The Role of Small Satellites in NASA and NOAA Earth Observation Programs (2000)

With few exceptions, small satellite programs are currently implemented using traditional management approaches. However, realizing many of the potential benefits of a small satellite approach requires innovations in programmatic style. That is, as with larger satellites, small satellite programs can be costly or slow to implement. Similarly, within some limit, larger satellite programs can be managed in ways that result in comparatively low-cost, short-development-time missions. (Clearly, very large and complex missions require more time for development and integration.) The discussion below examines some of the management and programmatic issues related to small satellite missions and presents committee views on how to reap the potential benefits of a small satellite approach.

The fundamental attribute of small satellite missions is their potential to shorten development time; all other benefits flow from this basic property. With longer development cycles, total system costs increase. Technological advances take place, but they cannot be pursued because the design is completed long before the mission is scheduled for launch. Improvements in scientific understanding that may occur during the development of the mission often have little impact on sensor design.

There are many ways to achieve shorter development times through changes in management practices. One way is for the government to limit the oversight burden by reducing the number of documents and reports that must be delivered by the contractor and the number of formal reviews that must be undertaken. For example, it has been estimated that a single major review of the Earth Observing System Data and Information System (EOSDIS) core system being built for the National Aeronautics and Space Administration (NASA) produced over 5,000 pages of material at a cost of several person-years of effort.¹ TRW estimates that the Critical Design Review for the Total Ozone Mapping Sensor, now in orbit as part of the Earth Probes program, required 10 person-years of effort. A lower cost approach is to reduce the required paperwork to sufficient information to allow a knowledgeable engineer or manager to evaluate progress and risks. Such an approach has been dubbed "insight versus oversight."²

Although there is anecdotal evidence of too much required paperwork, the committee is not aware of any formal study of optimal levels of oversight.³ In any event, reduction of oversight does not mitigate the importance, or lessen the involvement, of government personnel, as they play a valuable role in ensuring mission success if they are a working part of the team rather than just project observers.

Many programs—Earth Observing System (EOS) AM is an example⁴—have incorporated NASA personnel into the product development teams. The Air Force has also been well integrated into several programs that used the product team approach. However, it is difficult to incorporate these plans into a proposal, as the government's evaluation of cost and effort can be adversely affected if it is believed that the contractor intends to use the government personnel to make decisions or accomplish tasks that should be done by the contractor. Closer integration between the technical personnel and the end user could also achieve shorter development times. This approach conceivably could extend into the proposal phase as well as the development phase. In this way, there would be fewer breaks in communication between producers and users, and new developments on either the technical or scientific side could be rapidly incorporated into the mission.

Building on experiences in the commercial manufacturing industry, contractors could rely more heavily on integrated product development teams rather than dividing the effort, as is typically done, into separate mission components. This approach could extend to all parts of the mission, including the ground, launch, bus, and sensor systems and might reduce failures due to lack of communication as well as stimulate more creativity and innovation. Notable examples of this approach can be seen in the automobile industry and in much of the Japanese manufacturing industry where individual companies regularly work together to produce a complete system. The General Motors Saturn automobile plant follows this model, as worker teams participate in the end-to-end manufacturing process.⁵

There are many risks to the management approaches outlined above. Reduced government oversight may substantially increase the risk of failure by not providing an outside viewpoint during the development process. Shorter development times (and perhaps lower profit margins) may encourage companies to take shortcuts and higher risk (but lower cost) options without a thorough analysis. None of these potential problems is an expected or proven consequence of reduced oversight; indeed, some might occasionally occur with present levels of government oversight.

Satellite development and launch operations are complicated, and seemingly innocuous errors can propagate through the system. For example, one of the early Pegasus XL missions failed because designers relied only on numerical modeling of vehicle aerodynamics rather than actual wind tunnel tests. When the model was run with incorrect numbers, the resultant control software became flawed. Fleeter (1997) reports that a similar reliance on models rather than actual tests resulted in the near failure of the Clementine mission.⁶

Government procurements are often based on rapid analyses by contractors. This leads to the additional risk of not detecting errors that might otherwise be found in a longer government review. In the best scenario, these errors are discovered when the actual development effort begins. Costs may escalate, but the program is not compromised. In a more pessimistic scenario, the errors are not found and program failure may result.

Current government funding and procurement practices may be out of step with the style of proposal and system development best suited to small satellite missions. For example, detailed procurement rules may significantly delay acquisition of critical parts or may force contractors to acquire parts much earlier than needed. Funding profiles may not reflect the need for rapid, early delivery of funds necessary to support a short development schedule. Requirements for long-lead-time space-qualified parts may actually decrease mission reliability by compressing the time available for assembly, integration, and test.

More subtle issues arise on the contractor side. Integrated development teams may require that several companies work closely together on the overall system. Such close cooperation depends on open communication, but this will be difficult if concerns over proprietary information are present. The U.S. space industry is not presently structured as a set of independent component builders. Few companies are involved in only one aspect of satellite missions, and many have business units that could provide end-to-end solutions from launcher to bus to sensor. Thus companies may be unwilling to disclose proprietary information to a partner that may be a potential competitor on a future project. Given industry trends of reduced profit margins and increased competition, these concerns are likely to increase.

The computer industry is often cited as an example of the benefits of a "faster, better, cheaper" approach. It is notable, however, that few companies manufacture all of the major subsystems of a computer—central processing unit (CPU), disk, backplane, operating system, and applications software. Thus a chip manufacturer such as Intel is more willing to work with an operating system vendor such as Microsoft on the details of a new CPU because it knows Microsoft will not be competing in the CPU marketplace. Tighter alliances between competitors in the space industry may run counter to traditional business practices as well as to government regulations.

Desirable characteristics of small satellite missions, such as more rapid development schedules and lower costs, are also associated with an increased likelihood of mission failure. Understanding, mitigating, and responding to the risk of failure are thus central issues in small satellite programs.

Space launch and space operations are inherently risky. One response to failure is to launch another mission immediately. The formerly classified Corona program of photo-reconnaissance missions in the late 1950s and 1960s was based on this principle. It took 12 failures before there was a successful mission (Ruffner, 1995). This strategy must be based on a firm scientific and political recognition of the vital importance of the specific mission. Other failures may be less dramatic. A contractor may simply run out of funds during development or may fall far behind schedule. In such cases, the response may be to cancel the mission outright (which has rarely been done, but see the discussion of the Clark mission in Appendix D) or to increase funding in the hope that the program will recover.

Recent government practice has been to encourage contractors to deliver within a fixed cost. The government perception is that larger companies more easily absorb cost overruns. This further reduces profit margins and makes it difficult to participate in such contracts, because the profits are also limited on the upside of a successful program. Such risks must be justified to corporate investors, who may be unwilling to risk the downside potential of such investments. The net effect may be to filter larger companies out of the process, creating a reliance on small companies (which have limited capital) to participate in small missions.

If the government is not willing to accept failures in the development process, then it must be prepared to either manage the risks in this process more closely or expend more funds to keep the development process on track. A similar issue is whether contractors will undertake designs with substantial risk if there is a perception that such failures (perceived or real) will damage their credibility for subsequent competitions. Such a response could result in small satellite missions that resemble more traditional satellite programs. The many potential benefits of smaller missions cannot be realized if they are managed as traditional "large" missions, with the only difference being the size of the budget.

There are costs associated with the management style discussed above that are frequently not considered and which comprise a special category of risk for small satellite missions.

The first such hidden cost involves the tendency of small, short development cycle missions, operating on low margins, to live off technical developments produced by much larger programs.⁷ Small, low-margin missions cannot afford to explore and develop new technologies that have not yet been proven. They are, however, very adept at capitalizing on opportunities presented by other programs. For example, Microwave Monolithic Integrated Circuit technology, developed as an offshoot of various Department of Defense programs to improve millimeter-wave radar, is now being incorporated in the EOS Microwave Limb Scanner. Unless there is a continual research and development activity that can provide such new technologies, this source will eventually run dry, as the small missions typically extract these new developments but do not replenish them. Development of new technology requires time and funding. Larger programs have provided the development in the past because they had the margins and funds. Small programs can also provide the development, but only if the schedule and funding are made available. Cost, schedule, and technological innovation cannot all be optimized simultaneously.

In the past, much of the Earth remote sensing research and development was done under the auspices of NASA's Research and Analysis (R&A) program. This particular role of the R&A program has diminished considerably in the last decade.⁸ Instead, the science community often focuses on what can be measured, rather than looking at the scientific questions that need to be answered with an eye to matching these with appropriate technology. Thus technology development often drives science rather than the other way around. New NASA programs, such as the Instrument Incubator concept, are attempting to bring together technologists and scientists in new ways so that technology is driven by an understanding of science issues and questions.

The second hidden cost is the human impacts of tightly integrated, small rapid development teams. While such a work structure can be exciting, it does exact a toll on workers through the demand for long hours, intense activity, and strong pressures for success (NRC, 1997a,b). Some people thrive in this "pressure cooker" environment, but many involved in small satellite missions report that they are unwilling to commit again to such projects for several years. This reluctance leads to turnover issues, as follow-on efforts could require entirely new teams—with the attendant learning curve costs and risks.

The small satellite approach carries with it several risks concerning scientific return:

1. Rapid development missions are often focused on "small" problems. Missions are not designed for long life and are sometimes viewed as "one shot" opportunities.

2. Missions employing small satellites are more likely to be developed as part of a program of technology demonstrations as opposed to a program in which the science return is paramount.

3. Small missions require a well-defined focus in order to keep them simple and costs low. This approach may not work well for scientific studies that require measurements of many processes.

4. Data processing and distribution may be related to relatively lower priority, thus making it difficult for nonproject scientists to gain access to the data. This problem could be exacerbated in the case of missions led by a principal investigator (PI) should research investigations become centered on an individual's personal scientific interests.

5. With more single-sensor missions, the proportion of funds spent on satellite hardware and launch costs will increase. Such funds might otherwise be spent on scientific research.

Measurements that are collected by small, focused missions will tend to be the vision of an individual or team rather than the vision of the broader Earth science community. While this approach has resulted in significant scientific progress in the past and will likely continue to do so, the increasing complexity and interdisciplinary nature of Earth science research requires larger and more broadly based scientific teams (NRC, 1997c). In fact, the scientific research community increasingly views remote sensing systems as shared resources rather than the province of remote sensing specialists.

Small satellite missions provide an increased level of programmatic flexibility throughout the funding process. Missions can be delayed or accelerated more easily because both the scientific constituency and the scope of the program are smaller. However, the relative ease with which small programs can be delayed can have scientific consequences that extend beyond a particular mission—for example, when an integrated observing system is being developed that depends on satellite clusters or constellations, or when anticipated scientific returns are driven by an assumed schedule of satellite missions. The danger is that critical measurements will not be made with sufficient quality, or for a sufficient period of time, or that necessary complementary data sets will not be available.

Mitigation of the scientific risks associated with small satellite operations requires innovative program management. Indeed, an effective response to the challenges posed by inclusion of small satellites in mission plans can lead to a more effective mission as measured by the science return.

A calibration and validation strategy should be developed for every mission, as well as for the overall suite of observing systems. In an effort to lower costs and to accommodate the mission on a small platform and small launch vehicle, on-board calibration systems as well as extensive prelaunch characterization are sometimes omitted or cut short. However, many important processes in the Earth system have long characteristic time scales so that separation of natural variability from variability in the sensor system is a difficult process. This does not necessarily mean that every mission must have complicated (and costly) on-board calibration systems. Instead, a well-developed (and well-supported) plan to ensure scientifically based levels of calibration and validation must include all aspects of an observing strategy. This may include prelaunch characterization of the sensor, strategies to ensure dynamic continuity of the data stream as sensors change and evolve, and field programs to quantify data quality.

The science community should also assess science needs for precise calibration of individual sensors against the need for more intensive temporal and spatial sampling, especially in the context of constellations of small satellites. The total error of any data set will be a convolution of the quality of the individual measurements that go into the set and the sampling error of the data (e.g., unresolved processes because of inadequate sampling, biases because of cloudiness, etc.). Many Earth science data sets must be continued for the foreseeable future due to either operational requirements or the presence of long-term (interannual to decadal) fluctuations. A strategy is required to ensure that new technologies and new measurements can migrate to the more slowly evolving series of operational satellite systems, such as the National Polar-orbiting Operational Environmental Satellite System (NPOESS). This includes, for example, a means to ensure a balance between the use of innovative technologies and of flight-proven systems.

Data processing and distribution must be an integral part of any mission, large or small, although these mission elements may result in higher costs to accommodate the on-board storage and downlink capabilities necessary to acquire global data. For example, the Lewis mission design was severely restricted in its ability to collect data (roughly 200 scenes), which essentially precluded its use as a global observing platform. Significant processing capabilities (either on board or on the ground) may be needed to reduce the overall costs of a constellation approach, such as data assembly to provide global observation from narrow swath sensors or compressing of downlinked data streams.

Adoption of a constellation approach for an Earth observation mission will require a far different management and budgeting structure within NASA than that currently in place. While the operational agencies have experience in maintaining small constellations, NASA does not. An observing strategy based on a dozen or so small satellites

that would need to be maintained for many years—such as Clouds and the Earth's Radiation Energy System type observations of Earth's radiation processes—will require different planning, development, and operating strategies than the typical NASA mission. There are many precedents, such as the Global Positioning System, the Defense Meteorological Satellite Program, and the Polar-orbiting Operational Environmental Satellite, that could provide important lessons if NASA decides to pursue such a course.

Increasing the number and decreasing the size of missions can have beneficial effects, but the increased burden on both contractors and government agencies should be recognized. For example, there will be a significant increase in time and effort for proposal preparation and evaluation. Smaller, shorter-time-duration missions imply more opportunities to propose new ideas and technologies. Since it is essential that the science community be involved in the evaluation process, this will mean more time being spent on reviews and panels. In addition, the overall observation strategy must be continually updated and evaluated against new opportunities to ensure that it remains relevant. Thus the benefit of having more missions is likely to be accompanied by an increased burden on the science community, as well as increased development and proposal costs.

With the trend toward flatter management structures in small satellite missions, PIs are being urged (if not required) to assume more and more responsibilities for the end-to-end system. In addition, with smaller profit margins, vendors of the various mission elements (launcher, satellite bus, etc.) are less willing to spend significant amounts of time in proposal development and support. Instead, basic information is provided, and it is up to the scientist to evaluate and assemble the components. The recent Earth System Science Pathfinder process resulted in over 85 initial proposals. Twelve proposals were selected for further development in a second round of evaluation. Finally, two proposals (and one backup) were selected. Although the two-stage proposal process is meant to reduce the proposal burden, this is often not the case. The amount of effort on the part of the PI to develop a credible initial phase proposal is significant. Given the low success rate in both phases, it is in the interest of the investigators to make the initial proposal as complete as possible.

The loss in scientific productivity from these increased burdens is difficult to quantify, but is likely not negligible. Any complete evaluation of the benefits of small satellite missions should attempt to account for this cost.

Management innovations are needed to exploit the potential advantages offered by the small satellite approach. Maintenance of science quality must be foremost in implementing these changes, however. This in turn will require a science-driven (versus technology-driven) approach to small satellite missions, as well as development and implementation of strategies to maintain dynamic continuity between sensors on successive satellites. An overall strategy for Earth observation is needed to serve as the benchmark against which to evaluate new missions, especially if research and operational observing systems move toward a constellation approach.

Use of small satellites in either a smaller and faster or constellation manner will require management to rethink how it assesses and manages risk. Compared to large missions, management will need to tolerate higher levels of risk and develop a more flexible response to failure. The management of small satellite programs also needs to adopt a more streamlined and less hierarchical approach than is typical for larger missions. It is advantageous if interactions between contractor and government emphasize insight rather than oversight. Finally, smaller product development teams may lower costs, but this should be achieved by improving processes and increasing risk tolerance—not by increasing pressure on the team.

The Earth science community must adjust to these new approaches. Sampling strategies must be placed on an equal footing with the drive to improve sensor quality. The community must be willing to streamline its proposal development and review procedures. Operational observing systems such as NPOESS will play an increasingly important role in Earth system research along with the traditional NASA research missions, and the research community must evaluate the full spectrum of Earth remote sensing missions in the context of a coherent observing strategy.

Fleeter, R. 1997. Mr. Murphy on small spacecraft and rocket reliability. Launchspace 2:14-16.

National Research Council (NRC), Space Studies Board. 1997a. Lessons Learned from the Clementine Mission. Washington, D.C.: National Academy Press.

National Research Council (NRC), Space Studies Board and Board on Atmospheric Sciences and Climate. 1997b. Scientific Assessment of NASA's SMEX-MIDEX Space Physics Mission Selections. Washington, D.C.: National Academy Press.

National Research Council (NRC), U.S. Global Change Research Committee. 1997c. Pathways. Washington, D.C.: National Academy Press.

National Research Council (NRC), Space Studies Board. 1998. Supporting Research and Data Analysis in NASA's Science Programs: Engines for Innovation and Synthesis. Washington, D.C.: National Academy Press.

Ruffner, K.C. 1995. CORONA: America's First Satellite Program. Washington, D.C.: Central Intelligence Agency History Staff. Available through the National Technical Information Service.

Shaiken, H., S. Lopez, and I. Mankita. 1996. Experienced Workers and New Ways of Organizing Work: A Case Study of Saturn and Chrysler Jefferson North. National Center for the Workplace Working Paper Series, No. 5. Berkeley, Calif.: Institute of Industrial Relations, University of California, Berkeley. Available online at <http://violet.lib.berkeley.edu/~iir/>.