UC San Diego

Climate change mitigation is a multi-faceted problem that requires both technological and behavioral solutions. This dissertation covers three distinct strategic projects for decarbonizing energy systems. The respective strategies are linked by their use of data-informed modeling to represent complex systems.

The first project integrates high-resolution geostationary satellite data from the GOES-16 advanced baseline imager (ABI) for cloud identification and solar forecasting models. The current temporal and spatial resolutions of remote sensing imagers offer immense potential for real-time and highly accessible modeling. The work assesses transfer learning fitness among meteorological stations of varying climates and demonstrates the need for strategic placement and investment in ground stations. Because models are not equally portable, meaning origin training matters to performance at the target destination, having ground station data available at more trainable sites is important for a future in which real time geostationary data is regularly fed into solar forecasting models.

Project two is the work concerning models for clear sky emissivity and transmissivity. The longwave radiation component in climate balance models underpin determinations of climate forcing and impacts of rising carbon dioxide levels in the atmosphere The contributions of this work are the proposed methodology for data preparation and quantification of model sensitivity, the outline of broadband and spectral wideband models for optical depth estimation per atmospheric constituent, and the demonstration of a spectral wideband model use and interpretation.

Project three concerns an immediate soft strategy for decarbonization using the University of California as a proving ground. Significant carbon emissions can be avoided by shifting the fall term so that it ends before Thanksgiving week, thus preventing short-term two-way trips. This shift has numerous benefits for health and emission purposes with strong potential for immediate impact. This carbon accounting study contributes to the pathway toward greater sustainability through behavioral change.

This body of work addresses challenging problems of climate change mitigation through study of machine learning model applicability, physical parameters of radiative transfer, and non-technological (soft) strategies for decarbonization. Each of these efforts contribute to the larger effort to understand, model, and adapt to abrupt climate change.