Continental-scale patterns in modern wood cellulose δ18O: Implications for interpreting paleo-wood cellulose δ18O

The isotopic composition of ancient wood may be a useful archive of some climatic or geochemical conditions of the past, but presently there are many uncertainties that constrain such interpretations. We sampled naturally growing, predominantly native trees in forested regions of North America and the Caribbean to evaluate the strength of the relationships among cellulose δ18O (δ18Ocel), relative humidity (RH), precipitation δ18O (δ18Oppt), and mean annual temperature (MAT) at the continental scale, and the general range of variability in δ18Ocel associated with site hydrologic conditions and species differences. We found up to 4‰ differences among different species growing at the same site, that conifer cellulose at a site is more enriched than angiosperm cellulose by 1.5‰ (p < 0.001), and that differences in landscape position, reflecting differing access to the water table, produced differences of <1‰ in δ18Ocel. At the continental scale, δ18Ocel was strongly influenced by modeled δ18Oppt (R2 = 0.80, p < 0.001). Average summer minimum RH (RHmin) combined with δ18Oppt explained more of the variability (R2 = 0.93, p < 0.001) in δ18Ocel across North American and Caribbean forests. MAT and δ18Ocel were also strongly correlated across North America (R = 0.91 and 0.95, p < 0.001, for angiosperms and conifers, respectively). The difference between δ18Oppt and δ18Ocel is not constant (varying from 35-44‰) and is inversely correlated with δ18Oppt. The relationships among δ18Oppt, RHmin, δ18Ocel, and MAT established for North America and the Caribbean applied reasonably well when δ18Ocel was used to estimate MAT and δ18Oppt in Asia, Europe, and South America, but there were important exceptions. The most accurate predictions of MAT and δ18Oppt from δ18Ocel require RHmin. Predictions of δ18Oppt and MAT made from δ18Ocel alone produced errors of up to 8‰ and 16 °C, respectively.