Methodology to integrate hyperspectral Remote Sensor data with GIS for decision support systems:- A case of hail storm damage in Sydney



  1. Spectral library of urban surface materials

    A full range (0.350µm - 2.500µm) spectro-radiometer from ASD was used to extract reflectance values from selected urban materials mainly consisting of roofing materials, pavers, bitumen, metals, corrugated fibro and slate. Roofing materials such as slate, corrugated fibro and terracotta were identified as highly vulnerable to hail storm (see table 1). The urban materials were examined under artificial illumination and the spectra were extracted from each of the sample materials. A spectral library of pure urban materials was created and resampled to the 126 channel HyMap image configuration. The spectral characteristics were viewed and studied using a spectral display software Viewspecpro. The second batch of spectra was generated under natural light. A clear and sunny day was chosen for collecting spectra from surface materials. This exercise was carried out between 12noon and 1 pm in order to reduce the effect from the sun angle. The two urban spectral libraries were named (USAL) urban spectra artificial light and (USNL) urban spectra natural light. Fig 3 A & B shows some of the sample spectra taken in the laboratory by using a spectroradiometer and in the field under natural light.

  2. Image calibration

    In order to directly compare hyperspectral image radiance spectra with field reflectance spectra, the encoded radiance values in the image must be converted to reflectance. The Empirical Line method was preferred over the Flat Field calibration and the Internal Average Relative Reflectance (IARR) techniques. The Flat Field calibration technique is normally applied to spectrally homogeneous areas whereas the IARR calibration technique is used to normalize images to a scene average spectrum. This technique is particularly effective in an area where no ground measurements exist and little is known about the scene. This technique was not chosen for analysis since the study area comprised of hetereogenous urban surface consisting of a variety of features.

    The Empirical Line calibration technique is used to force image data to match selected field reflectance spectra (Roberts et al., 1985; Conel et al., 1987; Kruse et al., 1990). Field reflectance spectra must be acquired from two or more uniform ground target areas which should have widely different brightness and be large enough to recognize in the image. A known dark (concrete) and bright target (sandstone) from the Hymap image were selected and paired with the spectra of the same objects from the reference spectra for calibration. The reference spectra were then forced with the spectra from the image by employing the empirical line technique. A linear regression was calculated between the reference spectra and the image spectra. The regression line was used to predict the surface reflectance spectrum for each pixel from its original image spectrum (ENVI user manual, 1999). The encoded radiance values of the image were thus converted to apparent reflectance by the empirical line technique.




    Figure 3A & B Spectra from Field and Image

  3. Interpretation of Urban roof spectra

    The mapping of urban roof types is one of the key factors in understanding the vulnerability to the damage potential of hail storm hazard in urban environments. Therefore a detailed mapping and interpretation of the urban surface materials and their spectral characteristics were essential. Generally, the total reflectance of a given object across the entire visible region (also termed albedo) is strongly related to the physical condition of the relevant targets (shadowing effects, particle size distribution, refraction index etc.) whereas the spectral peaks are more related to the chemical condition of the sensed target (specific absorption). Several urban related chromophores do provide significant absorption features in the VIS-NIR, such as chlorophyll (at 0.68µm), iron oxides (at 0.50µm, 0.56, 0.88) Hunt et al, 1971). In addition to these specific absorption features the shape of the spectral curve also holds importance in distinguishing urban surface materials (Bhaskar et al, 2000)

    Metal roofs show a high reflectance in the visible range due to the presence of steel and aluminium unlike concrete and terracotta. Terracotta is mainly composed of clay which shows strong absorption peaks in the VIS range and in the far infrared region at around 2160 nm due to the presence of hydroxide ions in Kaolinite, a naturally occurring mineral found in clay based materials. This absorption feature at the 2160 nm is notably absent for other roof types such as metal, slate, and concrete. However, older terracotta shows higher absorption when compared to newer terracotta due to the weathering and fading of the original colour. Concrete tile roofs show high reflectance in the VIS but generally are featureless throughout the spectrum. Roofing slate is a dense natural material that is practically non-absorbent. The colour of slate is determined by its chemical composition. Because these factors vary from region to region, slate is available in a variety of colours. These same factors also influence how susceptible slate is to changing colour upon exposure to the weather. In the study this character of slate made it difficult to detect slate roofs accurately. The presence of slate or similar material such as bitumen, asphalt made it difficult to distinguish slate from the other materials.

  4. Mapping Vulnerable regions by Supervised Classification (Spectral Angle Mapper)

    The spectral angle mapper is an automated method for comparing image spectra to individual spectra or a spectral library (Boardman, unpublished data; CSES, 1992; Kruse et al., 1993). The spectral angle mapper (SAM) classification was used for comparing image spectra to the reference spectra. The SAM algorithm determines the similarity between two spectra by calculating the spectral angle between them as unit vectors in spectral space with dimensionality equal to the number of bands (ENVI user manual, 1999). A classified image was produced by supervised classification in which each pixel was assigned to a class (roofing material, Fig 4).

    The study area is dominated by terracotta tile roofs both old and new varieties. It may be deducted that the predominant land-use is residential, interspersed with some commercial establishments along the major highways and some industrial activities which had metal roofing. Some of the residential dwellings were also made up of concrete structures which appear in a random manner in the classified image. Some educational institutions which fell in the study area had roofing materials consisting of concrete and metal where their compounds were made up of concrete pavers. From the pattern of the land use and distribution of surface materials one can assume that the main threat from hail storm hazard would be directed to residential areas since terracotta roofs are susceptible to hail stones.

    In many instances by a combination of hyperspectral analysis and scientific visualization (photo-interpretation using elements such as shape, colour, pattern, association) the land use may be determined which in turn also assists the spectral analysis of urban materials. Two twin structures belonging to the department of housing were initially recorded as concrete roofs, but the spectral analysis showed that they were indeed some type of metal roof. This was confirmed later to be corrugated iron during the field verification process. Reliable local knowledge was used wherever the access to the roofs was impossible.
Page 4 of 6
| Previous | Next |