Ch 3 - Spatial data Modelling

advertisement
Ch 3 - Spatial Data Modelling
Address spatial form vs spatial processes
 Form = structure & distribution of features (now)
 Process = how features interact (ch 7)
So here we deal with "models of spatial form" (hence modelling title)
Construction of model - a series of stages or steps of "data abstraction"
Steps cause us to move from reality to somehow representing reality in the
computer…..interesting "backwards approach"…deconvolve the real world
and make it fit into a computer.
3 steps they focus upon:
1. Conceptualize your model - identify real features and determine how
(pts, lines, polygons) to represent them..(example in book is that at their
scale they decide to use a point to represent a hotel, not a polygon or
square)
2. Represent the conceptual model with a spatial model - select the
system you'll use (points, lines, or polygons in vector; a cell or series of
cells in raster)
3. Store the spatial data model by using a spatial data structure that codes
raster cells or defines pts and line segments
Fig 3.1 represents the process, authors note at each step how you get
further away from reality
A. Step 1 - identify the features / entities of interest and choose the mode of
representation - 5 types
Pts , lines, polygons… and networks and surfaces Fig 3.2
Problems can be encountered in this simplifying step for several
reasons
 World is dynamic, ever-changing
 Sometimes tough to ID discrete vs continuous surfaces
 Scale may be not appropriate for your efforts
Use the example of trying to describe a forest…collection of points? Or is it
a polygon? And how might the entity change through time?
Another issue crops up in placing boundary between forest and non-forest.
This makes us wonder about continuous vs discrete mapping.
Bottom line: Don't minimize the importance of the decision concerning entity
type.
Ch 3 p.2
B. Develop a spatial data model
Begin by re-emphasizing how good the human eye & brain are at
recognizing shapes and forms, but we are also reminded that "computers
require unambiguous instructions".
The 2 basic ways to create spatial data models are with raster and vector
systems. Fig 3.4
Raster
Basic building block is the cell. All pts line polygons networks and
surfaces are represented by cells. Cell dimensions very important to the
look of your map….Fig 3.5
Vector
Basic building block is the point. Pair of points makes a line, more
points can make a polygon. Cartesian coord system (x,y pairs) provides
locations. Number of points used to make lines and polygons very
important to the look of your map. Fig 3.5
C. Create a spatial data structure
Defined: "structures provide the information required for computer to
reconstruct the spatial data model in digital form"
"many diff data structures used in GIS". This helps explain why it's
tough sometimes to move from one GIS to another GIS…
analogy - Microsoft Word vs Microsoft Works - one format can't read the
other!
 Raster data structures
Fig 3.6 - most common method for "coding" raster data structures is to
assign a numeric value to a cell that corresponds to a code you develop to
describe a "layer" of data
Thus you need 5 coded data "layers" to reproduce the 5 different spatial
entities shown in Fig 3.4
Fig 3.7 shows good example of coding land values to represent:
1. residential area
2. forest
3. farmland
Ch 3. P. 3
A BIG problem with raster data is the amount of storage space required,
especially when you reduce the cell size to increase resolution.
16x
4x
original
There are compaction techniques available…we won't deal with these here.
 Vector data structures
Fig 3.10 shows the concept, where the points are given ID numbers, and
polygons can be defined with those points as corners.
The storage is fine with small amts of data, but with large amounts of
polygons it gets unwieldy. Thus the system keeps track of shared
boundaries, using "topology".
Topology is the science of describing connectedness of features. They use
a food analogy (spaghetti) to describe how unconnected polygons can
appear to be connected if they overlay each other on the computer screen,
but are not tied together internally in the data structure.


Start with a point. All it needs for reference is an x,y coordinate ….
A line is an ordered set of points known by various names (arc, segment,
chain) that has a beginning point and ending point (nodes). Line
direction is automatically known because you know beginning and end.
Ch 3 p.4

An area (polygon) is defined using the knowledge developed inputting
pts and lines, with a couple guiding principles:
 Nodes & line segments NOT duplicated
 All polygons have unique identifiers
 Islands and holes can be represented
Fig 3.11 shows variety of polygons, how they're linked with each other
topologically
Qualifier at end of section - many diff data structures exist in GIS - these are
but a few examples……
D.




Other aspects of modelling to be considered in this chapter:
Surfaces
Networks
The third dimension
The 4th dimension
a. modelling surfaces
remember how we view a surface…it's a 3D surface that we represent in
2D…like a topo map is the 2D representation of a 3D surface of a mountain.
But the surface can be a map of anything, like rainfall per year in inches, or
pollution at a given site in milligrams per liter.
Book uses the example of DTM (digital terrain modeling) as a way to
descirbe a surface - in this case it IS terrain topography (but the principles
apply to any surface you're trying to map…)
Raster and vector systems handle this problem differently, so we need to
view the 2 approaches.
 Raster - pretty simple - put a height value in for each cell, this represents
the height at that point. Good resolution if cell size is a foot or two, not
so good if cell size is 100' x 100'.
0 5
10
5
0
5
10 15
10
5
0
5
5
0
10
3D view:
Ch 3 p.5

Vector - a bit more complex than raster. Simplest version is equallyspaced grid of point heights (Fig 3-12a), but more common and accurate
version is the TIN - triangulated irregular network. Because 3 pts ( a
triangle) define a plane, triangles of varying shapes and sizes are used
to model a surface as a collection of different planes (Fig 3-12b).
b. Modelling networks
Network defined: "set of interconnected linear features through which
materials, goods, and people are transported, or along which
communication is achieved"
Here, the vector system works much better than raster.
(Fig 3-13) shows vector components to a network…you can see how the
flow directions can be modeled, as well as junctions, beginnings, and ends.
What can networks represent? Many things…
-
Transportation (road, rail, air)
Utilities (power lines, pipelines, cable TV, etc)
Hydrogeology too (rivers, streams)
Network characteristics can be stored as attribute info for both points and
lines in a database (eg, one-way street going North).
2 key characteristics of networks that can be defined or modelled are:
1. impedance (like resistance in an electrical circuit) Fig 3.14
2. supply and demand
1. Impedance can be viewed as "cost" - delivery van example where costs
include time of travel, fuel cost, and driver cost in terms of pay.
 Factors dictating the actual impedance value could be things like:
- Traffic volume
- Number of one-way streets
- Topography (uphill more costly than downhill)
- Weather (like wind speed, or ice will slow things down…)
2. Supply and demand - supply refers to amount of resource available to
the network, while demand refers to amount that the network consumes.
 Example used is a local hospital with fixed number of beds serving a
community….supply is the number of beds, demand is the No. of
people wanting to use the beds.
Ch 3 p.6
Good example of the importance (or lack) of correct
representation…London subway (New York for that matter)…subway
maps themselves are generalized, probably not to scale in terms of
distance, but has the important connecting and station information that
allows a novice user to get around…..
E. "Building computer worlds" - we use objects and layers, along with the
data structures, to build the computer worlds that represent reality.

Layered approach in Fig 3.15. note that you have multiple data layers
(shapefiles or themes in Arcview parlance), but you don't have to bring
all of them into your view.
Each layer is "thematic", representing a theme. We show a feature of the
same area (land use, soil type, hydrogeology, etc.) with each layer. This
reminds us that these features overlap and co-exist in the same space.

A newer approach is the object-oriented approach. This entails grouping
of objects, not layering Fig 3.16. The OO approach tries to keep
relationships more "normal", but still tries to describe characteristics and
relationships between objects.
2 ways to classify objects in this scheme: put the object in it's real world
class, or put the object in it's computer world class. Example in book is a
polygon is known as a "car park" in the real world, and an "area" in the
computer world.
F. Modelling the 3rd dimension
Tough to do, but GIS improving. Arcview has a 3D analyst extension that
can allow for 3D perspective views. Examples of 3D view in Plate 1, 2
(p.160).
G. Modelling the 4th dimension
Typical way to model right now is to simply make layers that represent
snapshots at one instant in time, like the single frame of a cartoon or a
movie camera.
Download