*5.3.1. Fuzzy approach and weighting*

**5.1. Object-based classification**

214 Sustainable Urbanization

can be found trying different scale factors [2].

**5.2. Accuracy assessment of LUC classifications**

used for the accuracy assessment of 2002 and 2015 LUC.

LUC, LUA, distance from roads, and distance from built-up areas.

**5.3. GIS-based MCE process**

Many complex land covers exhibit similar spectral characteristics, making separation in feature space by simple per-pixel classifiers difficult, leading to inaccurate classification. Therefore, an object-based classification is a potential solution for the classification of such regions. The specific benefits are an increase in accuracy, a decrease in classification time, and that it helps to eliminate within-field spectral mixing [25]. Basically, there are three steps in object-based classification: segmentation, classification, and per field integration. An image is divided into segments dependent on pixel spectral similarities, structure of the image, and surface texture characteristics. This progress is up to variables such as scaling factor, smoothness versus compactness, and shape factors. Each segment contained a group of pixels, and scaling factor is defined as the minimum pixel counts that have similar spectral characteristics in a segment. Compactness and smoothness are important for creating pixel groups. Shape factor deals with the boundary of a segment. Scale factor is variable according to the study scale and ideal scale

The accuracy of the LUC classifications is tested by applying error matrix and *κ* statistic. The error matrix approach is the most widely used in accuracy assessment [26]. After the generation of an error matrix, other important accuracy assessment elements, such as overall accuracy, user accuracy, producer accuracy, and *κ* coefficient, can be derived. *κ* is the difference between the observed accuracy and the chance agreement divided by 1 minus that chance agreement [27]. Ground truth data collected from field trip and old topographical and forest maps are

The general information on this process is discussed in Section 1. However, criteria selection, standardization, weighting, and mapping progress and methods are introduced in urban sprawl suitability extent in this study. Factors are defined based on similar literature, data accessibility, and regional variations. Seven factors are defined: elevation, slope, hillshade,

Weighted linear combination (WLC) is one of the most used MCE techniques to define suitability degrees for continuous factors. With WLC, factors are combined by applying a weight to each followed by a summation of the results to produce a suitability map [28]:

where *S* is suitability, *Wi* is the weight of factor *i*, and *Xi* is the criteria score of factor *i*.

*S WiXi* <sup>=</sup> å , (1)

Nothing is exact in nature. Fuzzy mentality is based on this simple rule. In a landscape, boundaries are flexible and natural characteristics of a land do not change suddenly, such as slope and soil texture. A fuzzy set can define a factor with transition zones according to the suitability degree of the factor [29].

Factors are of different data ranges and units. Weighting a factor linearly needs a standard input data set to define the priority degree of the factors. Standardization process is applied to the factors separately. However, ideal and non-ideal values of the each factor have to be defined primarily. In this study, ideal regions for urban sprawl are defined based on urban development between 2002 and 2015. Changed regions are accepted ideal areas for urban sprawl. The ideal and non-ideal values of each factor are defined considering the existing urban growth. It does not mean that the existing urban growth is ideal for sustainable development, but, finally, Urban was developed to these regions and the spatial characteristics of the changed areas are good indicators for the future urban sprawl. All factors are reclassified based on ideal intervals, which are defined according to data frequencies, and ideal data ranges are found for fuzzy standardization.

The weights of each factor are obtained using ideal data diversity in the categories of the factors. For example, slope is reclassified to five intervals: very high, high, medium, low, and very low. If urban change is located in all areas homogeneously, it means that the slope is not important for urban sprawl; if urban change is monitored heterogeneously, it means that the slope is important because only one or two slope ranges are suitable for urban growth. The standard deviation (SD) of the each factor diversity is shown as data homogeneity with a standard result. If the SD value is high, weight should be high. The weights of each factor are defined according to the SD values.

### *5.3.2. Scenario development*

Three scenarios were evaluated in the study: economic, ecologic and sustainable. In the economic scenario, only water bodies, existing city areas, security areas, and historical protection areas were analyzed and defined as restrictive regions. There was not any limitation on LUC or LUA. Therefore, the trend of the last 13 years was considered to define urban suitability in the economic scenario. Ecologically important areas, such as wetlands, were ignored.

In the ecologic scenario, important areas, such as wetlands (bulrush) and highly available lands for agriculture and grasslands, were protected strictly from urban sprawl.

In the sustainable scenario, urbanization demands were considered using urban change in time. According to urban change results, agricultural and wetland areas were protected partly in addition to other restrictive areas, so that a balanced urban spread was provided for human life and nature in the future.
