The use of a resemblance function in the measurement of climatic parameters from the physiognomy of woody dicotyledons

Abstract

Abstract Wolfe compiled a data set (CLAMP) of the physiognomy of woody dicotyledon leaves with a view to using it to extract climatic information from fossil floras. Attempts made, using these data, to identify and describe the relationship between plant physiognomy and climate, have all searched for a global relationship applicable throughout data space. We investigate the possibility that the relation between climate and physiognomy is so non-linear that only local relationships should be sought. We use a resemblance function to select sites that are physiognomically similar to each unknown site; these similar sites, alone, yield a climatic estimate for the unknown site. Data collected from New Zealand and Australia are then added to Wolfe's data set in order to test this new method. We show that the use of a resemblance function enables us to produce Mean Annual Temperature estimates for all modern sites that lie within 2.6°C of meteorological data for 90% of the sites. This method also allows the objective identification of outliers based on their distance in physiognomic space from their nearest neighbours. When a method is used that searches for a global relationship between physiognomy and climate, the misfit of estimated Mean Annual Temperature for the combined Southern Hemisphere and CLAMP sites is almost twice as great (5°C) as that to the CLAMP sites alone. In contrast, when the resemblance function is used, the RMS misfit of estimated Mean Annual Temperature increases by only 0.2°C. This result demonstrates the robustness of the new method.