Urban areas constitute a major source of greenhouse gas emissions, yet they also contain significant areas of vegetation that can take up CO2 via photosynthesis and release CO2 via respiration. However, urban vegetation is seldom accounted for in urban CO2 exchange models because the magnitude of, and the controls on ecological CO 2 fluxes in cities are not well quantified. The eddy covariance method is increasingly used to directly measure CO2 fluxes above urban areas but, importantly, it calculates the net CO2 flux ( FC), including both anthropogenic emissions, and fluxes from vegetation. These ecosystem fluxes are highly spatially variable due to heterogeneity in urban land use types and at the same time, exhibit a strong diurnal and seasonal cycle due to meteorological forcing and the state of the ecosystem. Here, I developed a data based approach from three different perspectives to assess both the temporal and spatial variation in a three year time series of flux measurements, collected from a 150m tall radio broadcast tower in a suburban neighborhood of Minneapolis-Saint Paul, Minnesota, USA: 1) A machine learning framework that accounted for spatial heterogeneity in the flux source area to fill gaps in the observed FC time series; 2) A separation and quantification of the biogenic flux components of FC from dominant anthropogenic CO2 emissions at 30-minute resolution; 3) A spatial investigation into the relationships between flux components and flux source area weighted land cover fractions, in order to up-scale local fluxes from trees and turfgrass lawns near the tower to the larger metropolitan area.

The spatial variation in the flux magnitude when modeling annual sums of FC was up to a factor of two depending on major land use types in different wind directions, including residential and recreational areas. Gross primary production had the largest magnitude of all separated urban FC components both on a monthly and on a diurnal temporal scale during the growing season, and was also the most seasonally dynamic flux (compared to vehicular traffic emissions, natural gas emissions from space heating as well as water heating and cooking, and ecosystem respiration). The modeled biogenic and anthropogenic fluxes were significantly related to source area weighted fractions of green cover and impervious cover, respectively. Finally, I calculated the first estimates of net CO2 exchange from urban trees in residential areas directly based on eddy covariance measurements and scaled them up to the larger metropolitan area throughout the growing seasons of 2007 and 2008. The modeling studies in this work provide estimates of CO2 release and CO2 uptake from urban trees and turfgrass lawns in a suburban neighborhood at half-hourly, daily, monthly and annual levels. New insights on the controlling factors of these biogenic fluxes can lead to improvements in capturing the function of urban greenspace in carbon and climate models at metropolitan, regional, and global scales.