Quantifying uncertainty of past pCO2 determined from changes in C3 plant carbon isotope fractionation

Ying Cui, Brian A. Schubert

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

Knowledge of the past concentrations of atmospheric CO2 level (pCO2) is critical to understanding climate sensitivity to changing pCO2. Towards this, a new proxy for pCO2 has been developed based on changes in carbon isotope fractionation (δ13C) in C3 land plants. The accuracy of this approach has been validated against ice-core pCO2 records, suggesting the potential to apply this proxy to other geological periods; however, no thorough uncertainty assessment of the proxy has been conducted. Here, we first analyze the uncertainty in the model-curve fit through the experimental data using a bootstrap approach. Then, errors of the five input parameters for the proxy are evaluated using sensitivity analysis; these include the carbon isotope composition of atmospheric CO213CCO2) and that of the plant material (δ13Corg) for two time periods, a reference time (t=0) and the time period of interest (t), and the value of pCO2 at time t=0. We then propagated the errors on the reconstructed pCO2 using a Monte Carlo random sampling approach that combined the uncertainties of the curve fitting and the five inputs for a scenario in which the reference time was the Holocene with a target period for the reconstructed pCO2 during the Cenozoic. We find that the error in the reconstructed pCO2(t) increases with increasing pCO2(t), yet remains <122% (positive error) and <40% (negative error) for pCO2(t)<1000ppmv. The error assessment suggests that it can be used with confidence for much of the Cenozoic and perhaps the majority of the last 400 million years, which is characterized by pCO2 levels generally less than 1000 ppmv. Towards this, an application of this uncertainty analysis is presented for the Paleogene (52-63Ma) using published data. The resulting pCO2(t) levels calculated using this method average 470 +288/-147ppmv (1σ, n=75), and overlap with previous pCO2(t) estimates determined for this time period using stomata, liverwort, and paleosol proxies. The analysis presented here assumes that the paleoenvironment in which the plants grew is unknown and is determined to be the largest source of error in the reconstructed pCO2(t) levels; errors in pCO2(t) could be reduced provided independent determination of the paleoenvironmental conditions at the fossil site.

Original languageEnglish
Pages (from-to)127-138
Number of pages12
JournalGeochimica et Cosmochimica Acta
Volume172
DOIs
StatePublished - 1 Jan 2016

Fingerprint

Carbon Isotopes
C3 plant
Fractionation
carbon isotope
fractionation
liverwort
stomata
uncertainty analysis
paleoenvironment
ice core
paleosol
Paleogene
Uncertainty analysis
sensitivity analysis
Uncertainty
Ice
Curve fitting
Holocene
fossil
Sensitivity analysis

Cite this

@article{d441b2a09030460b86b06cf47d771dfc,
title = "Quantifying uncertainty of past pCO2 determined from changes in C3 plant carbon isotope fractionation",
abstract = "Knowledge of the past concentrations of atmospheric CO2 level (pCO2) is critical to understanding climate sensitivity to changing pCO2. Towards this, a new proxy for pCO2 has been developed based on changes in carbon isotope fractionation (δ13C) in C3 land plants. The accuracy of this approach has been validated against ice-core pCO2 records, suggesting the potential to apply this proxy to other geological periods; however, no thorough uncertainty assessment of the proxy has been conducted. Here, we first analyze the uncertainty in the model-curve fit through the experimental data using a bootstrap approach. Then, errors of the five input parameters for the proxy are evaluated using sensitivity analysis; these include the carbon isotope composition of atmospheric CO2 (δ13CCO2) and that of the plant material (δ13Corg) for two time periods, a reference time (t=0) and the time period of interest (t), and the value of pCO2 at time t=0. We then propagated the errors on the reconstructed pCO2 using a Monte Carlo random sampling approach that combined the uncertainties of the curve fitting and the five inputs for a scenario in which the reference time was the Holocene with a target period for the reconstructed pCO2 during the Cenozoic. We find that the error in the reconstructed pCO2(t) increases with increasing pCO2(t), yet remains <122{\%} (positive error) and <40{\%} (negative error) for pCO2(t)<1000ppmv. The error assessment suggests that it can be used with confidence for much of the Cenozoic and perhaps the majority of the last 400 million years, which is characterized by pCO2 levels generally less than 1000 ppmv. Towards this, an application of this uncertainty analysis is presented for the Paleogene (52-63Ma) using published data. The resulting pCO2(t) levels calculated using this method average 470 +288/-147ppmv (1σ, n=75), and overlap with previous pCO2(t) estimates determined for this time period using stomata, liverwort, and paleosol proxies. The analysis presented here assumes that the paleoenvironment in which the plants grew is unknown and is determined to be the largest source of error in the reconstructed pCO2(t) levels; errors in pCO2(t) could be reduced provided independent determination of the paleoenvironmental conditions at the fossil site.",
author = "Ying Cui and Schubert, {Brian A.}",
year = "2016",
month = "1",
day = "1",
doi = "10.1016/j.gca.2015.09.032",
language = "English",
volume = "172",
pages = "127--138",
journal = "Geochimica et Cosmochimica Acta",
issn = "0016-7037",
publisher = "Elsevier Ltd",

}

Quantifying uncertainty of past pCO2 determined from changes in C3 plant carbon isotope fractionation. / Cui, Ying; Schubert, Brian A.

In: Geochimica et Cosmochimica Acta, Vol. 172, 01.01.2016, p. 127-138.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Quantifying uncertainty of past pCO2 determined from changes in C3 plant carbon isotope fractionation

AU - Cui, Ying

AU - Schubert, Brian A.

PY - 2016/1/1

Y1 - 2016/1/1

N2 - Knowledge of the past concentrations of atmospheric CO2 level (pCO2) is critical to understanding climate sensitivity to changing pCO2. Towards this, a new proxy for pCO2 has been developed based on changes in carbon isotope fractionation (δ13C) in C3 land plants. The accuracy of this approach has been validated against ice-core pCO2 records, suggesting the potential to apply this proxy to other geological periods; however, no thorough uncertainty assessment of the proxy has been conducted. Here, we first analyze the uncertainty in the model-curve fit through the experimental data using a bootstrap approach. Then, errors of the five input parameters for the proxy are evaluated using sensitivity analysis; these include the carbon isotope composition of atmospheric CO2 (δ13CCO2) and that of the plant material (δ13Corg) for two time periods, a reference time (t=0) and the time period of interest (t), and the value of pCO2 at time t=0. We then propagated the errors on the reconstructed pCO2 using a Monte Carlo random sampling approach that combined the uncertainties of the curve fitting and the five inputs for a scenario in which the reference time was the Holocene with a target period for the reconstructed pCO2 during the Cenozoic. We find that the error in the reconstructed pCO2(t) increases with increasing pCO2(t), yet remains <122% (positive error) and <40% (negative error) for pCO2(t)<1000ppmv. The error assessment suggests that it can be used with confidence for much of the Cenozoic and perhaps the majority of the last 400 million years, which is characterized by pCO2 levels generally less than 1000 ppmv. Towards this, an application of this uncertainty analysis is presented for the Paleogene (52-63Ma) using published data. The resulting pCO2(t) levels calculated using this method average 470 +288/-147ppmv (1σ, n=75), and overlap with previous pCO2(t) estimates determined for this time period using stomata, liverwort, and paleosol proxies. The analysis presented here assumes that the paleoenvironment in which the plants grew is unknown and is determined to be the largest source of error in the reconstructed pCO2(t) levels; errors in pCO2(t) could be reduced provided independent determination of the paleoenvironmental conditions at the fossil site.

AB - Knowledge of the past concentrations of atmospheric CO2 level (pCO2) is critical to understanding climate sensitivity to changing pCO2. Towards this, a new proxy for pCO2 has been developed based on changes in carbon isotope fractionation (δ13C) in C3 land plants. The accuracy of this approach has been validated against ice-core pCO2 records, suggesting the potential to apply this proxy to other geological periods; however, no thorough uncertainty assessment of the proxy has been conducted. Here, we first analyze the uncertainty in the model-curve fit through the experimental data using a bootstrap approach. Then, errors of the five input parameters for the proxy are evaluated using sensitivity analysis; these include the carbon isotope composition of atmospheric CO2 (δ13CCO2) and that of the plant material (δ13Corg) for two time periods, a reference time (t=0) and the time period of interest (t), and the value of pCO2 at time t=0. We then propagated the errors on the reconstructed pCO2 using a Monte Carlo random sampling approach that combined the uncertainties of the curve fitting and the five inputs for a scenario in which the reference time was the Holocene with a target period for the reconstructed pCO2 during the Cenozoic. We find that the error in the reconstructed pCO2(t) increases with increasing pCO2(t), yet remains <122% (positive error) and <40% (negative error) for pCO2(t)<1000ppmv. The error assessment suggests that it can be used with confidence for much of the Cenozoic and perhaps the majority of the last 400 million years, which is characterized by pCO2 levels generally less than 1000 ppmv. Towards this, an application of this uncertainty analysis is presented for the Paleogene (52-63Ma) using published data. The resulting pCO2(t) levels calculated using this method average 470 +288/-147ppmv (1σ, n=75), and overlap with previous pCO2(t) estimates determined for this time period using stomata, liverwort, and paleosol proxies. The analysis presented here assumes that the paleoenvironment in which the plants grew is unknown and is determined to be the largest source of error in the reconstructed pCO2(t) levels; errors in pCO2(t) could be reduced provided independent determination of the paleoenvironmental conditions at the fossil site.

UR - http://www.scopus.com/inward/record.url?scp=84944810913&partnerID=8YFLogxK

U2 - 10.1016/j.gca.2015.09.032

DO - 10.1016/j.gca.2015.09.032

M3 - Article

AN - SCOPUS:84944810913

VL - 172

SP - 127

EP - 138

JO - Geochimica et Cosmochimica Acta

JF - Geochimica et Cosmochimica Acta

SN - 0016-7037

ER -