A paper presented at the Pecora 13 Symposium,
Sioux Falls, South Dakota,
August 22, 1996
William E. Stoney
Mitretek Systems Corporation
This paper provides an overview and analysis of the 46 high resolution earth observation satellites (equal to or better than the current Land sat's 30 meters) planned to be launched in the decade starting in 1996. All imaging instrument types are represented, panchromatic film cameras, panchromatic, multispectral and hyperspectral scanners, and radar imagers. A detailed summary and analysis of the data characteristics of each system, including their spatial resolutions and spectral and temporal coverage is presented and shows both the breadth and depth of the data products that could be available.
It would be naive to believe that all of the plans presented will come to pass. The tightening world economies will likely affect both the government and commercial plans for new satellites. But it would be equally naive not to recognize that the plans represent a real and growing worldwide interest in the potential of satellite remote sensing by both governments and the private sector. The next decade will certainly see at least three wide area Landsat like systems in orbit simultaneously and equally surely at least one and probably two of the commercial high resolution systems. And the continuation of at least two radar systems also seems assured.
The message to the users and especially to the research and value added communities is that now is the time to aggressively develop the new and more useful information products that the multiple satellite environment will make possible for the first time. The future for satellite remote sensing really lies in the hands of those responsible for developing scientifically and/or economically useful information products from the diverse and robust data sets that can be available if the economics are shown to be viable. The number and kind of satellites that are actually launched and will continue to be operated is in the "invisible hands" of the users.
In the late nineteen sixties, William Pecora, while serving in a number of positions at the Department of the Interior, including Undersecretary, convinced a reluctant bureaucracy to initiate the Landsat series of land observing satellites. This was at a time when detailed views of the land from space were restricted to the military alone, and when the earth science community had little or no understanding as to the scientific and practical value satellites could bring to their disciplines. Fortunately his vision prevailed and the Landsat program launched its first satellite in 1972. As shown in Figure 1, that first satellite was followed by four more U.S. satellites and by several more from France, India, Japan, Russia and ESA. The growing community of earth scientists now have 10 national and international satellites currently in orbit that meet Dr. Pecora's land imaging goals.(1) The principal purpose of this paper is to show potential users that there are satellite program plans that could expand that number to 20 or more in the next decade, most of which will have measurement capabilities that are well beyond Dr. Pecora's fondest dreams.
This paper covers only the civil, domestic and foreign, land observation satellites current or planned for the next decade with thirty meter (the current Landsat resolution) or better ground resolution. While this consists of a large number of satellites, it is only a small part of the total number of earth observing satellites which are or will be measuring the earth in various other ways, from the daily images (with 1 Km ground resolution) and atmospheric soundings of the NOAA and foreign weather satellites to the numerous scientific measurements of NASA's Mission to Planet Earth and its associated foreign collaborators. This subset has been chosen because it is most likely that satellites which have the ability to define human scale objects will prove most useful to those who attempting to record and understand the details and scope of the interactions between man and his habitat.
2.0 The Satellites
Nearly all the following data were gathered by a questionnaire sent to the participants at the American Society of Photogrametry and Remote Sensing (ASPRS) September 1995 conference "Land Satellite Information in the Next Decade" or from presentations made at that conference. Corrections have been made where required and several programs initiated in the interim have been added.(2)
Figure 2 lists 46 satellites from nine countries that either have been placed in operation since 1990 or will be launched by 2004. It lists the satellites by the four major sensor types carried. It also provides the sponsoring country, the source of funding (government or private), the name of the program, the sensor names including the lower resolution wide angle sensors where they are used, the launch dates and the sensor types carried. Figure 3 provides the launch schedule and probable operating lifetimes of all the systems noted on Figure 2.
As can be seen on Figures 2 and 3, the satellites group naturally, first into the two sensor types, optical and radar. Second, the optical systems fall into two groups based on the characteristics of the data they are designed to provide: the first group following closely in the footsteps of the Landsat system, incorporates wide swaths and intermediate resolution, with broad spectral coverage including SWIR, and the second group, made possible by the governments decision to allow high resolution commercial systems, have high resolutions, narrow swaths, and PAN and VNIR sensors only. The charts also note the five experimental systems being developed by the US, which are discussed below.
3. The Year 2000
Perhaps the best understanding of the basic data characteristics of the next decade's planned satellites can be seen on Figure 4 which lists those satellites with launches planned for '96 to the year 2000 and thus those which should be in orbit in that year. This chart provides the resolution of each of the color bands measured by the sensor, the type of stereo coverage possible, the swath width and the nadir global repeat which results from that swath width.(3) The color bands are listed under the Thematic Mapper's band numbers. While the individual band ranges do vary somewhat as shown on Figure 5, the differences are negligible for the majority of data uses.
Note that the two operational optical groups are each fairly homogenous in terms of the number of bands, the range of the resolutions and the swath widths in each group and that these attributes are distinctly different between the two. It is apparent that the systems are focused on two different sets of applications and that neither can effectively serve both.
3.1 The Landsat Group
The upper group on the chart, all but one of which is government funded, is configured for multispectral monitoring of the changes to the land cover of large areas, i.e. the globe, continents, countries, states or Texas scale ranches. The data must be acquired over these large areas often. For example the Global Land Change Science Program would like global coverage of all the land areas at least once every season. Equally important the sensor must include at least the lower SWIR band which is considered vital for many multispectral analysis based functions. As the chart shows, all but ADEOS (4) include at least the lower SWIR band.
The one commercial venture included in this group, Resource21(5) is unique in several ways. Their target market is support of the farmer to enable "precision farming" technologies. This requires delivery of products at least once per week which in turn requires 4 satellites to provide a reasonable chance of ensuring the weekly data delivery in spite of missing coverage due to cloud cover. Also the data must be very well calibrated to be useful in identifying meaningful changes in the field's growing patterns. As a result their sensor includes a solar calibration design similar in capability to that being developed for Landsat 7.(6)
3.2 The High Resolution Group
The second group is essentially commercial(7) and is focused on mapping and GIS products at scales currently provided by the airborne sensor industry. They project that their primary products will be panchromatic 1 to 3 meter scenes and stereo pairs. Their high resolution is essential to the development of new markets for imagery capable of identifying and measuring changes at the human scale of houses, roads and civil engineering projects. It is well understood that this scale of imagery could be very useful to those countries that may feel the necessity of being kept aware of their neighbor's potentially threatening activities. While they cannot provide the global wall to wall coverage currently targeted by the Global Land Change Program, their ability to point and return quickly to the same scene may be ideal for statistical sampling methods using both their PAN and VNIR bands for specific land change science and land change management functions.
3.2.1 The Experimental Systems: The US is funding four test oriented missions which will be operating in the year 2000. One will carry a Japanese developed experimental sensor, ASTER. The data from the experimental sensors will be used to evaluate the potential of increasing the accuracy of spectral identification of earth cover classes by increasing the number of spectral bands measured, either by adding additional discrete bands in the SWIR and TIR regions as in ASTER, or by using a hyperspectral sensor as in Lewis. EO-1 will test several advanced sensor and satellite technologies. In particular current plans call for two hyperspectral sensor approaches, one using a wedge filter and the second a grating spectrometer. The exact sensor characteristics of Warfighter 1 have been left up to the proposers whose task is to meet a detailed set of target identification tasks. This system bears watching by the civil users because the contractors are being encouraged to develop a system which will have civil uses as well.
3.2.2 The Radar Group: Radars are at once the most promising and the least understood of the satellite land observation sensors. The ability to see through clouds makes radar imaging very attractive to users living in cloudy areas, which probably explains ESA's early and continuing interest in this technology. The radar satellites can supply images with resolutions from 5 to 100 meters for swaths from 20 to 500 kilometers and thus can cover a wide range of applications. At the same time radar images are very complex compared to optical images and data interpretation techniques are not as well understood or available as those for optical data.
There are still technical questions concerning the choice of band which should be used, ESA choosing the C band and Russia the S. Japan, whose past and proposed radar satellites did not make the date window cut of Figurefour, selected the L band for their systems. The US is notably absent from this group. However NASA teamed with Germany and Italy to develop and fly SIR C on the Shuttle. The missions included three different bandwidth radars, C, L, and X. The preliminary results indicate that the ability to register images from several radar bands may prove as valuable for interpreting radar images as multispectral interpretation has proven for optical systems.
4. Data collection variables
It is readily apparent from the data characteristics presented for the satellites on Figure 4, that there are wide variations in the image features supplied by the different systems. The major features of interest to the user are spatial resolution, temporal resolution and spectral coverage.
4.1 Spatial resolution
In late 1994 congress and the administration created a revolution in civil remote sensing by providing rules for commercial satellites which did not put limits on the resolution that could be used. The private sector immediately requested and got licenses for first a three meter satellite (to be launched late in '96) and then after considerable discussion between the administration and the congress, one meter resolution data licenses were granted to four companies (to be launched in '97, '98 and '99). The fact that even then 2 meter panchromatic scenes could be obtained through commercial channels (and still can) from a Russian satellite was a motivating factor in the decision process.
Doubling the resolution requires four times the data for a given area. Not only do the costs of data management go up dramatically, the sensor and satellite costs force system designs to be limited to narrow swaths. In practice the systems intended for the identification of land cover and land use and thus broad area coverage have focused on moderate resolutions between 5 and 30 meters and swaths of 100 to 200 kilometers, while the high resolution satellites are designed with 1 to 3 meters resolution and 4 to 40 kilometer swaths.
4.2 Spectral coverage
As noted the three optical groups differ in spectral coverage. Figure 5 provides a graphic picture of the spectral bands for the year 2000 satellites. It shows the band information from Figure 4 in a form in which the similarities and differences are readily apparent.. (It does not show the higher resolution sharpening PAN bands carried by all but the Resource21 and the ASTER systems.)
The major use of the color bands is automated classification of land cover types. The lower SWIR band has been shown to be nearly essential in such analysis and thus is featured in the global coverage type systems. The upper SWIR has proven of great value in the discrimination of surface mineralogy. The TIR band measures the radiation from heat sources and has not been generally used in land classification projects, partially because of its lower, 120 meter resolution.
The high resolution systems carry VNIR bands only because of the extra costs of adding bands above that region. Since their major use will be in mapping and the identification of man made objects, they are not as concerned with having the capability of computer classification of the natural environment.
The color bands of the two experimental satellites are pointing to the increased interest in doing more and better computer identification and classification by increasing the amount of spectral information measured, either by adding many more individual bands as in ASTER or by measuring the spectrum continuously as in the hyperspectral sensor on Lewis.
4.3 Temporal coverage
There are two types of temporal coverage; global repeat coverage, which is the frequency with which all locations can be visited, and is a direct function of the swath imaged by the sensor; and site revisit period, which is the time required to revisit any given site and is a function of the side pointing capability of the satellite and or instrument. All of the satellites except Landsat have cross track pointing and thus revisit times of 2 to 3 days at the equator. Because all the satellite orbital paths cross near the poles, both global repeat and site revisit times are halved at 60 degrees latitude and get rapidly better above that latitude.
Figure 6(8) provides a graphical look at the relation between the broad area and the high resolution satellite swaths and the resulting global coverage times. It illustrates the three system configurations that have been developed to cover the trade offs between spatial coverage, temporal coverage and resolution. ( MODIS & AVHRR kilometer resolution, 1 to 2 day global repeat systems; Landsat like 10 to 30 meter, 2 to 3 week global repeat systems; and high resolution 1 to 3 meter, 4 to 12 month global repeat systems)
4.4 Multiple satellites
The above discussion of scene attributes has been focused on the capabilities of a single satellite. Because many potential uses require data to be taken over very large areas frequently and at particular times it is apparent that no single 30 meter or better satellite can meet such needs. Figure 7 illustrates that the situation in the year 2000 could solve the resolution/frequency trade off by the availability of multiple satellites. The top section shows the number of times any point on the equator will be overflown in a hundred day period by one of the six government satellites noted. These satellites are from five different countries. If their orbits and their repeat periods had been coordinated the Figurewould have shown a constant repeat interval of 3 days. When the other satellites planned for the year 2000 are added, nearly every other day coverage results. It is obvious that this is due principally to the four coordinated satellites of Resource21 with their own 4 day repeat cycle.
4.5 The effect of clouds
The user must be aware that the above overpass figures do not directly represent data acquisition opportunities because of the frequency of clouds. All of the above figures must be corrected for the effect of clouds. Figure 8 presents the results of 'flying' the satellites over a global map of Landsat scenes of the world each of which have been filled with the percentage of cloud free area for each day of the year.(9) A scene acquisition plan which accepts all scenes which are better than 85% cloud free and any series of scenes that have a high probability of being 85% cloud free when mosaiced together, was carried out for a single satellite, for two satellites and for four satellites.
The Figure 8 graphs show the cumulative percent of the acceptably cloud free global land scenes achieved by each of the three configurations for each of six 16 day global cycles. The increased cloud free scene acquisition capability of multiple satellites is obvious as is the inability of a single satellite to meet the science goal of quarterly global coverage.
But cumulative global coverage figures are not too helpful if your location of interest doesn't represent the average weather conditions. Figures 9, 10, 11 and 12 map the cloud free percentage of each scene acquired in the first global pass of the one, two, three and four satellite configurations. These are mapped onto a grid of the 11,680 Landsat 185x170 kilometer scenes that make up a complete global coverage. It is apparent that again multiple satellites are critical to obtaining cloud free data in many of the most ecologically interesting areas of the world, i.e. the major growing areas.
It is up to the users now.
If only half of the satellites listed in this paper are actually launched, Pecora's vision of a robust and capable global land observation space system will have been fulfilled. The 'rocket scientists' have done their job. The political (and soon the entrepreneurial) parts of society have allocated multi billions of dollars to place satellite systems in orbit that can assess the state of the world's vegetation almost daily or will be able to provide detailed maps with the latest roads and buildings at almost any frequency the user might want. It really will be technically possible for anyone to observe, record and measure the outdoor building activities (from backyard pools to nuclear plants) of anyone else, any place in the world, nearly anytime.
The value of these systems in now in the hands of the user community. For what and how well they will be used literally depends on the imagination and enterprise of the scientific disciplines concerned with understanding our world and the impact our numbers and activities are making on its possibly fragile systems.
The science users still face critical problems. They must decide which data will be useful for what science objective. This will necessarily involve interdisciplinary and international teams in an iterative process that will extend years if not decades. The need for funding must be addressed now if there are to be any significant programs in place when the new millennium satellite systems are in place.
There is at least one more problem that must be addressed by the science user community as soon as possible. Hopefully it is one that may not require much funding. Landsat, Spot, IRS-1C and ADEOS are in orbit now. A forceful science community could surely devise a scientifically viable land data acquisition plan using all of these satellites that would ensure that a global archive would be available for the future. Data acquisition and storage of raw data should involve only marginal cost increases and thus be acceptable to the government owners and operators as a small price to pay for an invaluable contribution to future generations of global science investigators.
The bulk of this work was performed under NASA contract NAS5-3206 and NOAA contract 50-SPNA-4-00023
(1) See also other Pecora 13 papers (e.g. Obenschain, et al., Williams, et al., Komar, et al.) which discuss the Landsat program, its history and its potential future.
(2) The author is grateful to ASPRS for permission to use much of the material originally published as part of the ASPRS summary conference report in this paper.
(3) Note that with the exception of Resource21, the repeat days are for a single satellite. Do not confuse global cover repeat times with the revisit times provided by those satellites which can point their sensors off axis. As the chart notes all of the systems with stereo capability can provide 2 to 3 day repeat visits at the equator and better than that at the higher latitudes. Global cover repeat times are a measure of the ability of the satellite system to gather a complete global data set if there were no clouds. Simulations have shown that the repeat days on the chart must be multiplied by at least 4 to 6 times to define the acquisition times for a reasonably complete global cloud-free data set.
(4) ADEOS is included in the upper group, but its lack of a SWIR band, its fairly low swath width and middle range resolution places it in between the two groups
(5) Resource21 is a proposed partnership of two aerospace companies, Boeing Commercial Space Company and GDE Systems and three agro-business companies, Agrium, LTD., Farmland Industries, and Pioneer Hi- Bred International. They have stated that they will make their final go no-go decision by 8/96.
(6) Sensor calibration, which is an important issue for many multispectral applications, was not provided in great detail in the questionnaire responses nor discussed in the conference presentations. The critical issue for multi-year land change applications is the long term stability of the sensors and/or the inter sensor comparability of successive sensors in a measurement program. Most current sensors depend on on-board calibration lamps. Landsat 7, Resource21, and ADEOS are the only systems that noted their plans to use the sun as a source to improve the precision of their long term calibrations.
(7) Russia's Spin and Almaz programs are posturing themselves as totally commercial, albeit their original systems were government funded. India is selling their current data through EOSAT.
(8) This figure is presented with the kind permission of its creator, Dr. Darrel Williams of GSFC.
(9) The cloud data are from an Air Force analysis of the cloud conditions for the year 1997 and were provided to Mitretek by the then GE Aerospace Corp.