Cyclone Evacuation Route Identification Using GIS for Coastal Communities in Mirzaganj Union , Patuakhali , Bangladesh

Cyclone Evacuation Route Identification Using GIS for Coastal Communities in Mirzaganj Union, Patuakhali, Bangladesh Maruf Billah , Sraboni Sarker 2 and Mehedi Ahmed Ansary 2, 1 Graduate student, Institute of Water and Flood Management, BUET, Dhaka, Bangladesh; marufbillah413@gmail.com 2 Undergraduate student, Faculty of Disaster Management, PSTU, Patuakhali, Bangladesh; srabonisarker8829@gmail.com 2 Professor, Department of Civil Engineering, BUET, Dhaka, Bangladesh; ansary@ce.buet.ac.bd

more than half of the population (52%) are poor and about 41% is under the age of 15 (R. M. Islam, 2008).In spite of the poverty and being the most vulnerable to different kinds of coastal disasters, with their traditional indigenous knowledge and perceptions, these coastal people are coping with different kinds of disasters and passing their lives with relentless struggle (Parvin, Takahashi, & Shaw, 2008).Two destructive cyclones occurred in 1970 and 1991, with more than 500000 and 140000 causalities, respectively.However, in the last 20 years, Bangladesh has been able to reduce deaths and injuries from the cyclones.For example, the recent major cyclones like Sidr of 2007 and Aila of 2009 caused 3406 and 190 deaths only, about a 100-fold reduction of lives compared to the deadliest 1970 and 1991 cyclones (A.Ali, 1996;Haider, Rahman, & Huq, 1991; U. Haque et al., 2012;Bimal Kanti Paul, 2009a).
As controlling the occurrence of natural disasters is almost impossible, the loss of life and property from these disasters can be minimized or prevented if proper precautions and preventive initiatives are taken timely.
For saving human lives the fictitious solutions is to build the habitation above the surge level and construct them strong enough to resist high wind speed and the surges (Mallick, Rahaman, & Vogt, 2011).Nevertheless, the majority of coastal residents are living under poverty level in poorly constructed houses that are vulnerable to cyclone (Bimal K. Paul, Rashid, Islam, & Hunt, 2010).Existing literatures shows the importance of constructing raised platforms, reinforcement of homestead structures, overall cyclone proof houses in reducing casualties (Ahmed, Haq, Nasreen, & Hassan, 2015).Besides, focus has been given on infrastructures stability and different infrastructures types to respond to cyclone like temple shaped house and elevated houses (S.K. Paul & Routray, 2011).Locations and pattern of settlements are the most important factor to decrease vulnerability of the coastal people.Scattered settlements in the coastal areas are more susceptible to cyclone (Alam & Collins, 2010).In the recent past, GIS and RS technology have increased the accessibility in the cyclone occurrence stage in order to develop map and assess the vulnerability of lives and infrastructures of cyclone-prone areas (Chisty, Islam, & Samad, n.d.;Rana, Gunasekara, Hazarika, Samarakoon, & Siddiquee, 2010).In developing countries it is badly in need to develop good quality and quantity GIS database for evacuation management.GIS and RS technology can contribute the management process in other hazard prone area like drought, storm surge, flood, land slide, salinity intrusion.So, there is a scope to work with the application of GIS and RS in disaster management especially in the country like Bangladesh.
The next solution is evacuation of the coastal people to the high ground above the surge level (Mallick et al., 2011).Evacuation is considered as the primary preparedness action against any large-scale disasters like the cyclone.An early warning system also plays a valuable role in the evacuation of vulnerable people during the occurrence of the disasters along with minimizing the loss (Dhar & Ansary, 2012).The effective early warning system provided by the government in advance of Cyclone Sidr, Aila, Mohasen, Komen, Roanu, Mora in 2007, 2009, 2013, 2015, 2016and 2017 enabled successful evacuation of coastal communities that resulted in fewer than expected deaths.Besides the early warning systems, different activities like construction of cyclone shelters, embankment, and timely evacuation have contributed to decreasing the death rates in the coastal areas (S. S. Ali, Rahman, & Chowdhury, 2012;U. Haque et al., 2012).In spite of the improvements in early warning systems, pre-cyclone evacuation remains a challenge (Dhar & Ansary, 2012).Some factors like illiteracy, lack of awareness, structural vulnerability and communication problems including poor road networks are the obstacles that hamper the evacuation system.Instead of moving to cyclone shelters, people in coastal areas still believe in a wait-and-see approach (Bern et al., 1993;Chowdhury, Bhuyia, Choudhury, & Sen, 1993;S. K. Paul & Routray, 2013).On the other hand, access to road network, distance of cyclone shelters from the homestead, and access to cyclone shelters are the factors that influence the evacuation decision (S.K. Paul, 2014).Keast, 2014;Bimal Kanti Paul, 2009b).Before the destructive cyclone of 2007, Bangladesh had only 1500 cyclone shelters in the coastal districts as the infrastructural support during cyclone.After Sidr, Bangladesh Government has constructed around 2000 new cyclone shelters for the population of the coastal belt.The World Bank surveyed and further estimated that there is still a need for 5500 new MPCS (Multi-Purpose Cyclone Shelter) facilities for the purpose of a disaster management strategy (A. S. Islam, Bala, Hussain, Hossain, & Rahman, 2010;Nateque Mahmood et al., 2014).GIS-based map has been produced to know about the coverage area of a shelter and need for new shelters (Mallick, 2014).
Despite of completing these initiatives, the evacuation route to cyclone shelters from any community remains a challenge and is not well defined.People living in the coastal areas are not aware of the appropriate evacuation plan and route during cyclones.They can't take the decision during cyclones in which shelter they should go and which shelter is within their reachable distance.It is common that computing shortest paths over a network is an important task for evacuation.With the development of geographic information systems (GIS) technology, network and transportation analyses within a GIS environment have become a common practice in many application areas.Different factors like easy accessibility of the community people to the shelter with maximum walking distance for an individual is restricted to 2.25 kilometers, covering the entire population of the region by producing safe zones and evacuation zones, suitability of the shelters for other community activities are needed to follow (Dalal, Mohapatra, & Mitra, 2007;Gupta, Gupta, Gupta, & Gupta, n.d.).There exists digitized road map including road class, distance and shortest optimum route (Devlin, McDonnell, & Ward, 2008).Many network flow model including logistics coordination model has been built for identifying optimal lane-based evacuation routing plans in a complex road network by using mixed-integer programming solver (Cova & Johnson, 2003;Murray-Tuite & Mahmassani, 2003;Yi & Özdamar, 2007).There exist gaps in the above papers for identifying the shortest distance analysis which is needed during an evacuation process.In the reviewed papers, shortest distance have been analyzed in a map for a vast area without presenting numerical values of the distance from evacuation point to the shelters.Even how many people could be evacuated from each point has not been defined.As Bangladesh is a cyclone prone country, proper evacuation route is a need to reduce the number of affected people.This paper presents an evacuation route planning for Mirjaganj union of Patuakhali, Bangladesh using the shortest distance analysis for each shelter as well as shelter capacity analysis.

Materials and Methods
The research is based on both secondary and primary data collected through household questionnaire survey.The total sample size for the household questionnaire survey has been considered 3220 households.In questionnaire survey different variables have been used to analysis the existing structural condition against cyclone hazard in which they live in.Different variables have been discussed in details in table 2. After collecting the data, ArcGIS10.1 and Microsoft excel have been used to analyze the collected data.ArcGIS10.1 has been used in developing maps, geographic data analysis, editing, data management, and geo-processing activities and also been used in finding the closest shelter facility location, using shortest path analysis method.
A realistic evacuation planning with the shortest route to shelters has been proposed in this study and has been validated using field data.

Results and Discussion
Mirzaganj union is bounded by the Payra River on the south and east, the Srimanta River on the west as can be seen in Figure 1(b).According to the Jatiyo Batayon, and there are 11 villages in this union.Total population is 22100.Among them male is 11073 (50.10%), female is 11027 (49.89%).After surveying 3220 households, it has been found that average 7 people live each in household.The literacy rate of the union is 70% including 13033 voters.Most of the people are farmers and small shopkeepers.

Condition of the Existing Cyclone Shelters and Infrastructures
In Mirzaganj area there are only two usable cyclone shelters (Figure 2).The shelters are designed to withstand a wind speed of 260 km/hour and placed on a higher ground to avoid the effect of the maximum  1).The shelters buildings are primarily used as schools.Shelters are designed in such a way that both human and livestock can take refuge there.

Existing Road Network System
Network Analysis in GIS is based on graph theory and topology.It is often used to find solutions to transportation problems by using either vector or raster models to represent the real world.The vector-based model appears to be more suited to analysis of precisely defined paths such as roads and rivers (Bukholm, Husdal, Nesland, Langerød, & Bukholm, 2003).For a proper evacuation, a good road infrastructure is needed.Road network in coastal villages are still in rudimentary stage and path to the cyclone shelters remain unsafe prior to the landfall of cyclone because of excessive rainfall, high wind velocity, flooded terrain and tidal waves.
Most of the time, roads are damaged by storm surge and make an obstacle for community people from going to safer places.

Evacuation Route Analysis
It has been already mentioned that there is no delineated evacuation plan in this union.While warning system of Bangladesh is improving day by day but evacuation remains as a challenge because of lack of government plan.If there is a shortest route earmarked for every dense area, people will be able to reach to the cyclone shelters properly.Before analyzing the shortest distance we have considered cyclone shelters as inlet where people will take refuge and in every dense area we assumed a point as outlet from where maximum people can gather and evacuate to the shelters.In this study, based on the accessibility to shelters and distance from the main roads, 21 evacuation points (outlets) are marked around the study area.These outlets are situated at or near to the significant roads.
Figure 6(a) shows the outlets in the densely populated areas from where maximum people can be evacuated to the shelters and 6(b) shows the shortest distance from every outlet to the nearest shelter.From every outlet to inlet shortest distance has been analyzed using Network Analysis Tool in ArcGIS 10.As there are two inlets, the total outlets are divided into two areas as can be seen in Figure 6. Green colored routes are presented as the shortest distances to cyclone shelter no 1 and red colored routes are presented as the shortest distances to cyclone shelter no 2.
(a) (b) For shelter no 1, the shortest distance is 351 meters from outlet which is marked as CS1O3 (which means, CS1 is Cyclone shelter 1 and O3 is outlet 3) and for shelters no 2, the shortest distance is 518 meters from outlet which is marked as CS2O3.
In Figure 8 buffer zones have been developed keeping the outlets as the centers with 600 m radius to assume how many people are covered for evacuation.It shows that 16465 people (based on field survey) can be evacuated to shelters, which is70% of total population.Time is also an important factor for real time evacuation.The shortest distance along with the shortest time can be the best solution for assuming how much time it will take to move to the shelters through the shortest distance from each outlet.However, from the field  Figure 9shows the various bent nodes and connectivity nodes of the road network that cause delays.
The placement of bent nodes and connectivity nodes are put in those places from where people would take turn to other roads and where roads have crossed each other.
In Table 3, we have developed a map of bent node network in which we have assumed 2 seconds delay for the bent nodes and in the map of connectivity node network we have assumed 4 seconds delay for the connectivity node for every shortest length to shelter.In Table 3, walking distance has been assumed 3.6 kilometers per hour or 60 meters per minute.(Hasnat, 2016).It has already been mentioned that 1300 people can take shelter in each cyclone shelter.In reality about 4000 people take shelter in each of them and there is always overcrowding as reported by LGED (2014).
Maximum 8000 people can use two shelters.On the basis of these data, two scenarios (with 1300 capacity and with 4000 capacity) have been developed to show how many people can be safe from cyclone by taking asylum in the shelters separately.
Figure 11 shows population departure from different outlets for the shelters (having 1300 capacity).
In this figure, Green color represents the outlets from where people will be able to take refuge to the shelters since the time for evacuation from these outlets are the shortest.On the other hand, Red color represents the outlets from where people will be unable to take refuge to the shelters since the time for evacuation from these outlets are relatively high that means the shelters will be fully occupied before the people from these outlets reach the shelters.Figure 12 shows population departure from different outlets for the shelters (having 4000 capacity).when capacity is 4000people in per shelter.
Figure 13 show that having capacity of 1300 people in each cyclone shelters can cover only 12% of the total population of the union.On the other hand, if the capacity is 4000 persons per shelter, it may cover 36% of the total population union.From Figure 4, it has been seen that about 5% (169 infrastructures) of total infrastructures are of low and very low vulnerability that can also be used as shelter during cyclones.These infrastructures include primary schools, mosque, homesteads (high plinth level with sturdy construction, as can be seen in Figure 3a).It has been assumed that average 50 persons can take shelter in these low vulnerable infrastructures covering 39% of the total population.If the capacity of the cyclone shelter has been considered as 1300 then with low vulnerable infrastructures, the capacity can be increased up to50% of the total population.
And if we consider maximum capacity of the shelters to be 4000, along with low vulnerable infrastructures the capacity can be increased up to 75% the total population.

Figure 1
Figure 1(a) shows the past records of cyclones that hit Bangladesh.Among the Unions of Patuakhali district, Mirzaganj Union is one of the most vulnerable unions to cyclone hazard.Area of Mirzaganj Union is 34 square km, located in between 22°13' and 22°29' north latitudes and in between 90°08' and 90°19' east longitudes.Total amount of road is near about 90 kilometers, among these paved roads are 20 km, herring bond roads are 10 km and earthen roads are 70km.There are 7secondary schools, 23primary schools, and 16madrasa.Main source of income is agriculture with 8527 acres of cultivable land.Due to the geographical location that union has been severely affected by 2007 cyclone Sidr and 2009 Aila.Figure 1(b) shows the location of the

Figure 4 Figure 4 :
Figure 4(a) shows the vulnerability map of the studied union.Figure 4(b) indicates that among 3220 surveyed infrastructures 24 are with very low vulnerability (1%), 145 are with low vulnerability (4%), 1114 are with moderate vulnerability (35%), 1887 are with high vulnerability (59%) and 50 are with very high vulnerability (1%).It also shows that about 5% households are with low to very low vulnerable to cyclone and 95% households are not strong enough to withstand cyclone.

Figure 5 Figure 5 :
Figure5(a) presents the map of road types of Mirzaganj Union.In this map three types of roads (earthen road, herring bond road, and paved road) have been shown.In Figure5, three types of roads and lengths of the particular roads are presented.Among those 72.47 kilometers are earthen road (71%), 24.49 kilometers are paved roads (24%) and 5.4 kilometers are herring bond roads (5%).It can be clearly seen that half of the total road is of earthen type and is the most vulnerable during a cyclone.For this reason during a cyclone earthen road may be a snag for community people for evacuation.Besides width of roads play an important role during evacuation.Wider roads allow more people to evacuate than congested roads.

Figure 6 :
Figure 6: Outlets and Inlets in Mirzganj Union (a), Shortest Distance Analysis of Mirzaganj Union (b) Figure 7 presents the shortest distances from the different outlets to the shelters.The shortest distances for shelter no 1 are colored with green and the shortest distances for shelter 2 are colored with red.
found that during cyclone different types of factors like connectivity nodes and bent nodes may affect the travel time.

Figure 9 :
Figure 9: Connectivity Nodes Map (a) and Bent Nodes Map (b)

Figure 10 Figure10:
Figure10shows time to reach to the shelters from the outlets using the shortest routes considering bents along the path for the two shelters.

Figure 11 Figure11:
Figure 11(a) shows the number of people departing from each outlet that cover the capacity (1300 people) of the cyclone shelter number one.It is observed that 1064 people can depart from outlet CS1O3 and only 236 people can depart from outlet CS1O9 to fully occupy shelter one.Similarly for Figure11(b), it is observed that 665 people can depart from outlet CS2O3 and only 635 people can depart from outlet CS2O6 to fully occupy shelter number two.

Figure 12 (Figure12:
Figure 12(a) shows the number of people departing from each outlet that cover the capacity (4000 people) of the cyclone shelter number one.It is observed that 3500 people can depart from outlet CS1O5, CS1O3, CS1O15, CS1O11 and only 500 people can depart from outlet CS1O9 to fully occupy shelter one.Similarly for Figure 12(b), it is observed that 3451 people can depart from outlet CS2O1,CS2O3, CS2O6, CS2O8, CS2O10 and 549 people from CS2O12 to fully occupy shelter number two.

Figure13:
Figure13: Population demand versus capacity under various scenarios

Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 22 October 2018 doi:10.20944/preprints201810.0344.v2 Cyclone
shelters play an important role during the evacuation.It has been estimated that around 1.5 million people have been evacuated by CPP (Cyclone Preparedness) volunteers to the multi-purpose cyclone shelters (MPCS) when Cyclone Sidr hit the coastal parts of Bangladesh in 2007 (Nateque Mahmood, Prasad Dhakal, &

Table 1 :
Details about Existing Cyclone Shelters

Table 2 :
Infrastructure Vulnerability Scoring Management and Relief (MODMR) and the Bangladesh University of Engineering & Technology (BUET).

Table 3 :
Shortest Time Analysis considering factors