ALBERT

Description: Over the past 50 years, the forests of western Washington and Oregon have been extensively harvested for timber. This has resulted in a heterogeneous mosaic of remaining mature forests, clear-cuts, new plantations, and second-growth stands that now occur in areas that formerly were dominated by extensive old-growth forests and younger forests resulting from fire disturbance. Traditionally, determination of seral stage and stand condition have been made using aerial photography and spot field observations, a methodology that is not only time- and resource-intensive, but falls short of providing current information on a regional scale. These limitations may be solved, in part, through the use of multispectral images which can cover large areas at spatial resolutions in the order of tens of meters. The use of multiple images comprising a time series potentially can be used to monitor land use (e.g. cutting and replanting), and to observe natural processes such as regeneration, maturation and phenologic change. These processes are more likely to be spectrally observed in a time series composed of images taken during different seasons over a long period of time. Therefore, for many areas, it may be necessary to use a variety of images taken with different imaging systems. A common framework for interpretation is needed that reduces topographic, atmospheric, instrumental, effects as well as differences in lighting geometry between images. The present state of remote-sensing technology in general use does not realize the full potential of the multispectral data in areas of high topographic relief. For example, the primary method for analyzing images of forested landscapes in the Northwest has been with statistical classifiers (e.g. parallelepiped, nearest-neighbor, maximum likelihood, etc.), often applied to uncalibrated multispectral data. Although this approach has produced useful information from individual images in some areas, landcover classes defined by these techniques typically are not consistent for the same scene imaged under different illumination conditions, especially in the mountainous regions. In addition, it is difficult to correct for atmospheric and instrumental differences between multiple scenes in a time series. In this paper, we present an approach for monitoring forest cutting/regrowth in a semi-mountainous portion of the southern Gifford Pinchot National Forest using a multisensor-time series composed of MSS, TM, and AVIRIS images.

Description: A simple mixing model employing reference endmembers (green vegetation, non-photosynthetic vegetation, soil and shade), and using 180 AVIRIS bands, was used to establish an interpretive framework for a forested area in the Pacific Northwest. A regrowth trend, based on changes in the endmember proportions, was defined for conifers that extends from clearcuts to mature forest, and by implication to old growth. Deciduous species within replanted forest plots caused the fractions to be displaced from the main coniferous regrowth trend and to move toward the green vegetation fraction. The results indicate that the spectral information in AVIRIS can be inverted to estimate approximate stand age and relative proportion of deciduous species in the context of the area studied. Using AVIRIS we measured a 3 to 5 percent increase in woody material in old-growth forest, as distinct from other mature forest. This result is consistent with a predicted increase in NPV in old-growth forest, based on field observations. Previous application of the mixing analysis to a TM image of the same area separated old growth based solely on the shade fraction; however the approach required successful removal of shade introduced by topography. Our new results suggest that with the high spectral resolution and high signal-to-noise of AVIRIS images it may be possible to characterize and map old-growth forests in the Northwest using both the NPV fraction and shade.

Description: An important application of remote sensing is to map and monitor changes over large areas of the land surface. This is particularly significant with the current interest in monitoring vegetation communities. Most of traditional methods for mapping different types of plant communities are based upon statistical classification techniques (i.e., parallel piped, nearest-neighbor, etc.) applied to uncalibrated multispectral data. Classes from these techniques are typically difficult to interpret (particularly to a field ecologist/botanist). Also, classes derived for one image can be very different from those derived from another image of the same area, making interpretation of observed temporal changes nearly impossible. More recently, neural networks have been applied to classification. Neural network classification, based upon spectral matching, is weak in dealing with spectral mixtures (a condition prevalent in images of natural surfaces). Another approach to mapping vegetation communities is based on spectral mixture analysis, which can provide a consistent framework for image interpretation. Roberts et al. (1990) mapped vegetation using the band residuals from a simple mixing model (the same spectral endmembers applied to all image pixels). Sabol et al. (1992b) and Roberts et al. (1992) used different methods to apply the most appropriate spectral endmembers to each image pixel, thereby allowing mapping of vegetation based upon the the different endmember spectra. In this paper, we describe a new approach to classification of vegetation communities based upon the spectra fractions derived from spectral mixture analysis. This approach was applied to three 1992 AVIRIS images of Jasper Ridge, California to observe seasonal changes in surface composition.

Description: The ability to map changes over large surface areas over time is one of the advantages in using remote sensing as a monitoring tool. Temporal changes in the surface may be gradual, making them difficult to detect in the short-term, and because they commonly occur at the subpixel scale, they may be difficult to detect in the long-term as well. Also, subtle changes may be real or merely an artifact of image noise. It is, therefore, necessary to understand the factors that limit the detection of surface materials in evaluating temporal data. The spectral detectability of vegetation and soil in the 1990 July and October Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data of Jasper Ridge, CA was evaluated and compared.

Description: Telescopic and spacecraft spectroscopic and geochemical data have been used to constrain the surface mineralogy of Mars and to yield clues about past and present Mars surface weathering/alteration scenarios. Based primarily on their visible to near-IR reflectance properties, several terrestrial iron-bearing minerals have been either identified on Mars or proposed as Mars spectral analogs. Among these are crystalline hematite, pyroxenes, as well as poorly crystalline materials like nonophase hematite and palagonite. Other iron-bearing minerals include nontronite, magnetite, jarosite, and goethite and the have been proposed as Mars surface constituents based on Viking Lander measurements or geochemical modeling. If present on Mars, these materials likely appear as spectral mixtures at the coarse spatial resolution of remotely sensed data. The detectability of any of these components must be evaluated relative to the other components with which they occur. The primary goal of this study is to determine how much of any given mineral would have to be present for it to be detectable in remotely-sensed data. Here, the detectability of iron-bearing minerals was evaluated using detection threshold analysis (DTA), an analytical technique based on spectral mixture analysis.

Description: The conditions that affect the spectral detection of target materials at the subpixel scale are examined. Two levels of spectral mixture analysis for determining threshold detection limits of target materials in a spectral mixture are presented, the cases where the target is detected as: (1) a component of a spectral mixture (continuum threshold analysis) and (2) residuals (residual threshold analysis). The results of these two analyses are compared under various measurement conditions. The examples illustrate the general approach that can be used for evaluating the spectral detectability of terrestrial and planetary targets at the subpixel scale.