Despite the wide error bars, Figure O-4 was misinterpreted by some as indicating the existence of one “definitive” reconstruction with small century-to-century variability prior to the mid-19th century. It should also be emphasized that the error bars in this particular figure, and others like it, do not reflect all of the uncertainties inherent in large-scale surface temperature reconstructions based on proxy data.
A more recent and complete description of what we know about the climate of the last two millennia can be gleaned from an inspection of Figure O-5, which was prepared by this committee to show the instrumental record compiled from traditional thermometer readings, several large-scale surface temperature reconstructions based on different kinds of proxy evidence, and results from a few paleoclimate model simulations. Figure O-5 is intended only to provide an illustration of the current state of the science, not a comprehensive review of all currently available large-scale surface temperature estimates.
The instrumental record shown in panel A is compiled from traditional thermometer readings that measure the temperature of the air just above the land surface (or, for ocean points, the temperature of the water just below the ocean surface). Panel B shows a global surface temperature reconstruction based on changes in the lengths of many mountain glaciers, which shrink when the climate warms and grow when the climate cools, and also a global surface temperature reconstruction based on borehole temperature measurements. Panel C shows a compilation of several recent multiproxy-based and tree-ring-based Northern Hemisphere surface temperature reconstructions, each performed by a different paleoclimate research group using its own selection of proxies and its own calibration and validation protocols. Panel D shows results from two climate model experiments forced with time-varying estimates of natural climate forcings over the last 1,000 years plus anthropogenic forcing since the start of the Industrial Revolution.
Each of the curves in Figure O-5 has different uncertainties and somewhat different geographical and seasonal emphasis; no one curve can be said to be the best representation of the actual variations in Northern Hemisphere or global mean surface temperature during the last 1,100 years. Nor is it possible to assign error bars to either individual reconstructions or the ensemble of reconstructions that reflect all of the uncertainties inherent in the conversion of proxy data into large-scale surface temperature estimates.