Regional principal components for temperature and precipitation PNAS

Geospatial-temporal analysis reveals two trends in global climate during the last deglaciation

The last deglaciation, which stretched from 19,000 to 11,000 ago, was characterized by increases in surface temperatures of 10-15 °C punctuated by millennial-scale warming/cooling periods, pulses of increasing atmospheric carbon dioxide and asynchronously increasing methane, and an overall sea level rise of 80 meters. In this geospatial-temporal study the authors compiled and analyzed sea surface temperatures and precipitation from ice cores, sea floor sediments, pollen, cave calcite records, and sea phytoplankton records. A principal component analysis revealed two important trends responsible for the variability in temperature during this transition. The dominant warming trend, which ultimately brought temperatures to pre-industrial levels, is global and correlates strongly with increasing atmospheric carbon dioxide levels. The second trend revealed by the analysis is geographically complex and is responsible for regional millennial-scale warming/cooling periods. Evidence from Pa/Th ratios in ocean sediments suggest the latter trend is associated with changes in the strength of the primary Atlantic north/south current.


The global climate underwent massive changes from the end of the Last Glacial Maximum (LGM) approximately 19,000 years ago to 11,000 years ago. During the LGM temperatures in East Antarctica were approximately 9–10 °C lower than today. In Greenland average temperatures were 15 °C lower.  Both surface temperature and atmospheric CO2 concentration began to increase about 17,500 years ago. Over the period of the deglaciation, atmospheric CO2 concentrations increased in pulses ultimately increasing by about 80 parts per million (ppm).

The last deglaciation was punctuated by several short term warming and cooling events which averaged about 1,500 years in duration. The largest were the Oldest Dryas cold period (19-15,000 years ago), Younger Dryas cold period (13-12,500 years ago) and intervening Bølling-Allerød warm period. The very abrupt warmings in Greeenland at the end of the Oldest Dryas and beginning of Bølling-Allerød raised average temperature on Greenland by about 9 °C.  At the end of the Younger Dryas temperatures increased by about 10 °C at the beginning of the Holocene, the current warm period.


Sea and land surface temperature and precipitation patterns were analyzed from 166 geographically distributed paleoclimate records of proxies for temperature (sea surface or continental) and precipitation for the interval 20-11,000 years ago. Proxies came from a variety of sources including ice cores, sea floor sediments, pollen, cave calcite records, and sea phytoplankton records. For example, alkenones, which are organic compounds produced by a particular type of phytoplankton, are used as a proxy for sea surface temperature. The ratio of oxygen 18 to oxygen 16 (delta-oxygen-18) is used as a proxy for land surface temperature. Other proxies include ice core delta-oxygen-18, pollen, cave calcite delta-oxygen-18, sea floor sediments, and others. Only records with resolution better than 500 years and with dating determined from at least several radiometric sources were included. Radiometric dating, including radiocarbon, potassium-argon, and uranium-lead dating, is based on the decay on naturally occurring radioisotopes and is considered the most reliable form of dating.

Geospatial-temporal analysis

The analytical technique used to analyze the geographically distributed sea surface temperature data to identify geospatial and temporal patterns during the last deglaciation is called principal component analysis. It is a way of decomposing geospatial-temporal data sets to separate physical processes from background noise. By analogy to Fourier analysis of radio signals, it is conceptually similar to extracting low frequencies that carry a message from high frequencies that represent noise. The principal component analysis identifies the dominant spatial patterns and principal time series which explain the greatest amount of variability in surface temperatures record during the last deglaciation.  Principal component analysis was applied to sea surface temperatures, land and sea surface temperatures and to different regions of the Earth’s surface.

Principal component analysis of temperature variability during last deglaciation PNAS
(A) Principal component PC1 based on all of the sea surface temperature records (solid blue line), only on alkenone (dashed light-blue line) and Mg/Ca records (dashed orange line). The percentage of variance explained by PC1 is 49%. (B) Principal component PC2 based on all of the sea surface temperature records (solid blue line), only on alkenone (dashed light-blue line) and Mg/Ca records (dashed orange line). The percentage of variance explained by PC2 is 29%.
LGM Last Glacial Maximum; OD Oldest Dryas; BA Bølling–Allerød; YD Younger Dryas.

69 high-resolution sea-surface temperature proxy records spanning the period 20–11,000 years ago were compiled. The data shows that warming trends were smallest at low latitudes (1–3°C)and higher at higher latitudes (3–6°C). This data was subjected to a principal component analysis to extract the dominant trends, temporal and geospatial, that accounted for the greatest sea surface temperature variability. The analysis revealed that two trends that together for 78% of the variability in the global sea surface temperature during the period 20-11,000 years ago.

  • The first trend is geographically uniform. Its associated principal temporal component (PC1) displays a two-step warming pattern with the first extending 18–14,300 years ago, followed by a plateau, and the second warming resuming during 12,800–11,000 years ago. It is the the most important component and accounts for nearly 50% of the variance in the sea surface temperature.
  • The second trend is more complex geographically, but its associated principal temporal component (PC2) oscillates with decreases during the Oldest and Younger Dryas cooling events and separated by an increase during the Bølling–Allerød warming period. It accounts for less of the variance than PC1, explaining less than a third of the variance.

Applying a similar analysis to land surface temperature data sets result in principal components for land surface temperature during the last deglaciation.  The analysis indicates that, as in the case of sea surface temperature, two trends explains much of the variability (64–100%) in regional and global climate during the last deglaciation.

Primary trend: Global warming

The principal component analysis identified two major physical trends that influenced surface temperatures during the last deglaciation. The most important that accounts for 50% of the variance is a global warming trend that started at about 18,000 years ago and is composed of two warming periods between 18,000–14,300 and 12,800–11,000 years ago which brought global temperatures to pre-industrial levels. This trend correlates strongly with increasing greenhouse gas levels.  CO2 concentrations started to rise approximately 17,500 years ago.   At 14,700 years ago CO2 levels plateaued and then began rising again between 12,900–11,700 years ago. CH4 concentrations also began to rise starting at approximately 17,500 years ago, but followed a different path from CO2.  There was an abrupt increase 14,700 years ago, followed by an abrupt decrease about 12,900 years ago followed by a rise at approximately 11,700 years ago.  The contrasting CO2 and methane profiles suggests that the source of atmospheric methane is different from that of CO2.

Secondary trend: Millennial scale regional warming and cooling

Principal components compared to CO2 and AMOC intensity during last deglaciation PNAS
A) Comparison of the global temperature PC1 (blue line) with record of atmospheric CO2 from Antarctica ice core (red line) (B) Comparison of the global temperature PC2 (blue line) with Pa/Th record (a proxy for Atlantic overturning current) Also shown are freshwater fluxes from ice-sheet meltwater, Heinrich events, and routing events. (C) Comparison of the global precipitation PC1 (blue line) with record of methane (green line) and radiative forcing from greenhouse gases (red line).

The second physical trend that accounts for less than a third of the variance was found to correlate with the North Atlantic Pa/Th ratio record.  Sedimentary protactinium/thorium ratios (Pa/Th) ratios provide a proxy for the strength of past ocean circulation.  The Atlantic Meridional Overturning Circulation or AMOC brings warm surface waters from the Southern Ocean to the north, where increasing saltiness (salinity) causes it to drop to the ocean floor to form deep water which then flows south.  The Pa/Th ratios provide evidence that the strength of the Atlantic overturning current is weak during Northern Hemisphere cooling periods such as the Oldest Dryas and Younger Dryas.

A plausible mechanism is that large inflows of meltwater in the north from melting glaciers reduced the strength of the Atlantic overturning current. A reduced Atlantic current results in cooling in the north and warming in the south – referred to as the hemispheric sea-saw. For example, the large reduction in the Atlantic overturning current during the Oldest Dryas can be explained as a response to a large meltwater pulse (Heinrich event HS-1) about 19,000 years ago from Northern-Hemisphere ice sheets. Similarly the reduction in the strength of the current during the Younger Dryas was likely caused by a freshwater pulse through the St. Lawrence River and is associated with Heinrich event HS-0.

Supplementary Notes

Measuring climate change during the last deglaciation


Global climate evolution during the last deglaciation, Peter U. Clark, Jeremy D. Shakun, et al., Proceedings of the National Academy of Sciences 2012 vol. 109 no. 19 E1134–E1142