Northern Hemisphere hydroclimatic variability over the past twelve centuries

Ljungqvist FC, Krusic PJ, Sundqvist HS, Zorita E, Brattström G. & Frank D

Nature, vol. 532, 94-98, 2016

This paper is a contribution to the 2k Network

Additional information:

Access the PAGES press release here

Fredrik Charpentier Ljungqvist
Department of History
Bolin Centre for Climate Research
Stockholm University
This email address is being protected from spambots. You need JavaScript enabled to view it.
Mobile phone: +46 70 662 0728


For further clarification about anything related to publicity, please contact one of the Nature offices, as indicated below:

For North America
Neda Afsarmanesh (Nature, New York)
Tel: +1 212 726 9231; This email address is being protected from spambots. You need JavaScript enabled to view it.

For China
Piao Li (Nature, China)
Tel: +86 21 2422 5042; This email address is being protected from spambots. You need JavaScript enabled to view it.

For Japan, Korea, Singapore and Taiwan
Mika Ishida (Nature, Tokyo)
Tel: +81 3 3267 8751; This email address is being protected from spambots. You need JavaScript enabled to view it.

For the UK
Rebecca Walton (Nature, London)
Tel: +44 20 7843 4502; This email address is being protected from spambots. You need JavaScript enabled to view it.

Michael Stacey (Nature, London)
Tel: +44 20 7843 4795; This email address is being protected from spambots. You need JavaScript enabled to view it.


Fredrik Charpentier Ljungqvist (Corresponding author)

Paul J. Krusic

Hanna S. Sundqvist

Eduardo Zorita

Gudrun Brattström

David Frank


1. What is the main significance of the new study?

Our article presents the first Northern Hemisphere hydroclimate multi-proxy reconstruction extending more than a millennium back in time. Secondly, our study shows that there have been larger long-term changes in hydroclimate anomalies during earlier centuries of the past twelve than during the 20th century. We also reveal that the natural long-term variability of hydroclimate is larger than often previously assumed. Our most significant result, however, is likely the finding that state-of-the-art climate model simulations tend to overestimate the increase of wet and dry anomalies during the 20th century compared to what we find in the reconstruction. Whereas the 20th century is the most “extreme” century in terms of wet and dry centennial-scale anomalies in the model simulations there is nothing really special about the 20th century mean wet and dry anomalies in the reconstruction.

2. What happened to hydroclimate in the 20th century compared to previous centuries according to the new reconstruction?

We see a tendency for wetter conditions during the 20th century in much of North America – despite dry decades such as the “dust bowl” in the 1930s – and in parts of China. But the change in hydroclimate is not larger than during some of the earlier centuries. Besides the hydroclimate reconstruction, we also produced a new multi-proxy temperature reconstruction using the same methods as for the hydroclimate reconstruction. We indeed find the 20th century to be likely the warmest century of the past 12 centuries, although the warming is not significantly different from the also very warm 10th century (the peak of the Medieval Warm Period/Medieval Climate Anomaly). But when it comes to hydroclimate it seems like the 20th century did not have the strongest wet and dry anomalies. However, wet anomalies were only more geographically widespread and intensive in the warm 10th century than in the warm 20th century.

3. What implications do the new findings have for rainfall under climate change in the 21st century?

Our study suggests the current generation of climate models may have a more limited ability to predict which regions will get drier and which regions will get wetter with global warming than previously assumed. Moreover, our findings indicate that the unforced natural variability of hydroclimate is very large and that the connection between temperature changes and hydroclimate changes is likely weaker and less linear than we have assumed. But global warming will affect hydroclimate in a variety of ways in many parts of the world. The issue here is that it might be harder than we have thought to predict them skillfully. Nevertheless, palaeoclimate data do agree with models that in some regions, like over southern Italy and parts of the southwestern United States, it is will likely as global temperatures rise these regions will experience more severe droughts. Our reconstruction also agrees with climate model predictions that some parts of the high latitudes are likely to get wetter when it is warmer.

4. Does the study show that there has been no recent increase in precipitation extremes?

No, it does not. We are looking at centennial-scale anomalies, relative to the mean conditions of 1000–1899, and to some extent also decadal-scale anomalies relative to the mean conditions of 1000–1899. Our result that there has been no unprecedented intensification and/or increase in wet and dry anomalies during the 20th century are with regard to the 20th century mean conditions. This conclusion, however, also holds for when we look at decadal time-scales. But we do not assess annual, seasonal or daily extremes in the article, thus we cannot say anything about whether there have been more or less short-term hydroclimate extremes during the 20th century. Moreover, the period we studied ends at 1999 precluding us from saying anything about the hydroclimate changes in the past 16 years. In summary, our study only more reliably compares 20th century average conditions with past centennial-scale average conditions.

5. How do the findings compare with observational studies showing a recent increase in rainfall extremes with rising temperatures?

Our reconstruction cannot assess the evaluation of extreme precipitation during single days or weeks. For example, Westra et al. (2013) found an increase in parts of the world of maximum amount of precipitation during single days with rising temperatures. We do not question this finding, and consider it physically plausible. We cannot, however, with the proxy data assess if such changes have occurred in the past or not. Only a handful of our proxies do reflect extreme flooding events but the variability is usually large in them and it seems hard to assess if recent decades have seen more flooding than some earlier periods. The data are simply not good enough to assess this and the data coverage of records reflecting flooding is too sparse.

6. Are some recent droughts and flooding events unprecedented in nature?

It might well be so. Our study can only compare, with fidelity, 20th century average conditions with past centennial-scale variability. And it is important to remember that 20th century mean conditions are not at all as anomalous as the c. post-1990 conditions in terms of climate change.

7. Does the study capture precipitation changes?

No, the hydroclimate signal is more mixed than a pure precipitation signal. We infer the signal as mainly reflecting relative water/moisture availability in a broad sense. The results presented in the article were the best we could obtain, with the admittedly sparse data coverage, in an attempt to reconstruct hydroclimate of the past. It is simply not possible yet, for lack of sufficient data, to reconstruct twelve centuries Northern Hemisphere precipitation in millimeters, indeed, not even five centuries back.

8. Has the number of extreme precipitation events increased in many parts of the world?

Yes, it seems, in some parts of the world like the maximum amount of precipitation falling during a single day – or during a few hours – during heavy rainfalls has indeed increased with global warming in recent decades. But this, admittedly important, issue is outside the scope of our study due to the number and type of proxy records we have used.

9. Why didn’t you investigate changes in daily or monthly extremes in precipitation and drought?

We simply do not have proxy data that adequately capture hydroclimate variability and extremes at such short time-scales. Only historical documentary data can do that, to some extent, and such records are very limited in space and time. We also decided not to study inter-annual variability since the use of only annually resolved proxies would have resulted in a much more geographically restricted reconstruction.

10. You lack data for reconstructing the whole Northern Hemisphere, right?

Yes, regrettably this is the case. Unfortunately, the geographical coverage of hydroclimate proxy records is still very limited – or nonexistent – in many parts of the world. Our conclusions would of course be even stronger if we had better data coverage. Still we have by far much more millennium-long data than in previous studies of hydroclimate. We can cover key regions such as China, Europe and most of Central America and North America as well as small areas over Africa. Our study highlights the need for more fieldwork and the development of long hydroclimate proxies in different parts of the world with little data presently available. It would be especially important to have hydroclimate data for dry regions as Central Asia and the Middle East. We have recently had, and still have, severe droughts in parts of the Middle East, and presently we have neither meteorological nor proxy records long enough to assess if these droughts are outside the range of natural variability.

11. Are you planning to develop a similar reconstruction for the Southern Hemisphere?

We do not have any such plans right now. There is simply not enough hydroclimate proxy data in from Southern Hemisphere land areas to do this yet. It will likely take many years until the proxy network in the Southern Hemisphere allows for such a study.

12. How might potential dating uncertainties in the proxy record affect the reconstruction?

The issue regarding dating uncertainties is an important one. If we had too large dating uncertainties, the common signal could potentially be cancelled out and, for example, dampen the amplitude of anomalies in the 20th century as well as the anomalies during other centuries. To begin with, this problem could be expected to be smallest in the 20th century since the proxies are expected to contain less uncertainties in the 20th century: sediment records have a better constrained dating usually in the more recent past, and tree-ring records are usually much better replicated (e.g. more samples) in the last few hundred years. Secondly, the uncertainties are quantitatively tested in the reconstruction by A) only using at least decadal resolved proxies with good dating control, and B) testing what happens if all proxies of a certain type (e.g. tree-ring records or sediments records) are removed: the major results remain basically the same.

We are convinced that the “wiggles” (i.e. wet and dry anomalies) cannot be moved by a hundred years or more in either direction for the following reasons:

1) About half of the proxies have annual resolution and absolute dating control – these proxies show the same centennial-scale patterns as the lower resolved proxies with dating uncertainties.

2) We have tested removing all the lake sediments proxies – about a third of the records – that is the proxy type with the largest dating uncertainties. The results still remain essentially the same; the differences are rather a result of less proxy coverage than the removal of substantial dating errors.

3) We made a separate reconstruction from more strictly screened proxies, only including those with at least decadal resolution and a well-constrained dating control. The results are the same as that of the full proxy reconstruction (r = 0.97, p. = <0.01 between the two).

4) Our main results are in agreement with the more geographically restricted spatial reconstructions of summer drought derived from only tree-ring data (i.e. proxies with no dating uncertainties): the North American Drought Atlas (western and central Unites States), the Monsoon Asia Drought Atlas (southeastern Asia), and the Old World Drought Atlas (Europe). In all three atlases one finds the 20th century wet and dry anomalies are not unprecedented.

13. Do the proxy records underestimate extreme events in the 20th century?

First, we do not assess individual extreme wet or dry extremes but centennial-scale – and to a lesser extent – decadal-scale wet and dry anomalies compared to the long-term mean. Secondly, since the proxies seem to be able to capture even larger “extremes” in terms of more intensive and geographically extensive wet and dry anomalies prior to the 20th century we find no plausible explanations of why the proxies would not also be able to do this equally well in the 20th century. The ability of proxy records to capture short-term variations is irrelevant for our study since we only look at hydroclimate anomalies at longer time-scales.

> PAGES Press Release (English, 6 April 2016, PDF)

> Stockholm University Press Release (Swedish, 6 April 2016, PDF)

> Stockholm University Press Release (English, 6 April 2016, PDF)

> Swiss Federal Research Institute WSL Press Release (German, 6 April 2016, PDF)

> Swiss Federal Research Institute WSL Press Release (English, 6 April 2016, PDF)

> Swiss Federal Research Institute WSL Press Release (French, 6 April 2016, PDF)


Press picture Nature 1 English


Figure 1: At the top the trend in annual measured precipitation between 1951 and 2010 is shown from the last scientific report of the Intergovernmental Panel on Climate Change (IPCC). Below the relative hydroclimate changes for four different centuries are shown from the new hydroclimate reconstruction covering the past twelve centuries. The stronger the green colour is the wetter it was and the stronger the brown colour is the drier it was. The anomalies are shown with regard to the long-term mean between AD 1000 and 1899. Copyright: CC BY-SA 4.0 J.P. 2k Network.


Press picture Nature 2 English



Figure 2: The percentage of the Northern Hemisphere land area with wet and dry anomalies, respectively, during the past the twelve centuries. The stronger the green colour is the wetter it was and the stronger the brown colour is the drier it was. The anomalies are shown with regard to the long-term mean between AD 1000 and 1899. Copyright: CC BY-SA 4.0 J.P. 2k Network.

Press picture Nature 3 English small


Figure 3: Drought sensitive 1000-year old trees from the mountains of Greece. Indirect recorders of past precipitation and drought variability such as tree-ring width data were used by the researchers to reconstruct twelve centuries of Northern Hemisphere hydroclimate variability. Photo credit: Paul J. Krusic.


The 2k Network is made up of nine regional working groups. Each regional group collects and processes the best time series and spatial reconstructions of important state variables of the climate system (e.g. surface and 500 hPa geopotential, temperature and precipitation).

> Africa2k (African climate of the last 2 millennia)
> Antarctica2k (Antarctic climate of the last 2 millennia)
> Arctic2k (Arctic climate of the last 2 millennia)
> Asia2k (Asian climate of the last 2 millennia)
> Aus2k (Australasian climate of the last 2 millennia)
> Euro-Med2k (European and Mediterranean climate of the last 2 millennia)
> LOTRED-SA (Long-Term Climate Reconstruction and Dynamics of South America)
> NAmerica2k (North American climate of the last 2 millennia)
> Ocean2k (Marine climate of the last 2 millennia)

More at:

Past Global Changes (PAGES) was established in 1991 to facilitate international research into understanding past changes in the Earth system to improve projections of future climate and environment, and inform strategies for sustainability. It receives funding mainly from the Swiss and US national science foundations. PAGES is a core project of Future Earth and a scientific partner of WCRP.

More at:


Please find a selection of chosen articles below.

6 April 2016 

Large variations in precipitation over the past millennium


Scientist: No evidence of extreme drought and floods in the twentieth century


Variationer i torka och nederbörd större förr


La disponibilité en eau a moins varié au 20e siècle


Variationer i nederbörd större förr


Variationer i nederbörd större förr


Wasser: Weniger Schwankungen als angenomme


Climate change might not be as bad as first thought


El clima extremo en el hemisferio norte fue menor en el siglo xx que en los últimos 1.200 anos


Ny forskning kan förbättra klimatmodeller


Större variationer i nederbörd och torka förr


Climate data since Vikings cast doubt on more wet, dry extremes


Dados climáticos que remontam aos vikings lançam dúvida sobre futuros eventos extremos

7 April 2016

Stora variationer i nederbörd och torka


Factcheck: Are climate models ‘wrong’ on rainfall extremes?


Ny forskning om regn kan förbättra klimatmodeller


Variationer i nederbörd större förr


Climate forecasts may be flawed, says study


ÎNCĂLZIRE GLOBALĂ Precipitații sau secetă: modelele de prognoză se dovedesc parțial defectuoase (studiu)




Klimatet: Regn vanligare förr


Variationer i nederbörd större förr


Naturliga klimatvariationer större än man trott


Variationer i nederbörd större förr


Climate-change models wrong on predicting rain, drought extremes: study


Climate predictions may be wrong, new study finds


Nem is volt olyan szélsőséges a 20. század éghajlata, mint amilyennek láttuk


Le 20e siècle moins extrême que suppose

8 April 2016

Global warming climate forecasts may be flawed, says study


Forsker: Ingen evidens for ekstrem tørke og oversvømmelser i det 20. århundrede


Wissenschaftler sagen einen grundlegenden Wandel des Klimas voraus

Category: Highlights | Journal articles