Understanding Spatial Variations in Earthquake Vulnerabilities of 1 Residential Neighborhoods of Mymensingh City, Bangladesh: An AHP-GIS 2 Integrated Index-based Approach

: Mymensingh city is highly earthquake vulnerable due to its geological setting, existence of three 8 faults, viz., Dauki Fault, Madhupur Blind Fault and Sylhet-Assam Fault in its close vicinity, and liquefaction 9 susceptible soil type. Recently an attempt has been made to assess earthquake risk of the city by Comprehensive 10 Disaster Management Programme II, of Government of Bangladesh using FEMA developed HAZUS tool which 11 requires usage of enormous resources and expertise. Poorly resourced city planning authorities of developing 12 countries are seldom equipped with such financial and human resources, and as a result, the inclusion of 13 earthquake risk analysis, more specifically, information regarding spatial variations of earthquake risk is very 14 often found missing in their physical planning exercises. This paper aims to assess the spatial variation of 15 earthquake vulnerability of residential neighbourhoods of Mymensingh city, employing an index-based low cost 16 approach which could provide a reasonably accurate result with minimum resource and expertise requirements. 17 Analytical Hierarchy Process and Weighted Linear Combination are combined with a Geographical Information 18 System to prepare a composite index considering 23 different parameters, stemming from geological, structural, 19 socio-economic and systematic dimensions of earthquake vulnerability. The findings of the reseach show that out 20 of 241 residential neighbourhoods of Mymensingh city, 51 are observed to be highly vulnerable, while, 123 and 21 67 are medium and low vulnerable respectively. Besides, the spatial distribution of earthquake vulnerable 22 neighbouhoods in Mymensingh City, observed in the current study has also been compared with spatial 23 distributions observed in two similar previous studies and observed found to be reasonably close. This justifies 24 the validity of the current low cost approach for wider application in cities of resource starved developing 25 countries.


Background 30
Bangladesh, the largest delta of the world, is prone to numerous natural catastrophes due to its geographical are making Mymensingh City vulnerable to earthquake and puts lives and assets of its citizen at risk. evaluated the seismic risk of Barcelona using the vulnerability index method and capacity spectrum-based method previous earthquake but requires every detail of building stock which is very time and resource consuming. There

Selection of Parameters of Earthquake Vulnerability Assessment 168
In this study, 23 influential earthquake vulnerability parameters have been selected based on diligent literature 169 review, expert opinion and by analysing available data, under four vulnerability dimensions, viz., geological, 170 structural, socio-economic and systematic vulnerability.

Geological earthquake vulnerability parameters 172
Geological parameter refers to the factors related to the earth that affects the earthquake vulnerability of an area.

Systematic Earthquake Vulnerability Parameters 179
One of the influential earthquake response issues in cities is the accessibility of residential neighbourhoods to

Structural Earthquake Vulnerability Parameters 188
Structural earthquake vulnerability parameter refers to the factors that relate to the built up environment such as 189 buildings, bridge, road, etc. Structural parameters have a great influence on earthquake vulnerability and damage 190 potential of a neighbourhood. In this study, eight most influential structural parameters are considered to assess 191 the earthquake vulnerability of Mymensingh city which is shown in Table 3.    Table 4.

231
L represents the Eigenvalue of the pairwise comparison matrix, and RI is the random inconsistency index, which 232 has some developed value and depends on the number of vulnerability assessment parameters (N). The variations 233 of RI value for different parametersare shown in Table 6.

236
WLC technique is an additive weighting method in which a weight is assigned to each factor at the initial stage.

237
The weight of vulnerability parameters determined by using AHP method based on expert opinions is used with   Table  7, Table 8,Table 9 and Table 10). The aggregated comparison matrix of earthquake vulnerability assessment 249 used in this study is shown in Table 11.    In this study 24 vulnerability parameters are weighted on a scale of 0 to 1. It is essential to assign a weight to 261 every sub-category of the abovementioned 24 parameters. Providing different weight to every sub-factor is a

Data Source 281
In

Data Analysis and Vulnerability Maps Preparation 290
In this study, the Analytical Hierarchical Process has determined weights of different factors and sub-factors of 291 seismic vulnerability. All gathered data has been processed in the following sequential order: Firstly, the socio-292 economic data and vulnerability scores of earthquake vulnerability of Mymensingh city has been stored in SPSS 293 environment and converted into Microsoft Access database to make them usable for analysis in GIS software 294 (ESRI product ArcGIS has been used). Secondly, neighbourhood wise data of structural and geological earthquake 295 vulnerability of Mymensingh city have been extracted using geo-processing in the ArcGIS environment. Then, 296 the databases are joined with the residential neighbourhood map of Mymensingh city map in vector-based GIS.

297
The centre points of each residential neighbourhoods are delineated using the conversion tool in ArcGIS. In the 298 next step, the maps have been reproduced for determining systematic vulnerability parameters using closest

Validation Methods Adopted 307
used to assess the performance of the classifier, is a metric that compares an Observed Accuracy with an Expected

Result and Discussion 332
The spatial variations of vulnerabilities are analyzed and shown in maps in 3 vulnerability zones, viz., high, 333 medium and low. From the city planning context for better understanding of the priorities of risk mitigation 334 activities, it is also essential to identify the relative importance of vulnerability parameters influencing earthquake vulnerability of the neighborhoods and therefore, have also been discussed in the following section as well.

342
is observed that Soil type has the highest (0.5) influence among the parameters followed by PGA (0.32). Shear

343
Wave Velocity (0.18) has the least influence among the three parameters used in this analysis.

Systematic Vulnerability 345
The distances of the hospital, fire station, emergency shelter and emergency evacuation route from the geometric 346 centre of each neighbourhood are considered and analysed in ArcGIS environment to assess the spatial variation The parameter wise assessment of systematic earthquake vulnerability of Mymensingh City on a scale of 0-1 is 368 shown in Fig.8. According to Fig.8, most of the residential neighbourhoods in Mymensingh City are highly

Structural Vulnerability 372
From the analysis, it is found that eight residential neighbourhoods of Mymensingh city are highly structural

381
It is critical to know which parameter has the most influence onthe structural vulnerability to prioritise city 382 planning implications.  Mymensingh City shows that mainly the city is socio-economically earthquake-vulnerable due to the high    Fig.15 in which the earthquake sensitivity of Mymensingh city is classified 431 into two categories viz. 1 st degree and 2 nd -degree earthquake sensitivity. According to CDMP-II, 1 st -degree earthquake sensitivity explicates the areas with high earthquake hazard risk, and 2 nd -degree earthquake sensitivity 433 indicates the areas with low earthquake hazard risk.

449
The earthquake sensitivity map developed by CDMP-II mainly considered geology and infrastructure related

467
The comparison of these two results is done only for residential cells. The confusion matrix score shows that there 468 exist 71% agreement in defining the highly vulnerable zones and 90% agreement in determining low vulnerable 469 zones (Fig. 18). The normalised confusion matrix shows that there exists 57% disagreement in defining a medium 470 vulnerable area which slightly misclassified as low vulnerable in the result.

Conclusion 472
Understanding spatial variability of earthquake vulnerability of a city in the earthquake susceptible zone is of 473 paramount importance for deciding on appropriate planning and development control interventions. Incorporating 474 earthquake risk in the city planning process for developing countries like Bangladesh is even more challenging (CDMP) for data support for this research work. Authors also would like to thank Dr. Ishrat Islam, Professor,

507
Department of Urban and Regional Planning, Bangladesh University of Engineering and Technology, Dhaka,