|6.1 Scientific Theory of Measurements|
When solar energy strikes an object, five types of interaction are possible. The energy is:
The Landsat-7 system is designed to collect 7 bands or channels of reflected energy and one channel of emitted energy. A well calibrated ETM+ enables one to convert the raw solar energy collected by the sensor to absolute units of radiance. Radiance refers to the flux of energy (primarily irradiant or incident energy) per solid angle leaving a unit surface area in a given direction. Radiance corresponds to brightness in a given direction toward the sensor, and is often confused with reflectance, which is the ratio of reflected versus total power energy. Radiance is what is measured at the sensor and is somewhat dependent on reflectance.
The eight bands of ETM+ data are used to discriminate between Earth surface materials through the development of spectral signatures. For any given material, the amount of emitted and reflected radiation varies by wavelength. These variations are used to establish the signature reflectance fingerprint for that material. The basic premise of using spectral signatures is that similar objects or classes of objects will have similar interactive properties with electrmagnetic radiation at any given wavelength. Conversely, different objects will have different interactive properties. A plot of the collective interactive mechanisms (scattering, emittance, reflectance, and absorption) at wavelengths on the electromagnetic spectrum should, according to the basic premise, result in a unique curve, or spectral signature, that is diagnostic of the object or class of objects. A signature on such a graph can be defined as reflectance as a function of wavelength. Four such signatures are illustrated in Figure 6.1
Figure 6.1 Spectral Reflectance Curves of Four Different Targets
ETM+ data can be used to plot spectral signatures although the data are limited to eight data points within the spectral range of .45 to 12.5 µm. More useful is plotting the ETM+ spectral signatures in multi-dimensional featire space. The four surface materials shown in Figure 6.1 are plotted in Figure 6.2 using just two of the ETM+ spectral bands. (GL representing grasslands, PW representing pinewoods, RS representing red sand, and SW representing silty water) may be characterized as distinct.
Figure 6.2 Spectral Separability Using Just Two Bands
Each of the materials has been plotted according to its percent reflectance for two of the wavelengths or spectral bands. When more than two wavelengths are involved, the plots in multi-dimensional space tend to increase the separability among different materials. This spectral separation forms the basis for multispectral analysis where the goal is to define the bounds of accurately identified data point clusters
|6.2 Spatial Characteristics|
Spatial resolution is the resolving power of an instrument needed for the discrimination of features and is based on detector size, focal length, and sensor altitude. More commonly used descriptive terms for spatial resolution are ground sample distance (GSD) and instantaneous field of view (IFOV). The IFOV, or pixel size, is the area of terrain or ocean covered by the field of view of a single detector. The ETM+ ground samples at three different resolutions; 30 meters for bands 1-5, and 7, 60 meters for band 6, and 15 meters for band 8. Figure 6.3 illustrates the ETM+ IFOV for bands 1-5 and 7 relative to other sensors and a football field. IKONOS, the recently launched Space Imaging sensor, has an IFOV of 1 meter. The French SPOT panchromatic sensor an IFOV of 10 meters whereas the SPOT multispectral (XS) sensor has an IFOV of 20 meters. ETM+ has an IFOV of 30 meters for bands 1-5, and 7 of 30 meters while the Indian Remote Sensing Satellite (IRS) has an IFOV of 36.25 meters.
Figure 6.3 ETM+ Spatial Resolution Relative to Other Sensors
A standard WRS scene covers a land area approximately 185 kilomenters (across-track) by 180 kilometers (along-track). A more precise estimate for actual scene size can be calculated from the 0R product image dimensions. These are listed in table 6.1
|Table 6.1 Image Dimensions for a Landsat 7 0R Product|
It is natural to assume that one could determine a scene's spatial extent by multiplying the rows and columns of a scene by the IFOV. This would lead to a scene width of 198 kilometers (6600 samples * 30 meters) and a scene length of 180 kilometers (6000 lines * 30 meters). While this calculation applies to scene length, the scene width calculation is more complicated due to the presence of image buffers and the staggered image bands in the 0R product. Left and right image buffers were placed in the 0R product to accommodate a possible increase in scan line length over the mission's life. The staggered image bands result from the focal plane design which LPS accounts for by registering the bands during 0R processing. The end result is an increasing amount of zero-fill preamble according to the band order on the ground projected focal plane array.
The detector offsets determine the amount of zero fill preamble for each band. These are listed in Table 6.2 and can also be found in the CPF. Coincident imagery for all 8 bands starts at pixel location 247 for the 30 meter bands. One need only to look at at the reverse scan odd detector offset for band 6 to see that this is true. This number, 116, is actually in 60 meter IFOVs which translates to 232 30 meter pixels. Another 14 pixels must be added to this number to account for the seven minor frames of image data pre-empted by time code. Coincident imagery for all 8 bands ends at pixel location 6333 for the 30 meter bands. This number is determined by looking at the reverse even detector offset for band 8. Add to this number the value 12,626 which represents the number of band 8 pixels per line (6313 minor frames times 2). The total, 12,666, is halved to put the ending pixel number into 30 meter units. The number of coincident images pixels in a scan is therefore 6087 (6333 - 247 + 1). The nominal width for a scene is therefore 182.61 kilometers (6087 * 30 meters).
|Table 6.2 ETM+ Detector Shifts|
|6.3 Temporal Characteristics|
Figure 6.4 Sun Synchronous Orbit of Landsat 7
A fixed mean sun time does not mean that local clock time will remain fixed for all points at a given latitude, since discrete time zones are used to determine local time throughout the world. The local time that the satellite crosses over a given point at latitudes other than at the equator also varies due to the time the satellite takes to reach the given point (nearly 99 minutes are required for one complete orbit), and the time zones crossed by the satellite relative to its equatorial crossing point.
While the orbit of Landsat 7 allows the spacecraft to pass over the same point on the Earth at essentially the same local time every 16 days, changes in sun elevation angle, as defined in figure 6.5, cause variations in the illumination conditions under which imagery is obtained. These changes are due primarily to the north-south seasonal position of the sun relative to the Earth (figure 6.6).
Figure 6.5 Sun Elevation Angle
The actual effects of variations in sun elevation angle on a given scene are very dependent on the scene area itself. The reflectance of sand, for example, is significantly more sensitive to variations in sun elevation angle than most types of vegetation. Atmospheric effects also affect the amount of radiant energy reaching the Landsat sensor, and these too can vary with time of year. Because of such factors, each general type of scene area must be evaluated individually to determine the range of sun elevation angles over which useful imagery can be realized.
Figure 6.6 Effects of Seasonal Changes on Solar Elevation Angle
Depending on the scene area, it may or may not be possible to obtain useful imagery at lower sun elevation angles. At sun elevation angles greater than 30 degrees, one should expect that all image data can be fully exploited. A sun elevation angle of 15 degrees, below which no imagery is acquired, has been established for the Landsat 7 mission.
Apart from the variability of scene effects, sun elevation angle is itself affected by a number of perturbing forces on the Landsat orbit. These include forces such as atmospheric drag and the sun's gravity. They have the effect of shifting the time of descending node throughout the year, and this results in changes to the nominal sun elevation angle. The effects of orbit perturbations, however, can be considered minor for most applications.
Repeat imaging opportunities for a given scene occur every 16 days (see Chapter 5 for details). This does not mean every scene is collected every 16 days. Duty cycle constraints, limited onboard recorder storage, the use of cloud cover predictions, and adherence to the Long Term Acquisition Plan make this impossible. The goal, however, is to collect as much imagery as possible over dynamically changing landscapes. Deserts do not qualify and thus are imaged once or twice per year. Temperate forests and agricultural regions qualify as dynamic and are imaged more frequently. Figure 6.7 illustrates archived imagery during the mission's first 112 days. Although the mission is still young, certain trends are emerging. The U.S including Alaska is quite green because every imaging opportunity is expoited. North Africa is mostly desert and appears red. Northern Asia is mostly red and yellow due to recorder constraints.
|Figure 6.7. Landsat 7 data archived during the first 112 days of operation.|
The importance of imaging dynamically changing landscapes frequently is illustrated in Figure 6.8. The image on the left was acquired over Salt Lake City on August 14, 1999 while the other was acquired four cycles later on October 17, 1999. The band combination for boths images is 5-4-2. The dramatic color changes in the mountains to the east of Salt Lake City indicate the montaine growing season is over. A multi-temporal analysis using images such as these allows one to resolve, with greater accuracy, key landscape components such as biomass, species components, and phenological growth patterns.
Figure 6.8. August 14, 1999 (left) and October 17, 1999 (right) images of the Salt Lake City area
|6.4 Radiometric Characteristics|
The ETM+ images are acquired in either a low or high gain state (Figure 6.9). Gain selection for a scene is controlled by the MOC and is performed by changing the reference voltage for the analog to digital convertor. This occurs in the preceding scene. The science goal in switching gain states is to maximize the instrument's 8 bit radiometric resolution without saturating the detectors. This requires matching the gain state for a given scene to the expected brightness conditions. For all bands, the low gain dynamic range is approximately 1.5 times the high gain dynamic range. It makes sense, therefore, to image in low gain mode when surface brightness is high and in high gain mode when surface brightness is lower. Table 6.3 lists the target/specification minimum saturation levels for all bands in both the low and high gain states.
Figure 6.9 Design ETM+ Reflective Band High and Low gain Dynamic Ranges
|Table 6.3. ETM+ Dynamic Range
watts / (meter squared * str * µm)
|Low Gain||High Gain|
|Band||Minimum Saturation Level||Minimum Saturation Level|
The currently implemented gain setting strategy for the Landsat-7 Long Term Acquisition Plan (LTAP) took effect July 13, 2000 (2000/195:22:40z) supersceding the at-launch plan and the December 2, 1999 operational decision to acquire band 4 in low gain when the sun elevation angle exceeds 45 degrees. It consists of a fixed categorization of the surface cover types of the Earth, and gain setting rules that are surface cover and sun angle based. Within the LTAP, the gains for bands 1-3 are currently always changed together as are the gains for bands 5 and 7. The earth surface categories are as follows:
1. Land (non-desert, non-ice)
5. Sea Ice
Each Landsat-7 WRS Path/Row location is categorized into one of these six types. For each surface cover type the gain setting rules are different:
1. Land (non-desert, non-ice):a) Bands 1-3 set to high gain
b) Band 4 set to high gain except where sun elevation is greater than 45°(set to low gain) - to avoid dense vegetation (reflectance > 0.66 ) saturation. (At this sun angle high gain in band 4 saturates at about a reflectance of 0.66, so switching to low gain keeps targets at this reflectance or below from saturating.)
c) Bands 5,7 set to high gain
d) Band 8 set to low gain
2) Desert:a) Bands 1-3 set to high gain except where sun elevation is greater than 28° - to avoid bright desert target (reflectance >0.65 in band 3, [>0.66 in band 1 , >0.71 in band 2]) saturation
b) Band 4 set to high gain except where sun elevation is greater than 45°(set to low gain) - to avoid bright desert (reflectance > 0.66 ) saturation.
c) Band 5,7 set to high gain except where sun elevation is greater than 38° -- to avoid bright desert target (reflectance >0.70 in band 5, [>0.68 in band 7] saturation
d) Band 8 set to low gain
3) Ice/Snow and Sea Icea) Bands 1-3 set to high gain except where sun elevation is greater than 19° - to avoid snow ice (reflectance > 0.95 in band 3, [ >0.94 in band 1, >1.03 in band 2]) saturation.
b) Band 4 set to high gain except where sun elevation is greater than 31° - to avoid snow/ice (reflectance >0.92) saturation.
c) Band 5, 7 set to high gain
d) Band 8 set to low gain
4) Water/Coral Reefs
a) Bands 1-5,7 set to high gain
b) Bands 8 set to low gain
5) Volcano/Night - nighttime imaging (sun elevation < 0) is only routinely performed for sights identified as "Volcano"a) Bands 1-4, set to high gain
b) Bands 5,7 set to low gain to reduce saturation of volcanic hot spots
c) Band 8 set to low gain
The actual saturation reflectances corresponding to a given sun angle are influenced by the Earth-Sun distance, which varies by ±1.5% over the year producing a ±3% irradiance variation. The current implementation does not incorporate this variability. Band 8 is in low gain for all routine acquisitions as the noise level in this band is such that high gain provides very little improvement in performance. This implementation is currently under review and the band 8 gain settings may be altered in the future.
Maps applying to the first day of each month are illustrated in Table 6.5. The nominal gain file (ASCII) contains a complete list of gain states, orgainized by month and WRS location, for daytime scenes.
|Table 6.5 ETM+ Monthly Gain Settings for Daytime Scenes|
|6.5 Landsat Niche|
The civilian space-based remote sensing industry is characterized by the ground footprint, spatial resolution, and the spectral channels of today's sensors. On one end of the scale are the low-resolution, large footprint, multi-spectral sensors such as NOAA's polar orbiters that have one-kilometer resolution and a 2000-kilometer swath or footprint. On the other end are high resolution, small footprint, panchromatic snapshot sensors such as IKONOS which was recently launched by Space Imaging. As depicted in Figure 6.22, Landsat 7 occupies a unqiue niche between these two extremes.
Figure 6.22 Landsat Sensing Characteristics Relative to Other Satellite Systems
The horizontal bars represent proportionately scaled footprints of the sensors on the left. Listed with each sensor is its spatial resolution, spectral coverage, and radiometric calibration accuracy. The right side of the chart lists the sensor's temporal resolution and pointing capability. Upon careful examination of this chart, one can quickly ascertain Landsat's unique niche. No other sensor can match Landsat's uncommon characteristics which include repetitive, broad-area, and global coverage at high spatial resolution in all four passive optical regions of the electrmagnetic spectrum (i.e. visible, near IR, short-wave IR, and thermal IR regions), and accurate radiometric calibration. In adddition, Landsat's retrospective archive stretches back 25 years.