IMPROVED PRECIPITATION ESTIMATES USING LEAF MORPHOLOGY

A database of leaf morphology and climate variables from fourteen modern floral localities along the East Coast of North America has been created to test precipitation equations. The floras are located from New Hampshire to southern Florida, and experience mean annual precipitation between 92 cm and 140 cm, and mean annual temperatures between 4.5°C to 22.3°C. The floras belong to a common geographic province and are floristically similar, thus limiting the chance that variation in leaf morphology between floras is due to major differences in topography or phylogeny between sites.

Are the equations that predict precipitation amounts based upon leaf morphology precise enough to differentiate average precipitation between modern sites in Eastern North America? The relationship between leaf morphology and average precipitation has been quantified using approaches calculated from modern leaf morphology and climate data. The original equations were created to apply to the fossil record for paleoclimate estimates. These equations have not been tested themselves against an independent sampling of leaf morphology from modern sites with known precipitation and temperature.

Four published equations that predict precipitation from leaf morphological characters were tested on the modern floras sampled to determine the validity of these equations. Results show that average absolute error for all equations ranges from 30-61 cm, with an average 95% predictive confidence interval ranging from 117-175 cm. In general, the more morphological characters used to determine mean annual precipitation, the more accurate the temperature estimates were.

A new multiple regression equation using eight morphological characters to predict precipitation amount is introduced based on leaf morphologic and climate data from 138 modern sites from the Americas. When this equation was tested on the same 14 sites, the average absolute error is 15 cm, with an average 95% predictive confidence interval of 52 cm. This new equation improves upon published precipitation prediction equations, and should be considered for use in paleoclimate reconstruction.