Scientists use climate models to understand current climate conditions, investigate natural processes and the anthropogenic activities that can affect them, and project future climate scenarios. A key challenge is testing the ability of climate models to accurately capture the natural variability of climate, which can dominate long-term climate change on decadal to centennial timescales.
Capturing natural variability requires data sets that are sufficiently long (centuries to millennia) to adequately characterize past variations in temperature and other variables. The instrumental record is too short to accomplish this, so researchers rely on proxy records from ice cores, tree rings, corals, and other paleoclimate archives.
They then use a number of methods to reconstruct spatial “fields” of data—such as precipitation, surface air temperature, and sea level pressure—interpolated from the information derived from those sparse proxy records. Most such methods rely on purely statistical relationships and do not take advantage of physical relationships embedded in climate models.
Here Hakim et al. fuse both sources of information—climate models and observations—taking a data assimilation technique commonly used for forecasting the weather and adapting it to paleoclimate analyses. The results are the first from their National Oceanic and Atmospheric Administration–funded project called the Last Millennium Climate Reanalysis (LMR). LMR aims to use paleoclimate proxies to extend gridded estimates of climate fields well beyond the 150-year instrumental record.
The results, based on a sparse, preliminary collection of proxy records, show good agreement with earlier reconstructions of the Northern Hemisphere’s average near-surface air temperature. This includes an overall cooling trend over the past 2000 years, reversing abruptly during the Industrial Revolution, as well as multicentury fluctuations consistent with the nominal Medieval Warm Period and the Little Ice Age.
When compared with the 1880–2000 instrumental temperature data, the results are most skillful in reflecting the tropics. A comprehensive error analysis was used to quantify uncertainty, including verification of the results against independent (not assimilated) proxies.
To demonstrate the advantages of their approach, the authors examined the climate reconstructions of the 1808–1809 “mystery” volcanic eruption, which is known about from ice core records but not from direct observations. The results clearly show an abrupt cooling in 1809, which is enhanced by the Pacific/North American teleconnection pattern. The dynamical conditions revealed by the analysis are also consistent with those of documented volcanic eruptions.
The results demonstrate the ability of the integrated data assimilation approach to capture climate fluctuations on multiple timescales and in multiple climate variables, providing confidence in its reconstructions. The approach also demonstrates the potential to improve historical climate estimates by combining climate modeling with proxy information. (Journal of Geophysical Research: Atmospheres, doi:10.1002/2016JD024751, 2016)
—Terri Cook, Freelance Writer