Spatial source attribution of measured urban eddy covariance CO2 fluxes
- First Online:
- Cite this article as:
- Crawford, B. & Christen, A. Theor Appl Climatol (2015) 119: 733. doi:10.1007/s00704-014-1124-0
- 509 Downloads
Interpretation of tower-based eddy covariance (EC) carbon dioxide flux (FC) measurements in urban areas is challenging because of the location bias of EC instruments. This bias results from EC point measurements taken above a complex CO2 source/sink surface that is spatially heterogeneous at scales approaching or exceeding those of the turbulent flux source areas. This makes it difficult to accomplish traditional measurement objectives such as calculating spatially unbiased ecosystem-wide cumulative FC totals or objectively comparing FC during different environmental conditions (e.g., day vs. night or seasonal differences). This study uses a multiyear FC dataset measured over a residential area of Vancouver, BC, Canada from a 30-m flux tower in close proximity to a busy traffic intersection on one side. The FC measurements are analyzed using surface geospatial data and turbulent flux source area models to exploit location bias to develop methods to statistically model individual emissions and uptake processes in terms of environmental controls and surface land cover. The empirical relations between controls and measured FC are used to spatially and temporally downscale individual CO2 emissions/uptake processes that are then used to create high-resolution maps (20 m) and calculate ecosystem-wide FC at temporal resolutions of 30 min to 1 year. At this site, the modeled ecosystem-wide annual net FC total is calculated as 6.42 kg C m−2 year−1 with traffic emissions estimated to account for 68.8 % of the total net emissions. Building sources contribute 27.9 %, respiration from soil and vegetation is 5.5 %, respiration from humans 5.0 %, and photosynthesis offsets are −7.2 % of the annual net total. The statistical models developed here are then tested by direct comparison to independent EC measurements using land cover scalings derived from 30-min source area models. Results are also scaled to ecosystem-averaged land cover to compare results to independent emissions/uptake models.