Mapping Global Environmental Change Under Current Conditions and Projected Future Climate Scenarios with Machine Learning

Anthropogenic climate change is the greatest threat our world faces. Because the impacts of climate change are linked to location, mapping is a meaningful way to convey results. I map 3 different phenomena and investigate methodological improvements and the associated uncertainty of acid deposition, heat waves, and soil organic carbon. My research on acid deposition updates the 2010 global budget for reactive nitrogen and sulfur components, improving the results of models from the second phase of the United Nation’s Task Force on Hemispheric Transport of Air Pollution (HTAP-II). My analysis is a step towards the World Meteorological Organization’s goal of global products for mapping harmful air pollution. Acid deposition is also relevant to the future climate; one potential response to climate change is stratospheric aerosol injection (SAI), where sulfur dioxide is injected into the stratosphere to block incoming solar radiation. I use outputs from the Geoengineering Model Intercomparison Project (GeoMIP) to track sulfur deposition from SAI through comparison with historical climate and two future Shared Socioeconomic Pathways (SSPs). My research emphasizes the lack of agreement between models and the importance of resolving these conflicts. The most common and recognizable climate change indicators are those related to temperature. However, most studies rely on a single dynamically downscaled model or an ensemble of statistically downscaled models. My work evaluates future heat wave risk in the US with an ensemble of dynamically downscaled models and an ensemble of statistically downscaled models. My results emphasize the importance of state-of-the-science modeling techniques for fine-resolution, domain-specific climate projections. In order to mitigate climate change, decarbonizing many industries will need to be a priority. A requirement of this work is an understanding of baseline soil organic carbon (SOC), before it is modified. My research focuses on SOC in the US from the 1980s to the present day. I incorporate satellite imagery, climatic and land use variables, and apply machine learning methods to produce high resolution, temporally and spatially continuous maps. Overall, my research aims to elucidate the human and environmental costs of climate change and bring clarity to complex, multidimensional data through mapping techniques and thorough analysis.