Urban green provides a multitude of ecosystem services (e.g. air pollution filtering, reduction of urban heat island effect, runoff reduction, carbon sequestration, biodiversity conservation and recreation), thereby significantly enhancing the quality of life in our cities. Due to these numerous benefits, it is increasingly recognized and incorporated as a key component in sustainable urban planning and management. Urban green however exists in many forms (e.g. street trees, gardens, green roofs), each type exhibiting a different set of ecosystem services with a different magnitude and scale, additionally depending on its properties (e.g. size of a tree) and local context (e.g. proximity to water body). The quantification and subsequent optimization of urban ecosystem services hence requires a highly detailed and up-to-date inventory of all urban green, including information on type, context and both biochemical and structural properties.
Figure 1: There is a clear need for detailed monitoring of urban green in the context of sustainable urban planning and management
Traditionally, data on urban green is collected using labor-intensive sampling campaigns, which produce a scattered and incomplete dataset of the area of interest. The latest developments in remote sensing technology show clear potential in this respect. Hyperspectral sensors on board of airplanes allow us to derive detailed information on vegetation status and health, whereas airborne LiDAR acquisitions show structural details of the Earth’s surface in high resolution. These highly detailed data sources suffer from one major drawback, i.e. their limited spatial coverage. Frequently recurring multi- or even hyperspectral observations from satellite platforms (e.g. Sentinel 2, EnMAP) can help us to build extensive and continuously updated datasets of urban green.
Figure 2: Data types used in UrbanEARS for urban green characterization
Our general aim is to derive detailed and spatially continuous information on the type, context and properties of urban green from remote sensing data, which can then serve as input for urban biophysical modelling (e.g. urban heat and water regulation).
More specifically, we intend to explore the potential of high resolution hyperspectral and LiDAR data to:
Classify functional urban green types, which account for plant type, main properties and context;
Derive chlorophyll content, water content and Leaf Area Index (LAI) of urban trees.
Additionally, we will assess to what extent these objectives can be fulfilled using lower resolution multi- and hyperspectral data. The main challenge when working with low resolution data is to deal with the increased occurrence of spectral mixtures, especially in urban areas. Therefore we will:
Explore and compare techniques to address for undesired background effects, i.e. obtaining the pure vegetation signal from a mixed pixel.
With regard to functional urban green mapping, we intend to combine both spectral and structural variables into an object-based classification algorithm for an optimal distinction between different urban green types. This classification is preceded by a detailed spectral separability analysis, which will reveal the most suitable spectral features to separate between each pair of urban green types.
Figure 3: Three main components of our approach to map functional urban green types
A first step for deriving urban tree properties is to develop an automated segmentation algorithm to delineate individual urban trees. This algorithm will be based on an NDVI map combined with a highly detailed digital surface model, derived from airborne LiDAR data. Next, different signal unmixing approaches will be tested to extract the pure vegetation signals from the (potentially mixed) spectral data. Finally, we will evaluate the performance of different existing vegetation indices to correctly determine biochemical and structural properties based on the pure vegetation signals. In case of structural properties (Leaf Area Index), this approach will additionally be compared to the direct estimation of the same properties from LiDAR data. The extracted tree properties will finally be fused into an objective tree health indicator.
Figure 4: Three main components of our approach to derive urban tree properties
Degerickx, J.; Hermy, M.; Somers, B. (2017). Mapping functional urban green types using hyperspectral remote sensing. IEEE Conference Proceeding. Joint Urban Remote Sensing Event (JURSE), Dubai, March 2017.
Degerickx, J, Hermy, M., Somers, B. Assessment of urban tree helalth using airborne remote sensing technology. Ecology Across Borders Conference, December 2017, Ghent, Belgium.
Degerickx, J., Roberts, D.A., McFadden, J.P., Hermy, M., Somers, M. (2018). Urban tree health assessment using airborne hyperspectral and LiDAR imagery. International Journal of Applied Earth Observation and Geoinformation, 73, 26-38.
Wetherley, E.B., McFadden, J.P., Roberts, D.A. (2018). Megacity-scale analysis of urban vegetation temperatures. Remote Sensing of Environment, 213, 18-33.