The urban sprawl dynamics: does a neural network understand the spatial logic better than a cellular automata?



It should be observed that, from a statistical point of view, the results are quite good for
the residential use, a little less for the industry and not so good for the commerce. The
difference is probably due to the different sample size for the three land uses. Since the
recent urbanisation process in the south of Milan concerns mainly residential sprawl,
many records are “good examples” for this land use. On the contrary the commercial
use, which is the less frequent, gives the worst results. This is shown on the scatter
diagram of observed (on the
x axis) and calculated values (on the y axis) for each land
use (figures 6 a, b and c).

The spatial representation of “errors” allows to evaluate the spatial logic of the SANNs
output.

Figure 7 shows the errors concerning the residential land use. Errors are measured in
ratio over the whole cell surface. One large underestimation is evident, just in the centre




(a)                                                         (b)

Figure 6 - The scatter diagrams of observed and calculated values of land use in each cell

of an agricultural area totally not urbanised and not infrastructured. This is due to an
entirely new settlement for affluent people, “Milano 3”, which is the result of a
negotiation between big investors and the local municipality. Evidently it was
impossible for the SANN to predict an event which is totally extraneous to his logic.

Other errors are mainly due to planning constraints, often forbidding a “natural” growth
and forcing the development elsewhere. As mentioned earlier in this experiment road
infrastructures have been ignored, whereas their topical role emerges in the already
mentioned SOM classification.

13



More intriguing information

1. Non-causality in Bivariate Binary Panel Data
2. The name is absent
3. Structural Conservation Practices in U.S. Corn Production: Evidence on Environmental Stewardship by Program Participants and Non-Participants
4. The name is absent
5. LOCAL PROGRAMS AND ACTIVITIES TO HELP FARM PEOPLE ADJUST
6. IMPROVING THE UNIVERSITY'S PERFORMANCE IN PUBLIC POLICY EDUCATION
7. Evidence of coevolution in multi-objective evolutionary algorithms
8. The Dynamic Cost of the Draft
9. Alzheimer’s Disease and Herpes Simplex Encephalitis
10. The name is absent
11. Incorporating global skills within UK higher education of engineers
12. Quality Enhancement for E-Learning Courses: The Role of Student Feedback
13. Who runs the IFIs?
14. Fiscal Policy Rules in Practice
15. The name is absent
16. Private tutoring at transition points in the English education system: its nature, extent and purpose
17. On the Integration of Digital Technologies into Mathematics Classrooms
18. Behavior-Based Early Language Development on a Humanoid Robot
19. The Clustering of Financial Services in London*
20. Reconsidering the value of pupil attitudes to studying post-16: a caution for Paul Croll