Australia is a big user of spatial digital data. This has stemmed from the large size of the continent, its low average population density and the technological sophistication of its economy.
However, any careful analysis of the pattern of usage of these datasets will reveal that most have been, and still are, concentrated in large government agencies, research institutions and large corporations. There has been very little trickle down' to smaller users. Farmers, landcare groups, bushfire brigades, local shire councils, catchment management committees and general consumers have been largely unable or unwilling to use them.
There are several reasons for this. The tyranny of distance plays a part. Users need to travel to the offices of data providers so they can physically browse through product catalogues. For users in the major metropolitan centres this is inconvenient. For would-be regional and remote users it's a major disincentive.
The advent of Internet-based browsing facilities has alleviated some of this. However, the user can only view a degraded product with such facilities, and cannot be assured of the final quality.
A second problem concerns timeliness. Conventional delivery mechanisms require people in the data provider agency to extract the data ordered by a user from the corporate data archive and write it on appropriate media. This is then generally dispatched by mail.
For data which is only slowly changing with time such as cadastral information or geological maps, a delivery mechanism which takes several days may be acceptable to a user. For data with a short use-by date, such as thermal images of bushfires, rapid delivery is mandatory for all except retrospective studies. Indeed, most remote sensing studies of dynamic events such as bushfires and floods have to be retrospective for just this reason.
Another problem concerns the need for user software and training. It is probably self-evident that digital spatial data cannot be used without appropriate hardware and software, and the knowledge of how to use both.
The hardware problem has been serious in the past because the kinds of computers and graphics cards needed to work with spatial data were extremely expensive. This has been solved by the advent of cheap PCs, but the software issue remains serious.
Thus although most computer users today are familiar with windowing environments and graphical user interfaces, the level of sophistication in software that is needed to handle and process digital spatial data is generally well above that of the average word processor or Internet browser. This has ramifications both for the cost and the ease of use of the software required by the fledgling user of digital spatial data.
Price is an issue. Spatial data is still prohibitively expensive, and generally beyond the reach of small users. The key problem here is the cost of manually processing orders. These costs are so high that data must normally be sold in large chunks to recover them.
Users also have to navigate licence agreements and conditions of use. As the majority of data providers are government organisations, these agreements vary, not only across levels of government but also within. For small businesses there is the added complication of the huge variation in redistribution policies.
With these problems in mind, a consortium was formed in 1997 to develop a prototype of a digital spatial data delivery system, called Australian Earth Data On-Line. The consortium consisted of the Australian National University and the ACT Emergency Services Bureau.
The development of AEDOL and its commercialisation into Earthinsite.com Pty Ltd has been possible thanks to a number of grants. The program was the recipient of a COMET grant (Commercialisation of Emerging Technologies). It also received direct grants from the Information Technology Online program of the National Office for the Information Economy, AusIndustry, the ACT Chief Minister's Department, the International Decade of Natural Disaster Reduction and the Australian National University.
A working online prototype system (AEDOL Version 0) was completed in September 1998 (it's still available at http:/ aedol.anu.edu.au), and has undergone development since then. During 2001 it has been going through the transformation into the fully commercial Earthinsite system.
For ease of use, the system does not require any installation routines or plug-ins. It only requires Internet connection and standard browser software such as Netscape or Internet Explorer. To minimise slow and costly Internet traffic, some simple local processing capabilities are provided on the user's computer. To limit the development task and provide state-of-the-art functionality, a commercial back-end is used. This removes the need for ongoing development and maintenance of the main processing engine.
Earthinsite uses a mix of fast local processing on the user's computer and a powerful remote image processing engine based on ER Mapper. Local processing capabilities are provided by cross-platform Java applets. These permit the user to make rapid improvements to the images without the need for Internet traffic.
At the back end of Earthinsite, ER Mapper is used to perform image processing and GIS overlays, without reliance on specialist software or training.
Earthinsite is based on three different programming environments. First, two specialised Java applets provide intelligent processing capabilities on the user's desktop. These applets have been written for maximum compatibility on different hardware (PC, Mac, workstations) and different browsers (Netscape 3 and 4, Internet Explorer 3, 4 and 5).
Second, structured Perl programs provide the middleware that handles multiple transactions of concurrent users and formulates user requests for processing by the Earthinsite back-end.
Finally, the back-end uses custom programming routines to call ER Mapper through the ER Mapper batch engine. These routines give the user the ability to perform image processing on selected images.
Earthinsite delivers jpeg files to the end user. These files are the end result of the processing chosen by the user. Thus the user can zoom in to an area of interest and change it by, for example, modifying the image transform, applying a spatial filter or choosing a different look-up table.
Figure 1: The AEDOL Geoselection Tool is a Java applet that allows a user to select a dataset for processing. It provides visual representations of the footprints of the data available to the user (e.g: the green box in the figure). The range of datasets is controlled by parameters that can be customised by the user, including the spatial extent of interest and the type of data required (such as AVHRR, Landsat TM and so on). Time-series data can be searched using a Set Date function. This facility enables the user to quickly filter out data that is outside the time window of interest.
A future enhancement to the system will allow users to download portions of the original data and the associated ER Mapper processing algorithm. These files could be viewed with ER Viewer, which is free (available at www.ermapper.com.au). Wholesale implementation of this data download capability will, however, require new pricing and licensing arrangements with data providers.
The system includes user registration with layered access to datasets. Since different datasets have widely differing values and public availabilities, it is necessary for any system that hosts a wide range of data to have layered access controls. Thus some datasets are publicly available, while others are only available to certain users.
The Earthinsite system allows each dataset and each dataset type to be independently identified in terms of access. It can be given open access, group access, or restricted to an individual user.
By providing access to small user-customisable areas from within large and expensive datasets, and adopting a per-pixel pricing policy, such a system, when commercialised, can significantly reduce the cost of data to small users. This is illustrated in Table 1.
For small property sizes such as the 1000 hectare example in Table 1, an automated commercial processing system like Earthinsite reduces the cost of digital spatial data to as little as 2% of its current price. In practice, a minimum credit transaction charge would apply, but this would still amount to huge savings for the end user.
The main data providers in Australia all have medium-term plans to develop individual data delivery systems to on-sell their own data. As far as we know, none has plans for a system to host data from other institutions or allow the combination of these different data types. Plans for the development of on-demand data processing capabilities have also been fairly limited.
The development of a national system which is a common entry point to different data provider holdings is preferable, both from the user point of view (only one system needs to be learned) and from the perspective of the data providers (no unnecessary duplication of system development need occur).
Rather than being competitive, Earthinsite is expected to be complementary to current commercial operators. It will stimulate the industry. Increasing demand for geo-data services creates a win-win situation for all stakeholders.
A business plan and market research study are currently underway, as an aid to obtaining funding for the commercial development of Earthinsite.
Further details can be obtained from Earthinsite's general manager John Payne john.payne@earthinsite.com on +61-2-6125-3224, or visit http://earthinsite.com.
