High-Resolution Urban Vegetation Classification for Biospheric CO2 Flux Quantification in Los Angeles, California

Urban areas are known hotspots of fossil fuel carbon dioxide (CO₂) emissions, however, the influence of the biosphere on carbon fluxes is not well understood and represents a source of uncertainty in urban carbon budgets. Los Angeles and the surrounding South Coast Air Basin (SoCAB) is a spatially heterogeneous region, containing highly managed and developed urban regions, as well as unmanaged mountainous and non-urban ecosystems. Current landcover classification maps in the SoCAB region either cover large areas with poor spatial resolution of 30 m or higher, or cover small areas with high spatial resolution. Understanding the influence of biosphere on CO₂ fluxes in Los Angeles is limited by the spatial resolution of these urban landcover classification maps, as well as the ability of current CO₂ flux models to accurately quantify urban biospheric fluxes.

We will present a new method for vegetation classification in a spatially heterogeneous megacity using 2016 National Agriculture Imagery Program (NAIP) imagery at 60 cm resolution. Our approach uses superpixel segmentation for object-based classification and supervised machine learning to differentiate urban landcover types. This method distinguishes between landcover classifications of impervious surface, tree, irrigated turfgrass, chaparral/shrub, non-photosynthetic vegetation, and water. A 30 m fractional urban vegetation classification map developed from the 60 cm imagery will be used to run a version of the Vegetation Photosynthesis and Respiration Model (VPRM). This model will help constrain the magnitude and fluxes of biospheric CO₂ emissions at a sub-daily temporal resolution and resolve uncertainties in carbon ecosystem dynamics of complex urban areas.