Theodore C. Smith
American River College, Geography 350: Data Acquisition in GIS; Winter 2012
7053 Enright Drive, Citrus Heights, CA 95621
Strom-triggered debris flows are a significant hazard in various parts of the world.
This project was undertaken to demonstrate proficiency in acquiring and generating GIS data. The selected project is related to my geology expertise but should not be used for site-specific design and construction.
In about 1976, I developed a model debris flow hazard map for a four square mile area of the Montara Mountain Quadrangle between the City of Pacifica and the community of Montara. The model was developed by interpreting several vintages of aerial photographs to identify areas exceeding 20% slope with convex upward curvature. [See Smith (1988) for a more complete description of the methodology.] The initial analysis was completed by hand and eye, requiring days to weeks to complete. The results were presented at a meeting of the Geological Society of America (Smith, 1977). Note that all of the hillslopes in the study area are underlain by rock units that produce sandy soil.
During a 32-hour period on January 3-5, 1982, a major rainstorm triggered about 85 debris flows in the study area, presenting an opportunity to review and refine the method (Smith, 1988). Those results suggested that the methodology was valid and might be useful for mapping mudslide hazards as part of the National Flood Insurance Program. However, unlike 100-year flood zone maps, the method relied heavily on the interpretive abilities of individual scientists. Unless a way could be found to standardize the map creation process, adoption was unlikely.
Within a few years, staff at the US Geological Survey began experimenting with a GIS version of the analysis. However, only one arcsecond (~30 m resolution) data and limited GIS programs were available. Earl Brabb (personal communication, about 1990) noted that those data sets were created by cartographers who, because of high production rate requirements, inadvertently introduced errors in the data that were difficult to detect and correct. First attempts to standardize the process were somewhat disappointing. Although small-scale susceptibility zone maps were produced for the San Francisco Bay region, the small-scale landslide maps coupled with the limited resolution GIS data resulted in only 82% of debris flow sources aligning with the most likely source zones in the calibration areas and 53% in the entire bay region (Ellen et al., 1997).
Today, one-third arcsecond (~10 m) data are available. This project digitizes one of the Smith (1977) maps and qualitatively compares a GIS model using elevation data and ArcGIS 10.
Steps in this project were as follows:
Photographed images were difficult to register due to paper stretch, folds, and distortion. Accuracy issues are illustrated in Figure 1. Time did not permit correction of these data for this preliminary analysis, but that will be done in the future.
Figure
1. The contours of the 1977 map are identical to those displayed on the USGS
quadrangle that was imported into ArcGIS. The match is excellent in the center
of the image, but duplicate lines indicate skew and stretch of the photographed
debris flow source map.
The hand-created debris flow susceptibility map has very smooth boundaries (Figure 2). In comparison, the 1 ArcSec (30-m resolution) elevation raster obtained by USGS produced a noticeable pixilated version of a hillshade and curvature map (Figure 3). Figure 4 shows the effect of using 1/3 ArcSec (10-m resolution) elevation data.
Figure
2. Hand-created debris flow susceptibility map from Smith (1988, plate 13D).
Figure
3. Hillshade and curvature map produced using 1 ArcSec elevation data from USGS.
The 30 m resolution produced noticeable pixilation. The separation between convex
and concave slopes was purposefully narrowed in this analysis. Note that masks
for areas with slopes less than 20% or underlain by alluvium were not applied.
Figure
4. Hillshade and curvature map produced using 1/3 ArcSec elevation data from
USGS. The 10 m resolution reduces pixilation. As in Figure 3, the separation
between convex and concave slopes was purposefully narrowed in this analysis.
Note that masks for areas with slopes less than 20% or underlain by alluvium
were not applied.
The 10 m elevation data produces a noticeable improvement in the resolution of the hillslope analysis. While there are green pixels within red swales, and red pixels on convex ridge tops, they tend to be much smaller. Additional geoprocessing and adjustment of the curvature break parameters might remove those artifacts.
Higher resolution elevation data produces a noticeable change in the detail of the analytical output. Rather than requiring days to weeks to analyze a four square mile area, this type of analysis takes minutes, depending on the extent of the area and whether digitized geologic maps are already available for the area to be analyzed.
The next steps in the investigation are to redigitize the debris flow sources, then compare those locations with the 1/3 ArcSec curvature maps to help refine curvature break values. Masks also need to be applied to eliminate slopes of less than 20% and alluvial areas, as well as using ArcGIS hydrology tools to the map stream network. Additional analysis also needs to develop a method of automatically identifying debris flow fans. Finally, higher resolution LIDAR data might be useful for determining whether small artifacts in the 1/3 ArcSec elevation data are real, whether a buffering function might help smooth the product and produce regulatory maps that are easier to apply, and whether the 1/3 ArcSec data have a resolution that is good enough to develop maps for National Flood Insurance Program purposes.