornomorsk terminal operations were suspended after infrastructure damage, and major container lines suspended calls.

But our weather series says something useful about what that is not. Neither Black Sea wheat region had a bad July. Ukraine posted its coolest July in the sixteen-year record at −1.07°C with rainfall at 93% of normal; Russia ran near-normal on temperature (+0.33°C, with ten of fifteen prior Julys hotter) and a low water-stress reading of −1.06, and although its July rain came in at 58% of normal, July sits after the May–June fill window that governs Russian wheat yield. Both regions closed the month with an excess-wet final dekad.

So several different kinds of risk got priced through the same tape in mid-July, on very different clocks, and it is worth separating them rather than calling the whole thing a supply shock. The French crop loss is settled. The heat has already happened; whatever it cost the wheat is fixed, and the corn number can still get worse but only from here forward.

The Black Sea is not one variable, and this is where a glib reading goes wrong. Part of it genuinely is posture — strait closures, insurance rates, vessel routing — the kind of thing that can change on a decision and has repeatedly through this war. But a second part is destroyed capacity, and that does not reverse on any horizon a trader is positioned for. Ukraine's strike on Taman on July 30 hit a grain export terminal rated at roughly 5 million tonnes of annual capacity along with a sunflower oil facility at the same port; Russian strikes have taken out terminal infrastructure and silos on the Ukrainian side, including at Chornomorsk. Concrete, gantries and silo capacity are not a policy setting. Rebuilding them is a multi-year capital commitment that nobody makes at pace while the facility remains a target, which is the reason export capacity in this basin has been ratcheting down for years rather than oscillating.

That distinction cuts against the comfortable version in both directions. A trader treating the Black Sea premium as a headline that will fade is underrating the part that is physically gone. A trader treating it as permanent is overrating the part that is only a closure. And the week after the surge is the useful check: European and US wheat gave back 5% with every element of the disruption still in place — the strait shut, Taman burning, Chornomorsk still offline. That tells you what the market had actually been paying for, and it was mostly the weather.

The other side Why we would not put a yield number on this

Sixteen years is a short record, and "unprecedented" is doing light work. Our French series begins in 2011. A once-in-sixteen-years July is a real signal, but it is not a once-in-a-century claim and we will not dress it as one. France has hot, dry summers in its longer history that our window simply does not contain.

More importantly, our own reference material argues against confident extrapolation. The two worst French crops of recent decades — 2016 and 2024 — were excess rain years, and both rank as the wettest May–June periods in our record. The documented French failure mode is waterlogging: a saturated autumn that blocks sowing, or a wet fill that invites disease. We have no in-sample analog for a hot, dry French crop failure, which means we can tell you this July is unlike any other in our data and simultaneously that we cannot convert it into a tonnage estimate with any honesty. Novelty and quantifiability are different things, and conflating them is how weather analysis goes wrong.

And a regional box average is not a yield model. Our region is a rectangle drawn around the French cereal basin; it does not weight by planted area, does not distinguish irrigated corn from rainfed, and does not know that the harvest ran three weeks early — which shortens the exposure window for wheat and complicates any like-for-like comparison against a normal calendar. The attaché has field observations and the official French condition assessment. We have physics. Those agree here, which is the reason to take it seriously, but our contribution is corroboration and mechanism, not a better number.