Estimation of rainfall variables from leaf characters in tropical Africa

Abstract

Regression models are defined for leaf physiognomic and climate variables from 30 modern equatorial African plant communities. Herbarium specimens provide physiognomic data for published species lists for the 30 sites; climate data are from a variety of published sources. By limiting the sampled region to equatorial Africa, where annual amount and seasonality of rainfall are critical determinants of vegetation, leaf physiognomy is found to be related linearly to moisture variables. Principal components analysis demonstrates that the pattern of variation among leaf variables is primarily a function of mean annual and wet months (total of all months with ≥50 mm) rainfall. Multiple regression models are derived for mean annual rainfall, wet months rainfall and the number of dry months. Accuracy of each model is tested by comparing the adjusted predicted values (the value for a case when it is not included in calculation of the regression line) to the actual climate values at the 30 sites. Leaf size is the strongest correlate of moisture; the mesophyll size class has the most significant positive relationship with annual and wet months rainfall. Entire-margined species increase with wet months rainfall and temperature. Acuminate tip and length/width ratio are also positively related to mean annual and wet months rainfall. Rounded base, emarginate tip, and acute tip are inversely related to rainfall variables. Regression models derived from modern plants and climate can be used to reconstruct seasonal and annual rainfall from Cenozoic African leaf fossils.