Navigating the Spearhead Traverse
Subjective Decision Making
Another part of this project that has to be considered carefully is the fact that throughout the entire project there were several important decisions that were made somewhat subjectively. I like to think that the decisions were all informed by my own research and understanding of avalanches and avalanche terrain, and that significant amount of thought did go into each important decision, but in the end they were subjective none-the-less. The most notable of these subjective decisions were deciding how slope angle should be weighted against landcover in evaluating slopes for potential avalanche danger, and choosing friction values for the potential avalanche danger and landcover surfaces.
For the weighting of slope angle vs. landcover I can justify my decision for a 4:1 weighting with two reasons. Firstly, slope angle is known to be significantly more important than landcover in terms of influence on avalanches (CAA). Even though this is true, it's hard to say that it is exactly 4x more important, but this is where my second justification comes in. Most of the landcover on the alpine Spearhead Traverse is open snow anyways. Knowing this, the landcover component of the MCE is less influential to the final outcome anyways so the weighting of one versus the other does not have a significant effect.
Conclusion and Discussion
I have been comparing my project to John Baldwin's well known map of the Spearhead Traverse, and also my own knowledge of the traverse gained from field experience, as a means to analyze how successful my attempts have been. I am surprised by how closely my proposed route matches Baldwin's overall. I understand that my project has some shortcomings and I think mostly these can be attributed to incomplete data, my lack of technical mastery of the software used for this project, as well as some subjective decision making on my part.
Incomplete Data
All in all, this project was based mostly on terrain and landcover data. The fact is avalanches are vastly more complicated than just two factors and an argument could be made that such a simple rendering of such a destructive phenomenon could be dangerous. The problem is many of the factors that influence avalanche activity are variable in time and thus difficult to map. I could have included the influence that the dominant winds have in snow-loading slopes on particular aspects, but if a different weather system came in and inverted the trend my map would even more inaccurate and possibly even more dangerous. I could have tried to use historical averages to display things like snow cover but I think getting into more specifics than just terrain and landcover is almost beyond the scope of this project. What I set out to do was make a map that would be useful for backcountry skiers and travelers who wanted to do the Spearhead Traverse. Only displaying terrain and landcover, my map is useful all season long and for all different snowpacks. It provides a useful base of information for experienced backcountry users to build upon.
I do admit however, I should have included glaciers into the cost surface. Glaciers are not always a concern when the snow cover is deep, but if my map was to be truly useful to people attempting the Spearhead early or late in the season then glaciers would have been a good addition to the map. Usually parties on the Spearhead have to cross several glaciers anyways, but its better to know that you're crossing one than not knowing. Also, with more time it would have been a good idea to use satellite imagery to fill the holes in the landcover data. There were many "shadows" along the traverse that could have been filled in as flat snow or jagged peaks (which is what I suspect the shadows usually were since they were mostly associated with steep slopes ending in a peak). Because the shadows were mostly associated with steep slopes the suggest route usually avoided them anyways, but a commercial version of this map would obviously need those holes filled in.
Flaw in the Analysis
There was one crucial flaw in the analysis that I can only attribute to my lack of technical mastery of ArcGIS. While I maintain that my analysis did successfully identify dangerous avalanche slopes and steer users away from such terrain, it failed to identify areas downslope of such dangerous avalanche slopes as also being dangerous. This problem is displayed in a clipping of the final map to the left. Here you can see that 'cost path' tool correctly chose to steer the path around Mt. Fissile instead of attempting to cross over it, avoiding its aptly deemed "high risk" slopes. The problem is that the software, under the guidelines I developed, sees no problem skirting along the edge of Mt. Fissile, right below the slopes that are designated as dangerous. If an avalanche were to slide when a party was underneath those slopes they would certainly be caught in that possibly fatal slide. I'm not sure how to correct this using the software, hence the reason field experience in the area is invaluable to a map like this. If I were to produce this map for commercial purposes I would manually override the software to circle around the gentler south side of Mt. Fissile.
Mt. Fissile
The Success of the Path Distance Analysis
The path distance analysis I was a very crucial part to this assignment that I think turned out very well. Trying to make this map without including any thoughts on how the terrain would affect a traveling party would be ignoring a huge component of backcountry safety so the terrain had to be considered. Before using the 'path distance' tool I tried considering slope as it's own surface to be weighted and used in the MCE. Using this approach didn't allow for considering going up a slope as energy draining and going down a slope as energy saving. In the end the 'path distance' tool was the only one that allowed for the type of analysis I thought was fitting for the project, and when compared to other methods I had tried there is a noticeable difference in the path that was produced by each one. Displayed on the right is the analysis performed in four different ways.
The Pink Line considers the avalanche danger only. Placed on top of the DEM we can see that it has decided that the safest way to get to the hut is by getting off of the Spearhead Traverse completely, traveling through the valley, and then climbing up to the hut on a gentle and safe slope. This doesn't take into account how much further the route is, and how much more draining all that elevation loss and then gain would be. The Black Line considers the avalanche danger as well as the terrain but was constrained by only allowing up to +/- 30° slopes to be crossed. It chose a route similar to the pink line; too far, to much climbing.
The Purple Line considers distance alone and in doing so crosses several dangerous avalanche paths and several incredibly steep slopes. Ultimately not feasible for a ski traverse.
The Blue Line is the line that considers avalanche danger with terrain up to +/- 60° slopes. This is the line that was used as the final analysis.
For the weighting of the potential avalanche danger and landcover surfaces I think that my weights were well calculated. It's hard to say whether an extremely dangerous slope is exactly 100x more dangerous than a flat surface, but the point of my map is to be somewhat conservative when it comes to safety. Under my parameters traveling on a PADS category 3 slope (defined as "considerable risk / caution advised") on the most efficient (least costly) "open land" landcover surface, the associated costs would still be higher than travelling through the slowest (most costly) "dense trees" landcover category on the safer category 2 slope (defined as "low risk / likely safe"). Basically, except for between the two safest potential avalanche danger categories, the nuissance of travelling through less efficient terrain is never weighted as being important enough to override the importance of avalanche danger. So while the weights might be somewhat subjective, their outcome is successful in keeping traveling parties safe.
Strange Pixelation in the DEM
There was another issue with the data that did have a minor effect on the final outcome of the product. For some reason the DEM elevation data was somewhat banded. In effect, the DEM was depicting small flat plateaus interrupting what I know to be continuous slopes along the range. This effect is best displayed in the rendered slope data depicted on the right (top). Scanning the final product I concluded that this didn't have any major effects on the project except for in a few areas. This banding affected the multi-criteria evaluations because they weren't recognized as being slopes and thus, small bands of "safety" existed across potential avalanche slopes. This in turn affected the cost path by allowing the path to go through what would be considered costly areas by following the small areas of safety. This can be seen happening in the clipping on the right (bottom), where the path is set going across an avalanche slope by squeezing through the relatively flat area that the DEM and slope data produced by accident. Luckily this only happened in 2 areas on the map and by studying them it appears that the cost path tool didn't have any other great options anyways; even with accurate slope information the path might have done a similar thing. To correct this problem I would likely have to find a DEM that was collected at a larger scale and thus more accurate for specific slopes represented at a relatively large scale.
Conclusion
As this project has worked to create what I consider to be a realistic proposition for the safest, most efficient, while still fun route along the Spearhead Traverse. I consider it to be a successful project on the basis that it has kept safety as the most important factor throughout the data handling and analysis, it considered terrain in a way that I believe to be realistic for a alpine ski traverse, and the final product was a route that is consistent with other maps and trip reports. There were some shortcomings in this project that changed the final outcome from what it should have been in the perfect world. If I were to continue with this project to try and create a commercially available map for backcountry skiers then I would need address some of these issues, primarily the fact that my analysis does not include avalanche paths and runouts. Finding a more precise DEM to eliminate the bands of "safe terrain" along the traverse would also be necessary. After those changes, all that would need doing is cleaning up the map to make it visually appealing and legible to the average backcountry user.
All in all I think this project was a success. I think the important lesson to be learned by this project however, is that a map such as this one cannot be made completely from one's office or lab. These maps require field observations, experience in the terrain, and secondary sources to identify areas where the data is not working as it should. I think GIS is a good start for a project like this but it does leave some ways to go in terms of being accurate and reliable.