INTRODUCTION


Study Rationale


The Working Draft of the Chicago Wilderness Biodiversity Recovery Plan calls for creation of large complexes and mosaics of several different natural community types (forest, savanna, prairie and wetlands). It also recommends activities to improve the scientific basis for ecological monitoring and management of the recovery and status of natural communities in the region. The use of hyperspectral scanning (HSS) remote sensing offers the potential of being a cost effective method of monitoring large-scale community restoration efforts in the Chicago region. Data recently collected as part of an intensive study to characterize dolomite prairie communities in the lower Des Plaines River valley (Sluis and Tandarich, 2000) provided an opportunity to 1) test the use of HSS in mapping native /natural sites on and near Midewin National Tallgrass Prairie (Midewin) and to 2) develop and verify this approach as a monitoring tool. This approach has the potential for significant savings in time and money for monitoring at several points during the restoration effort and for long-term monitoring and management of natural communities.


The characterization of vegetation and soils of reference deep and shallow soil grassland -- the latter type locally called dolomite prairie -- sites was recently completed under a Chicago Wilderness grant (Sluis and Tandarich, 2000). Ten grassland sites within the Des Plaines and Kankakee River watersheds within and adjacent to Midewin were studied. Recent wetland delineation and ecological surveys of Midewin revealed less than 100 acres of dolomite prairie. According to the Natural Resources Conservation Service (NRCS) soil map, there are approximately 2000 acres of soils on Midewin which may have supported some type of dolomite prairie in the past and which could be successfully restored to this uncommon prairie type. However, a restoration of a large areal scope as Midewin requires cost-effective monitoring techniques. Ground methods only would be both time-consuming and costly.


This study used an aerial component incorporating the new hyperspectral scanning (HSS) remote sensing imagery to characterize grassland environments on Midewin. This is a necessary step prior to the ability to monitor ongoing prairie restoration efforts. This method, which has been proven to work effectively with agronomic crop plants, could significantly reduce ground time and enable a synoptic, yet detailed view of restoration areas. Ultimately, this method could be expanded to large-scale restoration efforts at Midewin, and elsewhere in the Chicago region.


Remote Sensing: Advantages of the Hyperspectral System


Spectral resolution is a problem when attempting to map native plant groupings. Color infrared (CIR) film is emulsion-based and has the attendant problems of exposure, processing, enlargement and grain. Multispectral scanning (MSS), being digitally based, has fewer problems than film. However, both CIR film and MSS images show a blending of reflectance, called spectral response, within broad spectral bands covering a limited portion of the electromagnetic spectrum. This means that the images contain the standard wavelength band separations, blue (0.4-0.5 µm), green (0.5-0.6 µm), red (0.6-0.7 µm), and near infrared (0.7-1.3 µm) (µm or micrometer is a unit of wavelength of 0.000001 meter) (Figure 1). Due to the digital nature of MSS, these bands can be viewed and analyzed uniquely. While these bands may be useful for broad-scale vegetation mapping, such as trees versus grass, there is a problem when trying to discriminate between many specific plant species. However, it is known that the spectral response in the blue, green and red regions is sensitive to the chlorophyll content, and the response in the near infrared region is sensitive to the plant and leaf structure (Figure 2). The advantage of HSS is its ability to break the spectral bands into smaller, more discrete units to better capture these unique responses.


The imagery scale of CIR and MSS is an additional problem. The enlargement of CIR film results in "graininess" (the silver grains coating the film surface becoming visible) that inhibits pattern recognition. While this is less of a problem for the digital imagery, the most common and readily available spacecraft MSS spatial resolution is 10-30 meters. This means that the smallest area that can be scanned on the ground and resolved by the sensor on a digital image is a square area 10-30 meters on a side that is 0.025 to 0.25 acres. This has made mapping complex native plant community groupings very difficult.


Within the last 15 years, a remote sensing system has been developed that allows the spectrum to be separated into 120 to 240 discrete bands, depending upon the sensor. This system is called hyperspectral (literally, beyond the normal spectrum) remote sensing (HSS). The sensor in our study, the ITD - Spectral Visions Sensor, has 120 bands (Appendix B). This gives spectral bandwidths of 0.03-0.05 µm that allows many discrete spectral response separations. In addition, the spatial resolution of this sensor produces l m picture elements (pixels). When used at this large-scale resolution, HSS enables discrimination of small areas of vegetation types based on their unique spectral responses not possible with a multispectral scanning (MSS) sensor such as carried on the Landsat 7 satellite.