Separating Natural Climate Variability and Anthropogenic Contributions in the Observed Global Temperature: A Study from 1950 to 2019.

Supervisors: Carlos Garcia-Soto, Jon Saenz (BEGIK Joint Research Unit IEO-UPV/EHU)
While there is no agreement on the degree to which natural factors influence the presentclimate, it is generally accepted that they can influence global-mean temperature by amplifyingor reducing the temperature increase at di↵erent temporal and spatial scales. Here, we seekto identify natural factors which significantly contribute to global-mean surface temperaturefor the period from 1950 to 2019. A multiple linear regression analysis of global-mean surfacetemperature using 14 natural factors and total anthropogenic forcing as possible explanatoryvariables is used. A procedure which outlines a multistep model selection method is proposed,resulting a model containing the anthropogenic factor, the Atlantic Multi-decadal Oscillation(AMO), El Ni˜no-Southern Oscillation (ENSO), volcanic forcing, the Pacific-North AmericanTeleconnection Pattern (PNA) and the Indian Ocean Basin Mode (IOBM) as explanatory variables.The anthropogenic factor accounts for 91%(±3.5) of the total temperature variability andthe collection of natural factors account for 51% (±19) of the residual temperature variabilitynot explained by the anthropogenic factor. The contributions to global-mean temperature fromeach of these natural factors are quantified and the global-mean temperature data is adjustedby removing estimated contributions from these natural factors. This allows for the calculationof the net anthropogenic global warming trend with reduced noise.Furthermore, we use principal component regression analysis in order to verify the resultsobtained from MLR and to quantify the contributions to global-mean temperature from all 14natural factors. Results from MLR and PCR agree for the most part; the PCR analysis suggeststhat MLR overestimates contributions from ENSO and the anthropogenic factor (suggesting thenet anthropogenic global warming trend needs to be adjusted). The results suggest that themodel from MLR does a better job at explaining the temperature variability. The globalmeantemperature data adjusted using models from PCR was found to have a trend whichis significantly di↵erent from the unadjusted data at a 97.5% confidence level. However, thissignificance should be taken lightly and future work is needed to verify the results in this study.