Improved Optimal Fingerprinting Based on Estimating Equations Reaffirms Anthropogenic Effect on Global Warming

Abstract The optimal fingerprinting approach is central to detecting and attributing climate change. It utilizes a regression model with covariates that have measurement errors, linked by a shared covariance matrix with the regression error up to a known scale. The inferences about the regression coefficients are vital for making reliable detection and attribution statements, as well as for quantifying uncertainties in outcomes like attributable warming. Traditionally, this has involved the total least squares (TLS) method, which depends on accurately estimating the covariance matrix of the regression error. However, inaccuracies in this matrix’s estimation can lead to skewed scaling factor estimators and overly optimistic confidence intervals, potentially misrepresenting the accuracy of detection and attribution statements. The recent advent of an estimating equations approach, which offers more efficient point estimation with smaller possible variance and precise uncertainty quantification, prompts a critical reassessment of past climate change detection and attribution analyses. By applying this advanced method to HadCRUT5 observational data and CMIP6 multimodel simulations, our study reevaluates temperature detection and attribution at global and regional levels, strengthens the existing detection and attribution conclusions at the global scale, and provides evidence of the effect of anthropogenic forcings in various regions. Significance Statement Optimal fingerprinting plays an essential role in quantifying human impacts on climate change by facilitating the statistical analysis of estimated scaling factors, which is fundamental for detecting and attributing climate change to external forces. A well-acknowledged challenge in this process is the estimation of the covariance matrix of the regression error, which is critical in correcting the bias due to errors in variables and quantifying the uncertainty of the resulting estimator. The challenge remained unsolved until the recent advancement of an estimating equations approach, which ensures efficient unbiased point estimators and reliable confidence intervals. Using this method, this work reassesses the previous climate change detection and attribution studies with the latest datasets, HadCRUT5 and CMIP6. This reevaluation confirms earlier findings on the global scale and unveils human impacts in various regions.