APPLYING PALEOBOTANICAL PALEOCLIMATE PROXIES: A CASE STUDY OF TROPICAL AFRICAN FLORAS

The climatic conditions of a region dictate the size and shape of woody dicotyledonous angiosperm leaves in a predictable pattern. For example, colder climates result in leaves with more teeth, while wetter climates result in larger leaves. Paleobotanists have used these relationships to develop proxies for paleoclimate. The univariate methods leaf margin analysis (LMA) and leaf area analysis (LAA) respectively use the relationship between the proportion of species with teeth and temperature and the size of leaves and precipitation to reconstruct mean annual temperature (MAT) and mean annual precipitation (MAP). Digital Leaf Physiognomy (DiLP) is a multivariate model that uses multiple characteristics of leaf size and shape to estimate MAT and MAP, and offers some improvement in climate estimates over the univariate methods. Studies of African floras have shown that they have a different climate-leaf shape relationship than floras from the temperate Northern Hemisphere. However, the DiLP model is best calibrated for the temperate Northern Hemisphere and the current database does not include any African floras, thus the applicability of DiLP to African floras is unclear. To test the effectiveness of DiLP for predicting climate using African floras, we measured modern leaves from multiple sites across tropical Africa. The DiLP model tended to overestimate MAT and underestimate MAP when compared to actual climate. However, the discrepancy between predicted and actual was not consistent: MAT overestimates varied from 2 to 10°C, while MAP underestimates varied from 800 to 1500 mm. In comparison, LMA and LAA more accurately estimated climate, but the methods still overestimated MAT and underestimated MAP. This inaccuracy has important implications for paleoclimate estimates using DiLP. Because the model’s calibration dataset is primarily restricted to the temperate Northern Hemisphere, it cannot accurately estimate climate in tropical Africa. Additionally, these results suggest that LMA and LAA cannot be reliably applied to African floras. Therefore, the DiLP calibration dataset must be updated to include African floras and/or a regional model for Africa must be developed before DiLP, LMA, and LAA can be reliably used to estimate the paleoclimate of tropical Africa.