Effects of Vegetation on Urban Thermal Anisotropy

Urban areas are characterized by rough, three dimensional surfaces which, due to the multitude of surfaces involved, pose significant challenges to the measurement of surface temperature using thermal remote sensing. The intersection of a remote sensor field of view with an urban surface means that only a subset of surfaces are viewed from any one viewing position. Urban surface temperatures are further influenced by sun geometry and patterns of sunlit and shaded surfaces. The impact of the surface and sun geometry makes remotely sensed estimates of surface temperature directionally-dependent. This directional-dependence has been termed the ‘effective thermal anisotropy' of an urban surface since it is not the product of any single urban surface facet, which may be thermally-isotropic, but is instead the result of the three dimensional urban surface viewed by narrow field-of-view remote sensing instruments. This thermal anisotropy limits the usefulness and reliability of the remotely observed temperatures for applications such as studying the urban heat island, estimating the urban energy balance, and evaluating numerical model output.

To date, observations and simulations of urban thermal anisotropy have focused on urban settings with low amounts of vegetative cover to simplify the assessment of building geometry influence on the resultant anisotropy. There has been no study that directly assesses the impact of urban tree canopies on anisotropy. This despite the fact that large fractions of many urban areas have significant tree cover. We hypothesized that tree canopies influence the directional component of surface temperature in two ways. Firstly, in areas of low building plan area and low canyon aspect ratio (i.e. short, widely spaced buildings), the shadows cast by larger vegetated elements may increase the magnitude of urban thermal anisotropy due to tree canopy shadows increasing the contrast between sunlit and shaded surfaces. Secondly, large vegetated elements may shade otherwise sunlit built surfaces which, prior to the introduction of vegetated elements, helped generate thermal anisotropy. This could decrease the observed urban thermal anisotropy by reducing the contrast between sunlit and shaded surface temperatures.

To test these, we use a numerical modelling approach, relying on the surface-sensor-sun urban model (SUM). SUM uses solid angle geometry and view factor analysis to calculate an integrated surface temperature for a simulated surface consisting of an array of patches. Patches representing the surface are assigned a spatial co-ordinate (x, y, and z) with an additional category for holding patch characteristics (e.g. surface type, surface temperature, etc.). For the current research, SUM has been modified to include the ability to incorporate tree canopies. This process uses a gap probability approach whereby the probability of gap within tree canopy envelopes is calculated using a modified Beer-Lambert-Bouguer law. This represents tree canopies as volumes of turbid media and is used to weight the view factor of vegetation to the simulated remote sensor. In order to determine the contribution of both sunlit and shaded leaf elements to the integrated surface temperature (i.e. view factor for sunlit and shaded foliage, separately), it is necessary to calculate the proportion of both as ‘seen' by the remote sensor. This model accounts for the hot-spot effect whereby the proportion of seen, sunlit foliage is greatest when the sun and sensor align (i.e. when the sensor views only sunlit surfaces) and decreasing as they diverge.

The modified SUM is validated using airborne thermal infrared imagery from the Sunset residential area in Vancouver, B.C. and is used to investigate the influence on effective urban thermal anisotropy of three main urban tree canopy parameters: 1) the height of tree canopies relative to building height, 2) the distance of tree canopies from building walls, and 3) the overall fractional coverage of tree canopies. Preliminary results indicate the ability of urban tree canopies to both increase and decrease the effective thermal anisotropy in remotely sensed images. This is largely a function of the temperature and location of tree canopies relative to built components. For example, when tree canopies are situated near building wall facets and are close to the shaded wall temperature, the result is a decrease in effective thermal anisotropy as canopies shade previously sunlit walls reducing the contrast between viewing directions. Sensitivity tests to determine the control of various tree canopy foliage parameters, such as leaf area and leaf angle distribution, on effective thermal anisotropy will also be presented. These results will help us better understand and potentially correct for surface temperature measurement bias resulting from remotely-sensed measurements made from a single viewing position.

The modified SUM model (SUMveg) greatly increases its applicability and therefore presents a viable alternative to the relatively high expense and complex requirements of observational campaigns.