Paleoclimate of highly diverse, well preserved early Eocene floras from North and South America: A test for digital leaf physiognomy.

CARIGLINO, B; WILF, P; ROYER, D

Bilbao

Conferencia; Climate and Biota of the Early Paleogene; 2006

Laguna del Hunco (LH, Chubut, Argentina) and Republic (Washington, USA) are midlatitude sites with well-preserved fossil floras from the warm, equable, early Eocene that have been intensively sampled. The floras are important because they each have well over 100 plant species and have been subjects of paleoenvironmental, paleoclimatic, systematic, and plant-insect associational studies. They are both lacustrine floras from volvanic settings, although LH is considered to be a sea level assemblage, and Republic was a montane flora at uncertain elevation. LH dates from 51.91 +/- 0.22 Ma, and a previous study using leaf margin analysis estimated the mean annual temperature (MAT) of LH as 16.6 +/- 2.0 °C. Analysis based on nearest living rlatives was concordant with this result or with higher temperatures.The Republic flora has been dated at 49.4 +/- 0.5 Ma. Several paleoclimatic studies on this flora have converged on similar results: Republic represents a warm-temperate forest with mixed broadleaved deciduous and evergreen elements, based on bioclimatic studies (MAT = 13.5 +/- 2.2 °C), climate-leaf analysis multivariate program (~11 °C), and leaf margin analysis (~9.4 °C). Thus both floras are well understood paleoclimatically and can be used to test and refine nex paleoclimate proxy methods.Here, we present new paleotemperature estimates for both fossil sites obtained from the first tests of digital leaf physiognomy (DILP). A study by Royer et al presented preliminary results showing potential of DILP to estimate temperatures. Using Adobe Photoshop, leaf images are processed by digital extraction of the fossil leaf from the matrix, restoration of margins whenever applicable, and petiole removal. A separate image is then generated with the teeth removed from the margin. Measurements are made using an image analysis program (SigmaScan Pro) on the complete leaf (Image I), the leaf blade without its teeth (Image II), and the teeth without the blade (Image III).The physiognomic variables that proved to be most applicable to the preserved portions of incomplete fossil leaves were ratios such as tooth area to blade area and perimeter ration (of Image I to Image II). Applying DILP to Laguna del Hunco using the "fossil" regression model of Royer et al, which uses margin percentage, perimeter ratio, and tooth area to blade area ratio as variables, resulted in a MAT estimate of 7.8 +/- 2.0 °C. This was lower than expected based on abundant evidence from the nearest living relatives indicating much higher temperatures. The low estimate results from the LH flora having much higher values for the perimeter ratio variable than any floras in the Royer et al calibration data set. Substituting the perimeter ratio variable in Royer et al model with the ratio of tooth area to internal perimeter, the estimated temperature rose to a value more in line with the floral composition (19.4 +/- 2.7 °C). This suggests that regression models from DILP need further revision, with the addition and analysis of a much greater geographic range of calibration data, before application can be made to floras from the Southern Hemisphere. The Republic flora provides a first test of DILP for a North American fossil flora, as well will demonstrate (work in progress at abstract deadline). With further calibration work, digital leaf physiognomy has the potential to provide significant improvements in the precision of paleotemperature estimates.