‹header›
‹date/time›
Click to edit Master text styles
Second level
Third level
Fourth level
Fifth level
‹footer›
‹#›
This presentation highlights some of the uncertainty visualization techniques used in the scivis community. It also describes a richer representation of uncertainty and a methodology for visualizing uncertainty using the new representation.
This presentation highlights some of the uncertainty visualization techniques used in the scivis community. It also describes a richer representation of uncertainty and a methodology for visualizing uncertainty using the new representation.
This presentation highlights some of the uncertainty visualization techniques used in the scivis community. It also describes a richer representation of uncertainty and a methodology for visualizing uncertainty using the new representation.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
These pictures are molecular surfaces of HIV Protease.
Top left:  uncertainty glyphs – uncertainty in direction mapped to width of arrow head.  Uncertainty in magnitude
can be mapped to extra bars on the arrow head (not shown).
Top right:  contour lines showing uncertainty.  Solid contour lines represent high certainty, while
dashed and broken contours represent places with higher uncertainty.  Color can be used for other
fields.  Other possibilities include line thickness, transparency, etc.
Bottom set:  Shows volume rendering of ocean circulation model data using a 2D transfer function.
The transfer function on the left is actually a scatterplot of data vs uncertainty.  The user then specifies
what color to render different parts of the scatterplot.  In this example, low data values with low uncertainties
are colored green, while high data values with low uncertainties are colored red.  The images on the right
shows volume rendered images of the circulation data.  The dark regions correspond to places where the
uncertainty values are higher – which also happen to be in the vicinity of a shelf break region where there
is higher standard deviation of different physical fields such as temperature, salinity, etc.
Greg Schmidt (NRL).  Experiments with various aspects of display of uncertainty.
Other examples that I’ve found on the web.
This one is from: “WEATHER FORECAST UNCERTAINTY MANAGEMENT AND DISPLAY”
Randy J. Lefevre*, Jonathan Pfautz, and Kenneth Jones.
Use of transparency and hue over a 2D spatial domain and 1 scalar variable.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
These pictures are molecular surfaces of HIV Protease.
Same data as previous slide.
Binwise addition is used instead of convolution addition.
Same seedpoint is used.
Streamline shows additive properties of uncertainties with binwise addition.
From Steve Feiner’s course notes.
Uncertainty in storm center track over time.
These pictures are molecular surfaces of HIV Protease.
These pictures are molecular surfaces of HIV Protease.
Visualizing dynamic molecular confirmations.
Measurements uncertainties does not allow us to know the precise state of a single molecule.
From: “Visualizing Dynamic Molecular Conformations”
Johannes Schmidt-Ehrenberg Daniel Baum Hans-Christian Hege
Zuse Institute Berlin (ZIB)
Positional uncertainty shown via transparent display of density of metastable conformations.
Rheingans (UM Baltimore).  Probabilistic surfaces.  Hybrid point-base and polygon-base representations. The more uncertainty, the more pointilistic or broken up the surface appears to be. From: “Probabilistic Surfaces: Point Based Primitives to Show Surface Uncertainty
Gevorg Grigoryan Penny Rheingans – Vis 2002
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Currently, uncertainty is usually represented by scalar quantities.
At times, a pair of numbers may also be used e.g. min/max.
Additional second moment statistics may also be used.
Using such representations, we have a suite of techniques that are available to present them together with the data.  Box plots and uncertainty glyphs show information at discrete locations, while contours, surface renderings, and volumetric renderings provide a more continuous spatial presentation.  Animations can also be used to show uncertainty in
the data.  Examples of these to follow in the next few slides.
1.Multivalued data sets are N times larger than uni-valued data.  Visualization may not be the
   right/immediate solution.  We need to do feature extraction first – analyze then visualize.
   Feature extraction will also need to work with multivalued data.
2. We just scratched the surface in terms of visualizing multivalued data.
   Need to explore other ways of visualizing it;
3. Need to train/evaluate users in these visualization.
   Choice of operators and metrics is large and can be drawn from different
   fields e.g. signal processing, info theory, statistics, etc.
   Need to know which is the appropriate one to use for different context.
4.The current multivalue representation does not preserve the spatial
   correlation information in the different realizations.  The visualization
   and analyses algorithms can try to recover that info.  This requires
   working with a matrix of values rather than a column of values for
   each scalar variable at each location in space!

Web Accessibility