SECTION 20

OUTLOOK FOR THE FUTURE

Land Sensing Satellites in the Year 2000

A paper presented at the
International Geoscience and Remote Sensing Symposium (IGARSS),
Singapore, August 7, 1997

By:

William E. Stoney
Mitretek Systems
7525 Colshire Drive, McLean VA. 22102-7400
703 610 1768, (1767 fax), wstoney@mitretek.org

In the year 2000, if all the current plans for earth viewing satellites come to pass, there will be 31 satellites capable of providing data of 30 meters or better resolution in orbit at the same time. 14 of these will be privately funded by US corporations, all with resolutions of 10 meters or better. 8 optical systems and 2 radar satellites will be funded by foreign countries. The US government will launch 2 multispectral systems and 3 hyperspectral satellites by then. While it is probable that the reality will not be as bountiful as the plans indicate, it is certain that the number of high resolution satellites in operation will be significantly greater than the current supply.

This paper presents a summary of the high resolution systems currently planned to be operating in the year 2000. The systems can be usefully classified into four groups, (a) broad area coverage, 5 to 30 meters resolution and multiple color bands, (b) narrow swaths, 1 meter or less panchromatic resolution, and VNIR color only, (c) Hyperspectral sensors with 30 meter resolution and (d) Radar with 5 to 10 meter resolution. Their capabilities are described and compared in detail, including their spectral bands and resolutions, and their coverage capacity.

Purpose

This paper provides an overview of the explosion in land observing satellites planned for the next decade. It is intended as a wake-up call and planning tool for all who are interested in knowing and keeping track of the details of what is going on with the surface of our planet and in particular for those who are developing the skills to measure and understand the breadth and detail of the information that analysis of the satellite data could make available to us for the first time. The amount and quality of the land information data which the land observing satellite fleet in 2000 will be capable of providing could revolutionize both our scientific knowledge and our practical management of our earth's resources. The satellites are however only the first step. Their value can only be realized through the ingenuity and efforts of the users.

The Satellites

Among the many predictions for the new millennium are the orbiting of 31 satellites in polar orbit providing land cover data at resolutions of one to thirty meters. These satellites are summarized in figure 1, the program names, system funders and launch dates are presented in figure 2 and their launch and operational schedules are presented in figure 3. They fall into four functionally separate classes.

Landsat-like: The thirteen Landsat-like satellites have the middle resolution, broad area and multispectral coverage characteristic of the current satellites, Landsat, SPOT and IRS. These current programs are being extended and expanded. As can be seen on figure 3, the Indian program, with plans for the flight of four satellites through this period is the most operationally robust of the government group. The group will be joined by two satellites created by a cooperative program between China and Brazil and one, four satellite, private system.

High Resolution: The twelve high resolution systems will provide an order of magnitude improvement in ground resolution, at the expense of less area and multispectral capability. With the exception of one Indian and one Russian satellite, these satellites are all funded and operated by private corporations. The almost exclusive interest of the private sector investors in the high resolution systems indicates their belief that this is the space capability required to create commercially valuable information products.

Hyperspectral: The three government funded hyperspectral satellites and the proposed private system will explore the potential for the development of new multispectral analysis based applications by providing near continuous radiometry over the visible, near IR and short wave IR spectrum.

Radar:The current Canadian and ESA radar programs will be continued into this period as well. Radar's all weather capability makes it the instrument of necessity for many observational problems and it will become increasingly valuable for general problems as better techniques for analysis are developed, including the integration of radar and optical data.

Technical Overview

The best way to understand the scope and variety of the data which will be available from the new millennium fleet is to look at the three principal observational dimensions of its data, ground resolution, land coverage frequency and spectral coverage. They are tied together in sometimes unfortunate ways, (from the user's point of view), by the laws of optics, orbital mechanics and the ultimate decision maker, economics. No one system can provide all the measurement features needed by the user community.

The following discussion will present three summary maps of the data scope and variation which will be provided by the 31 satellites; land coverage and ground resolution, the spectral position of measured bands and the ground resolution of each band.

Land coverage and ground resolution: All but two of the satellites will cover the total land mass since they are in polar sun synchronous orbit. The two exceptions are SPIN-2 which is in a 65 degree orbit and QuickBird for which a 52 degree orbit inclination is being considered.

Land coverage frequency must be considered in two ways, the frequency with which the system can provide images of the total globe, and the time it takes to revisit a given site. Because global coverage frequency is inversely proportional to the sensor's ground field of view or swath width, this parameter will be presented as one measure of coverage capability in the following discussion.

Figure 4 presents the ground resolution and the ground swath width for all of the satellites noted above. This plot provides a graphic illustration of the difference in coverage and resolution between the four classes of satellites. Only two sensors escape the boxes. The IRS C,D Pan sensor flies on a satellite that is in the Landsat-like box, but lies outside that category because it sacrifices swath width for its higher resolution. However, it can be pointed off the orbit path which allows 2 to 4 day revisits to specific sites. SPIN-2's escape from its box is described below.

The Landsat-type satellites are designed to provide fairly frequent global coverage by choosing the sensor ground swath and orbit parameters so that they will cover the complete equatorial surface each orbital repeat cycle. The current and planned satellites achieve this by having ground swaths between 120 and 200 kilometers. Their orbital periods and thus global coverage times, vary from 16 days for Landsat to 22, 24, and 26 days for the Indian, French and China/Brazilian satellites. Taking singly, even these repeat cycles are too long for many applications. However, the similarity of the sensor data from the 12 satellites in this group can, for many applications, make it possible to use the data from all of the satellites interchangeably and thus have available the one to seven day coverage rates illustrated on figure 5.

Figure 5 shows the number of the Landsat-like satellites that you could see ( or more precisely the number of satellites that could see you at an equatorial [1] site) on any day over a randomly selected 100 day period for the three satellites now in orbit, for the 8 government satellites in orbit in the year 2000, and for the 12 satellite fleet [2] resulting from adding the four Resource21 birds. The aperiodic nature of the second plot cries out for effective international cooperation to optimize the spacing of the coverage opportunities. (There is no indication that this is likely to happen).

The High Resolution Group: The much narrower ground swaths of the high resolution sensors, 4 to 36 kilometers, can only achieve total global coverage in periods ranging from 4 months to 2 years. Since the high resolution sensors being planned generate communication rates between 20 to 100 times that of Landsat this design limitation is caused by the practical and economic limits of the data collection systems. SPIN 2 avoids this problem since its data collection system is film return which places it in its unique position on the chart. However, for many users the good news is that the satellites are designed to be capable of quickly pointing off nadir and thus can see any given site in 2 to 4 days. Thus, even two high resolution satellites properly synchronized could provide daily repeat coverage nearly anywhere.

WARNING-Clouds severely effect the above quoted repeat times: The above discussion of the repeat times should not be used without at least doubling the numbers quoted to provide some sense of the effect of clouds on the actual ability to get cloud free images, i.e. to actually see the desired targets. Figure 6 presents the results of a simulation which recorded the best cloud free percentage images for each Landsat WRS site [3] collected over a 16 day period in early spring by using one, two, three and four satellites orbiting over a WRS grid containing the % of cloud cover in 5% increments for every day of the year [4].

The fact is plain, our planet is cloudy and the clouds obstruct our satellite land view more than we would like for many of our time critical applications. The message is equally plain, multiple satellites (or radar) are required if we need assured land coverage in short time periods of weeks to months. (Note the four maps also represent very closely the collection capabilities of one satellite for 16, 32, 48 and 64 days.) As the maps make plain, the problem is geographically focused and as would be expected the agricultural belts, where frequent data are most required, are the cloudiest. It remains to be seen whether the small target areas and pointability of the high resolution systems will provide higher cloud-free data returns than those calculated for the large area non-pointing systems illustrated above.

The Hyperspectral Group: the US government is launching three satellites to test the full potential of multispectral analysis for the identification of both man-made and natural surface elements. Because of the very high data rates required by the hyperspectral sensors, the resolution of these systems has been restricted to thirty meters. There is also a sense that thirty meters may well be more than sufficient to characterize the majority of at least the natural targets, i.e. mineral and vegetative cover. The Australian government is stimulating interest in the private sector for the commercial development and operation of a near hyperspectral system, since the sensor uses two groups of 32 bands instead of the spectrometers of the other systems.

Radar: The current and proposed radar satellites can provide data in a variety of resolution, swath combinations. The values on the figure represent their high resolution capabilities. Again, the practical limits on data rate have been an important factor in their resolution/swath tradeoffs. It is beyond the scope of this paper to define the large range of resolution/coverage products available from these satellites and the potential user is urged to contact his friendly data supplier.

Spectral Coverage and Ground Resolution: Figure 7 presents the bands measured by each of the multispectral instruments. Figure 8 provides the resolution of each band. (The bands are listed under their Landsat 7 band counterparts).

The Landsat-like Group: As can be seen on figure 7, except for the EOS AM-1 ASTER instrument, the bands from all the satellites are very close to those used by Landsat. This is of course because the Landsat bands were placed in nearly all the wavelength windows free of severe atmospheric absorption. The Landsat-like satellites emphasize multispectral coverage, all of them having at least the lower SWIR band and three including both the upper SWIR and TIR bands. ASTER will provide even greater spectral definition in the upper SWIR and TIR regions.

It is important to note that all multispectral data may not be equally usable for all applications even when the same bands are available. For analysis that are critically dependent on measuring the absolute reflected radiation over years to decades, sensor calibration becomes a critical parameter. Landsat 7 and Resource21 systems will have sun and moon based calibration capabilities while the other systems will rely on internal lamps and ground targets for their calibration. Equally important to such applications is the ability to adjust the measured radiation for the varying atmospheric conditions. NASA is planning to operate Landsat 7 and AM-1 in very close proximity to measure the atmospheric input using an AM-1 sensor (MODIS).

As shown on figure 8, the multispectral resolutions range from ten to thirty meters with the exception of the six meter sensor on IRS-P5 and 2A which achieve their higher resolutions by reductions in swath width, (see figure 4). The panchromatic sensors of interest in this group range from 6 to 20 meters. Experience with integrating the ten meter panchromatic data and the twenty meter multispectral data from SPOT has shown the value for many applications of the use of the pan band in sharpening the color bands.

The High Resolution Group: In contrast to the Landsat-like group, half of this group has limited multispectral coverage, while the other half has none at all. It is obvious that as a group the critical measurement is the ground resolution which is essential for identifying man-made objects and for updating maps and GIS data bases. Whereas in the Landsat-like group the pan bands are used to sharpen the color bands, in this group the color bands will probably be used to add additional information to the pan band data.

The Hyperspectral Group: The hyperspectral satellites are being flown to explore the potential of using the full spectral response over the VNIR and SWIR spectrum. Note on figure 8, that hyperspectral is being defined as sensors with 32 to 256 bands per VNIR or SWIR range.

Radar: While the current and planned radar satellites will have only one frequency, they do have several polarization options and thus have a multidimensional analysis possibility analogous to the optical system's multispectral analysis. Again, the reader is advised to contact the radar data providers to get an understanding of the full range of data products their systems are capable of providing.

Data Availability: The good news is that there are plans for 31 satellites capable of providing a wide range of land data information products. The amount, sophistication and variety of the land data that COULD be available for analysis is staggering. It is probably equally good news that all the data, government and private, except Landsat 7 data, will be available commercially at market determined prices. It's even better news that current US law requires that Landsat 7 data be made available to all at the " cost of furnishing user requests". However, to be available the data must be first acquired. Landsat 7 is the only system which plans to acquire and archive multiple total land cover data sets each year. The other government systems will be collecting for their own purposes and for orders acquired by their commercial sales outlets. The private systems' acquisition plans will be totally market driven.

Why So Many Satellites?: 31 satellites may seem to be more than a few too many for needs of the earth observing community. Before making that judgment however, it may be useful to consider the following points.

As noted above, none of the planned satellites will provide all of the data characteristics needed by the broad range of user requirements. Thus at least four systems would be needed to provide the different data types the fleet is currently planning. The day of the battlestar galactica, single satellites with suites of many instruments, appears to be over [5].

The need for multiple satellites was also discussed in the section on coverage frequency, which emphasized the negative effects of the world's 50% cloud cover.. Resource21 is planning a four satellite system to meet their customers need for weekly observation of crop conditions. The Global Change Science goal of global of seasonal coverage requires a minimum of three to four satellites. The use of satellite data for disaster analysis and relief planning can be very effective but only if the satellite can acquire imagery almost immediately after the event, a possibility only if two or more pointable sensors are in orbit. For weather related disasters, radar is often the only system which can see the ground. Again multiple radar satellites would be required for sufficiently rapid coverage.

Finally there is the need to assure operational stability. In the last two years, two land observation satellites failed to make orbit, Landsat 6 and SPIN-2, and two failed on orbit prematurely, SPOT 3 and ADEOS. Obviously more than one system must be available to provide the operational assurance required if users are to be able to make the data a requirement for their activities. India, EarthWatch and Resource21 are planning operationally robust systems of four satellites each. CBERS and all of the high resolution satellite providers are planning two systems each.

Concluding remarks

This paper attempts to provide the best current information available on the world wide, public and private, plans for operating land observation satellites in the new millennium. it also provides charts summarizing the types of measurements and their parameter ranges the satellite/sensor fleet will be capable of providing. Much of the data presented in the previous charts have been combined and provided in a full service reference chart in figure 9.

[NOTE: Figure 9 is an Excel file.

Linking to the file will give you the opportunity to download the file to your system.
You will have to download the file and have a recent version of Excel (Version 5.0 or better for Mac) on your system to view it.
To download the file, press "SHIFT" ("OPTION" on a Mac) while selecting figure 9.]

The goal of the report is convince the land information user community and especially the so-called "value added" experts in industry and academia, that their cup is about to runneth over. The satellites are really coming, though probably not in the numbers presented in this paper. However half are government funded and most of these are in or on the path toward construction. If only half the proposed commercial satellites make orbit there will be 20 satellites in orbit in 2000.

The really big bucks, literally billions, required to create the satellite systems are being spent by both the public and private sectors.. It is now up to the users, public and private to invest in the development of the analysis technologies, the information products and the applications that will generate the dollars that will keep the new millennium satellites flying. The question is are you, they, anybody ready for the deluge?

NOTES

[1] Since polar orbits cross near the polls and have constant width ground swaths, the ground overlap between orbits increases with latitude. At 60 degrees the overlap is 100% and thus the equatorial coverage rate doubles.
[2] I have taken the liberty in the third plot of assuming that the two Indian satellite series will eventually be planned to have the same orbital period in order to create the advantage of a total periodic period of 6 days in place of the aperiodic 11 and 12 day periods currently quoted (which also assumes that the two satellites of each series are placed in orbits halving their 22 and 24 day return periods.)
[3] The maps are the world as seen by the constant swath Landsat image and thus are greatly distorted at the higher latitudes. The Landsat World Reference System (WRS) maps the world in 30,107 185x170 kilometer squares.
[4] The WRS cloud data were created by the Air Force from a global data set for the year 1977 and represent the cloud coverage at the 9:30 AM local crossing time of the Landsat satellite.
[5] It is however worth noting that four of the Landsat-like systems, SPOT, IRS, CBERS and AM-1, do carry one or two other wide field of view sensors to provide daily to weekly coverage to supplement their main sensor data. (See notes and box on figure 9 for details).

---

[an error occurred while processing this directive] [an error occurred while processing this directive]