ABSTRACT
Paleoclimatology is the study of past climate. In paleoclimate reconstruction problems, climate scientists differentiate between two types of prediction or hindcast problems: index reconstructions and climate field reconstruction. This chapter reviews the statistical methodologies underlying paleoclimate reconstruction. Informally, a paleoclimatic reconstruction produces predictions or hindcasts of an underlying climate process, given direct observations of the climate system and indirect observations of climate-sensitive paleo proxies. There is a large variety of climate proxies, including, measurements on tree rings, pollen assemblage, ice core, corals, boreholes, and speleothems. The noise structures of different proxies are characterized differently. For example, the error terms in the tree ring–climate relationship largely reflect that trees are imperfect recorders of local climate. Gaussian spatial and temporal models are often used for characterizing proxy measurements such as tree ring width and density, and chemical measurements such as isotope ratios. This is certainly common for hindcasting past temperatures.
