Quantifying Uncertainty in Climate Sensitivity Metrics Using Large Ensemble Observational Datasets
Abstract
Gridded surface temperature observational datasets are an important line of evidence for estimating climate sensitivity metrics. Climate sensitivity metrics are often derived from comparing historical warming in these observations against climate model simulations, an approach that is known to be sensitive to the choice of observational dataset. With observational products now providing large ensembles that sample their own methodological uncertainty, we use these large ensembles, together with a spatial-temporal model-observation framework, to assess the sensitivity of climate sensitivity metrics to observational uncertainty. We find that the uncertainty within a single dataset is now the dominant source of observational uncertainty, five to six times larger than the typical uncertainty captured when only the median estimate from each data product is considered. The additional datasets from these observational products further allow us to isolate where this uncertainty originates in the construction of a dataset. Uncertainty in how the land record is constructed contributes at least seven times more than uncertainty in the construction of the ocean record for climate sensitivity metrics, and we find that the choice of whether to spatially infill the observational record shifts estimates by about 0.2◦C. Applying the large ensembles to a global, time-mean energy balance framework, we find that uncertainty from a single dataset can produce a difference of about 1.1◦C in the median estimate of equilibrium climate sensitivity.
Related
- Quantifying and attributing CO emissions over Central Asia using TROPOMI and explicit observational uncertainty
- Uncertainty analysis of a machine learning-aided ensemble forecast for coastal flood maps
- Tail Skill and Ensemble-Configuration Sensitivity in Six-Hour Precipitation Postprocessing
- Integrating Observational Datasets and CMIP6 Projections for Multi-Scale Drought Characterization: Evidence from Western Türkiye
- Threshold Sensitivity and Reference Uncertainty in Sentinel-1 Inundation Mapping: The July 2021 Mahad Flood
BibTeX
@misc{sapkota2026quantifying,
title = {{Quantifying Uncertainty in Climate Sensitivity Metrics Using Large Ensemble Observational Datasets}},
author = {Vikrant Sapkota and Chris E Forest},
howpublished = {EarthArXiv},
year = {2026},
month = sep,
doi = {10.31223/x5sz2z},
url = {https://eartharxiv.org/repository/view/14718/},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Quantifying Uncertainty in Climate Sensitivity Metrics Using Large Ensemble Observational Datasets
AU - Vikrant Sapkota
AU - Chris E Forest
PY - 2026
DA - 2026/09/01
PB - EarthArXiv
DO - 10.31223/x5sz2z
UR - https://eartharxiv.org/repository/view/14718/
AB - Gridded surface temperature observational datasets are an important line of evidence for estimating climate sensitivity metrics. Climate sensitivity metrics are often derived from comparing historical warming in these observations against climate model simulations, an approach that is known to be sensitive to the choice of observational dataset. With observational products now providing large ensembles that sample their own methodological uncertainty, we use these large ensembles, together with a spatial-temporal model-observation framework, to assess the sensitivity of climate sensitivity metrics to observational uncertainty. We find that the uncertainty within a single dataset is now the dominant source of observational uncertainty, five to six times larger than the typical uncertainty captured when only the median estimate from each data product is considered. The additional datasets from these observational products further allow us to isolate where this uncertainty originates in the construction of a dataset. Uncertainty in how the land record is constructed contributes at least seven times more than uncertainty in the construction of the ocean record for climate sensitivity metrics, and we find that the choice of whether to spatially infill the observational record shifts estimates by about 0.2◦C. Applying the large ensembles to a global, time-mean energy balance framework, we find that uncertainty from a single dataset can produce a difference of about 1.1◦C in the median estimate of equilibrium climate sensitivity.
ER -