Question

The Statewide Landslide Information Database for Oregon (SLIDO, Figure 1) is a GIS compilation of point data representing past landslides. Each point is associated with a number of attributes, including repair cost, dimensions, and date of occurrence. For this exercise, I asked whether or not SLIDO could be subjected to a hot spot analysis, and if so, could the results be insightful.

Figure 1: SLIDO (version 3.2).

The Tool

Hot spot analysis is a tool in ArcGIS that spatially identifies areas of high and low concentrations of an inputted weighting parameter. The required input is either a vector layer with a numerical attribute that can serve as the weighting parameter, or a vector layer whose features indicate an equal weight of occurrence. Outputs are a z-score, p-value, and confidence level bin for each feature from the input layer.

Description of Exercise

Performing the hot spot analysis was more than simply running the tool with SLIDO as an input with the weighting field selected. Selecting an input field was easier said than done, as the SLIDO attribute table is only partially completed. Based on a survey of fields in the SLIDO attribute table, it was clear that repair cost was the best choice. All points having a repair cost were then exported to a new layer, which was then inputted into the hot spot analysis. An important note is that this step greatly reduced the number of points, and their scatter, and the output map looks slightly different than Figure 1.

Outputs

The output of this exercise is a comparison of SLIDO points colored by their repair cost with SLIDO points colored by confidence level bin (Figure 2).

Figure 2: Comparison of coloring by hot spots to simply coloring by cost.

Discussion

The second map in Figure 2 shows the presence of a major hot spot and a major cold spot regarding landslide costs in Oregon. The figure shows that, on average, landslides in the northwest corner of the state are more expensive. This observation can only be made because there appears to be a similar density of points, located at similar distances away from their neighbors, across the entire network of points. The figure also shows that single high-cost landslides do not play a major role in the positioning of hot spots, which is a nice advantage of the hot spot analysis.

In general, I think that the hot spot analysis did a good job illustrating a trend that may not have been obvious in the original data.

Bonus Investigation

In the hot spot documentation, it is stated that the analysis is not conducive to small datasets. An example of performing a hot spot analysis on a small dataset is provided in Figure 3. While there may be a trend in points colored by normalized infiltration rate, the hot spot map shows not significant points.

Figure 3: Hot spot analysis on a small dataset.