Digital leaf physiognomy: a multivariate technique for reconstructing climate from fossil leaves. Geological Society of America Annual Meeting, Huston, Texas, USA.

DANA L. ROYER; DANIEL J. PEPPE; WILF PETER; BARBARA CARIGLINO; KIRK JOHNSON; ARI IGLESIAS

Huston, texas, USA

Congreso; Geological Society of America annual meeting; 2008

Geological Society of America

A major export of paleobotany is climate reconstructions. The ‘grand-daddy´ of all leaf-climate methods is leaf-marginanalysis, which is based on the positive correlation between mean annual temperature (MAT) and the fraction ofuntoothed woody dicot species. Another important method is leaf-area analysis, which relies on the relationshipbetween leaf area and mean annual precipitation (MAP). Although these two methods have been largely successful forreconstructing MAT and MAP from fossil plants, a potential drawback is that each is based on only one aspect of leafphysiognomy. To address this shortcoming, several multivariate techniques have been developed, most notably CLAMP,which describes leaf physiognomy with ~30 categorical variables. However, CLAMP typically does not outperform theunivariate approaches, probably because some of its variables are difficult to score in a reproducible manner. In recentyears, we and other colleagues have developed a multivariate technique for reconstructing climate called digital leafphysiognomy. This method is based on characters that can be scored in a reproducible manner, such as tooth number,tooth area, and perimeter-to-area relationships. Further, digital leaf physiognomy is based mostly on continuous (notcategorical) variables, and thus can better differentiate the spectrum of leaf physiognomic information. An earlier studycalibrated the new method with 16 U.S. east coast floras and one Panamanian flora. Results from this study werepromising: multivariate models for estimating MAT were more accurate than leaf-margin analysis, and the methodcould be applied to fossils. Here, we expand the present calibration to over 40 sites, including floras with highlydiffering phylogenetic histories and climate (including MAP). Application to four early Paleogene floras (Republic andBonanza from western North America, and Laguna del Hunco and Salamanca from Argentina) yields estimates of MATand MAP that are similar to estimates from more traditional proxies, but with considerably less error.