Progress and Grand Challenges of Marine GIS
Grand Challenge: Data Access and Exchange
A recent report assessing the geospatial data needs of the Integrated Ocean Observing System (IOOS; Hankin et al., 2003), estimated that the annual flow of oceanographic data collected to support assessing, modeling and monitoring coasts and oceans will exceed ~2.9 terabytes per year. As such, we are faced with new challenges involving the synthesis, visualization and analysis of these disparate data types to maximize the utility of past, present and future marine data collection efforts. To meet these challenges, the marine science and management community will need to implement common data standards and protocols promises to allow for more efficient data sharing, higher quality analysis, and more direct linkage of spatial and temporal events in the marine system.
There are also different user communities that will need to collaborate more closely in the future. The research oceanography and bio-informatics communities are making advances in large information systems programs but tend to use mathematical scripting languages (e.g., MATLAB, Generic Mapping Tools, etc.) to process spatial and temporal data. The "end user" marine management and conservation communities tend to use desktop commercial GIS packages. In order to bridge the gaps between these communities, efforts need to be made to develop more appropriate and interoperable software and data models for marine applications (e.g., Wright et al., 1998; Goldsmith, 2000).
As these varying communities interact, there will be a continuing need to formalize concepts and terms (i.e., ontologies) that will be used to aid the user in more effective searching and analysis of data and information (e.g., McGuinness, 2002) . For example, in the search for data and resources, one may use interoperable terms such as coastline vs. shoreline, seafloor vs. seabed, ecological resilience vs. robustness, scale vs. resolution, wetland buffering vs. GIS buffering, etc. Here the development of ontology repositories for marine data will be important, along with "semantic integration and interoperability" (e.g., Goodchild et al., 1999; Egenhofer, 2002; Kuhn, 2003), to aid in fully describing the context in which data were collected for its proper use, or for appropriate legacy uses beyond the initial mission or target of the data collection. At the very heart of this, and providing working examples, guides, cookbooks, and tools, is the new, international Marine Metadata Interoperability initiative (www.marinemetadata.org).
Grand Challenge: Representation of Marine Data and Common Data Models
One of the most powerful features of a GIS is the ability to combine data of various types simply by assigning coordinates and displaying these "layers" together. Of course, this representation runs into difficulty if the data are dynamic, with constant changes in location or attribute, and best viewed that way, when the data represent entities of different scales, or when its dimensionality is three, four, or greater. Marine applications, with tides (and hence shifting shorelines), upwellings, ships and vehicles moved by waves and currents, shorelines, and the like demonstrate all of these difficulties.
Indeed, much has been written about the importance of error and uncertainty in geographic analysis (e.g., from Chrisman, 1982 to Heuvelink, 1998), and with the challenge of gathering data in the dynamic marine environment from platforms that are constantly in motion in all directions (roll, pitch, yaw, heave), or in tracking fish, mammals, and birds at sea, the issue of uncertainty in position is certainly critical. We must accept that no representation in two-, three-, or four dimensions can be complete.
Data models lie at the very heart of GIS, as they determine the ways in which real-world phenomena may best be represented in digital form. A data model for marine applications must undoubtedly be complex as modern marine data sets are generated by an extremely varied array of instruments and platforms, all with differing formats, resolutions, and sets of attributes. Not only do a wide variety of data sources need to be dealt with, but a myriad of data "structures" as well (e.g., tables of chemical concentration versus raster images of sea surface temperature versus gridded bathymetry versus four-dimensional data, etc.). It has become increasingly obvious that more comprehensive data models are needed to support a much wider range of marine objects and their dynamic behaviors.
An example is ArcMarine, a data model involving a collaboration of Oregon State University, Duke University, and the Danish Hydraulic Institute with ESRI (dusk.geo.orst.edu/djl/arcgis; support.esri.com/datamodels). The common marine data types within this model extend current GIS data structures (points, lines, polygons, and rasters) to include more temporally referenced data structures that will allow for better representation of spatially and temporally dynamic marine data. This ongoing project seeks to provide the international marine GIS user community with a generic template for easier and faster input and conversion of data, better map creation, and most importantly, the means for conducting more complex spatial analyses by capturing the behavior of real-world objects in a geodatabase.
Grand Challenge: Dynamic Modeling in Space and Time
Probably the most interesting of the grand challenges facing marine GIS is the development of more dynamic models representing marine processes in space and time. The dynamic processes we are interested may be geophysical, ecological, resource management or economic in nature, but all of them will require fundamental adaptations to the way we collect, process, analyze and validate our data and our assumptions. It is still very difficult to imbed dynamic oceanographic models seamlessly into a GIS environment.
The questions that managers and policy makers are asking are becoming increasingly specific. More than ever now geospatial analysts are being asked to provide information to help forecast change over time. Parallel to the constraints we find representing a four dimensional ocean environment with two dimensional maps, our ability to forecast complex relationships at short time-intervals is constrained by statistical modeling approaches that were often originally developed for more static analyses. New developments in time-series and spatio-temporal modeling approaches are going to be crucial to completing the analytical framework of marine geospatial analysis. Many of these may be borrowed and adapted from the geocomputation, including diffusion modeling, time series regression, cellular automata and network, extensions, differential equation modeling, and spatial evolutionary algorithms (e.g., Box, 2000; Yuan, 2000; Peuquet, 2002; Albrecht, 2003; Green and King, 2003b)