USGS Spectroscopy Lab
http://speclab.cr.usgs.gov
Derived from (use the following reference):
Trude V.V. King, Roger N. Clark, and Gregg A. Swayze,
Applications of Imaging Spectroscopy Data: A Case Study at
Summitville Colorado, in
INTRODUCTION
From 1985 through 1992, the Summitville open-pit mine produced gold from low-grade ore using cyanide heap-leach techniques, a method to extract gold whereby the ore pile is sprayed with water containing cyanide, which dissolves the minute gold grains. Environmental problems due to mining activity at Summitville include significant increases in acidic and metal-rich drainage from the site, leakage of cyanide-bearing solutions from the heap-leach pad into an underdrain system, and several surface leaks of cyanide-bearing solutions into the Wightman Fork of the Alamosa River. In general, drainage from the Summitville mine moves downslope into the Wightman Fork, a small tributary of the Alamosa River, which in turn flows east into the Terrace Reservoir before entering the agricultural lands of the San Luis Valley. The increase in the trace-metal burden of the Alamosa River watershed due to the mining activities at Summitville is of concern to farmers and fisherman, as well as Federal and State of Colorado agencies having responsibility for land stewardship. >
The environment of the Summitville area is a result of 1) its geologic evolution, that culminated in the formation of precious-metal mineral deposits; and 2) previous metal mining activity. Mining accentuates, accelerates, and pertubates natural geochemical processes. The development of underground workings, open pits, mill tailings, and spoil heaps and the extractive processing of ore enhances the likelihood of releasing chemicals and elements to the surrounding areas and at increased rates relative to unmined areas. Both mined and unmined mineralized areas can produce acid drainage from the formation and movement of highly acidic water rich in heavy metals. This acidic water forms principally through the chemical reaction of oxygenated surface water and shallow subsurface water with rocks that contain sulfide minerals, producing sulphuric acid. Heavy metals can be leached by the acid solution that comes in contact with mineralized rocks, a process that may be enhanced by bacterial action. The resulting fluids may be highly toxic and, when mixed with groundwater, surface water, and soil, may have harmful effects on humans, animals, and plants. Thus, understanding the geologic and hydrologic history of this area is a critical piece of the environmental puzzle in the Summitville area.>
The Summitville mine operators had ceased active mining and begun environmental remediation, including treatment of the heap-leach pile and installation of a water-treatment facility, when it declared bankruptcy in December 1992 and abandoned the mine site. The U.S. Environmental Protection Agency (EPA) immediately took over the Summitville site under EPA Superfund Emergency Response authority. >
Summitville has focused public attention on the environmental effects of modern mineral-resource development. Soon after the mine was abandoned, Federal, State, and local agencies, along with Alamosa River water users and private companies, began extensive studies at the mine site and surrounding areas. These studies included analysis of water, soil, livestock and vegetation. The role of the U.S. Geological Survey (USGS) was to provide geologic, hydrologic and agricultural information about the mine and surrounding area and to describe and evaluate the environmental condition of the Summitville mine and the downstream effects of the mine on the San Luis Valley (King, 1995). >
>
IMAGING SPECTROMETER DATA
To address geological, hydrological, and agricultural problems, approximately 2100 km2 of imaging spectrometer data for the Summitville mining district, adjacent areas in the San Juan Mountains, and agricultural areas in the neighboring San Luis Valley were collected on September 3, 1992 (Figure 1). >
Imaging spectroscopy is a departure from traditional remote sensing concepts in that the data represent continuous, narrow-band spectral coverage over a selected portion of the electro-magnetic spectrum. Spectroscopic processing delineates absorption features in reflectance due to individual chemical bonds in surface materials and in the atmosphere and, when used with image analysis, maps their occurrence and distribution.
The narrow spectral channels of an imaging spectrometer form a nearly continuous sampling of the reflectance spectrum of the Earth's surface, in contrast with the 4 to 7 discontinuous broad channels of the other imaging instruments , such as Landsat Thematic Mapper (TM) and Multispectral Scanner (MSS). Imaging spectroscopy duplicates the capabilities of Landsat by distinguishing brightness and slope differences in the reflectance spectrum of the surface. However, imaging spectroscopy can also resolve absorption bands in the spectrum which can be used to identify spectral components. Analysis of imaging spectroscopy data allows minerals, vegetation types, manmade materials, water, snow, and many other materials to be mapped if they have unique and identifiable absorption features in the 0.4 to 2.45 m spectral region (e.g. see Clark et al., 1992 , Clark et al., 1993).>
The imaging spectroscopy system used for this study was the NASA "Airborne Visible Infra-Red Imaging Spectrometer" (AVIRIS) instrument. AVIRIS acquires data in the spectral range from 0.4 to 2.45 m in 224 spectral channels. The instrument is flown in an ER-2 aircraft (a modified U-2) at 19,800 meters (65,000 feet). The pixel size is 20 meters with 17 meters spacing between pixel centers; the swath width is about 10.5 kilometers (614 pixels). The swath length can be as much as 1000 kilometers, if necessary, with the current tape recorder capacity on the aircraft.>
A combined method of radiative transfer modeling and ground calibration were used to correct the AVIRIS radiance data to surface reflectance (Clark et al., 1998). This method, called Radiative Transfer Ground Calibration, or RTGC, corrects for variable water vapor and other components in the atmosphere, produces surface reflectance spectra free of unwanted artifacts, with spectral channel to channel noise approaching the true signal-to-noise ratio of the data. In the RTGC method, the initial step is to compute a radiative transfer model of atmospheric scattering and absorption for each pixel and remove these effects from the data. Secondly, a ground target is measured to characterize the spectral properties of a known area. The ground target characterization is used to correct any minor absorption residuals remaining after the radiative transfer model calculation and correction has been applied to the AVIRIS data. For this study, the primary calibration site was a plowed, vegetation-free field near the center of the AVIRIS data set. Approximately 12 individual samples from the calibration site were obtained on the day of the overflight and spectrally characterized on a modified Beckman-5270 laboratory spectrometer (Clark et al, 1990b). These spectra then were averaged together to represent the spectral signature of the overall calibration site. The site soil samples were spectrally bland and, thus, suitable to use as calibration standards. >
DATA ANALYSIS
Subsequent to the RTGC process, the data are analyzed to determine the materials present in the imaged scene. Clark et al., (1990a, 1991, 1992, 1995, 1998) developed a new analysis algorithm, Tetracorder, that uses a digital spectral library of well-characterized materials and a fast, modified-least-squares method of determining if spectral features for a given material are present in a pixel of the imaged data. >
The current version of Tetracorder uses stepwise analyses that build on previous analytical results and directs subsequent analyses. In analyzing imaging spectroscopy data sets, many tens of materials can be found in a single scene and, perhaps, hundreds can be found if appropriate standards existed. Materials can be minerals, snow, vegetation (and different vegetation species), pollution, man-made objects, etc. Because so many different spectral signatures exist in a single scene, a spectral feature identification algorithm (Clark et al., 1990a) was developed as a first step in analysis. Using this feature matching algorithm, the analysis was expanded by Clark et al. (1991) to simultaneously analyze multiple spectral features in each unknown spectrum. Tetracorder applies multiple tests to match the spectrum under evaluation with standard spectra by comparing continuum-removed spectral features from the imaging spectrometer data set to corresponding continuum-removed spectral features from a reference spectral library (Clark et al., 1990; Clark et al., 1991). The continuum removal process isolates an absorption band from "background absorptions" so that spectral features, including depth, shape, and wavelength position can be accurately analyzed and compared. To facilitate comparison, a straight-line continuum is removed from the library reference spectrum using data points (spectral channels) on each side of the absorption feature that is to be mapped. The same methods are used to remove the continuum from the observed spectrum in the flight image data. Several channels on each side of the absorption feature can be selected as continuum end-points to reduce noise in the continuum. The continuum is removed from the spectrum by division because of the non-linear effects of scattering and Beers Law. Adjacent spectral regions are analyzed simultaneously, resulting in multiple identifications when multiple spectral components are present. Depending on the results of this analysis phase, Tetracorder may choose additional analyses.
Other Tetracorder capabilities include fit, band depth, and continuum thresholding. For example, fit thresholding evaluates the value of the correlation coefficient from the least squares analysis. If the fit is too low to reasonably identify any of a given set of materials, the pixel is rejected from further analysis. Band depth thresholding works in a similar fashion. Continuum thresholding prevents material identification in pixels with very low continuum levels where noise could resemble absorption features. By knowing the continuum value, it is possible to determine if the spectrum represents an area in a shaded region, over water, or in a region obscured by a cloud.
The analytical methods take an all-inclusive approach to the observed electromagnetic spectrum, treating each spectrum as a single data object, rather than as a collection of 'channels' or discrete wavebands, thus allowing a more robust solution. Traditional analytical methods require a set of spectral end members (e.g. - calcite, dolomite) to sum to one, which requires some of the end members to be identified in all pixels. Tetracorder analysis has no such requirement, and only finds materials if diagnostic absorption features are present.
VERIFICATION OF IMAGING SPECTROMETER DATA AND RESULTS
Imaging spectrometer data allow several methods of verification, surpassing the capabilities of other remote sensing data. Common remote sensing data, such as MSS and TM, have a few spectral bands (data points) and each measured value can be influenced by a multitude of factors. Many factors, including, but not limited to, atmospheric absorptions, calibration errors, and surface material absorptions, affect the measured radiance, but the degree to which each factor influences the measurements cannot be determined. In contrast, a continuous spectrum over a wide wavelength interval (e.g. the ultraviolet to near-infrared wavelength region), such as from AVIRIS or a field-spectrometer, produces sufficient information to resolve the individual spectral contributions related to: (1) calibration errors, (2) atmospheric absorptions, (3) model artifacts, and (4) material surface absorptions. The spectral effects of each contribution then can be evaluated and corrected.
Data accuracy is a measurement of how well the imaging spectrometer data represent the true reflectance characteristics of the surface. Verification of the accuracy of imaging spectroscopy data, such as AVIRIS, can be accomplished by several means, including self-verification. In the self-verification method, spectra are extracted from the corrected image and examined. The wavelength calibration of the data set can be verified on the basis of known absorption features that do not shift in wavelength position (for example, atmospheric gases and selected materials). The correction of the imaging spectroscopy data to surface reflectance also can be confirmed using self-verification. Errors in the correction model or method used to derive the surface reflectance values can produce residual positive or negative spectral features that are recognized by an experienced spectroscopist as anomalous compared to the known spectral signatures of surface materials and atmospheric gases. Thus, the identified errors can be corrected to derive a viable surface reflectance.
Verification of imaging spectrometer material distribution maps, including minerals, vegetation, and man-made materials, can be accomplished by (1) traditional field verification methods (see below), and (2) by direct examination of the imaging spectrometer data. By extracting a spectrum from the data set, diagnostic absorption features can be identified and material identification confirmed based on their wavelength position and shape. This direct examination of the image can be done without field checking. This latter method only works for those materials having diagnostic absorptions that are separated from the spectral features of material mixtures. In many instances, the individual components of material mixtures can be identified even if overlapping absorption features exist. By visually examining a spectra or by using modern spectral identification algorithms, individual minerals and mixtures, such as kaolinite and hematite, montmorillonite and goethite, jarosite and muscovite, calcite or dolomite, can be positively identified. However, some material mixtures can produce absorption features that can be confused with other materials (e.g., a kaolinite and smectite mixture is spectrally similar to a spectrum of halloysite). Consequently, field checking is often required to properly determine the composition.
Traditional field checking methods require a sample from a specific geographic area represented by a pixel in the image. The sample then is analyzed by a method (e.g. X-ray diffraction or a micro-beam technique), that provides (1) mineralogy, (2) mineral composition; and (3) mineral abundance. These data then are compared to the spectral information. Given that the imaging spectroscopy data represent only the surface of the exposed material, care must be taken during the laboratory phase to ensure that the materials analyzed in the laboratory are the same as those mapped by the imaging spectrometer. Different analytical techniques have different levels of detection for specific materials. For example, spectroscopy is more sensitive for the identification of clays, iron oxides, and amorphous materials, sometimes by factors of ten compared to X-ray diffraction methods. In contrast, X-ray diffraction analysis commonly can discriminate between kaolinite and smectite mixtures, or kaolinite and muscovite mixtures, and halloysite, which are difficult to discriminate using spectroscopy data. Failure to evaluate the detection limitations of each method can result in a disagreement between different analytical techniques. Field-checking verifies both the accuracy of the mapping algorithm and the spectral standards and is the most rigorous type of validation. Unfortunately, field-checking can be time-consuming, costly and, in some instances, may not be logistically feasible.
MAPPING MINERALS
Reflectance spectra can provide information on the chemistry, mineralogy and crystal structure of materials. Spectroscopy can identify both crystalline and amorphous materials. Subtle changes in composition or structure will result in changes in shape and/or wavelength position of characteristic absorption features. Laboratory investigations have quantified the causes of changes in position and shape of absorption features for particulate minerals (Hunt, 1977 and references therein; Adams, 1975, Gaffey et al., 1993), carbonates (Gaffey, 1986, 1987), silicates (Cloutis et al., 1986, Cloutis and Gaffey, 1987; King and Ridley, 1987), and phyllosilicates (Crowley et al., 1988; King and Clark, 1989; Clark et al., 1990). Advancements in remotely-sensed data acquisition platforms and computer analysis techniques now allow a direct comparison between laboratory and remotely sensed data. Thus, well-characterized laboratory spectral databases can serve as "standards" to map the spatial distribution and composition of remotely-sensed materials.
Reflectance spectroscopy identifies and maps specific chemical bonds in materials which are present in the upper 1-2 mm (in most instances) of an exposed surface. Mineralogically, this material commonly represents the layer which has been exposed to physical and chemical weathering. These minerals may all be secondary weathering products, depending on the composition of the parent mineral/rock and the physio-chemical conditions of weathering. The ability to define and map secondary minerals and amorphous, or poorly crystalline, materials is important because they are common and are sources of easily-available trace metals and anions. Thus, by combining the spatial distribution and mineral chemistry data gained from the imaging spectroscopy data, with supporting field studies, it is possible to identify sources of pollution, monitor mineral transport and fate, predict metal geoavailabitlity, and assess mineral associations.
Absorption bands in the visible and near-IR portion of the spectrum (~0.4-1.0 m) are caused by electronic processes, including those due to crystal field effects, charge transfer, color-centers, and conduction bands. The absorption features resulting in this portion of the spectrum often involve elements of the first transition series, which have an outer unfilled electron orbital d-shell. The energy levels are determined by the valence state, coordination number, and site symmetry of the element. Differences in these parameters between materials are manifested as individual diagnostic absorption bands in the visible and very near infrared wavelength regions. Absorptions in this wavelength region commonly are associated with the presence of iron (Fe) and other transition elements (Mn, Cr, Ti, etc.) in the mineral structure. The intrinsic strength of these absorptions is quite strong. Therefore, reflectance spectroscopy is well suited to study the varied Fe-bearing oxides, sulfates, and hydroxides produced by typical weathering associated with unmined and mined mineralized areas.
Near-infrared radiation (1-2.5 m, in this study) absorbed by minerals and other materials is most commonly converted into molecular vibrational energy. The frequency or wavelength of the absorption depends on the relative masses and geometry of the atoms and the force constants of the bonds.
Absorption features in the 2.2 to 2.3-m region are commonly used to make mineral identifications. Many of the absorption features in this wavelength region result from a combination of the OH-stretching fundamental with either the Al-O-H bending mode (usually absorbing at approximately 2.2 m), the Mg-O-H bending mode (absorption usually near 2.3 m), and to Fe-O-H bending mode absorbing near 2.25 m. At high spectral resolution, these bands are recognized to be complex absorption features. Based on previous work (King and Clark, 1989, Clark et al., 1990b, Swayze and Clark, 1990, Clark et al., 1993, and others), the strength, position and shape of these features has been found to be a function of the mineral chemistry. Similarly, an overtone of the fundamental asymmetrical stretching mode of C-O in the carbonate ion produces an absorption feature in the 2.3 to 2.34-m wavelength region.
For the present study, minerals were mapped based on the presence of absorption features in the ~0.45 to 2.45 m wavelength region (i.e., the visible and near-infrared portions of the electro-magnetic spectrum). A laboratory standard spectral database totaling 130 minerals was used as a basis for comparison with the remotely sensed data. The database included pure minerals, mineral mixtures, and materials collected from the Summitville study area. From this database, a search was conducted for 50 minerals with absorption features at wavelengths near or less than 1.0 m and 14 of these 50 potential minerals were mapped in the Summitville/San Luis Valley region. The 14 mapped materials are predominately crystalline and amorphous iron-oxides, thus their spectral identification was based primarily on the wavelength position and shape of the continuum removed 0.9 m absorption feature. Continuum removal, as previously mentioned, allows subtle differences to be distinguished between similar, but distinct, absorption features. The minerals detected include: two types of goethite, two types of jarosite, two different Fe-bearing minerals, nanohematite, coarse- and fine-grained hematite, amorphous Fe-oxide, ferrihydrite, wet amorphous Fe-oxide, Cu-bearing precipitate, and sediment-bearing water.
To evaluate the presence of minerals that have absorption features in the 2.2-2.3 m wavelength region, 80 laboratory mineral standards were used. These standards included phyllosilicates, sulphates, carbonates, and cyanide compounds. Eleven (11) different minerals of significant areal extent were detected in the AVIRIS data. Subtle spectral differences allow discrimination between K and Na alunites and between poorly-crystalline and highly-crystalline kaolinites. However, because of spectral similarities and limitations of the mapping algorithm, some of the material mapped as Na-montmorillonite may be muscovite or sericite.
Spectral data contained in the AVIRIS pixels are very similar to the spectral standards measured in the laboratory. Figure 2 shows the spectrum of a mixture of alunite, kaolinite and goethite detected in the AVIRIS data from near Alum Creek compared to laboratory standard spectra of an alunite and kaolinite mixture and goethite. Breaks in the spectra of the AVIRIS data occur at the wavelengths where absorptions from atmospheric gases occur. The absorption features in the spectrum of the mineral mixture from the AVIRIS data can be correlated with the diagnostic absorption features in the laboratory standard. The absorption features near 0.7 and 1.0 m in the pixel spectrum result from the presence of iron in goethite. The absorption features between 1.4 and 1.7 m and the one near 2.3 m are due to the presence of alunite. Those near 2.2 m result from the presence of kaolinite in the pixel spectrum.
Spectral processing and analysis of AVIRIS data using absorption features in the 2 m wavelength region identified areas of hydrothermal alteration minerals. Data covering the Summitville mine showed mineralogical differences between the open-pit and the heap-leach-pile at the mine site. Hydroxyl-bearing materials, including clays, show discrete distribution patterns at both the mine site and within the Iron Creek, Alum Creek, and Bitter Creek basins (Figure 3). The data also defined discrete mineralogical boundaries in one other basin (Burnt Creek) which is the principle drainage for mineralized regions that are not included in the over-flight area.
An interesting observation is that the Summitville mine, at the time of the over-flight, apparently did not contribute OH-bearing minerals via the Wightman Fork to the Alamosa River. In contrast, the mineralized areas in Iron Creek, Alum Creek and Bitter Creek basins do contribute OH-bearing minerals to the Alamosa River. This observation is based on the spectral characteristics of the exposed fluvial sediments along Alum Creek and Bitter Creek and lack of exposed OH-bearing fluvial sediment along the Wightman Fork. The unmined mineralized areas are believed to contribute OH-bearing materials to the Alamosa River due to the porous character of the well-exposed hydrothermally altered bedrock, which allows altered materials to be eroded easily and carried downstream. The paucity of phyllosilicates being transported from the mine site to the Wightman Fork may be attributed to on-site remediation efforts. Conversely, the material may be carried as a very fine-grained aqueous suspension, which does not settle onto the creek banks, or as aqueous components that may precipitate under higher pH conditions downstream.
From the AVIRIS data, it appears that only a small amount of material having vibrational absorptions is being transported from all upstream locations into the Terrace Reservoir via the Alamosa River. At the time of data acquisition, no material (or amounts below the spectral detection limit) having absorption features near 2.2-2.3 m were being discharged from the Terrace Reservoir. Analogously, there is no spectral indication of alteration materials having 2.2-2.3 m absorption features being transported via La Jara Creek to the San Luis Valley (Figure 1).
Spectrally detectable quantities of OH-bearing materials are present in some plowed and/or harvested fields in the San Luis Valley. On-site field-checking of the plowed fields revealed small to fist-size clumps of alunite, suggesting that at some earlier time the alunite and associated minerals were transported from areas of exposed altered material, presumably near the Alum Creek, Bitter Creek, or Iron Creek basins, to the valley floor.
AVIRIS images show that the Summitville mine and Iron Creek, Alum Creek, and Bitter Creek basins are sources of iron (Fe)-bearing sediments to the Alamosa River (Figure 4). Field observations show that these sediments give a reddish-brown color to stream banks, a characteristic typically associated with acid drainage, and are potential carriers of heavy metals to locations downstream. The AVIRIS data show that the Fe-bearing materials enter the Terrace Reservoir via the Alamosa River. Consequently, in assessing the environmental impact of mining near Summitville, it is important to recognize that both the Summitville mine site and the local drainage basins associated with unmined mineralized areas are contributors of Fe-bearing sediments and aqueous components.
Based on the AVIRIS data analysis, there are no other sources of alteration minerals from mineralized areas of substantial size that significantly influence the distribution of Fe-bearing sediments. However, two large alluvial fans associated with the Alamosa River and La Jara Creek distribute Fe-bearing sediments to the Valley floor (Figure 1). Both of these alluvial fans show distinct lobes, some of which have been incorporated into cultivated fields. The wide-spread aerial distribution patterns of the Fe-bearing sediments incorporated in the fans indicates that they are the natural weathering products of the volcanic rocks in the San Juan Mountains and, for the most part, are not related to mining activities at the Summitville mine.
Reservoirs and lakes (including La Jara and Terrace Reservoirs) in the data set have been mapped as specific minerals, which is inaccurate. However, based on spectra that have been extracted from the remotely sensed data set, these water bodies are different than our standard water spectra. The La Jara Reservoir and some lakes (Big Lake) in the data set map similarly, but the Terrace Reservoir appears anomalous in mineral maps. The AVIRIS data may be detecting suspended sediments, algae, or bottom sediments, but the exact character of the materials detected in the bodies of water and why they map as minerals are subjects of ongoing studies.
MAPPING VEGETATION
Obtaining quantitative information about vegetation has proven difficult. To first order, all vegetation is chemically similar, and most healthy plants are green and have similar absorption features. However, the human eye sees plants as shades of green, thus allowing subtle, but significant, spectral differences between plant species to be distinguished. Quantifying these differences through spectroscopy provides the ability to map plant species, determine the water abundance in a species, and determine the relative health of a species or community.
The primary spectral features used for vegetation identification result from the presence of chlorophyll, the organic material that gives plants their green color. The human eye sees different plants as shades of green, because of the "green peak" of reflected light. The "green peak" results from the absorption of most wavelengths of light, except those that appear as green to the human eye, in the visible portion of the electro-magnetic spectrum. The broad absorption features that are responsible for the overall similarity in the green color of vegetation spectra result from the presence of chlorophyll and other pigments. However, absorptions due to bending and stretching of the O-H bond in water, as well as the presence of carbon and nitrogen in the plant structure, result in absorption features at other wavelengths as well (Danks et al., 1984; Murray and Williams 1987; Curran, 1989; Curran et al., 1992).
Although the spectra of plants are sufficiently different to allow species identification, the spectra of an individual species can vary. Spectral variations probably result from the amount of chlorophyll and water in the plant, a complex relation between the stages of a growing cycle and the health of a plant. The health of a plant can be affected by many factors, including the amount of water available (too much or too little) and metal toxicity. These factors influence the shape and depth of the characteristic absorption features.
The Tetracorder algorithm is very sensitive to the shape of spectral features and has the potential to distinguish more subtle differences in the visible spectrum of plants than the human eye. The continuum-removal portion of the algorithm is an important step in detecting and mapping vegetation, particularly when a pixel contains spectral information from green plants, dry vegetation, and soil. By isolating the absorption features with continuum removal, the position and shape of the continuum removed-spectral feature remains constant, although its depth changes as a function of the areal extent of the vegetation in the pixel.
The extreme changes in elevation from ~3960 m near the Summitville mine to ~2300 m in the San Luis Valley, is reflected by diverse vegetation communities, ranging from alpine to irrigated agricultural environments. Much of the mapped area is above tree-line at the higher elevations, and grades downward into mixed Lodgepole and Ponderosa Pine, Douglas Fir, Aspen and deciduous growth at intermediate elevations.
At the lower elevations, the study area includes farmland producing potatoes, alfalfa, barley, oat hay, canola, and fields containing chico and other unidentified weeds. Ideally, data analysis could use a digital spectral library of reference spectra for all plant species likely to be encountered in the study area. However, such a library does not exist for vegetation as it does for minerals (e.g. Clark et al., 1993). Producing a spectral library for vegetation will be much more complex than for minerals, as the vegetation is not static; stress factors and the number of spectra, as a function of growing season, that would be required to adequately represent a plant species is not known.
Reference spectra were obtained for sites of known species directly from the AVIRIS data. AVIRIS, having been well calibrated to surface reflectance, acts as an excellent field spectrometer, providing data for large areas and averaging over many plants to reduce spectral variations within one species. The reference spectra obtained from the AVIRIS data set are shown in Figure 5. The alfalfa, canola, oat hay, and Nugget potato spectra (Figure 5a) show the plants to be green and healthy. The barley had lost all its chlorophyll signature because of natural senescence (Figure 5b). The Norkotah potatoes were not being irrigated as they were about to be harvested, and consequently they showed a weak chlorophyll and cellulose absorptions as well as clay absorptions from exposed soil. These potatoes were also being sprayed with a defoliant in preparation for harvest and should show decreased chlorophyll absorption, along with a shift of the red edge of the absorption to shorter wavelengths due to drying. The chico and pasture spectra show combinations of chlorophyll and cellulose (dry vegetation) absorptions, which can be attributed to seasonal and species variability.
Differences in the shape of the absorption features in the continuum-removed (chlorophyll-containing) crop spectra enabled differentiation using the AVIRIS data (Figure 6). The continuum-removed spectrum of each pixel in the image was examined and compared to the standard crop spectra to produce a color-coded distribution map (Figure 7). Farmers in the San Luis Valley rely heavily on irrigation waters applied by center-pivot irrigation systems, thus most farmed lands appear as circular areas in the AVIRIS data.
Field verification data were supplied by Maya ter Kuile of Argo Engineering (1993, personal communication), and as part of this investigation. Of 43 verification fields (not including chico/pasture areas), 7 included the sites of reference spectra and were, of course, identified correctly. Of the remaining 36 fields, 33 were identified correctly and another 3 were identified as mixed by the Tetracorder analysis (but were indicated as one crop type in the field data); no fields were misidentified. The mixed fields were identified as consisting of two species, one of which was correct in each case. If a score of half is given to the three fields identified as mixed, the 96% score (34.5 of 36) implies the method is accurate.
The accuracy is even more impressive considering several fields were mapped correctly even though they had already been harvested suggesting that low concentrations of plant residue are sufficient to make correct species identification. The harvested areas are identified in Figure 7 by circular plots where the colored pixels are sparse and/or low in intensity. The spectra of the harvested areas are also identified in the senescence/stress map (see below).
Canola was mapped in many of the areas known to be chico/pasture. While it is possible that canola seeds have been blown into surrounding fields and are thriving, it is more likely that the canola spectral signature is similar to other plants in those uncultivated fields for which standard spectra were lacking.
The distribution of vegetation species/communities based on the shape of the chlorophyll absorption features near the Summitville mine are shown in Figure 8. Fifteen generic vegetation types are identified within this AVIRIS scene. However, because of a paucity of standard reference vegetation spectra for this region and the rapid onset of winter after the data were collected, supporting field-work was not an option. Consequently specific species names can not be given for this portion of the study. Thus, Figure 8 is indicative of vegetation communities rather than individual species. It is interesting to note the differences in communities on north and south facing slopes. These differences may be because of differences in community distribution or indicative of differences in fall-colors (natural senescence) between north and south facing slopes or both.
Figure 9 shows a green-vegetation/water-abundance map for the area near Summitville, based on the depth of 0.95 and 1.2 m water absorption features and the chlorophyll absorption feature. These features are indicative of the dryness of vegetation in the AVIRIS data. Low abundances of chlorophyll or water, as indicated by either the 0.95 or 1.2 m water absorption features or chlorophyll absorption, can produce similar spectral affects. If water abundances are low, the mixing of the three end member colors, representing the three absorption features, will produce red to reddish-brown colors as seen on Figure 9. Because the data were acquired in September, the natural senescence of the vegetation at the higher elevations had begun and preferential "drying" was taking place on the south-facing slopes. Comparably, the agricultural crops in the San Luis Valley show different stages of maturation and, consequently, dryness.
SENESCENCE/STRESS MAPPING
The long-wavelength side of the chlorophyll absorption (~0.68 to ~0.73 m) forms one of the most extreme slopes found in spectra of naturally occurring common materials, plants or minerals. The absorption is usually very intense, ranging from a reflectance low of less than 5% (near 0.68 m) to a near infrared reflectance maximum of ~50% or more (at~0.73 m). The properties of the reflectance spectra of plants often indicate that this absorption band is "saturated". If the absorption feature is saturated, the wavelength position of the absorption band minimum will not differ significantly, but the wings, and consequently the width, of the absorption will change. When the chlorophyll absorption in the plant decreases, the overall width of the absorption band decreases. The short wavelength side of the chlorophyll absorption can not be observed in reflectance spectra because of other absorptions in the ultraviolet (UV) wavelength region. In spite of these observational limitations, decreases in the strength of the chlorophyll absorption (less chlorophyll in the plant) causes a shift of the long wavelength side of the absorption feature toward shorter wavelengths. This has popularly become known as the "red-edge shift" or the "blue shift of the red edge" and has been used by researchers as an indication of senescence or stress-induced chlorosis (e.g. Milton et al., 1983; Rock et al., 1985; Miller et al., 1987; Milton et al., 1989). However, questions remain about the limits of these measurements for canopies and leaves, as well as the exact causes of these shifts (Curran et al., 1991).
To determine if a red-edge-shift occurred, where it occurred, and the relative amount of shift within the data set, we used field spectrometer spectra from the San Luis data set to compute a ratio cube. The ratio of two spectra, each having steep "red edge" spectral slopes, which are shifted in wavelength relative to the other, will produce a spurious value even if there is only a small relative shift between them. If a "blue" shifted spectrum is divided by an unshifted spectrum, a peak will be observed in the ratio. For a spectrum of green vegetation (from Figure 5), a 1 nm shift will produce a residual feature of approximately 6% (Clark et al., 1998). The AVIRIS data have a signal to noise ratio of several hundred in this spectral region, so red-edge shifts of less than 0.1 nm can be detected. Figure 10 shows differences in the senescence/stress of the vegetation from the Summitville mine site to the San Luis Valley based on differences in the position and shape of the chlorophyll absorption feature.
Many red-edge shifts are observed in the data set (Figure 10). However, to be useful as an indicator of stress, the red-edge-shift map should be compared to the community/species map to reduce interpretative bias, as red-edge-shift can vary as a function of species. Comparison between these two maps suggests that no major red-edge shifts can be related to mining activities. However, the time of the data collection (September) was not optimum for studies focusing on the spectral detection of vegetation stress. Natural senescence and agricultural processes (defoliation) could conceal the spectral identification of plant stress related to the uptake of toxic materials. It should be noted that if an area of "stressed vegetation" was suggested in the AVIRIS data, additional field work and analysis would be necessary for confirmation.
CONCLUSIONS
The unique utility of imaging spectroscopy in mapping mineral and vegetation distribution on both local and regional scales at the Summitville mine and adjacent portions of the San Juan Mountains and San Luis Valley has been demonstrated. Imaging spectroscopy data provides mineralogical and chemical data unavailable from any other remote sensing method, and is an excellent tool for environmental assessments, mineral mapping and exploration, vegetation communities/species and health studies, and general land management applications. For geologic studies, imaging spectroscopy data can be used to detect and map sources, pathways of transport, and fate of materials in an area having unmined hydrothermal alteration and mining activities. In addition, applications of imaging spectroscopy data for several aspects of botanical investigations have been demonstrated.
In the Summitville mining region, the mine site does not contribute clay minerals to the Alamosa River, but does contribute Fe-bearing minerals, and the Fe-bearing materials are being transported into the Terrace Reservoir. Such minerals have the potential to carry heavy metals and their transport and fate need to be monitored as they can serve as a source of contamination. The imaging data suggest that if hydroxyl-bearing and sulphate-bearing materials are being transported in the Wightman Fork, they are in suspension and have not been deposited in bank-materials in this reach of the Summitville watershed, but may be deposited further downstream due to variations in water chemistry.
The AVIRIS data suggest that alteration materials, including alunites, montmorillonites, and kaolinite/smectites have previously been transported to the valley floor from the area near the Summitville mine and/or Alum Creek, Iron Creek, or Bitter Creek basins as suggested by the spectral signature of some plowed fields. Alluvial material has been widely dispersed on the Valley floor from both the Alamosa and La Jara drainages. Because of its widespread distribution and the lack of either unmined mineralized or mining sources in the La Jara basin, these materials are believed to be primarily related to the natural weathering of the volcanic rocks in the adjacent San Juan Mountains and are not environmentally significant as sources of heavy metals.
The utility of imaging spectroscopy as a source of botanical data for environmental applications, to monitor vegetation cover and its health, and define the distribution of vegetation communities and specific species has been demonstrated. The data presented here illustrate the advantage of using continuum-removed spectra for general detection and mapping of agricultural crops, as well as defining differences between closely related species. The ability to remotely detect and map differences in plant species could be used for more specific, and potentially more accurate, crop yield predictions.
The AVIRIS data for the Summitville area and San Luis Valley show that some vegetation spectra have a shift in the wavelength position of the chlorophyll absorption feature (a red-edge-shift). However, when compared to vegetation community maps, these changes in the indigenous vegetation and agricultural crops cannot be attributed to metal loading by either the mined or unmined mineralized areas in the region.
ACKNOWLEDGEMENTS
This manuscript has benefitted from the helpful comments and reviews of Dan Knepper, Ian Ridley, and Eric Livo. We thank Robert Green for his help in acquiring the AVIRIS data.
REFERENCES
Adams, J.B., 1974, Visible and near-infrared diffuse reflectance of pyroxenes as applied to remote sensing of solid objects in the solar system. J. Geophys. Res., 79, pp. 4829-4836.
Clark, R.N., A.J. Gallagher, and G.A. Swayze, 1990a, Material absorption band depth mapping of imaging spectrometer data using a complete band shape least-squares fit with library reference spectra: Proceedings of the Second Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) Workshop, JPL Publication 90-54, pp. 176-186.
Clark, R.N., T.V.V. King, M. Klejwa, G. Swayze, and N. Vergo, 1990b, High Spectral Resolution Reflectance Spectroscopy of Minerals: J. Geophys Res. 95, pp. 12653-12680.
Clark, R.N., G.A. Swayze, A. Gallagher, N. Gorelick, and F. Kruse, 1991, Mapping with Imaging Spectrometer Data Using the Complete Band Shape Least-Squares Algorithm Simultaneously Fit to Multiple Spectral Features from Multiple Materials, Proceedings of the Third Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) Workshop, held in Pasadena, California, on May 20-21, 1991, JPL Publication 91-28, pp. 2-3.
Clark, R.N., G.A. Swayze, C. Koch, A. Gallagher, and C. Ager, 1992, Mapping Vegetation Types with the Multiple Spectral Feature Mapping Algorithm in both Emission and Absorption: Summaries of the Third Annual JPL Airborne Geosciences Workshop, Volume 1: AVIRIS Workshop. held in Pasadena, California, on June 1-5, 1992, JPL Publication 92-14, pp.60-62,
Clark, R.N., G.A. Swayze, A. Gallagher, T.V.V. King, and W.M. Calvin, 1993, The U. S. Geological Survey, Digital Spectral Library: Version 1: 0.2 to 3.0 m, U.S. Geological Survey, Open File Report 93-592, 1340 pages.
Clark, R.N., T.V.V. King, C. Ager, and G.A. Swayze, 1995, Initial vegetation species and senescence/stress mapping in the San Luis Valley, Colorado using imaging spectrometer data. Summaries of the Fifth Annual JPL Airborne Earth Science Workshop, January 23- 26, R.O. Green, Ed., JPL Publication 95-1, pp. 35-38.
Clark, R.N., G.A. Swayze, K. Heidebrecht, G.A.F.H. Goetz, 1998, Calibration to Surface Reflectance of Terrestrial Imaging Spectrometry Data: Comparison of Methods, Applied Optics in review.
Cloutis, E.A., and M.J. Gaffey, 1991, Pyroxene spectroscopy revisited: Spectral-compositional correlations and relationships to geotherometry. J. Geophys. Res., 96, pp. 22,809-22,826.
Cloutis, E.A., M.J. Gaffey, T.L. Jackowski, and K.L. Reed, 1986, Calibrations of phase abundance, composition, and particle size distribution of olivine-orthopyroxene mixtures from reflectance spectra. J. Geophys. Res., 91, pp. 11,641-11,653.
Curran, P.J., 1989, Remote sensing of foliar chemistry, Remote Sens. Environ. 30, pp. 271-278.
Curran, P.J., J.L. Dungan, B.A. Macler, and S.E. Plummer, 1991, The effect of a red leaf pigment on the relationship between red edge and chlorphyll concentration, Remote Sens. Environ. 35, pp. 69-76.
Curran, P.J., J.L. Dungan, B.A. Macler, S.E. Plummer, and D.L. Peterson, 1992, Reflectance spectroscopy of fresh whole leaves for the estimation of chemical concentration. Remote Sens. Environ. 39, pp. 153-166.
Danks, S.M., E.H. Evans, and P.A. Whittacker, 1984, Photosynthetic Systems: Structure, Function and Assembly. Wiley, New York.
Gaffey, S.J., L.A. McFadden, D. Nash, and C.M. Pieters, 1993, Ultraviolet, visible, and near-infrared reflectance spectroscopy: Laboratory spectra of geologic materials. In, Remote Chemical Analysis: Elemental and Mineralogical Composition, C.M. Pieters and P.A.J. Englert, Eds., Cambridge University Press, pp. 43-77.
Gaffey, S.J., Spectral reflectance of carbonate minerals in the visible and near-infrared (0.35-2.55 ): Anhydrous carbonate minerals. J. Geophys. Res., 92, pp. 1429-1440.
Gaffey, S.J., Spectral reflectance of carbonate minerals in the visible and near-infrared (0.35-2.55 microns); Calcite, aragonite, and dolomite. Am. Mineral., 71, pp. 151-162, 1986.
Hunt, G.R., Spectral signature of particulate minerals in the visible and near infrared. Geophysics, 42, pp. 501-513, 1977.
King, T.V.V. and R.N. Clark, 1989, Spectral Characteristics of Chlorites and Mg-Serpentine Using High-Resolution Reflectance Spectroscopy. J. Geophys. Res., 94, pp. 13,997-14,008.
King, T.V.V., and W.I. Ridley, 1987, Relation of the spectroscopic reflectance of olivine to mineral chemistry and some remote sensing implications. J. Geophys. Res., 92, pp. 11,457-11,469.
King, T.V.V., 1995, Environmental Considerations of Active and Abandoned Mine Lands: Lessons From Summitville, Colorado, Editor, U.S. Geological Survey Bulletin 2220, 38 pages.
Miller, J.R., E.W. Hare, A.B. Hollinger, and D.R. Sturgeon, 1987, Imaging spectroscopy as a tool for botanical mapping, In, Imaging Spectroscopy II, G. Vane, Editor, International Society for Optical Engineering, Bellingham, WA, pp. 108-113.
Milton, N.M, W. Collins, S.H. Chang, and R.G. Schmidt, 1983, Remote detection of metal anomalies on Pilot Mountain, Randolf County, North Carolina, Econ. Geol., 78, pp. 605-617.
Milton, N.M, C.M. Ager, B.A. Eiswerth, and M.S.Power, 1989, Arsenic and selenium-induced changes in spectral reflectance and morphology of soybean plants, Remote Sens. Environ., 30, pp. 263-269.
Murray, I., and P.C. Williams, 1987, Chemical principles of near-infrared, In, Near Infrared Technology in the Agricultural and Food Industries, P. Williams and K. Norries, Eds., American Association of Cereal Chemists, St. Paul, MN, pp. 17-37.
Rock, B.A., D.J. Vogelmann, D.L. Williams, A.F. Voglemann, T. Hoshizaki, 1986, Remote detection of forest damage: Bio. Sci. 36, 439.
Swayze, G.A., and R. N. Clark, 1990, Infrared spectra and crystal chemistry of scapolites: implications for Martian mineralogy: J. Geophys. Res., 95, pp. 14,481-14,495.
FIGURE CAPTIONS
Figure 1. Schematic map showing the location of the AVIRIS flight lines in the San Juan Mountains and San Luis Valley of Colorado.
Figure 2. Spectra of muscovite and goethite measured in the laboratory and muscovite and goethite mixtures detected by AVIRIS near Alum Creek. Note the match between the AVIRIS pixel and goethite in the 1m region and the laboratory muscovite and the AVIRIS pixel in the 2m wavelength region. The gaps in the AVIRIS spectrum are wavelengths where atmospheric gases absorb.
Figure 3. AVIRIS image of the Summitville mine and surrounding area showing the distribution of minerals having absorption features in the 2.0 m-2.4 m wavelength region. The image covers an area from approximately 37o 22' 30"N to 37o 27' 30"N latitude, 106o 30'W to 106o 37'30"W longitude.
Figure 4. AVIRIS image of the Summitville mine and surrounding area showing the distribution of minerals having Fe-absorption features. The image covers an area from approximately 37o 22' 30"N to 37o 27 30"N latitude, 106o 30'W to 106o 37'30"W longitude.
Figure 5. Reference spectra used in the mapping of vegetation species. The field calibration spectrum is from a sample measured on a laboratory spectrometer; all others are averages of several spectra extracted from the AVIRIS data. Note that the noise is extremely low, comparable to the lab spectrum of the field calibration site. In Figure 5a (top), each curve has been offset from the one below it by 0.05. In Figure 5b (bottom), each spectrum has been offset by 0.04 from the one below it, except the top spectrum is offset 0.06. The offsets are cumulative, so the field calibration spectrum is offset a total of 0.18 for clarity.
Figure 6 The continuum-removed chlorophyll absorption spectra from Figure 5 are compared. Note the subtle changes in the shapes of the absorption between species.
Figure 7. The Tetracorder analysis of vegetation species. See text.
Figure 8. The Tetracorder analysis of vegetation species/communities near the Summitville mine based on the shape of the chlorphyll absorption.
Figure 9. The Tetracorder analysis of vegetation near the Summitville mine showing relative dryness of the species based on the 0.95 and 1.2 m absorption features.
Figure 10. The Tetracorder analysis of vegetation from near the Summitville mine showing differences in the senescence/stress levels.
U.S. Geological Survey,
a bureau of the U.S. Department of the Interior
This page URL= http://speclab.cr.usgs.gov/PAPERS/PAPERS/leadville99/ldv99.html
This page is maintained by:
Dr. Trude V.V. King tking@speclab.cr.usgs.gov and
Dr. Roger N. Clark rclark@speclab.cr.usgs.gov
Last modified May 24, 2000.