Climate Leaf Analysis Multivariate Program (CLAMP)

Section contents

Paleoclimate estimation with plant fossils

1. Nearest living relative (NLR) methods
2. Univariate leaf physiognomic methods
3. Climate-Leaf Analysis Multivariate Program (CLAMP) ←
4. Digital leaf physiognomy (DiLP)

The Climate Leaf Analysis Multivariate Program (CLAMP) has its roots in the work of Jack A. Wolfe (1993, 1995).

Leaf characters & scoring

The first step in a CLAMP leaf physiognomic analysis is scoring the leaf morphotypes in a flora. Scoring the leaf morphotypes somewhat resembles scoring taxa in a morphological matrix for a phylogenetic analysis. CLAMP characters have states that are assigned values. In some cases, the values are fixed. In others, they vary depending on the number of categories (character states) that apply to a given morphotype. The CLAMP characters below have thus been broken into two tables that reflect how they are scored.

CLAMP Scoring System: Characters with Fixed Scores¹

¹Characters after Wolfe (1993) and CLAMP Online.

CLAMP Scoring System: Characters with Variable Scores²

The categories below are divided into three or more characters. Each characters is scored separately, with a total score that sums to one (1) for the whole category. A morphotype should received a fractional score for all characters to which it belongs for a given category. The fractional score for each characters = 1 ÷ (the total number of characters found in that morphotype). For example, if a morphotype has four characters in a given category, it receives a score of 0.25 (= 1 ÷ 4) for each characters. Characters not found in a morphotype receive a score of zero (0).

²Characters after Wolfe (1993) and CLAMP Online.

CLAMP Analysis

CLAMP uses canonical correspondence analysis (CAA) in order to estimate climate parameters for a given flora. As with other paleoclimate methods based on leaf physiognomy, CLAMP estimates the paleoclimatic parameters of fossil floras based on datasets drawn from modern vegetation. The basic approach behind CLAMP is not that different than the basic approach behind leaf margin analysis (LMA) or leaf area analysis (LMA). All three types of analyses try to answer the basic same question: Given the known leaf physiognomic variable(s) of a fossil flora, what are the estimated climate variable(s)?

As in LMA and LAA, CLAMP uses datasets from modern floras in order to develop equations that can be used to mathematically relate leaf physiognomy to climate parameters. In CLAMP, however, the relationship between physiognomy and climate parameters is more complicated; rather than just one leaf character, 31 leaf characters are being used simultaneously to relate leaf physiognomy to each of 11 climate parameters!

Calibrating CLAMP

In order to run a CLAMP analysis, the program must be calibrating using a dataset based on modern vegetation. Happily, several different calibration datasets have already been compiled for CLAMP (available for download from the CLAMP website); these are regional (Asia, northern hemisphere, southern hemisphere) or global. The choice of calibration dataset depends on the geographic region under study and the approach the investigator wishes to take.

The data points used in CLAMP calibration datasets are modern sites or sampling localities. For each site, data are compiled on climate and leaf physiognomy. For leaf physiognomy, each woody dicot species at a site is scored using the CLAMP characters listed above. (At least 20 species must be score per site.) Then, a percentage score is calculated for each leaf physiognomic character for each site. The percentage score for each character is calculated as follows:

Percentage score = (sum of all species scores) ÷ (number of species scored) × 100

Thus, two data matrices are compiled to calibrate a CLAMP analysis:

Matrix 1: sites and their percentage scores for each leaf physiognomic character.

Matrix 2: sites and their climate variables.

The calibration data matrices are used to calibrate CLAMP, or, in a nutshell, to mathematically relate leaf physiognomic and climatic variables using datasets from modern vegetation. In order to do this, the sites are plotted in a cloud that positions them relative to one another based on their leaf physiognomic scores. This cloud is physiognomic space. Vectors representing each of the climatic variables are drawn through physiognomic space to summarize how each climate parameter relates to leaf physiognomy. The vectors are the basis for equations (regression models) that relate the leaf physiognomic variables to the climate variables at each site.

Analyzing a fossil flora with CLAMP

For a fossil flora, only leaf physiognomy can be described through observation; climate variables must be calculated. In order to apply CLAMP to a fossil flora, leaf physiognomic characters must be tabulated for at least 20 woody dicot taxa in the fossil flora. The percentage score for each leaf physiognomic character is then calculated (see the equation above).

In a CLAMP analysis of a fossil flora, CLAMP places the fossil flora into physiognomic space among the modern floras used to calibrate CLAMP. Then, it uses the equations based on the climate vectors to estimate the climate variables of the fossil flora. CLAMP can be used to analyze more than one fossil flora simultaneously. If more than one fossil flora is analyzed simultaneously, the percentage scores must be tabulated for each fossil flora.

Climate parameters

Which climate variables can be estimated using CLAMP? The calibration datasets include 11 variables, each of which can be estimated for fossil sites. The table below lists the variables and their abbreviations as used in the CLAMP calibration datasets.

Climate Parameters Included in CLAMP Datasets³

³List of variables after Yang et al. (2011) and CLAMP Online.

Pros & cons of CLAMP

CLAMP has been and continues to be widely used to estimate the climate parameters fo fossil floras. The method, however, is not without its critics. The scoring system for CLAMP is somewhat arbitrary, and questions have been raised as whether and to what extent some of the included leaf characteristics correlate with the CLAMP climate parameters. CLAMP may be subject to variation due to differences in how users interpret and score the characters. Furthermore, the results of CLAMP may not be an improvement on the results produced by the simpler analyses (LMA, LAA). Finally, the use of canonical correspondence analysis (CCA) and the scoring of the CLAMP calibration datasets have been critiqued.