THE “MARINE GIS” — DYNAMIC GIS IN ACTION 
Christopher Gold (1), Michael Chau (2), Marcin Dzieszko (1), Rafel Goralski (1) 
(1) Dept. Land Surveying and Geo-InformaticsHong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong SAR, PRC. 
(2)Hong Kong Marine Department, Pak Sha Wan, Hong Kong SAR, PRC. 
Christophergold@Voronoi.com 
KEY WORDS: GIS, Oceanography, Navigation, Algorithms, Data Structures, DEM/DTM, Graphics 
ABSTRACT: 
The sea moves: the land usually stays still. It is not surprising that the underlying structure of a land-based GIS is rarely appropriate 
for marine applications. Add the third spatial dimension and it is clear that an attempt to simulate the sea requires a major overhaul 
of the appropriate algorithms and data structures. 
It has seemed obvious to us for some time that spatial data structures need to adapt locally to change, and that the Voronoi diagram 
provides a conceptually simple framework for which dynamic and kinetic algorithms may be developed. The opportunity to work on 
real marine data for the Hong Kong area provided the incentive to put our ideas into practice. The challenge was to produce a 
dynamic three dimensional equivalent to the classical “Pilot Book”, which contains the rules for navigation in the proximity of 
individual harbours. 
While we have done some work on true dynamic three dimensional data structures, as required for marine profiling, the Pilot Book 
application could be achieved with a kinetic two dimensional structure, but in several layers. The terrain (above and below the sea 
surface) was modelled with the dual Delaunay triangulation, and the coastline at any particular tidal time was captured by its 
intersection with the current local sea level. This, together with the locations of individual ships and other surface features, was used 
to form a two dimensional dynamic Voronoi diagram at the sea surface for proximity and collision detection. Other layers were used 
to indicate fairways, marine markers, submarine contours, etc. 
However, in order to provide a realistic simulation, we needed to take concepts (and models) from 3D games development and 
provide marine markers such as lighthouses and buoys, and simulate fog and darkness. We also needed to provide a variety of 
camera views: overhead and on board a selected ship — a deck view, above and behind, below and behind. This required an 
appropriate scene graph structure to manage the scales, objects, lights and cameras in order to give us the flexibility required for 
realistic simulation. The result, while still requiring work (especially on ship navigation) may provide a feasible replacement for the 
Pilot Book, especially for practice simulations. 
1. INTRODUCTION tides, and potential obstacles. Therefore we need to look at 
dynamic (or, more properly, kinetic) data structures to represent 
The title “Marine GIS” gives us two contradictory thoughts. On our features and their spatial relations. 
the one hand, a “GIS” refers to a land-based static 
representation of a two-dimensional surface (maybe with hills). 
On the other hand, “Marine” implies three dimensional, 
dynamic representation and analysis. So far, we have used the 
first as an approximation of the second. This paper attempts to 
go a little further. 
While we are also working on fully three-dimensional 
volumetric modelling, for such things as changes in salinity 
with depth, this paper is primarily an extension of the ideas of 
Gold and Condal (1995), which worked with one or more two- 
dimensional surfaces simultaneously. They suggested the 
simple Voronoi diagram/Delaunay triangulation for the 
bathymetric model, and the kinetic moving-point Voronoi 
diagram for the sea-surface navigation and collision-detection 
layer. We would like to report progress on that approach. 
Paper maps are two dimensional and even with modem 
technology, so are computer screens (stereo systems, although 
available, still appear awkward). Thus visualization issues 
become critical for three-dimensional models: either we work 
with a volume representation (usually voxels) that are either In brief, we started with the ideas of Gold and Condal, but felt 
sliced or transparent in places, or else we work with a surface 
representation (more familiar to most of us) — again with issues 
of visibility and transparency. This is the mind-set of 3D 
modelling and games, and many fundamental techniques have 
been developed in recent years. 
that the visualization issues mentioned previously had to take 
priority. Thus our first step was a well-designed graphical 
interface, and a well-designed object-oriented spatial data 
model. Of particular concern were the issues of flexible data 
structure design, and of a user navigation system within the map 
space that was intuitively easy to understand. We believe that 
we have achieved these objectives, which then made the 
implementation of the spatial models relatively simple. We 
found our work greatly simplified by the use of OpenGL (Woo 
et al., 1999) for the visual display (especially with the use of 
Thus a 3D GIS should be dynamic, which is entirely consistent 
with the marine imperative. However, in the marine case the 
argument for full dynamism of features as well as observers 
becomes compelling. Ships move, as do fish, pollution plumes, 
688 
Interna 
transpa 
of its € 
prograt 
(Guiba 
relatior 
primal 
In rece 
using a 
the firs 
Davis 
dimens 
represe 
Lockw: 
give a: 
land-ba 
objects 
over tii 
are mo: 
at sea) 
(data ai 
others). 
should 
major [ 
built fo 
Structur 
adjacen 
by com 
extend - 
databas 
real pro 
topolog 
Marine 
objects 
an obje 
now wi 
just an « 
Most ( 
architec 
databas: 
cannot 
these pr 
that car 
same da 
and Gol 
tempora 
or freel 
diagram 
Condal 
GIS wh 
the *tilii 
entirely 
perform 
traditior 
Goodch 
publishe 
to the n: 
dynamis