The world’s cities account for up to 70 % of global carbon dioxide (CO2) emissions, while covering less than 2 % of the Earth’s surface. Achieving global goals of keeping temperature increases below 2° requires dramatic reductions in emissions. In keeping with this goal, cities around the world are implementing strategies to reduce carbon dioxide emissions. To support this effort, observations and analyses that assess attribution of emission reductions to specific mitigation strategies are needed. However, monitoring and attributing carbon dioxide emissions in cities are challenging since numerous emission sources are densely presented in cities with complex topography and turbulent mixing.In this dissertation, I present a novel approach to understanding urban carbon dioxide and to attribute emissions to specific source sectors using a near-surface, high-density urban monitoring network. The Berkeley Environmental Air-quality and CO2 Observation Network (BEACO2N) includes ~70 nodes in the San Francisco Bay Area distributed at ~2 km horizontal spacing. I show that the relationship between CO2 concentration and highway traffic flow is coherent throughout the network. Using a Gaussian plume model to represent the dispersion from the highways, I show that the observations constrain the decrease in emission rate per vehicle from 2017 to 2019. Increased fuel efficiency and electrification of the vehicle fleet are among the primary tools in California’s greenhouse gas reduction plan and this assessment suggest these plans are on track. Second, I leverage the Gaussian plume model to determine biogenic uptake of CO2 in the region. I find promising estimates of biogenic emissions that is comparable to the daily and seasonal estimates based on SIF Finally, I describe the implementation and evaluation of other trace gas sensors (O3, CO, NO, and NO2) for source attribution. I demonstrate the use of the relationship between trace gases that are co-emitted from combustion to characterize various emission sources.