ul 2004
mp.
C.
M
| year
| year
| year
| year
10 year
1 year |
| year
for 1995.
ess, QAA
curacy of
Temp.
: Acc. 3
ATM
%
0.326
0.331
0.275
0.255
0.324
0.313
0.290
mpartment
for 1993.
vol = beach
assification
> accuracy,
NR = not
construct à
in a beach
interactivity
1ynamically
andle high
risualization
ti temporal
ent process
ations.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXX V, Part B4. Istanbul 2004
3.1 Dynamically Linked Views
With dynamically linked views we mean that graphs (e.g.
multivariate visualization techniques) and maps are displayed
separately but dynamically linked. If one element in a map is
clicked, the corresponding elements in other maps or graphs
will be highlighted. Conversely, by clicking on these
information elements in a graph, the particular object will be
highlighted in the map. Dynamically linked views increase the
user interactivity and are now considered indispensable for
supporting data exploration.
In the beach nourishment application, these information
elements concern the quality elements from table 2, which can
be activated by pressing the mouse button at a compartment. A
separate graphic will show the quality values for that particular
compartment. Also, by clicking on a phenomenon in a graphic,
the compartment concerned will be highlighted in the map.
32 Multivariate Visualization Tools
There are several methods to visualize spatial data quality
(McGranaghan, 1993; Lucieer and Kraak, 2002; Van der Wel et
al, 1994) Here, we focus on multivariate visualization
techniques for illustrating the quality elements for beach
nourishments derived from the ontological approach.
Over the past decade, many different visualization techniques
have been developed (Card et al., 1999). For geographic data,
visualization techniques can be categorized in geometrically
transformed displays or iconic displays. The geometrically
transformed displays include the scatterplot matrix, a
commonly used method in statistics, and the parallel coordinate
plot (Inselberg, 1985), that is a popular technique in exploratory
visualization. Star plots (Chambers et al., 1983) and Chernoff
faces (Chernoff, 1973) are techniques of iconic displays that
visualize each data item as an icon and the multiple variables as
features of the icons. From all these techniques, the dynamic
parallel coordinate plot has been demonstrated as a powerful
multivariate visualization technique (Spence, 2001) and has
been used in several applications (McGranaghan, 1993; Lucieer
and Kraak, 2002).
The dynamic parallel coordinate plot was introduced in 1985 by
Inselberg (Inselberg, 1985). The display is obtained by taking
dimensions as vertical axes thereby arranging them parallel to
each other. The individual data values are then marked off for
each dimension onto corresponding coordinate with the highest
data value as maximum value and the lowest as minimum.
From the structure of the resulting display one can draw
conclusions for the relationship of the corresponding data
values. A group of lines with a similar gradient can, for
example, indicate that their data records correlate positively.
Furthermore, outliers in values are easy to detect.
33 Temporal Ordered Space Matrix
To deal with multi-temporal datasets, we propose a novel
visualization technique, whereby linear elements (as with
coastline. compartment) are portrayed against the temporal
Variations of a particular user-defined quality element.
Therefore, we construct a matrix of squares, named temporal
Ordered space matrix. In horizontal direction, the matrix is a
Sort of schematized map and represents each compartment as a
Cll in identical order as in Cartesian space.’ Hence,
compartments in the west of the study area are depicted in the
matrix left of compartments in the east. The decision maker
needs to reflect each quality element with a conformance
quality level, i.e. a threshold value. When quality elements
fulfil the threshold value, they are shown as green cells.
However, if they fail, they are show as red cells. Quality
elements close to the threshold value are shown as orange cells.
The time is projected in vertical direction. Hence, the evolution
of quality elements can be easily interpreted from the temporal
ordered space matrix.
The main advantage of using ordered space is the preservation
of the adjacency relations between the compartments. This will
assist the decision maker in understanding and detecting areas
where effects of quality elements are important.
4. PROTOTYPE
Using the results of table 2 and visualization techniques as
mentioned above, we designed a prototype for multivariate
visualization, composed of three sections (see figure 3). The
first section is the main interface that displays a base map of the
study area. A pull-down menu gives the user the selection of
maps illustrating the beach volume or particular quality
elements for each compartment.
The second section consists of a parallel coordinate plot (PCP).
In the PCP each line represents a compartment. The vertical
axis shows the quality elements derived from table 2 for each
compartment. Here, the lower a line is located in the plot, the
better the dataset meets the user's preferences. In the PCP the
user can also the possibility to display the available datasets for
all times. Consequently, the PCP will display a particular
quality element as a time series, with the recorded date on the
vertical axis. A menu next to the display will give the
opportunity to display the quality elements per compartment or
as a timeserie.
The third section is the temporal ordered space matrix. Next to
the matrix is an entry menu, where the user can select a
particular quality element. Additionally, the user can enter a
threshold value for a particular quality element. The distribution
of this quality element in (ordered) space and time, as well as
its ability to meet the threshold value, will be depicted in this
matrix.
The prototype enables interactivity between the sections. The
user can explore the dynamically linked displays by pointing
the mouse on a compartment in the map and the corresponding
PCP-axis or matrix cell becomes highlighted. Vice versa, the
user can point on specific line or matrix cell leading to the
highlighted contour of the corresponding compartment. Even
more, the user can select a dataset by a moving window in the
temporal ordered space matrix. Besides, the user can select a
threshold value, changing the color legend of the map and the
PCP (see figure 3).
5. DISCUSSION
The prototype for multivariate visualization of spatial data
quality elements can be useful for interactive and explorative
purposes. Further elaboration of the prototype will help in
understanding datasets and their quality elements by instant
view. Its effectiveness towards insight of the data and their
shortcomings related to their quality, help decision makers to
1193