Вы находитесь на странице: 1из 12

Nat Hazards DOI 10.

1007/s11069-011-9852-2 ORIGINAL PAPER

A DEM-based evaluation of potential ood risk to enhance decision support system for safe evacuation
Jinyoung Kim Yuji Kuwahara Manish Kumar

Received: 5 January 2011 / Accepted: 8 May 2011 Springer Science+Business Media B.V. 2011

Abstract A practical, DEM-based practical method is proposed to enhance ood risk management in uvial areas by quantifying relative risk as a function of vulnerability to inland and evacuation difculty. Both measures are based mainly on the topography of the region, so the method does not require detailed data on the physical characteristics of the land. First, we use the deterministic 8-node method on a digital elevation map (DEM) to trace storm waterways. Second, we repeat the process on a reversed DEM to trace evacuation routes that avoid the waterways and zones dangerously close to the rivers. Finally, on the basis of such two ow lines of evacuee and storm water, we proposed the protocol to evaluate the ood risk at every point on the map taking into account both the minimum time required for oodwater to arrive and duration of an evacuation from that location. The time that must be allocated for safe evacuation is dened as the potential ood risk of evacuation (PFRE). The method is demonstrated on a uvial area of the Kaki River in Nagaoka city, Japan. In addition, we illustrated the application of the PFRE map to divide the region into areas of greater or lesser evacuation urgency. Keywords Digital elevation model (DEM) Evacuation Flood preparedness Flood hazard Fluvial area Potential ood risk of evacuation (PFRE)

1 Introduction The Asia and Pacic countries are situated in the worlds hazard belts and are subject to oods, droughts, cyclones, earthquakes, windstorms, tidal waves, and landslides. The
J. Kim (&) Department of Urban Engineering, The University of Tokyo, Tokyo 113-8656, Japan e-mail: gamnamuu@gmail.com Y. Kuwahara Ibaraki University, Ibaraki, Japan M. Kumar Tezpur University, Napaam, Assam 784 028, India

123

Nat Hazards

region has suffered 50% of the worlds major natural disasters (ESCAP 1995). Japan, particularly, not only suffers from devastating earthquakes as the case of March 11, 2011, but also receives twice of the world average annual precipitation (due to frequent occurrence of torrential rainfall caused by typhoon). In such situation, Japan has to heavily rely on the emergency preparedness and response for minimizing the casualty and increasing its combating efciency. In general, ooding accounts for about one-third of all deaths, injuries, and property damage caused by natural disasters (Akew 1999). All river plain areas are physically, structurally, and socioeconomically vulnerable, so require meticulous care for efcient ood management. For example, although ood-prone river plains constitute only 10% of the Japans entire area, they contain 50 and 75% of the nations total population and property, respectively (Guidelines for ood hazard mapping 2005). Therefore, immense effort has been spent on the prevention and/or mitigation of oods in recent decades, largely based on the ideas of risk assessment and risk management. Previous researchers have used various denitions for ood risk and guidelines for risk management, chosen to suit their specic purposes and regions of study (Parthasarathy et al. 1987; Trinic 1997; Oriola 1994; Hogue et al. 1997; Okoduwa 1999; Schanze et al. 2008). Among the many products of this research, hazard maps and evacuation plans are the tools most likely to offer practical and accessible information for residents and support disaster management authorities improving their preparedness and response programs. However, few studies have attempted to link ood hazard maps directly to evacuation planning. Many have focussed on the physical, structural, or socioeconomic aspects of a ood: e.g., trafc conditions, shelter locations, evacuee behaviors, and the capacity of shelters (Feng and Wen 2003; Sherali et al. 1991). Although people are supposed to initiate evacuation before a ood becomes critical, in accordance with a pre-evacuation plan established by the authorities, river plains always face the risk of unexpected inland inundation and river ooding, during evacuation. Such problems could arise due to an embankment failure or an error in the monitoring/forecasting system, for example. In addition to their high latent ood risk, uvial areas are currently facing increased uncertainty and vulnerability due to many factors. For instance, urbanization and climate change are pushing Japans disaster authority and residents to renovate and improve the conventional ood-ghting strategy. Recent dynamic models take into account many physical aspects of ooding (Liu et al. 2006; Hsu et al. 2003), but require considerable expertise in hydrological data evaluation, model design, and the efcient management of large spatial and temporal dataset. Furthermore, the uvial plains of surrounding small- and medium-sized rivers often lack sufcient physical data to carry out a dynamic simulation. These limitations have prevented dynamic modeling from gaining widespread acceptance as an efcient tool for ood risk assessment and management. Further, their reliability and usability are still in question (Merz et al. 2010). Even if a ood hazard map and evacuation plan are prepared based on the result of a dynamic simulation, these products are seldom understood correctly by the residents. (People tend to judge ood hazard only in terms of the ood depth depicted in a ood hazard map, ignoring other factors; see Katada et al. 2004). Geographic Information Systems (GIS), being practical and widely available, are gaining favor as a technique to assess ood risk (Ologunorisa and Abawua 2005). In particular, digital elevation models (DEMs) have made it possible to estimate hydrological ow without recourse to detailed simulation (Boonklong et al. 2007; Mark 1988). This study presents a practical, DEM-based method to assess the risk due to river oods in a context of potential inland inundation and evacuation activity. Compared with a dynamic simulation, this method greatly simplies the ood ow model and evacuee

123

Nat Hazards

behavior patterns. While it does not investigate physically accurate values or attempt to predict the exact magnitudes of torrent processes, the proposed method does provide qualitative and explicit criteria for ood preparedness. It can be used to set priorities in an evacuation plan to coordinate pre-rescue activity and to enhance the risk awareness of residents by presenting the expected time scale of the ooding at their location.

2 Study area This study concerns the most urbanized region of Nagaoka city, with a surface area of 20.7 km2 (the central coordinate is approximately 37270 900 N, 138510 200 E). The study area is located in the alluvial plain of the Shinano River, Japans biggest river in terms of length and ow rate. The administrative district and terrain of Nagaoka city are represented in Fig. 1a; the study area is shaded. There are three other rivers in and around the study area: the Kaki, the Suyosi, and the Ota. All are tributaries of the Shinano River, as shown in Fig. 1b. The Kaki River basin is highly vulnerable to ood damage due to its central location, highly urbanized surroundings, and low conveyance of channel as depicted in Fig. 1c. In addition, the evacuation routes in case of ooding restricted by the presence of the other three rivers. From the viewpoint of Kaki River ood management, these Shinano, Suyosi, and Ota present serious barrier not just to evacuation, but also to rescue operations and relief work. Flood hazard maps for the four rivers were released to the public in 2008. The maps contain information on ood depth and evacuation directions to relatively hilly inland areas. The disaster prevention authority has designated 75 buildings as evacuation facilities in the study area. Among these, schools account for more than 44 percent of the total number. The hydrological contents of the ood hazard maps are summarized in Table 1.

Fig. 1 Terrain of Nagaoka city and location of study area (a), rivers and evacuation facilities in the study area (b), and aerial photography of the study area (c)

123

Nat Hazards Table 1 Brief of hydrological contents in ood hazard map of four rivers River name Shinano River Ota River Suyosi River Kaki River
a

Flood deptha (m) 2.05.0 0.5 0.5

Design rainfall 171 mm/2 days 245 mm/day 261 mm/day 184 mm/day

Return period 150-years 100-years The biggest rainfall ever 50-years

Dominantly expected ood depth in the study area

3 Materials and methods This study employs a digital elevation model (DEM) of the area. Surface noise from buildings, yovers, and trees was removed by ltering the data based on aerial laser scanning with 2-m spatial resolution. The data were produced by the Geospatial Information Authority of Japan in 2004 and include all rivers wider than 5 m and larger than 25 m2 in terrestrial water area. The GIS software package used was ERDAS Imagine Professional (version 8.7) (ERDAS Inc. 5051 Peachtree Corners Circle Norcross, GA 30092-2500 USA). 3.1 Denitions and assumptions Throughout this paper, we use the following notation. Flood arrival time (AT): the time required for ooding to move from a river point to a processing cell. It is calculated for all river points that could contribute to ooding in the cell. Flood risk from river (FT): the shortest ood arrival time AT for a processing cell. Inland inundation: A linear or polygonal area ooding based on water ow rate and the local elevation map. Evacuation route: the line connecting a processing cell to its highest neighboring cell. Evacuation time (ET): the cumulative time required to shift from a cell of interest to a processing cell. Evacuation ending time (ETT): the total time required to nish evacuation from a cell of interest to the nearest shelter. Intermediate risk of evacuation (IRT): the time difference between the cumulative evacuation time (ET) and the ood risk from river (FT) at a processing cell; IRT = ET - FT. Potential ood risk of evacuation (PFRE): The maximum value of IRT from a cell of interest to evacuation facility. The assumptions that were made in this study to isolate the mechanism of overland ow distribution from the physical surface conditions, in order to simplify our model of hydraulic processes based on the DEM, are as follows: (1) (2) The ow from a grid cell to its neighboring cell is dependent upon the topography of the surface. The physical conditions of the surface (inltration capacity, land use, evaporation) were not considered.

123

Nat Hazards

DEM based model


Aspect of inland stormwaterflow River stage Topographic condition Aspect of riverflood Evacuation route minimizing inland flood risk

Computation of the risk encountering river flood on evacuation route

Potential flood risk of evacuation

Fig. 2 Scheme of DEM-based potential ood risk assessment

(3)

River water ows over embankments with equal hydraulic water head over the whole river.

Our approach to DEM-based ood risk assessment is represented in Fig. 2. While water level has been used to indicate danger in traditional ood control strategies, our scheme integrates the movement of water and people with the topography of the surface. 3.2 Storm waterways prediction using the D-8 method Lin et al. (2002) proposed that a watershed can be described as the locus of points within an area where runoff produced inside the area perimeter moves into a single watershed outlet. This approach presumes that the topographic form of a surface facet dictates the ow distribution and is known as the single ow direction (SFD) method. The ow direction is the steepest downward slope among the eight neighboring facets in a 3 9 3 pixel window centered on the cell of interest. The deterministic eight-node (D8) method is a variant of the classical SFD method. In order to be able to construct a spatially connected series of cells along the ow from any processing cell to an outlet, any sinks and peaks within the watershed (i.e., for which all eight neighbors are higher or lower in elevation, respectively) are considered noise. The elevations of such cells are replaced with the lowest elevation among the eight neighbors. The ow line determined by the D8 method is regarded as the route taken by surface storm water. 3.3 Identication of evacuation routes through reversed DEM The most common safety measure against ooding is to run to higher ground as swiftly as possible. The most likely evacuation route is therefore a line connecting hills, such as a ridge crest. MacMillan and Shary (2004) proposed an algorithm to extract boundaries between hills and valleys according to threshold of slope. In that algorithm, the original DEM data were reversed by multiply it with -1. In order to determine the line that connects a cell of interest to the highest place within the same watershed in this study, we modied the DEM reversing formula as in Eq. 1 and once more applied the D8 method into reversed DEM data. 1 RDEMi;j DEMMAX DEMi;j 1

RDEM(i,j) is the reversed elevation at processing point (i, j), DEM(i,j) is the original elevation of processing point (i, j), and DEMMAX is the maximum elevation in the study area.

123

Nat Hazards

DEM

Reversing process

67 56 49 46 50

0 11 18 21 17 14 23 30 29 19 9 12 45 36 43 6 20 46 51 48 14 33 55 56 55
Filled reversed DEM

Reversed DEM

53 44 37 38 48 58 55 22 31 24

Fill sinks and peaks Filled DEM Filled reversed DEM Establish restriction D8 method Direction to the lowest neighbour cell Storm waterways Direction to the highest neighbour cell Evacuation routes

61 47 21 16 59 53 34 12 11 12
Filled raw DEM

Direction of flow

Direction of evacuation

Strom water ways Evacuation route

Fig. 3 Procedures to determine storm waterways and evacuation routes (left) and illustration of its application in raster DEM (right)

n=n+1

Point n IRT(1) = FT(1)


k=k+1

2 1

IRT(start) = FT(1) = 2

Start
at Point (n)

t=5

Computation of E T(k) and F T(k)

2
IRT(k) = ET(k) F T(k)

ET(2) = 5 FT(2) = 6 IRT(2) = 5-6 = -1

6
Is there evacuation facility available within 300m?
Yes No

t=5

3 7

ET(3) = 10 FT(3) = 7 IRT(3) = 10-7 = 3 ET(k) = 25 FT(k) = 20 IRT(k) = 25-20 = 5


Goal

PFRE(n) = Max (IRT(k), k=1,k) n = End of map?


Yes

t=15 k 20

End

PFRE = Max ( IR T(k), k=1, k ) = 5

Fig. 4 Flow for the calculation of PFRE and illustration of the sequential step for PFRE calculation

Although storm water ow and evacuation routes are opposing processes on a raster DEM, the two paths can be obtained by the same algorithm. By reversing the elevation data, the algorithm identies the steepest gradient toward higher ground. In this study, the storm water ow paths are taken as spatial constraints on potential evacuation routes when applying the D8 method. In this way, no evacuation route crosses a waterway. Our procedure to determine evacuation routes in raster DEM is illustrated in Fig. 3. 3.4 Assessment of river ood hazard (AT) Consider the point labeled n in Fig. 4. The ood arrival time (AT) from point R(1) on the river is calculated as follows:

123

Nat Hazards

AT 1 RTf R Ts

Tf stands for the time required for the ood to ll each concave surface between the river and point n. Ts represents the time required for oodwater overowing a concave surface to run downhill to the next concave surface. Ts is then calculated as per the Eq. 3: Ts RV = QIN 3

where V is the total volume of a concave surface between the river and point n. QIN is the discharge volume owing over the riverbank, which can be represented as a sharp-crested rectangular weir. The depth of overow from the river is assumed to be 0.5 m higher than the elevation of the cell adjacent to the river. This is adopted from dominant ood depth in the ood hazard map of Nagaoka city. The discharge over the weir (QIN) is dened as per the Eq. 4: QIN
1 2 1 cd b 2g2 h2 3

where QIN = ow rate (m3/s), cd = coefcient of discharge (1.84), b = thickness of weir (2 m; this is the size of a raster DEM cell), h = head above crest of weir (0.5 m), g = gravitational constant (9.81 m/s2). Tf is calculated from the Manning equation presented as Eq. 5: Q 1 2 1 A R3 S2 2 5

where Q = ow rate (= QIN, m3/s), n = Mannings roughness coefcient, R = hydraulic radius (m), S = slope. After repeating these calculations for each river point (i.e., from 1 to 7 in Fig. 4), the minimum AT value is regarded as the highest potential ood risk (FT) from the river at point n which can be represented as Eq. 6. FT n Min AT m; m 1; 7 6

In other words, FT(n) is the smallest estimated time for oodwater originating from a point on the river to arrive at point n. 3.5 Calculation of evacuation ending time The total evacuation time of a cell is dened as the time required for a person on foot to reach the nearest safe facility within 300 m of an evacuation route. The 300 m buffer is added because evacuation facilities are marked with a point in the GIS data, not a cell. In this way, we also avoid generating any evacuation routes that do not lead to a facility. Assuming an average walking speed of 1.42 m/s (Browning et al. 2006), the travel time to a horizontal or vertical neighbor cell is 1.4 s and the travel time to a diagonal cell is 2.0 s. 3.6 Assessment of potential ood risk of evacuation The potential ood risk of evacuation (PFRE) at a cell is dened as the time required between the onset of ooding and complete evacuation with no person encountering any river ood risk on the evacuation route. It is selected as the maximum time difference

123

Nat Hazards

between the cumulative evacuation time (ET) and the ood risk from river (FT). Figure 7 describes in detail the algorithm used to calculate the PFRE at an arbitrary cell (n).

4 Results 4.1 Evacuation route considering hazard of inland ood The ow direction is calculated in every cell of the study area. Figure 5a shows the main waterways (gray lines). In Fig. 5b, we shade the regions within 300 m of the Kaki River and within 100 m of the other two rivers. These areas are considered sufciently dangerous that during an evacuation, people are not allowed to approach. This rule is based on an existing guideline for ood hazard maps in Japan. The 100-m margin around other rivers was established to prevent embankments from being used as evacuation routes. Taking the dangerous zones adjacent to rivers and inland water ows as constraints, we then calculate the evacuation direction for each cell using the reversed DEM. Fig. 6 shows the evacuation routes overlapped with the given constraints such as inland water ows and dangerous zones. The enlarged part of Fig. 6 shows a bridge being used as an evacuation route. 4.2 Flood arrival time and evacuation time at every point Figure 7a shows the distribution of ood risk (FT) over the whole study area. Areas marked with a dotted line are considered prone to ooding in the existing ood hazard map. Although ooding is not expected in the northern part of the study area and marked as less vulnerable region in existing hazard map, our algorithm reveals that the longer evacuation times from this region increase the risk to inhabitants when we considered ood risk at upstream. Figure 7b shows the distribution of evacuation times from each cell. The thumbtacks mark facilities used for evacuation. Since the evacuation facilities were designed to serve
R(1)

R(2)

QIN

R(3) R(4) n

R(5)

R(1)

QIN

Tf1

Ts1 Tf2 Ts2 Tf3


n

River
R(6)

V1
R(7)

V2

V3

(Section 1)

Fig. 5 Conceptual illustration for evaluating river ood risk

123

Nat Hazards

Channel

Channel

Bridge

Evacuation routes Storm waterways Dangerous area of 300 m from Kaki River 100 m from other rivers

Fig. 6 Evacuation routes detoured around inland storm waterways and river-side area

in various kinds of disasters, not every facility is appropriate for oods in terms of elevation and geographical location. In this analysis, only sixteen facilities out of seventy-ve were found to be useful for ood evacuation. 4.3 Potential ood risk of evacuation at every point The PFRE calculated over the whole study area is illustrated in Fig. 8a. As an example of how this information might be applied, the study area is classied into 10 regions and 7-risk levels, based on PFRE contour line separated by 10 min. Since the PFRE value is based entirely on topography, the boundaries between the risk regions have been adjusted to consider the efciency of using road networks in a real situation. Specically, each segment of the PFRE contours is replaced with the nearest road having more than two lanes. Fig. 8b shows all ten risk regions. The letters L and R stand for the left and right bank of the Kaki River, and the number represents the evacuation priority. The contents of each area are summarized in Table 2. The yellow region with PFRE values less than zero is established as a buffer zone between the ooded and safe areas. Advance bases can be

123

Nat Hazards

Flood area expected in existing hazard map Shelters used in evacuation Time legend (min) (FT)
0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 than 60

(ETT)
than 60 50 - 60 40 - 50 30 - 40 20 - 30 10 - 20 0 - 20

(a)
Fig. 7 Flood arrival time from river (FT,) (a) and total evacuation time (ETT) (b)

(b)

established in this zone to control the trafc of refugees and support early ood management activity.

5 Discussion This paper has proposed a practical, DEM-based method to assess the ood risk associated with water ow and evacuation times in a local river plain. It is well suited to small- and medium-sized rivers, which commonly lack sufcient data for detailed ow simulations. Our algorithm is based upon the premise that while river ooding is a process with some uncertainty and unpredictability, by applying uniform ow conditions over a region we can fairly evaluate the relative risk. In this study, we emphasized explanatory depiction of ood risk that requires less physical data and expertise in hydrological process than those of existing dynamic modeling. Instead of predicting physically accurate ooding processes, this approach is designed to provide a straightforward criterion of ood response for individual as well as community-based risk management in terms of the spatial distribution of potential risk. The criterion is based on the coupled information of evacuation activity and ood risk by river and inland inundation.

123

Nat Hazards

Kaki River PFRE legend 112 -421 Road Road used for evacuation

(a)
Fig. 8 PFRE map (a) and an example of its application (b)

(b)

Table 2 Example of area management using potential ood risk

Left bank area L5 L6 L7

Right bank area R1 R2 R3 R4 R5 R6 R7

PFRE value 60 5060 4050 3040 2030 1020 010

Urgency High

Within our study area, the potential ood risk of evacuation (PFRE) varies between -421 and 112. This range implies that approximately 2 h can be afforded to conduct emergency preparedness and response activities before the ood event. However, the qualitative distribution of potential ood risk on a regional map is much more signicant information than the broad physical estimate of 2 h. The area can be classied into several groups on the basis of risk, and evacuation activities prioritized to regions with greater risk. This quantitative basis for risk management can also benet residents by strengthening awareness and motivating them to take preventive actions.

123

Nat Hazards

In conclusion, we wish to emphasize that ood risk assessments should take into account both physical hazard and the vulnerability of the population. In this paper, we include evacuation times in the risk equation by analyzing the topography of the region. GIS techniques can also integrate geomorphological, hydrological, meteorological, and socioeconomic variables, thereby extending the approach to other vulnerabilities such as land use and population.
Acknowledgments This work was supported by the Institute for Global Change Adaptation Science (ICAS) of Ibaraki University, Japan. We thank Mr. Miyagi for his initial cooperation. All the concerned parties of Nagaoka city are also gratefully acknowledged.

References
Akew AJ (1999) Water in the international decade for natural disaster reduction. In: Leavesley et al (ed) Destructive water: water-caused natural disasters, their abatement and control. IAHS. Publication No. 239 Boonklong O, Jaroensutasinee M, Jaroensutasinee K (2007) Computation of D8 ow line at Ron Phibun area, Nakhon Si Thammarat, Thailand. Proc World Acad Sci Eng Tech 21:377380 Browning RC, Baker EA, Herron JA, Kram R (2006) Effects of obesity and sex on the energetic cost and preferred speed of walking. J Appl Physiol 100:390398 ESCAP 1995. Asian and pacic ministerial conference in preparation for the world summit for social development: Proceedings (ST/ESCAP/1475) Feng CM, Wen CC (2003) Trafc control management for earthquake-raided area. J Eastern Asia Soc Transp Stud 5:32613275 Guidelines for ood hazard mapping (2005) River bureau, Ministry of land, infrastructure, transport and tourism, Japan (in Japanese) Hogue MM, Abdullah A, Khan MSA (1997) Storm surge ooding in Chittagong city and associated risks. IAHS Publications No. 239, 115122 Hsu M, Fu J, Liu W (2003) Flood routing with real-time stage correction method for ash ood forecasting in the Tanshui River, Taiwan. J Hydrol 283:267280 Katada T, Kodama M, Saeki H (2004) Study on residents recognition of a ood hazard map and promotion measures of its recognition. Ann J hydraul Eng 48:433438 (in Japanese) Lin CY, Lin WT, Chou WC (2002) Soil erosion prediction and sediment yield estimation: the Taiwan experience. Soil Tillage Res 68:143152 Liu Y, Hatayama M, Okada N (2006) Development of an adaptive evacuation route algorithm under ood disaster. Ann Dias Prev Res Inst, Kyoto Univ 49B:189195 Mark DM (1988) Network models in Geomorphology, automated basin delineation from digital elevation data. Geo Process 2:299311 Merz B, Hall J, Disse M, Schumann A (2010) Fluvial ood risk management in a changing world. Nat Hazards Earth Syst Sci 10:509527 Okoduwa AI (1999) An application of GIS to ood prediction: a case study of Benin city, Nigeria. Unpublished B Sc. Dissertation, Department of Geography and Planning University of Benin, Nigeria Ologunorisa TE, Abawua MJ (2005) Flood risk assessment: a review. J Appl Sci Environ Manag 9(1):5763 Oriola O (1994) Strategies for combating urban ooding in developing countries: a case study from Ondo. The Environmental 14(1):5762 Parthasarathy B, Sontakke NA, Monot AA, Kothawale DR (1987) Droughts/oods in the summer monsoon season over different meteorological sub-divisions of India for the period 18711982. J Climatol 7:5770 Schanze J, Schwarze R, Cartensen D, Deilmann C (2008) Analyzing and managing uncertain futures of large-scale uvial ood risk systems. In: Managing ood risk, reliability and vulnerability, Proceedings of the 4th International Symposium on Flood defense, Toronto, Canada, 33, 68 Sherali HD, Carter TB, Hobeika AG (1991) A location-allocation model and algorithm for evacuation planning under Hurricane/ood conditions. Transp Res B 25(6):439452 Trinic D (1997) Hydrological analysis of high ows and oods in the Sava river near Zagreb, Croatia, IAHS Publications No. 239, 5158

123

Вам также может понравиться