Global warming is one of the biggest threats humanity faces today. Carbon dioxide (CO2) emissions from fossil fuels are warming our planet in a dangerous and irreversible way. Quantitative scientists must be able to quantify exactly and robustly the amount of CO2 emitted by humans to inform and support effective climate action. Although such an exact CO2 account seems obvious and necessary, it is currently done using approaches that lack scientific rigour and fail to properly account for uncertainties.

This PhD research will explore the challenges associated with the reliable quantification of CO2 emissions using the latest cutting-edge uncertainty quantification methods combined with modern digital and generative tools. These methods will take explicit and quantitative account of the randomness, vagueness and limited quality of CO2 data. Current methods for CO2 calculations are deterministic and entirely based on coarse grid conversion factors with little--if at all--mention of the uncertainty that such calculations can incur. This research seeks to bridge this gap, providing a more rigorous framework for CO2 emissions assessment.

Strathclyde researchers are teaming up with Mavarick to revolutionise how CO2 calculations are done, moving towards a more modern and rigorous account of the uncertainties.

Methods that rigorously quantify the uncertainty include Gaussian processes and neural networks for regression and forecasting, imprecise-probability methods to model data quality and to propagate coarse measurements through the calculations.

The use of generative tools to build knowledge representations of the CO2 grid factors will also be explored to tackle ambiguity and classification uncertainty.