Toward New Partnerships In Remote Sensing: Government, the Private Sector, and Earth Science Research (2002)

Because the private sector and the scientific community (including those in universities, government, and other research organizations) have different requirements and operate in different ways, forging an effective collaborative relationship will take careful work. The situation is even more difficult because neither community can exercise substantial economic force. The scientific community depends, for much of its research funding, on a decentralized process based on annual appropriations for science agency budgets, distributed through competitive scientific review processes. The commercial remote sensing industry is also decentralized and remains economically fragile. For the industry to survive, it must make a financial profit in a commercial market. It is unlikely to succeed by serving the needs of small, niche markets such as the scientific research community. Despite these obvious differences, those in both the scientific community and the commercial remote sensing industry believe that significant advantages are to be gained by working together.

A robust remote sensing industry can provide valuable new data, such as the high-resolution data used in many commercial applications, to the scientific community. High-resolution data, essential for certain types of research, are currently unavailable except through the private sector. Production of these data could probably not be justified by a government science agency because of their high cost and because the number of scientific users is still low. Because of their value to other customers, however, these data can be produced by the private sector for a commercial market, and they are also useful to scientists. In short, the data

needs of the applied commercial-user community are serving to expand the data resources available to scientists.

The use of commercial remote sensing data by scientists can, in turn, benefit the commercial remote sensing industry. The private sector often looks to scientists and engineers for innovations in the construction, use, and application of new sensors and new algorithms; for research that eventually leads to the development of new applications; and for training technical experts who will work in applied settings and be responsible for using commercial imagery.¹

Although both the commercial remote sensing industry and the scientific research community are important for policy and economic reasons in the United States, and government science agencies are important to each, significant differences and some incompatibilities still exist in the ways these groups operate. Such differences complicate the process of building partnerships. Workshop participants from both the scientific community and the private sector emphasized that the need to generate a return on investment drives the private sector. Individual firms must identify a viable commercial market consisting of actual or potential customers who have well-defined requirements and the resources necessary to purchase data.² According to a workshop participant from the private sector, defense and national security applications present commercial image providers with a mature market possessing well-known requirements. Civil infrastructure applications represent a potential second market, but this market is fragmented and its requirements are still evolving. Two other potential markets are the management of natural resources and of the environment; scientific research is a component of each of these. These primary and secondary markets (defense, national security, civil infrastructure, natural resources, and the environment) are all, to a large extent, government markets. Commercial markets also exist—for applications such as agricultural monitoring and crop forecasting—but these are not well developed.

Workshop participants identified critical differences between the data requirements of the science community and the commercial practices of data providers. Table 4.1 identifies these differences in data archiving and management, data continuity, resolution, instrument specifications and calibration, satellite tasking, cost, and data access. These issues are discussed later in the chapter.

Most commercial data providers acquire data by tasking on demand, that is, data are obtained when a specific request is received from a customer for the data. Additional data may be collected in areas of anticipated use, such as urban planning, if estimates of future sales justify the investment both in operations and processing and in archiving. This approach is in contrast with that of the government, typified in Landsat and the Earth Observing System (EOS), which is to obtain observations systematically with a goal of full coverage, regardless of current demand, and to archive the data on a routine basis for future use. These data become part of a historical or “heritage” archive that can be used by scientists for research on global change (which by definition requires data at multiple time periods) and regional trends. According to scientists at the workshop, data continuity in space and time is a basic requirement for global change research, and public-private partnerships have not addressed this issue.

Another area in which the requirements of the private sector and the scientific community differ is related to instrument design and development of algorithms. The remote sensing industry understandably takes a proprietary approach to instrument design and algorithms. They are part of the unique intellectual and technological property of a firm and cannot be distributed publicly without undercutting the firm’s competitiveness. This approach leads to what scientists call the

black box, in which technical details of instrument design and performance are not shared with customers. As one of the participants at the workshop noted, black boxes prevent scientists from obtaining technical information that is necessary to their research—specifically in such areas as calibration and validation that require openness to assure data quality.

The science community operates in a way that often appears diametrically opposed to the modus operandi of the commercial remote sensing industry. Scientists generally have very specific data requirements, demanding greater assurance of data quality than most commercial users. Specifically, scientists require the following information about remote sensing data:

• The sensor mechanisms that collect the data;

• The sensor reference and light sources used to provide calibration of the collected data;

• The changes of the sensor instrument and calibration sources over time; and

• The software algorithms used to make adjustments to the raw collected data.

In addition, scientists using remote sensing data span a host of disciplines (e.g., environmental sciences, coastal sciences, oceanography, forestry, global change research, and atmospheric sciences) and have widely different research interests. Their data requirements are not standardized and can vary greatly from project to project in terms of geographic coverage and the required spatial and temporal resolution. Scientists often require wide-area coverage and access to archival data to address trends. Scientists attending the workshop reported that data of low spatial resolution may be sufficient for global science that requires long-term observations over large areas. Their evolving data needs are more often met by systematic data acquisition or monitoring programs that obtain remote sensing data on a sustained and regular basis than by commercial tasking practices that require specification of data before tasking and delivery of the data.

At the same time, the availability of higher-resolution data may be creating new research opportunities. Government science agencies have supplied scientists working on global change research, for example, with large-area data sets that have been obtained at lower and medium resolution (e.g., 15 to 30 meters) and at wider swath widths than the high-resolution commercial data available today. Yet, workshop participants noted that the demand for higher-resolution data for science could grow if scientists are able to integrate the new data into long-term observational frameworks.

Most scientists at the workshop also mentioned the need for open access and sharing of data as a problem when using commercially produced remote sensing

data. Because science operates on the basis of collaboration, public review, and evaluation of research, scientists make their data openly available for working with collaborators, testing hypotheses, and replicating scientific results. The intellectual property practices of the private sector often run counter to scientists’ ways of operating.

Significant differences exist between the scientific community and the private sector in data management, including archiving and data processing. In general, scientists require well-calibrated, consistent data records at multiple points in time for intertemporal comparisons. Management of these data involves the following: minimizing deterioration or loss of data; providing access to “heritage” data for current and future research; and making all data available through an active data center for immediate use, and through a long-term data archive for future use. Scientists also require verification and validation of remote sensing data according to open protocols.

Private sector data providers respond to a different set of requirements, which are a result of their need to operate a commercially successful enterprise. Many commercial clients do not require validation and verification, or if they do, they do not need to know the protocols that are used. Once the data are delivered, the commercial firm archives the data for potential resale to another client at a later time. Ownership of the data may be transferred to the user in some cases but not in others. The treatment of intellectual property is not standardized across the private sector, and NASA reported to the steering committee that the agency negotiated separate intellectual property agreements with each of the companies involved in the Science Data Buy.

In response to scientific requirements for data management, NASA, NOAA, and the USGS have established data centers and archives for government-produced data. Guidelines for data archiving are available in other National Research Council (NRC) reports, such as Adequacy of Climate Observing Systems,³ and are monitored by various NRC committees.⁴

As yet, no provision has been made for data management for subsequent scientific use of data obtained through the public-private partnerships discussed at the workshop, other than the stipulations outlined in the U.S. Department of Commerce licensing regulations (see the subsection “Long-Term Archiving,” below). Workshop participants suggested that steps be taken to improve the management of private sector data for subsequent scientific research. Workshop participants suggested that data management protocols should be part of the overall design of all operational data production systems, including public-private partnerships. Data management encompasses quality assurance, validation and verification, data processing, and data archiving. When done according to standards of best practice, according to workshop participants, data management can be costly, amounting to as much as 5 to 10 percent of observing system costs.⁵ This amount does not include the costs of validation and verification, which are currently provided by the government in public-private partnerships.

If science data management were mandated at the time public-private partnerships were negotiated, it would add to the immediate cost of the data, but over the long term would result in making more data available for scientific research at lower unit costs. From a scientific perspective, data management issues to be taken into account in these negotiations would include the cost of data management and archiving, ownership of the data, and future access restrictions. Intellectual property provisions in public-private partnerships might also be standardized (see the subsection “Intellectual Property and Access to Data,” below). Both parties in the partnership might seek ways to provide scientists with ongoing access to data obtained in the partnership for multitemporal comparisons.

Scientists who use commercial remote sensing data for research purposes are always concerned about data quality and processing. The first real requirement is accurate knowledge of the observation location on Earth’s surface. These data can be translated to specific map projections, as needed. The locational precision desired is typically some factor less than the pixel size of the observatory. For example, with Landsat 30-meter observations, it would be desirable to know pixel center location to at least 15 meters. Today this is generally not achieved, even with postacquisition processing of the observations, such as those provided by Earthsat with its orthorectified Landsat data product. The accuracy requirement becomes more difficult with the increased spatial resolution of current

commercial systems with 1- to 4-meter resolution. Accurate geometric correction allows scientists to compare images obtained on multiple dates, to detect change, and to identify the processes at work.

The second requirement that scientists are concerned about is the radiometric processing of the data. When remotely sensed data are radiometrically processed to meaningful physical units such as “percent reflectance,” the spectral measurements obtained on different dates can be accurately compared in order to detect change.

Many scientists who attended the workshop praised the quality of the data processing provided by NASA under both the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and the Science Data Buy (SDB). Many said they believe that the most important government contribution to the use of commercial remote sensing data by scientists has been the validation and verification of the remotely sensed data. In the SeaWiFS program, the quality of the data for science users was assured by having government-supported research scientists intimately involved with OrbImage personnel in the design of the data collection and processing components. This resulted in data that are calibrated to meet science requirements. The SDB instituted an entirely different validation and verification effort to assess the quality of the commercial data—by establishing a validation and verification center at NASA’s Stennis Space Center to undertake these tasks. Stennis Space Center set up an elaborate sensor test-bed facility in which data buy vendors can test their remote sensing instruments according to engineering specifications. In addition, the validation and verification team assures the quality of every SDB data set prior to passing the data to the scientist for research purposes.⁶ If the data do not meet specifications, they are not sent to the scientist. SeaWiFS and SDB validation and verification efforts have also resulted in the creation of detailed metadata that are valuable for current research and future scientific uses of the data.

Because data archiving has both scientific and public benefits, it may be a more appropriate responsibility for the public sector than for the private sector at this time. If the private sector discovers a market for time-series data, private sector interest in long-term archiving could change. Remote sensing data archives contribute to the nation’s information infrastructure and provide benefits to global change research, environmental monitoring and projections, environmental and food security, public health, disaster management, and sustainable management of natural resources. Workshop participants stressed that all raw data and related

metadata should be archived, because the future scientific value of environmental and climate data records is difficult to predict. Participants at the workshop said that plans for archiving should also require that metadata be archived with the remote sensing data, including sufficient information to support future scientific applications, and that back-up data files should be held in a separate location to avoid single-point failure. The growing volume of remote sensing data is likely to place steadily increasing demands on archiving facilities, although improvements in data storage and compression may lead to more cost-effective archiving procedures in the future.

Data archiving presents a different challenge in the context of public-private partnerships. Given the cost of archiving and its long-term public rather than short-term commercial benefits, it is unlikely that the private sector will assume this responsibility. At present, firms that obtain commercial licenses through NOAA are required to provide, at a reasonable cost, land observation data to the National Satellite Land Remote Sensing Data Archive maintained by the USGS in Sioux Falls, South Dakota, before they purge the data.⁷

Data produced or obtained for scientific use under a public-private partnership are in a somewhat different category. These data have not only present scientific value, but future scientific value as historical or “heritage” data for use in multitemporal comparisons. At present, however, the data in the SDB are licensed or restricted, not purchased. It is not clear what the government’s rights are regarding the archiving of these data, except for the stipulations made in the U.S. Department of Commerce licenses issued to commercial remote sensing satellite operators.⁸ Moreover, since separate licenses or data distribution rights were negotiated with each firm participating in the SDB, there may be no single answer.

Workshop participants emphasized that, at a minimum, common standards are needed for documentation and preparation of data, obtained through either the public or the private sector, that are to be placed in permanent archives. The development of standards is most effectively completed early in the design of data systems, and it is most effective when it involves government, the private sector, and the scientific community. These standards could cover metadata, formats, and portability of data, and adherence to them might be required as part of the award process for public-private partnerships or the licensing process.

Data access was the primary focus of one breakout group at the workshop. Participants recognized that the scientific community and the private sector have different perspectives on data rights that provide access to commercial remote sensing data. The commercial remote sensing industry, which is premised on obtaining financial profit through data sales, regards remote sensing data as its intellectual property and protects its right to resell the data. In contrast, the scientific community tends to view large-scale databases as a public or collective commodity that should be openly available for use in research.⁹ Yet if this approach were extended to the private sector, the practice of making data universally available could compromise the property rights of the data producer and could seriously affect, if not destroy, the commercial base of the industry. Many of the issues related to the interests of the scientific community, the public sector, and the private sector were explored in the NRC report A Question of Balance: Private Rights and the Public Interest in Scientific and Technical Databases.¹⁰

There is no easy solution to this problem, and both the scientific and commercial remote sensing communities have debated the issue of data access in the context of a potential Landsat data continuity mission.¹¹ Workshop participants suggested that because the value of data changes over time, private sector data producers might reduce the cost of older commercial data to the research community as a way both to protect their intellectual property and to meet the needs of scientists. Imposing a 14-day delay in providing scientific access to data to ensure the value of the data for commercial customers has been the practice in the SeaWiFS project. This waiting period usually does not cause a problem for scientific data users and could be applied in connection with future science data buy programs.

A related concern expressed by scientists is that of obtaining access to instrument calibration details contained in the “black box.” This information is used to develop algorithms for processing the data and for distinguishing data that correspond to the performance of the instrument from data that have scientific meaning. One possible solution might be “third party” certification that data meet scientific quality standards and calibration as specified by researchers. Scientists may not need to “see” inside the box if they are convinced that the data meet specific quality standards. The federal government could provide this certifica-

tion, much in the way that the National Institute of Standards and Technology provides calibration information for manufacturing, or that the Environmental Protection Agency safeguards proprietary data obtained from the chemical industry (data used to report publicly the industry’s compliance with environmental regulations). Alternatively, a nongovernmental third party could provide certification, much like the assurances supplied by the Underwriter’s Laboratory for consumer electrical devices.

The process of licensing commercial firms to launch Earth observation satellites also affects data rights. NOAA, the U.S. government agency responsible for licensing commercial remote sensing systems, can encourage private sector firms to employ favorable terms for use of the data in research.¹² NOAA may also require that an applicant have a nondiscriminatory access policy and provide broadly defined data rights for systems developed with significant government funding. The licensing process is probably not an appropriate vehicle for imposing operational data management and intellectual property requirements on commercial firms, however, because it does not encompass issues related to markets and applications of the data. In addition, the commercial sector would prefer a faster licensing process; however, there was no suggestion at the workshop that the current process was excessively long, given the requirements of licensing. If development of a remote sensing system involves foreign participants, either from the science community or the private sector, export control licenses are required under the International Traffic in Arms Regulations (ITAR). ITAR requirements can add more time to the licensing process.

Several workshop participants mentioned problems posed by the lengthy time involved in completing government processes associated with establishing public-private partnerships. Given the complexity of the issues under negotiation, the length of the contracting process has created a financial hardship for private sector firms involved in the SDB and SeaWiFS. Specifically, long procurement processes with changing schedules created cost burdens and uncertainties that decreased the real value of data contracts. Two factors are important here. First, if a company must maintain staff to work on a contract and there are delays in start-up, substantial unanticipated costs can be incurred that cannot be recovered. Second, for projects in which private sector investments and capital

are required, the cost of money to cover these long lead items in the face of schedule uncertainty can create a financial hardship. Thus, lengthy contracting procedures may create a private sector disincentive to producing scientific data with government funding.

Another area of great importance is the length and stability of the public-private contract. Space remote sensing systems require substantial front-end investments prior to launch. To finance these systems at reasonable rates, a public-private partnership of less than 5 to 10 years is not likely to be attractive to a commercial firm. Such a long-term commitment may also require that a government agency seek all of the funding for its purchase of scientific data in a single fiscal year or that the government create a means to make financial commitments for long-term contracts with the private sector.

There is value in both ongoing and periodic evaluations of public-private partnerships to facilitate scientific access to commercial data. Ongoing evaluation involves the establishment of performance metrics and their use in program management. Periodic evaluation also involves the appointment of an external group to review the program. Emphasis on performance metrics in the government has existed since the passage of the Government Performance and Results Act of 1993.¹³ This act requires that agencies develop performance standards and periodically report to Congress on their progress in reaching these standards. The experience of government agencies in the years after passage of the act has been that developing accurate and appropriate standards is very difficult, particularly for metrics that measure the quality or utility of scientific research.¹⁴ Nonetheless, despite the difficulty, the discipline of identifying performance metrics for public-private partnerships in remote sensing data for science can be valuable both in making the operational goals of the partnership explicit and in the management of the program.

Some performance measures would be applicable only to the public sector, the private sector, or the scientific community, although others would transcend these boundaries. For example, is the partnership cost-effective for government,

industry, and science? What is its track record in data quality and the efficiency of data processing, delivery, and contracting procedures? Are significant scientific results generated from the data? Specifying and measuring attainment of specific goals require collaboration among all parties early in the process of establishing a public-private remote sensing data partnership, and could involve independent review by external groups.

The true costs of public-private partnerships can go significantly beyond the purchase price of the data. Both the government and the private sector may have to assume expenses in a partnership that they do not face in other transactions. These added expenses are generally not calculated as part of the cost of establishing a public-private partnership.

In the SDB, NASA selects the scientists who will receive data under the program through a competitive process open to its current grantees. This requires the agency to cover the cost of staff time to set up the program and to conduct the competition for data. Prior to redistributing the data to scientists, NASA also conducts validation and verification checks on data received from the vendors, because commercial image providers do not routinely provide such information. In SeaWiFS, calibration, validation, and verification are supported by NASA’s Goddard Space Flight Center.

From the private sector side, the costs of procuring a government contract and responding to contract changes can raise the cost of the transaction to the private sector partner, the government, or both, above the agreed purchase price of the data. In public-private partnerships that involve the launch of satellites, additional expenses are also encountered in engineering, prelaunch, and quality assurance.

Other reports of the Space Studies Board have urged NASA to use full-cost accounting on an agencywide basis,¹⁵ and progress to this end is slow. Realistic cost accounting for public-private partnerships is a separate issue. Because it involves both the public and private sectors, accounting information required for evaluating the true costs of a public-private partnerships may be different from general financial data that are used in agency accounting practices.

A number of the cross-cutting issues discussed above—data archiving and management, data continuity, resolution, intellectual property, licensing, performance metrics, government processes, and realistic cost accounting—are related to the different approach or orientation of scientists and the private sector. These issues must be reconciled if public-private sector partnerships are to play a significant role in producing data for scientific research. The steering committee prioritized these issues, as shown in Table 4.2, according to their significance for public-private partnerships and the difficulty in resolving them.

Two of the most important issues for scientists are data management, including archiving and data processing, and data continuity. The steering committee judged them to be low in complexity because they can be resolved easily, provided sufficient financial resources are available. Intellectual property issues and the impact of government processes on the effectiveness of public-private sector partnerships were judged to be high in significance and high in complexity, largely because there is no convergence on these issues between parties with legitimate but different needs. An example of a high-significance/high-complexity issue is the redistribution of private sector remote sensing data. Scientists believe that the open distribution of research data is critical to the advance of science. Private sector firms producing remote sensing data, on the other hand, must charge scientists for access to their data, both to recover their initial investments and to make a profit for their investors. If the use of commercially generated remote sensing satellite data by the government and the scientific community increases, the need to resolve intellectual property issues will intensify.

Two other issues were of lower significance but were still considered highly complex. These are the development of metrics for evaluation purposes and the issue of licensing data. Finally, realistic cost accounting, although of lower significance and complexity than other topics, is critical in evaluating the costs and benefits of alternative means of acquiring and distributing scientific data.