Spatiotemporal Mapping of Geo-hazards and Flood Inundation Modelling in the Himalaya Region of Bhutan, India, and Nepal

Investigator: 
Sushant Banjara
Advisor: 
Karen Seto
Start Date: 
May, 2018
Description: 

I intend to create a hazard database of the Himalaya region for last two decades. The hazards I will work with are fire, flood, landslide, and earthquake. Of the four hazards the one requires most amount of remote sensing is Flood, the methods for which is described below:

The bi-daily MODIS image provides high probability of acquiring cloud free images without losing much temporal information, through ‘best-pixel’ composite (MOD09A1). However, for a coarse resolution image of 500m, Open Water Likelihood (OWL) (Guerschman et al., 2011) as shown in the following equation has been found to perform better than popular indices such as mNDWI in delineating water from non-water(Chen et al., 2011; Overton et al., 2010).
OWL = 1/ (1 + exp(f))
f = a0 +∑aixi  (i=1 to 5)
where x1= SWIR Band 6 reflectance; x2= SWIR Band 7 reflectance; x3= NDVI;  x4=NDWI;
and x5= MrVBF  (an index representing degrees of valley bottom flatness (Gallant & Dowling, 2003)).
  While OWL gives the fraction of water within a pixel, there isn’t a set threshold to discriminate between flooded and unflooded pixels(Ticehurst et al., 2015); much so for populated and rugged topography of Nepal. Therefore, inundation maps of 2017 Nepal flood events will be created for varying thresholds (5% to 20%) and will be compared against other high resolution satellite image (for example, Landsat).