Review of NASA's Distributed Active Archive Centers (1999)

The Alaska Synthetic Aperture Radar (SAR) Facility (ASF) DAAC is located at the University of Alaska in Fairbanks. Its mission is to process, distribute, and archive SAR data collected at present exclusively from foreign spacecraft—the European Remote Sensing Satellites 1 and 2 (ERS-1,2); the Japanese Earth Remote-Sensing Satellite-1 (JERS-1); and the Canadian RADARSAT-1. As specified in Memoranda of Understanding between NASA and the foreign space agencies, only limited quantities of data are acquired by the Alaska SAR Facility and distributed to NASA-approved investigators. The data that are the most accessible are largely from the Alaska and McMurdo station masks, with

the result that polar researchers are usually satisfied with the DAAC, whereas researchers in other disciplines often are not. In addition, the authority and budget for the development of software and hardware necessary for the ASF DAAC to succeed is not vested with the DAAC itself, but with the Jet Propulsion Laboratory (JPL) developers. However, responsibility for satisfying users rests with the ASF DAAC. A key recommendation is therefore that NASA grant authority for the operation and development of the facility to the DAAC. Finally, the demand for SAR data from U.S. researchers far exceeds the supply, and the panel urges NASA to formulate and help implement a long-term national policy for the acquisition, processing, and use of SAR data for civilian purposes (i.e., research and commercial operations).

The Alaska Synthetic Aperture Radar Facility was established in a Memorandum of Agreement between the University of Alaska and NASA in 1986. Its mission is to establish, operate, and maintain a receiving, image processing, analysis, and archiving facility for SAR data, which are collected exclusively by foreign spacecraft. JPL designed and installed the data acquisition and management system, and SAR data have been acquired and distributed since 1991. The ASF DAAC was created in 1990, and it now handles the data processing, distribution, and archive for the Alaska SAR Facility. The DAAC's current holdings include data from ERS-1 and 2, JERS-1, and RADARSAT-1 missions.

SAR data are useful for applications ranging from sea-ice dynamics to volcanology to ecosystem change (Box 6.1). Consequently, the user community is small, but growing. Its growth, however, is hindered by data restrictions, which are specified by MOUs between NASA and foreign space agencies.

The ASF DAAC is unique within the EOSDIS system, not only because of its international character, but also because the processing information systems are being developed at JPL, rather than by the EOSDIS Core System (ECS) contractor. Because of delays in the ECS, JPL was tasked with developing an interim information system; the final system will be provided by the ECS contractor if sufficient funds are available. However, the JPL and ECS systems are not interoperable, so the ASF DAAC faces a difficult transition period several years from now if current ECS plans are adhered to. Finally, unlike most other DAACs, the ASF DAAC is currently managing large data streams. As such, it is the first DAAC to try to ''drink from the fire hose,'' and its experience may well be a preview of what other DAACs will face.

The Panel to Review the ASF DAAC held its site visit on December 18–19, 1997. At that time, the management of the ASF DAAC was in transition. This transition has not yet been fully completed. However, many of the fundamental issues raised by the panel pertain more to NASA's long-term development, management, and use of the facility than to the DAAC's operation and will retain

their relevance beyond the transition period. The following report is based on the results of the site visit and e-mail discussions with DAAC staff in June through September 1998, a meeting with JPL developers in January 1998, and a meeting with Paul Ondrus (ESDIS) in April 1998.

The ASF DAAC is critical to NASA's Earth Science Enterprise and the U.S. Global Change Research Program because it is the primary source of SAR data for U.S. researchers. These data are essential for answering important scientific questions in a variety of disciplines, as well as for detecting and monitoring natural hazards. Further, SAR data become even more useful when combined with other remotely sensed and ground-based data. The DAAC's current holdings are listed in Box 6.2.

The ASF DAAC uses an international standard developed by the Committee on Earth Observation Satellites, rather than the EOSDIS standard (Hierarchical Data Format [HDF]-EOS), because SAR data are recorded as complex numbers, a data type not defined in the HDF-EOS standard.

The supply of SAR data to U.S. researchers is limited by two factors: the operations of data recorders on-board the satellites and MOUs between NASA and the flight agencies of Europe, Canada, and Japan. The DAAC has little control over either. All three MOUs restrict the DAAC from distributing data to unapproved users, and the DAAC is therefore hamstrung when trying to meet the demand of U.S. researchers.

Without an operational on-board recorder to capture the data, the satellite must transmit the data to a station on the ground. This requirement defines the station "mask," the area in which the antenna can track the satellite. For the ASF DAAC, this is a circular area of roughly 3,000-km radius around Fairbanks. Most of the DAAC's holdings image this area.

Neither ERS-1 nor ERS-2 spacecraft carry an on-board data recorder. The MOU between NASA and the European Space Agency (ESA) permits the DAAC to distribute ERS data within the Fairbanks mask to approved investigators. Outside this mask, approved U.S. users must order data from ESA at a cost equal to the processing cost. These costs are paid by the approved user. Unapproved users must pay the market price (about $1,600 per scene). In addition, the ESA has periodically made limited amounts of data available at no cost to U.S. researchers through announcements of opportunity. On the one hand, this process limits access to data by new investigators, and on the other hand, a number of principal investigators are now reaching their quotas.

Similarly, the MOU between NASA and the National Space Development Agency of Japan allows NASA-approved investigators to request JERS data within the Alaska station mask. Prior to the failure of the on-board tape recorder, which was turned off in August 1997, limited amounts of data from outside the Alaska mask were also available from the DAAC. The quantity of data that can be ordered depends on the allocation limit of the project, which is set in advance. In addition, some U.S. researchers have obtained JERS data free of charge by submitting proposals directly to the Japanese government. The JERS satellite ceased operations in October 1998.

For scientists seeking to obtain data outside the ASF mask without paying market prices, RADARSAT may be the most viable choice. RADARSAT is the only civilian radar satellite with a functioning data recorder on-board. Approved users may order data outside the Alaska station mask from the appropriate ground station. These costs, which are paid from the DAAC budget (not by the approved user), may reach $1 million during this coming year. The total U.S. allocation for RADARSAT SAR time is 1,519 minutes per 24-day cycle, of which only 114 minutes of on-board recorder time is allotted to U.S. users. Here again, the MOU between NASA and the Canadian Space Agency restricts the ASF DAAC from distributing data to unapproved investigators. The ASF DAAC is also responsible for ensuring that the above limits are not exceeded. Thus, the U.S. allotment is a

precious resource, which is both oversubscribed (by a factor of four) and a source of contention.

SAR data are processed on demand to Level 1 (Table 6.1), even if the product has been processed before for someone else. (About 30% of JERS data ordered are requested more than once, and the percentage is higher for RADARSAT.) The DAAC adopted on demand product generation because the RADARSAT upgrade, during which the system was reconfigured to support RADARSAT processing, did not include a Level 1 archive. Since then, the MOUs with the Japanese and European flight agencies have increased the data rates considerably (by a factor of 10, according to DAAC staff). As a result, only a small fraction of the DAAC's holdings are actually processed, which presents scaling problems for the DAAC as the demand grows. Furthermore, because of this practice, users cannot be guaranteed that the data ordered are exactly what they need and indeed cover a specific geographic area of interest. This processing policy therefore places a practical limitation on "data mining."

The DAAC has occasionally given high priority to large processing requests,

FIGURE 6.1. Mosaic consisting of more than 3,000 RADARSAT-1 images, which were Quick-Look processed by the Alaska SAR Facility and personnel from NASA's Jet Propulsion Laboratory and the Byrd Polar Research Center. This joint U.S.-Canadian project has yielded a new view of the ice-covered continent, which will enable quantitative analysis of the glaciology, geology and coastal processes of both East and West Antarctica. SOURCE: Richard R. Forster and Kenneth C. Jezek, Byrd Polar Research Center. Copyright Canadian Space Agency 1997.

such as the RADARSAT Antarctic Mapping Project. An initial step was to synthesize a Quick Look mosaic of the rim of Antarctica (Figure 6.1), which enabled verification of complete coverage and permitted preliminary research prior to complete processing. When the DAAC is engaged in such a large processing project, however, the processing requests of all other users are put on hold. Despite their scientific value, large processing projects are a guaranteed source of discontent.

The ASF DAAC currently archives two copies of raw signal data, and will

continue to archive them for 10 years past the end of ESA missions and for 15 years past the end of the Canadian mission. At present, there appear to be no plans to migrate the data to NOAA or to maintain them at the Alaska SAR Facility beyond the 10-or 15-year period following a mission. The issue of long-term custody of SAR data is particularly complex in this case because such a large fraction of the data is subject to restrictions imposed by foreign agencies, most of which are under pressure to recover their investment by charging a fee for the data.

The contents and structure of any long-term archive holding other data products are tightly coupled with the processing strategy. Should the DAAC abandon the on demand processing strategy in response to increasing demand and choose instead to process all data to Level 0, then a Level 0 archive will have to be designed and implemented. In addition, a suitable plan will have to be devised to reprocess systematically all existing holdings. At the site visit, this possibility was mentioned, but no specific plan or budget for this task was presented to the panel.

Approved researchers are able to obtain SAR data from the ASF DAAC at below-commercial prices. Consequently, researchers in a variety of subdisciplines—sea ice, oceanography, glaciology, geology, geophysics, and land applications—are the primary users of the ASF DAAC. In addition, certain government agencies with operational needs (e.g., sea-ice monitoring within shipping lanes) constitute another group of important customers (see "Special Processing," below).

Users can be divided into in-mask and outside-of-mask users. As a consequence of the ASF's geographic location, in-mask users are mainly polar-ice researchers; researchers in most other disciplines must obtain data from outside the mask. The ASF DAAC also processes data from McMurdo station in Antarctica, which further enhances the polar focus of the DAAC. The polar focus, however, does not include Alaskan volcanoes. The panel was surprised that the Volcano Observatory, which is located in the same building as the DAAC, is not moving more aggressively toward using SAR data from the ASF DAAC to monitor volcanoes in the Aleutians, especially with interferometry. The panel suggests that the DAAC encourage scientists at the Volcano Observatory to use DAAC services for this purpose.

The function of the ASF DAAC User Working Group (UWG) is to hold biannual meetings, prepare and maintain a five-year plan, and advise the DAAC

on user needs. At the time of the review, the UWG had not met since the summer of 1996 and was being reconstituted. (A new UWG met in May 1998.) Consequently, the panel spoke only with the former chair of the UWG, Peter Mouginis-Mark. The previous UWG felt that the DAAC was not responsive to its suggestions, so members stopped attending the meetings. In particular, the UWG felt that the priorities of the DAAC should be on the production of data sets that the community wants, rather than on data acquisition, which is a primary ASF function but not a DAAC function. The panel concurs.

The panel strongly believes that an effective UWG must be established. Experience with such groups at other DAACs indicates that they can provide valuable advice and guidance from an interested segment of scientists. Ideally, the membership will span a wide range of disciplines and types of users. The effectiveness of the UWG depends on regular meetings and regular communication of DAAC activities in between meetings. Such communication should document DAAC data distribution, staff activities both locally and with larger EOS functions, and track progress on action items identified at meetings. Typically, the DAAC scientist and an elected member of the UWG co-chair the UWG and run the meetings. The meetings are best designed to elicit interactions rather than be passive information transfer sessions.

As noted above, the DAAC serves users working on problems that fall either in its mask or outside its mask. This dichotomy has resulted in two levels of support and satisfaction among its user community: in-mask users are generally served well and are satisfied, whereas outside-of-mask users are usually not satisfied. Few outside-of-mask users receive the data they need on time, if at all. The absence of a working scheme for tracking data orders (see "User Services," below) only aggravates their impatience. Several dissatisfied users have even gone so far as to call the director of the host institution, the Geophysical Institute. The DAAC is quite aware of the problem, and the current reorganization is intended in part to solve it.

Because in-mask data are more readily available to study polar processes, scientists at the Alaska SAR Facility and the Geophysical Institute are understandably concentrating on polar studies. Many of these studies are regional in nature and do not necessarily require collaboration outside Alaska. The panel encourages the Alaska SAR Facility to take a more outward-looking approach to research and thereby rebuild a satisfied national and international constituency.

One of the most demanding in-mask users is the National Ice Center, which requires Level 1 data within six hours of acquisition to forecast sea ice in the

Arctic region. Its requirements for speed and volume are affecting other users, which may have contributed to the panel's perception of favoritism toward polar studies. The Antarctic Mapping Mission, which monopolized processing resources at the ASF DAAC for several weeks, also contributed to this impression. The Amazon Basin mapping effort, although not focused on polar problems, is another example of a large project that monopolized much of the DAAC resources and affected users.

The DAAC devotes nearly 10% of its budget and five people to User Services. It has conducted several user surveys, the most recent of which is excellent and includes information such as how scientists use ASF DAAC data in their research. It also includes a table of who received data products, how many products they received, and when they received them. These metrics could be tracked over time to help the DAAC measure changes in user satisfaction.

However, the survey and a brief visit to the user services group by the panel indicate that user services staff do not provide an adequate level of scientific and technical support to users. Improving horizontal communication between the divisions of the Alaska SAR Facility, particularly between scientists and the user services division, should help the user services staff to answer questions from the DAAC's customers.

The absence of a data request tracking and feedback system at the DAAC must be addressed soon. An on-line server should allow users to determine the status of their project at any time. Milestones such as scheduling, acquisition, scanning, and processing should be listed when completed. At present, users must first telephone the DAAC to request that data be collected and then telephone again, possibly many times, to obtain the processed data.

The panel praises the efforts of the science team at the Alaska SAR Facility to develop and distribute software for interferometric analysis of SAR data. This type of analysis has perhaps the greatest potential for fostering rapid progress in earth science studies because it is possible to measure crustal deformation and ice flow with a spatial sampling 10 to 100,000 times denser than previous surveying techniques with a comparable precision. Although several space agencies are

actively developing such software, the Alaska SAR Facility is the only group willing to distribute source code for its working software.

The ASF software suite offered to users was supported with funds from the ASF DAAC. The majority of tools contained in the software can handle data from all four existing radar satellites, and the DAAC is committed to keep the software updated. This suite of programs was developed by the ASF science team. In addition, a SAR correlator and interferometry suite of programs was developed in consultation with Howard Zebker, formerly at JPL and now at Stanford University. The roots of the Alaskan software appear to lie in Zebker's version of the JPL Fortran code (circa 1996), which was not widely available to the scientific community outside JPL. The ASF package appears to be a major improvement because it is written in the more flexible C language and is cleaner and more portable. Indeed, the "processor" part of the software runs on the Cray T3D at the University of Alaska.

Several specialized software modules are restricted, however. As noted on the ASF Web page for interferometry software (http://www.images.alaska.edu/index.html), "the programs which make up the interferometry package are at a restricted ftp site. Restricted programs are denoted by (*R) following the name in the Description section. For information on becoming a registered user in order to access restricted programs, see the ASF Software Agreement.... Note: This concern may be relaxed by ASF upon completion of software classification by the U.S. State Department. NASA HQ has requested that we limit international distribution until this can be clarified." In view of the scientific value of the SAR processing software, the panel suggests that the ASF and NASA explore ways to improve access to scientific software.

Some concerns about the services provided by the ASF DAAC to U.S. users stem fundamentally from the lack of a U.S. national policy or program for civilian SAR. The United States operates no satellite-borne SAR, and neither NASA nor the NSF seems willing to purchase data from the three foreign flight agencies that operate civilian radar satellites in volumes that would foster major scientific advances and at the prices quoted by these agencies. User demand far exceeds the amount of data available under the terms of existing MOUs. Unless NASA negotiates more favorable terms for U.S. researchers in future MOUs or purchases SAR data from foreign space agencies, the ASF DAAC will not be able to satisfy its user community any better in the future. For example, the panel heard numerous anecdotes that the Canadian Space Agency has placed severe restrictions on the amount of data U.S. researchers can obtain affordably—a situation that was not anticipated by many U.S. researchers. This unfortunate circumstance is compounded by the poor performance of the SAR processor for RADARSAT data (see "Processing Software," below).

Most of the ASF's production hardware and software is selected, developed, and installed by JPL. At the present time, it appears that JPL has the lead role in developing an agenda for facility development and that the ASF decides on the priorities within the resulting list. In talking to the ASF production staff, however, it became clear that the staff have the responsibility to make this facility work, but not the authority to ensure that it is working as effectively as possible. This issue is discussed in more detail below (see '' Relation to the JPL Developers'').

The production facilities of the ASF consist of four components:

1. satellite data acquisition facilities,

2. data processing facilities,

3. data distribution facilities, and

4. data archiving facilities.

In evaluating the DAAC's responsibilities within the production facilities, the panel tried to ignore the first component because data acquisition is the role of the Alaska SAR Facility, not the ASF DAAC. Yet the line between data acquisition and processing and dissemination is somewhat artificial and can be drawn in different places according to what criteria are used (e.g., budget, personnel, hardware, data product level). Still, data acquisition appears to be the strongest part of the facility. Indeed, the panel was impressed with the ASF's success in tracking satellites at the 97–98% level.

With the exception of the satellite antennas, the four components are all housed in a single, cramped computer room. Although this is not necessarily a problem, it does mean there is little room for expansion or for the installation of alternative equipment during a transition phase. The equipment being used, which was selected or built by the JPL developers, is also quite heterogeneous. Although this too is not necessarily a problem, it does raise concerns about the maintenance costs associated with the disparate pieces of software across the various platforms. Finally, there exists a long-standing problem associated with the obsolescence of certain critical components, such as high-performance tape

drives no longer supported by the original manufacturers, which can be maintained only by using custom-made replacement components.

One item of particular concern is the hardware processor, denoted the Alaska SAR Processor (ASP), which is a custom-built, special-purpose processor devoted to SAR data processing. As this equipment ages, the costs associated with the maintenance of both its software and its hardware will inevitably rise. In addition, the electronic components will not be replaceable before long.

Figure 6.2 illustrates the flow of data through the ASF processing system. Currently, raw signal data from the archive is converted to Level 1 products by the hardware and software processors. Neither processor can ingest data in Level 0 format. The hardware processor cannot easily be made year 2000 compliant, so it will be decommissioned on December 31, 1999. The software processor, on the other hand, will be modified to ingest both Level 0 and raw signal data.

At the time of this review, the archival data storage consisted of open tape racks that were accessed manually. The ECS contractor has since installed a StorageTek mass storage silo and the software to operate it, but the equipment, which was procured early because of a budget opportunity at ESDIS, will remain idle until the JPL developers procure a Level 0 processor and integrate it into the system, probably by the end of 1999. Consequently, the open tape racks will continue to serve as a data archive for at least the next several years. A prototype Level 0 processor has been procured, and the ASF DAAC, JPL developers, and ESDIS are currently determining how many will be necessary to attain a parallel processing capability that exceeds the raw signal data rate. The goal is to simultaneously carry out Level 0 processing of the raw signal data and complete migration of existing data from the open tape racks in the year 2000.

The uncertain health of the primary radar processor at the ASF and the current processing structure that employs one executable code for processing RADARSAT, ERS, and JERS data require serious attention by combined efforts in Alaska and at JPL. Specific improvements would be to provide 16-bit data and to process whole strips instead of single scenes for users requiring regional coverage.

Accessibility of the ASF DAAC's holdings is poor. The DAAC has no plans for distributing data on CD-ROM as the ESA now does, and the Web pages are largely informational instead of permitting convenient data access. Users should be able to browse images, download data sets and software, and order data and publications via the Web. Some of these features are available through the Version 0 Information Management System, but the interface is poor compared with

FIGURE 6.2. Schematic illustration of the flow of data through the ASF processing system. SOURCE: ASF DAAC.

modern Web technology. In addition, browse capabilities would enable timely access to SAR data, thereby allowing their use in near-real-time applications, such as sea-ice forecasting and volcano monitoring. The User Working Group has been unable to get the DAAC management to devote people and resources to the development of a fully functional Web interface, and the panel agrees that greater use of the Web would help build the DAAC's user community.

One problem affecting the ASF's ability to communicate with its user community is the relatively limited bandwidth between the SAR facility and the rest of the world. All Internet connections are limited to T1 speeds (approximately 1.5 mbps), which limits the amount of data that can be conveniently exchanged. There are however, plans underway to install fiber-optic connections between

Alaska and the lower 48 states, possibly in 1998. This would provide for much higher bandwidth and thus reduce the current relative isolation of the SAR facility.

This review took place during a time of transition, which is not yet complete. Syun-Ichi Akasofu, the director of the Geophysical Institute, has formed a faculty committee to evaluate proposals to reorganize the Alaska SAR Facility. The proposals have been evaluated, and a search committee has been formed to hire a permanent director, who will also serve as DAAC manager and chief scientist. The reorganization is scheduled to be completed in the fall of 1998. In the meantime, Craig Lingle, a faculty member at the University of Alaska, serves as acting director. The need to reorganize was apparently driven by dissatisfaction among many users, strain within the DAAC management team, and persistent tensions in the relationship between the ASF DAAC and the JPL developers.

The Alaska SAR Facility has two components: (1) a satellite-receiving ground station, which is responsible for data acquisition and antenna operation, and (2) the DAAC, which is responsible for data processing, distribution, and archiving. The DAAC component accounts for nearly 80% of the budget. The receiving ground station is funded by NASA via the Wallops Flight Facility. The DAAC is funded by NASA via the Goddard Space Flight Center. In practice, the Alaska SAR Facility and ASF DAAC operate as a seamless organization from the point of view of budget and personnel.

At the time of this review, the activities of the Alaska SAR Facility were carried out by nine divisions, including data management, data systems, engineering and maintenance, management, operations, planning, science, systems coordination and development, and user services. (At present, the organization has seven divisions.) In the panel's view, the large number of divisions presents a potential management problem by making it difficult to oversee the end-to-end operations of the DAAC. The panel also observed that the functional partitions between the divisions are firm and the boundaries between divisions are quite impervious, which inhibits sharing of technical resources and collegial or cooperative approaches to problem solving.

At the time of the review, an acting DAAC manager had just been named, and it was difficult for the panel to evaluate personnel issues. Nevertheless, it was clear that the main tensions were between ASF DAAC staff and JPL developers, at both the management and the working levels (see "Relation Between ASF

DAAC and JPL," below). Responsibility and authority are split between the ASF DAAC and the JPL developers, exacerbating the normal tensions between operations and development.

The DAAC's relationship with its ECS liaison, on the other hand, seems more positive. The ECS science liaison is treated as an ASF employee, rather than as an outside contractor. Consequently, she may become an effective bridge between the ASF DAAC and the ECS contractor if and when the ECS is delivered.

The ASF DAAC's total budget is $13.4 million in FY 1998 (Table 6.2). Development by the ECS contractor (8% of the budget in FY 1998) and the JPL developer (40% of the budget in FY 1998) accounts for approximately half of the budget. The ECS costs, which are largely associated with hardware and software acquisition and maintenance, will become insignificant by FY 2000. JPL development costs likewise decline from 68 to 36% of the DAAC's budget over the nine-year period shown. It should be noted that some of the JPL development efforts are directed toward the ground receiving station, which is not part of the DAAC. Therefore the costs of the JPL component shown below are maximum values. Overall, the DAAC's total budget will decline to $9.8 million by FY 2002, the latest year for which projections were provided to the panel.

The only measure of cost-effectiveness presented to the panel is the average cost of a Level 1 data product delivered to a user. It is calculated as the ASF

budget averaged over some time period, divided by the number of images produced and distributed during the same time. This metric is misleading because the ratio of Level 1 products ordered to Level 0 products archived is quite low (less than 10%). Nevertheless, the cost per Level 1 product, measured this way, has dropped substantially due to recent major increases in processing volume, which led to a large increase in the number of images flowing through the system. The increases in processing volume were related to the debugging of RADARSAT processing code, system changes and upgrades, and management decisions on the RADARSAT processing workload and user priorities. As SAR imagery becomes more popular, these data will become increasingly important (and therefore valuable) over time.

In absolute terms, however, the budget of the DAAC is large compared with the small number of users (about 400). A high priority of the DAAC should therefore be to increase the size of its user community, thereby increasing its cost-effectiveness.

Quantitative metrics are a useful tool for measuring and monitoring the performance of a DAAC. Indeed, without them, it would be difficult for a DAAC to show that it has successfully fulfilled its goals. The ASF DAAC appeared to possess few metrics. At a minimum, metrics should be developed for determining how much data are acquired, how much are distributed, and how much are actually lost. (The panel heard informal stories about data losses that could not be documented or assessed.)

Another concern of the panel is that there appears to be no strategic plan for the evolution of the facility. The JPL developers sees themselves as providing "sustaining engineering and technology insertion." In other words, they do not see future planning as one of their responsibilities. One key issue is a replacement for the ASP, which, as noted earlier, will become increasingly expensive to maintain as time goes on. Some experiments by the ASF Science Division using the Arctic Region Supercomputing Center's Cray T3E are quite promising in this regard as an alternative. The JPL developer is also considering a "software architecture" for future developments. However, no one appears to be coordinating these parallel efforts.

The ASF DAAC is considered to be an important part of the Geophysical Institute. The director, Akasofu, is quite aware of DAAC activities and instituted a reorganization and personnel changes to improve its operations. Akasofu is also

seeking to improve the relationship between the ASF DAAC and the JPL developers.

Surprisingly, despite its importance to the Geophysical Institute, the DAAC has been allocated very little space for hardware. Space is tight, and new equipment such as the StorageTek archive will have to be located off-site. Unless the University of Alaska and the Geophysical Institute increase their commitment to the DAAC and allocate sufficient space for its operations, NASA may have no choice but to relocate the DAAC to more spacious quarters.

Several scientists from the Geophysical Institute attended the review, but the chief scientist and the acting ASF director are the only scientists on the DAAC staff. The DAAC's main interaction with scientists at the university is through the science division—about six scientists at the Geophysical Institute contribute regularly to the DAAC. On the other hand, the DAAC has acted as a catalyst for remote sensing research at the university, particularly the Geophysical Institute. The panel hopes that this productive interaction will continue in the future.

Paul Ondrus at ESDIS is responsible for the ASF DAAC and the JPL development contract. During an interview with the panel, he confirmed that JPL has the technical lead for the ASF DAAC and that JPL considers him to be the ASF DAAC manager. Before his arrival, several other people at ESDIS were responsible for the ASF DAAC. DAAC personnel have opined that the high turnover rate at ESDIS has led to a lack of continuity in long-term planning for the DAAC and may have adversely effected the DAAC's budget.

At a higher level, the DAAC feels somewhat ignored by NASA because it holds data from foreign spacecraft, rather than from NASA instruments or field experiments. There is, for example, no NASA project office that deals with SAR data.

To be most useful, the data sets of the ASF DAAC should be combined with other types of data. Many of these complementary data will be provided by the other DAACs. For example, scientists using SAR data for interferometry will frequently request digital elevation models (DEMs) from the EDC DAAC, and scientists interested in improving the DEMs may want to use SAR data from the ASF DAAC. Similar examples abound in other disciplines (polar science, oceanography), and linkages with other DAACs (e.g., NSIDC DAAC, PO.DAAC) should be formalized to ensure that scientists can obtain the data they need. At present, the integration of other EOS data sets with SAR data held by the ASF DAAC is possible for sophisticated users, but the convenience of such a task is far from the standards of EOSDIS. As noted above, the DAAC is not on the

delivery schedule for the ECS, and there are no plans to make the JPL and ECS systems compatible.

Delays in the ECS have caused the ECS contractor to focus almost exclusively on those DAACs that will be receiving data from the AM-1 platform. The other DAACs, including the ASF DAAC, will not receive the ECS in the near future, if at all. When the ASF DAAC was on the ECS delivery schedule, DAAC personnel attended ECS planning meetings and participated in many phases of ECS planning and development. Funding for these activities was provided by ESDIS, over and above the DAAC budget. These activities have now ceased. In the panel's view, the lack of coordination between the ASF DAAC and the ECS will lead to increasing isolation of the SAR user community from the more integrated EOS community. It also affects improvements needed at the ASF DAAC, such as software links to the broader community.

Because the priorities of the ECS contractor have shifted to the AM-1 DAACs, ESDIS asked JPL to build an "interim solution" for processing data and to keep the system going until the ECS arrives. The JPL system, however, is incompatible with the ECS, and JPL developers argue that this is because they were neither required nor funded to make the two systems compatible or to make the ASF DAAC interoperable with the other DAACs. Moreover, ESDIS has apparently directed JPL away from exploring interoperability issues with the ECS contractors until after the ECS is delivered to the AM-1 DAACs.

ESDIS funds the ASF DAAC and the JPL developers under separate contracts. A five-year enterprise plan outlines the roles and responsibilities of each component. The JPL developers function as the instrument team for the DAAC because they provide Product Generation Executables (PGEs), a software element that accepts low-level data products and ancillary information and outputs a higher-level data product. They are also the equivalent of the ECS for the ASF DAAC.

The relationship between ASF and the JPL developers has apparently been strained for some time. From the DAAC's point of view, JPL's focus has been on optimizing processors, not on building an integrated system, as wished by the DAAC. Other DAAC requests for fixes are similarly ignored by JPL. From the

JPL developers' point of view, however, the DAAC asks for more than JPL is budgeted to provide—JPL considers the requests, but it can only act on the highest-priority items.

Panel visits to the JPL developers and Ondrus subsequently clarified the decision process: Ondrus sets the priorities for the JPL developers. The JPL developers have no plans to understand better the needs of the ASF DAAC by detailing staff to the DAAC for short periods of time, for example. The ASF DAAC hopes that the new five-year plan developed with ESDIS will alleviate the tension between the DAAC and JPL. The plan calls for the DAAC to establish operational needs and requirements, and for the JPL developers to meet these needs. In view of the poor performance of the SAR processor to date, NASA should evaluate the option of using alternate developers or fund the ASF DAAC to do system development in-house.

The panel views this structure as one that perpetuates a classical management problem arising when authority is not vested in the same persons or organizations that are burdened with the responsibility. In this instance, since the ultimate mission of the DAAC is to satisfy its users, this point of view is inevitably that the DAAC should be given the appropriate authority to discharge this responsibility.

Synthetic aperture radar is a spectacular technique for studying the Earth. At present (except for a few airborne and shuttle-based experiments of limited duration), it is collected exclusively by foreign space agencies, which, by agreement with NASA, allow limited amounts of SAR data to be acquired and distributed to NASA-approved investigators at below-commercial prices. The lack of a national SAR policy, however, has led to problems in data acquisition. Nevertheless, as the primary source of affordable SAR data to U.S. researchers, the ASF DAAC provides an important resource. The DAAC's role in EOSDIS is almost equally important to researchers because SAR data are most useful when they are integrated with other types of scientific data. By serving as part of an EOSDIS system, rather than as an independent data center, the ASF DAAC has the potential to foster the type of multidisciplinary research for which EOSDIS was designed.

As illustrated in its excellent user survey, the ASF DAAC has had mixed success in meeting the needs of its scientific user community. Scientists who need data from the Alaska station mask tend to have good interactions with the DAAC and to be able to obtain the data they need. These scientists, many of whom work at the Geophysical Institute, concentrate largely on ice motion studies. Some of these studies require access to near-real-time data, and the ability of polar scientists to obtain them bodes well for scientists seeking to monitor volcanoes. Other studies, such as those supported by the Antarctic Mapping Mission, require a large amount of processing time. The impressive Quick Look product of the rim of Antarctica demonstrates the DAAC's ability to "drink from the fire hose," a challenge that the other DAACs will not face until the launch of the EOS satellites.

On the other hand, scientists who need data from outside the Alaska station mask are commonly frustrated, mostly because NASA has not concluded the necessary data acquisition agreements with foreign ground stations. Nor is it willing to fund U.S. researchers to purchase the data from foreign agencies at market prices. However, even if data can be acquired through the ASF, the DAAC has no tracking and feedback system for monitoring user requests. Consequently, users have no choice but to call the DAAC repeatedly to learn the status of their data acquisition and/or processing request. By implementing a tracking and feedback system, the DAAC will improve its relation with out-of-mask users, better serve the needs of the in-mask user community, and increase the size of its user community overall.

Another immediate challenge for the DAAC is to improve its relationship with ESDIS and the JPL developers. Currently, responsibility and authority are distributed among three organizations: (1) the ASF DAAC is responsible for operations and for satisfying user needs; (2) the JPL developers are responsible for creating the data acquisition and processing systems; and (3) ESDIS has authority for setting development priorities and overall budgets. This separation of authority from responsibility hinders the ASF DAAC from fulfilling its mission of serving the polar and earth science communities.