Author(s): Dadiyorto Wendi; Shie-Yui Liong; Yabin Sun; Chi Dung Doan
Linked Author(s):
Keywords: SRTM; Hydrology; Landsat 8 and Artificial Neural Network
Abstract: Digital elevation model (DEM) plays an absolutely essential role in hydrological study, from understanding the hydrological characteristics of a watershed to setting up a hydrological model, which includes rainfall-runoff or hydrodynamics simulations. However, access to high resolution DEM is very costly. Although Shuttle Radar Topography Mission (SRTM) is a publicly available DEM, its resolution is known to be rather coarse (92m outside US). SRTM, obtained through radar based remote sensing, and suffers from inaccuracy especially on area covered by canopy as the 5.6 cm wavelength used does not penetrate vegetation well. This presents challenges to hydrologists in modeling catchments that are subject to canopy coverage. This paper considers a forested area in Singapore and presents a proof of concept of an approach to improve the SRTM dataset. The approach makes full use of (1) the introduction of Landsat 8 data, of a resolution of 30m, into SRTM data; and (2) the Artificial Neural Networks to flex its known strengths in pattern recognition. The study shows a series of significant improvements of the DEM for the forested area, for examples: (1) a reduction of about 68%in RMSE in elevation; (2) their spatial 10m contour map; (3) a large reduction of about 90%and 75%in DEM’s mean error and RMSE respectively, in the longitudinal profile; (4) a much clearer delineation of drainage network. The proposed approach has been demonstrated to be very promising to improve the low resolution DEM, directly obtained from publicly accessible SRTM data, by cleverly amending SRTM data with Landsat 8 data and then use the strengths of pattern recognition power of ANN.
Year: 2014