When are leaves good thermometers? A new case for Leaf Margin Analysis

Research output: Contribution to journalArticle

Abstract

Precise estimates of past temperatures are critical for understanding the evolution of organisms and the physical biosphere, and data from continental areas are an indispensable complement to the marine record of stable isotopes. Climate is considered to be a primary selective force on leaf morphology, and two widely used methods exist for estimating past mean annual temperatures from assemblages of fossil leaves. The first approach, Leaf Margin Analysis, is univariate, based on the positive correlation in modern forests between mean annual temperature and the proportion of species in a flora with untoothed leaf margins. The second approach, known as the Climate-Leaf Analysis Multivariate Program, is based on a modern data set that is multivariate. I argue here that the simpler, univariate approach will give paleotemperature estimates at least as precise as the multivariate method because (1) the temperature signal in the multivariate data set is dominated by the leaf-margin character; (2) the additional characters add minimal statistical precision and in practical use do not appear to improve the quality of the estimate; (3) the predictor samples in the univariate data set contain at least twice as many species as those in the multivariate data set; and (4) the presence of numerous sites in the multivariate data set that are both dry and extremely cold depresses temperature estimates for moist and nonfrigid paleofloras by about 2°C, unless the dry and cold sites are excluded from the predictor set. New data from Western Hemisphere forests are used to test the univariate and multivariate methods and to compare observed vs. predicted error distributions for temperature estimates as a function of species richness. Leaf Margin Analysis provides excellent estimates of mean annual temperature for nine floral samples. Estimated temperatures given by 16 floral subsamples are very close both to actual temperatures and to the estimates from the samples. Temperature estimates based on the multivariate data set for four of the subsamples were generally less accurate than the estimates from Leaf Margin Analysis. Leaf-margin data from 45 transect collections demonstrate that sampling of low-diversity floras at extremely local scales can result in biased leaf-margin percentages because species abundance patterns are uneven. For climate analysis, both modern and fossil floras should be sampled over an area sufficient to minimize this bias and to maximize recovered species richness within a given climate.

Original languageEnglish (US)
Pages (from-to)373-390
Number of pages18
JournalPaleobiology
Volume23
Issue number3
DOIs
StatePublished - Jan 1 1997

Fingerprint

Thermometers
thermometers
Temperature
Climate
leaves
temperature
climate
flora
fossils
species richness
analysis
fossil
leaf morphology
marine record
sampling
species diversity
paleotemperature
Isotopes
multivariate analysis
Datasets

All Science Journal Classification (ASJC) codes

  • Ecology, Evolution, Behavior and Systematics
  • Ecology
  • Agricultural and Biological Sciences(all)
  • Palaeontology

Cite this

@article{60b0e9661088483480636f667d947f30,
title = "When are leaves good thermometers? A new case for Leaf Margin Analysis",
abstract = "Precise estimates of past temperatures are critical for understanding the evolution of organisms and the physical biosphere, and data from continental areas are an indispensable complement to the marine record of stable isotopes. Climate is considered to be a primary selective force on leaf morphology, and two widely used methods exist for estimating past mean annual temperatures from assemblages of fossil leaves. The first approach, Leaf Margin Analysis, is univariate, based on the positive correlation in modern forests between mean annual temperature and the proportion of species in a flora with untoothed leaf margins. The second approach, known as the Climate-Leaf Analysis Multivariate Program, is based on a modern data set that is multivariate. I argue here that the simpler, univariate approach will give paleotemperature estimates at least as precise as the multivariate method because (1) the temperature signal in the multivariate data set is dominated by the leaf-margin character; (2) the additional characters add minimal statistical precision and in practical use do not appear to improve the quality of the estimate; (3) the predictor samples in the univariate data set contain at least twice as many species as those in the multivariate data set; and (4) the presence of numerous sites in the multivariate data set that are both dry and extremely cold depresses temperature estimates for moist and nonfrigid paleofloras by about 2°C, unless the dry and cold sites are excluded from the predictor set. New data from Western Hemisphere forests are used to test the univariate and multivariate methods and to compare observed vs. predicted error distributions for temperature estimates as a function of species richness. Leaf Margin Analysis provides excellent estimates of mean annual temperature for nine floral samples. Estimated temperatures given by 16 floral subsamples are very close both to actual temperatures and to the estimates from the samples. Temperature estimates based on the multivariate data set for four of the subsamples were generally less accurate than the estimates from Leaf Margin Analysis. Leaf-margin data from 45 transect collections demonstrate that sampling of low-diversity floras at extremely local scales can result in biased leaf-margin percentages because species abundance patterns are uneven. For climate analysis, both modern and fossil floras should be sampled over an area sufficient to minimize this bias and to maximize recovered species richness within a given climate.",
author = "Wilf, {Peter Daniel}",
year = "1997",
month = "1",
day = "1",
doi = "10.1666/0094-8373-23.3.373",
language = "English (US)",
volume = "23",
pages = "373--390",
journal = "Paleobiology",
issn = "0094-8373",
publisher = "Paleontological Society",
number = "3",

}

When are leaves good thermometers? A new case for Leaf Margin Analysis. / Wilf, Peter Daniel.

In: Paleobiology, Vol. 23, No. 3, 01.01.1997, p. 373-390.

Research output: Contribution to journalArticle

TY - JOUR

T1 - When are leaves good thermometers? A new case for Leaf Margin Analysis

AU - Wilf, Peter Daniel

PY - 1997/1/1

Y1 - 1997/1/1

N2 - Precise estimates of past temperatures are critical for understanding the evolution of organisms and the physical biosphere, and data from continental areas are an indispensable complement to the marine record of stable isotopes. Climate is considered to be a primary selective force on leaf morphology, and two widely used methods exist for estimating past mean annual temperatures from assemblages of fossil leaves. The first approach, Leaf Margin Analysis, is univariate, based on the positive correlation in modern forests between mean annual temperature and the proportion of species in a flora with untoothed leaf margins. The second approach, known as the Climate-Leaf Analysis Multivariate Program, is based on a modern data set that is multivariate. I argue here that the simpler, univariate approach will give paleotemperature estimates at least as precise as the multivariate method because (1) the temperature signal in the multivariate data set is dominated by the leaf-margin character; (2) the additional characters add minimal statistical precision and in practical use do not appear to improve the quality of the estimate; (3) the predictor samples in the univariate data set contain at least twice as many species as those in the multivariate data set; and (4) the presence of numerous sites in the multivariate data set that are both dry and extremely cold depresses temperature estimates for moist and nonfrigid paleofloras by about 2°C, unless the dry and cold sites are excluded from the predictor set. New data from Western Hemisphere forests are used to test the univariate and multivariate methods and to compare observed vs. predicted error distributions for temperature estimates as a function of species richness. Leaf Margin Analysis provides excellent estimates of mean annual temperature for nine floral samples. Estimated temperatures given by 16 floral subsamples are very close both to actual temperatures and to the estimates from the samples. Temperature estimates based on the multivariate data set for four of the subsamples were generally less accurate than the estimates from Leaf Margin Analysis. Leaf-margin data from 45 transect collections demonstrate that sampling of low-diversity floras at extremely local scales can result in biased leaf-margin percentages because species abundance patterns are uneven. For climate analysis, both modern and fossil floras should be sampled over an area sufficient to minimize this bias and to maximize recovered species richness within a given climate.

AB - Precise estimates of past temperatures are critical for understanding the evolution of organisms and the physical biosphere, and data from continental areas are an indispensable complement to the marine record of stable isotopes. Climate is considered to be a primary selective force on leaf morphology, and two widely used methods exist for estimating past mean annual temperatures from assemblages of fossil leaves. The first approach, Leaf Margin Analysis, is univariate, based on the positive correlation in modern forests between mean annual temperature and the proportion of species in a flora with untoothed leaf margins. The second approach, known as the Climate-Leaf Analysis Multivariate Program, is based on a modern data set that is multivariate. I argue here that the simpler, univariate approach will give paleotemperature estimates at least as precise as the multivariate method because (1) the temperature signal in the multivariate data set is dominated by the leaf-margin character; (2) the additional characters add minimal statistical precision and in practical use do not appear to improve the quality of the estimate; (3) the predictor samples in the univariate data set contain at least twice as many species as those in the multivariate data set; and (4) the presence of numerous sites in the multivariate data set that are both dry and extremely cold depresses temperature estimates for moist and nonfrigid paleofloras by about 2°C, unless the dry and cold sites are excluded from the predictor set. New data from Western Hemisphere forests are used to test the univariate and multivariate methods and to compare observed vs. predicted error distributions for temperature estimates as a function of species richness. Leaf Margin Analysis provides excellent estimates of mean annual temperature for nine floral samples. Estimated temperatures given by 16 floral subsamples are very close both to actual temperatures and to the estimates from the samples. Temperature estimates based on the multivariate data set for four of the subsamples were generally less accurate than the estimates from Leaf Margin Analysis. Leaf-margin data from 45 transect collections demonstrate that sampling of low-diversity floras at extremely local scales can result in biased leaf-margin percentages because species abundance patterns are uneven. For climate analysis, both modern and fossil floras should be sampled over an area sufficient to minimize this bias and to maximize recovered species richness within a given climate.

UR - http://www.scopus.com/inward/record.url?scp=85056225756&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056225756&partnerID=8YFLogxK

U2 - 10.1666/0094-8373-23.3.373

DO - 10.1666/0094-8373-23.3.373

M3 - Article

VL - 23

SP - 373

EP - 390

JO - Paleobiology

JF - Paleobiology

SN - 0094-8373

IS - 3

ER -