Geochimica et Cosmochimica Acta
We measured the δ18O of cellulose (δ18Ocel) extracted from fossil wood collected at 9 sites in the northern and southern hemispheres as a potential source of information about precipitation δ18O (δ18Oppt) in the past and paleotemperatures. The samples had been buried in fluvial sediments for periods of time ranging from ca. 45 million to 250 years. At the oldest localities (high latitude, Eocene- through Pliocene-age sites in Canada and Russia), mean annual temperature (MAT) estimates derived from the modern relationship between MAT and δ18Ocel are 6–16°C lower than the MAT estimates derived from other biological proxies. Estimates of Pleistocene and Holocene mean annual temperatures are close to the modern values at those sites. These results are consistent with other recent findings that the MAT/δ18Oppt relationship across North America was not constant throughout the Cenozoic. Paleo-δ18Oppt estimates derived from fossil cellulose and the modern North American relationship between δ18Ocel and δ18Oppt are within the current annual range of δ18Oppt values at all locations. The middle Eocene δ18Oppt we determined from arctic cellulose samples (−21.9‰) is consistent with river water δ18O determined in two other studies (−19.1‰ to −22‰). These findings provide some evidence that a precipitation δ18O signal may be retained in wood cellulose during millions of years of burial, and that latitudinal patterns in δ18Oppt may not have changed much during the past 45Ma. These interpretations depend, of course, on the assumption that the isotopic composition of the cellulose has not changed during burial, an assumption for which it is difficult to gather direct evidence. XRD analysis shows that the crystalline form of the fossil cellulose we used to estimate paleoprecipitation δ18O and paleo-MAT is the same as that of modern cellulose, and that the samples are free of quartz and iron oxide contaminants that result in negative errors in measured δ18Ocel.