GEOG 410 ASSIGNMENT THREE
Introduction
Caused by a variety of both natural and human factors, landslides occur throughout the U.S. in every state (Department of Homeland Security). Due to the coastal and cascade ranges, Oregon is particularly prone to landslides which can range from moderate infrastructure impact to the complete demolition of roads, bridges, and buildings (USGS). Because of the threat landslides pose to businesses and residents alike, numerous governing bodies have monitoring systems in place to detect and record landslide occurrences. There are several natural and geographic factors that make areas more/less prone to landslides, and the more of these factors that exist in a single area, the more likely that area will experience a landslide event. These factors include proximity to geologic activity (earthquakes and volcanoes), low amounts of vegetation, steep slopes, and high levels of water runoff (GeoScience Australia). The destructive potential of a landslide is dependent on the landslide’s length with shorter slides (sometimes only a few meters) being less destructive than longer ones (can be multiple kilometers). Though the destructiveness of individual landslide events is the result of numerous factors, landslides with a length of approximately 200 meters or more tend to cause significant infrastructural damage (Waythomas & Walder).
For this project, we wanted to statistically describe the the general state of landslides in Oregon. This included finding the average length of landslides as well as finding the probability of an Oregon landslide reaching the critical length of 200 meters mentioned above. Additionally, we wanted to quantitatively describe the spatial distribution of landslides in the state which was accomplished using both nearest neighbor and quadrat analysis.
Methods
The study site of this project was the state of Oregon. Landslide data was collected from the Oregon Department of Geological and Mineral Industries (DOGAMI) which included landslide locations as well as other spatial attributes. Before performing our analysis, we first removed any landslides with duplicate locations as these would cause an error in the SpatStat package. We first performed a quadrat analysis. The state was divided into sixteen quadrats and the number of points in each quadrat was determined. These values were then multiplied by the number of cells so that now we had all the values (x, f, and fx2) needed to determine the spatial mean, variance, and variance-mean ratio. We also conducted a chi-squared test to test for the statistical significance of the spatial distribution. The next analysis was nearest neighbor. The bulk of this was conducted via the SpatStat nndist feature. Nearest neighbor distances for all points were calculated and then used to determine the NND-mean.
Results
Quadrat Analysis (m): 
VAR = 308629.3
MEAN = 284.375
VMR = 1085.29
CHI-SQUARE = 16279.35
P-VALUE = 0.00001
Nearest Neighbor Analysis (m):
NND-bar = 625.3597
NND-r = 3743.315
SIGMA-NND = 29.00813
R = 0.1670604
Z = -107.4856
P-VALUE = 0.00001
Length Attributes (ft):
MEAN = 296.279
STD DEV = 525.0094
Z = 1.96
95% CONFIDENCE = (281.0238, 311.5342)
CRITICAL LENGTH = 656
Z = 0.6851705
P(>656) = 0.253382
Discussion
Both the quadrat and the nearest neighbor analysis show that there is a statistically significant degree of clustering of Oregon landslides. The VMR of the quadrat analysis is much greater than one, indicating a high level of clustering and this result is significant at the 1% level. This result is echoed in the nearest neighbor analysis with the value of NND-bar much lower than the value of NND-r. Additionally, the standardized value (R) of 0.167 is indicates a high amount of clustering and these results are also significant at the 1% level. The analysis of landslide length shows yields a sample average of approximately 296 feet with a 95% confidence interval from approximately 281 feet to 312 feet. Based off of this sample data, there is about a one in four chance (25.33%) that an Oregon landslide will supersede the critical threshold of 656 feet (200 meters) above which major structural damage may occur.
The implications of this study is that landslides do not occur in spatially random distributions. In other words, it suggests that there are causal geographic factors that, when present, make an area much more prone to landslide than it would be otherwise. This conclusion is consistent with the current understanding of landslide and mass movement processes which claim that factors such as slope, water runoff, and soil type all affect the likelihood of a landslide event. Some work that could be done to enhance this study would be different delineations of the quadrats. By dividing the entire state into only sixteen quadrats, we had nearly 80% of the points in only three of the quadrats with multiple quadrats having zero points. This could be addressed either by dividing the state into more quadrats or focusing in on a smaller area. The same approach could help enhance the nearest neighbor analysis too. For example, only focus on the western part of the state with the coastal and cascade ranges – you could then see how the distribution of landslides varies in the areas that are already landslide-prone.
Sources
“Applying Geoscience to Australia’s Most Important Challenges.” What Causes Landslides? GeoScience Australia, n.d. Web. 16 Apr. 2016. <http:// www.ga.gov.au/scientific-topics/hazards/landslide/basics/causes>.
“Landslides & Debris Flow.” Landslides & Debris Flow. Department of Homeland Security, n.d. Web. 16 Apr. 2016. <https://www.ready.gov/landslides-debris-flow>.
“Landslides 101.” Landslide Hazards Program. United States Geological Survey, n.d. Web. 16 Apr. 2016. <http://landslides.usgs.gov/learn/ls101.php>.
Waythomas, Chris, and Joseph Walder. “A Catastrophic Flood Caused by Drainage of a Caldera Lake at Aniakchak Volcano, Alaska, and Implications for Volcanic Hazards Assessment.” Research Gate. USGS Alaska Volcano Observatory, n.d. Web. 16 Apr. 2016. <https://www.researchgate.net/figure/ 240670867_fig9_Fig-9-Length-and-width-values-for-landslide-dams- that- have-experienced-breaching-and>.