Quantifying Tsunami and Earthquake Hazards using Remote Sensing and GIS

On Septem 28, 2018 a tsunami struck the Palu, city on the Indonesian island of Sulawesi, after the earthqauke. The combination of the earhquake and the tsunami led a death toll of more than 2 thousand people. Not only Indonesia but also countries like Japan, Thailand, Chile and many more are vulnurable to tsunami. Although prevention of water flood should be the main concern, we also need to find effective ways to quantify the tsunami (and earthquake) hazards to better prepare oursleves for the future.

Comparison methods

Our aim was to highlighting and quantifying the damaged area. We can divide change detection in remote sensing into five groups (Mouat, Mahin, & Lancaster, 1993).

1. Visual interpretation

● True color comparison

2. Image Algebra (difference images and ratioed images)

● Single Band


● Euclidean Distance

3. Transformation and data reduction


4. Classification

● Supervised Classification

5. Statistical (This method was kept outside the study scope due to time limit)

As the essence of change detection is to compare pixels from before and after images the

following preprocessing steps are very important (Campbell & Wynne, 2011). The images

should be;

● Acquired by the same sensor

● Acquired during the same season

● Well co registered

● Free of clouds

● Atmospherically corrected

Preparing the Satellite Data

September (left) and October (right) satellite imageries.

September (left) and October (right) satellite imageries.

For this project I used radiometrically and geometrically corrected Sentinel Level 1C data to analyse before and after images from Palu. As the Sentinel images are provided in top-of-atmosphere reflectance I converted them to surface reflectance using ENVI`s QUACK tool. The before and afer (September and October) images were also free of cloud which made the process much easier.

Workflow to Calculate the Hazard Area

  1. Input required atmospherically corrected bands (i.e NIR and Red) for both September

    (before) & October (after)

  2. Use a formula (i.e NDVI or Euclidean Distance) or use a tool (like PCA) for both

    September & October

  3. Subtract the resulting two raster from each other to see the how a given method

    reveals different parts of the imagery in before and after images. The subtraction

    most likely will indicate the location of substantial changes and the hazard.

  4. Convert the resulting raster that has float values to integer

  5. Extract values that are important (for NDVI > 0)

  6. Convert raster to polygon

  7. Import shapefiles for roads and buildings

  8. Intersect the shapefiles with the polygon file that shows us the possible affected


  9. Calculate the area of predicted damaged zones, building and roads for each method



  1. Mouat, D. A., Mahin, G. G., & Lancaster, J. (1993). Remote sensing techniques in the

    analysis of change detection. Geocarto†International†, 8†(2), 39–50.


  2. Campbell, J. B., & Wynne, R. H. (2011). Introduction†to†Remote†Sensing¨†Fifth

    Edition†. Guilford Press.