Translator Disclaimer
29 October 2007 Comparison of multi-resolution SRTM data for morphometric features identification using neural network self-organizing map (case study: Eastern Carpathians)
Author Affiliations +
The Shuttle Radar Topography Mission (SRTM) was launched on 11 February 2000 and 3 arc second data were publicly released in July 2004. Easy availability of SRTM 3 arc second data, covering almost 80% of the land surface on earth, has resulted in great advances in morphometric studies and numerical description of landscape features. In this study we introduce a new procedure using Neural Network - Self Organizing Map - to characterize morphometric features of landscapes.. We also investigate the effect of two resolutions for morphometric feature identification. Specifically we investigate how the SRTM 3arc second latitude / longitude data projected to UTM coordinates with 90 meter respectively 28.5 m grid, corresponding to Landsat TM data resolution, affect the morphometric characterization. Morphometric parameters such as slope, maximum curvature, minimum curvature and cross-sectional curvature are derived by fitting a bivariate quadratic surface with a window size of 5×5 for the 90 m data (450 m on the ground) and 9×9 for the 28.5 m data (about 250 m) . Kohonen Self Organizing Map as an unsupervised neural network algorithm is employed for the classification of these morphometric parameters into 10 exclusive and exhaustive classes. These classes were analyzed and interpreted as morphometric features such as ridge, channel, crest line, planar and valley bottom for both data sets based on morphometric signatures, feature space and 3D inspection of the area. The difference change detection technique was used between two DEMs (DEM-90 and DEM-28.5 m) to analyze differences in morphometric features identification. The results showed that the introduced method is very useful for identification of morphometric features. Increasing spatial resolution from 90 meter to 28.5 meter, can produce digital elevation models (DEMs) allowing more precise identification of morphometric features and landforms. Increasing spatial resolution overcomes the main constrains for morphometric analysis with SRTM 90 m data, such as artifacts, unrealistic feature presentations and isolated single elements in the output map. Increased spatial resolution together with the smaller window size emphasized local conditions but main morphometric features were preserved. An overall change of 66.36 % is observed for morphometric features in the 28.5 meter DEM. The most and least frequent changes occurred for class no.6 (moderate slopes, channel) with 82.74% and class no.7 (Gentle slope to flat, valley bottom, planar) with 43.31% respectively. Increasing spatial resolution can be applied also to watersheds studies like drainage modeling.
© (2007) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Amirhoushang Ehsani and Friedrich Quiel "Comparison of multi-resolution SRTM data for morphometric features identification using neural network self-organizing map (case study: Eastern Carpathians)", Proc. SPIE 6749, Remote Sensing for Environmental Monitoring, GIS Applications, and Geology VII, 67491J (29 October 2007);

Back to Top