Night Time Damage Estimation

Night Time Damage Estimation

SHARE

Masayuki Kohiyama, Haruo Hayashi, Norio Maki, Shin Hashitera

Reduction in night time light intensity in the images can be used as an effective tool for estimating earthquake damaged areas

In a large earthquake disaster, rapid detection of spatial distribution of damaged areas is indispensable information for effective emergency response, relief and recovery activity. The authors are developing the Early Damaged Area Estimation System (EDES), which provides the information as to the estimated impact areas within the first 24 hours after any significant earthquake based on night time lights observed by the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS). The estimation method is based on the detection of significant reductions or loss of lights in nighttime images following the event, for it can be expected that city lights will observably decrease after a large earthquake due to various reasons such as electricity failure, building collapses, evacuation to shelters and the suspension of commercial activities. The research purpose is to provide the geographic information of damaged areas, which will be useful for disaster managers and volunteers of NGOs and NPOs.


Fig. 1a: Pre-earthquake image


Fig. 1b: Post-earthquake image

Damage estimation using DMSP/OLS
The DMSP satellite images are suitable for the early identification of the damaged areas for following reasons:

  • Low light imaging capabilities permit the detection of faint light sources present at the earth’s surface.
  • The nighttime images may be observed twice or more times in a single night due to orbital overlaps and multiple day-night DMSP satellites.

These mean that it is generally possible to analyse nighttime OLS data for significant reductions in nighttime lights immediately following earthquake events worldwide. The DMSP/OLS imagery has spatial resolution of 2.7 km, lower than that of the Landsat/TM or the SPOT/HRV.


Fig. 2: Map of preliminary estimated result

But the repeat periods of these satellites with high-resolution sensors may be more than two weeks making the probability to acquire imagery of the impacted area immediately after a disaster event quite low.

Damaged area estimation method
EDES uses the following processing steps as shown in Figure 3a:

  • Acquiring the earthquake hypocenter data through the Internet, which is for example announced by the USGS.
  • Determining the sampled area of images by the attenuation formula of earthquake ground motion.
  • Selection and download of the corresponding nighttime DMSP/OLS data from National Oceanic and Atmospheric Administration, National Geophysical Data Center (NOAA/NGDC).
  • Estimating the damaged areas by analysing the statistically significant decrease in light intensity of the images.
  • Creating the thematic map to present the estimation results as geographic information.
  • Disseminating the result map widely and quickly through the WWW.

The estimation method is basically based on significant test of the reduction of lights more than normal fluctuation. Figure 3b shows conceptual estimation criteria using histogram of differences in Visible-near infrared (VNIR) band digital number between pre-event and post-event images.


Fig. 3a: Flow chart of estimation

Application to Gujarat earthquake of January 26, 2001

We applied the estimation method to the Gujarat, India Earthquake of January 26, 2001. Pre-event and post-event nighttime images are shown in figures 1a and 1b, respectively. The red line in the figure 1b is of noise. We estimated damaged areas using these two satellite images. Figure 2 is a preliminary estimation result, which contains some extent of errors due to clouds, observation angle, and water reflection influence. The areas where the nighttime-light intensity became significantly weaker after the earthquake are colored in red (confidence coefficient > 99%) and yellow (confidence coefficient > 95%) in the map. The gray spots in the map show the areas where the brightness after the earthquake exceeds the measure range of the OLS sensor, and the light intensity difference cannot be figured out in the areas. The pale purple areas in the map show cloud existence where the city lights were influenced and the light intensity difference cannot be figured out in the areas, either.


Fig. 3b: Estimation criteria

As far as the estimation result indicates, damaged areas distribute even 300 km far from the epicenter. Most of news reports have been focusing on heavily afflicted areas near the epicenter. But we think much of equity of disaster relief that all the disaster victims should be supported, who might live in forgotten, non-media reported towns. We hope relief communities make use of this spatial information for their activities, and wish this information assist recovery of all the suffering people.