254
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
detection and mapping using satellite imageries included the following steps (Khan
et al. 2011a).
1. Terra MODIS near real-time subsets covering the study region were
retrieved from the National Aeronautics and Space Administration (NASA)
web site http://rapidfire.sci.gsfc.nasa.gov/subsets.
2. Color composite images were downloaded for image processing. The false
composite of MODIS bands 1, 2, and 7 (red, near-infrared, and shortwave
infrared) has a resolution of 250 m. The true color composite of MODIS
bands 1, 3, and 4 was used for visual interpretation.
3. False color composite images were the subset to the region of interest, and
ISODATA classification was performed (20 classes and 3 iterations).
4. All of the water classes were combined into one water class.
5. The raster-type images were exported in a geographical information system
(GIS)- compatible format for further processing.
6. The images obtained in step 5 were overlaid on the true color image to
remove the cloud contamination and shadows that were falsely classified as
water.
7. The final product overlaid in the GIS environment under a reference water
layer (Shuttle Radar Topography Mission [SRTM]-based water bodies) was
used to identify the current flooded areas.
11.2.2 hydRologic Modeling
A distributed hydrologic model (coupled routing and excess storage [CREST]) developed by Wang et al. (2011) was used to generate modeled flood areal extents for
comparison with the satellite-based flood inundation maps. The distributed CREST
hydrologic model is a hybrid modeling strategy that has recently been developed by
the University of Oklahoma (hydro.ou.edu) and the SERVIR Project Team in NASA
(www.servir.net). CREST simulates the spatiotemporal variation of water fluxes and
storages on a regular grid, with the grid cell resolution being user defined. The scalability of model simulations is accomplished through subgrid-scale representation
of soil moisture variability (through spatial probability distributions) and physical
process representation. CREST can also simulate inundation extent in an effort to
obtain spatial and temporal variation of floodwater within a grid-based domain. For
more information regarding the CREST model, please refer to the work of Wang
et al. (2011).
To apply CREST over the study area at a 1-km spatial resolution, local drainage direction and accumulation were established using a 30-arc-second-resolution
SRTM digital elevation model from HydroSHEDS data. The precipitation forcing
data are the Tropical Rainfall Measuring Mission (TRMM)-based multisatellite
precipitation analysis 3B42 real-time (TMPA 3B42RT) products (Huffman et al.
2007). The subscript ‘RT’ refers to real time, which in reality refers to pseudo real
time where data are available via the Internet with an 8–16 h latency for the end
user.
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
detection and mapping using satellite imageries included the following steps (Khan
et al. 2011a).
1. Terra MODIS near real-time subsets covering the study region were
retrieved from the National Aeronautics and Space Administration (NASA)
web site http://rapidfire.sci.gsfc.nasa.gov/subsets.
2. Color composite images were downloaded for image processing. The false
composite of MODIS bands 1, 2, and 7 (red, near-infrared, and shortwave
infrared) has a resolution of 250 m. The true color composite of MODIS
bands 1, 3, and 4 was used for visual interpretation.
3. False color composite images were the subset to the region of interest, and
ISODATA classification was performed (20 classes and 3 iterations).
4. All of the water classes were combined into one water class.
5. The raster-type images were exported in a geographical information system
(GIS)- compatible format for further processing.
6. The images obtained in step 5 were overlaid on the true color image to
remove the cloud contamination and shadows that were falsely classified as
water.
7. The final product overlaid in the GIS environment under a reference water
layer (Shuttle Radar Topography Mission [SRTM]-based water bodies) was
used to identify the current flooded areas.
11.2.2 hydRologic Modeling
A distributed hydrologic model (coupled routing and excess storage [CREST]) developed by Wang et al. (2011) was used to generate modeled flood areal extents for
comparison with the satellite-based flood inundation maps. The distributed CREST
hydrologic model is a hybrid modeling strategy that has recently been developed by
the University of Oklahoma (hydro.ou.edu) and the SERVIR Project Team in NASA
(www.servir.net). CREST simulates the spatiotemporal variation of water fluxes and
storages on a regular grid, with the grid cell resolution being user defined. The scalability of model simulations is accomplished through subgrid-scale representation
of soil moisture variability (through spatial probability distributions) and physical
process representation. CREST can also simulate inundation extent in an effort to
obtain spatial and temporal variation of floodwater within a grid-based domain. For
more information regarding the CREST model, please refer to the work of Wang
et al. (2011).
To apply CREST over the study area at a 1-km spatial resolution, local drainage direction and accumulation were established using a 30-arc-second-resolution
SRTM digital elevation model from HydroSHEDS data. The precipitation forcing
data are the Tropical Rainfall Measuring Mission (TRMM)-based multisatellite
precipitation analysis 3B42 real-time (TMPA 3B42RT) products (Huffman et al.
2007). The subscript ‘RT’ refers to real time, which in reality refers to pseudo real
time where data are available via the Internet with an 8–16 h latency for the end
user.