The last time it was this dry

Summer 2018

"Unprecedented in 2,110 years"#

In 2021 a team led by Ulf Büntgen measured the stable carbon and oxygen isotopes in the wood of oaks from the Czech Republic and southern Germany, and concluded that the dry summers from 2015 to 2018 were unprecedented in the past 2,110 years.

The same year a team led by Monica Ionita, averaging a drought atlas made from tree rings over central Europe, found longer and more severe droughts in about 1400–1480 and 1770–1840.

The two answers come from different wood and areas of different size, so neither refutes the other. This page asks the question for every place on the map, with the summers since 2013 included.

Year 0 to 1902

Two thousand summers, read from the rings of trees#

In trees that are short of water in summer, the width of each year's ring can reflect how much water they had. Edward Cook and colleagues used tree-ring records from across Europe to reconstruct a summer drought index for every year from the year 0, on a grid of half a degree: the Old World Drought Atlas (2015). How many trees stand behind each estimate, and how good it is, change from place to place and century to century.

The index is the self-calibrating Palmer Drought Severity Index, a standard measure of how wet or dry the soil is against the local normal: zero is a normal summer, minus 3 a severe drought, minus 4 an extreme one. Browns on the map are summers with drier soil than usual, teals wetter.

The atlas was fitted to an instrument-based version of the index over 1928–1978 and checked on 1901–1927. After 1978 it holds instrument-based values, not tree rings, so this page uses the current series of the Climatic Research Unit (University of East Anglia) for every summer from 1903 to 2025.

2,026 summers

Every summer, stacked#

Here the map is cut along the latitude of Berlin. The cut face holds every summer at that latitude, one thin row a year: 2025 at the top, the year 0 at the bottom. Above the pale line the rows come from instruments; below it, from the trees.

Dry runs cross the face as brown bands, some a decade wide. The question is whether the band at the top is the deepest one.

Two readings

Why the trees cannot be read at face value#

Read at face value, the atlas makes the recent four-summer run (the average of four summers in a row) the driest on record in 15% of the land on this map, in yellow. That reading misleads, in both directions.

Over 1903–1978, where both exist, the atlas and the instrument record agree only in part: a median correlation of 0.52, where 1 would be perfect agreement and 0 none. The atlas picks the instrument record's driest summer in only 21% of places. In some regions its swings are wider than the instrument record's, so its past droughts look deeper; in others narrower, so they look milder.

Take Germany. Read at face value, the atlas makes the recent run the driest in 0% of the country; with its errors allowed for, in 54% over the whole record. Iberia goes the other way: 54% at face value, 15% with the errors.

With the trees' own errors

From Berlin to Timișoara, the driest four summers since at least 1500#

For each place, 500 versions of the past were simulated. Each follows the atlas, with errors as large as those it made in 1903–1927, years it was not fitted to, and as persistent from one year to the next as in its fitting years. The colour is the share of versions in which the recent run is drier than every earlier four-summer run, the instrument-based ones included.

These shares are not the chance of a historical fact. They hold only under the model's assumptions: that the trees erred in the past as they did in the early twentieth century, with errors of the same size and persistence, and that the fitted relationship between atlas and instruments is exact.

Since 1500, the recent run is the driest in more than half the versions in 27% of the land, and in more than nine out of ten in 15%. Around Berlin >99%, Warsaw 99%, Timișoara 97%, Prague 96%, Budapest 95%, Rome 91%, Vienna 91%. By country: 98% of Poland, 96% of Hungary, 95% of Czechia, 74% of Germany, 48% of Italy.

The record begins in the year 0 in 61% of the land; in the east and south-east it starts later, in 15% after the year 1000 (around Athens in 967). Where the atlas agrees poorly with the instruments (a correlation below 0.3, 11% of the land) the map is hatched and the page gives no answer.

Back to the year 0?

Two thousand years: only if the oldest trees were as good as the newest#

Over the whole record the shares are lower. The recent run is the driest in more than half the versions in 17% of the land: around Berlin >99%, Warsaw 93%, Budapest 86%, Timișoara 85%, Prague 81%.

Those shares assume that the trees of the first millennium erred no more than those of the early twentieth century, and nothing measures that. With errors half as large again before 1500, the land falls from 17% to 5%. Berlin holds at 83%; Warsaw falls to 12%, Prague to 10%. On the 2,000-year question for central Europe, this atlas cannot decide.

After 1500 there is a check. Written records can test the trees before the instruments: in two regions with long series, the atlas agrees with them about as well over 1500–1900 as over 1903–1978. Swiss summer rainfall reconstructed from documents: 0.55 and 0.53; Czech grape harvests: 0.31 and 0.21. In Andalusia the atlas agrees with neither the documents (0.09) nor the instruments. These two correlations do not show how large the error of any single drought was, nor do they test the simulated shares; they are simply no worse before 1900 than after.

The last time

Where it has been this dry within living memory#

In much of western and northern Europe the answer needs no trees. Around London the instrument record holds a drier four-summer run in 1995–1998; around Paris in 1947–1950, Madrid 1992–1995, Milan 2003–2006, Stockholm 1973–1976.

On this map, pale means a drier run in the last century, purple a drier run centuries ago, red none since the trees' record begins.

Romania

Romania: the west, not the east#

Around Timișoara the recent run (summers 2022–2025) is the driest since 1500 in 97% of versions, around Oradea in 90%, around Cluj in 68%.

Around Bucharest it is a coin toss even since 1500 (50%); the trees there reach back only to 967. Around Iași the trees hold drier runs, the latest with a median date around 1864.

A page on this site compares the trees with the droughts and famines written down in the Romanian chronicles from 1473 to 1738.

Dry soil

Drier soil, not only less rain#

The index tracks water in the soil, which depends on rain and on heat: a warmer summer dries the soil faster. For 2022, a study in Nature Geoscience by Emanuele Bevacqua and colleagues estimated that human-induced climate change contributed 31% of the total soil-water losses in Europe that summer (the middle half of climate models gave 8–51%).

The atlas and the instrument series both estimate the same kind of index, in different versions; the comparison adjusts for the difference in their average relationship, not for every difference, such as how snow is handled. And it is not a comparison of rainfall: a place can have had wetter summers than in 1540 and still drier soil.