Flood vulnerability analysis using the frequency ratio method with the watershed ecosystem in Bulukumba Regency, South Sulawesi Indonesia

Bulukumba Regency is located in the Province of South Sulawesi. It is one of the main tourist destinations and phinisi boat industry that provides much economic added value for the community and local government. Apart from these advantages, the problem of flooding is an obstacle and has a detrimental impact on the regional economy. Drainage problems, weather anomalies, and land function experts in the upstream area are factors in the occurrence of floods in Bulukumba Regency. This study used remote sensing and geographic information systems (GIS) combined with the Frequency Ratio (F.R.) method to create flood vulnerability maps. The parameters in this study are based on literature studies, data availability, and research site conditions such as rainfall, earth curvature, river distance, marbles, altitude, topographic wetness index (TWI), stream power index (SPI), soil texture, soil permeability, geology, and land use/land cover (LULC). The results of the identification of flood events obtained a total area of flood events, namely 6,345 ha, which was identified with the highest F.R. value in the closure of pond land and coastal sand beds, sand soil texture, and lithology, namely alluvium rocks. Validation was obtained for the success rate with a value of 0.895 and the prediction rate with a value of 0.887. It shows the weight that falls into the good category. The area of insecurity has a high of 7.20% and high of 1.69%.


Introduction
One of the earliest and most noticeable changes detected due to recent global warming is the increase in extreme weather events.Weather patterns have become more varied in the last half-century, with rainfall more frequent and intense [1,2].The climate is a crucial factor influencing environmental systems, socioeconomic conditions, and the availability of water resources [3,4].Due to varying weather patterns, the increase in hydrometeorological disasters, namely floods in the Asia Pacific region, especially in Indonesia, is very vulnerable.Flooding in Indonesia is a significant and complex problem [5].Flood 1230 (2023) 012044 IOP Publishing doi:10.1088/1755-1315/1230/1/012044 2 overflows can occur throughout the rainy season, and flooding can last more than three days.It causes enormous losses, with a population of about 640,000 each year.The National Disaster Management Agency (2022) said that 1,932 flood events, or 42.1% of the disaster events, occurred throughout 2021 [6].
Flooding is the accumulation of water that overflows beyond the normal limits of a river or other body of water in an ordinarily non-submerged area and part of an extreme weather event [7,8].Floods are affected by various characteristics of the climate system.It is mainly from rainfall patterns (intensity, duration, amount, time, phase-rain or snow) and also temperature patterns [9] which endangers the lives of living populations, especially in watercourses and their surroundings, as well as less marbled areas [10].Flooding is also influenced by the conditions of drainage basins, such as the water level of rivers, snow and permafrost, soil characteristics (permeability, soil moisture, and its distribution), urbanization, and the presence of embankments, dams, and reservoirs.If close to sea level, river flooding will coincide with storm surges or extreme tidal events [11].
The case study is in Bulukumba Regency, with flooding problems that often occur every year.Bulukumba Regency is located in South Sulawesi Province, Indonesia.It is one of the main tourist destinations and phinisi boat industry that provides much economic added value for the community and local government.Apart from these advantages, the problem of flooding is an obstacle and has a detrimental impact on the regional economy.Flooding occurs after heavy rains and causes damage mainly in urban areas [10,12].Rural and urban drainage problems, weather anomalies, and forest land function experts becoming massive agricultural areas in the upstream area are factors in the occurrence of flood disasters in Bulukumba Regency.The exploitation of watershed ecosystems in the upstream area plays a role in environmental damage and causes disasters [13].Sugianto (2022) said that it would cause changes in water systems as a response to nature and result in the interaction of nature and humans, so it has been the problem that the water system has to use watershed units [14].Mitigating flood damage can be done by mapping flood vulnerability with a watershed ecosystem approach.
Flood management consists of four stages: prediction, preparation, prevention, and damage assessment [15].Many researchers have conducted studies on mapping flood vulnerability using Remote Sensing and Geographic Information Systems [16].This remote sensing method is quite interesting in mapping flood vulnerability due to the availability of data periodically and the low cost of observing the earth's surface from various spatial resolutions [17,18].This study combines the Frequency Ratio (F.R.) method with Remote Sensing and Geographic Information Systems.The F.R. method is relatively new to flood vulnerability analysis, and it is widely used in analyzing other natural disasters, such as landslide vulnerability mapping [19][20][21].Several researchers used this method in mapping flood vulnerability with research on Shariri and Roslee (2022) in Kota Belud, Sabah, Malaysia, validating 89% model success and 82% prediction of flood insecurity [22].The results of research by Saha et al. (2021) in Raiganj City, East India, validated 91% model susceptibility and 86% prediction of flood insecurity [23].Tehrany et al. (2017) revealed that applying the F.R. method in mapping flood vulnerability showed that this method is very efficient for flood vulnerability modeling [24].They state that this new methodology is effective in recognizing flooded areas.Therefore, this study aimed to determine flood vulnerability in Bulukumba Regency using the F.R. method with a watershed ecosystem approach.

Study area
Bulukumba Regency is located in the southern part of Sulawesi, about 153 km from Makassar City, South Sulawesi, Indonesia.Moreover, geographically located at coordinates 5°20' -5°40' South and 119°58' -120°28' East.The Administrative Area of Bulukumba Regency consists of 10 districts with a total area of 1,167.53km 2 .The topographical state of this area is dominated by flat marbles and ramps in the downstream area and steep upstream.The soil type in Bulukumba Regency consists of alfisol, entisol, Inceptisol, and ultisol types.The average monthly rainfall varies with the lowest values in August, with values of 13.9 mm, and the highest in December and January, with values of 287.8 mm and 265.7 mm-the rainfall intensity in this area is <13.61 mm/day.The location of the study area can be seen in Figure 1.The watershed approach was used in this study.The watershed as the basis for the study area in Bulukumba Regency refers to the hydrological process in the watershed ecosystem.The imbalance of water using in the watershed ecosystem, both the infiltration process, precipitation, and hydrological processes, results in disasters, one of which is flooding.Changes in watershed landscapes, such as urbanization, logging, and agriculture, have caused rainfall runoff to become major floods [25].This research uses the watershed approach to assess flood events in Bulukumba Regency by viewing it as a unified ecosystem.The collection of information on the history of flood events is compiled to determine the location of the flood distribution.This information significantly affects the accuracy of mapping flood vulnerability later [26], as seen in Table 1.The floods that hit some areas in Bulukumba Regency caused hundreds of heads of families (K.K.) to feel the impact.There are even 25 households or 138 people who have to be evacuated to the office hall of the Bulukumba Regional Development Planning Agency (Bappeda), Jalan Kenari

2021
Ujung Bulu, Rilau Ale, Kindang, Bulukumpa and Gantarang Five housing units were heavily damaged, and another 600 were affected.Four bridges were cut off, and one other was destroyed.In addition, there was 57 cattle-type livestock dragged by the river current.They were soaking 10 ha of rice fields and 40 ha of gardens on the banks of the Bijawang River.Soaking 60 ha of rice fields and 30 ha of the park on the banks of the Kirasa river.
There is also an overflow on the Balantieng river

Material and methods
In this study, there are three main stages, namely data preparation, data analysis, and validation which can be seen in detail in Figure 2. can evaluate the possibility of future flood events [27].A record of the timing of flood events is available in Table 1.The guide to the area of distribution of flood events using Sentinel-1B and Sentinel-1A SAR images was obtained from ESA (European Space Agency), which can be downloaded on the http://scihub.copernicus.eu/website.Spaceborne Synthetic Aperture Radar (SAR) Systems are an effective tool in flood monitoring because it has near real-time (NRT) properties [28].There are two main steps in obtaining flood event inventory data, namely: Where σ0 is sigma naught, m is the master (before the flood), and s is the enslaved people (after the flood).The formula is applied using SNAP software with a band maths toolbox.After the process, flood data management is continued in ArcGIS software to carry out a vectorization and generalization process to obtain flood distribution data.In some vulnerability model analyses, the data is divided into training and testing/validation data in point form [29,30].This training data is used as data processing on the weighting of the Frequency Ratio value in each factor that affects flood events with a portion of 70% (446,409 pixels) and testing data of 30% (191,186 pixels) [31,32].The distribution of flood point distribution can be seen in Figure 3.

Flooding causative factors.
Step to create a flood vulnerability map, determining factors causing floods is associated with regional characteristics, the availability of appropriate data, and literature studies to obtain the best results [33][34][35].The causative factors that are the parameters in this study include: 3.1.2.1.Rainfall.Frequent and extreme rainfall is one of the leading causes of flood disasters [36].Rainfall data in the Bulukumba Regency area, obtained through rainfall data from the PERSIANN-CCS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks -Cloud Classification System) satellite product from CHRSthe University of Arizona available from 2003 with a spatial resolution of 0.04 o and a temporal resolution of 1-hour [37].
3.1.2.2.Curvature.This curvature value represents the morphological shape of the topography.In this case, the earth's curvature is associated with the condition of inundation after heavy rains.Curvature in marbled areas contains more water and holds water from high rainfall longer [38,39].Making maps using tools from ArcGIS, namely Contour.This map has 3 (three) classes: concave, flat, and convex.

Distance from the river.
Areas along the River are very vulnerable to flooding both under normal circumstances and flash floods due to the movement of water flowing from the highlands and then accumulating in low areas in the watershed [40].In making river distance maps, the data used is the National DEM and is carried out in ArcGIS using the Hydrology tool.This process will produce river nets, and further management is carried out in the Euclidean Distance tool to create distances from river nets with classes of 0-20 m, 20-40 m, 40-60 m, 60-80 m, 80-100 m and >100 m.
3.1.2.6.Soil texture.Soil texture is the essential part of the soil that considerably impacts flooding [41] because it affects water distribution on the soil surface.The soil's structure and texture determine water retention and infiltration [44].

Stream power index (SPI)
. SPI is based on the value of the strength of the erosion rate on surface runoff within the watershed [45].With the increase in catchment and Slope of the Slope, the amount of water contributed by the slope area and the speed of the water flow increase, so the flow strength index and the risk of erosion increase.It affects flood events that occur in areas affected by height and Slope.

Topographic wetness index (TWI)
. TWI is a method of quantification of topographic control of hydrological processes.The TWI value describes the tendency of water accumulation on a slope based on the gravitational force that controls water flow [46].The spatial distribution of hydrological conditions can be mapped using this method.TWI can effectively identify flood-prone areas by mapping areas that experience inundation [47].TWI assessment was implemented using DEM in the form of a Digital Terrain Model (DTM).

Geology (Lithology).
Geology is related to the strength of rock materials because the composition of lithology and structure varies for different types of stones and resistance to propulsive movements depending on the strength of the rocks [21,48].Lithological maps are obtained from data extraction results from geological maps issued by the Indonesian Geological Research and Development Center.There are 5 (five) classes on this map, namely Baturape-Cindako Volcanic Rocks (Tpbl), Camba Formation (Tmc), Parasitic Eruption Results (Qlvp), Basal and Basalt Hacks (b1), and Aluminum Deposits (Qac).

Land use/land cover (LULC).
Land use/land cover (LULC) affects the infiltration process, evapotranspiration, and the magnitude of surface flow [49].Vegetation affects the volume of water entering rivers and lakes into the soil and groundwater reserves.The type of land use/cover will affect vegetation, and this will also affect the infiltration rate and the magnitude of the surface flow, which will cause the extent of the flood [50].In this study, the interpretation of Sentinel 2A imagery to obtain data on land use/land cover.The data that has been collected is converted into raster data which will be used as input for the analysis of flood vulnerability maps with a pixel resolution of 10 m x 10 m.A spatial overview of each factor is presented in Figure 4.

Data analysis
Data analysis is used with quantitative methods, namely the F.R. method, widely applied to various disaster vulnerabilities such as landslides, forest fires, and groundwater potential [21,51,52].FR value for each causal factor is calculated as the area of the landslide occurrence; and the ratio in each class on each factor to the total size of the study area [53].If the balance is more significant than 1.0, the relationship between the flood event and the causative factor is higher.If the ratio is less than 1.0, the relationship between the flood event and the causative factor is low.The ratio value in each class shows the degree of relationship of the frequency ratio value calculated by the formula: Fr is the frequency ratio value in each class at a specific factor.PxcL is the number of pixels with flooding in class n of the parameter m (nm), Pixel is the number of pixels in class n of parameter m (nm), ΣPnxL is the total Pixel of parameter m, and ΣPnx is the whole Pixel of the Area.
Step to create a Flooding Susceptibility Index (LSI), all causal factors are mapped in the form of a raster map of F.R values and then summed using the formula: FSI is the Flooding Susceptibility Index, and Fr1, Fr2, and Frn are the raster frequency ratio map for each flood factor.

Validation
The results of the flood vulnerability analysis were then validated using flood data obtained by dividing 70% of the training data for model success and 30% of the validation data for the model prediction rate.This validation will show how well the model predicts flooding.The results of this validation show the value of prediction accuracy based on AUC (Area Under Curve) through analysis (Receiver Operating Characteristic) using SPSS software.The AUC method's vulnerable values ranging from 0.5 to 1.0, assess the model's accuracy [54].The success rate value of the AUC curve was obtained from training data, and predictions were calculated from the validation of ROC data with values close to 1.0 and those close to 0.5, indicating inaccuracies in the model [55].

Frequency ratio calculation
In disaster vulnerability analysis, independent variables, namely causal factors, play a specific role in presenting disaster vulnerability maps [31,56,57].Table 2 shows the correlation of the relationship between flood events and the classes of each factor causing flooding.This relationship will result in a spatial comparison of flood-prone areas with their causal factors.The lowest correlation is indicated by the F.R. value of <1 and the highest correlation of >1.This study shows the highest value in the closing/land use factor, namely in the pond class (8.47).Then on this factor, successively, the value of >1 was shown in the class of coastal sand beds (6.46), rice fields (3.05), and secondary mangrove forests (2.07).The following is the overall data of the calculation of the F.R. value in each factor in Table 2. Flooding is one of the ecological disturbances in the watershed due to a large amount of rainfall that the soil cannot absorb.Vegetation helps increase soil capacity to reduce surface runoff, such as forests, grasslands, shrubs, and food crops on agricultural land [58].From the frequency ratio analysis results, the closure of secondary mangrove forest land, ponds, stretches of beach sand, and rice fields has an F.R. value of >1, meaning a high probability of flood events.Land closure in coastal areas (those in mangrove forests, ponds, and stretches of beach sand) is vulnerable to inundation by tidal floods in Bulukumba Regency.It is due to several things, such as mangrove forests that function as catchment areas and embankments/natural barriers from disaster threats [59], which have suffered damage due to human activities in excessive use, such as clearing pond land and logging mangroves for coastal 1230 (2023) 012044 IOP Publishing doi:10.1088/1755-1315/1230/1/01204411 communities.In addition to coastal areas, paddy fields have a high probability of flooding.The paddy field area provides a relatively slow level of soil permeability.In contrast, coarse-fractional soil with low organic matter content has a low permeability value, so the puddle will flow slowly [60].In addition, paddy fields and ponds are at high risk of flooding due to their location not being too far from the coast.
The texture of the soil has an indirect effect on flooding.In the soil texture factor, soil conditions with sand texture (6.45) showed probabilities in influencing the occurrence of floods relatively high and the surface of dusty loam soils (1.57).Texture and moisture content is essential in soil skin determination [40].In the study area, the texture of sand soil has a high probability of flooding and is spread in the coastal area of Bulukumba Regency and dusty clay texture.Soils with a texture consisting of sand and dust will tend to form a rusty and rough soil surface to reduce the infiltration rate [35].In addition, soil infiltration capacity affects the state of soil texture due to a decrease in infiltration capacity so that there is an increase in surface runoff [61].
The highest rainfall F.R. values are in the class 1,390-1,696.08mm/year and the class 1,696.08-1,846.77mm/year, with the distribution being in the downstream and urban areas of Bulukumba Regency.In addition, the distribution of floods comes from upstream regions with very high rain levels.Increased rainfall is the main factor that triggers flooding [62,63].Heavy rainfall over a long period trigger flooding in the downstream area, derived from an increase in surface runoff that leads to a rise in the surface of the river flow and an increase in sediment in the downstream area [64][65][66].
The curvature is classified into Concave, Flat, and Convex.The calculation results show that the Flat class has a higher probability value for flooding than other classes, with an F.R. value of 1.38.In the case of curvature, flooding occurs in areas that have a flat curvature.It is because flat terrain is the most suitable area for flooding.The marbled hills pushed the water flow into the lower, flat areas, which caused flooding [24].
Slopes in the flat class become a class with a high probability of flooding.The Slope of the Slope in the flat class is spread downstream.The effect of this Slope is a controlling factor in surface runoff where low-gradient slopes are particularly susceptible to flooding [41,51].In addition to slopes, altitude is another topographical factor that indirectly affects flooding.Altitude is connected to high rainfall increasing river discharge downstream [67].This condition allows the downstream area to become a place for water accumulation and sites that often occur in floods [68].
The flood vulnerability lithology factor is one of the conditions that must be considered because it directly affects the permeability and soil runoff.The analysis results show that the type of alluvium sedimentary rock, namely the Qac class (6.77), and the brection stone, namely the Qlvb class (1.35), is highly probable for flooding in Bulukumba Regency.The characteristics of alluvium rocks are relatively capable of holding water formed due to sedimentary deposits that affect flood events [69,70].
Distance from the River is one of the essential factors in determining the level of flood vulnerability [71].Flood intensity decreases in locations far from rivers and is at higher risk in areas near riverbanks [49].The results showed the highest effect of flooding was a distance of 0-20 meters with a value of 1.31.and it is affected until the distance of 80-100 meters from the river.The river's water level will increase with high rainfall, so flooding will cause water overflow to the area closest to the riverbank [24].
TWI is a form of hydrological modeling in mountainous and hilly terrain that can be used to identify flood-prone areas [71].The highest TWI is recorded in the medium, high, and very high classes.A significantly very high TWI value indicates a high chance of flooding in the area.It is supported by rock factors where the nature of brection it difficult to absorb water into the soil (infiltration), so the tendency to flooding will also increase.TWI predicts water accumulation at each point and shows a movement to gravitational forces in water displacement and infiltration influenced by soil characteristics such as the influence of water pressure by soil pores and permeability [72].
The SPI demonstrated river erosion strength and slope stability in the study area [71].The F.R. analysis above shows the impact of SPI on flood events.The highest ratio values are indicated on the classification of very weak, weak, and medium classes.The stability of the Slope is affected by the SPI, which directly reflects the erosive force of a current.The relationship between geomorphic factors and floods is claimed to be influenced by the strength of flow power in a catchment area [73].

Validate model
Validation is carried out with ROC (Receiver Operating Characteristic) analysis which produces predictive accuracy values based on AUC (Area Under Curve).Running a model and using the model's results must be validated for accuracy in testing the model's reliability.For support model predictions, the area Under the Curve (AUC) is considered with an AUC value of 0.5 or less, so the model is unsuitable for flood hazard mapping.If the value is perfect, this model is most suitable for estimating flood trends.If the AUC value is > 0.80, flood predictability is well-received [49].Based on the results of the ROC analysis, there are two validation results, namely validation of the model's success rate from the F.R. value and validation to determine the model's prediction level for flood events.The curve of validation results obtained from SPSS software with ROC analysis in Bulukumba Regency can be seen in Figure 5. From the results obtained, it is known that the validation level of success obtained has a value of 0.895, and the validation of the prediction level shows a value of 0.887.It offers the value that falls into the good category.AUC values between 0.8 -0.9 indicate the model's ability to be entirely accurate and appropriate [57], so the relationship between flood events and factors causing floods correlates with each other in forming vulnerability areas.The level of flood vulnerability is obtained from all raster maps of the factors driving the flood with frequency values.This Flood Susceptibility Index (FSI) value is classified based on natural breaks into five classes of vulnerability, which can be seen in Table 3.Based on the data in Table 3, the level of flood insecurity includes very low, low, medium, high, and very high grades.The high flood insecurity in this area is influenced by 10 (ten) factors that have a relationship or correlation to the occurrence of floods.Reclassify Index value shows the summation of F.R. values on all factors used.Then flood insecurity can be seen in Figure 6 through visualization and spatial distribution of the vulnerability class.

Conclusion
Through the results of the F.R., the disclosure of the factors of flooding in Bulukumba Regency can be known.This model is highly efficient in flood hazard management with the use of remote sensing with the development of disaster early warning systems.It is also proven to evaluate the number of watersheds in Bulukumba Regency with the worst levels in the Bijawang, Bialo, Ujung Loe, Lo He, Raowa, Aparang, and Kirasa watersheds.The level of flood insecurity is divided into 5 (five) classes: very high, high, moderate, low, and very low classes concentrated in the downstream area.Assessment of F.R. results using the ROC curve by looking at the AUC and showing the best results on the built model (success rate of 0.895 and prediction rate of 0.887).

Figure 2 . 4 IOP
Figure 2. Flow chart of the work.

Figure 3 .
Figure 3. Distribution of flood event points.

Figure 5 .
Figure 5. AUC of the flood vulnerability ROC using the F.R. method: (a) Success rate accuracy and (b) Prediction level accuracy.

Table 1 .
History of flood events in Bulukumba Regency.

Table 2 .
Results of calculation of F.R. values on each factor causing flooding.