The key role that atmospheric carbon dioxide plays in climate warming makes it particularly relevant to understand the human and natural processes that increase or decrease carbon in the atmosphere. Contributing fluxes from the ocean, land, and various anthropogenic sources have been estimated by previous research but large uncertainties remain. As the only component directly under our control, the human contribution is critical to understand for policy decisions seeking to mitigate and prevent climate impacts. In this dissertation I have focused on a few aspects of understanding future fossil fuel emissions on both short and long timescales.
I started with the long term, exploring carbon feedbacks on fossil fuel emissions. In my first research chapter I conceptualize and estimate the size of several economic mechanisms that generate a carbon-climate feedback, using the Kaya Identity to separate a net economic feedback into components associated with population, GDP, heating and cooling, and the carbon intensity of energy. In a fossil fuel intensive future scenario, I found that such decreases in economic activity due to warming reduced fossil fuel emissions by 13% this century, lowering atmospheric CO2 by over 100 ppm in 2100. The natural carbon-climate feedback, in contrast, increased atmospheric CO2 over this period by a similar amount, and thus the net effect including both feedbacks was nearly zero. Importantly, although these impacts of climate warming on the economy may offset weakening land and ocean carbon sinks, a loss of economic productivity will have high societal costs, potentially increasing wealth inequity and limiting resources available for effective adaptation.
The uncertainty in my estimation of a potential economic carbon feedback effect is high, however, motivating a need for improved understanding of these feedbacks within more sophisticated models. To that end, in my second study I have extended a framework previously developed for calculating natural carbon feedback parameters to include anthropogenic feedback effects, so it can be used to compare a larger set of carbon feedbacks across models as more human-driven mechanisms are incorporated. I then illustrate some of these calculations using a the model from the previous chapter and a modified version of the Dynamic Integrated Climate-Economy (DICE) model. This work demonstrates a framework that can be applied to evaluate model representation of both anthropogenic and natural feedbacks in integrated assessment models, aiding further model development and improving policy-relevant model outputs.
In my final research chapter I turn to a shorter-term analysis of fossil fuel emissions, exploring the use of autoregression models to make forecasts of emissions in the United States over intervals of a few months to a few years. I focus on freely available and frequently updated predictors including climatic and socioeconomic variables and test several different modeling approaches across all subsets of the predictors. The approach with the most predictive power for out-of-sample forecasts of up to a few months was a vector autoregression (VAR) model, which had a mean absolute percent error of 3.2% for an out-of-sample forecast one month ahead and was able to outperform an existing annual forecast from the EIA by an average of 20%. The model demonstrates the potential of simpler statistical models for short term emissions forecasts and provides a foundation for producing similar global forecasts.
The combined results from these analyses help support improved modeling efforts on short and long timescales of a critical climate driver: carbon dioxide emissions from fossil fuels.