This week a 78-man team of international researchers organized as the PAGES (Past Global Changes) 2k Network published a report for Nature Geoscience reconstructing Earth's climate for the past 2,000 years. The report is the most comprehensive of its type yet published, and its focus on detailing regional (vs. hemispheric or global) temperatures is unique. Each region is examined in 30 year increments, allowing us to spot regional variations usually lost in the global climate reconstructions.
|Temperature changes for six continents, 0-2000 AD. The graph is divided into bars representing 30 year periods; color represents the extent of warming or cooling.|
Image Source: Adapted from Figure 2: Continental Scale Temperature Reconstructions in PAGES 2k Consortium. "Continental-scale temperature variability during the last two millennia." Nature Geoscience. 22 April 2013.
The PAGES 2k Network's "frequently asked question" page provides a concise summary of the results:
What are the primary conclusions of the study?
(1) The most coherent feature in nearly all of the regional temperature reconstructions is a long-term cooling trend, which ended late in the 19th century.
- The regional rate of cooling varied between about 0.1 and 0.3°C per 1000 years.
(2) Temperatures did not fluctuate uniformly among all regions at multi-decadal to centennial scales. For example, there were no globally synchronous multi-decadal warm or cold intervals that define a worldwide Medieval Warm Period or Little Ice Age.
- A preliminary analysis using a climate model indicates that the overall cooling was caused by a combination of decreased solar irradiance and increased volcanic activity, as well as changes in land cover and slow changes in the Earth’s orbit. The simulations show that the relative importance of each factor differs between regions.
- The period from around 830 to 1100 CE generally encompassed a sustained warm interval in all four Northern Hemisphere regions. In contrast, in South America and Australasia, a sustained warm period occurred later, from around 1160 to 1370 CE.
- The transition to colder regional climates between 1200 and 1500 CE is evident earlier in the Arctic, Europe and Asia than in North America or the Southern Hemisphere.
- By around 1580 CE all regions except Antarctica entered a protracted, multi-centennial cold period, which prevailed until late in the 19th century.
(3) The 20th century ranked as the warmest or nearly the warmest century in all regions except Antarctica. During the last 30-year period in the reconstructions (1971-2000 CE), the average reconstructed temperature among all of the regions was likely higher than anytime in nearly 1400 years. However, some regions experienced 30-year intervals that were warmer than 1971-2000. In Europe, for example, the average temperature between 21 and 80 CE was warmer than during 1971-2000. 
- Cooler 30-year periods between the years 830 and 1910 CE were particularly pronounced during times of weak solar activity and strong tropical volcanic eruptions. Both phenomena often occurred simultaneously. This demonstrates how temperature changes over large regions are related to changes in climate-forcing mechanisms. Future climate can be expected to respond to such forcings in similar ways.
I imagine a lot of the commentary on this study will focus on point #3, for it is the conclusion most relevant to current policy debates. As my interest in understanding the the dynamics of human civilization is just as strong as my interest in contemporary politics, my comments will focus on the first two conclusions, which have the potential to shed light on several larger historical questions.
The novelty of this report lies in its highly detailed breakdown of the last 2,000 years along regional lines. From this regional breakdown it is apparent that many famous climatic events, such as the Little Ice Age and Medieval Warm Optimum, were not global events, but regional ones. Even events of a hemispheric scale differed in intensity from region to region. The Medieval Warm Period, for example, sees warming across the Northern Hemisphere, but it is in Eastern North America and Europe (two regions whose temperature is influenced by the Gulf Stream) that warming is most dramatic. Other events rarely mentioned by historians also become apparent - more dramatic than the Medieval Warm Optimum are the temperatures recorded eight centuries before. Europe's climate in this "Roman Warm Optimum" (my term) was hotter than its climate during the more famous warm period period of the middle ages, and is comparable to Europe's climate during the 20th century. Another warm period, rarely mentioned in the historical literature, can be seen in Oceania and South America during the high middle ages.
Why does all of this matter?
In recent years numerous historians have tied the events they chronicle to the broader climatic conditions of the times. The implications of some of these assertions are far reaching. A good example is Mark Elvin's observation that the Chinese dynasties who managed to expand their influence into Manchuria, Korea, and the Northern steppe all rose to power during periods of warmth:
|Table 1 from Mark Elvin. The Retreat of the Elephants: An Environmental History of China.|
(New Haven: Yale University Press). 2006. p. 5.
An exhaustive review of the relationship between Chinese history and East Asian ecology deserves its own post; for the moment I am content to point out the research routes studies like this one open up to historians. However, these studies do have some serious limitations.
The most important of these is highlighted by existence of the study itself. The PAGES 2k Network built continental reconstructions instead of a single global reconstruction because they felt that previous reconstructions of Earth's temperature masked dramatic regional variation. Their suspicions were proven correct.
This same concern can be applied to to continental reconstructions just as easily as it is to global ones. Do continental reconstructions mask regional variation within continents?
Looking at the geographic distribution of the sites used by the 2k Network to create the study can help us answer this question. One of the largest sets of data was gathered for the Asia continental region. The data was taken from tree ring samples across the continent.
|Location of tree rings used to create temperature estimates for Asia. |
Image Source: Figure S.9 from "Supplementary Information" to PAGES 2k Consortium. "Continental-scale temperature variability during the last two millennia." Nature Geoscience. 22 April 2013. p. 42.
A quick glance at the map shows that the sites are not evenly distributed. The majority are found in the sparsely populated Altai, Qilian, and Himalayan mountain ranges or the island of Hokkaido. The Deccan Plateau and Yangtze River Basin have one site each, South East Asia claims less than three, and the Manchurian plain and Yellow River Basin have none at all. The reason for this seems fairly obvious: Asia's major population centers will have few century old trees to reconstruct temperatures from. Unfortunately for historians, these areas are also home to the region's oldest civilizations and were the center of the continent's greatest sedentary empires. Thus our question becomes: to what extent does the climatic history in Mongolia, Tibet, or Hokkaido reflect temperatures in Nanjing or Hyderabad?
Measurements of climate change during the last century provides a partial answer. As the study notes, the 20th century was marked by warming on a global scale. The figure at the top of this post breaks this warming down on a continent-by-continent basis. Because the data available for the 20th century far outstrips that of past centuries, this data can be analyzed with greater finesse. Consider the following representation of global temperature and precipitation change from 1950 to 2000:
|Changes in average temperature and precipitation, 1950-2011.|
Image Source: adapted from Geert Jan van Oldenborgh, Francisco Doblas-Reyes, Sybren Drijfhout and Ed Hawkins."Verifying Regional Model Trends." RealClimate. 15 April 2013.
Continental variation is easy to see. More interesting is regional variation within continents. Again taking Asia as the example, the degree of warming differed quite drastically from one region to another. Bangladesh, Sichuan, the Yellow River Plain, and the Russian steppe vary widely in the pace of warming seen during these 50 years. At least 2 degrees Celsius mark the difference between areas adjacent to the Bay of Bengal and those within the Eurasian steppe! PAGES 2k Network's decision to base their regions on continents (instead of biomes or another less arbitrary unit) limits the study's usefulness. 
With this limitation recognized, the study is still worth using and referring to. Many other studies of this type limit their claims to century long averages. The PAGES 2k team was able to report average temperatures on a thirty year basis, and promises that in the future it will be able to create annual averages for the entire time period studied. This means that for the areas like Western Mongolia, New Zealand, or Hokkaido where multiple sites exist, scientists can deduce the average temperature of any given year of the last 1,000. That is a truly remarkable achievement.
 PAGES 2k Consortium."2k Network FAQ." www.pages-igbp.org. accessed 22 April 2013.
The citation for the original study is:
PAGES 2k Consortium. "Continental-scale temperature variability during the last two millennia." Nature Geoscience. Published online 22 April 2013. doi:10.1038/ngeo1797
 This is true for the other continents studied as well. The majority of temperature reconstructions for the Australian continent, for example, were from New Zealand, while the few sites on the Australian continent itself were about equal to the sites on Pacific islands or atolls. The continent's interior had no sites at all. North America is another intriguing case. Here the data is divided between two sets: pollen counts, whose sites were centered on the Eastern United States, and tree rings, whose sites were centered on the Rocky Mountains. The variation between the two data sets probably reflects the different geographic distribution of each.