Introduction to Geographic Information Systems in Forest Resources
|
Introduction to Geographic Information Systems in Forest Resources |
|
|||||||||||||||||
|
|||||||||||||||||
Overview
Most of the data used in GIS are vector data converted from paper maps. Most of the data on hardcopy maps were gathered using "traditional" methods of direct measurement, such as plane and topographic surveys. The contrast to direct measurement of features is the remote sensing of features.
A large body of remotely sensed data exists, and is constantly growing. Much of the data on current hardcopy maps was developed by interpretation of aerial photographs, the most important source of remotely sensed imagery. Remotely sensed imagery can be used within a GIS for map display and analysis. Remote sensing and image processing are disciplines unto themselves, but deserve to be covered in this course because remotely sensed data is an important source of data for GIS. As more earth-observing satellites are placed in orbit, the amount of data available will increase at an advancing rate.
Measurement methods
Most of the data we use in GIS are derived from methods of direct measurement. Direct measurement is a process by which features are measured by physically placing measurement devices on the features themselves. Traditional topographic surveying is the form of direct measurement that produces the location of features used in maps and GIS. Features are measured with devices that measure length, angle, and elevation. These measurements are stored and transformed to generate numeric map data, which are input into a GIS to create vector features.
Ground-based surveys have been, and will always be, an important source of data in GIS. The geodetic control framework forms the basis of coordinate control in GIS. It is this coordinate control and registration of all themes in the GIS that allows the simultaneous display of multiple themes, or analysis of one theme against another.
These geodetic control systems have been developed by careful and accurate, ground based surveys. Without such geodetic control, registration of remotely sensed data would be impossible. Therefore, even the use of remote sensing is dependent, at least in part, on ground-based survey.
Remote sensing is a process by which objects are measured by devices that do not physically contact the objects. In GIS, remote sensing is most commonly performed with cameras (aerial photographs) or scanners which measure intensity in certain bands of the electromagnetic spectrum (satellite imagery).
Because direct measurement of objects takes a large amount of time, it is generally quite costly. Remote sensing, on the other hand, is less costly, but involves more post-processing time and effort than ground survey. Also, because remotely sensed data is not generated by direct physical measurement, there are problems with precision and accuracy usually not associated with ground survey. Furthermore, because ground survey methods have been developed over thousands of years, these methods are "tried and true." Image processing of remotely sensed data is a very new field of study, and its methodology is still being developed.
Electromagnetism is a form of energy. Electromagnetic energy has behavior described by both particle and wave physics. The function of electromagnetic energy is dependent on the wavelength and amplitude of an individual wave. Electromagnetic energy with very small wavelengths (<100 nm) has the potential to cause great damage to physical structures (gamma and X-rays). Electromagnetic energy with very long wavelengths (> 1 mm) have different effects (microwaves, radio waves). What we see as light is a very small slice of this spectrum of electromagnetic energy, specifically, within the range of 380-750 nm.
Within the visible part of the electromagnetic spectrum are the colors of the rainbow, with violet at the small-wavelength end of the visible band, and red at the large-wavelength end. In terms of energy, above the violet end of the visible band lie the ultraviolet (UV) bands, and below the red end of the visible band are the infrared (IR) bands.
All objects that are exposed to electromagnetic radiation reflect some of that energy, and absorb the rest. The energy absorbed from high-energy ultraviolet waves can cause rearrangement of the nucleic acid sequences in DNA, leading to genetic mutations. Energy absorbed in the infrared bands causes objects to heat up. Different chemicals absorb and reflect different wavelengths of electromagnetic energy.
Chemicals have color because they absorb the bands we do not see, and reflect back the bands we do see. The metabolically active chlorophyll contained in vegetation absorbs in the violet and red bands, and is therefore green in color. Chlorophyll uses the energy from that radiation to convert water and carbon dioxide to carbohydrate.
Because different objects have different chemical makeup, they reflect back specific bands of the electromagnetic spectrum. Most objects are composed of many different chemicals, and therefore absorb and reflect specific parts of the spectrum. It is theoretically possible to discern objects based on the wavelengths which they reflect. The reflected radiation from objects is known as a "spectral signature." Some objects, such as snow and ice, have very distinct spectral signatures, and are easy to discern from other objets. Other objects are less easy to discern from one another, such as different species within the pine family.
Everyone is familiar with photographic imagery. But how do photographs "work?" Paper is covered with a film of gel and specific chemicals which undergo reactions when exposed to varying intensity and wavelength of light. Most of the aerial photographs in common use are prepared from black-and-white (otherwise known as panchromatic or monochrome) films. These films react to the overall amount of visible light reflected back from the object to the camera. Objects that are highly absorptive appear dark (e.g. certain types of bare mineral soil, coniferous trees), while others are highly reflective (snow, clouds, certain types of rock), and appear light in value.
Usually, photographic paper is manufactured to capture electromagnetic radiation within the visible bands. Any visible color can be reproduced with a combination of red, green, and blue light. Color photographs are produced from films that capture reflected radiation within the red/green/blue bands. Some photographic paper is manufactured with chemicals that absorb in the near-infrared (NIR) bands. This type of photographic paper is frequently used to map vegetation, since vegetation is reactive within these bands. The film is processed to show the NIR bands in one of the visible colors. These images are known as false-color images, because they do not recreate the true colors seen by the human eye and brain.
The most valuable type of photographs used in GIS are digital orthophotographs. Digital orthophotos are scanned aerial photographs that have been corrected for optical and relief distortion with the use of digital elevation models and complex mathematical transformations. Once corrected for distortion, the images are mosaicked together and saved as a single image.
Normal aerial photographs are not distorted at the center of the photograph, but landscape features appear outwardly, radially stretched toward the edges of the frame. Landscape features that are have a high elevation difference across a short distance, such as hills and valleys, and especially landscape features that are tall and lie near the edge of the frame, appear highly distorted. In orthophotos, these features do not appear distorted, or their distortion is minimized.
The human eye is responsive to electromagnetic radiation within the 380-750 nm range. However, there are also some ultraviolet and infrared waves that penetrate Earth's atmosphere. Although they are not visible to the human eye, some of the infrared bands are visible to other species, such as certain insects and birds. Scanners have been developed that sense the amount of electromagnetic radiation reflected from objects, both in the visible and non-visible parts of the spectrum. The scanners are set to detect only small slices of the spectrum. They record the intensity within that slice as a numeric value, usually in the range of 0 - 255, known as 8-bit (2^8 = 256). Objects with low reflectance have a low value, and objects with high reflectance have a high value. That value is stored on a physical storage device, such as a memory chip, or magnetic disk or tape. Typically, scanners are configured so that they simultaneously record reflected radiation in several different bands, much like color photographic film is configured to capture simultaneously red, green, and blue bands.
The scanners are arranged in long rows, and placed aboard earth-orbiting satellites. Each scanning element is focused to detect reflected radiation from one specific place. The entire row of scanning elements is set to detect reflected radiation for one narrow patch of ground at a time. Once the scanners take their readings for the first narrow patch of ground, they move to the next patch of ground, and so on, until the entire scene has been scanned and recorded. All remote sensing satellites that do not use photographic film use electronic scanners.
You can imagine the row of scanners as a piece of paper with a thin slit cut in it. If you look through the slit, you will only see a part of your normal field of vision. However, if you take a photograph through the slit, then move the paper across your field of vision, taking another photograph with each move, you will be able to recreate the normal field of vision by cutting and pasting together the individual photographs.
When a scene has been scanned and recorded, the stored data are translated to a radio wave that is broadcasted at a specific radio frequency. Highly sensitive radio receivers on earth are set to this frequency, and receive the radio signal. The radio signal is translated back to a digital file, which can be recreated as an image on a CRT display or sent to a printer.
There are many different satellites currently in orbit, sending image data back to Earth. The new data are archived and made available (usually at a fairly high cost) soon after it is received. Data acquired from satellites in the past is archived and also available.
The data format for satellite imagery is referred to either by the name of the imagery or by the name of the satellite. This section is included just so that you get a general understanding of the different remote sensing platforms and what type of products they provide. This section is provided more for reference than for you to remember in detail.
Landsat multispectral scanner (MSS) data provide a historical record of the Earth's land surface from the early 1970's to the early 1990's.
A sample MSS Image of Boston, MA in false color (visible RGB bands are not displayed as RGB):
Here is an excerpt from the Landsat
MSS page at the USGS, describing the specific characteristics of the MSS:
Data Characteristics
Since 1972, the Landsat satellites have provided repetitive, synoptic, global coverage of high-resolution multispectral imagery. The characteristics of the MSS and TM bands were selected to maximize the band's capabilities for detecting and monitoring different types of Earth resources. For example, MSS band 1 can be used to detect green reflectance from healthy vegetation, while MSS band 2 is designed for detecting chlorophyll absorption in vegetation. MSS bands 3 and 4 are ideal for recording near-IR reflectance peaks in healthy green vegetation and for detecting water/land interfaces.
MSS Bands 4, 2, and 1 can be combined to make false-color composite images, where band 4 controls the amount of red, band 2 the amount of green, and band 1 the amount of blue in the composite. This band combination makes vegetation appear as shades of red with brighter reds indicating more vigorously growing vegetation. Soils with no or sparse vegetation will range from white (sands) to greens or browns, depending on moisture and organic matter content. Water bodies appear blue. Deep, clear water appears dark blue to black in color, while sediment-laden or shallow waters appear lighter in color. Urban areas appear blue-gray in color. Clouds and snow appear as bright white, and they are usually distinguishable from each other by the shadows associated with the clouds.
MSS scenes from Landsats 4 and 5 have an instantaneous field of view (IFOV) of 68 meters in the cross-track direction by 82 meters in the along-track direction (223 by 272.3 feet, respectively). To understand this concept, consider a ground scene composed of a single 82- by 82-m area. The scan monitor sensor ensures that the cross-track optical scan is 185 km at nominal altitude regardless of mirror scan nonlinearity or other perturbations of mirror velocity. Cross-track image scan velocity is nominally 6.82 meters per microsecond. After 9.958 microseconds, the 82- by 82-m image has moved 67.9 meters. The sample taken at this instant represents 15 meters of previous information and 68 meters of new information.
Therefore, the effective IFOV of the MSS detector in the cross-track direction must be considered to be 68 meters which corresponds to a nominal ground area of 68 meters, by 82 meters at the satellite nadir point. Using the effective IFOV in area calculation eliminates the overlap in area between adjacent pixels.
Landsats 1 through 3 provided Earth coverage similar to Landsats 4 and 5. However, the higher altitude of Landsats 1
through 3 resulted in a different swathing pattern with the IFOV being 56 meters in the cross-track direction by 79 meters in the along-track direction (183.7 feet by 259.2 feet, respectively).The resolution for the MSS sensor is shown below:
Landsats 1-3 Landsats 4-5 (meters) Band 4 Band 1 79/82* Band 5 Band 2 79/82 Band 6 Band 3 79/82 Band 7 Band 4 79/82 Band 8** 237* As a result, the nominal altitude was 920 km for Landsats 1, 2, and 3. Nominal altitude for Landsats 4 and 5 is 705 km. The resolutions are approximately 79 and 82 meters respectively.
** Landsat 3 only.Temporal Coverage
Background information and status of Landsat satellites:
Satellite Launched Decommissioned Sensors Landsat 1 July 23, 1972 January 6, 1978 MSS and RBV Landsat 2 January 22, 1975 February 25, 1982 MSS and RBV Landsat 3 March 5, 1978 March 31, 1983 MSS and RBV Landsat 4 July 16, 1982 * TM and MSS *** Landsat 5 March 1, 1984 ** TM and MSS **** in standby mode used for range and command as of December 14, 1993.
** currently operational
*** MSS data acquisition suspended in 1992Spectral Range
The MSS sensors were line scanning devices observing the Earth at a right angle to the orbital track. The cross-track scanning was accomplished by an oscillating mirror; six lines were scanned simultaneously in each of the four spectral bands for each mirror sweep. The forward motion of the satellite provided the along-track scan line progression. All five Landsats have carried the MSS sensor which responds to Earth-reflected sunlight in four spectral bands. Landsat 3 carried an MSS sensor with an additional band, designated band 8, that responded to thermal (heat) infrared radiation.
The radiometric range of bands for the MSS sensor is shown below: (Handbook, 1979 and 1984, USGS).
Wavelength Landsats 1-3 Landsats 4-5 (micrometers) Band 4 Band 1 0.5 - 0.6 Band 5 Band 2 0.6 - 0.7 Band 6 Band 3 0.7 - 0.8 Band 7 Band 4 0.8 - 1.1 Band 8 10.4 - 12.6
The Landsat Thematic Mapper, often known simply by the acronym TM, is currently used widely by scientists across various disciplines.
A sample TM image of Koichi City, Japan, displayed in true color (visible RGB bands shown as RGB).
Here is an excerpt from the Landsat TM page at the USGS, describing specific characteristics of the Thematic Mapper:
Data Characteristics
Since 1972, Landsat satellites have provided repetitive, synoptic, global coverage of high-resolution multispectral imagery. The characteristics of the MSS and TM bands were selected to maximize detecting and monitoring different types of Earth resources. For example, band 1 of TM data penetrates water for bathymetric mapping along coastal areas and is useful for soil-vegetation differentiation and for distinguishing forest types. TM band 2 detects green reflectance from healthy vegetation, and TM band 3 is designed for detecting chlorophyll absorption in vegetation. TM Band 4 data is ideal for detecting near-IR reflectance peaks in healthy green vegetation and for detecting water-land interfaces. The two mid-IR red bands on TM (bands 5 and 7) are useful for vegetation and soil moisture studies and for discriminating between rock and mineral types. The thermal-IR band on TM (band 6) is designed to assist in thermal mapping, and is used for soil moisture and vegetation studies.
Typically, TM Bands 4, 3, and 2 can be combined to make false-color composite images where band 4 represents the red, band 3 represents the green, and band 2 represents the blue portions of the electromagnetic spectrum. This band combination makes vegetation appear as shades of red, brighter reds indicating more vigorously growing vegetation. Soils with no or sparse vegetation range from white (sands) to greens or browns depending on moisture and organic matter content. Water bodies will appear blue. Deep, clear water appears dark blue to black in color, while sediment-laden or shallow waters appear lighter in color. Urban areas appear blue-gray in color. Clouds and snow appear bright white. Clouds and snow are usually distinguishable from each other by the shadows associated with clouds.
Spatial Resolution
A Landsat-4 or -5 TM scene has an instantaneous field of view (IFOV) of 30 meters by 30 meters (900 square meters) in bands 1 through 5 and band 7, and an IFOV of 120 meters by 120 meters (14,400 square meters) on the ground in band 6.
The resolution for the TM sensor is shown below:
Resolution Landsats 4-5 (meters) Band 1 30 Band 2 30 Band 3 30 Band 4 30 Band 5 30 Band 6 120 Band 7 30
Temporal Coverage
Background information and status of Landsat satellites.
Satellite Launched Decommissioned Sensors Landsat 1 July 23, 1972 January 6, 1978 MSS and RBV Landsat 2 January 22, 1975 February 25, 1982 MSS and RBV Landsat 3 March 5, 1978 March 31, 1983 MSS and RBV Landsat 4 July 16, 1982 * TM and MSS Landsat 5 March 1, 1984 ** TM and MSS* in standby mode used for range and command as of December 14, 1993.
** currently operationalSpectral Range
The TM sensor is an advanced, multispectral scanning, Earth resources instrument designed to achieve higher image resolution, sharper spectral separation, improved geometric fidelity, and greater radiometric accuracy and resolution than the MSS sensor. The TM data are scanned simultaneously in seven spectral bands. Band 6 scans thermal (heat) infrared radiation.
Spectral range of bands and spatial resolution for the TM sensor are:
Wavelength Resolution Landsats 4-5 (micrometers) (meters) Band 1 0.45 - 0.52 30 Band 2 0.52 - 0.60 30 Band 3 0.63 - 0.69 30 Band 4 0.76 - 0.90 30 Band 5 1.55 - 1.75 30 Band 6 10.40 - 12.50 120 Band 7 2.08 - 2.35 30All TM bands are quantized as 8 bit data.
The Advanced Very High Resolution Radiometer (AVHRR) is a broad-band, four or five channel (depending on the model) scanner, sensing in the visible, near-infrared, and thermal infrared portions of the electromagnetic spectrum.
Here is a composite image of Arizona, where AVHRR channels 2, 1, and 3 are red, green, and blue. The image is from NOAA-14 satellite, 1996 Nov 12 20:29 UT.
For more complete information on AVHRR data, see the AVHRR page at the USGS. Here is an excerpt:
Spatial Resolution
The average instantaneous field-of-view (IFOV) of 1.4 milliradians yields a LAC/HRPT ground resolution of approximately 1.1 km at the satellite nadir from the nominal orbit altitude of 833 km (517 mi). The GAC data are derived from an on board sample averaging of the full resolution AVHRR data. Four out of every five samples along the scan line are used to compute one average value and the data from only every third scan line are processed, yielding 1.1-km by 4-km resolution at nadir.
Temporal Coverage
Satellite Launch Ascending Descending Service Number Date Node Node Dates -------- ------ ---- ---- ------------------ TIROS-N 10/13/78 1500 0300 10/19/78 - 01/30/80 NOAA-6 06/27/79 1930 0730 06/27/79 - 11/16/86 NOAA-7 06/23/81 1430 0230 08/24/81 - 06/07/86 NOAA-8 03/28/83 1930 0730 05/03/83 - 10/31/85 NOAA-9 12/12/84 1420 0220 02/25/85 - Present NOAA-10 09/17/86 1930 0730 11/17/86 - Present NOAA-11 09/24/88 1340 0140 11/08/88 - 09/13/94 NOAA-12 05/14/91 1930 0730 05/14/91 - Present NOAA-14 12/30/94 1340 0140 12/30/94 - PresentNOAA-B launched May 29, 1980, failed to achieve orbit. NOAA-13 launched August 9, 1993, failed due to an electrical short circuit in the solar array.
Spectral Range
Band # Satellites: Satellites: IFOV NOAA-6,8,10 NOAA-7,9,11,12,14 1 0.58 - 0.68 0.58 - 0.68 1.39 2 0.725 - 1.10 0.725 - 1.10 1.41 3 3.55 - 3.93 3.55 - 3.93 1.51 4 10.50 - 11.50 10.3 - 11.3 1.41 5 band 4 repeated 11.5 - 12.5 1.30 (micrometers) (micrometers) (milliradians)
AVIRIS is a remote sensing platform which gathers 224 contiguous spectral channels (also called bands) with wavelengths from 400 to 2500 nm. The instrument flies aboard a NASA ER-2 airplane (a U2 plane modified for increased performance) at approximately 20 km above sea level, at about 730 km/hr. AVIRIS has flown all across the US, plus Canada and Europe. AVIRIS data is often known as "hyperspectral" imagery, because it gathers data in 224 bands.
The most common complaints about satellite imagery is low spatial and spectral resolution. For example, the Landsat TM gathers 7 bands of data, but the bands are wide, and gaps exist in between several of the bands. This means that certain objects will either not appear in the image, or objects that reflect in different wavelengths may appear indistinguishable from each other. AVIRIS data provides a high degree of spectral resolution, meaning that it would record different values for those two objects that would be indistinguishable to a TM sensor. The AVIRIS pixel size, at 20 m on a side, offers better than twice the spatial resolution of the Landsat TM (20 m * 20 m = 400 m^2; 30 m * 30 m = 900 m^2).
Spectral analysis becomes potentially easier the more bands are present in the image data. As with all classification schemes, the more data we have about individual objects, the easier it is to discern objects.
For more information, go to the AVIRIS home page at the Jet Propulsion Laboratory, or read about Imaging Spectroscopy at the USGS Spectroscopy Laboratory.
Here is a classified mineral map based on an AVIRIS image of Cuprite, NV:
Radar data data hold a great deal of promise in analyzing the structural characteristics of landscapes. Radar waves are scattered differentially by objects such as tree boles, branches, and leaves. Radar waves penetrate through vegetation and atmosphere, and are frequently used to generate terrain data for vegetated areas. Certain types of radar waves can also penetrate the ground to reveal old river channels.
From the Jet Propulsion Laboratory's excellent paper on Spaceborne SAR, which contains dozens of examples of the use of SAR imagery:
SAR data provide unique information about the health of the planet and its biodiversity, as well as critical data for natural hazards and resource assessments. Interferometric measurement capabilities uniquely provided by SAR are required to generate global topographic maps, to monitor surface topographic change, and to monitor glacier ice velocity and ocean features. Multiparameter SAR data are crucial for accurate land cover classification, measuring above-ground woody plant biomass, delineation of wetland inundation, measurement of snow and soil moisture, characterization of oil slicks, and monitoring of sea ice thickness.
Also, for more information, such as a FAQ, see the Alaska SAR Facility's page.
Here is an SAR image of Washington, DC, courtesy of Sandia National Labs. Note how the texture of individual features shows clearly.
SPOT (Systeme Probatoire de l'Observation de la Terre)
The French government operates a group of satellites, global network of control centers, receiving stations, processing centers, and data distributors. SPOT image data have been quite popular in the past because of the small pixel size (20 m for red, green, and near IR; 10 m for panchromatic). Although the SPOT satellite images give a high degree of spatial resolution, the images do not contain the same spectral resolution or number of bands as some of the other satellite image types.
The USGS SPOT page describes the SPOT program in detail. Here is an excerpt:
Spatial Resolution
Mode Band Resolution ____ ____ __________ Multispectral (XS) 1 (Green) 20 meters 2 (Red) 20 meters 3 (Near Infrared) 20 meters Panchromatic (P) Not Applicable 10 meters
Temporal Coverage
Launch Decommission Satellite Date Date _________ ______ ____________ SPOT 1 2/22/86 12/31/90 SPOT 2 1/22/90 SPOT 3 9/26/93 SPOT 4 Anticipating launch in 1997. SPOT 5 Anticipating launch in 2001.Spectral Range
Mode Band Micrometers ____ ____ ___________ Multispectral (XS) 1 (Green) 0.50-0.59 micrometers 2 (Red) 0.61-0.68 micrometers 3 (Near Infrared) 0.79-0.89 micrometers Panchromatic (P) Not Applicable 0.51-0.73 micrometers
This is a small section of a SPOT image.
IKONOS is one of the newest imaging platforms. It captures very high resolution imagery in both gray scale and multispectral band combinations.
The following is an excerpt from Space Imaging's description of the IKONOS platform:
The camera system for each IKONOS satellite will simultaneously collect panchromatic (gray-scale) imagery with one-meter resolution, and multispectral data (red, green, blue, and near infrared) with four-meter resolution, across an 11 km swath of the Earth’s surface. The panchromatic imagery will provide highly accurate Earth imagery, enabling geographic information system (GIS) users to generate precision maps. The multispectral data will have a variety of scientific applications, including environmental and agricultural monitoring.
Here are the statistics for the IKONOS platform:
Spatial Resolution
Mode Band Resolution ____ ____ __________ Multispectral 1 (Green) 4 meters 2 (Red) 4 meters 3 (Near Infrared) 4 meters Panchromatic (P) Not Applicable 1 meters
Temporal Coverage
Launch Decommission Satellite Date Date _________ ______ ____________ IKONOS-1 11/24/1999 N/ASpectral Range
Mode Band Micrometers ____ ____ ___________ Multispectral (XS) 1 (Blue) 0.45-0.52 micrometers 2 (Green) 0.52-0.60 micrometers 3 (Red) 0.63-0.69 micrometers 4 (Near-infrared) 0.76-0.90 micrometers Panchromatic (P) Not Applicable 0.45-0.90 micrometers
Here are a few sample images:
An overall view of part of Washington, DC. The bounded rectangle is the Jefferson Memorial.
A more close view of the Jefferson memorial. Remember that this 1-m resolution
image was taken from over 400 miles out in space!
Light Detection and Ranging is a recent addition to remote sensing. Using many rapid small bursts of laser light, an aircraft-borne apparatus records reflection from multiple sources. The first reflection is generally from the vegetation canopy. Intermediate returns can come from other vegetation structures, with the last return from the ground surface. The results, when processed by sophisticated software, result in startling and detailed images.
Because of the rapidity of the laser pulses, it is possible to generate images that have extremely high resolution, in which individual tree crowns, downed logs, stream and road beds can be seen.
Here are a few 3-D images from Capitol Forest, WA, which were derived from data contracted by the US Forest Service. Click on the thumbnails to see the full images.
Forest
canopy (1st return)
Ground
surface (last return)
These images hold a lot of promise in all types of natural resource management and analysis applications.
For more information on LIDAR in terrain mapping, see EagleScan Incorporated's home page. For information on LIDAR and atmospheric analysis, see NOAA's ETL Division's page.
Remotely sensed data are viewed much like any other raster data in ArcView 3.x. Each pixel has a reflectance value. That value is mapped to a specific color or range of color or shade in the Legend Editor. Usually single-band remotely sensed imagery is displayed in shades of gray, where pixels of low reflectance are dark, and pixels of high reflectance are light in shade. For panchromatic photographic or satellite imagery, this looks very much like a black-and-white photograph.
Multi-band images are usually displayed with a composite of red, green, and blue. For each pixel, the reflectance in a given band is mapped to a value, exactly as with panchromatic images. When three bands are displayed, each pixel is mapped to a value for each one of the basic colors of light. The individual colors blend together to form a color composite. For data displaying three 256-value bands, the total number of possible colors is over 16 million (256^3 = 16,777,216).
Here are a few images to help explain the RGB color combination.
The RBG color model uses a cube where each pure color is at one corner of the cube. Any position within the cube is defined by the combination of the red, green, and blue values. As you can see, blue is at (0, 0, 255), and magenta is at (255, 0, 255). Shades of gray are defined by the set of coordinates that have the same value for red, green, and blue.
For example, in this Color Picker from Photoshop, the combination of R=210, G=90, B=255 results in a specific shade of purple. The colors of a remotely sensed image are generated for each pixel or grid cell based on the values of the individual bands displayed.
Here are the 3 basic colors of light mixed together:
The same color mixing applies to combinations of intensities in real reflectance data. Here are the red, green, and blue bands (3 top layers) merged together on the bottom as a color composite:
Depending on which bands are mapped to which basic color, different features are discernible. Some band combinations look quite close to true-color photographic images, while others look quite different.
In ArcView 3.x, multispectral images are loaded the same way as panchromatic images. When the Legend Editor is invoked, however, the option exists to alter which band is mapped to which basic color. The display of the different bands can be altered to increase the intensity of display for a given band. The display mapping can also be narrowed or stretched in order to achieve different effects.
A histogram of pixel values and counts is displayed for each band. The histogram is shown as a curve. The straight line represents a function that maps input pixel value with output display value. In the image above, this means that most of the red band has values from about 20 to 100. However, the output values are stretched, or rescaled, from 0 to 255, based on the linear relationship represented by the straight line.
Although the different platforms have different bands, the basic process of mapping values to colors is the same. This description, from NASA's Observatorium, describes the basic methods and reasons for color composites with Landsat images.
Three-band composites are created by using the measured reflected energy in each of three Landsat Thematic Mapper (TM) spectral bands to control the amount of blue, green, and red in a color output image. The way in which the seven TM bands are mapped to the three colors in the output image depends on what information is desired to be highlighted in the image. For some applications, it may be desirable that landcover classes be associated with familiar colors, e.g., grass is green. In other cases, contrasting colors are preferred to highlight objects of interest from the background. We give three examples of commonly used band combinations and describe how different features appear in each.
Note: The specific bands used in three-band composites are often identified by giving the band numbers used for red, green, and blue, respectively. Thus, an image using band seven for red, band four for green, and band two for blue would be designated (7,4,2). We use this same convention.
- True-Color Composite (3,2,1)
True-color composite images approximate the range of vision for the human eye, and hence these images appear to be close to what we would expect to see in a normal photograph. True-color images tend to be low in contrast and somewhat hazy in appearance. This is because blue light is more susceptible than other bandwidths to scattering by the atmosphere. Broad-based analysis of underwater features and landcover are representative applications for true-color composites.
- Near Infrared Composite (4,3,2)
Adding a near infrared (NIR) band and dropping the visible blue band creates a near infrared composite image. Vegetation in the NIR band is highly reflective due to chlorophyll, and an NIR composite vividly shows vegetation in various shades of red. Water appears dark, almost black, due to the absorption of energy in the visible red and NIR bands.
- Shortwave Infrared Composite (7,4,3 or 7,4,2)
A shortwave infrared composite image is one that contains at least one shortwave infrared (SWIR) band. Reflectance in the SWIR region is due primarily to moisture content. SWIR bands are especially suited for camouflage detection, change detection, disturbed soils, soil type, and vegetation stress.
Several different file formats are used to store remotely sensed data. All of these formats are raster formats, but the structure of the files is very different. There are many formats, of which several are supported by ArcView 3.x.
ArcView 3.x supports the following image formats as themes:
By default, ArcView 3.x looks for BSQ, BIL, and BIP images when they have these file extensions, respectively: .bsq, .bil, and .bip.
ArcView 3.x supports hot linking to the following image formats:
Image processing is a term given to the analysis of remotely sensed imagery. One of the most common uses of remotely sensed data is the classification of landscape features. Because remotely sensed imagery typically cover very large areas of the earth's surface, and frequently contain a large amount of data spanning several slices of the ER spectrum, a large amount of effort has been devoted to developing methods of landscape classification and feature identification based on this imagery.
Satellite image processing is essentially the process of applying mathematical operations on pixel values. There are many, many methods used for classification and identification, of which we will cover only a few, and only with the most cursory treatment. For a more complete treatment of satellite image processing, you should look for specific courses in the subject. Satellite image processing is a very important supporting technology to GIS, but generally not considered a core technology. Most experts in GIS have some rudimentary knowledge of RS, and vice versa.
For landscape features with widely different spectral signatures, such as concrete and grass, the comparison of values is easy. However, for features that have fairly close spectral signatures (e.g., old-growth Douglas-fir/hemlock forests, and mature Douglas-fir/hemlock forests), specific techniques are still being developed.
Image classification falls into the categories of supervised and unsupervised.
In unsupervised classification, the software uses specific algorithms to attempt to lump pixels together based on similar spectral signatures. Supervised classification relies on the expertise of the computer operator. The operator uses a mouse pointer to select single pixels or groups of pixels of known land cover type. The spectral signatures for these classes are used to analyze every pixel in the image. Based on user-defined tolerances of error, the image is classified into the groups originally specified by the operator.
Ground-truthing is the process whereby the classification is rated for accuracy. Field personnel will visit specific places on the ground to verify that the computer correctly placed pixels in the correct class. Most image classifications are called good if they attain accuracy of about 85% or better.
Other methods of feature characterization have been developed to derive a characterization of an entire image. Two of these techniques we will explore are the Normalized Difference Vegetation Index (NDVI) and the Tasseled Cap indices of wetness and greenness.
NDVI Analysis
What is NDVI? NDVI is a value calculated on a pixel-by-pixel basis, using the near infrared (NIR) and red bands from multispectral satellite images. High values indicate vigorous and copious vegetation, and low values indicate non-vegetated areas or dry vegetation. The resultant values are usually associated with graduated colors in the legend.
How is NDVI calculated? NDVI is calculated for each pixel based on this function:
which for Landsat TM images translates to
Why?
The reason NDVI is related to vegetation is that healthy vegetation reflects very well in the near infrared part of the spectrum. Green leaves have a reflectance of 20 percent or less in the 0.5 to 0.7 micron range (green to red) and about 60 percent in the 0.7 to 1.3 micron range (near infra-red).
The visible channel gives some degree of atmospheric correction. The value is then normalized to the range -1<=NDVI<=1 to partially account for differences in illumination and surface slope.
How is it useful?
NDVI provides a crude estimate of vegetation health and a means of monitoring changes in vegetation over time. The possible range of values is between -1 and 1, but the typical range is between about -0.1 (NIR less than VIS for a not very green area) to 0.6 (for a very green area).
Here is a sample NDVI image of Australia:
Note that this image composite is from December, which is summer south of the Equator. Most of Australia is in the low range, with the southernmost extents (Tasmania, Sydney, Melbourne, Canberra, and Perth) having the highest greenness value.
Tasseled Cap Analysis
The Tasseled Cap transformation was developed via multivariate statistical methods in order to simplify the multispectral qualities of MSS and TM images into a single value. The indices indicate brightness, greenness, and wetness, and haze. The original article describing the Tasseled Cap analysis is Crist, E., P., Cicone, R., C. (1984). A Physically-Based Transformation of Thematic Mapper Data - the TM Tasseled Cap. IEEE Transactions on Geoscience and Remote Sensing, GE-22, 256-263.
The indices are created by multiplying the constants, listed in the table below, for each band and the pixel value in that band. This produces 7 values for each pixel. The values are then summed to obtain a single value for each pixel in the image. The higher the index value, the greater the value of the property
|
|
|||||||
| Property |
|
|
|
|
|
|
|
| Brightness |
0.3037 |
0.2793 |
0.4743 |
0.5585 |
0.5082 |
0.1863 |
|
| Greeness |
-0.2848 |
-0.2435 |
-0.5436 |
0.7243 |
0.084 |
-0.18 |
|
| Wetness |
0.1509 |
0.1973 |
0.3279 |
0.3406 |
-0.7112 |
-0.4572 |
|
| Haze |
0.8832 |
-0.0819 |
-0.458 |
-0.0032 |
-0.0563 |
-0.013 |
|
|
|
|||||
| Property |
|
|
|
|
|
| Brightness |
0.332 |
0.603 |
0.675 |
0.262 |
|
| Greenness |
-0.283 |
0.66 |
0.577 |
0.388 |
|
| Yellowness |
-0.899 |
0.428 |
0.076 |
0.041 |
|
Here is the Tasseled Cap Wetness theme, where deeper blue colors indicate "wetter" vegetation:
Tasseled Cap Wetness is an indication of the overal moistuire content of the location. This includes the vegetation as well as soil moisture.
And a Greenness grid, where the deeper greens indicate "greener" vegetation.
Greenness is an indication of the amount of green vegetation at each pixel location.
Do you see how the older stands are greener, but the younger stands are wetter?
|
|||||||||||||||||
|
The University of Washington Spatial Technology, GIS, and Remote Sensing Page is provided by the College of Forest Resources and the College of Ocean and Fisheries Sciences through Unit-Specific UIF. Site administrator: Phil Hurvitz. |
Copyright © Phil Hurvitz, 1998-2003 |
Syllabus