Three Decades of Landsat Instruments Aram M. Mika

Abstract This paper traces the development history of the multispec- tral sensors for the Landsat series of satellites, from the first Multispectral Scanner aboard Landsat 1 to the latest variant of the Enhanced Thematic Mapper for Landsat 7. For each sensor, we begin with an overview of the design objectives and program context in which the instrument specifications were established. This is followed by a design description that outlines the operation of each sensor and highlights key technology features. The discussion for each of these instru- ments is concluded by a performance summary and opera- tional history.

Introduction The multicolored Landsat images that adorn the walls and research papers of nearly everyone in the Earth-science com- munity originated from a handful of spaceborne sensors de- veloped over the past three decades, and several of these sensors are still operating today. These instruments have transformed the way that we look at the Earth and have spawned a new remote-sensing industry that is poised to flourish in the years ahead.

While Landsat instruments are fundamentally just electro- optical transducers that ingest photons and eject a digital bit stream, this transduction relies upon the state of the art in nu- merous technologies including optics, precision electrome- chanics, detectors, advanced materials, cryogenics, and signal processing. The development of these transducers has fol- lowed a path of progressive sophistication that has exploited advances in these technologies. Each successive sensor-devel- opment effort has balanced technical risk, performance, and reliability to meet increasingly demanding mission objectives, but this technical progression has been tempered by budget and schedule pressures that have often played a decisive role in selecting the payload design, particularly for Landsats 6 and 7. Nevertheless, infusion of updated technology has added new internal capability and refinement to instruments that may appear outwardly similar. Spatial resolution, spectral coverage, radiometric sensitivity, calibration accuracy, and re- liability have all been upgraded over the years.

The Multispectral Scanner Design Objectives The concept for an Earth-resources technology satellite took shape in the early 1960s following the success of early weather-observation spacecraft such as TIROS - the Televi- sion-Infrared Observation Satellite. The U.S. Department of Agriculture (USDA), along with the Department of the Interior (DOI), began working with NASA to define the instrumentation for a satellite tailored for observation of the solid Earth.

Principal interests of the U ~ D A included synoptic moni- toring of agricultural activity to provide additional data for acreage control and management of support payments. Forest inventories and urban land-use assessment were additional topics of interest. Concurrently, DOI was interested in photo- geology, orthophoto mapping, photogrammetry, and map compilation. At the same time, NASA was engaged in a series of laboratory and ground studies, airborne measurements from a multispectral scanning instrument, high-altitude false- color infrared photography, as well as Earth-observation ex- periments on the Gemini and Apollo missions.

All of these activities crystallized in the definition of the payload instruments for the Earth Resources Technology Sat- ellite (ERTS), later renamed Landsat 1. The ERTS payload in- cluded two principal sensing instruments: the Return-Beam Vidicon (REW) system and the Multispectral Scanner (MSS). The R B ~ system, consisting of three coaligned television cam- eras, was initially considered to be the primary sensor on ERTS 1, while the MSS would serve as a secondary instru- ment. However, concerns about the geometric fidelity and ra- diometric repeatability of the R ~ V , coupled with the outstanding performance of the ms, changed the operating protocol once the spacecraft was in orbit: the MSS became the primary imaging instrument on ERTS in fairly short order. The spatial, spectral, and radiometric performance require- ments for both of these payload instruments are summarized in Table 1.

Spatially, both the REW and MSS on Landsats 1 and 2 provided a resolution (instantaneous field-of-view) of approx- imately 80 metres. Spectrally, the three-camera REW system was designed to cover the blue-green, yellow-red, and red/ near-infrared regime, while the MSS was slated to provide a similar band set, along with a fourth band to extend coverage further into the near-infrared spectrum. A later variant of the RBV system for Landsat 3 utilized two cameras with panchro- matic spectral response and higher spatial resolution (40 m) to complement the multispectral coverage provided by the MSS.

Upon closer examination, the spectral-band specifica- tions for the MSS seem rather curious (with benefit of 25 years of hindsight): the band selections progress in multiples of tenth-micrometre intervals. Surely, nature is not so coop- erative as to arrange the reflective spectral features of vegeta- tion and minerals at these uniform wavelength increments. In actuality, the MSS (and original ~ s v ) bands were selected from a very practical standpoint: to provide image products that would approximate false-color infrared aerial photogra- phy film. Because researchers already had some experience in evaluating and interpreting such images from airborne platforms, a spaceborne extension of that capability seemed to be a natural progression; at the time, there was a very lim-

Hughes Aircraft Company, 2000 E. El Segundo Boulevard, Bldg. El, MIS 150, El Segundo, CA 90245.

The author is presently with Lockheed Martin Missles and Space, 1111 Lockheed Martin Way, Bldg. 101, Sunnyvale, CA 94089.

Photogrammetric Engineering & Remote Sensing, Vol. 63, No. 7, July 1997, pp. 839-852.

0099-1112/97/6307-839$3.00/0 O 1997 American Society for Photogrammetry

and Remote Sensing

PE&RS July 1997 839

ited experience base for interpreting and analyzing images with other spectral-band combinations.

At any rate, the MSS progressed from design studies to subsystem-technology demonstrations to a full-scale flight- hardware development program commencing in 1967.

Design Description The M s s is a scanning multispectral imaging radiometer that produces radiometrically accurate images of the Earth utiliz- ing a scanning system that covers a 185-km swath across the orbital path of the satellite. This is accomplished by employ- ing an object-space scan mirror - so called because it is placed in front of the sensor's telescope, in the same "space" as the objects to be viewed. Operation of the scanner is dia- grammed schematically in Figure l . Light from the Earth is directed by the scan mirror through a reflective telescope to a fiber-optic array at the focal plane. The MSS operates by re- peatedly scanning this 24-element fiber-optic array from west-to-east across the Earth, while the orbital motion of the spacecraft provides a natural north-to-south scanning motion. The fiber-optic bundle at the focal plane fans out to six de- tector elements for each of the four spectral bands: 18 of these are photomultiplier tubes, while six are discrete silicon photodiodes.

The state of the art in 1966 technology drove the design decisions that led to the ~ s s ' configuration. At that time, in- tegrated detector arrays had not yet been developed, so fiber optics (very modern in 1966) were the only practical avenue for achieving relatively high effective detector density at the focal plane - a requisite for high spatial resolution. Because detectors were large, expensive, and required extensive sup- port electronics (especially in the case of photomultiplier tubes with their high-voltage power supplies), design trade- offs were conducted to minimize the number of detectors that would be required. In the limiting case, the ~ s s could have been designed with only one detector per spectral band, but such an approach would have resulted in unwork- able scan rates and electronic bandwidths. A sensor design with six detectors per spectral band proved optimal in the context of late 60's technology.

The scan mirror assembly was the key to providing wide-field, high-resolution coverage. The mirror follows a "sawtooth" scanning waveform at a frequency of 13.5 Hz (74-millisecond period) with active imaging taking place only during the forward scan. The net scan efficiency - the fraction of the total scan period devoted to active imaging - is 45 percent. The balance of the scan period is occupied by the retrace interval and settling time required for the mirror to return to high-fidelity linear motion.

Controlling the scan mirror to provide a precise, linear scan was a substantial technical challenge. The scan mirror had to be structurally rigid in order to maintain excellent dy- namic flatness to preserve optical performance, yet with min- imal mass and rotational inertia to accommodate the rela- tively high scan frequency. These conflicting requirements were satisfied by use of lightweight beryllium for the scan mirror: beryllium is an extraordinarily rigid and low-density metal, and mass was further reduced by removing additional material in the core of the mirror by using electric-discharge machining.

Mindful of the concerns regarding the reliability of elec- tromechanical subsystems, the MSS scan mirror was designed with flexure pivots in lieu of conventional bearings. These flex pivots are, in effect, torsional springs with high radial stiffness. When designed and utilized properly, they will ex- hibit infinite fatigue life, so that the mechanical elements of the scan mirror are not life limiting. Indeed, reliability analy- ses indicated that the servo-drive electronics would ulti- mately prove to be the life-limiting factor, and successful

Scan

Per Band

(b) Figure 1. MSS scanning approach. (a) Object-space scan mirror scans image west-to-east across track while orbital motion provides north-south scan. (b) Fiber-Optic array re- lays scanned image from focal plane to 24 discrete de- tectors. Active imaging only takes place in one scan direction.

year-after-year of on-orbit operation of these scan mirrors (without a failure to date) has proven this point.

While the design of the scan-mirror assembly was eso- teric for its day, the optical design of the ~ s s was quite straightforward, utilizing a Ritchey-Chretien telescope form with a 22.9-cm aperture diameter and a focal ratio of ft3.6. The Ritchey-Chretien is a high-performance variant of the conventional Cassegrain design: the Cassegrain utilizes a par- abolic primary mirror, while the Ritchey-Chretien employs a hyperboloidal figure for both the primary and secondary mir- ror to provide excellent resolution over a larger field-of-view. This latter feature is important for sensors that have ex- tended focal planes with multiple detectors - in this case, the %-element fiber-optic bundle of the Mss . The telescope mirrors are made of fused silica (quartz) glass and held in alignment by an optical metering structure made of Invar - a low-expansion steeltnickel alloy.

The signal-processing chain for the M s s is straightfor-

July 1997 PE&RS

ROTATINO SHIliTER

W N CALIBRATE MIRROR

SCAN MONITOR

TODIODE DETECTORJPREAWP

(b) Figure 2. Multispectral scanner hardware. (a) External view of MSS flight hardware. (b) Artist's cutaway view of MSS internal details.

ward, consisting of preamplifiers, analog-to-digital convert- ers, and a multiplexer that merges the digitized data from the 24 detector channels into a 15-megabit-per-second serial bit stream.

All of this hardware was packaged into a sensor assem- bly with dimensions of 53 by 58 by 127 cm, having a mass of 64 kg, and drawing 50W of power during imaging opera- tion, as illustrated in Figure 2.

Key performance specifications for the MSS include 80- metre spatial resolution and 6-bit radiometric quantization Although this performance is moderate by today's standards, global multispectral space imagery of that caliber was un- precedented in 1972, and the success of the M s s on Landsat 1 led to similar instruments aboard Landsats 2 and 3 and, later, to the genesis of the Thematic Mapper on Landsats 4 and 5. The radiometric performance for several delivered flight instruments is summarized in Table 2.

A total of six Multispectral Scanners were built. The first, a non-flight engineering model, is now on display at the Smithsonian's Air and Space Museum. The first flight model for ERTS was delivered in late 1971 and subsequently launched in 1972, and the second flight model was delivered two years later. Flight model three incorporated a thermal-in- frared spectral band in addition to the four visible and infra- red bands. This upgrade was initially slated for the second flight model but was deferred to flight three in order to pre- serve the launch schedule for the second spacecraft. The 10.4- to 12.6-ym thermal-infrared band on M s s 3 was designed to provide about 240-m spatial resolution by using two mercury-cadmium-telluride detectors that were radia- tively cooled to 90°K. The thermal band proved problematic and ultimately saw little use on orbit due to failure of one of the detectors and recurring condensation of moisture on the radiative cooler.

The MSS design was subsequently modified for the Land- sat D and D' program (Landsat 4 and 5). In contrast to Land- sats 1 through 3, which were designed to fly in a 909-km orbit, Landsat 4 was slated to fly in a 705-km orbit, so the telescope and scan-servo designs were modified to accommo- date the different altitude and resulting change in ground- track velocity (and, hence, scan period). The thermal infrared band that had been incorporated experimentally on Landsat 3 was dropped from the ~ s s on Landsats 4 and 5 because this spectral coverage would be provided by the new The- matic Mapper.

As an aside, the numbering convention for M s s spectral bands was also changed in the transition from the Landsat 1, 2, 3 series to the Landsat 4 and 5 series: in the former, the MSS bands were numbered 4 through 7 (4 through 8 for

TABLE 1. ERTS (LANOSAT 1) PAYLOAD SPEC~FICAT~ONS

Instrument Spatial Spectral Radiometric

Return-Beam Vidicon (RBV) 80-m IFOV (instantaneous Three spectral bands: System field-of-view)

1. 0.48 - 0.58 pm [Three coaligned cameras; one 185-km by 185-km 2. 0.58 - 0.68 pm for each spectral band.] Framing Cameras 3. 0.70 - 0.83 pm

Multispectral Scanner (MSS) 79-m IFOV Four Spectral Bands:

185-km Swath Scanning 4. 0.5 - 0.6 pm Sensor (Continuous strip 5. 0.6 - 0.7 pm image) 6. 0.7 - 0.8 pm

7. 0.8 - 1.1 pm

Analog video signal transmitted.

33-dB signal-to-noise ratio in bands 1 and 2; 30-dB in Band 3, all at max radiance (highlights). Digital video transmitted:

6 bits per pixel, linear coding; logarithmic coding also available on bands 4, 5 and 6.

I PE&RS July 1997 841

I

TABLE 2. SIGNAL-TDNOISE RATIO MEASUREMENTS FOR DELIVERED FLIGHT HARDWARE

Spectral Band

Spec MSS 1 - 3 MSS - 1

Spec MSS MSS - 2 4 - 5 MSS - 4

129 59 69 98 5 7 74 76 3 7 62

130 57 84 - - --

TABLE 3. MSS LAUNCH A N D OPERATING HISTORY

MSS Spacecraft Launch Date Operating Life Remarks

Landsat 1 (ERTS) Landsat 2

Landsat 3

Landsat 4

Landsat 5

-

23 Jul 72 5.5 years Deactivated after 5.5 years. 19 Jan 75

05 Mar 78

16 Jul 82

8.5 years 5.5 years

14.7+ years Instrument still fully functional; spacecraft on standby. Active operation for first -7 years.

01 Mar 84 13.1+ years Instrument and spacecraft still functional for direct- downlink service. MSS only activated when requested by customers; TM in daily use.

TABLE 4. THEMATIC MAPPER SPECTRAL BANDS WERE SELECTED FOR SPECIFIC APPLICATIONS

Color Application

Blue 0.45 - 0.52 Soillvegetation discrimination, deciduous/coniferous forest differentiation, clear-water bathymetry

Green 0.52 - 0.60 Growthlvigor indication for vegetation, sediment estimation, turbid-water bathymetry

Red 0.63 - 0.69 Crop classification, ferric iron detection, ice and snow mapping Near Infrared (NIR) 0.76 - 0.90 Biomass surveys, water-body delination Shortwave Infrared (SWIR) 1.55 - 1.75 Vegetation moisture, snow-cloud differentiation Shortwave Infrared (SWIR) 2.08 - 2.35 Hydrothermal mapping, rocklsoil type discrimination for mineral and

petroleum geology Thermal Infrared (TIR) 10.4 - 12.5 Thermal mapping, plant stress, urbanlnon-urban land-use differentiation

Landsat 3) because the RBV bands were designated as bands 1 through 3; in the latter, the MSS bands were renumbered as bands 1 through 4. This change in conventions bears watch- ing when comparing historical data from different Landsat satellites.

Operational History The first launch of the M S ~ occurred on 23 July 1972 aboard ERTS, and the results were spectacular. Originally considered to be an experimental payload with a one-year operating-life requirement, the first MSS operated for 5.5 years before being deactivated. Indeed, all of the MSS instruments have been ex- traordinarily long lived, as noted in Table 3. Landsats 2 and 3 have been decommissioned, but MSS instruments aboard both Landsats 4 and 5 are fully functional: Landsat 5 is in daily use at this writing, but assorted spacecraft infirmities have consigned Landsat 4 to standby status (the payload in- struments are in fine working order, but the spacecraft's power, communication, and other subsystems on Landsat 4 are marginal).

The Thematic Mapper

Design Objectives Even as the Multispectral Scanner was under development, the Earth-science community was beginning to define the re- mote-sensing objectives for the next generation Landsat in- strument. This definition process was catalyzed by the

launch of Landsat 1: on-orbit experience with the M s s from research studies such as LACIE (the Large-Area Crop Inven- tory Experiment), mission-requirements studies, and a num- ber of seminars and workshops engaging the Earth-science community led to the specifications for the Thematic Map- per. The Thematic Mapper was named because its images would be used to produce maps tailored to different Earth- observation themes, such as agriculture, hydrology, geology, and the like.

The TM represented a dramatic advancement in every di- mension of sensor performance: spatial, spectral, and radio- metric. Spatially, the TM would have a 30-m ground reso- lution in nearly all of its spectral bands - a factor of seven improvement in resolved area over the MSS [30 by 30 m (900 square metres) for the TM versus 80 by 80 m (6400 square me- tres) for the MSS], while covering the same 185-kin (100-nauti- cal-mile) swath width. In addition to revealing far greater detail in urban areas, the improved spatial resolution would permit much more accurate agricultural monitoring and crop classification of smaller fields under cultivation - a key to ac- curacy in many regions of the world where agricultural par- cels are much smaller than in the United States.

Spectrally, the TM would provide enhanced spectral coverage and spectral resolution. Unlike the MSS band set chosen by analogy to color infrared film, the TM bands were selected on the basis of comprehensive study and analysis of spectral reflection features for a variety of vegetation types and surface minerals. Spectral classification accuracy was a key determinant for selecting the specific band edges and

July 1997 PE&RS

TABLE 5. THEMATIC MAPPER DESIGN REQUIREMENTS

Spatial Radiometric Spectral Range Resolution Resolution

(pm) [m) (% NEAp) Band

1 0.45 - 0.52 30 0.8 2 0.52 - 0.60 30 0.5 3 0.63 - 0.69 30 0.5 4 0.76 - 0.90 30 0.5 5 1.55 - 1.75 30 1.0 7 2.08 - 2.35 30 2.4 6 10.4 - 12.5 120 0.5K NEAT

bandwidths. Initially, a total of six spectral bands were spec- ified for TM: three in the visible spectrum, one in the near infrared, one in the shortwave infrared, and one longwave (thermal) infrared (the latter spectral band would have a re- duced spatial resolution - 120 m). Each of these bands was selected with specific applications and discrimination capa- bilities in mind, as summarized in Table 4. Note the striking

contrast with the uniform, tenth-micrometre intervals for the MSS bands - the TM bands are anything but arbitrary.

This band selection process was difficult and conten- tious, with each scientific discipline lobbying for the band set that would be best suited for its research interests. In- deed, the process continued even beyond the beginning of the TM hardware development program. The geological com- munity felt that the TM band selections were biased in favor of agricultural/vegetation applications - to the detriment of geology and mineralogy. This case was pressed until a sev- enth spectral band (2.08 to 2.35 pm in the shortwave infra- red) was added to the TM specifications under a contract modification. By this time, development had proceeded to a point where the numbering scheme for the f is t six bands was already well entrenched, and this explains one of the enduring trivial oddities of TM: the bands are not numbered in strict order of increasing wavelength: band 7 is out of se- quence because it was added after design was already well underway.

The TM specifications also included a substantial im- provement in radiometric performance. Quantization to 256 levels (8 bits) was specified, corresponding to about 0.5 per-

Optical Scan-Line

Cooled Detectors and Flners (Bands 5-7)

Prime focal Plane

Cooled Focal Plane Bands 5.6. 7

Track Direction

Bands 1 2 3

30m

Scan .)Direction '17-We14 I B MSC

(b) Figure 3. Thematic Mapper scanning approach.

PE&RS July 1997 843

cent NEAp in key spectral bands (compared to 6 bits for MSS). The added sensitivity was needed to detect subtle re- flectance differences that were crucial for improved classifi- cation accuracy. Overall performance requirements for TM are summarized in Table 5.

Design Description and Performance Like the MSS, the Thematic Mapper is a cross-track scanning instrument utilizing an object-space scan mirror, but a num- ber of significant technological advances mark the TM as a second-generation instrument: bidirectional scanning, higher-density detector arrays employing small-scale integra- tion, extensive use of composite materials, as well as ex- tended spectral coverage, higher spatial resolution, and 8-bit

As presented in Figure 3, the TM'S bidirectional scan mirror sweeps the detector's line of sight in west-to-east and east-to-west directions transversally across track, while the spacecraft's orbital path again provides the north-south mo- tion. The bidirectional scan gives rise to a higher mechanical scan efficiency of 85 percent (versus 45 percent for the MSS) and a more moderate acceleration profile during mirror re- bound. The latter factor results in improved linearity, less vi- bration, and reduced mechanical stress, while the improved scan efficiency translates into additional detector dwell time that can be used to enhance radiometric sensitivity and/or improve spatial resolution.

Bidirectional scanning is not as simple as it initially ap- pears because the compound effect of along-track orbital mo- tion and cross-track scanning leads to significant overlap and underlap in ground coverage between successive scans. This problem was solved by employing a synchronous image-mo- tion-compensation system utilizing a pair of oscillating mir- rors in the optical path to introduce a compensatory motion that offsets the along-track orbital motion of the spacecraft. This phase-locked mechanism, called a scan-line corrector, rectifies the scan motion so that successive scan lines are par- allel, without overlap or underlap (as illustrated in Figure 4).

The scan mirror and associated servomechanism for the TM was even more challenging than that for the MSS because of the TM mirror's larger size (53 cm) and more stringent dy-

Spacecraft Travel I t Scan

a) Uncompensated

b) Correction For Orbital Motion

c) Compensated I

Figure 4. Scan-line corrector produces parallel scans in both directions.

namic flatness and scan-linearity requirements due to the TM's higher spatial resolution. Beryllium was again the mate- rial of choice because of its stiffness-to-weight ratio, but a more sophisticated lightweighting scheme was used to mini- mize mass and inertia. A solid billet of beryllium was cut into halves, each of which was electric-discharge machined with identical egg-crate patterns and then brazed back to- gether. A flex-pivot suspension was again utilized for the scan mirror, in conjunction with a magnetic-compensation system that effectively canceled the spring forces of the piv- ots. The mirror's dynamics were controlled by a fully redun- dant digital microprocessor control system.

The larger aperture, longer focal length, and higher reso- lution of the TM required special consideration to minimize thermal distortion in order to meet optical performance re-

July 1997 PE&RS

TABLE 6. MEASURED RADIOMETRIC AND SPATIAL PERFORMANCE FOR THEMATIC MAPPER FLIGHT MODELS I S SIGNIFICANTLY B ~ E R THAN SPECIFICATIONS. ' [A) Thematic Mauper Radiometric Performance Measurements

Scene Radiance Signal-to-Noise Ratio (SNR) Noise-Equivalent (specified) Raflectance Difference

[mW/cm2 - sr) At Minimum Scene Radiance At Maximum Scene Radiance N U p (%I

Measured Measured Measured Performance Performance Performance

Landsat Landsat Landsat Landsat Landsat Landsat Band Min Max Spec. 4 5 Spec. 4 5 Spec. 4 5

1 0.28 1.00 32 5 2 60 75 143 143 0.8 0.16 0.16 2 0.24 2.33 3 5 60 59 170 2 79 234 0.5 0.18 0.21 3 0.13 1.35 26 48 46 143 248 215 0.5 0.20 0.23 4 0.19 3.00 3 2 35 46 240 342 298 0.5 0.19 0.22 5 0.08 0.60 13 40 3 5 75 194 175 1.0 0.23 0.25 7 0.046 0.43 5 2 1 2 8 45 164 180 2.4 0.41 0.37 6 300K 320K NEAT= NEAT= NEAT= NEAT= NEAT= NEAT=

0.5K 0.12K 0.13K 0.42K 0.10K 0.11K

(13) THEMATIC MAPPER SPATIAL-RESOLUTION PERFORMANCE MEASUREMENTS

Square-Wave Response at Nyquist Frequency [Band Average)

Measured

Band Specified Landsat 4 Landsat 5

quirements. A Ritchey-Chretien design with a 40.6-cm aper- ture and 244-cm focal length ( f / 6 ) was utilized, with mirrors made of ultra-low expansion (ULE] titanium-silicate glass. The primary mirror was extensively lightweighted by using an egg-crate structure. Further, graphite-epoxy composite ma- terial was utilized for the optical metering structure: both the metering structure and ULE mirrors have extraordinarily low coefficients of thermal expansion in order to maintain pre- cise optical alignment through temperature changes.

On the focal plane, detector technology had progressed to the point that solid-state detector arrays could be em- ployed in all spectral bands. Sixteen-element monolithic sili- con photodiode arrays were used for bands 1 through 4. During early design tradeoffs, charge-coupled-device detector arrays were also considered for the silicon bands, but the de- velopment cost and risk was considered prohibitive at the time of the TM's initial gestation.

Monolithic indium-antimonide photodiode arrays were specified for the shortwave infrared bands (bands 5 and 7), and the thermal infrared band utilized photoconductive mer- cury-cadmium-telluride detectors. All of these infrared detec- tors, which must operate at cryogenic temperatures, were cooled to 90°K by a two-stage passive radiative cooler based upon design techniques derived from the highly successful coolers developed for the Visible-Infrared Spin-Scan Radiom- eter (VISSR) for geostationary meteorological satellites.

Low-noise, wide-bandwidth analog electronics - a com- bination that is often mutually exclusive - were required for the TM'S analog signal processing. These circuits, along with the analog-to-digital converters and high-speed multiplexers, made extensive use of hybrid integrated circuits in order to maximize performance while minimizing size, mass, and power consumption. Over 250 hybrids were used in the TM electronics.

Structurally, graphite-epoxy composites were used throughout the instrument, along with Invar and beryllium. The overall size of the instrument is 2.0 by 1.1 by 0.7 m, with a mass of 258 kg and maximum power consumption of 335 W. Figure 5 shows external and cutaway views of the TM.

Three Thematic Mapper instruments have been built: a ground-based engineering model, a protoflight model for Landsat 4, and a flight model for Landsat 5. The flight instru- ments have performed markedly better than their specifica- tions: for example, measured NEAp is two to five times better than specified, while NEAT is about four times better than specified, as noted in Table 6 .

Operational History The protoflight TM was launched aboard Landsat 4 on 1 6 July 1982, and the second TM was launched aboard Landsat 5 on 1 March 1984. As of this writing, both of these instru- ments are fully functional. Indeed, the Landsat 5 TM is in daily use providing service to Landsat receiving stations throughout the world. These instruments were specified to have a two-year operating life, with a design goal of three years. To date, they have exhibited a combined longevity of over 27 instrument years - nearly seven times their speci- fied life, as summarized in Table 7 . At this writing, the Landsat 5 TM, for example, has provided over 28 million im- ages to ground receiving stations throughout the world (My- lod, personal communication, 1997). Further, it has not yet been necessary to activate any of the internally redundant systems on either of these instruments. i Enhanced Thematic Mapper (ETM) for Landsat 6 Design Objectives 1 The definition of the Enhanced Thematic Mapper payload for Landsat 6 was the result of a complex series of decisions that were ultimately driven by policy and economics, rather than technology. The story is somewhat convoluted, but its telling is necessary in order to reveal the origins of the specifica- tions for the Enhanced Thematic Mapper - indeed, for the very existence of the ETM.

The first Thematic Mapper was developed in the context of the Landsat D program - a program that originally encom- passed the procurement and launch of four identical satellites that would each carry a Thematic Mapper and Multispectral Scanner. In the same context, long-range plans (extending be- yond the Landsat D series) called for an advanced Landsat system canying third-generation instruments employing new technology. Concept definition for this third-generation system

TM Operating

Launch Design-Life Life to Date Spacecraft Date Requirement [April 19951 Remarks

Landsat 4 16 Jul 82 2 years 14.7+ years Instrument still fully functional; spacecraft on standby. Active operation for first -5 years.

Landsat 5 01 Mar 84 2 years 13.1+ years Instrument fully functional; spacecraft still in daily use.

began even as the Thematic Mapper was under development. NASA and the nascent space remote-sensing community recog- nized that the utility of Landsat data would be further en- hanced by improvements in spatial resolution and radiometric performance.

There was a broad consensus that this performance im- provement would be delivered by pushbroom Multispectral Linear Array (MLA) sensor technology: the MLA approach uses linear detector arrays that span the entire cross-track field of view so that an east-west scanning mirror is not re- quired. Only the natural orbital motion of the spacecraft is needed to provide a north-south scan of the detector array along the ground track - in a fashion that is reminiscent of sweeping with a pushbroom, as illustrated in Figure 6.

The MLA concept offered the promise of improved per- formance because of the increase in detector dwell time made possible by proliferation of detector elements and elim- ination of the scan mirror. This is, in essence, a classic serial versus parallel tradeoff: many parallel detector channels are utilized to simultaneously view the entire cross-track scene, in contrast to a scanning sensor where a relatively small number of detectors serially sample the cross-track field-of- view. The resulting increase in detector dwell time permits improvements in spatial, spectral, and radiometric resolu- tion.

The feasibility of this sensor concept was driven by the maturation of detector technology: by the mid to late 1970s, large-scale integrated circuit technology applied to the pro- duction of detector arrays made it feasible to produce extended linear focal planes at reasonable cost and technical risk. In this context, NASA sponsored several MLA instrument definition studies in 1981 to develop the advanced sensor concepts that would supplant the TM in future Landsats. Four "Phase-B" instrument definition studies were com- pleted in December 1981 as a prelude to a hardware-devel- opment program. However, budget and policy issues quickly came to the forefront to change the course of events.

Encouraged by the success of Landsats 1 through 3 and motivated by a desire to transition Landsat to commercial operation, the Administration canceled procurement of two of the four Landsat D series of spacecraft and suspended gov- ernment sponsorship of MLA sensors for Landsat. Over the next three years, from 1981 through 1984, the Administra- tion devised a commercial transition plan that would gradu- ally shift Landsat from the public sector to commercial operation. This transition included passage of the Land Re- mote Sensing Act of 1982 (Public Law 98-365) that author- ized this action to proceed under the auspices of the Department of Commerce.

The Department of Commerce subsequently issued a re- quest for proposals to commercialize the Landsat system over a period of 1 2 years, albeit with a significant initial gov-

ernment investment to help underwrite initial development costs until the revenue stream from future sales of Landsat data could sustain the commercial enterprise. A number of proposals were received from industry and several of these proposals included MLA sensor technology for future Land- sats.

Unfortunately, the proposals languished because budget constraints made it impossible for the Department of Com- merce to proceed with the procurement as originally planned. The budget for Landsat was subsequently reduced by a factor of two, whereupon some of the bidders withdrew. The reduced budget significantly increased the financial in- vestment and risk associated with fielding new technology for the next generation of Landsats. In this context, utiliza- tion of previously developed technology - based upon the Thematic Mapper - emerged as the most workable path for providing continuity of Landsat data, and the Earth-Observa- tion Satellite Company (Eosat - then a joint venture of Hughes Aircraft and RCA) was awarded a contract in 1985 to proceed with the development of Landsats 6 and 7.

Eosat had originally proposed a dual-sensor spacecraft, with an advanced MLA instrument alongside a TM - a com- bination of old and new sensors providing an orderly transi- tion to the advanced higher-performance instrument, much as NASA had done with the MSS and TM on Landsats 4 and 5. When the procurement budget was reduced, the MLA devel- opment had to be dropped, and Landsats 6 and 7 would then carry just a Thematic Mapper. However, Landsat 6, with its 30-metre spatial resolution, would be operating in the late 1980s and early 1990s when competitive systems would be providing 10-metre panchromatic and 20-metre multispectral data. Consequently, Eosat proposed to add a 15-metre resolu- tion panchromatic band to the TM, and to modify the high- speed multiplexer to permit simultaneous transmission of multispectral and panchromatic data. The combination of ETM'S extended spectral coverage (including SWIR and TIR), broader swath width, precision radiometry, and excellent op- tical performance at a 15-metre resolution would be highly competitive while still providing fully compatible data conti- nuity for Landsat's established (and growing) community of users. Thus, the Enhanced Thematic Mapper was born.

Key specifications for the ETM are summarized in Table 8. Note that the ETM would produce two 84.9-MBPS data streams during its operation, in contrast to the single 84.9- MBPS data stream hom the TM. This was due to the addition

Figure 6. Pushbroom concept utilizes linear detector arrays scanned by orbital motion.

July 1997 PE&RS

TABLE 8. ENHANCED THEMATIC MAPPER (ETM) DESIGN REQUIREMENTS bands was carried forward without change. Indeed, spare

Spatial components from the Landsat 415 program were available for Performance Radiometric On 6'

Spatial (MTF at Sensitivity There were significant changes to the electronics to ac- Specbal Range Resolution Nyquist (SNR) commodate the addition of the panchromatic band, provide

Band ( ~ m ) (m) Frequency) [high-gain state] improved radiometric performance, and upgrade redundancy for greater predicted reliability. The addition of a second 85- %bit quantization,

8 bits transmitted MBPS multiplexer to handle the added data rate, due to the P 0.50 - 0.90 1 5 0.124 15 pan band, required repackaging the electronics into two mod- 1 0.45 - 0.52 30 0.275 32 ules: the main electronics module mounted on the scanner as- 2 0.52 - 0.60 3 0 0.275 3 5 sembly (as before for the TM), along with a new auxiliary 3 0.63 - 0.69 30 0.275 26 electronics module mounted remotely. New analog-to-digital 4 0.76 - 0.90 30 0.275 3 2 converters were designed to provide 9-bit radiometric preci- 5 1.55 - 1.75 30 0.275 1 3 sion, and two gain states were incorporated so that only 8 bits 7 2.08 - 2.35 30 0.275 5 were transmitted. In this fashion, the sensitivity and dynamic 6 10.4 - 12.5 120 0.275 0.5K NEAT range of the instrument were extended: ground controllers

could select the gain characteristics to optimize radiometric performance for each scene. Numerous other rehements on a

of the 15-m panchromatic band: this band would produce subsystem-by-subsystem basis, such as enhanced redundancy data at four times the rate of a 30-m multispectral band - in the power supplies, made the ETM a more robust and capa- approximately two thirds of the total data rate of the TM. Al- ble instrument than its predecessor. Table 9 summarizes the though this could have been handled by a second data physical characteristics of the ETM, while Figures 7 and 8 stream of approximately 56 MBPS, it was more straightfor- show the flight instrument during assembly-and-test and, sub- ward and economical to produce two identical multiplexers sequently, as mounted on the Landsat 6 spacecraft. operating at 85 MBPS and utilize the additional capacity to Test measurements on the completed ETM showed that provide redundant transmission of two of the 30-m multi- the instrument's performance was substantially better than its spectral bands. Further refinements were also specified, such specifications, as noted in Table 10. The modulation transfer as %bit radiometric resolution by using two gain states. function for the panchromatic band was particularly notewor-

thy: the measured MTF at the nyquist frequency was 0.38, ver- ETM Design Description and Performance sus a requirement of 0.124. This outstanding performance Although many features and subsystems of the Thematic portended superb images from space. Note also that the MTF

Mapper carried forward to the ETM, there were a figures as listed are the worst-case data in the Scan direction; number of modifications and refinements associated with the performance is even better in the along-track d~ection- enhancements for Landsat 6.

The telescope optics and scan-mirror assembly were sub- ETM Operational History stantially similar to the TM, but the optical field-of-view was The operational history of the ETM for Landsat 6 is disap- slightly increased to accommodate the additional spatial ex- pointingly nonexistent because the spacecraft failed to reach tent of the focal plane that now included the detectors for its orbit. Although the launch on 5 October 1993 aboard a the panchromatic band. Additionally, the resolution require- Titan 11 booster proceeded smoothly, a probable failurelmal- ments associated with the 15-m panchromatic band, corre- function in the spacecraft's propulsion system led to orbit-in- sponding to a 21-yrad angular subtense, placed more strin- jection failure and loss of the spacecraft. The spacecraft was gent tolerances on telescope manufacture and alignment, as never located, but it is presumably somewhere on the ocean well as on the linearity and dynamic flatness of the scan mir- floor in the South Pacific. ror. The larger focal plane and increased field-of-view also affected the design of optical baffles.

The prime focal plane for the ETM was substantially up- Landsat 7 Payload graded by the use of a single, monolithic silicon detector ar- Design Objectives ray for all of the visible and near-infrared spectral bands The Landsat Program changed substantially between the (including the new 15-m panchromatic band). This approach commercialization initiatives of the early to mid 1980s and provided better band-to-band geometric registration and sta- the definition and development of Landsat 7 in the 90s. bility: because all of these detectors are on a common silicon The challenges associated with commercialization of the pro- substrate, their geometry is established with photolitho- gram led to a reevaluation of government priorities and poli- graphic accuracy. In contrast, the TM used four separate de- cies. Landsat's value in providing benchmark data sets for tector arrays for bands 1 through 4, and this required global-change research was becoming evident, so continuity precision OP~O-mechanical alignment. Although on-orbit per- of such data became increasingly important to the scientific formance has validated the original TM design approach, the ~ommunity. Further, there was increasing interest in Landsat monolithic ETM focal plane is inherently more producible by the Department of Defense: Landsat had proven valuable and stable. The cooled focal plane with indium antimonide for updated during the Gulf War, and the prom- and mercury cadmium telluride detectors for the infrared ise of a more capable Landsat system was of considerable in-

terest, so a "dual-use" (civil and defense) strategy was

TABLE 9. ETM PHYSICAL CHARACTERISTICS adopted for Landsat. All of these factors crystallized in the passage of the Land Remote Sensing Policy Act of 1992 (Pub-

Mass Scanner Assembly: 288 kg lic Law 102-555) and the formation of a joint Air-FO~C~/NASA Auxiliary electronic Module: 8 1 kg program office that was charged with the procurement, de-

Dimensions Scanner Assembly: 1.3 by 0.7 by 2.0 m velopment, and operation of Landsat 7. Auxiliary electronic Module: 0.5 by 0.7 by 1.0 rn This led to a request for proposal that included specific

Power (maximum) 490 W requirements for data continuity, along with a list of desired

Date Rate 2 by 84.9 MBPS enhancements to Landsat's capability - enhancements such as improved spatial resolution (5 m), stereo imaging capabil-

PE&RS July 1997 847

Figure 7. ETM flight hardware during assembly and test.

I Figure 8. ETM on the Landsat 6 spacecraft. I ity, and cross-track pointing for more frequent revisit oppor- tunities. Additionally, there was a requirement for upgraded absolute radiometric accuracy. This latter specification came from the recognition that Landsat 7 would effectively become viewing. This approach effectively partitioned the develop- a part of the Mission to Planet Earth sensor suite, with in- ment risk for the program by using proven technology for the creasing reliance on the fidelity and accuracy of its data for data-continuity mission, and a new sensor for the added ca- global-change research purposes. The Landsat 7 design pabilities. Again, this was a reinforcement of NASA's success- objectives established by the joint Air F o r c e l ~ ~ S ~ program office are summarized in Table 11.

Landsat 7 Payload Design Description In order to satisfy the needs of the broad user community for Landsat 7, the General ElectricIHughes Aircraft team pro- posed a system with two payload instruments: a further up- grade to the Enhanced Thematic Mapper (i.e., the ETM+) for data continuity with earlier Landsats, and a second sensor, the High Resolution Multispectral Stereo Imager (HRMSI) to provide high spatial resolution (5-m pan, 10-m multispectral) and agile pointing capability for both stereo and cross-track

TABLE 10. ETM MEASURED PERFORMANCE WAS SIGNIFICANTLY BETTER THAN THE SPECIFICATIONS

Specified Measured Specified Measured Band SNR SNR MTF MTF

20 45 55 40 4 7 2 1 19

0.13K NEAT

TABLE 11. LANDSAT 7 DESIGN OBJECTIVES

1. Data continuity - with data at least equal in quality and kind to that offered by the Thematic Mappers and Enhanced Thematic Mapper of Landsats 4, 5, and 6:

One Panchromatic Band with 215-metre resolution (Landsat 6) Seven multispectral bands spanning the visible through long- wave infrared spectrum

Six Visible through shortwave infrared bands at 230-metre resolution One Longwave infrared band at 5120-metre resolution

Radiometric performance (signal-to-noise ratio and calibration) at least equal to that provided bv Landsats 5 and 6 185-km ( ~ O O nautical'mile) swath width 16-day revisit cycle 5-year on-orbit reliability

2. Enhancements: The following, in priority order, represent addi- tional capabilities desired for Landsat 7, after meeting the data-con- tinuity requirements:

Improved spatial resolution Improved absolute calibration Stereo mapping capability Additional spectral bands Improved revisit time (cross-track pointing) Improved radiometric sensitivity Improved line-of-sight accuracy

July 1997 PE&RS

SPACECRAFT . Landant 6 upgndo

E l M c DATA CONTINUITY 16 m Panehrwnatlc

.30 m VWlft18W1~

.SO m LWIR

Cstlbntfon . t x f S M p M

HRUSI: HIWRIDRfFY ENHANCEMENTS . 6 m ~ r o w t u t I m .10 m VNIR: 4 Muitlspectni WMIs .BOkJnswmh . AlOtl#&& IltM.0 imPl)h . Crasstrack pdntIna id+'Jw mwtt

.2x')bMbps

.WaksteppdarterdsaienphBM - ixtwmr budqrt oonstrslnEb

Figure 9. Landsat 7 payload as originally planned. I

ful pattern of carrying old and new instruments alongside one another in order to effect a smooth transition and mini- mize risk to data continuity. Figure 9 depicts the originally planned payload suite for Landsat 7.

High-Resolution Multispectral Stereo Imager (HRMSI) Utilizing pushbroom sensor technology, the High-Resolution Multispectral Stereo Imager (HRMSI) was designed to provide several high-priority performance enhancements for Landsat 7:

Improved spatial resolution: 5-metre ground-sampling dis- tance (GSD) panchromatic band and four 10-metre GSD Visible and Near-Infrared (VNIR) bands. Improved radiometric sensitivity: The pushbroom sensor con- cept provides increased dwell time and improved signal-to- noise ratios compared to equivalent channels of the ETM, even at higher spatial resolution. Stereo imaging: Stereo imaging at a 5-metre spatial resolution with a variable in-track stereo angle from nadir to f 30". This same-pass stereo capability represents a significant improve-

TABLE 12. HRMSI DESIGN CHARACTERISTICS

Scanning Method Swath Width Pointing Capability

Telescope

Size

Mass

Power Data Rate (Compressed)

Pushbroom, Multispectral Linear Array 60-km (5" field-of-view from 705-km orbit) f 38" cross track for rapid revisit + 30" along-track for stereo imaging 18-cm aperture, unobscured reflecting triplet

Sensor Assembly: 82.5 by 81.3 by 64.8 cm Electronics Module: 0.1 m3 Sensor Assembly: 45.4 kg Electronics Module: 54.5 kg 125 w, 148W when slewing telescope Panchromatic Band: 75 Mbps at 4 bits per pixel VNIR Bands: 75 Mbps at 5 or 6 bits per pixel

Spectral Spatial Bandwidth Resolution

Band (pm) Detectors (m) SNR

Pan 0.50 - 0.90 6100 silicon photodiodes 5 20 1 0.45 - 0.52 3050 silicon photodiodes 10 3 2 2 0.52 - 0.60 3050 silicon photodiodes 10 3 5 3 0.63 - 0.69 3050 silicon photodiodes 10 2 6 4 0.76 - 0.90 3050 silicon photodiodes 10 3 2

PE&RS July 1997

ment over other systems that rely on multiple-orbit side-look- ing images to acquire a stereo image pair. Cross-track pointing: Cross-track pointing at f 38" provides a 3-day or less revisit frequency at the equator and more fre- quent revisit at higher latitudes. 60-km swath: A 112- by 112-km stereo image area can be ac- quired well within 90 days.

The HRMSI development effort exploited over a decade of de- sign work on multispectral linear array instruments. The tel- escope utilized an all-reflective three-mirror anastigmatic form that provided excellent optical performance over the 5" field-of-view, and the entire telescope assembly was mounted i n a two-axis gimbal to provide the requisite pointing agility. Linear photodiode arrays were specified for all of the detec- tor bands, and on-board data compression was utilized in or-

Rad-

- (Stowed)

FAC = Full Aperture Callbrator PAC = Partbal Aperture Calibrator SLC = Scan Line Corrector OBC = On Board Calibrator PFPA = Prime Focal Plane Assembly CFPA = Cold Focal Plane Assembly PS = Power Supply - ETM+ NewIModlfied Designs

Two 75 Mbps

Streams to

Pavload Correction

r=I Electronics Module Analog

Postamps Command 8 Telemetry

Figure 11. ETM+ block diagram highlights new or modified subsystems in bold out- line.

der to maintain compatibility with the 2- by 75-MBPS months short of the critical design review that would have capacity of the Tracking and Data Relay Satellite. These and marked the completion of the detailed design of the instru- other design features are summarized in Table 12, followed bv an artist's illustration of the instrument in Figure 10. "

Design of the HRMSI instrument began in ~ei tember 1992, progressing through a preliminary design review and well into the detailed design phase. Unfortunately, budgetary pressures and differing priorities placed the Landsat 7 pro- gram in a considerable state of flux in early 1994. This ulti- mately led to the Air Force's withdrawal from the program, and Landsat 7 emer~ed as a NASA-sponsored project under the aegis of ~ i s s i o n t o Planet ~arth: In this context, The- matic vMapper data continuity for global-change research be- came the raison d'etre for Landsat 7. and budeetarv constraints precluded further development of The ~ S I . As a result, work on HRMSI ceased in May of 1994, just a few

Scanning Method Bidirectional cross-track, Scan Frequency: 7 Hz

Swath Width 185 krn (15" field-of-view from 705 km orbit)

Telescope 40.6-cm aperture, Ritchey Chretikn

Size Scanner Assembly: 1.5 by 0.7 by 2.5 m Auxiliary Electronics Module: 0.4 by 0.7 by 0.9 m

Mass Scanner Assembly: 298 kg Auxiliar~ Electronics Module: 103 kg Cable ~ G n e s s : 20 kg

-

Power 510 W

Quantization 9 bit AID conversion, 8 bitslpixel transmitted (2 gain states)

Data Rate 2 by 75 MBPS, CCSDS format

Spectral Bandwidth

SNR Spatial (at min

Resolution signal Band (pm) Detectors (m) radrance)

Pan 0.50 - 0.90 1 0.45 - 0.52 2 0.52 - 0.60 3 0.63 - 0.69 4 0.76 - 0.90 5 1.55 - 1.75 7 2.08 - 2.35 6 10.4 - 12.5

32 Si photodiodes 15 16 Si photodiodes 30 16 Si photodiodes 30 16 Si photodiodes 30 16 Si photodiodes 30 16 InSb photodiodes 30 16 InSb photodiodes 30 8 HgCdTe photoconductors 60

manufacturing processes for enhanced producibility and reliability.

July 1997 PE&RS

Figure 13. Partial-aperture solar calibrator uses faceted prism assembly to provide calibration-ref- erence irradiance when spacecraft crosses the terminator.

ment. Landsat 7 was subsequently restructured as a single- payload program, carrying only the ETM+ for the data-continuity mission.

Enhanced Thematic Mapper Upgrade (ETM+) The ETM+ represents a further refinement of the ETM devel- oped for Landsat 6, with the addition of end-to-end on-orbit calibration and a 60-m, rather than a 120-m, long-wavelength infrared (LWIR) band. The 15-m panchromatic band and the six 30-m multispectral bands are carried forward from the ETM. Additional upgrades for ETM+ are focused on enhanced reliability. While this may seem unnecessary in view of the extraordinary on-orbit longevity of the Thematic Mappers on Landsats 4 and 5, the statistical probability of long-term op- eration can be bolstered by increasing redundancy in subsys- tems or by incorporating additional flexibility in on-orbit switching and cross-strapping of existing subsystems, and these upgrades become important in the context of a five- year design-life requirement. The ETM+ block diagram illus- trated in Figure 11 highlights the new or modified subsystems that differentiate the ETM+ from its predecessor, and Table 13 summarizes the key design characteristics of the instrument.

While the ETM+ can trace its lineage to the Thematic Mappers of Landsat 4 and 5, and the ETM of Landsat 6, many of the subsystems and components such as detectors, spec- tral-bandpass filters, high-speed electronics, and on-board calibration subsystems have been upgraded. The cooled focal plane assembly, incorporating the detectors for bands 5, 6, and 7, has been updated to reflect modern practice in detec- tor producibility. This includes backside-illuminated indium antimonide detectors for bands 5 and 7 and a quartz sub- strate for the focal plane, as shown in Figure 12. The spec- tral-bandpass filters have been manufactured with ion-deposition techniques to provide better control of band- edge characteristics and to minimize airlvacuum shifts in spectral response. Elsewhere in the system, the high-speed multiplexers have been redesigned to provide two 75 MBPS data streams in CCSDS (Consultative Committee for Space Data Systems) format for compatibility with the Earth Ob- serving System communication and data processing protocol. The reduction from 85 MBPS to 75 MBPS is feasible because

PE&RS July 1997

LaunchLock

(b) Figure 14. Full-aperture solardiffuser calibration refer- ence. (a) Full-aperture solar diffuser is designed to pro- vide an end-toend calibration reference upon ground command. (b) ETM+ hardware photo shows full-aperture calibrator in stowed position. Multilayer protective blan- kets cover back sid of diffuser; stacked rotary actuator is at lower right side of diffuser.

data buffers within the multiplexers allow continuous data transmission during the scan-turnaround interval, whereas previous TM instruments transmitted actual image data only during the active portion of the scan. The new multiplexers are also fully redundant with additional cross-strap switch- ing to enhance reliability.

Recognizing Landsat's importance in providing accurate measurements in support of Mission to Planet Earth, calibra- tion accuracy has been an area of special emphasis on ETM+. Accordingly, three independent on-board calibration systems are used to calibrate the panchromatic, visible and near-in- frared (VNIR), and short-wavelength infrared (SWIR) bands. They consist of

A full-aperture solar diffuser on the inner surface of the aper- ture door that illuminates the focal planes with diffusely re- flected solar energy when commanded into position; A partial-aperture solar reflector that illuminates the focal planes with attenuated solar energy, once per orbit; and Calibration lamps that project calibrated energy onto the focal planes via the main calibration shutter, once per scan, during the scan mirror turnaround.

Note that only the calibration lamps and shutter were uti- lized in the earlier TM and ETM instruments, so the ETM+ represents a significant step forward in absolute radiometric calibration accuracy. Figures 13 and 14 show the additional on-board calibrators for the ETM+.

TABLE 14. THREE DECADES OF LANDSAT I N S T R U M E N T S

MSS TM ETM ETM+

Spectral Bands 1 0.5 - 0.6 pm 1 0.45 - 0.52 pm 2 0.6 - 0.7 pm 2 0.52 - 0.60 pm 3 0.7 - 0.8 pm 3 0.63 - 0.69 pm 4 0.8 - 1.1 pm 4 0.76 - 0.90 pm

5 1.55 - 1.75 pm 7 2.08 - 2.35 pm 6 10.4 - 12.5 wm

Spatial Resolution

Radiometric Resolution

Data Rate

Mass

Average Imaging Power

Envelope

6 bits

15 Mbps

64 kg

30 m VNIR/SWIR 120 m TIR

8 bits

85 Mbps

258 kg

P 0.52 - 0.90 pm 1 0.45 - 0.52 l m 2 0.52 - 0.60 pm 3 0.63 - 0.69 pm 4 0.76 - 0.90 pm 5 1.55 - 1.75 pm 7 2.08 - 2.35 pm 6 10.4 - 12.5 pm

15 m Pan 30 m VNIR/SWIR

120 m TIR

9 bits (8 bits transmitted, 2 gain states)

2 X 85 Mbps

288 kg Scanner 81 kg AEM

490 W

P 0.52 - 0.90 pm 1 0.45 - 0.52 pm 2 0.53 - 0.61 pm 3 0.63 - 0.69 pm 4 0.78 - 0.90 pm 5 1.55 - 1.75 pm 7 2.09 - 2.35 l m 6 10.4 - 12.5 pm

15 m Pan 30 m VNIR/SWIR

60 m TIR

9 bits (8 bits transmitted, 2 gain states)

2 X 75 Mbps

318 kg Scanner 103 kg AEM

510 W

1.3 X 0.7 X 2.0 m Scanner 1.5 X 0.7 X 2.5 m Scanner 0.5 X 0.7 X 1.0 m AEM 0.4 X 0.7 X 0.9 m AEM

Aperture 23 cm 40.6 cm 40.6 cm 40.6 cm

At this writing, the ETM+ is in the final assembly and test process. Interim test results indicate that this instrument will meetlexceed its specifications, as has been the case for its predecessors, but definitive acceptance-test data are not yet available for the ETM+.

Operational History The ETM+ for Landsat 7 has nearly completed its manufac- ture and test at Hughes Santa Barbara Remote Sensing, and will subsequently be delivered to Lockheed-Martin for inte- gration with the spacecraft. Launch of Landsat 7 is scheduled for 1998.

Summary Three decades of Landsat sensor development have produced progressively more capable instruments for this important Earth-observation mission, as delineated in Table 14. The MSS first launched in 1972 led in turn to the Thematic Map- per, and three successive generations of Thematic Mapper in- struments have been produced for Landsats 4. The latest of these, the upgraded Enhanced Thematic Mapper (ETM+) is nearing completion. All of the sensors built to date have per- formed better than their specifications, and those that have successfully reached orbit have exhibited extraordinary on- orbit reliability.

Acknowledgments The author gratefully acknowledges the contributions of nu- merous colleagues at Hughes and NASA. Literally hundreds of talented engineers and scientists contributed to the design and development of the instruments described in this paper, and it is a privilege to serve as the chronicler of this collec- tive work. Key contributors to the development of the Mss included Virginia Norwood and Jack Lansing, with addi- tional engineering and project management by Ralph Wen- gler, Leroy Barncastle, and the late Tony Lauletta. The Thematic Mapper was developed under the technical leader-

ship of Oscar Weinstein and Jack Engel, with the late Warren Nichols providing program management and executive lead- ership of the team leading to the delivery of the Protoflight Model for Landsat 4. Dr. Vince Salomonson led the outstand- ing Landsat Thematic Mapper science team at NASA. The Landsat 5 TM was completed and delivered by a team led by Dr. Fletcher Phillips. For the ETM, technical development was again led by Jack Engel with contributions by Frank Malinowski and program management by Richard Ruiz. Lee Tessmer, Richard Roberts, and Roberto Diffoot have all been key figures in the development of the ETM+ for Landsat 7, with Rick Obenschain, Ken Dolan, and Dr. Darrel Williams providing project leadership at NASA. Finally, Virginia Trau- twein, Sharon Fullmer, Barbara Hoffman, Margaret Finlay, Kathy Cremeen, Greg Krueger, and Nancy Richards all con- tributed to exhuming the historical records that served as source material for this paper. The author extends apologies in advance for any errors of omission in these acknowledg- ments.

References Colwell, R.N., D.S. Simonett, and F.T. Ulaby (editors), 1983. Manual

of Remote Sensing, Second Edition, Volume 1, American Society of Photogrammetry, Chapter 12.

Jones, C.R., and J.L. Engel, 1981. Multispectral Scanner, Thematic Mapper and Beyond, Proceedings of Remote Sensing Sympo- sium.

Lansing, J.C., Jr., and R.W. Cline, 1975. The Four and Five Band Multispectral Scanners for Landsat, Optical Engineering, 14(4): 312-322.

Santa Barbara Research Center, 1981. MSS-D 4-Band 52000 F-1 Model, Serial No. 3 Consent to Ship Data, SBRC Reference No. HS248-6692.

, 1984. Thematic Mapper: Design Through Flight Evaluation, Final Report, SBRC Reference No.RPT41741.

UC Santa Barbara Extension and Santa Barbara Research Center, 1983. Landsat Short Course.

July 1997 PE&RS