What is CLAMP?

Climate Leaf Analysis Multivariate Programe (CLAMP) (Wolfe, 1990, 1993; Kovack and Spicer, 1995; Wolfe and Spicer, 1999; Spicer, 1999; 2000; 2007; 2008; Spicer et al., 2004) is a multivariate statistical technique that decodes the climatic signal inherent in the physiognomy of leaves of woody dicotyledonous plants. It was developed as an evolutionarily robust, accurate, and precise tool for direct atmospheric palaeoclimate determinations over land, and as such complements marine-based climate proxies such as oxygen isotopes. CLAMP calibrates the numerical relations between leaf physiognomy of the woody dicotyledons and meteorological parameters in modern terrestrial environments. Using this calibration, past climatic data are potentially determinable from leaf fossil assemblages provided that the calibration is robust over time and that the sampling of the fossil assemblage represents well the characteristics of the living source vegetation. CLAMP has been applied effectively to fossil floras up to 100 million years old (Herman and Spicer, 1996), but is an even more powerful and reliable tool for late Tertiary (Wolfe, 1995; Spicer, 1999) and Quaternary assemblages.

The statistical engine used in CLAMP is Canonical Correspondence Analysis (CANOCO) (Ter Braak, 1986) as this is robust for data that do not necessarily conform to normality and does not assume independence of variables (essential when leaf characters are the product of functional compromise, constructional efficiency and a finite genome). CANOCO is a direct ordination method, used widely in plant ecology, that orders samples, in this case vegetation sites, based on a set of attributes. In CLAMP the attributes are the scores of the 31 different leaf character states taken from more than 20 species of woody dicots in each vegetation site. The largest calibration dataset now routinely used consists of foliar physiognomic measurements and climate observations from 173, predominantly Northern Hemisphere, modern vegetation sites. Earlier versions used fewer modern sites (e.g. 103 in Herman and Spicer, 1996) and different datasets can be used for different purposes. For example if initial results indicate a warm climate then a subset of the full dataset that excludes the so-called "alpine nest" offers greater precision. (Wolfe, 1993; Wolfe and Spicer, 1999). Because climate variables for these sites are known by observation (>30 year averages are used from climate stations proximal to the sites) vectors for each of the measured climate variables can be positioned in physiognomic space and calibrated. Palaeoclimate variables can be quantified by scoring a fossil assemblage in the same manner as for the modern vegetation and positioning the fossil site in physiognomic space in a passive manner so it does not distort the calibration by its presence. By projecting the position of the fossil site on to the calibrated climate vector the ancient palaeoclimate can be determined.