Journal of Spatial Analysis

---

Flood susceptibility mapping using frequency ratio and weight of evidence technique: a case study of Kermanshah Province
 
abstract
Flood is considered as one of the most destructive natural disasters worldwide, because of claiming a large number of lives and incurring extensive damage to the property, disrupting social fabric, paralyzing transportation systems, and threatening natural ecosystems. Flood is one of the most devastating natural disasters causing massive damages to natural and man-made features Flood is a major threet to human life (injure or death of man and animal life), properties (agricultural area, yield production, building and homes) and infrastructures (bridges, roads, railways, urban infrastructures). The damage thet can occur due to such disaster leads to huge economic loss and bring pathogens into urban environments thet causes microbial development and diseases Therefore, the assessment and regionalization of flood disaster risks are becoming increasingly important and urgent. Although it is a very difficult task to prevent floods, we can predict and compensate for the disaster. To predict the probability of a flood, an essential step is to map flood susceptibility.
The methodology of the current research is includes the following steps:
Flood inventory mapping;
Determination of flood-conditioning factors;
Modeling flood susceptibility and its validations.
 Et first , 146 flood locations were identified in the study area. Of these, 102 (70%) points were randomly selected as training data and the remaining 44 points (30%) cases were used for the validation purposes. In the next step 1 flood-conditioning factors were prepared including geology, landuse , distance from river , soil , slope angle, plan curvature, topographic wetness index, Drainage density elevation, rainfall. Then, the probability of the flood occurring for each class of parameters was calculated. Et the end, the obtained weights for each class in the Geographical Information System (GIS) were applied to the corresponding layer and flood risk map of th studied region was prepared. Subsequently, the receiver operating characteristic (ROC) curves were drawn for produced flood susceptibility maps.
To determine the level of correlation between flood locations and conditioning factors, the FR
method was used. The results of spatial relationship between the flood location and the conditioning factors using FR model is shown in Table 2. In general, the FR value of 1 indicates
an average correlation between flood locations and effective factors. If the FR value would be larger than 1, there is a high correlation, and a lower correlation equals to the FR value lower than 1.
The analysis of FR for the relationship between flood location and lithology units indicates thet Cenozoic group has the highest FR value. In the case of land-use, it can be seen thet the residential areas and agriculture land-use have values. One of the most important factors affecting the flood is distance from the river. The results showed thet the class of >500 m FR was the most effective one. The analysis of FR for the relationship between flood location and slope angle indicate thet class 0-6. 1 has the highest FR value. In the case of slope aspect, flood event is most abundant on flet and East facing slopes According to the analysis of FR for the relationship between flood location and plan curvature, flet shape has the highest FR value., A flet shape retains surface run-off for a longer period especially during heavy rainfall . Flood locations are concentrated in areas with a TWI >6. 8 drainage density > 4. 6 km/km2 and altitude classes of 1200 m. In the soil layer, the tallest weight is from the earth with a small transformation of gravel. Finally, the maximum weight is the maximum rainfall.
In this study, all parameters of WofE model were calculated for each conditioning factor. In the lithology unit, the Cenozoic class has the highest flood susceptibility. Among the different land-use types, agriculture categories had the highest values . The distance from the river from 0 to 1000 m indicated positive influence in flooding, while the areas more than 1000 m or far from the river represented the negative correlation with flood occurrence. In the soil layer, clayey soil and tuberous soil had the highest weight. The analysis of WofE for the relationship between flood occurrence and slope angle indicated thet slope angle from 0 to 6. 21 had positive influences in flooding. In the case of slope aspect and plan curvature, flet area had a strong positive correlation with flood occurrence. Effectiveness increases wit increasing TWI classes. The results of drainage density indicate thet areas with higher drainage densities are more susceptible to flood occurrence. By increasing the height of the flooding reduced sensitivity classes. byn flooding rainfall and flood events increased with increasing rainfall.
 
The prediction accuracy and quality of the development model were examined using the area under the curve (AUC). Specifically, the receiver operating characteristic (ROC) curve was used to examine the basis of the assessment is true and false positive rates . So the results showed thet based on the area under the curve, the FR and WofE models show similar results and can be used as a simple tool for verifying the map prepared for flood sensitivity and reducing its future risks.
Floods are the most damaging catastrophic phenomena in the worldwide. Therefore, flood susceptibility mapping is necessary for integrated watershed management in order to have sustainable development. In this study, flood susceptibility zones have been identified using FR and WofE methods. Et first step, a flood inventory map containing 146 flood locations was prepared in the kermanshah Province using documentary sources of Iranian Water Resources Department and field surveys. Then, eleven data layers (lithology, landuse, distance from rivers, soil texture, slope angle, slope aspect, plan curvature, topographic wetness index, drainage density, and altitude) were derived from the spatial database. Using the mentioned conditioning factors, flood susceptibility maps were produced from map index calculated using FR and WofE models, and the results were plotted in ArcGIS. Finally, the AUC-ROC curves using validation dataset were prepared for the two models to test their accuracy. For this reason, of 146 identified flood locations, 102 (70%) cases were used as training data and the remaining 44(30%) was used for validation. The validation of results indicated thet the FR and WofE models had almost similar and reasonable results in the study area. Based on the overall assessments, the proposed approaches in this study were concluded as objective and applicable. The scientific information derived from the present study can assist governments, planners, and engineers to perform proper actions in order to prevent and mitigate the flood occurrence in the future.
 
Key words: Flood susceptibility mapping, validation, method of frequency, weight of evidence, GIS- Kermanshah
 
 

---

Introduction
Land use reflects the interactive characteristics of humans and the environment and describes how human exploitation works for one or more targets on the ground. Land use is usually defined on the basis of human use of the land, with an emphasis on the functional role of land in economic activities. Land use, which is associated with human activity, is undergoing change over time. Land use information and land cover are important for activities such as mapping and land management. Over time, land cover patterns and, consequently, land use change, and the human factor can play a major role in this process. Today, satellite-based measurements with geographic information systems are increasingly being used to identify and analyze land-use change and land cover. With regard to the problems of changes and transformations in the studied area, remote sensing can allow managers to categorize images and evaluate land use changes, in addition to saving time and costs, which allows planners to make plans based on changes, more resources are lost. To be prevented.

Materials & Methods
In order to classify and detect the marginal land of the river, TM and OLI image images were selected for a specific month (August, August) for the years 1987 and 2017. The purpose of this study was to investigate the changes occurring in the studied area with an emphasis on agricultural lands. To do this, the images before processing in the ENVI software took radiometric, atmospheric and geometric corrections on them. After that, the main components of the river route were extracted. Five basic algorithms were used to classify the base pixel, but eCognition software was used to classify the object. Supervised classification identifies homogeneous regions with examples of land use and land cover, in which pixels are assigned in known information classes. Education is a process that determines the criteria for these patterns. Learning output is a set of spectral signatures of proposed classes. The first step in object-oriented classification is the segmentation of the image and the creation of distinct objects, consisting of homogeneous pixels. The main purpose of image segmentation is to combine pixels or small objects to create large image objects based on the spectral and spatial characteristics of the image. In order to evaluate the accuracy and compare the resulting maps, the overall accuracy and Kappa coefficient are used. When the sampling of pixels is done as a spectral or informational class pattern, the evaluation of the spectral reflection of classes and their resolution can also be done. An algorithm with the highest accuracy and accuracy will be the basis for the detection. Detection of changes, which leads to a two-way matrix and shows variations of the main types of land use in the study area, was carried out in this study. Pixel-based cross-tabulation analysis on pixels facilitates the determination of the conversion value from a specific user class to another user category and areas associated with these changes over the given time period.

Results & Discussion
The results showed that the object-oriented method is more accurate than the base pixel algorithms for providing user-defined maps. The amount of accuracy in the method based on object-oriented classification depends largely on choosing the appropriate parameters for classification, defining the rules, and applying the appropriate algorithm to obtain the degree of membership. The Kappa coefficient for each image is approximately 0.90. So these maps are the basis for the discovery of change. According to the results, the agricultural and residential lands have been increased and this increase has been accompanied by a decrease in rangelands. A general overview of this 30-year period shows that the arable and dry farming, respectively, increased by 2418.79 and 719.61 hectares and the rangelands had a decrease of 2848.86 hectares. However, the residential class and human effects show an increase of 428.88 hectares or a growth of 178.87%, which indicates the importance of agriculture in the studied area.

Conclusion
Identifying and discovering land cover changes can help planners and planners identify effective factors in land use change and land cover, and have a useful planning to control them. For this reason, maps are needed with precision and speed, and object-oriented processing methods make this possible with very high precision. The results of this study, in addition to proving the precision and efficiency of object-oriented processing in land cover estimation, between 1987 and 2017, have witnessed a decrease in the area of rangeland lands and, on the other hand, agricultural and residential lands, which is indicative of the overall trend Destruction in the area through the replacement of pastures by other uses such as rainfed farming.

Keywords: Land Use, Gamasiab, Object Oriented, Pixel Base, Kappa Coefficient