B1775
Title: Machine learning techniques for bio-accelerated mineral weathering
Authors: Iris Janssens - IDLab (UAntwerp - IMEC) (Belgium) [presenting]
Thomas Servotte - University of Antwerp (Belgium)
Tim Verdonck - KU Leuven and UAntwerpen - imec (Belgium)
Abstract: The goal of staying below the 2-degree Celsius warming limit of the Paris Agreement requires safe and scalable negative emission technologies (NETs). NETs allow CO2 to be actively removed from the atmosphere, thereby offsetting climate change. Therefore, BAM!, a Horizon2020 FET (future emerging technology) funded project, aims to develop such a NET by bringing silicate weathering, a naturally occurring CO2 sequestration process, into a controlled environment and accelerating it through the use of biota. In early 2022, a series of batch experiments began, in which every 8 weeks, 200 batches containing combinations of minerals, organic katter and biota, were irrigated for 8 weeks. The inorganic carbon in the system was measured as a proxy for the carbon dioxide removal. BAM! aims to explore the parameter space as much as possible. However, the number of combinations that can be generated from the input variables is extensive and the number of batches that can be run is limited. Moreover, the measured data are very noisy. Nevertheless, identifying the conditions that favour weathering rates is crucial to optimising the carbon sequestration. Therefore, machine learning is applied to predict the inorganic carbon and investigate the role of biota in the weathering process. Some preliminary results are presented.
