Land AVHRR Data Users Group Workshop -
Towards Product Standards

David L B Jupp
CSIRO Earth Observation Centre


1. ABOUT THE CSIRO EOC


AVHRR has provided the world community with global data and a dense time series of earth observation now for at least 15 years. However, it is only relatively recently that the data have moved out of the research area to be evaluated in the 'workplace' of end user needs. This has come about in Australia from the development of operational receiving stations and programs of archiving and processing the data received over the continent and EEZ.

CSIRO Office of Space Science and Applications (COSSA) and Agencies such as the Bureau of Meteorology have played a significant role in this development and today the availability of daily near real time and archival time series of data for the Australian region is unprecedented. Nevertheless, the development of this capability has not been without its problems. Many of these were described in a review carried out on behalf of COSSA of its Data Acquisition and Utilisation multi-Divisional Program MDP-19 . In this report (often referred to as the Simpson Report from its principal author) many of the issues of fragmentation and dispersion of effort were outlined and remedies suggested.

One of the outcomes of the Simpson Report has been the formation of the Earth Observation Centre (EOC) which is developing to provide a focus and identity for CSIRO Earth Observation Research and Development as well as act to coordinate some of the activities that were identified in the Simpson Report as limiting to the full value being obtained from CSIRO activities. In the area of Land AVHRR use there is considerable focus and activity among CSIRO scientists who either as part of the formative EOC or from the Divisions with which it is working closely will outline their work to you this morning. They are among scientists who have been and will in the future engage in cross-Divisional projects in the fields of:

The knowledge base for the projects largely rests in the Divisions and the work will be aimed to support the Divisional needs in data and the Divisional interface with clients and users in the Sectors which the Divisions address. However, we see an underlying generic set of activities which are bigger in scope than can be handled by any one Division and which underpin the eventual acceptance and usefulness of the products to the real end users - people who want information about the world and the way it is changing.

The EOC is an initiative that has been developing for some time and is now fully and rapidly moving to address generic issues in the areas outlined above. It aims to maximise the benefit obtained from the activities of the excellent earth observation scientists in CSIRO and see it impact effectively on the development of the industries that can put the information to work. In order to do this fully, however, I believe we must all (researchers, producers and users together) consider the issue of data and product standards.

 

2. THE ISSUE OF STANDARDS IN AVHRR DATA USE


From the scientific viewpoint, there has been a significant progression from the type of research paper presented in the inaugural ARSC1 in 1979 to the present ARSC8 in 1996. The mass centre of such presentations has moved from descriptive work through exploratory research to a point where advanced measurement based models are being effectively used with remotely sensed data and widely reported. Remote Sensing is truly becoming a measurement based activity. Among the international scientific community this progress is well in train as you will hear from a number of the scientists speaking today.

However, at the user end, is there evidence of as much progress? If there is, then it would be seen in a progression from the research question of "What might AVHRR do for me?" to the operational issue of "What does AVHRR do for me?". This last question involves the development of well defined products and standards as well as well established methods for creating products of value to specific end users. This is not so much an issue for research as for specification.

Nevertheless, researchers cannot ignore the issue of standards. Standards are needed if the AVHRR user industry is to leave R&D 'school' and join the much more engineering oriented 'workplace'. Engineering based standards define what levels of processed data product are needed and what methods of value added processing ensure a product with a guaranteed accuracy or utility can be delivered. Such standard specifications are needed to protect data and product suppliers and users. They are also essential tools with which end users can cost the use of remotely sensed data and value adding into major projects. I claim that the lack of these is the biggest hurdle to acceptance of our efforts at this time and more vital to us than the need to do more advanced research.
Ideally, standard data products, data processing and interpretation methods should be:

The last of these is desirable rather than essential.

The standard procedures with such characteristics that are needed in our industry are also of two basic kinds. The first is for data delivery such as from an operational receiving station. These are typically set in a 'Levels' framework and for each level there are a set of specifications and measures of performance. Perhaps these are not as well developed and consistently deliverable in Australia as they should be. The second are for value added products. These may need manuals of standard practice which, if followed, provide products of a proven standard. Of course, the standard may seem quite low to many scientists but I believe the current lack of them is causing problems for researchers and users alike.

3. STANDARD PRACTICE AND RESEARCH EXCELLENCE


If there are well defined standards, it places research and development into context and protects it from being confused with operational production. Research is needed in cases where the standard processing is inadequate, where it does not solve the user's problems and also needed to create new products which may eventually become standard processes. The important advance for researchers in this situation is that when users ask researchers to try new methods because the standard ones are inadequate it is they, and not the researchers, who take on the risk that the result may not be any better.

Having a solid base of established standards also puts into context the difficult issue of "Best Practice". I think the name is the problem! It is always better to use "Best Current Practice" and even better just to say "Standard Practice". I think that 'best practice' is actually what the 'best people' do and always will be. It is impossible to specify the 'best' or create it by documentation and imposition of measures of success. What it is possible to specify is the currently acceptable standards and it is in everyone's interest that these standards exist and provide a measure against which new methods can prove their worth.

The area where research lives is in the quest for the best but it may not be that current research really is better. The measure of 'better' to the user lies as much in repeatability and consistency as it does in the incorporation of more 'realistic' components or complex physics. However, the researcher needs to have room to explore and can move the horisons to create what may be tommorrow's standards.

An example of such a slow but sure process can be seen in evaporation methods as practiced in Irrigation Engineering. In evaporation standard practice the scientifically superior Penman-Monteith method has taken many years (more than 30) to become included in the standard manuals. It has had to be proven in operational practice to be simply applicable, reliable, consistent and even conservative so that it could replace tried although maybe not very true methods in use for many years in operational calculations of "Consumptive" Water Use.

4. BRINGING PRODUCERS AND USERS TOGETHER


In order to bring about these standards, it is important that Australia work closely with any moves or existing progress in these directions in the international community. However, there is a need for Australian users and data or value added product producers to get together and define what the needs of Australian users are - whatever the progress or lack of progress internationally. In the case of Land AVHRR Data I think the need is critical. This Workshop is an ideal place to start the process as there are researchers, data providers, value adders and some end users all meeting together to discuss the issues facing the 'industry'. Let's not waste the opportunity.

In any transaction of the kind we are discussing, involving a producer and a user, there are two risks that must be minimised or at least balanced if the transaction is to be fair and even for it to keep occurring. These are the "Producer's Risk" and the "User's Risk". They assume there is a measure of quality available for the product and a strategy of testing. In this case, the producers risk is the probability that a perfectly good product will be rejected as poor and the consumer's risk is the probability that a product called acceptable is poor. Clearly, standards, measures and strategies of testing must all be developed by users and producers together to create a situation everyone can live with. In practice, the financial viability of a product to the producers may be an important component in the definition of a standard.

In our area of AVHRR data and products there are two forums for standards. These are the interface between data producers and data users (or value adders) and the interface between value adders and the end users. In the end, the two will form part of the same 'sentence' and need to be compatible:


"With data produced at certified Level x, using processing methods outlined in manual y, with staff and infrastructure of capability z it is possible to produce specified product p with accuracy q."


Perhaps this Workshop, with its useful blend of researchers, data managers, data suppliers, value adders and users, can outline the cases where Australian users can fill in the gaps in sentences like this one and help define what Level x is for the benefit of the data producers. Perhaps there is no existing sentence of this type of which we can be confident. In either case, what are the base standards, what can reliably and consistently be achieved and what measures can we put into place to ensure in the future (to the benefit of producers and users alike) that they are being met? I claim that although the answers to these questions do not always involve scientific issues of interest to leading edge researchers it is in their interest and essential for their long term survival that some answers be found.




Copyright CSIRO 1999 ©

Use of this web site and information available from it is subject to our
Legal Notice and Disclaimer