Bridging the Past with Today’s Microwave Remote Sensing: A Case Study of Long Term Inundation Patterns in the Mekong River Delta

Abstract:

Surface inundation extent and its predictability vary tremendously across the globe. This dynamic is being and has been captured by three general categories of satellite imagery: 1) low-spatial-resolution microwave sensors with global coverage and a long record of observations (e.g., SSM/I), 2) optical sensors with high spatial and temporal resolution and global coverage, but with cloud contamination (e.g. MODIS), and 3) in more ‘‘snapshot’’ form by high-resolution synthetic aperture radar (SAR) sensors. We explore the ability to bridge techniques that can exploit the higher spatial resolution of more recent data products back in time with the help of the temporal evolution of lower resolution products.

We present a study of long term (20+ year) inundation patterns in the Mekong River Delta using baseline observations from the Surface Water Microwave Product Series (SWAMPS), an inundation area fraction product derived at 25km scale from active and passive microwave instruments (ERS, QuikSCAT, ASCAT, and SSM/I) that spans from Jan. 1992 to Jun. 2015. Every hydrological basin has unique characteristics – such as its topography, land cover / land use, and space-time variability – thus, a downscaling algorithm needs to take into account these idiosyncrasies. We merge SWAMPS with topographical information derived from 30m SRTM DEM, river networks from USGS HydroSHEDS, and assess the best statistical procedure to “learn” from two sets of classified SAR data: (1) L-band imaging radar from ALOS PALSAR, 2007-2010, and (2) C-band imagery from the Sentinel-1 mission (2014 to present). We present a comparison of retrospective downscaled flood extent with Landsat imagery and recent observations from SMAP.

With a higher spatial resolution of past flooding extent, we can improve our understanding of how delta surface hydrology responds to local and regional events. This is important both in the short-term for accurate flood prediction, as well as on longer-term planning horizons.