Thames in London during the cold winters of the Little Ice Age, with the last one occurring in 1814. It would be quite wrong, however, to attribute their absence since that time solely to a rise in Northern Hemisphere winter temperatures: As London has grown and developed, the “urban heat island” effect has reduced the likelihood of frosts in the city center, and the replacement of the old London Bridge in the 1830s allowed greater uptide incursion of saltwater, which freezes less easily. Manley’s central England temperature series indicates that the winter of 1962–1963 was the third coldest since 1659, yet the Thames did not freeze below its tidal limit (Jones and Mann 2004b).
The problem of quality control becomes even more acute further back in time, so that—in contrast to natural archives such as ice cores or tree rings—historical records generally degrade in their utility as they become older. There are, for example, weather records preserved in Irish and Norse annals back to the middle of the first millennium A.D., but their dating is imprecise and descriptions of weather and climate often are exaggerated. Understandably, historical observations also tend to focus on extreme events rather than climatic averages. For example, it was major storm events that most concerned Venetian traders and mariners; their records were used by Grove (2004) to reconstruct the climate of Crete in the 16th and 17th centuries. Documentary evidence is one of the few kinds available that can register severe floods, hurricanes, and other natural disasters. Consequently, their analysis enables an investigation of the relationship between variations in climate and the frequency and severity of extreme events, a subject that is of major societal concern in relation to projected global warming.
Historical observations are typically discontinuous through time and, as such, one of their most valuable roles is in providing a cross-check on reconstructions based on other proxy records, such as tree rings, and on the validity of paleoclimate model simulations. For example, modeling experiments show marked warming in Siberia during the winters immediately following major explosive volcanic eruptions, such as that of Pinatubo (Shindell et al. 2003). The diaries of travelers passing through northern interior Asia in key years (e.g., 1815–1816, 1883–1884) would allow this prediction to be tested independently.
Europe and East Asia are the two regions of the world where long temperature series have been most successfully developed from documentary evidence in a repeatable and consistent way for periods of more than the last two centuries.
The documentary evidence for Europe as a whole has been reviewed by Brázdil et al. (2005) and for the Mediterranean region by Luterbacher et al. (2006). Seasonal temperature data have been compiled for most areas of central and western Europe back to 1500, and these show that the late 20th and early 21st centuries have been warmer, at high probability for three out of four seasons, than any time period in the last five centuries (Xoplaki et al. 2005) (Figure 3-1). By combining documentary evidence with other proxy data, Luterbacher et al. (2004) were able to map winter and summer temperature anomalies across Europe for individual years back to 1500, along