Assessment of Vulnerability to Extreme Flash Floods in Design Storms

Transcription

1 Int. J. Environ. Res. Public Healt 211, 8, ; doi:1.339/ijerp87297 OPEN ACCESS International Journal of Environmental Researc and Public Healt ISSN Article Assessment of Vulnerability to Extreme Flas Floods in Design Storms Eung Seok Kim 1 and Hyun Il Coi 2, * 1 2 Department of Civil Engineering, Sunmoon University, 1, Kalsan-ri, Tangjeong-myeon, Asan-si, Cungnam-do, , Korea; ydrokes@sunmoon.ac.kr Department of Civil Engineering, Yeungnam University, 214-1, Dae-dong, Gyeongsan-si Gyeongbuk-do, , Korea * Autor to wom correspondence sould be addressed; icoi@ynu.ac.kr; Tel.: ; Fax: Received: 3 Marc 211; in revised form: 29 June 211 / Accepted: 5 July 211 / Publised: 14 July 211 Abstract: Tere as been an increase in te occurrence of sudden local flooding of great volume and sort duration caused by eavy or excessive rainfall intensity over a small area, wic presents te greatest potential danger treat to te natural environment, uman life, public ealt and property, etc. Suc flas floods ave rapid runoff and debris flow tat rises quickly wit little or no advance warning to prevent flood damage. Tis study develops a flas flood index troug te average of te same scale relative severity factors quantifying caracteristics of ydrograps generated from a rainfall-runoff model for te long-term observed rainfall data in a small ungauged study basin, and presents regression equations between rainfall caracteristics and te flas flood index. Te aim of tis study is to develop flas flood index-duration-frequency relation curves by combining te rainfall intensity-duration-frequency relation and te flas flood index from probability rainfall data in order to evaluate vulnerability to extreme flas floods in design storms. Tis study is an initial effort to quantify te flas flood severity of design storms for bot existing and planned flood control facilities to cope wit residual flood risks due to extreme flas floods tat ave ocurred frequently in recent years.

2 Int. J. Environ. Res. Public Healt 211, Keywords: flas flood index; extreme flood; design storm; runoff ydrograp 1. Introduction A flas flood is local flooding of great volume and sort duration. Suc sudden local floods ave occured quite frequently in recent years due to eavy or excessive rainfall in a sort period of time over a small area. Te Korea Meteorological Administration (KMA) reported tat as tere as been an increase in summer precipitation and localized torrential downpours, te yearly precipitation as also increased approximately 19% in te period. Since te temporal and spatial fluctuations of te precipitation are expected to worsen, te climatic and geomorpic vulnerability of waterseds in te Korean Peninsula is exposed to flas flood azards caused by localized convective storms of sort duration over small steep slope regions. Te rapid runoff associated wit debris flow as inundated some watersed areas and te river flow altered by debris flow as resulted in some flood damage suc as bank erosion and bridge collapses, as reported in te annual natural disaster bulletin [1]. A rapid local flood poses te greatest potential danger treat to uman life, public ealt and property, te natural environment and ecosystems, water and oter natural resources, etc. Altoug suc types of dangerous flood damage now constitute national-wide natural disasters, te structural and non-structural alternative plans for flood mitigation ave been mainly carried out for large basins in Korea. Te Fourt Assessment Report (AR4) of te United Nations Intergovernmental Panel on Climate Cange (IPCC) [2] identified te following key priorities wic must be addressed to narrow gaps between current knowledge, and policymaking needs; quantitative assessment of te sensitivity, adaptive capacity, and vulnerability of natural and uman systems to climate cange, particularly canges in te range of climatic variation and te frequency and severity of extreme climate events. Terefore, it is necessary to assess flood mitigation ability or vulnerability of flood control facilities and to devise bot structural and non-structural flood mitigation measures against te residual risk of te extreme floods in ungauged small waterseds. Flas floods ave been studied mostly as climatological penomena, especially focused on te temporal and spatial caracteristics of rainfall [3-5]. From te ydrological perspective, tere ave been several studies on te caracterization of flas floods by analysis of te caracteristics of runoff ydrograps. Kyiama [6] was te first to caracterize flas floods from a runoff perspective using runoff ydrograps. Baskar et al. [7] presented a flas flood index using runoff ydrograp caracteristics suc as te rising curve gradient, flood magnitude ratio, and flood response time evaluated directly from observed runoff ydrograps of 3 flood events from four waterseds in eastern Kentucky. Jung [8] estimated te flas flood index for several flood events of te Bo-cung river basin in Korea following Baskar et al. [7]. In tese studies, te flas flood index was determined by te sum of te tree relative severity factors, eac using a different ordinal scale were class intervals are to some extent arbitrary. Kim and Kim [9] estimated te flas flood index for investigating te relative severity of flas floods in te Han River basin wit 11 flood events, and quantified te flas flood severity for some flood events caused by eavy rainfall in July of 26. However, tere was no attempt to quantify te severity of floods occurred in small catcments were usually observations are not available. Terefore, tis study as modified te flas flood index

3 Int. J. Environ. Res. Public Healt 211, presented by Baskar et al. [7] and developed a new flas flood index determined by te average of relative severity factors on te same scale ratio to te recorded maximum value. IPCC AR4 [2] pointed out tat we need to focus on improvement of systems and metods for long-term monitoring and understanding te consequences of climate cange and oter stresses on uman and natural systems. Te flas flood indexing metods are implemented by quantifying te caracteristics of ydrograps generated from a rainfall-runoff model, HEC-HMS (Hydrologic Engineering Center-Hydrologic Modeling System), for bot te annual maximum rainfall event series during 36 years and te probability rainfall data in return periods of 2, 3, 5, 1, 2, 5, 1, and 2 years and for duration times of 1, 3, 6, 12, 18, and 24 ours in a small ungauged basin, te Oui-mi River basin in Korea. Tis study as provided flas flood index-duration-frequency (FI-D-F) relation curves developed from rainfall intensity-duration-frequency (I-D-F) relations for te use of evaluating vulnerability to extreme flood conditions in a design storm in order to establis disaster countermeasures for residual flood risk in bot existing and planned flood control facilities. 2. Study Catcment Te caracterization of local flas flooding was applied to a small illy drainage catcment, te Oui-mi River basin, located between E~ E, and N~ N, as sown in Figure 1. Te Oui-mi River catcment is a natural basin comprised of 86.3% of woodlands, 12.2% of farmlands, and 1.5% of oter types. Tis study basin is 7.52 m long wit a size of km 2, and te average elevation is m above mean sea level, wit an average slope of 53.4% [1]. Figure 1. Basin map for te Oui-mi River. Location M t. Yong-do (E L m ) Tere is a rainfall gauge station around te study basin, te Jae-cun Gauge Station managed by te Meteorological Administration, were long-term ourly precipitation data is available. Te annual maximum rainfall data series during were collected from tis gauge station. Te annual

4 Int. J. Environ. Res. Public Healt 211, mean rainfall volume was 1,322.5 mm over te same period, and te recorded maximum dept of a single rainfall event is mm in Sep. 11, Flas Flood Indexing Metod To quantify te relative severity of flas floods in small ungauged catcments, tis study determines a flas flood index, FI, from flood runoff ydrograps simulated by a rainfall-runoff model for te annual maximum precipitation event series in a study basin. Baskar et al. [7] as caracterized te flas flood severity by defining a flas flood index evaluated directly from te observed flood ydrograp caracteristics suc as te rising curve gradient, flood magnitude ratio, and flood response time. Tis flas flood index from Baskar et al. [7] is obtained from te sum of tree caracteristics quantified by te relative severity factors. Te problematic issue is in quantifying tese relative severity factors by using eac different ordinal scale of assignment were te coice of class intervals is to some extent arbitrary. Hence, tis study computes all relative severity factors on te same scale ratio to eac recorded maximum value. Te flas flood indexes suggested in tis study are determined by te average of te relative severity factors. Details of te flas flood indexing procedure are presented below Flood Runoff Hydrograps Flood runoff ydrograps are generated from a rainfall-runoff model, HEC-HMS [11], using te annual maximum precipitation event series of te Jae-cun Gauge Station in te Oui-mi River basin for past 36 years ( ). Te NRCS (Natural Resources Conservation Service) curve number metod is used for te loss rate metod and te Clark unit ydrograp is used for te transform metod. Te NRCS curve number is averaged for te study basin as 7.1, and te storage coefficient is estimated as 1.18 rs. Table 1 (column 2) sows tat te maximum peak flood discarge of m 3 /s occurs in 5 August 27 and te minimum peak flood discarge of 14.5 m 3 /s occurs in 23 August 1974 among te 36 annual maximum simulated ydrograps in te Oui-mi River basin Rising Curve Gradient (K) Baskar et al. [7] described te rising limb of ydrograps as an exponential equation using te rising curve gradient, k: kt Qt = Qe (1) were Q is te specified initial discarge in te rising limb of ydrograps, and Q t is te discarge at a later time t close to te time to peak. Tis exponential function usually used for ydrograp recession curves not only as a problem to describe te rising curve gradient in cases of double-peak ydrograps, but also as a difficulty in defining te specified initial discarge, Q, for simulated ydrograps. Hence, tis study expresses te rising limb of te simulated ydrograp using a mean slope gradient approximation: ( Qp / A) K = (2) T

5 Int. J. Environ. Res. Public Healt 211, were K is a mean slope gradient of te rising limb, Q p is te peak discarge, T is a duration time between te starting time of a flood event and te peak flow occurrence time, and A is te drainage area. Te rising curve gradient, K, is computed for te specific discarge (discarge per unit area) wit a unit of mm/r 2. Table 1. Summary of runoff and flas flood indexing caracteristics for flood events along wit rainfall caracteristics in te Oui-mi River basin. No Flood Runoff Caracteristics Flas Flood Indexing Parameters Rainfall Caracteristics Flood event date (1) Flood peak discarge Q p (m 3 /s) (2) Time to peak discarge T(r) (3) Rising curve gradient K(mm/r 2 ) (4) Peak discarge magnitude M(mm/r) (5) Flood response time T (r) (6) RK (7) Relative Severity Factors RM (8) RT (9) (1) Flas Flood Index (11) Average rainfall intensity I a (mm/r) (12) Max. 1-ourly rainfall R 1 (mm) (13) Max. 2-ourly rainfall R 2 (mm) (14) Max. 3-ourly rainfall R 3 (mm) (15) 1 6/29/ /23/ /15/ /14/ /6/ /19/ /4/ /22/ /1/ /21/ /19/ /2/ /17/ /19/ /22/ /14/ /26/ /11/ /2/ /24/ /13/ /3/ /25/ /28/ /1/ /8/ /2/ /22/ /3/ /31/ /27/ /18/ /11/ /16/ /5/ /24/ average maximum minimum Total rainfall dept R t (mm) (16) Rainfall duration time D (r) (17)

6 Int. J. Environ. Res. Public Healt 211, Te large values of te parameter K can be associated wit a rapid local flood of great volume because te rising curve gradient represents te steepness of te rising limb of flood ydrograps. Te range of values for te rising curve gradient, K, is from.3 mm/r 2 to 5.63 mm/r 2 for te Oui-mi River basin as sown in column 4 of Table 1. Te relative severity for te rising curve gradient, K, is quantified as a dimensionless index, RK, wic is a ratio of eac flood s K i to te recorded maximum value, K max : K i RK = (3) K max 3.3. Peak Discarge Magnitude (M) Baskar et al. [7] presented te flood magnitude ratio, m, wic means a ratio of te peak flood discarge to te long-term average discarge: m = Q p / Q a (4) were Q a is te long-term average discarge. Because te long-term average discarge, Q a, is not only unavailable in ungauged catcments, but also canceled out later in obtaining te relative severity factor in Equation 6, te parameter m is replaced wit te peak specific discarge magnitude, M, as: M = Qp / A (5) Te values of te peak discarge magnitude, M, varied from 3.11 mm/r to mm/r for te Oui-mi river basin as sown in column 5 of Table 1. Te relative severity factor, RM, is also computed by a ratio of eac flood event s M i to te recorded maximum value, M max M max : M i RM = (6) 3.4. Flood Response Time (T) Te flood response time, T, can be measured directly from flood ydrograps. Te flood response time, T, varied from 3 rs to 24 rs for te Oui-mi River basin as sown in column 6 of Table 1. Because a low value of T is readily associated to a ig runoff velocity causing sudden local flooding, te relative severity factor, RT, is computed by a ratio of te inverse value of eac flood event s T i to te inverse value of te recorded minimum value, T min : 3.5. Flas Flood Index T RT = (7) Te relative severity factors need to be integrated for an overall value to evaluate flas flood severity for eac flood event. Baskar et al. [7] ave presented a flas flood index, RF, te sum of te tree relative severity factors on different ordinal scale values suc as RK = 1 ~ 7, RM = 1 ~ 16, and RT = 1 ~ 1 were te coice of class intervals is to some extent arbitrary. Altoug tey applied min T i

7 Int. J. Environ. Res. Public Healt 211, systematically eac severity factor to 3 flood events from four waterseds in eastern Kentucky, te flas flood index, RF, determined by te sum of te tree severity factors is often subjected to a certain factor wit te greater scale of measurement tan oter factors. Tis study presents a flas flood index, RF, by taking te average of te tree relative severity factors on te same scale ratio to te recorded maximum value: RK + RM + RT RF = 1 (%) (8) 3 In Equation 8, te two relative severity factors suc as RK for te rising curve gradient and RT for te flood response time may represent te similar caracteristics of a flood ydrograp because a low value of te flood response time can be associated wit a ig runoff velocity leading to te steep rising limb of flood ydrograps. Tis study presents anoter relative flood severity, FI, te average of te two relative severity factors, RK for te rising curve gradient and RM for te peak discarge magnitude: RK + RM FI = 1 (%) (9) 2 Table 1 sows te two flas flood indexes, RF and FI, along wit rainfall caracteristics for te study basin. 4. Comparison of Flas Flood Indexes Tis study examines te dependence of eac of two flas flood indexes, RF and FI, on te flood ydrograp caracteristics using scatter plots and regression equations as illustrated in Figure 2 and Table 2. Te flas flood index, RF, obtained by te average of tree relative severity factors suc as RK for te rising curve gradient, RM for te peak discarge magnitude, and RT for te flood response time sows a strong dependence on te rising curve gradient, K, were a coefficient of determination, R 2, is.95, wile te relation between te flas flood index, RF, and te peak discarge magnitude, M, is relatively weak wit R 2 of.428. Te reason is because te two similar ydrograp caracteristics, K and T, are in major controlling factors determining te flas flood index, RF, rater tan te contribution of M. On te oter and, te flas flood index, FI, expressed by te average of two relative severity factors suc as RK and RM is dependent at almost even level on bot parameters, K and M, wit R 2 values of.861 and.817, respectively. Table 2. Regression equations between eac ydrograp caracteristic and eac of two flas flood indexes. Flas Flood Index, RF Flas Flood Index, FI Regression Equations R² Regression Equations R² K RF = K FI = K M RF = 1.418M FI = 2.258M Tis study as also investigated te relationsip between flas flood indexes, RF and FI, and several rainfall caracteristics suc as te average rainfall intensity, I a, te maximum rainfall depts for te 1- our, 2-our, and 3-our durations, R 1, R 2, and R 3, respectively, te total rainfall dept, R t, and te

8 Int. J. Environ. Res. Public Healt 211, rainfall duration, D for 36 annual maximum rainfall event series in te study basin. Te average rainfall intensity refers to te total amount of rainfall for a storm event divided by te duration of te storm. Te scatter plots of eac of two flas flood indexes, RF and FI, versus eac rainfall caracteristic in te study basin are illustrated in Figure 3 Figure 5. Table 3 sows te regression equations between eac rainfall caracteristic and eac of two flas flood indexes, RF and FI, in te study basin. Figure 2. Te comparison of trends between eac of two flas flood indexes, RF, (left) and FI, (rigt) and eac of two ydrograp caracteristics: (a) te rising curve gradient, K; (b) te peak discarge magnitude, M, respectively, in te Oui-mi River basin K (mm/r 2 ) K (mm/r 2 ) (a) Relation between te flas flood indexes and te rising curve gradient M (mm/r) M (mm/r) (b) Relation between te flas flood indexes and te peak discarge magnitude Figure 3. Te comparison of trends between eac of two flas flood indexes, RF, (left) and FI, (rigt) and eac of two rainfall caracteristics: (a) te average rainfall intensity, I a ; (b) te 1-ourly maximum rainfall dept, R 1, respectively, in te Oui-mi River basin I a (mm/r) I a (mm/r) (a) Relation between te flas flood indexes and te average rainfall intensity

9 Int. J. Environ. Res. Public Healt 211, Figure 3. Cont R 1 (mm) R 1 (mm) (b) Relation between te flas flood indexes and te 1-ourly maximum rainfall dept Figure 4. Te comparison of trends between eac of two flas flood indexes, RF, (left) and FI, (rigt) and eac of two rainfall caracteristics: (a) te 2-ourly maximum rainfall dept, R 2 ; (b) te 3-ourly maximum rainfall dept, R 3, respectively, in te Oui-mi River basin R 2 (mm) R 2 (mm) (a) Relation between te flas flood indexes and te 2-ourly maximum rainfall dept R 3 (mm) R 3 (mm) (b) Relation between te flas flood indexes and te 3-ourly maximum rainfall dept

10 Int. J. Environ. Res. Public Healt 211, Figure 5. Te comparison of trends between eac of two flas flood indexes, RF, (left) and FI, (rigt) and eac of two rainfall caracteristics: (a) te total rainfall dept, R t ; (b) te rainfall duration, D, respectively, in te Oui-mi River basin R t (mm) R t (mm) (a) Relation between te flas flood indexes and te total rainfall dept D (r) D (r) (b) Relation between te flas flood indexes and te rainfall duration Table 3. Regression equations between eac rainfall caracteristic and eac of two flas flood indexes. Flas Flood Index, RF Flas Flood Index, FI Regression Equations R² Regression Equations R² I RF 3.252I FI 4.44I a = a = a R 1 RF.9R FI 1.195R = = R 2 RF.597R FI.83R = = R 3 RF.484R FI.682R = = R t RF =.81Rt FI = 195 Rt D RF =.164D 6.289D FI =.187D 6.495D Overall, te flas flood index, FI, sows a muc stronger relation to some rainfall data wit relatively ig coefficients of determination, R 2, as compared wit te relationsip of te flas flood index, RF, to rainfall caracteristics. Te flas flood index, FI, wic avoids double-counted relative severity factors wit similar caracteristics is adequate to estimate te relative flood severity in tis study basin. It is observed tat te Oui-mi River basin as a ig linear relation between te flas flood index, FI, and te 3-ourly maximum rainfall dept, R 3, wit te coefficient of determination, R 2 of.86, wile te total rainfall amount, R t, and duration, D, sow no evident relation to bot flas

11 Int. J. Environ. Res. Public Healt 211, flood indexes, RF and FI, as illustrated in Figure 5. Tis result supports te notion tat local flas floods in small waterseds are mainly caused by eavy or excessive rainfall in a sort period of time. Figure 6 illustrates long-term trends of te averaged rainfall dept in te study watersed for te rainy season during June September in wic all annual maximum flood events for te past 36 years occur. Te growing trend of te 7-year centered moving average (CMA) clearly reveals a tendency of steady increase in rainfall amounts over te years. Also, Figure 7 denotes interannual variation of te rainfall intensity, I, and te flas flood index, FI. Te interannual variability of te new flas flood index, FI, well follows te increasing trend of te rainfall intensity for te past 36 years during in tis study site. Figures 6 and 7 sow periodic trends tat increase in annual series of bot te averaged wet season rainfall amount and te maximum rainfall intensity, and Figure 7 demonstrates tat te rainfall intensity can be used as a key indicator to determine and forecast te relative flas flood severity. Since it implies tat flas flooding may occur more severely and frequently due to te increasing trend of te eavy rainfall intensity caused by global climate canges, te vulnerability assessment in design storms to te extreme flas flood is required to deal wit current and future flooding risks. Figure 6. Interannual variation of te rainy season rainfall dept during in te Oui-mi River basin. 5 4 Rainfall CM A of Rainfall Rainfall (mm) Figure 7. Interannual variation of te rainfall intensity, I, and te flas flood index, FI, during in te Oui-mi River basin. Year I FI 8 I (mm/r) Year

12 Int. J. Environ. Res. Public Healt 211, Assessment of Vulnerability in Design Storms Based on results sowing tat te rainfall intensity can capture te relative flas flood severity represented by te new flas flood index, FI, tis study as also determined te flas flood index from probability rainfall data in te study basin to provide flas flood index-duration-frequency (FI-D-F) relation curves developed from rainfall intensity-duration-frequency (I-D-F) relation curves. FI-D-F curves are intended to evaluate vulnerability in a design flood to te extreme flood condition and residual flood risk of bot existing and planned flood control facilities. Te Gumbel distribution is selected troug estimation of parameters and goodness fit test for te several probability distributions of annual maximum rainfall data in te Oui-mi River basin. Table 4 summarizes te probability rainfall data in return periods of 2, 3, 5, 1, 2, 5, 1, and 2 years and for duration times of 1, 3, 6, 12, 18, and 24 ours, respectively. Te Huff distribution [12] is used for te temporal distribution of te probability rainfall data. Te Huff distribution is presented as cumulative percentages of total duration and total rainfall accumulation tat consists of four quartile patterns for te bulk rainfall of te storm event as sown in Table 5. Different families of Huff distribution curves are applicable for different drainage waterseds, and te 3rd quartile is used for te design flood in te Oui-mi river basin [1]. Duration Table 4. Probability rainfall data in te Oui-mi river basin. Dept of precipitation (mm) 2 yrs 3 yrs 5 yrs 1 yrs 2 yrs 5 yrs 1 yrs 2 yrs 1 rs rs rs rs rs rs Table 5. Huff s cumulative rainfall curve in te Oui-mi river basin. Cumulative time (%) Cumulative Rainfall (%) 1st quartile nd quartile rd quartile t quartile Te flas flood index, FI, is computed by ydrograps generated from a rainfall-runoff model, HEC-HMS [11], for te probability rainfall data wit respect to eigt return periods (2, 3, 5, 1, 2, 5, 1, and 2 yrs) and six temporal durations (1, 3, 6, 12, 18, and 24 rs), as sown in Table 6. Te flas flood index, FI, is determined by te average of two relative severity factors suc as RK and RM, ratios of te rising curve gradient, K, and te peak discarge magnitude, M, from eac design flood ydrograp to te recorded maximum values, 5.63 mm/r 2 of K max and 39.41mm/r of M max,

13 Int. J. Environ. Res. Public Healt 211, respectively. Te maximum FI is 179.7% for a return period of 2 year and duration of tree rs, and te minimum FI is 14.89% for a return period of two years and duration of 24 rs among 48 design floods (eigt return periods and six temporal durations). Figure 8(a) indicates I-D-F curves, and (b) illustrates FI-D-F curves for te relation between te flas flood index, FI, and te rainfall duration, D, wit respect to te eigt return periods in te study basin. According to te basic plan report for te Oui-mi River maintenance works [1], some structural flood control projects are planned for te 1 year design flood in te Oui-mi River basin. In Figure 8(b), FI values are greater tan 1% for te 1 year design storm in duration of less tan 12 rs. It means tat te flood control facilities designed under tese conditions, for example river levee improvement works in te Oui-mi River basin, may ave countermeasure ability towards te recorded flas floods in tis region. It is also implied tat drainage pipe lines designed for 5 to 1 year floods may need non-structural flood mitigation plans as well as structural alternative plans in order to cope wit residual flood risk to te extreme flas flood tat occurs frequently in tese days. Table 6. Summary of runoff and flas flood indexing caracteristics for design floods along wit rainfall caracteristics in te Oui-mi River basin. Flood Runoff Caracteristics Flood Indexing Parameters Rainfall Caracteristics Design Flood Time to Rising Peak Relative Flas Average Total Rainfall No flood peak peak curve discarge Severity Flood rainfall rainfall duration discarge discarge gradient magnitude Factors Index intensity dept time (1) Q p (m 3 /s) T (r) K (mm/r 2 ) M (mm/r) RK RM I a (mm/r) R t (mm) D (r) (2) (3) (4) (5) (6) (7) (8) (9) (1) (11) 1 2year 1r year 3r year 6r year 12r year 18r year 24r year 1r year 3r year 6r year 12r year 18r year 24r year 1r year 3r year 6r year 12r year 18r year 24r year 1r year 3r year 6r year 12r year 18r year 24r year 1r

14 Int. J. Environ. Res. Public Healt 211, Table 6. Cont. Flood Runoff Caracteristics Flood Indexing Parameters Rainfall Caracteristics Design Flood Time to Rising Peak Relative Flas Average Total Rainfall No flood peak peak curve discarge Severity Flood rainfall rainfall duration discarge discarge gradient magnitude Factors Index intensity dept time (1) Q p (m 3 /s) T (r) K (mm/r 2 ) M (mm/r) RK RM I a (mm/r) R t (mm) D (r) (2) (3) (4) (5) (6) (7) (8) (9) (1) (11) 26 2year 3r year 6r year 12r year 18r year 24r year 1r year 3r year 6r year 12r year 18r year 24r year 1r year 3r year 6r year 12r year 18r year 24r year 1r year 3r year 6r year 12r year 18r year 24r average maximum minimum recorded max Figure 8. (a) I-D-F curves and (b) FI-D-F curves for te Oui-mi River basin. 1 (a) 1 (b) I (mm/r) YR 3 YR 5 YR 1 YR 2 YR 5 YR 1 YR 2 YR YR 3 YR 5 YR 1 YR 2 YR 5 YR 1 YR 2 YR D (r) D (r) 5. Summary and Conclusions Tis study as modified te flas flood index by Baskar et al. [7], and developed a new flas flood index determined by te average of relative severity factors wit te same scale ratios of eac flood event s caracteristics to te recorded maximum values in order to evaluate te relative severity of floods to extreme flas floods. New relative severity factors were presented in tis study to describe te caracteristics of simulated runoff ydrograps for ungauged waterseds; te mean slope gradient of te rising limb and te peak specific discarge are substituted for te exponential curve gradient and

15 Int. J. Environ. Res. Public Healt 211, te flood magnitude ratio, respectively, suggested by Baskar et al. [7]. Te new flas flood indexing metod was implemented by computing a dimensionless index for caracteristics of ydrograps generated from a rainfall-runoff model for te long-term observed rainfall data in a small illy ungauged catcment, te Oui-mi River basin in Korea. Te flas flood index, FI, te average of te two relative severity factors, RK and RM, provides a stronger relation to some rainfall caracteristics as compared wit te flas flood index, RF, including te two similar relative severity factors, RK and RT, along wit RM. Te trend between te flood flas index, FI, and te rainfall over a sort interval, 3-ourly maximum rainfall dept, sows te best-fit line, wile te flood flas index, FI, sows no evident relation to te total rainfall amount. It illustrates tat eavy or excessive rainfall in a sort period of time is a primary cause of local flas flooding in small waterseds. Te best-fit regression equation between te new flas flood index, FI, and a certain rainfall caracteristic can provide te basis database for forecasting a local flood severity directly from rainfall data in small ungauged catcments were te flood response time is quite sort. Tis study as also estimated te flas flood index, FI, from probability rainfall data wit respect to eigt return periods (2, 3, 5, 1, 2, 5, 1, and 2 yrs) and six temporal durations (1, 3, 6, 12, 18, and 24 rs) in te study basin. Te flas flood index, FI, for eac design storm is determined by te two relative severity factors, RK and RM wit ratios to te recorded maximum values, and ten te plot of FI-D-F relation is developed to assess vulnerability in a design flood to extreme flas floods. FI-D-F curves illustrate tat major flood control facilities designed for te 1 year frequency rainfall in a duration of less tan 12 rs may ave flood mitigation ability towards te recorded flas floods in tis study basin. However, small or midsize facilities suc as drainage pipe lines designed for 5 to 1 year frequency storms may ave significant vulnerability to extreme flas floods in tis region. Te FI-D-F relation curves suggested in tis study is expected to be one of scientific bases for decision makers to select structural or non-structural alternative flood mitigation plans against flooding disasters. Tis study is an initial effort to evaluate vulnerability in design floods for bot existing and planned flood control facilities in order to cope wit te residual flood risk of extreme flas floods. Te future study needs to incorporate various ydrometeorological perspectives, especially focused on te use of radar information, in te implementation of te new developed metodology for more precise and general flas flood predictions. Acknowledgements Tis researc was supported by te Yeungnam University researc grants in 29 (29A3816). References 1. Korea National Emergency Management Agency. Te Annual Natural Disaster Bulletin; Korea National Emergency Management Agency: Seoul, Korea, Intergovernmental Panel on Climate Cange (IPCC). Fourt Assessment Report, Climate Cange 27 Summary for Policymakers; IPCC: Geneva, Switzerland, Doswell, C.A., III. Flas Flood-Producing Convective Storms. In Proceedings of te U.S.-Spain Worksop on Natural Hazards, Barcelona, Spain, 8-11 June 1993; pp

16 Int. J. Environ. Res. Public Healt 211, Lapenta, K.D.; McNaugt, B.J.; Capriola, S.J.; Giordano, L.A.; Little, C.D.; Hrebenac, S.D.; Carter, G.M.; Valverde, M.D.; Frey, D.S. Te callenge of forecasting eavy rain and flooding trougout te Eastern Region of te National Weater Service. Part I: Caracteristics and events. Weater Forecast. 1995, 1, Opitz, H.H.; Summer, S.G.; Wert, D.A.; Snyder, W.R.; Kane, R.J.; Brady, R.H.; Stokols, P.M.; Kul, S.C.; Carter, G.M. Te callenge of forecasting eavy rain and flooding trougout te Eastern Region of te National Weater Service. Part II: Forecast tecniques and applications. Weater Forecast. 1995, 1, Kyiama, G.K.A. Monitoring and Caracterization of Flas Floods. MS tesis, Department of Civil Engineering, University of Louisville: Louisville, KY, USA, Baskar, N.R.; Frenc, B.M.; Kyiama, G.K. Caracterization of flas floods in Eastern Kentucky. J. Hydrol. Eng. 2, 5, Jung, J.C. Te Study on Estimation of te Flas Flood Index for te Bo-cun River Basin. MS tesis, Department of Civil Engineering, Suwon University: Kyeonggi Province, Korea, Kim, B.S.; Kim, H.S. Estimation of te flas flood severity using runoff ydrograp and flas flood index. J. Korea Water Resour. Assoc. 28, 41, Wonju City. Te Basic Plan Report for te Oui-mi River Maintenance Works; Wonju City: Gangwon Province, Korea, USACE. Hydrograp Modeling System; Tecnical Reference Manual, Hydrologic Engineering Center: Davis, CA, USA, Huff, F.A. Time distribution of rainfall in eavy storms. Water Resour. Res. 1967, 3, by te autors; licensee MDPI, Basel, Switzerland. Tis article is an open access article distributed under te terms and conditions of te Creative Commons Attribution license (ttp://creativecommons.org/licenses/by/3./).