Compact Airborne Spectrographic Imager (CASI) used for mapping biophysical parameters of boreal forests

During the Boreal Ecosystem-Atmosphere Study (BOREAS), which took place in Saskatchewan and Manitoba in 1994, the Compact Airborne Spectrographic Imager (CASI) acquired images of boreal forests. In this paper we present results of radiometric and geometric analysis of the CASI data for developing algorithms for retrieving leaf area index (LAI) and crown closure of the boreal forest. The images of over 30 sites, composed of black spruce, jack pine, and aspen stands, were acquired on different days and locations with various solar illumination and view geometries. The geometrical-optical model, named "4-Scale" [Chen and Leblanc, 1997, 1316-1337], was used to correct the images to a common solar zenith angle (35 degrees) and a common view angle (nadir). The 4-Scale model is also used for radiometric analysis based on spectral signatures of leaves and the background (moss, grass, and soil) acquired using various field and laboratory techniques. The red reflectance of all three cover types decreased with increasing LAI as expected. Similar but weaker decreasing trends were found in the near-infrared (NIR) band for conifer stands in contrast to previous findings for cropland and grassland. No significant NIR response to LAI was found for aspen stands. It is shown from 4-Scale that as LAI increases, the crown and ground shadow fractions of conifer forests increase, while the sunlit background fraction decreases and the sunlit crown fraction increases. The large change in the shadow fractions is the major factor controlling the behavior of red and NIR signals. Since boreal forests have abundant green moss and understory as the background, there is only a small difference in optical properties between the overstory and the background. The increases in the shadow fractions with LAI help strengthen the response of optical measurements to changes in LAI, providing a key mechanism for remote information retrieval. The implications of these findings on formulating/selecting vegetation indices and inversion models are discussed in this paper. Relationships of crown closure with CASI measurements are also included in the analysis.