Evaluation of the Prediction Capability of AHP and F_AHP Methods in Flood Susceptibility Mapping of Ernakulam District (India)

7 Floods are one of the frequent natural hazards occurring in Kerala because of the remarkably high annual rate of 8 rainfall. The objective of this study is to prepare the flood susceptibility maps of the Ernakulam district by 9 integrating remote sensing data, GIS, and analytical hierarchy process (AHP), and fuzzy-analytical hierarchy 10 process methods. Factors such as slope angle, soil types (texture), land use/land cover, stream density, water ratio 11 index, normalized difference built-up index, topographic wetness index, stream power index, aspect, sediment 12 transport index have been selected. The area of the final maps is grouped into five flood susceptible zones, ranging 13 from very low to very high. The major reasons for flood occurrence in Ernakulam district are the combined effect 14 of multiple factors such as excess silting, reduction of stream width due to human intervention, and changes in 15 land cover and land use pattern, lower slope, higher soil moisture content, lower stream capacity, and poor 16 infiltration capacity of soils. The prepared map was validated using the receiver operating characteristic (ROC) 17 curve method. The area under the ROC curve (AUC) values of 0.75 and 0.81 estimated by the ROC curve method 18 for the AHP and F-AHP methods is considered acceptable and excellent, which confirms the prediction capability 19 of the prepared maps. The very high susceptible zone constitutes around 19% of the district. This map is useful 20 for land-use planners and policymakers to adopt strategies which will reduce the impact of flood hazard and 21 damage in the future.


Introduction 24
Flooding is one of the natural hazards that often cause significant damage to property and loss of life (Merkuryeva 25 et al. 2015). This condition can arise from diverse hydrological processes, such as high tide levels, precipitation, 26 high groundwater levels, and high river flows (Acreman and Holden 2013). Fluvial floods can be defined as the 27 overflowing of streams or other water bodies of accumulated water over areas that are not normally inundated 28 with water (Pratomo et al. 2016). The frequency of flooding is expected to increase due to unscientific 50 Hepatitis, and Leptospirosis (Jobin and Prakash 2020). Leptospirosis has long been a major threat to Kerala with 51 more than 1000 cases reported annually (James et al. 2018). By providing the public with accurate information 52 on flood risk through flood susceptibility maps, the damages and losses can be minimized.

53
Remote sensing (RS) and GIS have made significant contributions in disaster management studies such as those

Data source 81
The

99
The slope of the study area was derived from the SRTM DEM using ArcGIS spatial analyst tools. The slope angle 100 of the Ernakulam district has been classified into five classes including, 0 -5.00, 5.00 -11.88, 11.88 -21.26,    1998). However, higher stream density need not imply a higher rate of runoff. This is because the stream capacity 131 depends on the width, depth, and length of drainage channels. In the study area, the distributaries of the 132 mainstream are narrow, tortuous, and shallow. Therefore, water retention leads to flooding. In addition to this, the 133 existing narrow stream channels in the lower parts of the study area are partially blocked due to heavy silting.

134
This explains the frequent flooding of the lower stretches of the Ernakulam district. The stream networks were 135 digitized from the SoI topographic maps and the stream density layer was prepared using ArcGIS spatial analyst 136 tools. The stream density of the study area is grouped into five classes ( Figure 5). They are 0-1.47 km/km 2 , 1.47-137 3.17 km/km 2 , 3.17-6.02 km/km 2 , 6.02-12.78 km/km 2 , and 12.78-26.94 km/km 2 .

144
The chance of flooding is high in areas with higher WRI.

146
NDBI is a satellite-derived index that represents urban built-up areas (Bhatti and Tripathi 2014). The NDBI value 147 close to 0 represents woodland, the NDBI value less than 0 represents a body of water, and the NDBI value greater

191
Where α is the area of the catchment (m 2 ) and β (radians) is the slope gradient.

192
The STI of the study area ranges from 0 to 247.77 and is grouped into 5 classes (Figure 11). The chance of flooding 193 will be high in areas with low SPI values, as these as depositional zones. The carrying capacity of stream channels 194 in these zones will be much reduced due to the deposition of sediments. 195

The AHP modelling 196
AHP is the most used decision-making method developed by Saaty (1980) to solve complex decision problems.

197
By reducing complicated decisions to a number of pairwise comparisons, AHP helps to make the right decision

200
Eigen vector and weighting coefficient (Table 1), and finally, calculation of consistency ratio to check the 201 consistency (Table 2).

215
Where RI is the random index 216   (Table 5), relative fuzzy weights of each factor (Table 6), and averaged and normalized 230 relative weights of factor (Table 7). The various steps involved with the F-AHP modelling are as follows:

235
The pair-wise contribution matrix is depicted in Eq. 12. Where k ij d % indicates the k th decision maker's preference of i th factor over j th factor (Ayhan 2013).
Step 2: When there is more than one decision-maker, the preferences (

241
Step 3: The pair-wise comparison matrix is modified based on the averaged preferences using Eq. 14.

247
Where i r% still depicts the triangular values.

248
Step 5: From the next 3 sub-steps, the fuzzy weight of each factor was computed.

249
Step 5a: The vector summation of each i r% was determined.

250
Step 5b: The (-1) power of the summation vector was computed, and the fuzzy triangular number was replaced, 251 to convert it into increasing order.

253
Step 5c: To compute the fuzzy weight of factors () ii iw % , each i r% was multiplied with the reverse vector as in Eq.

Results and discussion 274
The map layers of slope angle, soil, LULC, stream density, WRI, NDBI, TWI, SPI, aspect, and STI were combined 275 using ArcGIS tools to prepare the flood susceptibility maps of the study area. The flood susceptibility maps were 276 prepared using the weights derived by the AHP and F-AHP methods. The area of the flood susceptibility maps 277 has been grouped into five classes, namely very low, low, moderate, high, and very high (Figure 13 and 14). The   confirm that the result is acceptable and excellent for the AHP and F-AHP modelling, respectively ( Figure 15).

292
This finding confirms that the F-AHP method is more effective in predicting flood-prone zones and was thus 293 chosen as the best model. According to the F-AHP model, the very high susceptible zone covers 19.37% of the 294 study area.

310
Conflict of Interest / Competing interests -The authors declare that they have no known competing financial 311 interests or personal relationships that could have appeared to influence the work reported in this paper.

312
Availability of data and materials -Not applicable.