Home Articles Earthquake loss estimation using high resolution satellite imagery

Earthquake loss estimation using high resolution satellite imagery

L. Chiroiu1,2, G. André1,2, F. Bahoken2
1PhD Candidate, Université Paris-7 Denis Diderot
GHSS, CNRS UMR 8586 PRODIG, Case 7001, 2 place
Jussieu, 752251 Paris Cedex 05; France
[email protected] , [email protected]

2Géosciences Consultants, 157 rue des Blains, 92220 Bagneux; France
geosciences.consu[email protected]; [email protected]

Abstract
The recent progresses of remote sensing in terms of spatial resolution and data processing open new possibilities concerning the natural hazard assessment. Using a high resolution optical imagery available today, a damage detection could be performed inclusively in urban areas. A multidisciplinary approach based on high resolution satellite data and earthquake engineering was applied in order to estimate the damage after the Bhuj, (India), Earthquake of January 26 th , 2001. The study provide a fast loss estimation, in terms of physical damage and human casualties. A GIS has been used in order to display the spatial distribution of damages. The results could be very useful for the rescue teams deployed immediately after the catastrophe.

Introduction
A new generation of high resolution optical imagery is provided today by commercial satellites such IKONOS, launched in 1999, with 1 meter resolution in panchromatic mode and 4 meters in multispectral, or EROS A1, launched in 2000, with 1.8 meters resolution in panchromatic mode. Features like buildings, streets or cars become visible with a high ground resolution. In the recent future, an incredible under 1m resolution of Quickbird * satellite will be available for the civil applications (0.61 meters in panchromatic mode, 2.8 meters in multispectral). The high level of details makes possible a reliable damage detection to the buildings or to other structures.

This study is proposing to asses damages in the urban area of the city of Bhuj after the January 26 th earthquake by high resolution optical imagery. Using a 1 meter image taken after the event and a 2 meters image acquired before the earthquake, losses were recognized by mono and multi temporal approaches. The damaged area was analyzed in terms of surface, and the results were integrated into a GIS database.

Bhuj Earthquake
On January 26 th , 2001, at approximately 8:46 a.m. local time, a Mw 7.7 earthquake occurred in western India, where around 20 million people live and work. While the earthquake was felt as far as Nepal and in Pakistan, its most heavy destruction was in the state of Gujarat. The death toll stands at over 20,000 and about 167,000 people have been injured. It is estimated that nearly one million homes were damaged or destroyed. Some cities were completely destroyed, like Anjar or Bachau. The city of Bhuj, located at around 20 km from the epicenter, suffered important losses. A maximum intensity of X (MSK) was assigned by the local authorities.

Damage detection
There are two possibilities to detect damages using photo interpretation analysis: a mono temporal technique based on a post event image, and a multi temporal approach, where a before event scene is compared with an after event scene. The mono temporal procedure consists in the visual recognition of the damaged elements, and it is directly related with the image resolution. With a medium resolution (around 10 meters) only larges zones completely destroyed can be observed. The 1 meter resolution allows the detection of damaged buildings one by one, the building size being considerably greater than the pixel size. In this study it was applied a classical mono temporal photo interpretation method, the damage being detectable by a visual analysis. The results were verified by a multi temporal change detection approach.

In the south part of the town, the recognition of destruction was facilitated by the regular distribution of buildings (Figs 1 and 2).

    

Damage states are generally not recognizable by remote sensing, only buildings completely destroyed being detectable. However, a distinction was made between two levels of destructions: extensive damage and complete damage (Fig. 3 a, b and c). The extensive damage is considered when a building or an entire zone is damaged, but the building(s) is not totally collapsed (partial failure of the structure). The complete damage corresponds to a totally collapsed building (total failure of the structure).

    

FIGURE 3. a) A collapsed building 
classified completely damaged; 
b) An example of complete damage 
(in red) vs. extensive damage (in yellow); 
c) An entire zone considered with severe 
damage.

Loss Estimation
A GIS was used in order to display the spatial distribution of damages (Fig. 4). An important concentration of destructions can be observed in the north side of the city, where the high vulnerability of the old town due to the traditional architectural style had amplified the losses .


FIGURE 4. The spatial distribution of losses for the urban area of Bhuj city.

A statistical approach based on engineering judgments were applied in order to obtain a fast estimation of losses. First, the surface associated to each polygon corresponding to a damaged zone was computed using a GIS tool. Over the entire urban zone covered by the image it was founded an area of 0.43 km 2 considered as damaged, with around 0.25 km 2 of extensive damage (ED) and around 0.18 km 2 of complete damage (CD). Comparing with the total area of Bhuj, around 4.3 km 2 , 10% of the city is considered with severe damages.

The population of Bhuj, estimated at 121.000 people, was divided by the total urban area of the city, in order to obtain a local density of population.

TABLE 1. Computation of population density

Total Population 121000
Surface (km2) 4.3
Density (people/km2) 28140

Simplified hypotheses were applied in order to estimate losses. There are hypotheses based on various statistics and engineering judgments, this theme being the subject of a work in progress. The following suppositions are proposed:

  1. regarding the complete damage, given that the structure is completely destroyed, it was assumed that 80% of the occupants are dead, and 20% of the occupants are injured.
  2. regarding the extensive damage, it was assumed that 5% of the occupants are dead, and 60% are injured.

These ratios were applied to the persons affected by a damage level. A rapid computation of losses was performed, detailed in the following table:

A simplified statistical approach allowed a fast evaluation of losses. For the urban area of Bhuj city, the results are, in terms of human casualties, a number of 4404 dead persons and 5234 injured persons.

Validation
Most of the field reports after the Gujarat earthquake present a disastrous situation of the city: “The old town of Bhuj, completely destroyed…” (AFPS 2001), or “several towns, like Bhuj (…) sustained widespread destructions” (EERI 2001). The spatial distribution of damages estimated by our study shows a high level of losses in the north side of the city, in the “old town”. The dimensions of the damages in this part of the city are expressed clearly by some pictures taken after the earthquake (fig. 5, a and b).

One of the last official statistics of losses for the city of Bhuj are given by a UNDMT report of January 29 th , where the number of dead persons is 5,065 and of the injured persons is 10,925. After this date, the losses are quantified for the entire Katchch region, and there are no other statistics of losses only for the city of Bhuj. One of the last reports of losses gives a number of around 20,000dead persons and around 167,000 injured persons for the Kachch region, where cities like Anjar and Bachau (both with around 51,000 people) have suffered serious destructions.

The loss estimation performed in this work, in terms of human casualties, is not far from the last official statistics concerning the number of dead persons (4404 dead persons in this study vs. 5065 of the official statistics). A slight underestimation (lesser than 10%) may be due to the impossibility to recognize moderate and slight damage. The same constraint could explain the difference between the number of injured persons, which could be in reality between 5 and 10 times higher.

Conclusions
The high resolution satellite imagery offers new possibilities for the earthquake damage
assessment. A multidisciplinary approach combining remote sensing techniques, spatial analysis and earthquake engineering can provide a fast loss estimation. Photo interpretation techniques applied to 1 meter resolution images, in mono or multi temporal analysis, enable, for the case of Bhuj, a damage assessment with a high degree of accuracy. Losses can be quantified using a GIS, in terms of damaged area, and human casualties can be estimated applying computations based on simplified approaches.

Depending on the surface to be treated and the number of persons involved in the study, a near real time damage assessment could be possible. The feasibility of the application is conditioned by a fast imagery purchasing, which must be done immediately after the catastrophe and focused on the major urban zones; in the same time, the cloud coverage represent obviously an important factor constraining the acquisition. The information can be integrated into a GIS database and transferred via satellite networks or internet to the rescue teams deployed on the affected zone. The results of a fast damage assessment received by the field operators could help the civil protection in order to better coordinate the emergency operations.

References

  • Association Française de Génie Parasismique AFPS (2001). Le séisme de Bhuj (Gujarat, Inde) du 26 janvier 2001. Rapport de Mission, Paris, France.
  • EDM (2000, b). The Bhuj Earthquake of January 26, 2001. Consequences and Futurs Challenges. Earthquake Disaster Mitigation Center; The Institute of Physical and Chimical Research (RIKEN). Miki, Hyako Prefecture, Japan; www.edm.bosai.go.jp
  • EERI (2001). Preliminary Observations on the Origin and Effects of the January 26, 2001 Bhuj (Gujarat, India) Earthquake. EERI Newslettres, Vol. 35, N°. 4; Oakland
  • R. Eguchi et al (2000). Using Advanced Technologies to Conduct Earthquake Reconnaissance after the 1999 Marmara Earthquake; 2 nd Workshop on Advanced Technologies in Urban Earthquake Disaster Mitigation, DPRI, Kyoto University, Uji, Kyoto, Japan.
  • NIBS (1999). HAZUS – Earthquake Loss Estimation Methodology. Technical Manual, Vol. III, Washington, D.C.
  • M. Tralli (2001). Assessment of Advanced Technologies for Loss Estimation. Multidisciplinary Center for Earthquake Engineering Research; State University of New York at Buffalo Research Foundation, NY.
  • United Nations Disaster Management Team (2001). India Earthquake Reports N° 1 – 21. www.reliefweb.org World Bank and Asian Development Bank (2001). Gujarat Earthquake Recovery Program; Assessment Report; www.worldbank.org

TABLE 2. Loss estimation computation.

Damage Level Density (people/km2) Area (km2) Affected People Death Rate Injury Rate Injured Persons Dead Persons
ED 28140 0.25 7 035 5% 60% 4221 352
CD 28140 0.18 5 065 80% 20% 1013 4052
          Total 5234 4404