Last Millennium Reanalysis

Past Climate Variability and Paleoclimate Records

How can climatologists claim that the last decade has been warmer than any decade in the past few hundred or thousand years if satellites have been circling the Earth, measuring temperatures of the planet's surface, for only a few decades? We can use weather station and ship-board measurements of surface temperatures in the pre-satellite era, but this only extends our record of surface temperatures back to the late 1800's.

We have to rely on paleoclimate records to extend our knowledge of past climate into the pre-instrumental era. These proxy records, such as tree ring widths, or isotopes from corals, ice cores, or lake sediment cores, can be used to reconstruct past temperature or rainfall at a specific location or in a specific region.

For example, the figure below shows the paleoclimate record locations and types from the recent version 2 of the Past Global Changes 2k (PAGES2k) data set.

Although these records are geographically widespread, they do not cover every location around the globe. Figure is from the PAGES2k scientific data release paper published in Nature Scientific Data.

Last Millennium Reanalysis:

Combining Model Data and Paleoclimate Data

The Last Millennium Reanalysis (LMR: a multi-institutional effort, now available online) employs a data assimilation approach to reconstruct past climates. The LMR combines annually resolved paleoclimate data with climate model data to reconstruct climate fields of the last millennium.

This product can be used to 'spread' paleoclimate information geographically between records where there are no annually resolved proxy records (e.g., over much of the ocean, central Africa, and central South America) in a dynamically consistent way.

I am currently using the LMR to study past climate conditions coinciding with extended drought. I am also using the LMR to examine which geographic regions may be significantly influencing global mean surface temperatures.

For example, the figure above shows the relationship (using a multi-taper spectral coherence method) between local temperature variability at a given location and the global mean temperature 1500-1850CE. Dark red indicates there is a strong relationship, and white indicates a weak relationship. The black dots (stippling) show where there is a statistically significant relationship between local and global temperatures, and local temperatures slightly lead the global.