Spatio-temporal differences in model outputs and parameter space as determined by calibration extent

1
Spatio-temporal differences in model outputs and
parameter space as determined by calibration
extent

• 7th International Conference on Geocomputation
• Charles K. Dietzel
• Department of Geography
• University of California Santa Barbara

2
Acknowledgements

• Public Policy Institute of California
• Dr. Keith C. Clarke, Advisor, UCSB
• USGS Rocky Mountain Mapping Center
• Mark Feller (Geographic Analysis and Monitoring
  Program)

3
Presentation Outline

• Introduction
• Methods
• Results
• Conclusions
• Broader impacts

4
Introduction

• Increased computational power of computers
• More geospatial data sources
• Dependence on both of these to better urban and
  land use models
• Do outside areas influence the focal study area?
• To what degree?
• SLEUTH urban growth model (Clarke et al, 1997)

5
Methods

• SLEUTH (Slope, Land use, Exclusion, Urban,
  Transportation, Hillshade)
• Cellular Automata
• Rigorously tested and widely used
• Santa Barbara (Candau and Goldstein, 2002)
• San Francisco (Clarke et al, 1997)
• New York (Esnard and Yang, 2002)
• Washington-Baltimore (Clarke and Gaydos, 1998)
• Lisbon and Porto (Silva and Clarke, 2002)
• Brute force calibration of historical data

6
Methods

• Five parameters of calibration
• Diffusion
• Breed
• Spread
• Slope resistance
• Road gravity

7
Methods

• Calibrated San Joaquin Valley dataset at three
  extents
• San Joaquin County (local)
• Bay Area Influenced (regional)
• Entire San Joaquin Valley (global)

8
Results
9
Results
10
Results
11
Conclusions

• Calibration at different extents shows that
  local, regional, and global growth cannot be
  generally characterized
• Relative to regional and global growth, San
  Joaquin County is growing faster
• This method may be used to explore growth
  dynamics in any system

12
Broader Impacts

• Areas outside the focal study area can change the
  parameter space, and hence the forecasting, of
  models
• Inclusion of outside areas can characterize local
  growth relative to a larger scale
• Increased computing power warrants modelers
  including possibly influential areas in future
  efforts
• Modelers should spend more time considering the
  geographic extent of their models

13
References

• Candau, J., Goldstein, N.C. (2002). Multiple
  scenario urban forecasting for the California
  South Coast region. Proceedings of the 40th
  Urban Regional Information Systems Association,
  Chicago, IL, October 26-30, 495-506.
• Clarke, K.C., Gaydos, L. (1998). Loose-coupling a
  cellular automaton model and GIS long-term urban
  growth prediction for San Francisco and
  Washington/Baltimore. International Journal of
  Geographic Information Science, 12 (7) 699-714.
• Clarke, K.C., Hoppen, S., Gaydos, L. (1997). A
  self-modifying cellular automata model of
  historical urbanization in the San Francisco Bay
  area. Environment and Planning B Planning and
  Design, 24 247-261.
• Esnard, A.M., Yang, Y. (2002). Descriptive and
  Comparative Studies of 1990 Urban Extent Data for
  the New York Metropolitan Region. URISA Journal,
  14 (1) 57-62.
• Silva, E. A., Clarke, K.C. (2002). Calibration of
  the SLEUTH urban growth model for Lisbon and
  Porto, Portugal. Computers, Environment and Urban
  Systems 26 525-552.