Paleoclimatic analysis of three Eocene lacustrine floras (Laguna del Hunco, Chubut, Argentina; Republic, Washington, USA; and Green River, Utah, USA) using digital leaf physiognomy.

CARIGLINO, B; WILF, P; ROYER, D

Dallas, Texas, USA

Conferencia; The 24th Midcontinent Paleobotanical Colloquium, Dallas, Texas.; 2007

Fossil leaf size and shape have been widely used in proxies for past continental climates since Bailey and Sinnott [1,2] first noted the climatic distribution of toothed leaves. Among the most-used methods for estimating past mean annual temperatures (MAT) are leaf margin analysis (LMA), the climate-leaf analysis multivariate program (CLAMP [3]), and the nearest living relative analysis (NLR). Here, we test a new method called digital leaf physiognomy (DiLP [4,5]) using three mid-latitude fossil floras from South and North America. This technique has the potential to provide more accurate paleoclimate estimates because subjective and irreproducible discrete characters are replaced by continuous variables, generated mostly by the image analysis using fixed algorithms. Thus, instead of only quantifying margin percentage of more problematic discrete characters such as base or apex types, DiLP was developed to accurately measure preserved areas (e.g., blade area, tooth area), perimeters, and other variables, as well as their ratios. Importantly, the common underestimation of MAT produced by the riparian vegetation effects (very common in fossil floras, since these tend to be lake or river deposits [7]) is reduced when using DiLP versus LMA or CLAMP [5]. We used a total of 20 extant floras from the eastern USA and Central and South America to calibrate this method [5,6].DiLP was tested using recent collections from the Laguna del Hunco (51.9 Ma; Tufolitas Laguna del Hunco Fm; LH), Republic (49.4 Ma; Klondike Mountain Fm; RP), and Green River (47.3 Ma; Bonanza Fm; BZ) floras. These are ideal test floras because they are well-preserved and well-understood paleoclimatically; thus, they can be used to test and refine new paleoclimate proxies for use in less-understood floras.DiLP mean annual temperature (MAT) estimates were in line with and near the warm end of those obtained by previous methods (LHmat=16+/-2.3 °C; RPmat=14.2+/-2.3 °C; BZmat=21.2+/-2.3 °C), including NLR, LMA, and CLAMP. The "warm" results suggest that riparian bias was mitigated in all three cases and that DiLP is the more consistently accurate proxy. Importantly, DiLP estimates were not significantly affected by the use of partly preserved and even fragmentary leaves. This allows maximization of species diversity and total sample size for analysis, increasing the accuracy of DiLP as in all other methods. Although DiLP approach is labor intensive, this is justified by the decreased riparian bias and the ability to use characters other than margin state reproducibly.