Hurricane Felix Response Products
Hurricane Felix made landfall with Nicaragua on September 4, 2007 as a Category 5 storm. The following products are a result of the U.S. Geological Survey's previous work in Central America in response to Hurricane Mitch in 1998. These products provide scientific information that can focus well-informed decisions for relief and recovery efforts following natural disasters and other catastropic events.
ASTER Digital Elevation Model Production in Afghanistan
This project involves the creation of 30 meter resolution digital elevation models (DEM) in Afghanistan primarily for natural resource assessments. In this study, absolute DEMs were created and extracted using PCI Geomatica's Orthoengine software. Preliminary results and analysis exhibit common errors that occurred in image areas corresponding to cloudy and snow covered areas, lakes, steep slope areas, and southeastern facing slopes. As a result of these features, poorly correlated elevation values produced erroneous holes, large pits and spikes in the initial elevation output. Ongoing research indicates that erroneous elevation values corresponded with steep slopes and scenes collected with high, off-nadir pointing angles. To address these errors, multiple scenes were acquired with low off-nadir pointing angles and overlapping DEMs were produced and mosaicked to fill void areas. In addition, a progressive morphologic filter was applied as a post processing step to remove pits and spikes. These post-processed and mosaicked DEMs produce more accurate and visually appealing elevation models for landform classification, geologic structure analysis, and natural resource assessment applications.
Also, research is being conducted on the accuracy of ASTER generated DEMs by comparing the output elevation values to spot elevations from topographic maps, SRTM elevations and GPS points.
Development of a progressive morphologic filter to remove erroneous values from DEMs
This project focuses on improving the accuracy and quality of photogrammetrically derived high-resolution digital elevation models through the development of a raster-based progressive morphological filter.
To improve the quality of output DEMs, most software routines employ a low pass filtering technique to smooth elevation values. This technique reassigns a mean elevation value for a 3x3, 5x5, or 7x7 pixel window around all cell values. Calculating a mean elevation in a window containing a large pit or spike biases the values of all cells within that window and reduces, but does not eliminate the erroneous value. A progressive morphological filter was developed to target and filter only erroneous pit and spike data values in raw DEM data produced from a stereo auto-correlation process.
Our progressive morphological filter iteratively compares individual raw elevation values to a set of focal neighborhood statistics and a user defined threshold value.
The result is that only elevation values that exceed the defined parameters are replaced; all other values remain unchanged and the overall output quality is improved without degrading the high resolution fidelity of the DEM.