Page 722 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

1) Distance factors: distance to residential areas, distance to existing agricultural lands, distance to existing bare lands,
distance to trees and parks, distance to roads, distance to urban exit points, distance to urban centers have been
considered as the distance factors in the modeling process.
2) Boolean Factors: Includes unchangeable areas and Boolean maps for each user. These factors were used as the modeling
controllers.
3) Topographic factors: including height, slope, and slope direction that reflect the overall shape of the land in the study
area.
Finally, according to the results of Cramer's V test, 9 factors were considered as effective factors in changing each user
shown in Figures 3, 4 and 5.

Fig. 3. (a)Distance to Boston, (b)Distance to development space, (c)Distance to developed open space

Fig. 4. (d) Distance to ponds, (e)Distance to roads, (f)Distance to streams

Fig. 5. (g)Massachusetts towns, (h) Slope, (i)DEM

4.3. Results and validation
As mentioned earlier, the main purpose of this paper was to predict land-use/land cover change with the method of
selecting the best factors among the available factors. According to the initial analysis of the maps from 1996 to 2006,
a significant point is the loss of a huge volume of deciduous forests, considered here as a driving force and sub-models
formed based on this. In the first step, in the logistic regression training phase, by using the land cover maps in 1996
and 2006 as a bulletin, the effective coefficient of each factor (Figures 3, 4, and 5) in changing the land cover classes
was determined. The Cramer's V test for each factor is shown in Table (2) and the transition potential map for land
cover class is shown in Figure 6 (b). Then, by using these coefficients and the 2006 factor maps and the implementation
of logical regression forward, the 2016 land cover map (Figure 7 (a)) was produced.

6