![](bg3.png)
file format, that specify metadata like names and
types for every variable of the table. The data in
each data point have to match the number of
variables and records. Missing values are allowed.
The meteorological data collected by the
Meteorological Institute of Villa Clara represent a
time series of each of the variables collected from
1977 to the present. The meteorological variables
are collected over a ten-day period (dekadal), the
variables are: dekadal average mean, maximum and
minimum temperature, dekadal average of mean,
minimum and maximum relative humidity,
cloudiness (dekadal average in 1/8 cover sky),
dekadal mean wind speed, dekadal total rainfall,
dekadal insolation (average daylight hours), dekadal
water vapor pressure and dekadal average
atmospheric pressure. From each of the four weather
stations in the province, there is a time series of 13
variables.
Users can prepare their own multiparameter
dataset for each data point and use a wizard that
helps to customize datasets to be visualized. It
creates a configuration file that is used by the
coordinated visualization module.
There are several ways to visualize data with
multiparameter ScVis techniques. Some of them
allow visualizations over maps, where the
association of the data with the geographical area is
perceived. In particular, ScVis techniques for
multiparameter data do not necessary have to be
associated with a map. However, these techniques
can be applied to all variables in each data point to
obtain correlations between certain variables of
several points. It is possible to use for example,
Parallel coordinates to display the 13x4 (52)
variables corresponding to the 4 meteorological
stations.
Another implementation method is to display
data from each point separately, and show a map
where it is evident to which region each graphic is
related to. This is achieved with an arrow that
connects the region in the map with the visualization
panel and naming each panel with the same name of
the data point file visualized on the map.
Some ScVis techniques for multiparameter data
can be displayed directly over the map. Circle
segments, Recursive pattern, and icon-based
techniques are examples of such techniques (see Fig
1). With icon-based techniques is possible to show
an icon for each data source (a weather station in our
case), which represents the set of variables for an
observation at a given time. Some widgets (sliders,
scroll bars) are used to scroll over the time, changing
the icon on the map accordingly. This permits the
user to study the evolution of the data over time. The
same approach can be used to scroll the data with
respect to other variables.
All these methods were implemented in the
developed module for gvSIG. For example, for non
coordinated visualizations, that is, when the data do
not necessarily have to be associated with a map, the
system allows the user to load a data file to be
analyzed using any of the following techniques:
Parallel coordinates, Andrews graphics, Starfield,
Shape coding, Profile Glyphs, Circle segments,
Recursive patterns. Coordinated visualizations with
maps include the following techniques: Parallel
coordinates (in independent panels), Andrews
graphics (in independent panels), Starfield (a record
at a time), Shape Coding (a record at a time), Profile
Glyphs (a record at a time), Circle segments (all
selected records), Recursive patterns (all selected
records)
In our case study we obtained the best results
with the Recursive patterns technique. It seems to be
a very good technique to carry out spatio-temporal
analysis in GIS.
The developed module allows some
functionalities that are available for all the
techniques: Selecting attributes (only the selected
attributes are shown in the graphic), Selecting a
percentage of the records, Selecting according to an
attribute a range of values (makes a subsample of all
data in a range of values selected for a given
attribute), Showing legends (shows the global color
range for each variable, it takes minimum and
maximum values of all datasets, nominal values are
assigned a different color for each value, some
techniques like Profile glyphs show in the legend a
different color for each attribute), Reorganizing
attributes (a new order of attributes is given)
The ordering by functionality orders all datasets
according to a given attribute. Pixel-based
techniques and icon-based techniques use this
functionality to order all the values using this
attribute. By using the time as the attribute the user
can analyze data over time.
All developed pixel-based techniques also allow
to modify the length of the graphics. These
techniques can be visualized in independent panels
as well as over the map.
The developed geometric techniques were
designed to be visualized in independent panels, one
for each data point. They are also coordinated by the
main configuration panel. The coloring of the
records is showed with a color that is defined by a
given variable.
The ScVis module for gvSIG allows the user to
obtain visual information about the variables. In this
case study, meteorological data from the Villa
KDIR 2010 - International Conference on Knowledge Discovery and Information Retrieval
430