https://doi.org/10.20965/jdr.2021.p0329 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

Paper: Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

Kaito Kotone∗, Kenji Taniguchi∗∗,†, Koichi Nakamura∗∗∗, and Yuki Takayama∗∗

∗Division of Environmental Design, Kanazawa University Kakuma-Machi, Kanazawa, Ishikawa 920-1192, ∗∗Faculty of Geosciences and Civil Engineering, Kanazawa University, Ishikawa, Japan †Corresponding author, E-mail: [email protected] ∗∗∗Nihonkai Consultant Co., Ltd., Ishikawa, Japan [Received September 30, 2020; accepted November 30, 2020]

In Japan, flood disasters caused by record-breaking 1. Introduction heavy rainfall frequently cause significant damages. It is also great concern that heavy rainfall may in- In recent years, record-breaking heavy rainfalls have crease and occur more frequently due to global warm- frequently occurred in Japan. In 2019, a super typhoon, ing. In July 2013, a heavy rainfall event caused record- Hagibis, made landfall in Shizuoka Prefecture and caused flooding of the Kakehashi River in Ishikawa Prefec- severe damage across a wide area of the eastern part of ture. In this study, pseudo global warming (PGW) Japan. In 2018, a large linear heavy rainfall system de- experiments were implemented for the heavy rainfall veloped in the western part of Japan. Not only such in 1998 and 2013 around the Kakehashi River basin. large-scale heavy rainfall systems, but also a short-time Based on the results of PGW simulations, rainfall with concentrated heavy rainfalls have caused critical dam- different return periods were generated. Runoff anal- ages (e.g., the heavy rainfall event in the Toga River in yses and inundation simulations were carried out by 2008). Frequencies of extreme rainfall events (rainfall > forcings with multiple return periods, and the results 100 mm/day, 200 mm/day, 50 mm/h, 80 mm/h, etc.) are were used to estimate the economic losses due to flood reported to have increased in Japan during the 115 years inundation. Expected values of the economic losses from 1901–2015 [1]. Such heavy rainfalls exceed de- were calculated using two methods for multiple return sign level rainfalls and have increased inundation dam- periods. Differences between the two expected values ages. Moreover, due to climate change associated with indicates the importance of the weighting method for global warming, rainfalls are expected to become large- the result of each return period. In addition, varia- scale and concentrated for a short period of time, and tions of spatial distribution of houses and firms in a flood damage will become more serious. In response to city (i.e., urban structure) were simulated using a com- flood damage caused by unexpected rainfall, the Ministry putable urban economics (CUE) model for the area of of Land, Infrastructure, Transport and Tourism in Japan middle-lower reach of the Kakehashi River basin to (MLIT) announced “Implementation of disaster preven- examine its impact on economic loss due to flooding. tion and mitigation at a new stage” in January 2015 [2]. In the simulation using the CUE model, a more severe This announcement was presented as a goal to save lives flood inundation risk and an additional insurance bur- and avoid catastrophic socioeconomic damage. In May den for general households were added, and possible 2015, the Flood Control Act in Japan was amended, and it variations of urban structure were estimated around was stipulated that flood control areas should be expanded the lower part of the Kakehashi River basin. Under and announced to the public, assuming the maximum pos- the more severe risk condition, relocation proceeded sible rainfall. In the announcement by MLIT, an assumed from higher risk areas to safer areas, and possible eco- maximum scale rainfall is defined as that with a return nomic losses decreased in the target area. This result period of 1,000 years or more in the target river basin ac- indicates that proper recognition of risk can reduce cording to a statistical measure based on past observed flood damages. On the other hand, there were small rainfall. The return periods of the current design rain- variations in economic losses under the condition with falls are from about 80 to 200 years for first-class rivers in the additional flood insurance burden. Japan. There is a large gap between these two standards and estimation of intermediate scale rainfall is also important. Keywords: flood, climate change, inundation simulation, For long periods, prevention of flood damages using flood flood economic loss, computable urban economics (CUE) control facilities for design rainfalls has been a goal of model flood management in Japan. However, it is difficult to pre-

Journal of Disaster Research Vol.16 No.3, 2021 329

© Fuji Technology Press Ltd. Creative Commons CC BY-ND: This is an Open Access article distributed under the terms of the Creative Commons Attribution-NoDerivatives 4.0 International License (http://creativecommons.org/licenses/by-nd/4.0/). Kotone, K. et al.

vent flood damage by the assumed maximum scale rain- fall using only hardware measures. Taniguchi and Shibuo [3] calculated the inundation depth from runoff analysis and flood simulation with var- ious probable annual rainfall inputs and estimated the ex- pected inundation depth from the middle to lower reaches of the Kakehashi River basin in . In this study, reproductive and pseudo global warming (PGW) simulations are made for the heavy rainfall events in the basin in 1998 and 2013. The rainfall event in 2013 caused a record flood water level of the Kakehashi River. Results of numerical weather simulations are expanded to rainfall with multiple return periods. Using rainfalls of Fig. 1. (a) Target area for the seamless model. Red crosses multiple return periods, runoff analyses and flood simula- indicate cross sections for levee breach, (b) Kakehashi River tions are implemented to estimate inundation depths. Us- basin and river network for RRI model. ing the simulated inundation depth as inputs, economic losses are then estimated for each return period accord- ing to the Flood Control Economic Survey Manual (draft) path is 42 km long and the basin area is 271.2 km2.The published by MLIT [4]. By using estimated economic city of Komatsu covers most of the basin, but part of the losses for multiple return periods, expected value of flood Nabetani and Hatcho river basins are also covered by the economic loss is calculated for the area of the lower Kake- cities of Nomi and Hakusan. In addition, more than 70% hashi River including Komatsu City, the second largest of the main flow channel and basin area are situated in the city in Ishikawa Prefecture. mountainous area. As an urban planning approach, Takagi et al. [5] pre- dicted flood safety level and land use using the com- putable urban economics (CUE) model under the con- 3. Method dition of implementation of flood control measures and evaluated their effects. Seya et al. [6] discussed flood 3.1. Precipitation for Runoff Analyses vulnerability assessment using the extended CUE model and conducted scenario analysis considering the effects In this study, the results of Taniguchi and Shibuo [3] of global warming. Teramoto et al. [7] evaluated the cost- were used to prepare forcing data for runoff analysis. The benefit of two types of land use regulation and examined rainfall data used herein are simulation results based on the applicability of such regulation to reduce flood risk. the rainfall that occurred in September 1998 and the rain- This study uses the CUE model for the Kakehashi River fall event that occurred in July 2013. Weather Research basin and simulates variations of spatial distribution of and Forecasting model (WRF) [8] version 3.6.1 was used houses and firms in a city (i.e., urban structure) with dif- for numerical weather simulations. Initial and boundary ferent levels of flood risk and additional economic bur- conditions for reproductive simulations were generated dens in vulnerable areas. Then, impacts of these varia- using NCEP FNL [9] and NOAA OI SST [10]. To esti- tions on flood economic losses are evaluated. mate possible future heavy rainfalls, a PGW method [11] Section 2 provides an overview of the target river basin. was applied. To generate PGW conditions as initial and In Section 3, simulation forcing, models, and calculation boundary conditions for numerical weather simulations, method of flood economic loss are introduced. Flood eco- results by four global climate models (GCMs) in the nomic loss and its expected values are discussed in Sec- 5th phase of the Climate Model Intercomparison Project tion 4. The impacts of variations in urban structures on (CMIP5 [12]) were used. In the CMIP5 experiment, sev- flood economic losses are shown in Section 5. Finally, eral greenhouse gas concentration scenarios (representa- concluding remarks are presented in Section 6. tive concentration pathways: RCPs [12]) were adopted. To implement PGW simulation, future anomaly between RCP8.5 and historical experiments were added to NCEP 2. Target Basin FNL. Future anomaly was defined as the difference in climatological monthly mean value for RCP8.5 and his- The Kakehashi River is a first-class river flowing torical run. Climatological value for future and cur- through Ishikawa Prefecture. The river starts at Mt. Suzu- rent climate are calculated as averages of 2090–2100 and gadake of the Dainichizan mountain range of Mt. Haku- 1990–2000, respectively. For WRF simulations, a two- san, heads north and enters Nomi, Enuma hills and joins way three-level nesting method was adopted as shown in several branches (Goutani river, Kasukami river, Fuddaiji Fig. 2. The spatial resolutions of three domains (D01, river) and enters the plains. After that, the Nabetani, D02, and D03) were 30 km, 10 km, and 2 km, respec- Hatcho, and Mae rivers join the Kakehashi River. It then tively. An outline of WRF settings is presented in Table 1. flows into the . The Kibagata lagoon is the In addition to the PGW method, a simple ensem- headwater of the Mae River (Fig. 1(a)). The main flow ble simulation technique (the lagged average forecasting

330 Journal of Disaster Research Vol.16 No.3, 2021 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

Table 2. Pairs of the scale factors α and β.

(α, β ) (α, β ) (α, β ) (α, β )         1 1 2 1 2 2 1 − , 0, , , 3 3 3 3 3 3 3         1 2 1 1 1 2 2 − , ,− ,1 , 3 3 3 3 3 3 3         1 1 2 1 1 − ,1 ,0 ,− 1,− 3 3 3 3 3         1 1 1 2 1 0, , ,0 1, 3 3 3 3 3

Fig. 2. Target domains of the WRF simulations. Green, yel- low, and red areas indicate D01, D02, and D03, respectively.

Table 1. Outline of WRF settings.

Item Setting Model version 3.6.1 Spatial resolution 30 km, 10 km, 2 km Temporal resolution 120 sec, 60 sec, 12 sec Cloud microphysics Lin ice scheme Cumulus parameterization Betts-Miller-Janjic Radiation scheme RRTM scheme Surface layer scheme RRTM scheme Land surface model Monin-Obuloc Mello-Yamada Nakanichi and Boundary layer model Niino level 3 scheme Data assimilation Spectral nudging (only for D01) Fig. 3. Spatial distribution of two-days total rainfall for the three selected rainfall events. The white line indicates the (LAF) method [13]) was applied to increase the number Kakehashi river basin. The unit of the color bar is mm. of sample rainfalls. In the LAF method, three base state vectors X1, X2,andX3 for different times were initially obtained. Then, two difference vectors ΔX2 and ΔX3 were the Kakehashi River basin (indicated by a white line). generated as the difference between X2 and X1 and be- Two rainfall events have time variation of a centralized tween X3 and X1, respectively. Finally, new state vec- peak, and another has backward concentrated time varia- tors Xn are calculated as follows: tion (Fig. 4). The selected rainfall was stretched to correspond to Xn = X + αΔX + βΔX , ...... (1) 1 2 3 multiple return periods. In the Kakehashi River basin, where α and β are scale factors for ΔX2 and ΔX3, respec- 9-hour rainfall is used as a design rainfall for flood tively. A total of 19 ensemble members were generated prevention planning. Here, the 9-hour rainfalls were for the reproductive simulation and each PGW simulation. calculated for the period from 500 to 43,000 years. The pairs of the scale factors are listed in Table 2. From 500 to 3,000 years, the 9-hour rainfalls were esti- From the results of ensemble PGW simulations, three mated for each 500 years. The rainfalls were then de- rainfall events with relatively uniform spatial distribution fined with 5 mm intervals for the return period longer were visually selected to consider impacts of various flood than 3,000 years. Four probability distribution models patterns. Fig. 3 shows the spatial distribution of the to- were adopted to estimate the target rainfalls: Iwai, lognor- tal rainfall during two days in the simulation for the se- mal distribution 3 parameter quantile, lognormal distri- lected three rainfall events. There are areas with small bution 2 parameter (SladeI, L product-moment method), (< 200 mm) or large (> 320 mm) total rainfall, but spa- and lognormal distribution 2 parameter (SladeI, product- tial variations of the total rainfall are relatively small in moment method). The average value estimated from the

Journal of Disaster Research Vol.16 No.3, 2021 331 Kotone, K. et al.

(the seamless model [15]) that can consider various flood- ing factors. The seamless model consists of a one- dimensional river channel model, a sewer network model, and a surface flood model. As for the sewerage network model, since the sewerage data are not available for the target area, a simple drainage model [16] was constructed using the specifications of the drainage area and corre- sponding drainage facilities according to the Komatsu City drainage plan. The one-dimensional river model covers 12.4 km from the mouth of the Kakehashi River. River discharge calculated by the RRI model was given as a boundary condition to the upstream end of the one- dimensional river model. Interflow from the five tribu- taries (Kasukami, Fuddaiji, Nabetani, Hatcho, and Mae Fig. 4. Temporal variations of the basin mean rainfall in rivers) were given as a lateral boundary condition calcu- the Kakehashi River for the three selected rainfall events. lated by the RRI model. For the boundary condition at Results are values stretched to 43,000 years rainfall. the downstream end of the one-dimensional river model, the water level was calculated as the average of the ob- served tidal level at Mikuni and Kanazawa, which are lo- Table 3. 9-hour rainfall for each return period (mm). cated to the south and north of the mouth of the Kakehashi River, respectively. Flood flow caused by a levee breach Return 9-hour Return 9-hour and inland flood are calculated by a two-dimensional un- period rainfall period rainfall steady flow model. Kibagata lagoon, which is located in [year] [mm] [year] [mm] the southern part of the target area, is defined as a flood 500 177.2 7,000 225.0 control basin. Spatial resolution of a one-dimensional 1,000 189.6 9,000 230.2 river model and two-dimensional flow model is 200 m 1,500 197.0 12,000 235.6 and 40 m, respectively. 2,000 202.3 15,500 240.5 For each rainfall with different return period defined Table 3 2,500 206.4 20,000 245.3 in , a simulation without levee breach was imple- mented to define cross sections where the maximum water 3,000 209.7 26,000 250.3 level exceeds the planned high water level (HWL). Then, 4,000 215.0 33,500 255.2 inundation simulations were made with levee breach at 5,000 220.9 43,000 260.1 these cross sections. Levee breach was not adopted for the section more than 8.6 km from the river mouth because it has little effect on the downstream area. In Fig. 1(a),levee four probability distribution models was defined as tar- breach sites are marked with a cross. Levee breaches oc- get rainfall for each return period. The estimated rainfalls cur at a time when the water level in the one-dimensional are reported in Table 3. The assumed maximum scale model exceeds HWL, and the width of levee breach ex- rainfall for the Kakehashi River estimated by MLIT is tends from the initial width (25 m) to the final width 350 mm/9-hour. The rainfalls in Table 4 are smaller than (100 m) in one hour. It is also assumed that the levee the estimation by MLIT, but a return period of the maxi- will be destroyed to its base. mum rainfall by MLIT could be overlong. Therefore, the MLIT’s estimation was eliminated in this study. 3.3. Estimation of Economic Losses Due to Flood- 3.2. Runoff Analyses and Inundation Simulations ing For the runoff analysis, the Rainfall-Runoff-Inundation Economic loss is one of the important indicators to ex- (RRI) model developed by the Civil Engineering Re- press the flood disaster risk. In this study, economic losses search Institute and the International Center for Wa- under different conditions are calculated by considering ter Hazard and Risk Management [14] was used. Hy- the flooding due to rainfall of various return periods as- droSHEDS, which is a global hydrological data set, was suming climate change. At the same time, impacts of used to create topographical data for the RRI model. El- variations in urban structure on economic losses due to evation, slope direction, and cumulative discharge with a flooding are examined. Such losses are calculated in ac- spatial resolution of 15 seconds were used to create river cordance with the Flood Control Economic Survey Man- basin data. The spatial resolution of the generated basin ual (draft) published by MLIT [4]. Economic losses are data is 500 m. The model watershed and river network are calculated for damages to houses, household goods, de- shown in Fig. 1(b). preciation and inventory of businesses, and crops. The Flooding and inundation in low-lying areas in the mid- land use classification of the target area (Fig. 5)wasdeter- dle and lower reaches of the Kakehashi River were sim- mined using the land use subdivision mesh published by ulated using an integrated flood and inundation model MLIT. The calculation of economic loss was carried out

332 Journal of Disaster Research Vol.16 No.3, 2021 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

Table 4. Damage rate for houses (RH ).

Inundation depth Ground slope Above floor Below floor –50 cm 50–99 cm 100–199 cm 200–299 cm 300 cm– Group A 0.032 0.092 0.119 0.266 0.580 0.834 Group B 0.044 0.126 0.176 0.343 0.647 0.870 Group C 0.050 0.144 0.205 0.382 0.681 0.888 (Ground slope) Group A: –1/1000, Group B: 1/1000–1/500, Group C: 1/500–

Table 4, the elevation and slope mesh data in 2009 pub- lished by MLIT are used. For the unit value of house for each prefecture, the evaluation value of Ishikawa Pre- fecture in 2003 from the Flood Control Economic Survey Manual (draft) is used.

3.3.2. Economic Loss of Household Goods To estimate the economic loss of household goods (LHH), the asset value of household goods is calculated for each calculation mesh by multiplying the number of households (NH ) and the household goods evaluation value per household (EHH). Then, the damage rate (RHH) for each inundation depth (Table 5) is multiplied by the asset amount of household goods to obtain LHH. The pro- Fig. 5. Land use around Kakehashi River basin. cedure is expressed by the following formula.

LHH = NH × EHH × RHH...... (3) based on the assumption that damage to houses, house- For NH in the target area, the total number of popula- hold goods, and depreciation and inventory of business es- tion and the total number of households in 2015 published tablishments would occur in residential/urban areas, and by the Statistics Bureau of the Ministry of Internal Af- crop damage would occur in agricultural land. It was as- fairs and Communications in Japan is used. In addition, sumed that inundation on the floor occurs when the inun- for EHH, the evaluation value in 2003 published in the dation depth is 45 cm or more, and soil accumulation in Flood Control Economic Survey Manual (draft) is used. houses is not considered. The calculation methods of each damage amount are described below. 3.3.3. Economic Loss of Depreciation and Inventory Assets for Business Establishments 3.3.1. Economic Loss of Houses Economic loss of depreciation and inventory asset for The amount of house assets is calculated by multiply- business establishments is calculated for each industrial ing the floor area by the evaluation value per 1 m2. Then, category. Depreciation and inventory assets for each cal- the economic loss of a house is calculated by multiplying culation mesh is calculated by multiplying the number of the amount of house assets by the damage rate accord- employees and the unit depreciable and assessment value ing to the inundation depth (Table 4) for each calculation per employee (1,000 JPY/person) for each industry classi- mesh of the two-dimensional flood model in the seamless fication. Then, economic loss is estimated by multiplying model. The amount of economic loss to the house is cal- the damage rate (Table 6) and depreciation and inventory culated as follows: assets. For the i-th industrial category, the economic loss is estimated as follows: LH = AH × EH × RH , ...... (2) LD,i = NE,i × ED,i × RD, ...... (4) where LH , AH , EH ,andRH are economic loss of the house (JPY), floor area in the target mesh (m2), the unit value of where LD,i, NE,i,andED,i are economic loss of depreci- house for each prefecture (1,000 JPY/m2), and the dam- ation and inventory assets (JPY), number of employees, age ratio of house for each inundation depth, respectively. and the unit depreciable and assessment value per em- For AH , the total floor area of the building in 2010 accord- ployee for i-th industrial category, respectively. RD is ing to the Japan Construction Information Center is used. the damage ratio for depreciation and inventory assets. The spatial resolution of the total floor area data is 100 m. The industrial categories covered herein are mining, con- For the ground slope data, which is used to estimate RH in struction, manufacturing, electric supply, gas supply, heat

Journal of Disaster Research Vol.16 No.3, 2021 333 Kotone, K. et al.

Table 5. Damage rate for household goods (RHH).

Inundation depth Above floor Below floor –50 cm 50–99 cm 100–199 cm 200–299 cm 300 cm– 0.021 0.145 0.326 0.508 0.928 0.991

Table 6. Damage rate for depreciation and inventory asset (RD).

Inundation depth Asset Above floor Below floor –50 cm 50–99 cm 100–199 cm 200–299 cm 300 cm– Depreciation 0.099 0.232 0.453 0.789 0.966 0.995 Inventory 0.056 0.128 0.267 0.586 0.897 0.982

Table 7. Damage rate for agricultural products (RR, RF ).

Inundation depth Inundation depth –0.5 m 0.5–0.99 m 1.0 m– Inundation days 1–2 3–4 5–6 7– 1–2 3–4 5–6 7– 1–2 3–4 5–6 7– Paddy (RR) 21 30 36 50 24 44 50 71 37 54 64 74 Fields (RF) 27 42 54 67 35 48 67 74 51 67 81 91

supply, water supply, transportation, communications, lated as follows: wholesale, retail, finance, insurance, real estate, and ser- LR = AI ×YR × ER × RR, ...... (5) vice business. For NE,i in the target area, the number of employees in LF = AI ×YT × ET × RF , ...... (6) 2015 published by the Statistics Bureau of the Ministry where LR, AI, YR, ER,andRR are economic loss of rice of Internal Affairs and Communications in Japan is used. paddy (JPY), inundation area (km2), rice yield per unit For ED,i, the values in 2003 published in the Flood Control area (t/km2), unit evaluation value of rice (1,000 JPY/t), Economic Survey Manual (draft) is used. and damage rate of the rice paddy, respectively. LF , YT , ET ,andRF are economic loss of field (JPY), tomato 3.3.4. Economic Loss of Agricultural Products yield per unit area (t/km2), unit evaluation value of tomato (1,000 JPY/t), and damage rate of the field, re- The economic loss of agricultural products is calculated spectively. based on Machida et al. [17]. For rice paddies, the dam- Average annual yield and the unit evaluation value of age amount is calculated by multiplying the rice yield per Ishikawa Prefecture in 2003 was used for estimation. This unit area by the unit evaluation value of rice and the in- information is published in the Flood Control Economic undation area. Then, the damage ratio calculated by inun- Survey Manual (draft). dation depth and inundation days is multiplied to estimate the economic loss of rice paddy. The damage ratios for agricultural products calculated by inundation depth and 3.4. Simulation of Urban Structure Using CUE inundation days are reported in Table 7. In calculation of the damage ratio, inundation days was set to one because Model the simulated flood inundation was resolved in about one Variations of the urban structure around the lower reach day in this study. of the Kakehashi River basin were simulated using CUE Economic loss of fields for other agricultural crops is model. calculated based on tomatoes because the average value Fujita and Ogawa [18] developed an urban spatial of the major agricultural products is close to that of toma- equilibrium model considering interactions between firms toes. As with rice paddies, the economic loss is calculated and households and explain the endogenous formation by multiplying the yield per unit area by the unit evalua- of polycentric urban spatial structures. Nakamura and tion value of tomatoes and the inundation area. Then, the Takayama [19] developed a CUE model based on Fujita damage ratio (Table 7) is multiplied to estimate the eco- and Ogawa [18] to examine the stability of spatial equi- nomic loss of fields. libria and clarified that polycentric urban structures can Economic losses for rice paddies and fields are calcu- emerge. In the current study, the CUE model by

334 Journal of Disaster Research Vol.16 No.3, 2021 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

Nakamura and Takayama [19] was used to simulate the ferent cross sections, and three patterns of rainfall shown urban structures. in Figs. 3 and 4. Based on the composite of the maxi- In this study, a city consists of discrete locations mum inundation depth, the economic losses due to flood- ing were calculated by the method described in the previ- J = {1,2,...,I}...... (7) ous section for each return period (Fig. 7). Each location a ∈ J is endowed with Aa, units of a The economic losses in the case of a return period of fixed supply of land, and the distance between the lo- 1,000 years was the shallowest in all cases because the cations a and b is dab. There are households and firms inundation depth was the shallowest, and most of eco- in a city. The total number of households in the city nomic losses are less than 100,000 yen. The meshes in is N, and the number of households residing in location which the economic loss is less than 100,000 yen are a and working at location i, “households ai,” is denoted farmland, and the damage is mainly due to crop dam- as nai. The total number of households residing at loca- age. The total damage is about 7.2 billion yen in the tion a is Na = ∑i∈J nai. The number of firms in location i case of 1,000 years rainfall. According to MLIT, in the F is Mi = ∑a∈J nai Each firm requires s units of land and case of Typhoon No.10, which caused disasters in a wide one unit of labor to operate. The profit of a firm choosing area around Iwate Prefecture in 2016, the economic losses location i is given by in Iwate Prefecture and were 168 billion yen and 165 billion yen, respectively. In addition, the eco- π = αF F (M)+η − w − r sF , i i i i i ...... (8) nomic loss of the Kinugawa heavy rainfall event in 2015 F where α Fi(M)+ηi is the level of production, Fi(M)= was 159 billion yen. Since the Kakehashi River is one of −1 Ai ∑ j∈J exp(−τdij)M j/A j represents production exter- the smallest first-class rivers in Japan and the target area nalities, ηi denotes the production fundamentals, wi is in this study is relatively small, the order of the estimated wage, and ri is land rent in location i. The functional economic loss seems appropriate. In the case of a re- form of Fi(M) reflects the assumption that firms produce turn period of 2,000 years, the economic loss in the entire more goods when they are close to other firms. Each basin was larger than that in the case of 1,000 years, and household consumes sH units of land inelastically for res- the maximum damage amount per mesh was also large. In idential purposes. The utility uai of a household ai is addition to meshes with damages of 100,000 yen or less, H uai = wi − s ra −tdai + ηa where t is the commuting cost meshes with damages of 10,000,000 yen or less are in- for a unit distance and ηa denotes residential amenities. creasing. Economic losses expand across a wider area as With regard to utility maximization, profit maximiza- the return period becomes longer. On the other hand, the tion, and market clearing, the utility is expressed as economic loss in each mesh hardly changes even when the return period become longer. As shown in Fig. 6,the u (n)=αF F (M)+η + η −td ai i a i ai longer the return period, the larger the inundation depth sH N + sF M sH N + sF M tends to be. However, because variations of damage rate − βsH a a − βsF i i . 2 A 2 A (9) according to inundation depth are not so large for agricul- a i tural products (Table 7), once crops have already been lost The spatial equilibrium condition is as follows: with smaller flood, additional losses are small even if the   inundation depth increases. On the other hand, for gen- exp θuai(n) nai =  N...... (10) eral households and business establishments, the damage A θu (n) a ∑∑exp ai rate significantly changes depending on the inundation a∈I i∈I depth (Tables 4, 5,and6); hence, the amount of damage Here, θ is a scale parameter. in the mesh increases as the inundation depth increases Parameters in the CUE model were estimated using the with longer return period. Since there are many areas of method in Ahlfeldt et al. [20]. In the estimation of param- farmland around the Kakehashi River, the increase of eco- eters, existence of inundation and inundation depth are in- nomic loss due to the increase of inundation depth is not troduced to consider variations of inundation conditions. so large. At the same time, an additional burden of insurance pre- mium depending on inundation depth is also introduced 4.2. Expected Value of Economic Losses Due to in the CUE model. Flooding The expected value of economic loss due to flooding 4. Estimated Economic Losses from Flooding is calculated by two methods using the economic loss amount for each return period obtained in the previous 4.1. Economic Losses Caused by Rainfall of Multi- section. The two calculation methods and results are ple Return Periods shown below.

Figure 6 shows a part of the spatial distribution of the 4.2.1. A Method Using Cumulative Probability of Tar- maximum inundation depth for each return period by in- get Rainfalls undation simulations. Each result is a composite of the Based on the occurrence frequency of the maximum maximum inundation depth with levee breach at five dif- 9-hour rainfall in the ensemble numerical meteorologi-

Journal of Disaster Research Vol.16 No.3, 2021 335 Kotone, K. et al.

Fig. 6. Composite of the maximum inundation depth (m) for each return period.

Fig. 7. Flood economic losses (1,000 JPY) for each return period.

cal simulation, which is the basis of runoff analysis and period. Then, the expected value of flood economic loss inundation simulation, the weight of economic loss is cal- is calculated by adding up the results of all return periods. culated from the cumulative probability of each 9-hour The cumulative probability and weight W1 for each return rainfall. The weight is calculated by subtracting the cu- period are reported in Table 8. mulative probability of each return period from 1.0. The weight is multiplied by the economic loss for each return

336 Journal of Disaster Research Vol.16 No.3, 2021 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

Table 8. Weights and probabilities used to calculate expected value of economic losses due to flooding.

W P Return period 9-hour rainfall Cumulative Weight W 1 Probability Pi − Weight W [year] [mm] probability ϕ (= 1 − ϕ ) (×1100 4) 2 1,000 189.6 0.907 0.0925 90.0 1.0384 1,500 197.0 0.912 0.0681 3.333 0.0385 2,000 202.3 0.946 0.0541 1.667 0.0192 2,500 206.4 0.955 0.0449 1.000 0.0115 3,000 209.7 0.961 0.0382 0.667 0.0077 4,000 215.0 0.971 0.0294 0.883 0.0096 5,500 220.9 0.978 0.0217 0.682 0.0079 7,000 225.5 0.983 0.0170 0.390 0.0045 9,000 230.2 0.987 0.0130 0.317 0.0037 12,000 235.6 0.991 0.0095 0.278 0.0032 15,500 240.5 0.993 0.0070 0.188 0.0022 20,000 245.3 0.995 0.0052 0.145 0.0017 26,500 250.3 0.996 0.0038 0.115 0.0013 33,500 255.2 0.997 0.0027 0.086 0.0010 43,000 260.1 0.998 0.0019 0.066 0.0008

4.2.2. A Method Using Occurrence Probability of Tar- tant, but we focus on the differences of the spatial distri- get Rainfalls bution of the expected value between the two methods. W In another method, expected value was calculated by When the weight 2 is used, the expected value is less than 5 million yen in most meshes. In addition, there is an weight W2 based on interval probability pi. The interval probability pi for the rainfall with the return period of yi area around the central part of the target area (indicated is defined as the occurrence probability for the interval by a white circle in Fig. 8) where the damage amount is between yi− and yi and calculated as the following equa- about 10 to 50 million yen. On the other hand, when the 1 W tion. weight 1 is used, there are more meshes with a damage amount of 5 to 10 million yen, and the expected value of 1 1 p = − . damage amount is decreasing on the left bank of the mid- i y y ...... (11) i−1 i stream part. As seen in Fig. 6, in the area on the left bank Then, the expected value of the economic loss is esti- of the Kakehashi River, inundation occurred even during mated with the economic loss of each mesh for each re- rainfall with a smaller return period. In such areas, the turn period. For the shortest return period in Table 8 (i.e., expected value of the economic loss becomes large when 1,000 years), the interval probability between 100 years, the weight W2 was used. In areas where flooding occurs which is the target of flood control plan in the Kakehashi only due to rainfall with a long return period, the expected River basin, and 1,000 years was given. To compare with value tends to be smaller. Therefore, the spatial distribu- the method using the weight W1 based on the cumulative tion of the expected value becomes inhomogeneous. On probability (described in Section 4.2.1), the weight W2 the other hand, when the weight W1 is used, the total was determined by multiplying the interval probability pi amount of damage is smoothed and relatively homoge- by a constant K so that the total expected value of eco- neous over the entire area compared to the case using W2; nomic loss by the two methods becomes the same value. hence, the spatial difference in the risk of economic loss within the target area is reduced. To select areas for implementing additional flood disas- W 4.2.3. Comparison of the Expected Values of Eco- ter prevention measures, an evaluation method using 2, nomic Losses Due to Flooding Using the Two which sharpens the difference in risk, may be appropri- Methods ate. However, as previously mentioned, the effect of rain- fall with a large return period tends to be underestimated Figure 8 shows the spatial distribution of the expected when using W . To consider risks of rare heavy rainfall values of the economic loss from flooding calculated by 2 events, the method using W1 seems better. In the future, the aforementioned methods. In both results, the total it will be essential to make a detailed comparison of such expected value of flood damage in the target area was methods and to examine how to use them. approximately 10.7 billion yen. Since the weights W1 and W2 give the contribution rate of each return period and have little physical or socioeconomic meaning, quan- titative evaluation of the expected value is not so impor-

Journal of Disaster Research Vol.16 No.3, 2021 337 ™

Kotone, K. et al.

Fig. 8. Spatial distribution of expectation of flood economic loss. (a) Result based on cumulative probability (W1), (b) re- W sult based on occurrence probability ( 2). Fig. 9. (a) Inundation depth (m) under the condition of the possible maximum rainfall, (b) variation of numbers of households (%) under the higher inundation risk estimated using the CUE model, (c) same as (b) but for companies. 5. Impacts of Variations in Urban Structure on Economic Losses Due to Flooding

5.1. Variation of Urban Structure Caused by Up- in the economic loss due to flooding under the different dated Flood Risk Information and its Impact urban structure is investigated. Here, the changes in the on Economic Loss number of households and companies at each area are ex- To consider the impact of larger flood hazard on urban amined. Specifically, this study focuses on the growth rate structure, it is assumed that the possible maximum rain- of households and companies at each location. fall (defined by MLIT) occurs at the same frequency as Figure 9(a) shows the inundation depth at the assumed the design rainfall in the present climate. Under the con- maximum scale rainfall. The results obtained by the CUE dition of the possible maximum rainfall, inundation area model are presented in Figs. 9(b) and (c).Fromthisre- and depth increase and the amenity and productivity in sult, it can be seen that the number of both households the target area decline. As a result, relocation of residence and companies tends to decrease at the location where and work location occurs in that area. In this section, the the inundation depth is greater than 3 m. This is be- variations in urban structure under the different flood risk cause the amenity level and productivity of these areas is simulated using the CUE model. Then, the variations are relatively lower than those of other areas. In other ar-

338 Journal of Disaster Research Vol.16 No.3, 2021 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City

eas with the inundation depth smaller than 3 m, the num- bers of households or companies increase, and rent will rise. In such case, households and companies can move to a location with high convenience (e.g., high amenity and productivity levels) or lower land rent even if there is a high risk of inundation. However, the complexity of the CUE model gives different tendencies in variations of number of households and companies. In area I, inunda- tion depth is quite small, and many companies moved to the area. Then, the rent becomes higher and houses de- crease in area I. In areas II and IV, the inundation depth is greater than 3 m, but the variations from the original inun- dation depth under the present design level are small (re- sults are not shown), or the inundation risks did not vary significantly. At the same time, the companies increased in areas I and III. Then, the number of houses in areas II and IV increased due to residents commuting to areas I and III. For area II, lowering the rent caused by relocation of companies could also accelerate the increase in houses. Figure 10 shows the difference in economic losses be- tween the original and the altered urban structures under the more serious flood condition. Here, results for two dif- ferent return periods are shown. Positive values indicate that the flood damage amount increased with the change of urban structure. The total damage amount in the tar- get area is decreasing in the results for both return peri- ods. Results for all return periods revealed that the total economic loss in the altered urban structure is 14–23% smaller than that of the original urban structure. The eco- nomic loss decreases on both sides of the Kakehashi River in the middle reach (area A in Fig. 10) but increases on the left bank on the downstream side (area B in Fig. 10). The region with decreasing economic damage due to flooding is where the number of houses and companies are decreas- ing (Fig. 9). On the other hand, the numbers of house- holds and companies are increasing in the area where the flood damage amount is increasing, and the economic loss becomes larger when flooding occurs in such areas. As shown in Fig. 9, changes in flood disaster risk sug- gest that companies and households may move to safer areas, reducing flood damage in the entire basin. How- ever, if the possibility of inundation at the relocation site Fig. 10. Variations in flood economic loss under the con- is not zero, the amount of flood damage at the relocation dition of different flood risk (mil. JPY). (a) Result for the site will increase. In candidate areas of relocation, addi- return period of 2,500 years, (b) result for the return period tional measures such as enhanced drainage systems and of 33,500 years. raising residential land would be necessary to reduce the economic losses even the case of increasing number of households and companies. defined as the expected value of the maximum inunda- tion depth calculated for multiple return periods and the 5.2. Variation of Urban Structure Caused by Ad- weight of each rainfall (here, W1 in Table 8 was used). ditional Insurance Burden and its Impact on The insurance premium for each location set by the above Economic Loss method is as shown in Fig. 11(a). The flood disaster risk is calculated using results of in- In this study, it is assumed that the flood insurance pre- undation simulation for multiple return periods, and the miums are paid only by the household. Fig. 11(b) shows insurance premium is set according to the flood risk for the growth rate of households when only households are each area. The effect of insurance premiums to guide insured. The number of households has decreased at the households and companies to locations with low water location where insurance premiums are paid. In addi- disaster risk is investigated. The insurance premium is tion, the employment place changes because the com-

Journal of Disaster Research Vol.16 No.3, 2021 339 Kotone, K. et al.

Fig. 11. (a) Flood insurance premium for each location, (b) variation of numbers of households (%) with additional insurance burden estimated using the CUE model, (c) same as (b) but for companies. Fig. 12. Variations in flood economic loss (mil. JPY) under the condition of additional flood insurance burden. (a) Result for the return period of 2,500 years, (b) result for the return muting cost changes as the place of residence changes period of 33,500 years. (Fig. 11(b)). As a result, the number of companies also decreases at the location where insurance premiums are set (Fig. 11(c)). However, in some locations where the decreasing ratio of companies is large, the rent also de- due to flooding increased in almost all meshes. However, creases, and households increase despite insurance pre- the range of the change in economic loss in each mesh miums being set. is smaller than the results in Fig. 10 (see the legend of Figure 12 shows the difference in total economic loss Figs. 10 and 12). As seen in Fig. 11, the number of houses between the original and the altered urban structure un- decreased due to the burden of insurance premiums in ar- der the insurance burden imposed on households. Here, eas with a high risk of flooding, such as the Kakehashi the results for two return periods are shown. The total main river, and the number of companies has increased. economic loss in the target area increased in both results. The decrease in rent due to the decrease in the number of For 15 return periods, the total economic loss increased households is considered to be one of the reasons for the (results are not shown). In addition, the economic loss company relocation, but detailed consideration is needed

340 Journal of Disaster Research Vol.16 No.3, 2021 Estimation of Potential Economic Losses Due to Flooding Considering Variations of Spatial Distribution of Houses and Firms in a City in the future. In addition, in Fig. 12, the mesh with in- Acknowledgements creasing flood economic loss coincides with the region This study was conducted with the support of the Ministry where companies are increasing, and it is thought that the of Land, Infrastructure, Transport and Tourism in Japan, and concentration of companies is the cause of the increase Hokuriku Regional Management Service Association. The au- in the economic loss. On the other hand, the change in thors are also grateful to all of the data providers. The NCEP the number of houses and companies was small when the FNL Operational Global AnalysisdataareavailableattheRe- insurance premium was charged. The effect of the flood search Data Archive at the University Corporation for Atmo- insurance was small on relocation from higher risk areas spheric Research (UCAR). The NOAA OI SST products obtained from NCEP at NOAA. The CMIP5 products are distributed from to safer areas. In the future, it is necessary to consider several CMIP5 data portals. effective insurance premiums to promote relocation from high-risk areas. References: [1] Japan Meteorological Agency, “Clinate Change Monitoring Re- port 2015,” 2016, https://www.jma.go.jp/jma/en/NMHS/ccmr/ 6. Summary ccmr2015 high.pdf [accessed Nobemver 18, 2020] [2] Ministry of Land, Infrastructure, Transport and Tourism, “Imple- In this study, three patterns of rainfalls created by nu- mentation of disaster prevention and mitigation at a new stage,” 2015, https://www.mlit.go.jp/saigai/newstage.html (in Japanese) merical weather simulations were extended to multiple re- [accessed September 15, 2020] turn periods for the Kakehashi River basin. Flood sim- [3] K. Taniguchi and Y. Shibuo, “Estimation of expectation of flooding ulation was carried out using the seamless model. The water depth based on heavy rainfalls with different return periods simulated by pseudo global warming method,” J. of Japan Society economic losses due to flooding were estimated from the of Civil Engineers, Ser. B1, Vol.74, No.5, pp. I 1405-I 1410, 2018 maximum inundation depth distribution for each return (in Japanese). [4] Ministry of Land, Infrastructure, Transport and Tourism, period in Kakehashi River basin. In addition, the expected “The flood control economic survey manual (draft),” 2020, value of the economic loss was calculated using two meth- https://www.mlit.go.jp/river/basic info/seisaku hyouka/gaiyou/ hyouka/r204/chisui.pdf (in Japanese) [accessed September 15, ods. Finally, we calculated the economic loss under the 2020] conditions of altered urban structure due to various fac- [5] A. Takagi, S. Muto, and N. Ohta, “Economic evaluation of flood tors and compared it with the original economic loss. control countermeasures by using computable urban economic model,” Advances in River Engineering, Vol.7, pp. 423-428, 2001 Results of the expected values of economic loss due to (in Japanese). flooding estimated using the two methods indicated the [6] H. Seya, Y. Yamagata, K. Nakamichi, and M. Tsutsumi, “Flood vul- weight for each return period has significant influence on nerability assessment using a CUE model in the Tokyo metropolitan area,” Papers of Research Meeting on Civil Engineering Planning, the spatial distribution of the expected value. If the differ- Vol.44, 2011 (in Japanese). ence of weights for return periods is small as W1,thespa- [7] M. Teramoto, Y. Ichikawa, Y. Tachikawa, and M. Shiiba, “Study tial difference in the expected value of economic damage on applicability of landuse regulation strategies based on flood risk management,” J. of Japan Society of Civil Engineers, Ser. B1, is also small in the target area. On the other hand, if the Vol.66, No.2, pp. 130-144, 2010 (in Japanese). difference of weights between each return period is large [8] W. C. Skamarock, J. B. Klemp, J. Dudhia, D. O. Gill, D. Barker, M. W G. Duda, X.-Y. Huang, W. Wang, and J. G. Powers, “A description as 2, the economic loss with a large weight has a large of the advanced research WRF version 3,” NCAR Technical Note, impact, and there will be significant differences in con- NCAR/RN-475+STR, doi: 10.5065/D68S4MVH, 2008. tribution of economic loss with different return periods. [9] National Centers for Environmental Prediction (NCEP)/National Weather Service/NOAA/U.S. Department of Commerce, “NCEP Then, spatial distribution of expected value of economic GDAS/FNL 0.25 Degree Global Tropospheric Analyses and Fore- loss varies greatly. Based on the results, it is necessary cast Grids,” Research Data Archive at the National Center for At- mospheric Research, Computational and Information Systems Lab- to investigate the optimal calculation method of expected oratory, doi: 10.5065/D65Q4T4Z, 2015 [accessed December 2, value of damage amount in the future. 2018] When the urban structure in the Kakehashi River basin [10] R. W. Reynolds, T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, and M. G. Schlax, “Daily high-resolution-blended analyses for sea is changed considering variation of flood risk from the surface temperature,” J. Clim., Vol.20, No.22, pp. 5473-5496, 2007. planned scale rainfall to the assumed maximum scale rain- [11] T. Sato, F. Kimura, and A. Kitoh, “Projection of global warming onto regional precipitation over Mongolia using a regional climate fall, the economic loss of the river basin for all return pe- model,” J. of Hydrology, Vol.333, No.1, pp. 144-154, 2007. riods became smaller than that in the original urban struc- [12] K. E. Taylor, R. J. Stouffer, and G. A. Meehl, “An overview of ture. This result indicates that the awareness of the inun- CMIP5 and the experiment design,” Bull. Amer. Meteor. Soc., dation risk of residents and companies in the basin may Vol.93, No.4, pp. 485-498, 2012. [13] K. Taniguchi, “A simple ensemble simulation technique for assess- contribute to the reduction of flood damage and realize ment of future variations in specific high-impact weather events,” J. town planning with a small flood disaster risk. When the Geophys. Res: Atmos., Vol.123, No.7, pp. 3443-3461, 2018. urban structure was changed by burdening flood insurance [14] T. Sayama, G. Ozawa, T. Kawakami, S. Nabesaka, and K. Fukami, “Rainfall-Runoff-Inundation analysis of the 2010 Pakistan flood premiums on households in the area with inundation risk, in the Kabul River basin,” Hydrological Science J., Vol.57, No.2, the total economic loss in the basin increased slightly, but pp. 298-312, 2012. [15] H. Sanuki, Y. Shibuo, S. A. Lee, K. Yoshimura, Y. Tajima, H. there was no significant change. Furumai, and S. Sato, “Inundation forecast simulatoin in urban- In the future, it is necessary to establish a more appro- ized coastal low-lying areas considering multiple flood causing fac- tors,” J. of Japan Society of Civil Engineers, Ser. B2, Vol.72, No.2, priate calculation method of the expected value of flood pp. I 517-I 522, 2016 (in Japanese). economic loss and consider the effective economic mea- [16] K. Taniguchi and Y. Shibuo, “Difference of flood and inundation sures for inducing the relocation from higher risk areas to characteristics under various types of precipitation patterns,” J. of Japan Society of Civil Engineers, Ser. B1, Vol.74, No.4, pp. I 1489- safer areas. I 1494, 2018 (in Japanese).

Journal of Disaster Research Vol.16 No.3, 2021 341 Kotone, K. et al.

[17] S. Machida, S. Kawagoe, S. Kazama, M. Sawamoto, H. Yokoki, and K. Yasuhara, “Evaluation of flood damages for climate change,” Name: Proc. of the Symp. on Global Environment, Vol.15, pp. 155-160, 2007 (in Japanese). Koichi Nakamura [18] M. Fujita and H. Ogawa, “Multiple equilibria and structural tran- sition of non-monocentric urban configurations,” Regional Science Affiliation: and Urban Economics, Vol.12, No.2, pp. 161-196, 1982. Nihonkai Consultant Co., Ltd. [19] K. Nakamura and Y. Takayama, “Development of an agglomeration model considering interactions between firms and households,” J. of Japan Society of Civil Engineers, Ser. D3, Vol.74, No.5, pp.I 555- I 569, 2018 (in Japanese). [20] G. M. Ahlfeldt, S. J. Redding, D. M. Sturm, and N. Wolf, “The eco- nomics of density: Evidence from the Berlin wall,” Econometrica, Vol.83, No.6, pp. 2127-2189, 2015. Address: 2-216 Izumihonmachi, Kanazawa, Ishikawa 921-8042, Japan Brief Career: 2019- Nihonkai Consultant Co., Ltd. Selected Publications: • K. Nakamura and Y. Takayama, “Development of an agglomeration Name: model considering interactions between firm and households,” J. of Japan Kaito Kotone Society of Civil Engineers, Ser. D3, Vol.74, No.5, pp. I 555-I 569, 2018 (in Japanese). Affiliation: Master Course Student, Division of Environmen- tal Design, Kanazawa University

Name: Yuki Takayama

Address: Affiliation: Kakuma-Machi, Kanazawa, Ishikawa 920-1192, Japan Associate Professor, Institute of Science and En- Brief Career: gineering, Kanazawa University 2019- Master Course Student, Kanazawa University Selected Publications: • K. Kotone and K. Taniguchi,“Evaluation of Economic Loss from Flooding by Considering Future Depopulation and Variations in Urban Structure,” J. of Japan Society of Civil Engineers, Ser. B1, Vol.76, No.2, Address: pp. I 517-I 522, 2020 (in Japanese). Kakuma-machi, Kanazawa, Ishikawa 920-1192, Japan Academic Societies & Scientific Organizations: Brief Career: • Japan Society of Civil Engineers (JSCE) 2005- West Japan Railway Company 2011- Ehime University 2014- Tohoku University 2016- Kanazawa University Selected Publications: Name: • Y. Takayama, K. Ikeda and J.-F. Thisse, “Stability and sustainability of Kenji Taniguchi urban systems under commuting and transportation costs,” Regional Science and Urban Economics, Vol.84, 103553, 2020. • Affiliation: Y. Takayama and M. Kuwahara, “Bottleneck congestion and residential Associate Professor, Faculty of Geosciences and location of heterogeneous commuters,” J. of Urban Economics, Vol.100, Civil Engineering, Kanazawa University pp. 65-79, 2017. • Y. Takayama, “Bottleneck congestion and distribution of work start times: The economics of staggered work hours revisited,” Transportation Research Part B: Methodological, Vol.81, No.3, pp. 830-847, 2015. Academic Societies & Scientific Organizations: • Japan Society of Civil Engineers (JSCE) Address: • Applied Regional Science Conference (ARSC) Kakuma-Machi, Kanazawa, Ishikawa 920-1192, Japan Brief Career: 2005- The University of Tokyo 2008- Kanazawa University Selected Publications: • K. Taniguchi, “A Simple Ensemble Simulation Technique for Assessment of Future Variations in Specific High-Impact Weather Events,” J. of Geophys. Res: Atmosphere, Vol.123, No.7, pp. 3443-3461, 2018. • K. Taniguchi and Y. Tajima, “Variations in Extreme Wave Events near a South Pacific island under Global Warming: Case study of Tropical Cyclone Tomas,” Progress in Earth and Planetary Science, Article No.8, doi: 10.1186/s40645-020-0321-y, 2020. Academic Societies & Scientific Organizations: • Japan Society of Civil Engineers (JSCE) • Meteorological Society of Japan (MSJ) • Japan Society of Hydrology and Water Resources (JSHWR) • American Geophysical Union (AGU)

342 Journal of Disaster Research Vol.16 No.3, 2021

Powered by TCPDF (www.tcpdf.org)