Flood Susceptibility Mapping in Gending District by Comparison Frequency Ratio and Weight of Evidence for Mitigation Strategy

the WofE method was 90.4%. The flood susceptibility map illustrates that 14% of the area is in very high and high flood-prone zones, 23% is in the moderate zone, and 63% is in the safe zone. The appropriate mitigation strategy based on the highest FR value is creating drainage networks, and green open spaces, normalizing rivers in residential areas, and implementing selective logging and reforestation programs. The results of this study are used to reduce the impact and risk of future flood disasters.

Gending District is one of the areas in Probolinggo Regency, East Java, which often experiences flooding during the rainy season. One reason is the location of Gending District, which is in the lowlands, so the floods that occur tend to be sent from higher ground. In addition, the Gending sub-district has a gentle slope, so rainwater is difficult to absorb and flows quickly into the rivers around the area. Floods that frequently occur in Gending District have significantly impacted the lives of the local community. Therefore, it is necessary to have flood mitigation to minimize the risks posed.
Flood mitigation can be done with various efforts, such as normalizing drainage networks, building dams, and regulating land use [3]. However, a flood susceptibility map is needed to prepare appropriate mitigation actions. The map can be used to describe places that have the potential to experience flooding. In addition, flood susceptibility maps can be used to determine the location of flood control infrastructure development and identify influencing factors [4].
Flood susceptibility maps can be made using various methods such as frequency ratio [5], [6], analytic hierarchy processes (AHP) [7], [8], logistic regression [9], fuzzy logic [10], weights-of-evidence [11] and decision trees [12]. Matej Vojtek and Jana Vojteková used the AHP method to map Slovakia's flood susceptibility [13]. However, this method has the disadvantage of relying on expert judgment. While the frequency ratio (FR) method can produce a simple flood susceptibility map but provides a high area under curve (AUC) validation value. [14], [15] [16]. Besides that, the weights of evidence (WofE) method can also provide a high AUC validation value [17]. This shows that both methods can produce flood susceptibility maps [14]. The FR method is a statistical method that calculates the ratio between flood events and the factors that influence flood events in an area [18]. In comparison, the basic principle of the WofE method is to calculate the weight of the factors that influence flood events. Each factor related to flooding is given a weight based on the probability of flooding in the area affected by that factor [19].
Choosing the right method is crucial because each method produces different results at different locations. Comparison of the results of different methods in the same location is very helpful in assessing the method's reliability. The accuracy of a method is based on many things, such as the scale of the study area and the factors that cause the flooding used [20]. With an area of 34.74 km 2 in Gending District and using six factors that cause flooding, namely slope, elevation, land use, normalized difference vegetation index (NDVI), curvature, and rainfall, it is necessary to know the appropriate method for making a flood susceptibility map in Gending District. This research aims to make a flood susceptibility map in Gending District by comparing the FR and WofE methods and providing mitigation recommendations. This research will obtain a flood susceptibility map with the highest AUC validation value and an appropriate disaster mitigation strategy. So that it is expected to reduce the impact and risk of flood disasters in the future.

Research Method
This research was conducted in Gending District in Probolinggo Regency. The research data is a Flood Inventory Map, Slope, Elevation, Curvature, Land Use, Rainfall, and NDVI. The data is divided into 70% for training and 30% for validation. Then carried out training on the parameters used and reclassified using the natural break method so that a flood susceptibility map will be obtained, which will then be compared with the Receiver Operating Characteristic (ROC) method.

Data Requirements
In making the flood susceptibility map, 30 x 30 m resolution is used with DEM SRTM data. The derivative of DEM produces slope, elevation, and curvature layers. Rupa Bumi

Frequency Ratio Model
The Frequency Ratio (FR) is a straightforward probabilistic model that is simple to comprehend and implement by calculating the ratio of disaster events to the total area also the ratio of the probability of disaster events to the occurrences of disasters for the given attribute component [21]. This approach to mapping flood susceptibility can be described as the ratio of the research region's total area to the area where a flood susceptibility can occur or as the ratio of the likelihood of an occurrence to the absence of an affair [14], [21]. The FR value is calculated by the equation below (1): Where A is the Number of flood occurrence pixels for each parameter class; B is the Number of flood occurrence pixels in the study area; M is the Number of pixels for each parameter class; and N is the Number of pixels for the total study area. The flood susceptibility index (FFSI) is calculated by adding each weighting factor.

Weights-of-Evidence Model
Based on Bayesian statistical theory, the weights-of-evidence method is a quantitative data-based approach that integrates data sets [22]. According to various research that has thoroughly detailed the formulation of the Weights-of-Evidence technique, The weights in each class for each parameter are obtained according to the occurrence/non-occurrence of floods in an area (2), (3), [23]- [25]. Where P is the probability, B is the influence of the parameters that cause floods, ̅ Is the absence of the influence of the parameters that cause floods, A is the presence of floods, ̅ is the non-existence of floods and W+ also W-, which are the positive and negative weights of the parameters that represent the presence/absence of the influence of each parameter on the flood event. The two weights' differences are defined as contrast (C), used to measure and see the spatial relationship between the slide and the evidence feature.
Then, the final weight can be calculated using the following equation.
Where S(W + ) also S(W -) are the categories of positive and negative weights, also S(C) is the standard deviation of C.

Receiver Operating Characteristic (ROC)
Receiver Operating Characteristic (ROC) is a technique for visualizing, organizing, and classifying several categories determined in a statistical model based on their performance [26]. The ROC graph illustrates the relationship between the True Positive Rate (TPR) and the False Positive Rate (FPR). The relationship between TPR and FPR values is related to each other. If there is an increase in TPR, the FPR will decrease, and vice versa. ROC graphs can produce a diagonal line by determining a random classification called Random Performance [27]. When all classification data includes TPR and FPR, the data can be plotted onto the ROC graph. Each point representing the data from the classification can be connected to become a ROC curve [28]. The model's probability or accuracy level is calculated based on the Area Under Curve (AUC) on the ROC graph. The higher AUC value, the better the model is [29].
The Receiver Operating Characteristics (ROC) method is used in evaluating the accuracy of the results because this method is general and easy to understand [1]. The predicted success rate should be evaluated as a result of each modeling process [30]. This method is the right method to validate flood susceptibility maps for the FR and WofE method [1]. The data needed to validate the flood susceptibility map are flood inventory data for testing (30%) and the flood susceptibility map that has been formed.

Flood Susceptibility Mapping Using Frequency Ratio Model
Frequency ratio values for six conditions factors were derived according to how they related to flooding. There is a significant association between conditioning factors and flood events, as indicated by a larger frequency ratio [31]. Therefore, a relationship is considered strong or weak if its value is greater or lower than 1. The slope has a close relationship with the possibility of flooding. The slope influences the infiltration process, where the higher the slope gradient will increase surface runoff and slow down the infiltration process. As a result, water will stagnate and become flooded in areas with low slope gradients [14]. The slope parameter that has the highest FR value is in class 0 -1,57 with an FR value of 1,19 also, the lowest FR value is in class 9,19 -19,04 with an FR value of 0,00.
The elevation is a significant component that influences flooding. The frequency ratio typically decreases as elevation rises, meaning that the influence of floods is most and least affected by the lowest and highest elevation values, respectively [14]. This is shown in the FR calculation results for the elevation parameters above. The highest values are in class 5.07 -13.06, with an FR value of 1.65.
Land use is the most influential parameter in flood events. Land use in urban areas increases surface runoff because the soil is impervious to water. At the same time, agricultural lands increase surface runoff because no ground cover vegetation can control and also prevent the rapid flow of water to the ground surface. Urban areas and agricultural land are at risk of flooding also erosion, so these areas are the most prone to flooding [14]. This is indicated by the analysis results in which the settlement and ricefields have the highest FR values of 5,28, and 0,22, respectively.
A negative value for the NDVI implies water, and a positive value for vegetation. That means the NDVI value ranges from -1 to +1. There is less chance of flooding the higher the NDVI class rating. Conversely, the likelihood of flooding increases as the NDVI class value decreases [14]. with distance. This is shown in the FR calculation results for the rainfall parameter above, the 23 -32 highest value is in the lowest class with an FR value of 1,94, and the value will decrease as the parameter class increases.
After preparing the six factors that cause flooding and giving weight to each parameter using the FR method, a flood susceptibility map is obtained by adding all the factors. The value of flood susceptibility is divided into five sub-classes, which are deficient, low, medium, high, and very high. This class is obtained from the result of reclassification using the natural breaks method in ArcGIS.

Flood Susceptibility Mapping Using Weights-of-Evidence Model
Like the FR method, the weights-of-evidence technique establishes the connection between flood episodes and the flood triggering elements listed in Table 4.   negatively influence the incidence of flooding. The land use factor that influences the incidence of flooding is the settlement class, with a weight value of 14,54. In other classes, such as plantation forests, dry land farming, rice fields, and ponds, there is no significant relationship to the incidence of flooding and is marked with a negative weight value.
In the NDVI factor, in class 0,09 -0, 21      The flood susceptibility map illustrates that areas with a very high level of flood susceptibility are mostly located in areas with low slopes, flat curvature, and land use in the form of settlements.

Validation of Flood Susceptibility Maps
Map validation is important in identifying susceptible areas to determine their quality, so model results cannot be applied without validation. To validate the two flood vulnerability maps that are formed, a prediction level curve is used based on the location of the flood and each flood susceptibility map. The area under the curve predicted rate indicates how well the model predicts flooding [29]. Figure 4 shows the results of the prediction rate curve.  Based on these results, the FR method, which has the largest area under the curve and the highest accuracy (92.8%) compared to the WofE method, is the most precise way for creating flood susceptibility maps in the Gending District.

Flood Mitigation Strategy
Gending District is an area susceptible to flooding. To deal with floods that regularly occur every year, flood mitigation is carried out based on two main factors that cause flood susceptibility, namely land use and NDVI, which is based on the highest FR value. normalization activities can also be carried out on the Gending River to restore the width and depth of the river so that the river can still accommodate the overflow of water when the rain intensity is high enough.
2. NDVI analysis needs to be monitored before the rainy season comes to increase the greenness index. Higher NDVI and dense vegetation reduce and slow water flow. Dense vegetation gives time for water to seep into the ground so that the volume of water is reduced and the possibility of flooding is smaller. Implementation of selective logging and reforestation programs can inhibit surface water flow and optimize infiltration in the upstream area, resulting in decreased runoff and reduced erosion rates [32].

Conclusion
The proper method for making a flood susceptibility map in Gending District is FR, with an AUC value of 92.8%, while the WofE method is 90.4%. The flood susceptibility map illustrates that 14% of the area is in very high and high flood-prone zones, 23% is in the moderate zone, and 63% is in the safe zone. The appropriate mitigation strategy based on the highest FR value is to create drainage networks, green open spaces, and normalize rivers in residential areas. In addition, the implementation of selective logging and reforestation programs needs to be carried out to increase the NDVI value. The results of this study are used to reduce the impact and risk of future flood disasters.

Acknowledgement
The researcher would like to thank and support the University of Jember, particularly the Civil Engineering Department and Faculty of Engineering, for giving them a chance to conduct research and write reports so that they could develop expertise in applied science.