Dynamic Earth Observation Systems: a potential EU/US collaboration
Giovanni Aloisio August 1999

Earth Observation Systems: from the static to a dynamic approach

An Earth Observation System is made of a sensor mounted on a satellite, an airplane or a space shuttle that gathers images of the surface of the Earth and of a "Computational Grid" (a pool of heterogeneous distributed machines needed to perform the image pre-processing or post-processing and a distributed database in which these images are stored). Moreover, a suitable web interface must be provided to allow the user to choose from a web browser a geographic area of interest, to specify some form of processing on it and and to get the final result. In order to transform the raw data gathered by the sensor in an real image, a certain amount of processing is required. This processing is called pre-processing to distinguish it from the post-processing needed to extract some information from the pre-processed images. It should be noted that while for a given sensor there's just one kind of pre-processing, different post-processing algorithms can be used to produce output that is always an image, or to extract some quantitative information (returning a graph or just a number), with a different impact on the network band needed! Earth Observation Systems (EOSs) are supported by the EU/US National Space Agencies and provided over the years a tremendous amount of images of the Earth's surface. These images are a potential source of information for lots of users working in various fields, like archaeology, geology, maps drawing, ecology and others. To access this precious source of knowledge, "static" approaches are usually provided and nice user interfaces have been developed to extract the requested information from a distributed database. We define "static" the current EOS systems because they allow a user to request only pre-processed images. Static systems lack the intelligence that's needed to start automatically a pre-processing in case the requested image is in the raw format or to start a specific post-processing algorithm on the selected image. On the contrary, "Dynamic" EOSs (DEOSs) should allow a user not only to choose from a web browser a geographic area of interest (as in the static systems), and to specify the information to be extracted from the selected images but they should also be able to make autonomous decisions starting from requests formulated at high level by the user (like "show me the corn fields in the surroundings of Lecce"). These high level requests could entail different user transparent decision strategies, sophisticated resource management techniques and the use of intelligent software in order to deliver the requested information at the lowest cost and in the least time!

Then the system should be now able to solve a variety of problems:

1. Search of the image: the system has to look for an image that suits the user's needs, and the requested information must be delivered at the lowest cost and in the least time. That means that suitable metadata, containing a variety of information on the data, the geographic location, the format (raw data, pre-processed, post-processed in some way) and the position in the distributed database, must be defined to solve the issues related to an intelligent resource management. 2. On-demand processing: the system must be able to start a processing when "explicitly" or "implicitly" required by the user. An explicit processing request is done when an image in the raw format is selected and a pre-processing is needed. An implicit processing must be activated when the information required needs to be extracted by some form of processing on the selected image. 3. Resource management: dynamic information concerning the performance of both the machines and the network must be managed. The system has to select the proper strategy to let the program and the image meet (i.e. should the processing be transferred on the machine with the data or is the opposite more convenient?) to allow the user have what she wants at the lowest cost and in the least time! 4. Authentication and Security: the system must recognize the user who's trying to access the archive from the web. A mechanism for passing authentication to the other machines of the Computational Grid is also needed. 5. Remote control: another issue is the remote control of interactive programs. If the processing requires an user interaction, a suitable interface with the remote process must be provided. 6. New forms of Electronic Trade: The images of the earth are not for free and the computation time on the computers belonging to the Computational Grid isn't for free either. The system should tell the user in advance the cost of what she asked for. In the case of a static system this has a quite straightforward solution: the price of the image can be stored as a field in the related metadata. When the system is dynamic, instead, the processing time must be evaluated according to the requests made by the user. This is an open problem. The system should be able to provide a user with a choice of this kind: "you can have what you requested in this time and at this cost or in a lower time at a higher cost". 7. QoS: As we're trying to allow potentially a lot of users to access the power of a computational grid by a web browser, the Quality of Service will be also an important issue.

The workshop should address all these technological issues related to the creation of a hardware/software infrastructure (Computational Grid) for a Dynamic EOS. Moreover, questions related to how these DEOSs can be maintained and can evolve should be answered. Clearly, if a federated DEOS can be thought (for ex. a joint effort of different National Space Agencies for the creation of a SAR DEOS!), other issues must be addressed (such as, what standard for the metadata should be adopted, how the Space Agencies will cooperate to build the federated active digital library,Š..).

The workshop should also try to answer the following questions:

… Can a Dynamic Earth Observation systems be developed by a joint collaboration between EU/US?

… What problems (political! not technological) need to be solved to achieve this goal?

… Why don't we try to build an EU/US Computational Grid and test it on a global Dynamic EOS?

… Are the National Agencies (NASA, ESA, DLR, ASIŠ) prepared for this global effort?

The results of the NSF Workshop on "Interfaces to Scientific Data Archives" produced and hosted by the Center for Advanced Computing Research, Pasadena-California, 1998 March 25-27, (http://www.cacr.caltech.edu/isda), should be also considered in the context of Dynamic Earth Observation Systems. Here are reported some statements I extracted from the NSF Workshop on "Interfaces to Scientific Data Archives" useful to start a possible discussion:

"The Vice-President of USA has endorsed the interoperation of geospatial and other scientific data archives as a national goal in the address "The Digital Earth""

"There is an urgent need for software infrastructures to create, maintain, evolve and federate active digital libraries of scientific data: infrastructures that consider the newcomer learning to use the system as well as the seasoned professional"

"Federation of archives grows in synergy with the creation of metadata standards"

"The eventual aim is not simply an interface to a single data archive, but to federate such archives such that what is learned at one place works at another, and so that data can be effectively combined from multiple, independent archives"

"Metadata was an important topic: how to describe data objects, to make catalogs, to form relationships between data objects"

"If metadata is to be useful to other libraries, to search and discovery services, or for federation of archives, then it must be a consensus about the semantics, the structure, and the syntax of the metadata"

"The dublin Core metadata standard is an effective and viable way to provide the semantics and structure of these metadata records. What is needed in addition is the development of languages needed to exchange metadata, schemas and ontologies. For each of these levels of abstraction, a definition language and a manipulation language are needed"

"Text-based interfaced should not be rejected in favor of point-and click"

"Authentication and security are important as soon as the archive moves beyond the prototype"

"When users can produce derivative data products and put these back into the archive, then subsequent users need to know and trust the people or organizations that did the processing" 

Web Links

ISDA: http://www.cacr.caltech.edu/isda/

DIGITAL PUGLIA: http://www.cacr.caltech.edu/SDA/digital_puglia.html

EOSDIS: http://spsosun.gsfc.nasa.gov/New_EOSDIS.html

ISIS: http://isis.dlr.de/

CEOS IDN: http://gcmd.gsfc.nasa.gov/ceosidn/

ESA ATSR: http://192.111.33.173/ATSRNRT/