Académique Documents
Professionnel Documents
Culture Documents
Un Automatisme Cellular
aplicat al creixement urb.
Conceptes Bsics
Robert Colombo
June 6th , 2005
What is all about?
2
What Does SLEUTH mean?
Slope, (Pendent)
Land use, (s del sol)
Exclusion, (Exclusio)
Urban extent, (Extensio urbana)
Transportation, (Transportation)
Hill shade, ( Sombrejat)
3
Who created the Model?
The model has been applied in Cities around the world but basically
in USA. Organizations such as NASA use this software for their
predictions
4
A CA system usually consists of four basic elements:
Cells.
Set of states (occuppied/free, on/off,).
Neighborhoods defined by cells .
Rules.
Time periods.
5
Cells are the smallest units which must manifest some
adjacency or proximity .
6
The state of a cell can change according to transition
rules which are defined in terms of neighborhood
functions (Xia &Gar 1998 ).
7
Neighborhoods can be configured in several ways :
5 x 5 Moore
3 x 3 Moore
8
Each time step in a model could be called Cycle or
Iteration or Generation.
9
The SLEUTH model
10
Every Growing Cycle in SLEUTH is based in :
11
Growing Cycle in SLEUTH
12
Growth Rules:
A set of decisions, or growth, rules applied to the data
to simulate urban driven land cover change.
The Exclude land and Slope input layers limit the
areas suitable for urbanization
Spontaneous Growth
New Spreading Centers
Edge Growth
Road-Influenced Growth
13
Spontaneous Growth:
14
New Spreading Center Growth:
15
Edge Growth:
if there is a urban cell that has at least three free
neighbor ready to urbanize cells, it has a probability
(controlled by the Spread coefficient and the slope
coefficient) to become urban.
(t+1)
16
Road-Influenced Growth:
If a road is found in a radius (road gravity coeff) of
a new urbanized cell (Breed coef), a temporary
urbanized cell is placed in the point where the road
network is close to the urban cell.
This temporary cell randomly moves across the
urban network (moves defined by Dispersion
coeff). The final location of this new urban cell then
acts as a new urban spreading center. Possible in
more than one cell at the time
17
Self modification rules:
The coefficient values in a simulation do not have to
remain static. Due to rapid or to low growth rates
the coefficients may be increased or decreased by
the system.
The limits Critical high and Critical Low have
effect in three of the growth control parameters:
dispersion, breed, and spread, increasing or
decreasing the values.
Dispersion
Breed
Spread
Slope resistance
Road gravity
19
Dispersion coefficient:
20
Breed coefficient:
21
Spread coefficient:
22
Slope coefficient:
23
Road gravity coefficient:
24
Coefficient Selection:
25
Coefficient Selection:
Example of UNIX terminal message while Example of statistical results obtained in SLEUTH in original format (txt file, and
computing Sleuth. treated in Office-like program. Brute force method means All possible
combination will be tried, even if they are non realistic
Lee Salee Weighted sum
Best Run Diffusion Breed Spread Slope Road Gravity Best Run Diffusion Breed Spread Slope Road Gravity
Corase 90 1 1 75 75 1 298 1 50 25 100 100
Fine 341 1 11 70 70 40 893 1 41 25 90 70
Final 550 1 22 61 66 64 1317 6 10 25 78 58
Different results obtained using 2 different Coefficient value selection
26
Coefficient Selection:
file name created in mode flag dependent
LOG_# T, C, P yes
avg.log T, C, P yes
coeff.log T, C, P yes
control_stats.log T, C no T=Test
restart_file.data# T, C, P no C=Calibration
std_dev.log T, C, P yes P=Prediction
Metrics available in the control_stats.log statistical files, used for coefficient selection and for image information.
27
Model Inputs
Images:
The input images for SLEUTH need to follow some
requirements to be analyzed in the model:
b c d
e f
Three sets of urban layers 1983, 1987 1992 and 1997 100*100 b)200*200 c)400*400 all of them 30
reduced to the same size (not used in SLEUTH)
Excluded Hillshade
Roads Slope
Input data layers used in the Atlanta metro area. The colors are exaggerated. Note the binary code. 31
Excluded Hillshade
Roads Slope
Original Layers with original color codes.
32
Model Application
Scenario Files:
The Scenario files are
regular text files that
contain a specific
commands, basically
coefficient ranges, image
layer paths, color legend
patterns, output image
file flags (numeric or
alphanumeric choosing
values), and all the basic
information for a SLEUTH
model .
33
SLEUTH simulation Scheme:
Model inputs
1 Test phase
Error
Correct
2 Calibration phase
Coarse Calibration
Fine Calibration
Final Calibration
Deriving Calibration
Prediction 3
phase
Error
Correct
34
END
Model Phases and modes:
Test:
The phase was created as a way to generate
a set of historical data simulations, for a single
coefficient set without requiring a Start or Stop
coefficients to be set. Its now use, as the
name shows to run simple simulations to test
the images
35
Model Phases and modes:
Calibration:
The user will have to set the coefficients and
refine the selection for them several times to
acquire finally some coefficient values valid for
prediction mode, using Brute force methods.
36
Model Phases and modes:
Calibration:
Coarse: the range of the coefficient values to
derive is set to the maximum values (0-100)
using large increments (typical steps of 25). The
resolution of the data is normally of the total
resolution .
Calibration:
Final: using the best fit values from the
control_stats.log in the fine calibration, the
range of coefficients will be narrowed again.
38
Model Phases and modes:
Prediction:
This phase will lead to predict how an urban
area will grow with the coefficients derived
from the Calibration process. All the previous
phases are just processed to be able to get
the coefficient values for this final step
39
Model Phases and modes:
Coefficients obtained from calibrations for Project Atlanta
Test Coarse Fine Final Derive
Image Resolution 400 400 200 100 100
Seed 1 1 1 1 1
Monte Carlo Iterations 1 5 7 9 100
Diffusion Start 1 0 0 1 1
Diffusion Step 1 25 5 2 1
Diffusion Stop 1 100 25 10 1
Breed Start 1 0 25 1 1
Breed Step 1 25 10 5 1
Breed Stop 1 100 75 25 1
Spread Start 1 0 50 55 63
Spread Step 1 25 5 4 1
Spread Stop 1 100 75 70 63
Slope Resistance Start 1 0 25 55 55
Slope Resistance Step 1 25 15 4 1
Slope Resistance Stop 1 100 100 70 55
Road Gravity Start 1 0 25 25 40
Road Gravity Step 1 25 15 15 1
Road Gravity Stop 1 100 100 100 40
Time 0.5h 18h 25h 40h 0.25h
Example of coefficient result, time of computation and data used in every phase for the 1st
Colombo simulation of the Atlanta MA.
40
SLEUTH simulation Scheme:
Model inputs
Test phase
Error
Correct
Calibration phase
Coarse Calibration
Fine Calibration
Final Calibration
Deriving Calibration
Prediction
phase
Error
Correct
END 41
Results:
42
Results:
43
Results:
Output image representing, prediction of urban growth in 2015 for Atlanta Metro. A succession 45
of images creating an animation is also available.
New Urban pixels
1,600,000
1,400,000
1,200,000
1,000,000
800,000
3rdColombo coef
600,000 1st Colombo coeff.
Samples coeff
Yang&Lo coeff
400,000
200,000
46
Questions and Future job to develop:
Xiaojun Yang, (2000) Integrating image analysis and dynamic spatial modeling
with GIS in a rapidly suburbanizing environment, Thesis (Ph. D.)--University of
Georgia, 2000. Dir. C.P. Lo
Yang, X., and Lo, C.P. 2003. Modeling urban growth and landscape change in
the Atlanta metropolitan area. International Journal of Geographical Information
Science 17:463488.
Torrens, P M & O'Sullivan, D, 2001, "Cellular automata and urban simulation:
where do we go from here?", Environment and Planning B 28 163-168
Dietzel, Ch. and Clarke K. (2004)Determination of Optimal Calibration Metrics
through the use of Self-Organizing Maps. Presentation for the Integrated
assessment of the land system: The future of land useInstitute for Environmental
Studies, Amsterdam October 30th, 2004
Clarke K. and Dietze Ch.l (2005)A Decade of SLEUTHing: Lessons Learned
from Applications of a Cellular Automaton Land Use Change Model, Department of
Geography, University of California,Santa Barbara, USA
Benenson, I. & Torrens, P.M. (2004). "Geosimulation: object-based modeling of
urban phenomena". Ed Wiley. 49
Web sites:
http://www.ncgia.ucsb.edu/projects/gig/project_gig.htm
http://www.mirwoj.opus.chelm.pl/ca/ca_links.html
www.geosimulation.com
www.sprawlsim.org
Gracies!!
Robert.colombo@upc.edu
50