Statistical analysis of soil colour and spectroradiometric data for hyperspectral remote sensing of soil properties (example in a southern Italy Mediterranean ecosystem)
Munsell hue, value and chroma of 69 surface soil samples were both visually estimated by four observers under diffuse daylight and computed from laboratory reflectance spectra by applying the CIE 1931 standard method. Significant relationships were found between 'observed' and 'computed' colour components, and between the latter and some soil properties. Using a correspondence analysis, soil colour was shown to be important in differentiating between soil types. From the original spectra, the visible bands of the MIVIS hyperspectral sensor were simulated and related to the colour components through single and multiple regression analyses. The R2 values for hue, value and chroma were 0.58, 0.81 and 0.87 respectively. The results were compared with those obtained using simulated visible Thematic Mapper (TM) bands. For each sample, a curve was fitted to both the MIVIS and TM bands. From these curves, values of colour components were computed and compared with those obtained from the original spectra. Results showed a clear improvement in colour determination. Nevertheless, the complexity and variability of the best fitting curves makes this approach difficult to apply to the images. Remote sensing of soil colour is expected to improve with future launches of higher resolution hyperspectral sensors.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.