Laserfiche WebLink
<br />00845 <br /> <br />4. Spectral rellectallce from vegetatioll is affected by soil alld litter cover, illumination <br />angle. and shadows. Methods have beell devised to mitigate these effects (Lee alld <br />Marsh, 1995; Garcia-Haro et aI., 1996; Todd and Hoffer, 1998; Blackburn, 1999; Pillder <br />alld McLeod, 1999; Yu et al.. 1999; and Quackenbush et aI., 2000). <br /> <br />5. Vegetatioll classification accuracies greater Ihan 80% have been obtained using <br />remotely sensed data (Butt et aI., 1998; Purevdorj et aI., 1998; Coulter et aI., 2000). Use <br />of seasonal data improves c1assilication for deciduous vegetation (Grignetti et al.. 1997; <br />Mickelson et aI., 1998). Additional and narrower wavelength bands increase <br />classification accuracy (Elvidge and Chen. 1995; May et al.. 1997; Green et aI., 1998). <br />Airborne imagery provides better accuracy than spaceborne imagery due to its higher <br />resolution (Rawlinson et aI., 1999; Zhu et aI., 2000). <br /> <br />All of this research points to the distinct possibility thaI terrestrial vegetation surveys can <br />become more automated, extensive. and less expensive using remotely-sensed data and image- <br />processing algorithms. Therefore, the remote sensing initiative investigated various airborne <br />technologies for mapping Ihe community-level compositions and deriving accurate canopy <br />elevations of habitats. <br /> <br />3.2.1 Vegetation mapping <br /> <br />For vegetation composition at the community level. we evalualed different types and <br />resolulions of image data that were acquired during the remote-sensing initiative (Davis et aI., <br />2002c). The data that were evaluated included II-cm CIR film (July 2000). 30.5-cm CIR film <br />(March 2000), 28-cm CIR film (acquircd during overcast conditions in September 2000), 30.5-cm <br />digital CIR imagery (September 1999), and 100-cm digital, 9-band multispectral (July 2000) <br />image dala. Vegetation texture was derived from each data set and used with the color <br />information in various supervised image classifiers to produce vegetation maps at fIve study areas <br />that were prcviously mapped by ground surveys (Kearsley and Ayers, 2000). The study areas <br />were located al river miles 43.1. 51, 55.5,68.2, and 71.4. The vegetation maps produced using <br />the airborne image data were compared to the ground survey maps to determine the accuracics <br />and relative merits of the different types and resolutions of image data for mapping CRE <br />vegetation communities. The resulting classification maps produced for study area river mile <br />(RM) 68.2 are shown in Figures 11-17. The results of this investigation are summarized in the <br />following items that were extracted from Davis et al. (2002c). <br /> <br />I. The i"'.,',nsic retlectance of vegetation is an important factor in discrimination of the <br />riparian vegetation within the CRE. Thus, digital sensors that record a large dynamic <br />range and maintain radiomctric fidelity provide higher mapping accuracies than <br />photographic film. Although image data acquired under overcast sk.y conditions <br />produced less shadowing within vegetation, the resulting lower retleclance oflhe <br />vegetation reduced the classification accuracies from these data over that obtained <br />from image data acquired under clear sky conditions. The overcast data wcre <br />acquired in Seplember (whcreas our other data were acquired in June-July) when <br />chlorophyll contents of some vegetation specics was lower and when some species <br />were in some stagc of ""leaf off' condition. In addition, it is best to obtain image data <br />near the summer solstice in order to minimize shadows within vegetation, but also <br />shadows cast by the canyon walls. Even at the solstice, there are areas within thc <br />CRE that need to be acquircd within one hour of noon in order to minimize shadows <br />from steep canyon walls (Figures 18 and 19). <br /> <br />15 <br />