IlllllllillllllliHIWINIHHIIHIIIllll’llllllllllllllllUi

711153131293 01557 1577

mfl ill“.-

This is to certify that the

thesis entitled

CARTOGRAPHIC VISUALIZATION OF
FOUR-DIMENSIONAL DATA

presented by

Lauren Elizabeth Anderson

has been accepted towards fulfillment
of the requirements for

_MA_degree in W

W
Major professor

Date ———Jun¢21,_l.995___

0-7539 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

 

 

 

2 7 4v- v“—

_ T.——*_’___

CARTOGRAPHIC VISUALIZATION OF
FOUR-DI MENSIONAL DATA

By

Lauren Elizabeth Anderson

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Geography

1995

ABSTRACT

CARTOGRAPHIC VISUALIZATION OF
FOUR-DI MENSIONAL DATA

By

Lauren Elizabeth Anderson

Mapping time has challenged cartographers for centuries, and traditional solutions
have forced the map user to mentally interpolate the temporal dimension. The
development of cartographic animation techniques will allow time to be included in the
visualization of complex geographic data. Since 1959, researchers have examined the
power of bringing data on a two-dimensional plane to life. This research examines and
evaluates various cartographic representations of three-dimensional data changing over
time. Scientific visualization techniques which have been proven successful in numerous
other fields were applied to a four-dimensional geographic data set. Animations of three-
dimensional ocean temperature data were developed using desktop and workstation
technology to show the seasonal fluctuation of the mackerel habitat along the Northeast
U.S. Continental Shelf in the Atlantic Ocean. The results achieved show that not only is
it possible to effectively display four dimensions of geographic data simultaneously, but
also that this new mapping technique is a valuable research tool for exploratory data

analysis.

Capyri ght by
Lauren Elizabeth Anderson
1995

ACKNOWLEDGMENTS

My thanks go first and foremost to people at Michigan State University. Richard
Groop should be granted some great reward for his patience and help during the project,
as I know mine was not the easiest research to direct My committee members Jon
Burley and Daniel Hayes were invaluable resources and without their guidance 1 would
no doubt still be searching for answers. Members of the Department, especially Marilyn
Bria, Daniel Brown, Michael Lipsey, David Lusch, Judy Olson, Sharon Ruggles, Randy
Schaetzl, Judy Slate and Ellen White, have all provided important help for which I am
grateful. Special thanks to Mike McPherson, the AVS programming wizard, without
whom this project literally would never have been finished.

I will be forever grateful to the Geography Department at Dartmouth College for
providing me with a fabulous base of knowledge and research skills and for the
encouragement which continued long after I left Hanover.

My friends have come to my rescue on numerous occasions. Thanks to Patrick
McHaffie for inspiring me to pursue cartographic animation research; my fellow graduate
students at MSU for making the journey so enjoyable; my friends from Brandywine
Creek for always being just the excuse I needed to stop working and start having fun; and
all the email gang (you know who you are!) for dealing with my periods of temporary
insanity. Julie Gordon and Missy Antman are owed additional thanks for being my

Midwest Information Network.

iv

To all those at home, you know that your love, support and encouragement got me
through this. I don't know how anyone can really thank their family enough, sol will
simply say this: Mom, Dad and Megan...you guys are the best! Also, a big thank you to
The Michigan Andersons for always making me feel at home, and to the Kellys for all the
support and mail that never failed to put a smile on my face.

lastly, a special thanks to Brian Graber, who has been my constant reminder that
there is more to life than work. Wheneverl needed anything, I knew where to turn, and

for that I am forever grateful.

PREFACE

This work is accompanied by a CD—ROM which contains the animations that
were designed for the project. It also contains several examples of other animations,
which should enhance an understanding of the history of cartographic animation.
Animations can be viewed on the Apple Macintosh or on a PC running Windows using
software that is contained on the CD-ROM. These animations were designed on a 16
inch color monitor with 256 colors, and are best viewed in this environment. Several
figures in the text depict a single frame from an animation technique, and the
corresponding animation is referred to by that figure number. In cases where one figure
is representative of several movies, these animations are labeled with the figure number
followed by a letter. A list of figures and their corresponding animations, if any, has been

included. I hope this will eliminate any confusion.

vi

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................ viii

LIST OF FIGURES ......................................................................... ix
KEY TO ANIMATIONS ................................................................................................ x
1. Introduction and Background ...................................................................................... 1
Visualization ........................................................................................................ 2
Cartographic Representation of Time ................................................................. 5
Animation in Geography ..................................................................................... 9

2. Methods for Creating the Animations ......................................................................... 15
General Design Considerations ........................................................................... 16
Technical Development Issues ............................................................................ 2O

Dicer Animations .................................................................................... 20

AVS Animations ..................................................................................... 22

3. Evaluating the Animations .......................................................................................... 25
Discussion ........................................................................................................... 35
Impressions .......................................................................................................... 41

4. Conclusions ................................................................................................................. 46
Appendix A ..................................................................................................................... 50
Appendix B ..................................................................................................................... 51
Appendix C ..................................................................................................................... 56
' Appendix D ..................................................................................................................... 58
Bibliography .................................................................................................................... 72

vii

LIST OF TABLES

Table 1. Comparison of Cost for Scientific Visualization Systems ............................... 25

viii

LIST OF FIGURES

Figure 1. Time map of the Southern Margin of the Iaurentide Ice Sheet. .................... 6
Figure 2a. Ocean surface temperature on Northeast Continental Shelf (Spring

1991) .................................................................................................................... 7
Figure 2b. Ocean surface temperature on Northeast Continental Shelf (Fall

1991). .................................................................................................................... 8
Figure 3. Terrain Model of Sleeping Bear Dunes National Lakeshore. .......................... 13
Figure 4. General Diagram of Dicer Animation Process. .............................................. 18
Figure 5. General Diagram of AVS Animation Process. ............................................... 19
Figure 6. AVS network for slice animations. ................................................................. 23
Figure 7. AVS network for isosurface animations. ........................................................ 23
Figure 8. Dicer ocean temperature slice animation (surface to floor). ............................ 30
Figure 9. Dicer ocean temperature slice animation (along longitude). ........................... 30
Figure 10. Dicer ocean temperature slice animation (along latitude). ............................ 31
Figure 11. AVS ocean temperature slice animation (surface to floor). ........................... 31
Figure 12. Ocean surface temperature changing over time 1991-1994. ......................... 32
Figure 13. Temperature changes at 107 m (May l991-November 1992). ...................... 32
Figure 14. Full volume of ocean temperatures created in Dicer. .................................... 33
Figure 15. Mackerel habitat in three-dimensions, flipbook animation. .......................... 33
Figure 16. Mackerel habitat in three-dimensions, simulating temporal

interpolation. ....................................................................................................... 34
Figure 17. Eight degree Celsius isosurface changing over time l991-1994. .................. 34

ix

KEY TO ANIMATIONS

Figure 1. Time map of the Southern Margin of the Laurentide Ice Sheet.

Figure 3. Terrain Model of Sleeping Bear Dunes National Lakeshore

Figure 8. Dicer ocean temperature slice animation (surface to floor).

Figure 9. Dicer ocean temperature slice animation (along longitude).

Figure 10. Dicer ocean temperature slice animation (along latitude).

Figure 11. AVS ocean temperature slice animation (surface to floor).

Figure 12a. Ocean surface temperature changing over time May 1991-November 1992.
Figure 12b. Ocean surface temperature changing over time April l993-September 1994.
Figure 13. Temperature changes at 107 m (May l991-November 1992).

Figure 14. Full volume of ocean temperatures created in Dicer.

Figure 15a. Mackerel habitat in three-dimensions, flipbook animation l991-1994.
Figure 15b. Mackerel habitat in three—dimensions, flipbook animation Falls 1991-1994.

Figure 16a. Mackerel habitat in three-dimensions, simulating temporal interpolation
1991-1994.

Figure 16b. Mackerel habitat in three-dimensions, simulating temporal interpolation
Falls 1991-1994.

Frgure 17a. Eight degree Celsius isosurface changing over time May 1991-November
1992. '

Figure 17b. Eight degree Celsius isosurface changing over time April 1993-September
1994.

CHAPTER 1
INTRODUCTION AND BACKGROUND

... not only is a picture worth a thousand words but the Interpretation
of phenomena geographically depends upon visualization by means of

maps
— Allen Phllbrick, 1953

Geographic phenomena are multivariate in nature and change across time as well
as space. The ultimate goal of geographic visualization is to represent spatio-temporal
data in a way that conforms to human conceptualizations of the world (Peuquet, 1994).
This is why maps are crucial to geographic research. By way of a map, the researcher is
better able to interact with the geographic data set. Since temporal change is an important
aspect of many geographic problems, it is essential to understand not only how the map
represents the physical world, but also how time display is incorporated in cartography.

Reducing the complex three-dimensional relationships which occur in nature to
the two-dimensional surface of paper or the computer monitor causes distortion and
information is lost When the map user creates a mental picture, reconstructing the three-
dimensional world from the two—dimensional map in order to detect patterns and
underlying processes, there is much room for misinterpretation of the data. Removing
this step of mental visualization allows greater attention to be focused on analysis and
synthesis of the data, and this is now possible with the use of computers. Highly
sophisticated visualizations of geographic data can be generated which bring the data
closer to its natural form by simulating three dimensions. This can provide a level of
understanding not always achievable using simpler methods such as static two
dimensional maps (Robertson and Abel, 1993).

After the difficulty of spatial visualization has been removed, there is still the
question of how to display time. Time and space are inseparable qualities of phenomena,
and one cannot be mapped without the other (Wood, 1992). Whether the phenomena

2
being mapped occurs at one instant in time or is a process evolving over time, there is no

way to remove time from the map. For dynamic processes, static maps and series of
static maps have been used to depict change and movement in geographic data. But it is
often difficult to visualize the movement of such complex spatial data sets. Symbology
depicting change and movement leave a considerable amount of mental interpretation and
interpolation to the map user, especially when the in-between moments not depicted by
the maps are of importance. It is clear that there is room for error in this process and
there is no way for the cartographer to control for this error, since each map user has the
capacity for an infinite number of visualizations of what occurs at in-between moments in
slices of time.

As the sheer volume of geographic data continues to increase, it is becoming more
important to develop new methods for manipulating and analyzing this data. The
complexity of spatio-temporal data demands special tools to visualize multidimensional
changes which occur over time. Improved access to temporal maps could lead to an
improved understanding of geographic processes (Langran, 1992). While traditional
methods fall short of providing researchers with the cartographic output necessary to
effectively analyze spatio-temporal data, there is great potential for cartographic
animation to fill this void. Animated maps can show how an area has changed over time
or how well a process interacts with the area in which it occurs. Because the map user
does not have to make leaps through temporal space from one moment to the next,
animation allows change over time to be graphically represented in a more powerful
manner than static time series maps. The introduction of real and near real-time

modeling has numerous implications in examining geographic processes.

Visualization
Real world phenomena occur in three dimensional space and change over time.
Given that this is the inherent nature of geographic data sets, the ability to easily and

3
effectively detect patterns and study process when they have been reduced to a two-

dimensional surface is diminished. While it is impossible to recreate geographic
phenomena without some level of generalization, it is important that the map maintains
the integrity of the environment so that the map user is able to visualize the data with the
minimum amount of error.

Visualization is the process of creating an image from numerical or graphical
data. The image can be strictly mental or it can take a concrete form, an event which can
only occur once a mental picture has been formed. Visualization is not a new concept:
for centuries, researchers in all fields have sought a better understanding of data by means
of graphs, charts and diagrams. By portraying numerical data in a graphic form,
researchers have been able to see the functionality of their data sets more clearly. The
term visualization (or scientific visualization), however, has recently been used to
describe the phenomena of using computers to generate images from a data set. The best
definition comes from Wolff and Yaeger (1993), who maintain that '[s]cientific
visualization was invented as the process whereby humans use software to do the work of
converting number to image that nature does by physical processes.” By using
computers to represent numbers as images, the brain is freed from this responsibility, and
researchers can direct their attention to comprehending complex data structures. As the
quantity and complexity of available data incrcases, the role of the computer in pattern
identification in all arcas of research will become increasingly important.

While no one disagrees that scientific visualization is an important research tool,
there is some disagreement as to the role the tool should play in research. Larry L. Smarr,
Director of the National Center for Supercomputing Applications (University of Illinois
at Urbana - Champaign), believes that not only is this a powerful tool for scientists in
making discoveries, but that it is also a tool for communicating results to colleagues and

to the public. The power of visualization for presenting results is that the human visual

4
system can more easily comprehend an image than it can the underlying mathematics

which drive the world (Wolff and Yaeger, 1993).

Weber (1993) sees visualization as a way to alleviate information overload. He
points out that pattems in the data can be identified and analyzed more easily using
visualization because over half the neurons in the brain are dedicated to visual processing.
Weber's discussion focuses on applications of the technology in terms of the researcher,
and stresses that visualization is not about graphics but about understanding data.

Like Smarr, others concerned with visualization (DiBiase, 1990; Taylor, 1991)
place communication as an important result of the visualization process. Early work by
MacEachren et. al. (1992) agreed that presentation of results was part of visualization.
Since then his definition has evolved and MacEachren (1994) differentiates between
visualization and communication. His work defines visualization as one end of a
continuum which cuts through the space of map use, where communication is at the
opposite extreme. His view looks at visualization in terms of two broad types of map use
- maps used in searching for unknowns and maps presenting knowns. The line between
visualization and communication obviously is not hard and fast, because communication
is a part of every map, but MacEachren does believe the two are independent pursuits.

Despite these differences, it is agreed that the main objective of scientific
visualization is ”to develop techniques that make perceptible the most important
informational aspects of the data set” (Wolff and Yaeger, 1993). Visualization can be
used as a presentation tool as well as a research tool. In the end, how the resulting
product is used, be it for scientific inquiry or communicating results, is not the important
issue. Rather, what is important is that scientific visualization allows the researcher to
recreate reality for the purpose of better understanding a given data set.

There are a variety of components which together comprise a scientific
visualization system. The effectiveness of the visualization will depend on how various

parameters are manipulated to best suit the data set. Among the elements which can be

5
used to represent data are traditional graphic methods such as color, texture, and three-

dimensional graphic rendering; extending these options to a multimedia environment
brings the addition of sound, interaction and animation to the data set (Weber, 1993).
Animation in scientific visualization allows three dimensions of spatial data to change
over time, so the researcher is able to investigate the four dimensions of a complex
dynamic data set (four-dimensional data) simultaneously. When all this is happening in
an interactive environment where the user is able to rotate the image and view it from
another perspective, change the illumination on the image, get a two—dimensional slice of
a three-dimensional object, or control the pace of the animation, patterns which may have
otherwise gone unnoticed can be identified and analyzed.

With an estimated 50 percent of the brain's neurons associated with vision, it is
not difficult to understand why visual mapping is recognized as one of the most powerful
analytical tools (Wood, 1994; MacEachren, 1994; Wolff and Yaeger, 1993). As in
numerous other fields, the introduction of scientific visualization has given geographers a
powerful new tool with which to explore data. Traditionally the fundamental method for
visualizing geographic data was through mapping (MacEachren, 1994). Yet mapping
extends beyond geographic research and is essential to many sciences. Researchers have
successfully used scientific visualization to map three-dimensional models of neuron
activity in the brain, DNA particles, chemicals dispersion, wave functions for orbit
molecules, airflow dynamics and ozone depletion, and proven that mapping is not limited

to the 2-D surface (Hall, 1993).

Cartographic Representation of Time

The study of geographic data through maps has traditionally been restricted to one
instance in time. For geographers, the challenge has been the reduction of the ”four-
dimensional reality of time and three-space into little marks on paper“ (Tufte, 1990 p.

119). This generalization of the worid results in enormous simplification and the

6
tremendous loss of information. Stripping the world down to a two-dimensional surface,

visualization of geographic trends becomes increasingly arduous for the map user.

 

10.000 B.P.

 

 

 

 

Figure 1. Time map of the Southern Margin of the Laurentide Ice Sheet.

Spatio-temporal data is traditionally displayed using one of two strategies: single
static maps or time series "chess maps”. A single static map depicts limited aspects of
change in or across space by showing change over time in a particular place or place-to-
place movement over time (Figure 1). The second type, what Monmonier (1990) refers
to as ”chess maps”, show how a phenomena evolves over time by presenting several
snapshots, each for a discrete moment in time, juxtaposed so that the map user can
compare spatial positions across time. Figure 2 is am example developed by the
National Marine Fisheries Service (NMFS) in Woods Hole, MA. to display ocean
temperature data (see also Appendix D). The problem with these types of maps,
however, is that they represent states but not the events that change from one state to the
next, leaving the burden of dealing with temporal processes to the map user. (Langran
and Chrisman, 1988)

It was not until the introduction and widespread use of Geographic Information

Systems (GIS) and the subsequent necessity for depicting cartographic time that

 

' p ,
u +86 ‘r‘0 \‘ eggs: .
\,/ M9”
43- fis\ . ' ‘3'
42 can u I Ar.) , \.\j .
u- M: ‘ \”/\f\/_‘) (3:;
"’ "I (\k7/‘\ Amman"

DEL9105
SPRING GROUNDF'ISH SURVEY
6 MARCH - 16 APRIL 1991

SEC TEMPERATURE (DEG C)

 

 

 

 

7'3 fii 7‘1 7'0 We

Figure 2a. Ocean surface temperature on Northeast U.S. Continental Shelf (Spring
1991). Reprinted from Holzwarth-Davis and Taylor, 1992.

 

I .
42»

f n
... l 0%

' " xtd/ ’
40- NJ. II/\,,2€b‘>:°- f/Q:\’%\
.’ 12 " .'

\‘lv .
39- a“ /n C .

DEL9110

”(in

FALL GROUNDF'ISH SURVEY
10 SEPTEMBER - 24 OCTOBER 1991

SEC TEMPERATURE (DEG C)

 

 

 

 

Figure 2b. Ocean surface temperature on Northeast U.S. Continental Shelf (Fall 1991).
Reprinted from Holzwarth-Davis and Taylor, 1992.

9
traditional display methods were challenged. The importance of displaying the resulting

maps and statistics from GIS analysis is a natural outgrowth of the technology since the
fundamental function of GIS is the analysis of processes which change over time
(Koussoulakou and Kraak, 1992). However, as volumes of literature have emerged from
the GIS community citing the importance of time as an element of the geographical
database, a notable omission has been temporal display methods.

Currently, the most comprehensive work on the dimension of time in GIS focuses
on the conceptual and structural problems of representing geographic change, not on how
it is displayed (Langran 1992). Langran deemed single static maps and time series maps
insufficient for depicting the most important component of time - change. While she
does note that in addition to static maps, animation techniques can be employed to show
change over both time and space, this alternative display method is merely mentioned and

is not evaluated.

Animation in Geography

Animation is an effective solution for dealing with temporal data sets. In order to
process, investigate, and understand complex data, simple representation is not enough.
With the use of computers, the process of mentally visualizing the temporal dimension
from a numerical data set is replaced by a graphic rendering of these numbers. As
technology has advanced to this point, the map has evolved from a communication device
to a powerful visualization tool (MacEachren and Ganter, 1990). Geographic
visualization has redefined the map as a research tool whose goal is "to stimulate
scientific insight by facilitating the discovery of patterns and relationships in spatial data"
(MacEachren and Ganter, 1990). To better understand the merger of scientific
visualization with geography, it is important to understand the history of the animated

map.

10
Norman J. W. Thrower (1959, 1961) gave the geographic community its first

glimpse of the potential use of animation as an alternative form of cartographic
representation. Thrower realized the potential impact animation could have in the field of
geography; time could be incorporated and geographic processes could be portrayed as
dynamic entities. By using the same techniques as cinematographers, a film could be
generated from hand drawn cels which held images in a state of suspended animation.
Although animation offered cartographers a new tool for representing the temporal
aspects of geographic data sets, the cinematic production skills necessary to create such a
map prevented the process from being embraced. From what appeared to be a huge
breakthrough in data representation techniques, relatively little research followed. Nine
years passed before another article on animation was published in a geographic journal,
and had it not been for the availability of computers, the duration may have been longer.

Pioneering work by Corrrwell and Robinson (1966) proposed using the computer
to generate cartographic animations by using the CRT (cathode ray tube) as the drawing
board and exposing the film directly from the monitor. The problem with this method is
not the technology, rather the lack of software which will produce the desired images.
The authors saw computer generated film as the solution to the problems of animated
cartography and provide researchers with a tool which would ease the visualization of
data changing over time. Widespread use of computer animation did not emanate from
this article.

Tobler's (1970) research on urban growth in Detroit resulted in the first computer
generated animated map sequence. This crude isosurface animation was the first attempt
at showing a three-dimensional surface changing over time. A simulated statistical
surface showing population growth mutated with the passage of time but lacked any
spatial or quantitative references, and ultimately was unable to convey more than very
general information. The next application was not until Moellering (1973) used computer

animation to map traffic accidents on film as a movie. Different sized flashes were

1 l
displayed on a street network and changed hourly over the course of a week, with larger

flashes representing more serious accidents. From this animation it was possible to locate
dangerous intersections and identify when the most accidents occurred.

In an article chronicling the 30 year history of animated cartography, Campbell
and Egbert (1990) illustrate the advances that have come from these humble beginnings.
Even as technology in the 1980's increased the power, speed and efficiency of both
personal and super computers, the development of this tool progressed slowly. In the
computer environment, the animated map could not only show change over time, but also
gave researchers the ability to generate digital terrain models and explore surfaces from
different vantage points under various illumination conditions. With the use of the
computer, Moellering (1980a, 1980b) showed that a static surface can be modeled and
dynamically explored in three-dimensions. By manipulating the image in a highly
interactive real-time system, the animation allows the researcher to rotate an image and
zoom in to inspect areas of interest. The more complex the surface, the more useful this
approach could be. Moellering (l980b) saw the benefit of such a system as the additional
capabilities which could be developed to incorporate the temporal dimension. However,
these developments did not come to pass and the constraints of time and money made
animation unaffordable, and interest in research on cartographic animation again faded.

It was not until 1990 that there was a renewed interest in cartographic animation.
Computers were now commonplace in geography departments, and software
development had reached the point where creating animated maps entirely on the
' computer was technically and economically feasible. No longer was it necessary for the
cartographer to be well versed in computer programming or have to expend tremendous
amounts of time and energy to produce an animation. Gersmehl (1990) gave geographers
an invaluable research aid with his comprehensive discussion of animation techniques.
This framework for determining which type of animation would be most appropriate for

displaying a given spatio-temporal data set is based upon ”the identification of design

12
issues that are relevant to 'four-dimensional cartography' " (Gersmehl, 1990). This was

the first time that the term four-dimensional was applied to cartography, and although his
work did not specifically address the animation of a three-dimensional cartographic
image, Gersmehl sought to establish a common vocabulary to facilitate discussion
between cartographers and animators so that animated maps could effectively represent
the four-dimensions researchers were investigating.

Since Gersmehl's (1990), a majority of the works being produced in cartographic
research programs are focusing on improving dynamic displays on a two-dimensional
surface. Adopting the traditional practice of abstracting the three-dimensional physical
world to a two-dimensional surface, adding the temporal dimension sets time series maps
in motion. This removes the burden of mentally generating the intermediate steps and
setting the data in motion, allowing the brain to focus on uncovering patterns or looking
for anomalies in the data. Animations of this type have been created to display a wide
variety of geographic data. When MacEachren and DiBiase (1991) were asked to produce
a series of maps depicting the spread of acquired immune deficiency syndrome (AIDS) in
Pennsylvania, they realized that animation would be ideal for the intended application - to
convey information about the spatial characteristics of the disease to the public. Recent
projects include the use of animation of temperature contours for viewing the spatio-
temporal deviations and trends in climatic change across the United States over the past
century (Weber and Buttenfield, 1993); in maps depicting historic journeys and events for
a CD-ROM encyclopedia (DiBiase, 1994); as part of a multimedia visual supplement to
travel writing (Des Roches, 1994); and animated maps which are juxtaposed and linked
to statistical graphs and text (Monmonier, 1992). These and other two-dimensional
dynamic displays do shed light on many geographic problems, and allow researchers to
more clearly and easily distinguish patterns. But by creating a visualization that is
limited to a two—dimensional representation, geographers have not realized the full
potential of this tool.

 

Figure 3. Terrain Model of Sleeping Bear Dunes National Lakeshore.

The use of three-dimensional visualization tools is invaluable when the right
problem is being addressed (Dorling, 1992). Unfortunately, little research specific to
geographic data sets has been done in this area. Moellering (1980a, 1980b) gave a
glimpse of how three-dimensional animation could be used with his terrain model; but
until the invention of scientific visualization software, such a product was too costly to
produce in terms of time and money when the resulting graphics were so simple. Since
then, three-dimensional modeling and animation has found its niche in numerous fields as
a powerful visualization tool. Chemists, engineers, astronomers, film makers and
medical researchers all have benefited from the increased accessibility of powerful
workstations and the development of scientific visualization software (e.g. AVS ) that
now makes three-dimensional animation easy. Previously geographers interested in
employing visualization techniques were forced to choose between generalizing a surface
to two dimensions and creating a cost effective animation and having a more data intense

animation which was costly to create. As workstation technology permeates geographic

14
research, cartographic models will be less abstract and provide researchers with new

procedures for examining spatio-temporal data sets. Researchers have used visualization
software to model thunderstorms (Wolff and Yaeger, 1993) as well as generating realistic
terrain models (Figure 3) that go far beyond the geometric rendering Moellering (1980a)
produced. Visualizing three-dimensional temporal processes in an environment which
allows the researcher to interact directly with the data should provide new insights into
geographic problems.

The purpose of this research is to examine and evaluate various cartographic
representations of spatio-temporal change in a complex data set. To do this, three-
dimensional ocean temperature data will be animated to show the seasonal fluctuation of
the mackerel habitat along the Northeast Continental Shelf in the Atlantic Ocean. Several
different animation techniques representing various levels of technical sophistication will
be developed and evaluated. The goal is to find difficulties, limitations, advantages,
"costs“, and other characteristics of the various techniques and to provide

recommendations concerning their applications.

CHAPTER 2
METHODS FOR CREATING THE ANIMATIONS

This chapter describes the framework for the development of three-dimensional
map animations. Temporal change is an important aspect of many geographic problems,
but it is often difficult to visualize the movement of complex spatial data sets.
McCormick and others (1987) found that ”better visualization of a problem leads to a
better understanding of the underlying science, and often an appreciation of something
profoundly new and unexpected". With volumes of data collected daily in both the public
and private sector, it often is difficult to understand how the pieces fall together.
Researchers need a tool which allows data to be mapped in a way which fosters a search
for questions rather than merely presenting results (MacEachren and Monmonier, 1992).
Using scientific visualization techniques to explore both the spatial and temporal
dimensions of a data set gives researchers an alternative to numbers. For geographers,
this means reinventing the type of map used for exploratory data analysis of four-
dimensional data sets.

Serious work has yet to be undertaken in terms of exploring alternative display
methods for f our-dimensional data sets. Technology has now reached the point where the
researcher is no longer limited to creating "two and one half dimensional" displays where
the height (or depth) is a frame upon which different feature can be draped. Volumes can
now be rendered, restoring three-dimensionality to the data. Bivins and Palmer (1989)
created "smart maps" of ocean resources, three-dimensional sea temperature maps with
which the researcher could interact. Their project proposed adding the temporal
dimension to study dynamic oceanographic phenomena through three-dimensional

animation, but this never came to fnrition.

15

16
General Design Considerations

Several animations of ocean temperature on the Northeast Continental Shelf in the
Atlantic Ocean were developed to assess how well a complex four-dimensional data set
can be graphically displayed. It is important to understand how ocean temperatures
throughout the volume fluctuate during the year since fish habitats are defined, in part, by
temperature zones. For example, the mackerel is a semi-pelagic / pelagic offshore
migrant fish, which means that it will move up and down the water column following a
temperature gradient. Mackerel aremostly found in water between 8° - 20° C, from the
surface to approximately 100 fathoms (182.88 m), and breeding is most productive at 8°
C (Bi gelow and Schroeder, 1953). Since the goal of this research is to examine
geographic visualizations of data in a more 'natural' environment, the mackerel habitat is
a good subject to animate because the data is inherently four-dimensional. If the
animations are successful, the researcher will gain new insight into the data, and will do
so without the burden of creating a series of mental images to recreate the three-
dimensional world.

While there are an abundance of scientific visualization packages which run in the
workstation environment [Application Visualization System (AVS) (Advanced Visual
Systems), Advanced Visualizer, (Wavefront Technologies), IDL (Research Systems,
Inc.), IBM Visualization Data Explorer (IBM), Khoros (The Khoros Group”. there are
significantly fewer visualization software tools for the personal computer. More
visualization software has been designed to run on the Apple Macintosh [IDL, Director
(Macromedia), Transform and Dicer, (Spyglass, Inc.)] because it traditionally has been
regarded as the superior graphics platform; but as the graphics quality of the PC has
improved, many of these products [IDL, Director] are being made to run on either
platform. Slocum (1994) gathered reviews from several researchers of various
visualization packages and rated them according to documentation, ease of learning and

use, and product support. The packages were not compared and contrasted so these

17
results were unable to recommend one visualization package over another. However, the

evaluation was helpful in deciding which package use for this research. Purchasing
software was not an option since high end workstation visualization software costs
thousands of dollars. AVS and Wavefront products were available at Michigan State
University, and after examining of both, AVS was selected. Unlike several of the
desktop packages, Dicer was specifically designed as a visualization tool and appeared to
be the best option for developing animations of three-dimensional data.

Digital ocean temperature data was obtained electronically from the National
Marine Fisheries Service (NMFS), Woods Hole MA. Spring and fall bottom trawl
surveys from 1991 through 1994 were selected because the survey was consistently done
twice a year, over the same area, for 4 years. Each survey contains approximately 300
sampling stations across the Northeast Shelf from Cape Hatteras to the Gulf of Maine.
One of the cruise objectives on each of these surveys is to collect hydrographic data of
temperature, salinity and density (sigma-t) data measured in the water column from the
surface to the ocean floor, averaged into 1 m intervals on an irregular grid of stations
(Holzwarth-Davis and Taylor, 1992, 1993, 1994). Because data collection took place
over several weeks, the data were compiled into one temporal instance as if all samples
were collected at the same time. This allowed the animations to contain information for
the study area as a whole rather than as small parts along the shelf. Before this data could
be used to create animations, the header record containing country and ship codes, cruise
and cast number, time and date had to be removed. A Quickbasic program (Appendix A)
was used to extract latitude, longitude, depth (m) and temperature (0C) and convert the
files from NODC 80-column ASCII format into ASCII comma delimited format. NMFS
also provided the digital ARC/INFO coastline coverage which is used in several of the
animations.

After the data were reformatted, they were used as input in several animation

packages. Spyglass Dicer was used on a Macintosh Quadra 700 to create several

18
animations of the four-dimensional data set. Dicer is the first program for the Macintosh

that displays and animates 3-D volumes of data in color and that allows the researcher to
interactively create slices, isosurfaces, blocks and cutouts for viewing those data values
(Spyglass, Inc., 1993). The program is excellently documented in the manuals, and
Spyglass has a very knowledgeable technical support staff which can be tapped as an
additional resource. Since the program is designed to maximize the capabilities of the
Macintosh operating system, creating the animations did not require programming

experience or prior knowledge of the program and only took. a few hours.

 

ASCII data

J

Hierarchical Data chmat
(HDF)

Qtickl'ime Movie

 

 

 

Figure 4. General Diagram of Dicer Animation Process.

Figure 4 shows the general process which is used to create an animation using
Dicer. Additional data conversion is often necessary to ensure that the data is readable by
the program. Since the ocean temperature data was irregularly sampled, it was necessary
to convert it to a regular three-dimensional grid. This gridded volume could then be
displayed and manipulated within Dicer. Once the image to be animated had been

constructed, the f rames were exported from the program using the animation modules.

19
These frames were then brought into another animation program and linked together to

form Stand-alone animations.

Six additional animations were created on a IBM RISC System/6000 workstation
using AVS. AVS is a scientific visualization program "designed for end-user
extendibility and platform independence for high performance computing, graphics and
imaging applications” (p.11 AVS, 1992); and based on an extended form of the Data
Flow Architecture. The object oriented programming is excellent and easy to use given
that the researcher has experience with the program. However, the learning curve for
AVS is very steep. AVS has the power to do almost anything imaginable to a data set,
but the number of modules and complexity of many processes almost necessitates having
an AVS expert available at all times to achieve the desired results. Modules not native to
the program can be built in, but this requires strong knowledge of the program and
programming skills in either C, C++ or FORTRAN. Animations for this work were done
in consultation Mike McPherson, an experienced AVS user at Michigan State University,
and required over 40 hours of programming in AVS before they could be generated.

 

 

 

ASCII AVS ARC/INFO
data» File Descriptor «coverage
Fields
4

 

AVS
Data! t_'Mapping_’FW +Data0utput

 

Qu’ckTime Movie

4
Sparkle

Qtl'ckTime Movie

 

 

 

Figure 5. General Diagram of AVS Animation Process.

20
Figure 5 shows how raw data is converted into an animation using AVS. Again

the irregularly spaced data was fit to a regular grid, and was converted into a format AVS
could read. This data field was then input into AVS, where it flowed through several
different modules. These modules build the raw data into an image, generated the
animation and created an output file. The resulting animation was very large (over 15
MB) and needed to be crossed over to the Macintosh platform, so it was transferred
through another software package where the animation was converted and compressed

into a stand alone movie.

Technical Development Issues

This section gives an in-depth discussion of how the animations were developed.
It has been provided for those readers interested in the technical issues behind the
animations. In addition to listing the parameters specific to each step, this section will
provide some insight into how much time and energy is necessary to create animations

using Dicer and AVS.

Dicer Animations

The Data Utility converts a variety of data formats into Dicer compatible three-
and four-dimensional data sets. This utility was used to transform the ASCII files into
3D floating-point Hierarchical Data Format (HDF) files. The hydrographic data was
converted to a 357x30x30 target matrix, where the z-axis value of 357 was determined by
the deepest reading collected from the trawl surveys. The output matrix data locations
that were not set by input data values were interpolated using a 3x3 smoothing kernel
routine called simple fill. A value of -l was assigned to missing data values. The Data
Utility was then used to convert a folder containing a series of 3D HDF matrix files into a

single compatible 4D HDF file.

21
The HDF matrix file opens directly into Dicer. The view size was set to 1/2 times

the size of the grid for the z-axis, and 8 times for the x and y-axes. This allowed the
volume to be viewed at maximum size within the confines of the desired QuickTime
animation size. For all Dicer animations, axes and labels were added to the bounds of the
volume to provide some orientation to the viewer. A spectral color scheme was selected
to represent the temperature data. Making use of the full range of colors, with the coldest
water represented by black and purple and the warmest represented by red, allows for
easier pattern identification because the spectrum has logical order as a visual variable.
Two distinct types of animations were generated for the volume representation
using two of the options in the Save Image Sequence... command. Running space
animation and time animation resulted in a series of PICT files. Space animation
required a 3D HDF input file and the number of frames to be created to be defined. The
output was a series of PICT files which were then brought into the shareware program
MovieMaker. MovieMaker converts the PICT files into a QuickTime movie; this
application was also used to set the speed of the animation. Time animation does not
interpolate temporally, so the number of output PICT files is limited to the number of 3D
files used to create the 4D HDF file, which in this case was eight. A QuickTime MooV
for these eight PICT files was created in MovieMaker, but the resulting animation was
unsatisfactory (see animations 15a and 15b). In the absence of temporal interpolation,
the images created from the 4D data set were imported as castmembers into MacroMedia
Director. It was then possible in Director to assemble an animation using pixel dissolves
to simulate temporal interpolation. Director was also used to create the clock for the
temporal animations. These animations were then exported as QuickTime MooVs,

utilizing the animation compressor to reduce the size of the output file.

22
AVS Animations

All visualizations are created in object oriented execution models called networks,
which provides the "data processing pipeline”. Each piece in the network, or module,
plays a specific role in the creation of the visualization. Modules can be categorized by
function: data input, filter, mapper and data output. When wired together, the data flows
through the network sending the output from one module to the input of the next. By
manipulating module settings, the researcher is able to interact with the data.

Before assembling the network and animating, the data files were converted into
a readable AVS format Several modules not native to AVS were needed for the
conversions, which were acquired from the International AVS Center and written
specifically for this project (Appendix B). The coastline was converted from an
ARC/INFO coverage into an AVS geometry format using plot_xyz. Reading the ASCII
data files into AVS format required numerous steps. First, the AVS file descriptor
module was used to read the ASCII files (Appendix C). The trivar module, with a radius
of 1000 minutes and a weighting factor of 1/d2, was used to interpolate each of the eight
ASCII files on a 20x20x20 regular grid. A modified version of trivar was used to create
a mask field, showing which of the elements in the new regular grid covered a physical
space which actually contained data in the .csv file.

Figures 6 and 7 show how the AVS network compiles the animation. The
networks are identical, except for the mapper module portions, which convert the
numerical data into a description of a three-dimensional geometric object. Both
simultaneously read the spatially interpolated data and mask fields into the network. The
data field then runs through the timeslice module, written specifically for this project,
which does the temporal interpolation using the routine POUNT, which implements
Neville's algorithm for polynomial interpolation (Press, 1989).' Clamp is used to deal
with the sub-zero temperature artifacts introduced by the temporal interpolation process.

All values less than 0.1 are set to 0.1, since zero has been reserved for undisplayed points

23

I animated float I

 

 

  

read field

 

read field

 

 

  
 

field math I .
orthogonal sllcer
m-II 3;.
field to mesh
I read geom I
I

 

 

I generate colormap I

use.

color legend I

   
      

 

     

l
I

  

Figure 6. AVS network for slice animations.

 

 

I animated fleet I reed field reed field

 

 

 

 

I generate colormap I

 

    

 

 

 

 

color legend 1' A '

 

Ireed geom _ I I
J

Figure 7. AVS network for isosurface animations.

24
such as those on the land or beyond the study area. F ieldmap then joins the two input

fields by multiplying the temporally interpolated field and the mask field. The mask is
defined such that locations which are part of the study area are assigned a value of one,
others are assigned a zero value. When combined, the data is pared down to only display
ocean temperatures along the Shelf. The data then goes through either an isosurface
(Figure 7) or orthogonal slice (Figure 6) where generate colonnap applies a spectral
(high-red, low-purple) color scheme to the graphic image of the temperature data and

. this, along with labels, legend and the coastline file are displayed in the geometry viewer.

When all viewing parameters are set, the animation was generated by attaching
the write_MooV module to the output link on the geometry viewer. This writes a (very
large) QuickTime MooV file, which was then transferred to the Macintosh in binary
mode, with the file type set to 'MooV" and the creator type to "TVOD'. The files were
then compressed using the shareware utility Sparkle 2.3 to reduce the size from 16 MB to
less than 1 MB animation. Animations did not suffer any apparent loss in graphic quality
due to the transition.

In the next chapter, in depth descriptions will be given of each animation. The
animations will be evaluated based on their success as a tool for visualizing a complex 4-
D data set. Comparisons will be made between the animations created on the Macintosh
personal computer and those made on the IBM RISC/6000 workstation. Some factors
which will be considered are data manipulation power, graphic image creation,
equipment and time costs incurred to assemble the animations, and output quality. In
addition to this discussion, a dozen researchers who are familiar with geographic data
viewed the animations and provided their evaluations of the product. Animations were
briefly described and shown to individuals or to pairs of researchers, and the reaction of
each researcher was recorded for each animation. These researchers represent the various
fields of geography, and their subjective evaluations are summarized following the formal

discussion of results and evaluations.

CHAPTER 3
EVALUATING THE ANIMATIONS

Visualizing four-dimensions is one of the largest challenges researchers face as
they begin to explore their data sets. A general lack of knowledge about scientific
visualization software often prevents such techniques as three-dimensional modeling and
animation from being employed. This research seeks to remove this banier and
encourage the use of visualization technology to explore complex geographic data set by
displaying how effectively a dynamic geographic phenomena can be displayed in an

alternative environment.

Table 1. Comparison of Cost for Scientific Visualization Systems

 

 

 

Softwa'e Cost Hrsdwae Minimum System Average Cost
Requiranents
Spyglasss Dicer $695 Apple Macintosh Hard drive, 4 MB RAM, $2,000
System 605/607 or 7.0
Advmced Visualiznion $8,000 Most Unix 50 MB hard disk. 16 MB 312.500
Systems AVS Platforms RAM

 

 

 

 

 

 

 

The visualizations for this project were created using two software packages. One,
Spyglass Dicer, represents a low cost, easy to learn desktop tool for representing
volumetric data. The other, AVS, is a much more complicated package representing the
high technology end of the spectrum. A comparison of hardware requirements and costs
are shown in Table 1. Each package allows the researcher to interact with the data and
manipulate the graphic output, but in different ways. Animations will be discussed in
terms of how much of the 4-D data set they were able to successfully display, graphic
quality, interactivity, output, time cost, and the overall ability to convey information
which the researcher previously was forced to mentally interpolate from static
representations of the data. Following the descriptions and evaluations is a brief section

summarizing some of the comments I received as people reviewed the animations. This

25

26
section should serve to enhance the success or failures indicated in the previous section,

as well as generate additional suggestions for future improvements or changes which
would make these visualizations more successful.

When examining a complex data set, it is often desirable to reduce it to smaller
parts before examining all the dimensions simultaneously. This is often the case when
dealing with four-dimensional data Without an understanding of the three-dimensional
space, it is impossible to understand how that space evolves over time. Creating a mental
image of the volume and its attributes from a series of maps results in an inconsistent,
inaccurate representation which is difficult to explore. Figures 8, 9 and 10, created with
Dicer, show how visualization technology can be used to render three-dimensional space
at one instant in time. These animations show the temperature of the ocean along the
Northeast Continental Shelf from Cape Hatteras to the Gulf of Maine. Each shows the
movement of a slice moving through the volume of water and displays the temperature
using a spectral color scheme. The black area in all animations represent space beyond
the study area; the black on the left side of the cube is the land, and the area to the right of
the temperature display is deep water beyond the shelf. The bottom of the cube which
turns black at various depths is the point where readings stopped at the ocean floor.

Figure 8 shows a visualization of the volume by moving a slice along the depth
axis from the surface to the bottom which has been set at the deepest location in the study
area; thus each slice shows the temperature at a given depth. The pattern illustrated here
shows that in late September the water is considerably warmer in the shallower,
southernmost part of the Shelf. With the exception of a few canyons offshore of the mid-
Atlantic states, the volume of cold, deep water is primarily in the Gulf of Maine. It is
perhaps easier to display the ocean depth in Figure 9 as the slice moves through the cube
from west to east showing temperature and depth at along a line of longitude, or in Figure
10 where the slice moves from northward from the southern part of cube, where

temperature and depth are displayed by latitude. Since Figures 9 and 10 show

27
temperature and depth simultaneously in a graphic environment, the topography of the

ocean floor begins to be defined.

Visualizations created using AVS allow for similar representations of the data set.
It is possible to create a slice moving through the water on the shelf, this time with a
coastline added for reference. To display temperature for late June (Figure 11), the image
has been rotated and tilted at an angle so that the coastline and the volume can be shown.
Better spatial interpolation and graphics rendering makes it easier to identify ocean floor
features as the slice moves from the surface to the bottom.

While these four animations provide an example of ways to investigate the spatial
component of a f our-dimensional data set, they do not touch upon the temporal
dimension. Before embarking on this combination, it is often desirable to examine the
temporal component in a more conventional animation, one of a two-dimensional surface
changing over time. Animating temperature surfaces over is time not new, but doing so
in a visualization package as opposed to a graphic animation package allows for temporal
interpolation. Figures 12a and 12b depict changes in ocean surface temperature over four
trawl surveys, a time period of one and a half years. Again the temperature is displayed
with a spectral color scheme, which allows for logical interpretation of order. A
coastline, Lake Ontario, Lake Champlain and a portion of the St Lawrence have been
incorporated to provide the necessary orientation and to help to define the extent of the
shelf. Time has been indicated by a counter which starts with the spring trawl and
progresses through the next three surveys counting by day. The coloring that extends
beyond the land boundary is an artifact of the masking process, which in all cases was
unable to cleanly clip the coastline to the data set. The line extending from the shore to
the St. Lawrence is a line segment which could not be removed without destroying the
integrity of the coastline. The fuzzy bound along the outer edge is a result of pixel

dithering and is not intended to convey any three-dimensional effect

28
To visualize four dimensions the temporal dimension must be combined with the

three-dimensional spatial display. There are several ways this can be done. The slice
animations can be merged with the two-dimensional temporal animation to create a four-
dimensional visualization of temperature change along the shelf at a given depth, as in
Figure 13. For the time between the spring survey in May 1991 and the fall trawl in
November 1992 the temperature surface at 107 m below the surface has been animated.
Again the coastline is in place for reference, and the clock counts the days from the
beginning of the first survey through the end of the last survey. The dark ”holes” which
appear on Georges Bank off the coast of Cape Cod are areas where the topography of the
continental shelf is elevated.

If it were possible to effectively animate the entire volume of water to show
changed over time,the researcher would be inundated with information, especially if the
research focused on a small thermal range. For example, the mackerel habitat is confined
to the thermal region between 8-20 oC. Visualizing the mackerel habitat as it changes
over time would benefit from minimizing the output to only temperatures within that
range. Dicer allows the researcher to limit the amount of information displayed by
blanking colors outside a designated temperature range. Figure 14 shows how layers of
information can be peeled off the volume, leaving only the mackerel habitat portion of
the data opaque and available for investigation. Making use of this feature allows
animations specific to a user defined tolerance to be generated. Once a three-dimensional
graphic image has been defined, it should be relatively simple to apply the fourth
dimension. Unfortunately, one of the major limitations of the Dicer software is that it is
unable perform temporal interpolation. Time animation will take a four-dimensional data
file and generate images only of the three-dimensional records which comprise the file,
limiting the number of output images to the number of time records in the data set. The
result is a jumpy visualization that is no more than a slideshow of three-dimensional
renderings (Figures 15a and 15b).

29
However, if these output images are brought into a graphics animation package, it

is possible to simulate temporal interpolation by utilizing pixel dissolves as a transition
from one image to the next The full four year period of the data set can be animated
(Figure 16a), showing the changing water volume for the mackerel habitat. The layout is
similar to other Dicer animations, where the land is located along the left side of the cube,
and the water which does not fall in the designated temperature range is not displayed.
The legend is divided in half along the horizontal, displaying the full range of
temperatures as well as the opaque range. Temperatures displayed, the opaque region,
show the mackerel habitat The volume along the Shelf changes dramatically from spring
to fall, and the passing of time is indicated by the timer in the lower right comer. This
visualization is dramatically different from Figure 15a because time has been
incorporated. Similar visualizations can be developed to examine how stable a seasonal
habitat is from one year to the next (Figure 16b). Here the four-dimensional data set was
built containing only the four Fall surveys, but the graphic layout is identicalto the full
year animation. From one Fall to the next, the volume does not change drastically. An
inegular sampling and interpolation method account for some of the more dramatic depth
changes, but the temperature pattern essentially remains constant

In the spring and eariy summer Mackerel breed in the 8 0C thermal zone. The
location of this region has importance to the maintenance of the food chain. Two
visualizations (Figure 17a and 17b) show the movement of the 8 0C isosurface over a one
and a half year period. One color, and shading to indicate changes in depth and spatial
location, is used for display since only one temperature zone is being animated. The
coastline has been tilted to allow the volume to be viewed. The counter at the bottom is
used to gauge the passage of time. The ladder effect which appears is a result of the AVS
algorithms not knowing how to accurately interpolate across time and space when the

isosurface dips into a canyon on the Continental Shelf.

 

Figure 8. Dicer ocean temperature slice animation (surface to floor).

 

Figure 9. Dicer ocean temperature slice animation (along longitude).

 

Figure 10. Dicer ocean temperature slice animation (along latitude).

 

Figure 11. AVS ocean temperature slice animation (surface to floor).

 

Figure 12. Ocean surface temperature changing over time 1991-1994.

 

Figure 13. Temperature changes at 107 m (May 1991-November 1992).

Northeast Shelf ‘ mu I99!

 

Figure 14. Full volume of ocean temperatures created in Dicer.

 

Mackerel Habitat {Northeast Shelf l9? 2 43-41

 

Figure 15. Mackerel habitat in three-dimensions, flipbook animation.

 

Hacker-91 Habitat (Northeast Shelf l99 l 4941'

 

Figure 16. Mackerel habitat in three-dimensions, simulating temporal interpolation.

 

Figure 17. Eight degree Celsius isosurface changing over time 1991-1994.

35
Discussion

The success of any visualization is largely a function of the data set For this
research, visualization were generated from a real world four-dimensional data set. This
means that the unpredictable and uncontrollable anomalies which occur in nature cannot
be ignored. It would have been possible to create a theoretical data set and generate
relatively flawless visualizations, but that would have defeated the purpose of this
research. In order to examine the role of visualization technology in exploratory analysis
of four-dimensional geographic data, it is necessary to do so with real data, even if that
data does not lend itself to easy, clean visualizations of all four dimensions.

Data is rarely collected in the exact spatial or temporal format the researcher
desires. The NMFS hydrographic data is a good example of how the quality and
availability of geographic data varies. Numerous surveys are done throughout the year,
but the time and location of the sampling varies from year to year. The entire continental
shelf is sampled twice a year, while smaller regions are sampled more regularly.
Visualizing thermal zone fluctuations from real data presented numerous problems which
show up in the animations. These problems are not necessarily a limitation of
technology, but rather a reflection of the data under investigation.

Comparing visualizations to static time series maps (Appendix D) is impossible.
While they are all depicting the same information, they are two entirely different types of
display and research tools. When we make use of time series maps, we are essentially
creating our own animation similar to the output from a visualization package and there is
no way to compare the two. It is possible, however, to compare animated maps and
evaluate their success relative to output from other visualization packages.

Both AVS and Dicer are interactive visualization packages, and when the
animations are output from these programs, the visualization which results is one in
which all the decisions have already been made. This is beneficial for presentation

purposes, but the researcher needs to interact with the data. The animations which were

36
created for this research were designed to demonstrate how visualization software can be

used to represent f our-dimensional geographic data It is necessary to understand that
each of these packages has numerous powerful features which cannot be adequately
demonstrated by these animations. The major feature is the ability in interact in real time
with the geometric image which is generated. In both environments, the three spatial
dimensions of the data are displayed as part of a cube. In Dicer, the view orientation can
be rotated in 90° increments along any of the three planes. The workstation environment
allows the AVS user to freely rotate the volume about a center point, so any viewing
angle can be achieved.

Orientation is only one of the interactive parameters which is missing from the
animations. Both Dicer and AVS allow for the user to define which temperature range is
displayed, what color pallet is used and what time frame, if any, is animated. Since the
AVS time slice module will only interpolate time from four input time periods, which
times are used is again decided by the researcher. Each animation has been designed
with the intention of demonstrating the power of visualization technology and introducing
a new data analysis tool to geographers. Given these limitations, animations are discussed
in terms of design and data output quality, ability to extract information, and overall
comprehensibility.

It is difficult to discuss the animations in great detail because the purpose of
creating a visualization is to substitute graphic representation for figures and words. It is
easier to discuss what is missing and detracts from a clear and easy interpretation of the
data than to discuss the features which bring the patterns to life. All utilize color graphic
renderings of three-dimensional spaces and movement, through time and/or space, to
convey information about the f our-dimensional data set How well the data can be
explored depends on the options available in each visualization environment The

animations designed for this research all achieved success at some level; thus comparison

37
gives mixed results as to which visualization was 'the best'. Instead, the result is a clearer

picture of an ideal to strive for.

Very smooth animations, such as these, require more frames and consequently
take longer to assemble and require more memory. On average, Dicer animations are
about 1.2 MB, and took about 10 - 15 minutes to create images for each QuickTime
movie. AVS animations were significantly larger, about 15 MB, but took less than 5
minutes to create the same number of frames. It took about the same amount of time to
convert each file to a self -contained QuickTime movie. The Director movies took an
additional 30 rrrinutes to assemble after Dicer generated the images, and exporting as
QuickTime movies brought the total assembly time to almost one hour.

The overall color and graphics quality is sharper in the AVS animations, which is
to be expected since they were created on a workstation with 24 bit color as opposed to 8
bit color Dicer is designed to maximize. But when Sparkle is used to compress the AVS
QuickTime MooV file, they are reduced to 8 bits color, so some of the image sharpness is
lost, but the graphic quality is still better than what is achievable in Dicer. The 3-D
graphics are blockier in Dicer; when a smooth surface is desired, AVS is a better choice.

The goal of this project was to examine various cartographic representations of a
f our-dimensional geographic data set, and create animations which allowed the researcher
to examine all four dimensions simultaneously. With some of the visualization methods
this was impossible. If actual data-based four-dimensional visualization is desired, then
AVS must be used. If it is possible to work with simulated temporal interpolation, then
Dicer produces very nice results when the images are animated in Director, but that takes
a substantial amount of time and often would not be worth the effort The Dicer time
animation module images, when animated, were not at all smooth and failed to convey a
sense of movement in time. This was due to lack of data. If enough intermediate images
could have been input into the four-dimensional matrix, the results achieved in Figures
15a and 15b would be dramatically improved.

38
Dicer‘s inability to perform temporal interpolation imposed serious limitations on

the success of visualization from the Macintosh. But the fact that the program is easy to
learn, data input is straightforward and the cost is reasonable makes this an attractive
product. It was possible to generate several animations of 3-D space at one instance in
time using Dicer. Slices moving through a cube of 3-D space (Figures 8, 9 and 10)
makes more intuitive sense when moving from the surface to the bottom so that changes
in temperature can be associated with a change in depth without losing spatial orientation
in the x and y plane. Since Dicer does not allow a coastline to be superimposed on the
cube, it is difficult to relate temperature information to anything other than depth at points
in the same plane is it moves through the cube along the latitude or longitude. The
animation is a bit more successful as the slice moves northward from the southern
extreme of the study area. Each of these animations suffer from the same problem: it is
difficult to identify location within the cube. If it were possible to place additional ties on
the axes, the problem might be lessened. These three animations do succeed in bringing
the volume to life with sharp colors and conveying a sense of the ocean floor topology.
Since each slice is essentially a shaded isoline map, they are easy to understand because
they are not unlike maps we are already familiar with. This makes the animations easy to
understand because it is not necessary to adjust to a new graphic representation of the
data.

It is possible to create similar slice animations in AVS. Figure 11 also shows a
temperature slice moving through the volume from the surface to the ocean floor. This
animation was not set to record as many frames as the Dicer slice animation (Figure 8),
so it doesn't flow quite as well, but the volume is still defined as the slice moves from the
surface to the bottom. This animation makes it easier to see where the floor is elevated
because the temperature surface has been interpolated more closely. Identification of
temperature zone is easier, thanks to smaller classifications along the legend bar. The

view angle was selected so that the coastline would provide spatial reference as the slices

39
travels along the z-axis of the data cube. An area where this animation does not achieve

the same success as Dicer is that there is no way to gauge depth. As can be seen in all
AVS visualizations, this is a glaring downfall.

Reducing the cube area to display only the mackerel habitat allows the researcher
to focus on a selected temperature region. This feature shows only the regions of a
potential fish habitat at the time of a trawl. While it is easier to locate the habitat with
other temperatures removed, Dicer does not allow the bounding color to be made
transparent; thus, it is impossible to see how deep temperatures inside the volume extend.
However, when brought into Director and animated, these three-dimensional models of
the habitat do a very good job of showing how drastically the size of the potential
mackerel habitat changes over the course of a year. Since no temporal interpolation is
possible, the clock runs based on animation time and arrives at the appropriate season
when the full image appears from the dissolve. This “three and a half dimension"
animation, which almost shows a time dimension, is a good link between visualizing
three and four-dimensional data because it teaches the viewer about a new presentation
style. Once comfortable with how to interpret a three-dimensional graphic and assimilate
a pseudo-change over time, it is then possible to fully investigate a four-dimensional data
set

The surface temperature animations (Figures 12a and 12b) showing change over
one and a half years make it very easy to see temperature patterns. Here, the data cube is
being viewed from directly overhead, so only three of the dimensions of the data set are
being examined. This view is especially useful if viewing true temporal animations for
the first time, since the amount of information which must be processed is minimized.
Since the AVS color legend is nicely categorized, it is easy to focus in on specific
temperature. Time has been marked with a running count of days to show how drastic
changes can occur over short periods of time. This clock could be changed to represent

time in any number of ways since each tick of the clock is linked to data which has been

40
temporally interpolated. This is different than the Dicer/Director clock which was limited

by the length of time of the dissolve between one image and the next Pixel dithering
around the edge is a result of color compression from 24 to 8 bits, and is not intended to
convey any sense of the third dimension. It is unfortunate that generating data for the
time between samples was such a difficult problem to overcome. Without a considerable
amount of time and an AVS expert on hand, it would have been impossible to display the
temporal dimension. Even with this luxury of this technical support, it was unfeasible to
interpolate between more than four time periods.

Figure 13 is the simplest display of all four dimensions of the data. Temperatures
107 m below the surface are animated over a one and a half year period. This is an
excellent resource for pattern identification at a given depth, and indicates where the
ocean floor is elevated in the same way the AVS surface to bottom slice animation . If
this slice were viewed from directly overhead, as in Figures 12a and 12b, the graphic
would not make as much sense because the areas along the coast where the shelf is
shallow would be blacked out, and the temperature data would appear to be floating out
to sea. A lack of depth reference is only mildly troublesome here, since the researcher
sets the depth at which the temperature surface will be animated.

The isosurface animations (Figures 17a and 17b) provide the most detailed and
specific information of the four dimensional data set By setting the temperature to any
desired level, the researcher is able to investigate changes in volume as the thermal zone
evolves over time. Shading adds considerable understanding to the volume, defining the
shape. There is no way to tell the specific depth or location of a specific point'in the
volume, but this visualization provides researchers with insights into the patterns and can
stimulate thinking about which additional factors should be incorporated into the
analysis. Viewing the visualization in AVS gives the researcher the opportunity to
examine the volume from any angle, and watch the patterns evolve for different thermal

ZOIICS.

41
All the animations are successful in providing a clearer picture of a complex

geographic data set. Some convey more information than others, and there is definitely a
difference in output quality. The ideal design for a full four-dimensional visualization
would be one in which the volume of ocean data was placed within a representation of
the ocean floor topology that extended from the shoreline to the edge of the Continental
Shelf. Orientation would be less of a problem because the viewer would have multiple
vertical and horizontal cues. Unfortunately, this was not possible using any of the
visualization techniques. If it were, the AVS isosurface animations would be much more
effective, because the researcher could locate the thermal band in relation to the ocean
floor. The same would be true for the Dicer/Director animations of the mackerel habitat,

and for all of the animations showing three-dimensional space.

Impressions

A dozen geographers with different research interests were asked to view the
animations and share their impressions. All are familiar with traditional maps and the
role maps play as part of exploratory data analysis. The responses are summarized here
to provide additional evaluation of the results achieved.

The overall impression was extremely positive, everyone who viewed the
animations was impressed with the data display. Almost everyone commented on how
"cool” the visualizations were, almost as if they were surprised that data could be
displayed in such a manner. The most common observation was how easily patterns
could be detected. Once these patterns were identified, a majority began to question the
pattern and hypothesize about the processes driving the patterns. I was surprised at how
many people immediately began to explore the data, especially since no one was familiar
with the data set. It was also interesting that several people who viewed the
visualizations immediately thought of how this could be applied to the type of geographic

data sets they work with. I find these responses particularly encouraging since there is

42
strong indication that these visualizations are indeed a part of future geographic research.

As more and more people are introduced to the technology and the potential is realized,
visualization will play a powerful role in exploratory data analysis.

Comments were also directed at very specific animation issues. People responded
positively to the color scheme, and specifically remarked that they liked the colors in the
Dicer slice animations. Several mentioned that the animation speed was a bit fast, and
that they would really like to have more interactive control over the animations. Since the
animations were exported as QuickTime movies, the interaction was limited to looping,
looping forward and backward and stepping through the frames. The option to step
through the frames one at a time was utilized by the viewers almost every time a new
animation was shown. That time was not displayed with a clock counting by month and
day or Julian date was troublesome for some, others did not appear to notice the clock at
all. The problem of orientation was unanimously commented upon; at least one of the
following problems was mentioned by each person: the absence of a coastline in the
Dicer animations, the inability to reference depth in the AVS animations and the desire to
have the ocean floor topology represented.

Some individuals compared the slice animations to one another, and the
consensus was that Figure 9 was not understandable and therefore not very effective, but
that Figure 10 worked well and that the slice moving from the surface to the bottom
(Figure 8) was one of the best visualizations of the bunch. The animations dealing with
all four dimensions of the data set were judged a success, but there were several
suggestions as to how to make them more successful. These suggestions include showing
the whole volume of water when a temperature isosurface is highlighted to give a sense
of surrounding temperatures and the ocean floor, being able to interactively query the
image for exact locations or statistics, and linking the image to other data sets such as fish

population counts to model the three-dimensional environment over time.

43
The visualizations were also viewed by two researchers who are familiar with

hydrographic data, one of whom had previously worked in NMFS and had experience
with ocean dynamics on the Northeast continental shelf. Their comments were more data
specific but having a strong background and understanding of the data allowed them to
judge how useful the visualizations really are. Both researchers found the animations to
be very informative. They specifically mentioned that picturing the data is beneficial and
that this methods was better than flipping through a series of maps. They thought that a
picture can be easily remembered, and can be remembered more accurately than a long
description; and that animation is a good tool for getting an overall impression of data or
for describing patterns and processes to others. Since neither researcher was familiar
with the data set, their approach was to examine the data as if they would be using it for
their own research. After running the animations through once they were more closely
examined by stepping through the images, interaction which made it easier for individual
pattern identification. The omission of the subsurface and coastline and the fact that the
clock was set as a counter instead of displaying calendar time was also mentioned by
these viewers as troublesome. However, the comments specific to the animations were
quite revealing since they came from people who are familiar with this type of data set
The two researchers extracted a significant amount of information from the
visualizations. Based on knowledge of ocean system functions on the shelf, Figures 12a
and 12b were considered to be an accurate representation of thermal dynamics. Cold
patterns on the shallow Georges Bank occur because the water here gets very cold very
quickly. The warm pocket of water which appears south of long Island may be the
Hudson River plume. Circulation patterns also emerged and for someone familiar with
this data, the animation explained why fish assemblages change so much over time. One
explained that the east coast pattem from the Gulf of Maine to the New York bight is the
most compressed temperature range in the whole Atlantic Ocean and commented that this

visualization in particular does a good job of showing the evolution of water masses in

44
that area. Both liked the animation of the 107 m subsurface temperatures as it changed

over time, again commenting on the processes that drive the patterns.

The Dicer slice animations were described as interesting and informative to some
degree, and overall the researchers' descriptions matched those already expressed. From
Figure 9 it was possible to extract temperature zones and get a feel for the ocean floor,
but there was a problem understanding how the movement reflects the coastline. The
slice moving from the surface to the floor (Figure 8) again showed the pattern nicely and
gave a better understanding of the shelf tepology. Understanding the location of the
coastline was clearer when the slice moved from south to north through the volume, as
motion simulated traveling along the coast, a concept which was easier to mentally
picture.

While these animations were judged useful, both researchers were more excited
about the visualizations which included the temporal dimension. 1 think this is due to the
fact that the data sets they generally work with involve the temporal dimension but they
had been unable to visualize it They found the Dicer/Director animations (Figures 16a
and 16b) were a ”neat” way to visualize the data but that it was difficult to examine the
three-dimensional graphic since it was impossible to see inside the volume. This
animation technique, since it did not truly include the element of time, was thought to be
useful to look for persistent patterns over successive seasonal observations (Figure 16b).

The isosurface animations (Figures 17a and 17b) were the visualizations these
researchers seemed most excited about Despite the fact that it was a little hard to see
specific points on the map and to orient depth without some vertical scale, this
visualization successfully defined the thermal habitat As the volume mutates,
knowledge of mackerel migration patterns enhances the amount of information which can
be extracted from the visualization. Knowing that the fish migrate along the front edge of

the thermal habitat as it moves onshore in the spring after wintering in the warmer

45
offshore waters is an essential piece of information if you are interested in identifying

where the fish are likely to be located.

The researchers provided several suggestions which they felt would enhance their
exploration of the data. The first was that it would be helpful to look at temperature
deviations from the average for an extended time period. This would help to identify
anomalies which could provide complications, and they could be investigated before a
full analysis is performed. Uniting this data set to others and being able to interactively
query the visualization was another feature they would like to have when exploring the
data.

This additional input provided by two experts confirmed several hypotheses.
First, as effective as the visualizations are without knowledge of the data set, they are
even more so if the researcher if familiar with the type of geographic data being studied.
For organizations such as NMFS or the National Oceanic and Atmospheric
Administration (NOAA), where map making is one of the important things they do,

animation can be a powerful option for investigating four-dimensional data.

CHAPTER 4
CONCLUSIONS

Visualization is the process of recreating reality for the purpose of better
understanding a given data set In geography, this has traditionally meant extracting
many dimensions of information from a two-dimensional surface and mentally
assembling a representation of the real world. The inaccuracy of such a method
prompted an investigation into alternative methods of representing the 4—dimensions of
spatio-temporal data. What this research has shown is that it is possible to display
geographic data in a much more realistic manner by utilizing scientific visualization
techniques.

Since the map has become so vital to the research process, it is necessary to
continue to improve upon the ways geographic data can be displayed. Improving the map
means being able to display more information without making the map more
complicated. Reducing the amount of generalization brings the map display closer to a
model of reality. In this new mapping environment the goal is to maintain the integrity of
the 4-dimensional data set so that data exploration does not rely so heavily on mental
imagery. With computers it is now possible to render three-dimensional models, and
numerous fields have made use of this technology to explore various types of data. Until
now geographers have not explored the possibility that visualization can enhance the
understanding of geographic processes.

This research has shown that not only is it possible to simultaneously display
four-dimensional geographic data, but also that this new mapping technique is a valuable
research tool for exploratory data analysis. By looking at the data from a new perspective
it is possible to uncover many patterns which would likely go unnoticed if the only

resource available was mental interpolation from over generalized maps. It is much

47
easier to extract inf onnation about natural phenomena from a visualization which creates

a model of the environment in question and sets it in motion.

Sometimes it is impossible to display all four dimensions at once. This does not
mean that a better map cannot be created using visualization software. For example, the
Dicer slice animations provide more information about ocean thermal zones than can be
gleaned from static maps of surface and bottom temperatures, just as the AVS animation
of surface temperatures changing over time is more descriptive than static time series
maps. Individually these animations allow the researcher to investigate another
dimension without the burden of mental image creation. When the brain is freed from
this task, more attention can be focused on searching the data for patterns. The resulting
analysis will likely be more accurate because visualizations provide graphic output based
on mathematical interpolation. Multiple representations of the same data set enhance the
understanding of displays when one visualization cannot sufficiently display all
dimensions.

People who have had the opportunity to explore complex spatio-temporal data
with animation are very excited about the potential of this technology. The initial awe of
the technology and novelty of the technique quickly gives way to the realization that
patterns may be more easily identified when the data is animated. Once introduced to the
technology, it is difficult to stop thinking about other opportunities to explore similarly
structured data sets. As other geographers break away from traditional methods of
exploratory data analysis and begin to realize the potential of visualization, we will see an
increase in the use of animated three-dimensional maps for research and presentation
purposes.

Although it is currently time consuming to animate a four-dimensional data set,
that will not always be the case. For the moment, the question we must ask ourselves is:
do the animations we currently are able to generate add to our geographic knowledge? If

so, it is worthwhile to continue researching the effectiveness of four-dimensional

48
cartographic visualizations. For AVS, the quality of the resulting animations is well

worth the cost, especially when the results can be viewed in their native environment and
the interactivity is maintained. Dicer is a good first step at investigating a four-
dimensional data set, but if time is an important dimension to the study, as it often is, then
any scientific visualization package that does not allow for temporal interpolation is
insufficient However, technology changes so quickly that this software will be obsolete
within a year or two, thus it is not so important as to how to produce animations using
this software but rather what we have learned about animating four-dimensional data.
Perhaps the most important thing is that it is possible to model real spatio-temporal data
and create a visual representation of a three-dimensional geographic object which
changes over time. As important is that this technique is useful for research as well as for
presentation purposes.

As visualization software is made easier for geographers to use, its power as a
data exploration utility will be realized by more people, and its use will become more
common and therefore more excepted as part of research process. As the technological
age continues, it is not unlikely that these maps will replace traditional maps for
presentation and exploration of many types of data. If these animated maps do indeed
work better than anything we have ever had before and they become easier to construct,
then perhaps they may add to the traditional map as a method of data display

This research is only a first step in the development of a cartographic visualization
of four-dimensional data and more work needs to be done. Several animation techniques
have been demonstrated but many more are possible and need to be evaluated. It is also
possible that some animation techniques will work better for some geographic data sets
than others. A set of design standards also needs to be established so that researchers can
design animations which will convey information efliciently. Finding these parameters
will require testing map users to find the most effective visualization techniques. Further

development of this technique will benefit from cognitive testing; once researchers

49
understand how this information is interpreted and remembered, designs can be improved

to make the visualizations even more effective.

Until visualization packages are developed which deal specifically with four-
dimensional geographic data sets, it is also necessary to conduct more developmental
research in order to find ways around the problems that exist in the current software.
This may mean writing a program to perform temporal interpolation outside Dicer and
then seeing if a satisfactory animation is achieved after more images are generated, or
creating new modules for AVS which would allow the ocean floor topology to be
modeled. These new features should then be compared to previous methods and their
success evaluated.

The success of visualization in other fields highlights how important it is for
geographers not to turn away from this technology. A move away from traditional

mapping is not bad if new maps can teach us more about the environment we live in.

APPENDICES

so
Appendix A

Quickbasic Program
! Richard Groop
! Department of Geography
! Michigan State University
! converts NODC 80 column format to comma delimited
I January 23, 1995

cs = I!"

count = 0: counter = 0

CLS

OPEN ”filenamedat“ FOR INPUT AS #1
OPEN 'fllename.csv" FOR OUTPUT AS #2

' FOR i = 1 TO 200

WHILE NOT EOF (1)

id = VAURIGHT$ ( i113, l) )

ON id GOTO 10, 25, 3O

10 count = count + 1: LOCATE 10, 10: PRINT count
counter = 0
latdeg = VAL(MID$ (ins, 5, 2) )
latmin = VAL(MID$ (in$, 7, 3) ) I 10
londeg = VAL(MID$ (in$, 11, 2) )
lonmin = VAHMID$ (in$, 13, 3) ) I 10
year = VAL(MID$ (in$, 19, 2) )
mon = VAL(MID$ (in$, 21, 2) )
day = VAL(MID$ (in$, 23, 2) )
botdep = VAL(MID$ (in$, 34, 4) )

GOTO 20
30 counter = counter + I: LOCATE 20, 20. PRINT counter
FOR i = 2 TO 5
x3 = MID$(in$, i, l)
IFx$<> " 'THENBSELSE
l=i-1
35 NEXT

depth = VAL(MID$ (in$, 1, 1)): s = 1 + 2

temp = VAL(MID$ (in$, s, 4) ) : s = 84- 5

salin = VAL(MID$ (in$, s, 5)): s = s + 5

sigmat = VAL(MID$ (in$, s, 4) )
PRINT #2, count; c$, counter, cs, latde , c3, latrnin, c5, year, c$, mon, c3, day, c$,
botdep, c$, depth, c3, temp, c3, salin, , sigmat, c3

20 WEND

'20 NEXT

25 LOCATE 1, 1: PRINT ”Error"
CIDSE#1, #2

END

51
Appendix B

Formula for Temporal Interpolation

C
C POLINT
C From "Numerical Recipes (FORTRAN)"
C Given arrays XA and YA, each of length N, and given a value X, this routing
C returns a value Y, and an error estimate DY. If Plx) is the polynomial
C of degree N-l such that P(XAi)sYAi, i=1,...,N, then the returned value Y-PlXF.
C
subroutine polint(xa,ya,n,x,y,dy)
parameter (nmaXalo)
dimension xa(n). ya(n), clnmax), dlnmax)
ns=1
dif:abs(x-xa(1))
do 11 i=1,n
diftsabslx-xalil)
ifldift.lc.dif)then
ns=i
difadift
endif
cli)-ya(i)
dli)-ya(i)
11 continue
y-ya(ns)
ns=ns-1
do 13 M81,N-1
do 12 i=1,n-m
ho-xali)-x
hp:xa(i+m)-x
w=cli+1)-d(i)
deneho-hp
iflden.eq.0.)pause
denaw/den
d(i)=hp*den
c(i)=ho*den
12 continue
if(2*ns.lt.n-m)then
dysclns+ll
else
dy-d(ns)
ns-ns-l
endit
yay+dy
13 continue
return
end

QOOOOOODDGOOOOOOOOOOO

000

52

Appendix B

AVS Temporal Interpolation Module

mod_gen Version 1

Module Name:

Author:
Date Created: Mon Feb 27 13:46:01 1995

This file is automatically generated by the Module Generator (mod_gen)
Please do not modify or move the contents of this comment block as

"timeslice"
Lauren Anderson

(Mapper)

(Subroutine)

mod_gen needs it in order to read module sources back in.

input 0 "invector" field 3D 3-space rectilinear float REQUIRED
input 1 "slicetime" real REQUIRED

output "outscalar" field BD 3-space 1-vector rectilinear

param 1 "timeO" typein_real 0.00000 PLOAT_UNBOUND FLOAT_UNBOUND
param 2 "timel" typein_real 0.00000 FLOAT_UNBOUND FLOAT_UNBOUND
param 3 "timeZ" typein_real 0.00000 FLOAT_UNBOUND FLOAT_UNBOUND
param 4 "time3" typein_real 0.00000 PLOAT_UNBOUND FLOAT_UNBOUND
param S ”time4' typein_real 0.00000 PLOAT_UNBOUND FLOAT_UNBOUND
param 6 "timeS" typein_real 0.00000 FLOAT_UNBOUND FLOAT_UNBOUND
param 7 ”time6" typein_real 0.00000 PLOAT_UNBOUND FLOAT_UNBOUND
param 8 "time?" typein_real 0.00000 FLOAT_UNBOUND FLOAT_UNBOUND

8nd of Module Description Comments

tittfiiitftfttttitttttiifittiiiitfiiittfl'i

Module Description
tittrtittttittiitttttt'tiititffitittitttt
integer function timeslice_desc()
implicit none
include 'avs/avs.inc'

integer in_port, out_port, param, iresult
external timeslice_compute
integer timeslice_compute

call AVSset_module_name('timeslice’, ’filter’)
call AVSset_module_flags(single_arg_data)

Input Port Specifications
in_port . AVScreate_input_port('invector',
5 'field BD 3-space rectilinear float', REQUIRED)
in_port - AVScreate_input_port(’slicetime', 'real', REQUIRED)
Output Port Specifications
out_port . AVScreate_output_port('outscalar'.
$ 'field 3D 3-space 1-vector rectilinear')

Parameter Specifications

param a AVSadd_parameter('time0’,‘real', 0.0, FLOAT_UNBOUND,
$ FLOAT_UNBOUND)

call AVSconnect_widget(param, 'typein_real')

param . AVSadd_parameter(’time1','real', 152.08, FLOAT_UNBOUND,
$ FLOAT_UNBOUND)

call AVSconnect_widget(param, 'typein_rea1')

param . AVSadd_parameter('time2',’real', 304.17, FLOAT_UNBOUND,
$ FLOAT_UNBOUND)

call AVSconnect_widget(param, 'typein_rea1')

param a AVSadd_parameter('time3','real', 547.50,
$ FLOAT_UNBOUND)

FLOAT_UNBOUND,

000

00

0000 CHOU

53
Appendix B

AVS Temporal Interpolation Module (cont)

call AVSconnect_widget(param, 'typein_real’)

param a AVSadd_parameter('time4',’real’, 699.58, FLOAT UNBOUND,
s FLOAT_UNBOUND) -

call AVSconnect_widget(param, 'typein_real')

param s AVSadd_parameter('timeS','real'. 912.50, FLOAT_UNBOUND,
s FLOAT_UNBOUND)

call AVSconnect_widget(param, 'typein_real')

param . AVSadd_parameter('time6',’real', 1034.17, PLOAT_UNBOUND,
s FLOAT_UNBOUND)

call AVSconnect_widget(param, 'typein_real')

param = AVSadd_parameter('time7','real’, 1216.67, FLOAT_UNEOUND,
$ ELOAT_UNEOUND)

call AVSconnect_widget(param, 'typein_real’)

call AVSset_compute_proc(timeslice_compute)
----> START OF USER-SUPPLIED CODE SECTION #2 (ADDITIONAL SPECIFICATION INFO)
<---- END OF USER-SUPPLIED CODE SECTION #2

timeslice_desc - 1

return

end

itittiit*it'iif.*itiiittt*i***ififiititt

Module Compute Routine
titttttittittitittiiitittttiiiiiifiittti
integer function timeslice_compute( invector, slicetime,
$ outscalar, timeo, timel, timez, time3, time4, times. time6,
$ time7)
implicit none
include 'avs/avs.inc'
integer invector
integer outscalar
real slicetime
real timeO
real timel
real time2
real time3
real time4
real times
real time6
real time7

ADD YOUR OWN CODE BETWEEN TEE
FOLLOWING COMMENTS. THE CODE YOU SUPPLY WILL NOT BE OVERWRITTEN.

----> START OF USER-SUPPLIED CODE SECTION #3 (COMPUTE ROUTINE BODY)
integer dims0(3)
integer n.istat
real xa(8)
integer inoffset,outoffset
real inbase(2),outbase(2)
write(0,*)’Here we are in timeslice_compute'

Free old field data
if (outscalar .ne. 0) call AVSfield_free(outscalar)
Allocate space for new field output

write(0,*)’Calling AVSfield_get_dimensions'

C

54
Appendix B

AVS Temporal Interpolation Module (cont)

call AVSfield_get_dimensions(invector,dimsO)

write(0,')'Calling AVSdata_alloc'

outscalar - AVSdata_alloc(

’field 3D 3—space 1-vector rectilinear real',dimsO)

if (outscalar .eq. 0) then
call AVSerror(’Allocation of output field failed.’)
timeslice_compute . 0
return

endif

C Copy the coordinate information from the input field.
C

D

C

write(0,*)'Calling AVSfie1d_copy_points'
istat - AVSfield_copy_points(invector,outscalar)
if(istat.eq.0) then
call AVSerror(
'Unable to copy coordinates from input to output.’)
timeslice_compute - 0
endif

C Now compute the interpolated values.

C

000

$

-xa(1)-time0
xa(2)-time1
xa(3)=time2
xa(4)-time3
xa(5)-time4
xa(6)-time5
xa(7)-time6
xa(8)=time7

write(0,*)'Calling AVSfield_get_int for veclen'
n-AVSfield_get_int(invector,avs_field_veclen)

write(0,*)'Calling AVSfield_data_offset for in'
call AVSfield_data_offset(invector,inbase,ino£fset)

write(0,*)'Calling AVSfield_data_offset for out'
call AVSfield_data_offset(outscalar,outbase,outoffset)

write(0,*)slicetime

call timeslice_computez(inbase(inoffset+1),
outbase(outoffset+1),dimsO(1),dims0(2),dimsOlB),
xa,n,slicetime)

call AVSfield_reset_minmax(outscalar)

<---- END OF USER-SUPPLIED CODE SECTION #3

timeslice_compute - 1
return
end

iii“.ititiffi'iittftttffitifititifftifitiitiiiiififfiifiiiftttfittiiftfitiifitt

Initialization for modules contained in this file.
itiittitit9ttititiittttttttititfiittttttttttfitttitfitttt'tfltitifif'tt

subroutine AVSinit_modules
include 'avs/avs.inc’

55
Appendix B

AVS Temporal Interpolation Module (cont.)

external timeslice_desc

integer timeslice_desc

call AVSmodule_from_desc(timeslice_desc)
end

C ----> START OF USER-SUPPLIED CODE SECTION #4 (SUBROUTINES, FUNCTIONS, UTILITY

subroutine
$ timeslice_computez(indata,outdata,ix,iy,iz,xa,n,slicetime)

implicit none

include 'avs/avs.inc'
real slicetime

integer ix,iy,iz,i,j,k,l,n,istat,percent

real xa(8).ya(8),y,dy

real indataln.ix,iy,iz)

real outdata(ix,iy,iz)

integer nd

nd . 0
write(0,*)'Here we are in timeslice_computeZ'
write(0,*)ix,iy,iz,n,xa,slicetime
do 1003 ksl,iz
do 1002 j-1,iy
do 1001 i=1,ix
do 1004 l-1,n
ya(l)=indata(l,i,j,k)

1004 continue
iflnd.le.100)write(0,')i-1,j-1,k-1,(ya(l),l=1,n)
nd I nd + l ‘
call polint(xa,ya,n,slicetime,y,dy)
outdata(i,j,k)ay

D if(nd.1e.100)write(0,*)i—1,j-1,k-1,y,dy

1001 continue
1002 continue
percent . int((floatlk)/float(iz))*100.0)
D write(0,*)percent
call AVSmodule_status('Timeslice: processing...",percent)
1003 continue
return
end

DUO

CO

C <---- END OF USER-SUPPLIED CODE SECTION #4

S6
1Apmxuudbr<2

AVS File Descriptor Module

sz/usr/local/bin/perl

($min1at, $minlong, Smindepth, Smaxlat, Smaxlong, Smaxdepth) s
(999999, 999999. 999999, 0, 0, 0);

3
# Get a file name
#

itusmmvw1){
print STDERR "csvtospy requires a filename root as its single argument.\n";
print STDERR "If you want to process \"nov92.csv.Z\", type:\n\n";
print STDERR "\tcsvtospy.pl nov92\n\n";
exit;

)

fl

# Now, read the raw CSV file and compute the extrema of latitude, longitude,
# and depth.

#

if(!(@lines a ‘zcat $ARGV[0].csv.Z‘))
print STDERR "Whoops, file \‘SARGV[0].csv.Z\" doesn't exist!\n";
exit;

Scurrentstation - 0;
Snlines . 0;
foreach Sline (Olines) {
(sstation, Sreading, Slatdeg, Slatmin. Slongdeg, Slongmin, Syear, Smonth,
$day, $bottom. Sdepth, $temp, Ssalinity, $deltat) . split(/\s*,\s*/.$line)
# print STDERR "Sstation, $reading, Slatdeg, Slatmin, Slongdeg, Slongmin, Syea
$nlines +- 1;
if($station !- Scurrentstation) {
print STDERR "Station #Sstation\n';
if($currentline)
push(@bottomlines,$currentline);

if($two eq 'Ealse') {
push(0twolines,sprintf("%.2f %.2f %.2f %.2f\n",$long,$lat,—2.0,99.0));

if($fish eq 'justright”)
push(Oeightlines,sprintf('%.21 %.2f %.2£ %.2f\n',$1ong,Slat,0.0,8.0)):

} elsifleish eq 'toohot")
pushioeightlines,sprintf('%.2f %.2f %.2f %.2f\n',$long,$lat,0.0,8.0)l;
pushiOtwentylines,sprintf("%.2f %.2f %.2f %.2f\n",$long,$1at,0.0,20.0)):

Scurrentstation - $station;
$fish . "toohot';
$two - "False";

Slat . (Slatdeg * 60.0) + Slatmin;

Slong a (Slongdeg * 60.0) + Slongmin;

$depth - -$depth;

$temp a $temp / 100.0;

if($lat < $min1at) { $min1at - Slat; }
if($long < $minlong) { $minlong - $long; )
if($depth < $mindepth) { $mindepth . $depth; }
if($lat > Smaxlat) ( Smaxlat s Slat;

if($long > Smaxlongl ( Smaxlong 3 Slang; }

l

57
Appendix C

AVS File Descriptor Module

if($depth > Smaxdepth) { Smaxdepth - $depth;
Scurrentline - sprintf('%.2f %.2f t.2f %.2f\n",$long,$lat,Sdepth,$temp);
if(SdEpCh In -2.0)

push(@twolines,Scurrentline);

$two - "True“;

if(Stemp < 8.0)
if($fish eq "justright") {
push(Oeightlines,Scurrentline):
$fish - "toocold";

}
} elsif($temp < 20.0)
if($fish eq ”toohot'l {
push(0twentylines,Scurrentline);
Sfish - 'justright";

)

open(TWO,*>$ARGV[O].two.spy"):
print TWO Otwolines: -
close(TWO);

Open(EOTTOM,'>$ARGV[0].bottom.spy");
print BOTTOM Obottomlines;
closelEOTTOM);

open(EIGHT,*>$ARGV[0].eight.spy"l:
print EIGHT Oeightlines;
closeiEIGHT);

openiTWENTY,">$ARth01.twenty.spy");
print TWENTY Otwentylines;

close(TWENTY);

exit;

#

9 Now, read the raw CSV file and compute the extrema of latitude, longitude.
8 and depth. ;

#

Snlines - 0:
while($1ine . <>) {

}

(Sstation, Sreading, Slatdeg, Slatmin. Slongdeg, Slongmin, Syear, Smonth.
$day, $bottom, Sdepth, $temp, $salinity, Sdeltat) - split(/\s*,\s*/.$line)

print STDERR 'Sstation, $reading. $latdeg, Slatmin, Slongdeg, Slongmin, Syea

$nlines +- 1;

$1at - (Slatdeg * 60.0) + Slatmin;

Slong . -(($longdeg * 60.0) + Slongmin);

$depth - -$depth;

$temp - $temp / 100.0;

if($lat < $min1at) { $min1at - Slat; }

if($long < $minlong) { $minlong . Slang; }

if($depth < $mindepth) { $mindepth - $depth; }

if($lat > Smaxlat) { Smaxlat - Slat;

if($long > Smaxlong) { Smaxlong - Slong; }

if($depth > Smaxdepth) { Smaxdepth - Sdepth; }

pushlolines,sprintf("%.2f %.2f %.2f t.2f\n',$10ng,$lat,$depth,$temp)l;

print "$nlines\n";

print Olines;

# print STDERR "$min1at, $minlong, $mindepth, Smaxlat, smaxlong, $maxdepth\n";
exit;

58
Appendix D

Ocean Surface Temperatures on the Northeast U.S. Continental Shelf

 

. “5:15" n4...”

-4.- v? j u} E ‘ . \ I. sccru .j

.1416. 3“... S“:

f [a true .
435— \l \J

/ \
nss V‘ ./ \Kj/

42- ' .

CONN 1L . { W . _'
a .
v .

   
  
  
  

ALB 92—03
SPRING GROUNDF‘ISH SURVEY
3 MARCH - 16 APRIL 1992

SEC TEMPERATURE (DEG C)

 

Source: Holzwarth-Davis and Taylor, 1993.

 

59
Appendix D

Ocean Surface Temperatures on the Northeast U.S. Continental Shelf

 

 

 

 

44+-
43.
42-
41 )-
‘0»
39:
38 n .'
K I“ ALB 92-11
o/ o

. 1‘ :'
37%- .f a: FALL GROUNDFISH SURVEY

'> 9 SEPT - 29 OCT 1992
36- .3 .3 src TEMPERATURE (DEG c)

PROFILER AND XBT DATA
q.
35 a.) 9, T
4'
‘0‘...

3‘77 7‘6 73 7‘4 7‘3 7‘2 7‘1 7‘0 6? 6‘9 6‘7 3% as

 

Source: Holzwarth-Davis and Taylor, 1993.

60
Appendix D

Ocean Surface Temperatures on the Northeast U.S. Continental Shelf

 

Jo
.9 2.. V
./‘\r\
43f ( Kl '&
.. $529.59-
42- .. N/s ' ‘

CONN

  

ALB 93-04

   

SPRING GROUNDFISH SURVEY
9 MARCH - 30 APRIL 1993

SFC TEMPERATURE (DEG C)

 

 

7‘3 7‘2 ea 6

Source: Holzwarth-Davis and Taylor, 1994.

66

 

61

Appendix D

Ocean Surface Temperatures on the Northeast U.S. Continental Shelf

*fi

 

 

42-

41-

39-

37~

36

 

 

DEL 93-11
FALL GROUNDFISH SURVEY
7 SEPT - 26 OCT 1993

SEC TEMPERATURE (DEG C)

 

 

3‘; 7'3 72

Source: Holzwanh-Davis and Taylor, 1994.

 

7‘1 7‘0 69 68 6 6

62
Appendix D

Ocean Surface Temperatures on the Northeast U.S. Continental Shelf

45

 

DEL 94-03
SPRING GROUNDFISH SURVEY
1 MARCH - 27 APRIL 1994

SEC TEMPERATURE (DEG C)

 

 

 

 

3477 7‘5 7‘5 7‘4 7‘3 7‘2 7‘1 7‘0 69 6'9 6‘7

Source: Mountain, 1995.

65

Latitude

63
Appendix D

Ocean Surface Temperatures on the Northeast U.S. Continental Shelf

 

 

 

 

 

1 D
i. o
44 '- ‘0 \)
__ \/‘¢\ \2~‘2
’9/ \Z/QV"
“5/ .....
13 ‘ ..... -
42 _ \,_.,//,..,..,« ;
I, M \ \9 .- ,
. "a?” /< ”\5
.— .x I I ,'
’M’ <, \ \>/ A\ '\ ."
79\ 0 4 °' V. _.-, clc
40 r 29 ’,\"'?'\\ , - " ‘ 6"
:‘i’/
38
ALBQ‘OQ
SURFACE TEMPERATURE
36
34 i 1 1 1 1 1 1 1 1 1 1
-77 -75 -73 -71 -69 -67
Longitude

Source: Mountain, 1995.

64

Appendix D

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

 

 

 

 

 

45
441-
43-
42~ ,
41~ i
40' .1
39: ‘.' A :0...
38 i ’ 1 ' 4
\ ' // Q 111219105
37%- .. Ev SPRING GROUNDFISH SURVEY
1 ‘KS—K 6 MARCH - 16 APRIL 1991
1'22
xii; fl BT11 TEMPERATURE (DEG c)
. . 9’?
@771
95'
7‘1 72? Er is 6‘7 6‘6""65

 

3)

Source: Holzwarth-Davis and Taylor, 1992.

65
Appendix D

 

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

 

DEL9110

FALL GROUNDFISI'I SURVEY

10 SEPTEMBER - 24 OCTOBER 1991

BT11 TEMPERATURE (DEG C)

(2 DEGREE CONTOUR INTERVAL)

 

TOWGT

7‘1

 

 

 

45

 

Source: Holzwarth-Davis and Taylor, 1992.

66
Appendix D

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

 

 

f?
.../c

‘w .
«~13 flég‘F-CSJ' /

 

., V9, 1 \_,. \,-

41'-

  

401-

39-

.38
ALE 92-03

37 _ SPRING GROUNDFISH SURVEY

3 MARCH - 16 APRIL 1992

36- BTM TEMPERATURE (DEG C)

35

 

3;

Source: Holzwarth-Davis and Taylor, 1993.

l w
... K. «s-Jfi—fil/ °

01‘

J --—--_-—.-L—. .

. 9—..__....-__O__---._---.

...-.-.. -J .

m).

In

67
Appendix D

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

 

  

ALB 92—11

FALL GROUNDEISR SURVEY

9 SEPT - 28 OCT 1992

BTM TEMPERATURE (DEG c) .

PROPILER AND XBT DATA

 

 

 

Source: Holzwarth-Davis and Taylor, 1993.

68
Appendix D

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

 

 

‘5

441-

43- "

42-
com '4. (s v- .'

‘1» ° e.

39;
38
ALB 93-04

3% SPRING GROUNDI'ISH SURVEY

9 MARCH - 30 APRIL 1993

BTM TEMPERATURE (DEG C)

 

 

 

 

3‘17 7‘6 7‘5 7‘4 7‘3 7‘2 7‘1 fi‘o 6‘9 6‘3 6‘7 6‘6

Source: Holzwarth-Davis and Taylor, 1994.

69
Appendix D

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

45 e -

 

r 1 —T—— V 1' ~

 

DEL 93—11
FALL GROUNDFISH SURVEY
7 SEPT - 26 OCT 1993‘

BTM TEMPERATURE (DEG C)

J1

 

 

 

73 7‘2 7‘1 7‘0 6‘9 6‘? 6f SE

Source: Holzwarth-Davis and Taylor, 1994.

70

Appendix D

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

 

45

44-

43-

42-

41-

4o-

39:

38

37-

36

35

 

DEL 94-03
SPRING GROUNDFISH SURVEY
1 MARCH - 27 APRIL 1994

BTM TEMPERATURE (DEG C)

 

 

 

 

31

Source: Mountain, 1995.

 

7‘1

70 69 68 67 66

65

7 1
Appendix D

-65

 

 

 

Ocean Bottom Temperatures on the Northeast U.S. Continental Shelf

_ r _ c _ _, WK. *3

9409
TOM TEMPERATURE

ALB
BOT

«1'

 

 

0 8

4
663:3

44
42

34

-67

-69

-71

-73

—75

—77

Longitude

Source: Mountain, 1995.

BIBLIOGRAPHY

Bibliography
Advance Visual Systems Inc., 1992, AVS Technical Overview Manual.

Bigelow, Henry R. and William C. Schroeder. 1953. Fishes of the Gulf of Maine.
Washington, DC: U.S. Government Printing Office.

Bivins, Luther E and Harold D. Palmer. 1989. 'Smart Maps' ...A New Look at Ocean
Resources. Sea Technology, November, pp.49-56.

Campbell, Craig S. and Stephen L. Egbert. (1990) Animated Cartography / Thirty Years
of Scratching the Surface. Cartographica, vol. 27, no. 2, pp. 24-46.

Comwell, Bruce and Arthur H. Robinson. 1966. Possibilities for Computer Animated
Films in Cartography. Cartographic Journal, vol. 3, no. 2, pp. 79-82.

Des Roches, Shannon. 1994. The Armchair Traveler Plugs In: Multimedia Cartography
as a Visual Supplement to Travel Writing. Cartographic Perspectives, vol. 19,
pp. 20-25.

DiBiase, David. 1990. Visualization in the Earth Sciences. Earth and Mineral Sciences,
Bulletin of the College of Earth and Mineral Sciences, PSU. Vol. 59, No. 2, pp.
13-18.

DiBiase, David. 1994. Designing Animated Maps for a Multimedia Encyclopedia.
Cartographic Perspectives, no. 19, pp. 3-7.

Dorling, Daniel. 1992; Stretching Space and Splicing Time: From cartographic
Animation to Interactive Visualization. Cartography and Geographic
Information Systems, vol. 19, no.4, pp. 215-227, 267-70.

Gersmehl, Philip J. 1990. Choosing Tools: Nine Metaphors of Four-Dimensional
Cartography. Cartographic Perspectives, no. 5, pp. 3-17.

Hall, Stephen S. 1993. Mapping the Next Millennium. New York, NY: Vintage Books.

Holzwarth-Davis, Tamara and Maureen H. Taylor. 1994. Description of the 1993
Oceanographic Conditions on the Northeast Continental Shelf.
. NOAA/Nh/IFS/NESC: Woods Hole, MA. NEFSC [Northeast fisheries Science
Center] Ref. Doc. 94] l.

Holzwarth-Davis, Tamara and Maureen H. Taylor. 1993. Description of the 1992
Oceanographic Conditions on the Northeast Continental Shelf.
NOAA/NMFS/NEFSC: Woods Hole, MA. NEFSC [Northeast Fisheries Science
Center] Ref. Doc. 93-25.

Holzwarth-Davis, Tamara and Maureen H. Taylor. 1992. Description of the 1991
Oceanographic Conditions on the Northeast Continental Shelf.
NOAAINMFS/NEFSC: Woods Hole, MA. NEFSC [Northeast Fisheries Science
Center] Ref. Doc. 92-08.

72

73

Koussoulakou, A. and M.J. Kraak. 1992. Spatio—temporal Maps and Cartographic
Communication. Cartographic Journal, vol. 29, pp. 101-108.

Langran, Gail and Nicholas Chrisman. 1988. A Framework for Temporal Geographic
Information. Cartographica, vol. 25, no. 3, pp. 1-14.

MacEachren, Alan M. 1994. Visualization in Modern Cartography: Setting the Agenda.
In MacEachren, Alan M. and D. R. Fraser Taylor (eds.). Visualization in Modern
Cartography. Modern Cartography, Volume 2. Oxford, UK: Pergamon/Elsevier
Science Ltd.

MacEachren, Alan M. (in collaboration with Buttenfield, B., J. Campbell, D. DiBiase,
and M. Monmonier) 1992. Visualization. In Abler, Ronald E, Melvin G.
Marcus and Judy M. Olson (eds.). Geography's Inner Worlds. New Brunswick,
NJ: Rutgers University Press.

MacEachren, Alan M. and David DiBiase. 1991. Animated Maps of Aggregate Data:
Conceptual and Practical Problems. Cartography and Geographic Information
Systems, vol. 18, no.4, pp. 221-229.

MacEachren, A. M. and J. H. Ganter. 1990. A Pattern Identification Approach to
Cartographic Visualization. Cartographica, vol. 27, no. 2, pp. 64-81.

MacEachren, Alan M. and Mark Monmonier. 1992. Introduction : What's Special About
Visualization? Cartography and Geographic Information Systems, vol. 19, no. 4,
pp. 197-200.

McCormick, Bruce H., Thomas A DeFanti and Maxine D. Brown (eds.). 1987.
Visualization in Scientific Computing. ACM SIGGRAPH Computer Graphics,
vol. 21, no. 6, pp. 3-14.

Moellering, Harold. 1973. The Automated Mapping of Traffic Accidents. Surveying
and Mapping, vol. 33, no. 4, pp. 467-477.

Moellering, Harold. 1980a. The Real-Time Animation of Three-Dimensional Maps. The
American Cartographer, vol. 7, no. 1, pp. 67-75.

Moellering, Harold. 1980b. Strategies of Real-Time Cartography. The Cartographic
Journal, vol. 17, no. 1, pp. 12-15.

Monmonier, Marie 1992. Summary Graphics for Integrated Visualization in Dynamic
Cartography. Cartography and Geographic Information Systems, vol. 19, no. 1,
pp. 23-36.

Monmonier, Mark 1990. Strategies for the Visualization of Geographic Time-Series
Data. Cartographica, vol. 27, no. 1, pp. 30-45.

Mountain, David. 1995. National Marine Fisheries Service (NMFS), Woods Hole MA.
Personal communication.

74

Peuquet, Donna J. 1994. It's About Time: A conceptual Framework for the
Representation of Temporal Dynamics in Geographic Information Systems.
Annals of the Association of American Geographers, vol. 84, no. 3, pp. 441-61.

Press, William H., et. al. 1989. Numerical Recipes: The Art of Scientific Computing
(FORTRAN version). Cambridge, UIC Cambridge University Press.

Peterson, Michael P. 1994. Cognitive Issues in Cartographic Visualization. In
MacEachren, Alan M. and D. R Fraser Taylor (eds.). Visualization in Modern
Cartography. Modern Cartography, Volume 2. Oxford, UK: Pergamon/Elsevier
Science Ltd.

Philbrick, Allen K. 1953. Toward a Unity of Cartographical Forms and Geographical
Content. The Professional Geographer, vol. 5, no. 5, pp. 1 1-15.

Robertson, Philip K and David A. Abel. 1993. Graphics and Environmental Decision
Making. IEEE Computer Graphics and Applications, MARCH 1993 pp. 25-27.

Slocum, Terry A. 1994. Visualization Software Tools (Introduction). In MacEachren,
Alan M. and D. R Fraser Taylor (eds.). Visualization in Modern Cartography.
Modern Cartography, Volume 2. Oxford, U.K.: Pergamon/Elsevier Science Ltd.

Spyglass, Inc. 1993. Dicer Manual, Version 2.0 for the Macintosh. Third edition.

Taylor, D. R. F. (1991) Geographic Information Systems: the microcomputer and
modern cartography. In Taylor, D. R. F. (ed.), Geographic Information Systems:
the Microcomputer and Modern Cartography. Oxford, U.K.: Pergamon.

Thrower, Norman J. W. 1959. Animated Cartography. Professional Geographer, Vol.
11, no.6, pp. 9-19.

Thrower, Norman J. W. 1961. Animated Cartography in the United States.
International Yearbook of Cartography, vol. 1, pp. 20-28.

Tobler, W.R. 1970. A computer movie simulating urban growth in the Detroit Region.
Economic Geographer vol. 46, no. 2, pp. 234-240.

Tufte, Edward R 1990. Envisioning Information. Cheshire, CT: Graphics Press.

Weber, Christopher R. and Barbara P. Buttenfield. 1993. A Cartographic Animation of
Average Yearly Surface Temperatures for the 48 Contiguous United States: 1897-
1986. Cartography and Geographic Information Systems, vol. 20, no. 3, pp.
141-150.

Weber, Jack. 1993. Visualization : Seeing is Believing. Byte, April 1993.

Wolff, Robert S. and Larry Yaeger. 1993. Visualization of Natural Phenomena. New
York, NY: Springer-Vedag New York, Inc.

Wood, Denis. 1992. The Power of Maps. New York, NY: The Guilford Press.

75

Wood, Micheal. 1994. Visualization in Historical Context. In MacEachren, Alan M.
and D. R. Fraser Taylor (eds.). Visualization in Modern Cartography. Modern
Cartography, Volume 2. Oxford, U.K.: Pergarnon/Elsevier Science Ltd.

76
List of Hardware and Software used in Thesis

Hardware:
Apple® Macintosh Quadra® 700
IBM RS/6000 POWERstation 560

m

AVS“ 5
Advance Visual Systems Inc.
Waltham, Mass.

Macromedia Director® 4.0.3
Macromedia, Inc.
San Francisco, CA.

Spyglass® Dicer® Version 2.0

Spyglass, Inc.,

Champaign, IL

Quick'l‘irne""‘I 2.0 and QuickTime 2.0.2 for Windows"
Apple Computer, Inc.

Cupertino, CA

SPARKLE” by Maynard Handley 1994.

SPARKLE23.sit.hqx
gopher: //gopher.archive.umich.edu:7055/l 1/mac/util/graphicsutil

MovieMakerlé by Jesse Jones 1993.

Internet Resources:

AVS Modules (Write MooV, plot_xyz, trivar)

International AVS Center (IAC) at the North Carolina Supercomputer Center

f tp://avs.ncsc.org

Data Acquisition
National Marine Fisheries Service (NMFS), Woods Hole, MA.
ftp site: f tp.wh.whoi.edu directory: /pub/hydro/