Geography Reference
In-Depth Information
for exploratory visualization. They match well with the models of both DiBiase (1990) and
Pirolli and Card (2005); hence, in the next few sections we discuss issues of where we are
with exploratory geovisualization under these four headings.
3.2 Data preparation
In our information age data is being produced at a phenomenal rate. In some projects data
is being produced at such a rate that is only possible to store it, and with current tech-
nologies and techniques it is often impossible to analyse it fully. An ever-increasing amount
of data is at high resolution, with companies and agencies creating and processing high-
quality digital land use and socio-economic data with spatial attributes. With the ubiquity
of phones and global positioning devices, more spatial and time-stamped data is being
stored.
Additionally, such geospatial databases are collected by different researchers, created by
an assortment of sensors, saved in different formats, held physically at various locations
around the world and incorporate many variables and types of data. Hence, the user
is faced with several challenges when they wish to integrate the geographical data from
multiple databases, across different domains and potentially use the data for a purpose
that was not originally envisaged. Each of these challenges impinges upon Exploratory
Visualization.
3.2.1 Too much data!
The total quantity of data is often one of the most crucial challenges for a developer to
consider. Size certainly impinges upon the processing time. It takes longer to process huge
amounts of data, especially because users wish to mine similarities and structure within
the information, hence parts of the data need to be correlated with other elements. Friesen
and Tarman (2000) write: 'even at tomorrow's gigabit-plus networking bandwidths, sharing
these enormous data sets with distributed sites becomes impractical, and we'll need high
performance visualization resources to have any hope of near-real-time interaction or ob-
servation of this data'. Solutions to processing this huge quantity of data include: parallel and
distributed computing or remote computing solutions (Harwick, Coddington and James,
2003), and more recently interest has turned toward service- and grid-based architectures
(Aktas
et al.
, 2005; Brook
et al
., 2007). Spatial datasets certainly bring specific challenges
that high-performance computing methodologies can address, but in reality few researchers
have looked at solutions to these challenges. Reducing the size of the data obviously would
make the exploration more interactive.
Methods such as filtering generate smaller demonstration sets (shoebox datasets) that
would be processed more readily, take up less screen space and allow the user to focus
on relevant data. Spatial filtering may be achieved through cropping, sampling, averaging
(binning), partitioning, clustering and aggregation (Tang and Shneiderman, 2001). These
techniques can be achieved through pre-processing the data or closely linking the filtering
with the visualization (dynamically visualizing the results). Dynamic queries (Ahlberg and
Shneiderman, 1994) permit the user to directly interact with the visualization by adjusting
Search WWH ::
Custom Search