Application of Remote Sensing to Biological and Environmental Science.

1. What is Remote Sensing?

For the purpose of this short introduction to remote sensing, I will limit my discussion to two optical sensors; namely the photographic camera and the multispectral scanner. These can be mounted on airborne or space borne platform. The optical instruments of our interest here, make use of atmospheric windows in the range of 400-2500 nm, 3-5 mm, and 8-14 mm of the electromagnetic spectrum.

Go to bottom

1.1 Photographic Camera

The most common aerial cameras in use today, are single lens frame cameras. They are used for obtaining aerial photographs for remote sensing in general, as well as for photogrammetric mapping purposes. The single lens frame cameras employ a low distortion lens, between-the-lens shutter, and are designed to provide extremely high geometric image quality with accuracies approaching 1/10000 of the flying height (photography taken at 1000 meters will accurately locate an object within 10 cm of its position). The film is advanced automatically, based on the scene shooting frequency pre-set on the intervalometer. Rolls of film up to 120 m can be loaded into the magazine designed for the film format size of 240 mm (the frame size 230 mm x 230 mm). The  magazine is also equipped with a film flattening mechanism. Lenses with focal length 90 mm, 152 mm,  210 mm and 300 mm may be used. Commonly, lenses with focal length of 152 mm are used.

In order to take advantage of wavelength specific radiance emanating from various earth features, multilens frame cameras may be used to obtain several discrete photographs taken simultaneously from the same vantage point. Every lens of the camera could be equipped with a different filter combination.

Strip cameras record images by moving film past a fixed slit in the focal plane as the camera moves forward. The shutter for a strip camera remains open continually while the picture is made. The film advances in the speed proportional  to the aircraft ground speed.

With panoramic cameras, ground areas are covered by either rotating the camera lens or rotating a prism in front of the lens. The terrain is scanned from side to side, transverse to the direction of flight. The film is exposed along a curved surface located at the focal distance from the rotating lens assembly, and angular coverage of the camera can extend from horizon to horizon. NASA developed this camera mainly for defense purposes.

For many remote sensing applications, very high geometric image quality is not required. If one is more interested in monitoring changes of spectral reflectance to identify certain types of crops or other types of vegetation, or mapping algae bloom or an extent of water pollution, 35 mm or 70 mm cameras have been used successfully (Adams et al., 1977, Lillesand and Kiefer, 1987). Clegg and Scherz (1975) compared large format (9 in) frame camera, 70 mm and 35 mm systems for resolution, airphoto interpretation, and metric accuracy. They found the smaller formats to produce imagery of comparably good resolution for altitutes below 3000' AGL. Using a wide-angle 24 mm lens a 35 mm camera photo covering an area of  2223 x 1525 meters displayed metric accuracy of points within 2.5 m of identical control points on the 9 inch format photo. For vegetation airphoto interpretation, the smaller formats were actually prefered by a number of interpreters. The investigators concluded that medium and small format (70 mm and 35 mm) cameras can be used for environmental mapping as effectively as a large format (230 mm) system at a fraction of the cost.

Go to top / bottom

1.2 Photographic Films

The black and white film consists of a light sensitive photographic emulsion coated onto a plastic (polyester film) base. The emulsion consists of a thin layer of light sensitive silver halide crystals, or grains, held in place by a solidified gelatin. When exposed to light, the silver halide crystals within the gelatin undergo a photochemical reaction forming an invisible latent image in the form of silver molecules. Upon treatment with suitable chemical agents, the exposed silver salts are reduced to silver grains appearing black, and forming a visible image. The result is a record of the camera's view in which the film areas struck by the most light are darkened by metallic silver, while the areas struck by no light remain transparent due to lack of silver. The intermediate areas have varying amounts of silver creating shades of gray depending on the amount of light striking the film emulsion.

The emulsion can be made sensitive to only UV and blue light, UV, blue light and green light (orthochromatic film), or UV, blue, green and red light (panchromatic film). Most untreated silver salt emulsions are only sensitive to UV and shorter wavelengths of the spectrum. By treatment of film during manufacture with special dyes, it is possible to extend the emulsion's sensitivity to the near infrared (0.7 - 1.2 microns) region of the spectrum. The result is a black and white infrared film.

The film exposure at any point of the photograph is directly related to the reflectance of the object imaged. Theoretically, the film exposure varies linearly with the object reflectance. In practice however, the relationship between the radiance entering the camera lens and that recorded by the film depends on the particular film characteristic curve. Characteristic curves are different for different film types, for different manufacturing batches and even for films of the same batch. Manufacturing, handling, storage, and processing all affect the film's characteristic response. The following figures show typical components of a black and white negative film characteristic curve (A), and other important film characteristics such as density resolution, radiometric resolution and exposure latitude (B). 
 

Typical negative B&W film characteristic curve (A). Two films with different characteristic curves (B). (Log film exposure "H" on the X axis vs. log film density "D" on the Y axis)

The film characteristic curve is expressed as the ratio of log D / log H, where log D = logarithm of film density, and 
log H = logarithm of film exposure. Notice the three divisions to the curve. First the density slowly increases with the exposure at an increasing rate (toe). This is followed by a  relatively linear increase in density (straight line portion). Finally, the maximum density is attained through a sinusoidal plateau (shoulder). The slope of density increase in the straight line portion of the film characteristic curve is the gamma (g)of the film. Gamma is given as: g = DD/Dlog H.It is an important determinant of the contrast of the film. For most aerial films, under manufacturers` recommended processing conditions and corresponding speeds quoted for the films give gammas in the neighbourhood of two. For medium- and high-altitude aerial photography, a gamma of two or three is usually desirable. Gamma is a function of not only the film emulsion, but also of the film development condition. It can be varied by changing developer, development time and/or processing temperature. The next important characteristic of film is its speed. It comprises the level of exposure to which film would respond. Graphically, it is represented by the horizontal position of the film characteristic curve. In the figure (B) above, the curve on the left is that of a "fast" film (it accomodates lower exposure limits; i.e. it is more sensitive). On the other hand, the curve on the right belongs to a "slow" film (less sensitive). "Fast" films are characterized by larger film grains, thus tend to have reduced spatial resolution as opposed to less sensitive films. On the same curve for the "slow" film, notice its wider exposure latitude. It expresses the log H (level of exposure) range which would yield an acceptable image (wider range of exposure would give an image with good density resolution; i.e. discrimination between different features would be better over wider range of exposure). The radiometric resolution is inversely proportional to contrast, so for a given density resolvability, a higher contrast film is able to resolve smaller difference in exposure.

Thus in general, the slower the speed of the film, the higher are the resolving power and the gamma, while the granularity and the latitude are less. Low contrast films offer greater radiometric range (exposure latitude) at the expense of radiometric resolution.  High contrast films offer a smaller exposure range, but improved radiometric resolution (Lillesand and Kiefer, 1987).

Color film is an advancement over the black and white film brought about by addition of two more light sensitive emulsion layers. Like on the black and white film, all emulsion layers consist of silver salts. However, the layers are treated during the manufacture so that they are made sensitive to separate colours of the visible or near infrared spectrum. The top layer in most colour films is sensitive to blue light. The second layer is responsive to green and blue light, and the bottom layer to red and blue. A blue blocking filter is inserted between the top layer and the other two, to prevent the two bottom layers to be exposed by the blue light. This results in three emulsions, each sensitive to blue, green, and red light respectively.

To develop the film as a negative, it is initially put into the colour developer, where the dyes are formed as the exposed silver halide is developed. The silver and the silver halide are then removed leaving the dye image. After processing, the blue sensitive layer contains yellow dye, the green sensitive layer contains magenta dye, and the red sensitive layer contains cyan dye. The amount of dye present in each layer is inversely proportional to the amount of  its corresponding primary color present in the original scene photographed. When viewed in composite, the dye layers produce visual sensation of the color in the original scene. 
 
 

Sensitivity curves - Kodak Aerochrome Infrared Film 2443, and Kodak Aerocolor Negative Film 2445

The final film processing depends on whether the film is a negative, or reversal film. Negative film produces image in negative colors or shades of gray and requires print development to obtain the final product. In reversal film, whether color or black and white, the final color tone of the transparency is matched to the original color of the scene photographed.

False-color film

The relationship between the spectral sensitivity of  the color film layer does not have to be the same as it was described  above for films with normal-color rendition. If different from the above, the result is an infrared (IR) film with a false-color rendition. In infrared sensitive color films, one of the layers is made sensitive to the infrared spectral region, while the other layers keep their sensitivities in the visual spectral region. These films may be made for either negative or positive mode of processing.

From the above discussion about the color film, it may be remembered that normal color film is sensitive to blue, green and red spectral regions. associated with these sensitivities are the yellow, magenta, and cyan dyes respectively, which after processing combine to produce the colors blue, green, and red which closely match those of the original scene photographed. With the color infrared sensitive film, the individual layers are sensitive to green,  red and infrared radiation. Again, the same three dyes - yellow, magenta and cyan are associated with the image formation, but their sensitivities have been shifted toward longer wavelenghts so that the yellow, magenta, and cyan dyes produce green, red, and infrared represented image (Doyle F.J et al., 1983). 

With both, color and infrared color films, since each layer is sensitive only in its own wavelength, our data is recorded in three separate bands. It is possible to break down electronically the original image into three separate bands approximately 100 nm wide. A  panchromatic color film for example, would generate three bands: red (600-700nm), green (500-600nm), and blue (400-500nm). Thus here we have a simple wide-band multispectral sensor.

Go to top / bottom

1.3 Non-photographic Cameras and Multispectral Scanners

Solid-State Array Cameras use one- or two-dimensional detector arrays of charge coupled device (CCD) for image data acquisition. A CCD is a microelectronic silicon chip, a solid-state sensor that detects light. When light strikes the CCD's silicon surface, electronic charges are produced with the magnitude of the charge being proportional to the light intensity and exposure time. Every single microelectric silicon chip within the array produces the smallest unit of the resulting image - the pixel. (Lillesand and Kiefer, 1987).  The digital camera resolution depends on the number of light sensitive detectors (pixels) in each array, the size of the array, the lens used, as well as on the camera's image processing. The resolution of today's high-end 35 mm type DSLR cameras with two-dimensional array sizes of 14 Mega pixels or more and the sensor size of 36 x 24 mm is approaching that of a 35 mm film camera. However, few medium format digital cameras can approach the resolution of the medium format (70 mm) film. As of 2008 the 70 mm Hasselblad camera with Phase One digital camera back had a two-dimensional array with 60.5 Mega pixels (8984 x 6732) and an effective sensor size of 53.9 x 40.4 mm. 

In remote sensing applications, large format digital frame cameras are getting on the market as well. One example is a Vexcel UltraCamX aerial digital camera with a two-dimensional CCD of 216 MP (14430 x 9400) pixels in panchromatic mode and 4810 x 3140 pixels in RGB & NIR (400nm - 1000nm) mode. The ground resolution then depends on the size of the CCD, as well as the altitude of the sensor. For example the high-resolution-visible (HRV) systems on French SPOT-2 (Système probatoire d'observation de la terre) satellite operating in black and white (panchromatic) mode has ground resolution of 10 meters, and the ground resolution of 20 meters when operating in colour infrared (multispectral) mode. 

Due to the wavelength dependency of reflectance from various earth features, the need for narrow band multispectral sensors has been recognized from the early days of remote sensing. Therefore Multispectral scanners (MSS) were developed. They sense bands ranging from the UV, through visible to near-IR, mid-IR and thermal IR portions of the spectrum. MSS utilize electronic detectors and are designed to sense energy in a number of narrow spectral bands simultaneously. They are equipped with a scanning mirror and optics which directs the incoming energy to be separated into several spectral components that are sensed independently. A dichronic grating is used to separate the non-thermal wavelengths from the thermal wavelengths. The non-thermal wavelength component is directed from the grating through a prism that splits the energy into a continuum of UV, visible and IR wavelengths. By placing an array of detectors at the proper geometric position behind the grating and the prism, the incoming beam can be separated and measured independently in a multiple of narrow bands (Lillesand and Kiefer, 1987).

In general, digital imagery collected in panchromatic mode is of much higher resolution than the imagery obtained in multispectral mode. For this reason techniques have been developed to fuse co-georegistered high spatial resolution panchromatic images with a set of coarse (low) spatial resolution multispectral (colour) images to obtain a fine (high) spatial resolution colour images. Thus, the so called Pan-sharpening is a concept of combining multiple images into composite products, through which more information than that of individual input images can be revealed.

Go to top / bottom

1.4 Electromagnetic Energy in Remote Sensing

      An overview of the electromagnetic spectrum.

Some portion of solar radiation passing through the atmosphere does not reach the earth's surface, and is effectively absorbed by atmospheric constituents such as water vapor, carbon dioxide and ozone. The wavelengths of sun's radiation which are particularly transmissive are used for remote sensing and are referred to as atmospheric windows.

The fate of radiation incident upon the earth surface is well expressed by the following energy balance equation:
 

E_I(l) = Incident Energy at wavelength (l), E_R(l) = Energy Reflected, E_A(l) = Energy Absorbed, E_T(l) = Energy Transmitted

For most surface materials, a portion of the incident radiation will be reflected and scattered back away from Earth's surface, a portion will be absorbed and another portion may be transmitted through the material. Some absorbed radiation may be re-emitted at wavelengths longer than those absorbed (Wickland E.D., 1991). The proportions of energy reflected, absorbed, and transmitted will vary for different earth features, depending on their material type and condition. The wavelength dependency means that even within a given feature type, the proportion of reflected, absorbed, and transmitted energy will vary at different wavelengths. Therefore, two features may be indistinguishable in one spectral range and be very different in another wavelength band (Lillesand and Kiefer, 1987).

Out of  the energy reflected from a particular object, only some actually reaches the sensor. The remaining energy may be scattered throughout the atmosphere. Alternatively, some scattered energy may contribute to the background noise signal the sensors receive as well. Elevated background noise/signal ratios are common with sensors located at higher altitudes, or those receiving signal passing through atmosphere with higher concentration of energy scattering particles. Large differences in scattering can be observed between heavily humid and dust ladden tropical maritime atmosphere, and that of considerably dryer and clearer continental polar atmosphere. Scattering causes the atmosphere to have a radiance of its own. The atmospheric luminance at solar altitude angle in the neighborhood of 20^o to 30^o reaches a maximum of about twice the value reached at solar altitude of 90^o. The effect of atmospheric radiance on aerial photography (as well as on other remote sensors) is a function of many variables, such as: sensor altitude; concentration, size-distribution and nature of atmospheric aerosols; solar altitude; spectral sensitivity range of the sensor; angle of  view from nadir and its azimuth with respect to the Sun; and polarization of light.

Scattering plays an important role in remote sensing, be it the effect of atmosphere on the quality of  the received signal, or interference with the signal received from within aquatic environments. Rayleigh scatter is common when radiation interacts with atmospheric molecules and other particles that are much smaller than the wavelength of the interacting radiation. The effect of Rayleigh scatter is inversely proportional to the fourth power of wavelength. This implies that there is a much stronger tendency for shorter wavelengths to be scattered than longer wavelengths. A clear atmosphere with a large  component of nonselective scattering typically shows 50% more scattering in the blue spectral region than in the red. The wavelength dependence of atmospheric scattering accounts for the blue of the sky and the red of the Sun when seen through a long atmospheric path (Sunset). Another type of  scatter is Mie scatter.  This comes about when particle diameter essentially equal the energy wavelength being sensed. This type of scatter tends to influence longer wavelengths compared  to Rayleigh scatter (Doyle F.J et al., 1983; Lillesand and Kiefer, 1987). 

All radiation received by the sensor is termed radiance. The radiance received by the sensor, and corrected for atmospheric scattering and the angle of incident radiation (angle of solar radiation) using mathematical models is termed reflectance. As mentioned earlier, reflectance varies spectrally. It is directionally dependent and it also varies vith the direction of the incident energy. 

Go to top / bottom

1.5 Common Earth Observation Space-borne Platforms (Satellites)

Earth observation satellites are satellites specifically designed to observe Earth from orbit. They are intended for non-military uses such as environmental monitoring, agriculture, forestry, range resources, water resources, coastal resources, land use monitoring, mapping, and geology. 

in 1972 the United States started its Landsat program, the largest program for acquisition of imagery of Earth from space. The LANDSAT-1, -2 had a three-channel RBV (return beam vidicon) camera system and a four-channel MSS (multi-spectral scanner) system with 80 meter resolution on board. The LANDSAT-3 acquired one 505-750 nm, 30 m resolution RBV band and a fifth MSS channel (band 8) operating in thermal infrared region of the spectrum which failed shortly after launch. The following LANDSAT-4 and -5 satellites were equiped with essentially the same MSS as their predecessors, but in addition, they possessed a TM (Thematic Mapper); a highly advanced multi spectral scanner with GSD of 30 m resolution except the thermal band 10400-12500 nm operating with GSD of 120 m. The latest, LANDSAT-7 was launched in April 1999. It is equiped with an upgrade of successful TM - Enhanced Thematic Mapper or ETM+ sensor.
 

LANDSAT-7 is to have a design lifetime of five years. The overall objectives of the LANDSAT-7 Mission are:

• Provide data continuity with Landsats 4 and 5.
• Offer 16-day repetitive Earth coverage.
• Build and periodically refresh a global archive of Sun-lit, substantially cloud free, land images.
• Make data widely available for the cost of fulfilling a user request (called COFUR).
• Support Government, international and commercial communities.
• Play a vital role in NASA's Earth Observing System (EOS) by promoting interdisciplinary research via synergism with other EOS observations. (In particular, orbit in tandem with EOS-AM1 for near coincident observations.) 

SPOT - the Satellite Earth Observation System was designed in France by the CNES (Centre National d'Etudes Spatiales), and developed with the participation of Sweden and Belgium. The SPOT orbit is polar, circular, sun-synchronous and phased. A polar orbit In conjunction with the rotation of the earth around the polar axis, the inclination of the orbital plane (98 degrees) allows the satellite to fly over any point of the earth during a 26 day cycle.

SPOT 1 was launched on 22 February 1986 with 10m panchromatic and 20m multispectral capability. The satellite was withdrawn from active service on 31 December 1990.
SPOT 2 was launched on 22 January 1990 and is still operational.
SPOT 3 was launched on 26 September 1993. An incident occurred on 14 November 1997 and after 4 years in orbit the satellite stopped functioning.
SPOT 4 was launched on 24 March 1998 and includes an extra Short Wave Infrared band and a (low resolution) vegetation instrument.
SPOT 5 was launched on 4 May 2002 with 2.5m, 5m and 10m capability, as well as along-track stereoscopic sensors.

DigitalGlobe is a private enterprise providing commercial high-resolution satellite imagery. It presently operates a constellation of three advanced satellites offering high spatial and spectral resolution imagery.

QuickBird launched in October 2001, WorldView-1 launched in September 2007, WorldView-2 launched in October 2009
 
 

GeoEye set geospatial industry standards with the launch of IKONOS®, the world's first sub-meter commercial satellite. Currently, the scientific community is served by high-resolution imagery from three of its satellites. OrbView-2 was launched in 1997, IKONOS launched in 1999, and  GeoEye-1 launched in September 2008.
 

RapidEye AG is a German geospatial information provider focused on assisting in management decision-making through services based on their own Earth observation imagery. The company owns a five satellite constellation producing 6.5 meter resolution imagery. They were all launched in August 2008, and are equiped with a pushbroom multi-spectral imager, the Jena Spaceborne Scanner JSS 56.
 
 

Radar-based remote sensing works on principles different from those of optical systems. SAR (Synthetic Aperture Radar) systems transmit their own microwave energy towards the surface and record the reflections. Thus, Radarsat satellites can image the Earth, day or night, in any atmospheric condition, such as cloud cover, rain, snow, dust or haze. Radarsat-1 uses a SAR sensor to image the Earth at a single microwave frequency of 5.3 GHz, in the C band (wavelength of 5.6 cm). Each of Radarsat-1's seven beam modes offer a different image resolution. The modes include Fine, which covers an area of 50 km by 50 km (2500 km²) with a resolution of 10 meters; Standard, which covers an area of 100 km by 100 km (10,000 km²) and has a resolution of 30 meters; and ScanSAR wide, which covers a 500 km by 500 km (250,000 km²) area with a resolution of 100 meters. Radarsat-1 also has the unique ability to direct its beam at different angles. With an orbital period of 100.7 minutes, Radarsat-1 circles the Earth 14 times a day. The orbit path repeats every 24 days, this means that the satellite is in exactly the same location and can take the same image (same beam mode and beam position) every 24 days.

Radarsat-1 was launched in November 1995 and is operated by the Canadian Space Agency (CSA). Radarsat-2 was launched in December 2007 and is operated by the MDA Geospatial Services International.

RADARSAT International is RADARSAT-1, LANDSAT, LANDSAT 5, LANDSAT 7, IKONOS, IRS, ERS, QuickBird, and ENVISAT imagery distributor in Canada.

Go to top / bottom

2. Ecology and Ecosystem Management Applications

Throughout evolution, various species of animals have adapted to various combinations of physical factors and vegetation. The adaptations of each species suit it to a particular habitat and rule out its use of other places. The number and type of animals that can be supported in a habitat are determined by the amount and distribution of food, shelter, and water in relation to the mobility of the organism. By determining the food, shelter, and water characteristics of a particular area, general inferences can be drawn about the ability of that area to meet the habitat requirements of different wildlife species. Because these requirements involve many natural factors, remote sensing techniques described here may prove to be an indispensable tool for wildlife habitat evaluation through mapping land cover, soil, forests, wetlands, and water resources analysis. The two most common aspects for which airphoto interpretation can most readily provide useful information are wildlife habitat mapping and wildlife censusing (Lillesand and Kiefer, 1987).

Wildlife biologists recognized the potential of remote sensing for decades. Dalke (1937) was the first to report the use of aerial photography for wildlife cover mapping. Leedy (1948) has published an article in the Journal of Wildlife Management, on the suggested uses of aerial photography in wildlife management. Unfortunately, the application of remote sensing in ecology and ecosystem management has not progressed as rapidly as the advances in remote sensing technology, and its potential still remains somewhat underutilized and unappreciated (Roughgarden J., 1991). There seem to be a couple of reasons for this. The apparent increase in complexity resulting from technological developments has led to decrease in acceptance and use by ecologists and wildlife biologists (Best R.G, 1983). Afterall, remote sensing is a multidisciplinary science which requires a reasonable knowledge and understanding of processes involved. Secondly, there exists some perception that the costs involved are formidable. This view may actually be justified when one considers the acquisition cost of the latest, state of the art high-resolution imaging spectroradiometers or radar sensors. It may be comforting to know however, that satellite remote sensing data is sold on the per scene basis, and that the cost per area covered is actually cheaper than the cost of aerial photography. Furthermore, volume of literature describes the use of inexpensive small and medium format aerial photography for a successful high spatial resolution data collection (Clegg and Scherz,1975; Adams et al.,1977; Curran P.J., 1981; Best R.G, 1983; Lillesand and Kiefer, 1987.

Go to top / bottom

2.1 Wildlife Habitat Monitoring and Classification

Recently the trend has been toward the use of satellite data or combination of satellite imagery and aerial photography (multistage concept of remote sensing); reason being larger area, and repetitive coverage at cost lower than that of large format aerial photography. Unfortunately, this advantage is inevitably offset by lower spatial resolution of satellite imagery as well as higher noise/signal ratios. Another problem with imagery aquired by sensors located on spaceborn platforms is the cloud cover. The use of satellite imagery together with aerial photography offers the best of the two worlds. It provides wide coverage at reasonable cost, yet, areas where higher detail and spatial resolution is required are flown at the time of the satellite overpass to provide the necessary spatial resolution needed.

Many state and provincial government agencies produce habitat inventories based on computerized classification of satellite data and aerial photography interpretation. The interpretation criteria for distinguishing between different classes are similar on both types of imagery. Water bodies have the lowest reflectance resulting in the darkest tones with little or no texture on the imagery. Rivers and streams can be identified by dark tones and a linear meandering shape. They may have light tones if they carry heavy loads of  sediment. Forested land also appears as dark tones on black & white imagery. It can be distinguished from water bodies by its mottled texture. Forested land has deep red tones on color infrared imagery. Cultivated land is easily interpreted by the regular shapes and patterns of fields. Tone and color will vary depending on the crop type, phenology and density. Fallow fields may have very dark tones, but this is variable and depends on the soil type and surface moisture. Rangeland can also have variable tones but generally has a less defined shape or pattern than agricultural land. It may show slight textural differences if there are pronounced vegetative density differences, or if shrubs are present. Urban areas are distinguished by the regular patterns of buildings, parking lots, roads and streets (Best R.G, 1983).

Wildlife Habitat Mapping:

Use of LANDSAT data for moose-habitat analyses in Alaska (Arthur J. Laperriere et al.,1980)

Here, the authors used LANDSAT MSS to produce classified vegetation maps at approximately 13 million hectares of east-central Alaska in order to evaluate moose (Alces alces) habitat. In 1974, the project was organized by the Alaska Department of Fish and Game (ADFG), in cooperation with NASA and the Alaska Cooperative Wildlife Research Unit (ACWRU).  LANDSAT scenes were acquired on the basis of geographic coverage, minimal cloud cover and seasonal date. LANDSAT scenes were computer processed, georectified and classified. A single engine fixed wing aircraft was used for aerial reconnaissance, together with 35mm IR oblique aerial photography and in combination with site visits for ground truthing. The areas to be flown were selected according to the imagery classification so that each cluster class had several of its sites covered by aerial reconnaissance.

Overall description and ecological evaluation of each site were recorded in the field. This summary contained comments on seral succession, evidence of past fire, aspect, and use of the area by wildlife species. Data on tree story, shrub story, and ground cover were obtained at each site. Mean height and maximum, minimum, and mean diameters at breast height were estimated on forested sites. They also estimated percent cover by species and recorded presence/absence of tree species. The data obtained for shrub story were percent cover by species and extent and type of browsing (e.g., planeleaf willow [Salix planifolia], 50% cover,  1/3-2/3 browsed by moose; or Salix lanata, 15% cover, >2/3 browsed by snowshoe hare [Lepus americanus] etc.).

From studies of radio-collared moose, ADFG biologists determined seasonal use patterns of various vegetation types. Based on the seasonal patterns of use, biologists compiled a listing of habitat types used by moose. A synthesis of cluster classes was tabulated for each scene to fit the desired habitat classes. This synthesis resulted in 11 habitat classes: open water, bare ground, early succession, late succession, high brush, coniferous forest, spruce bog or heath, mixed forest, wet sedge meadow, and moist tundra. As a final product, 1:250000 color maps were produced for all data analyzed.

The authors conclude that the use of LANDSAT data permits mapping of large areas in a short time at feasible costs. Unfortunately, due to low MSS spatial resolution small stands of riparian willow occurring along minor drainages were not detected and classified. These riparian willow stands are important winter browse for moose. Nevertheless, LANDSAT data still provided a timely, cost-effective vehicle for synoptic, vegetation type mapping of large areas.

Habitat Classification, Habitat Utilization by Wildlife :

A simple wetland habitat classification for boreal forest waterfowl (Robert S. Rempel et al., 1997)

In this paper the investigators proposed a wetland habitat classification for boreal forest waterfowl that builds on existing classification systems, and which attemts to characterize fundamental properties of wetland structure with the objective of predicting its ecological function. This classification was prepared using interpretation of black and white aerial photographs of parts of 58691 km² Claybelt eco-region of Ontario, Canada (50^oN, 84^o15'W - 48^o15'N, 79^o30'W). The study area 
was gridded into 10x10 km blocks, and each block was further subdivided into 25 400-ha  (2x2 km) of which 140 were selected randomly for pair plots and 104 for brood plots.

The selected plots were delineated on 1:15840 or 1:20000 scale B&W aerial photographs, and each plot and distinct wetland within was assigned a number. Each wetland identified on the photograph was classified by habitat class (Table 1).
 

The classification was tested on 14 species of boreal waterfowl: mallard (Anas platyrhynchos), American black duck (A. rubripes), green-winged teal (A. crecca), blue-winged teal (A. discors), American wigeon (A. americana), ring-necked duck (A.collaris), wood duck (Aix sponsa), hooded merganser (Lophodytes cucullatus), lesser scaup (Aythya affinis), bufflehead (Bucephala albeola), common goldeneye (B. clangula), common merganser (Mergus merganser), Canada goose (Branta canadensis), and common loon (Gavia immer) .

Helicopter waterfowl surveys following the method of Gabor et al. (1995) were conducted over 3 years (1988-1990), 3 times per year for both pairs and broods. Over 3 years, 13 species of waterfowl (ducks and geese) and the common loon  were observed. For the details of the surveys, please refer to (Gabor et al., 1995).

Rempel R.S. et al (1997) evaluated patterns of habitat use by the above 14 species by comparing expected with observed waterfowl distributions, where expected distributions were based on a null model that the birds will use habitat in proportion to its availability. Chi-square statistics was applied to pairs and broods analysis, while Monte Carlo simulations of eigenvalues derived from canonical correspondence analysis (CCA) determined joint multivariate relations between the suite of waterfowl species and environmental variables.

The analysis showed that the habitat use deviated from the expected by random associations for both pairs and broods. Mallard, American black duck, green-winged teal, blue-winged teal, hooded merganser and wood duck pairs selected beaver pond marshes (R-BPM), while American wigeon and lesser scaup selected lake river delta marshes (L-RDM). The selection of P-SOF wetlands was also observed for mallard, ring-necked duck, blue-winged teal, wood duck and Canada geese pairs. The P-OF wetlands were selected by common loon, and bufflehead pairs. For broods, L-RDM habitat was used more than expected by mallard, common goldeneye, and American wigeon, while R-BPM habitat was used more by mallard, American black duck, green-winged teal, and blue-winged teal.

In the original paper, the authors used CCA plots to separate associations of the 14 different species with the 11 different habitats for pairs and broods. The plots showed a separation between divers and dabblers, their relation to the size of wetlands, speed of water movement, varying water debth, and decreasing vegetative cover.  They suggested that a simple classification of wetland habitat based only on identifiable structural characteristics can be used successfully to predict waterfowl community use. Apparently detailed site visits for physical measurements and identification of plant communities were not necessary to predict patterns of waterfowl use at the coarse level of precision addressed in this study.