Technical information, news, research, and opinion on avalanches, snow safety, and winter backcountry travel.

Saturday, April 7, 2012

Unintentional Pseudoscience

As soon as I get my head round you, I come around catching sparks off you—The Jesus and Marychain

This post will probably be unpopular, but I'm going to publish it anyway. I'm sure that I'll lose a few readers, but I'm willing to take that risk in order to say something that I believe is incredibly important.

NWAC and/or Friends of NWAC and/or Huxley College of the Environment ( referred to herein as "NWAC et al." ) recently introduced a feature that allows you to visualise the current avalanche forecast as an overlay on a topographical map. I first saw this feature at the Northwest Snow and Avalanche Summit in November, and I was immediately uncomfortable. Rather than write a post about it last November, I decided to think about it for a while and see if my feelings changed.

They haven't.

Figure 1.1. "Danger Rose" map produced by NWAC et al. Unless asked, I won't post detailed criticism, but I will say that the map is utterly divorced from reality. Look at the huge sections of "moderate" danger. Does anyone really believe that avalanche danger is moderate everywhere indicated by the yellow overlay? If so, when is the avalanche danger moderate? This is a classic case of transforming data from a large scale ( synoptic ) to a much smaller scale ( meso or micro ) apparently without the additional data ( or procedures ) required to perform such a transformation with any reasonable degree of accuracy.

My "understanding" of the avalanche problem is "very advanced". By this, I mean that as my "understanding" has grown over the years, it has become increasingly clear to me that there are many things that can't be known. Perhaps that's why I so often write about uncertainty and complexity.

I recently wrote an email to a researcher who is a colleague of mine. In the email I discussed my travels through the network of avalanche phenomena, and about how uncomfortable it was to realise that many of my old observation tricks rested on empty assumptions. These discoveries made me realise that my record of making "safe decisions" is based in luck far more often than skill. From the email:

Funnily enough, it only took a few videos to convince me that many of the data collection/observation techniques that I've been using in the backcountry are probably very weak, and possibly very dangerous. The inescapable conclusion is that natural phenomena, such as the chaotic interaction of air masses and rough mountain terrain, are complex enough to DEFY our ability to make anything beyond what amount to very simple inferences.

The fact of the matter is that the science of terrain, snowpack, and weather are not nearly advanced enough to allow anyone to create detailed avalanche forecast maps for a large area. NWAC et al. have gone through the motions of at least twice disclaiming the inapplicability of the maps at small scales, even while the map presents the data at small scales. Yes, they've limited the zoom feature, and while I certainly believe that they're trying to do the right thing, the fact of the matter is that the science says this map is dangerous and inaccurate.

I believe that a professor at the Huxley College of the Environment handled the programming, and since I'm not at all sure of his background, I'll keep my comments narrow: the capability to make a map of the avalanche forecast does not mean that it is a good idea, and anyone with a reasonable, working knowledge of the scientific realities of avalanche phenomena should understand why.

Figure 1.2. "Danger Rose" map with single kilometre quadrants. It's very obvious that this map inappropriately presents data at the kilometre and sub-kilometre scales. Each kilometre quadrant is 20x20 pixels, which means that each pixel is 50 metres. 50 metres IS approximately the scale of a single slope. Even though NWAC doesn't allow additional zooming, I think it's fair to say that the danger rose map presents data at the synoptic, meso, and micro scales.

The map CLEARLY shows a synoptic scale forecast that has been inappropriately transformed to the kilometre scale, and possibly to the slope scale as well. Think I'm just being picky? Ask yourself a question: where are the data for each coloured point shown in the map and how were the data generated?

In reality, there's no actual data for most of the points in the map, and the lack of a disclaimer for this "little" fact is misleading AT BEST. Shouldn't the disclaimer say something like "We made this map using data from a variety of computer models and a few weather stations. Even though the map looks complete, we actually have no data for most of the forecast area." That would, at least, be honest.

I feel perfectly comfortable assigning the bulk of the criticism to Huxley College of the Environment. Their logo is featured prominently at the bottom right corner of the map, and they handled the programming. It's Huxley's responsibility to get the science right, and this abject failure rests solely at their feet.

The science of avalanche phenomena tells us unequivocally that such maps cannot be accurate given the current state of the art. I understand the desire to make things easier for people to understand, but this is not the right approach. It's one thing to implement something as an experiment, or even as a tool for professionals, but it is another thing entirely to make it available to the general public, who may not understand the limitations.

As it stands, scientific impossibility relegates this map to the realm of unintentional pseudoscience. NWAC et al. should remove this feature from their web site immediately, and the good folks at Huxley College of the Environment should get a better handle on the science behind avalanche phenomena.

1 comment:

  1. Agreed. Well argued. And there's science to back up your assertion - see Karl Birkeland's PHD dissertation on avalanche danger variability at the mountain range scale:

    Birkeland, K.W. 1997. Spatial and temporal variations in snow stability and snowpack conditions throughout the Bridger Mountains, Montana. Ph.D. Dissertation, Department of Geography, Arizona State University, Tempe, Arizona. 205 pp.

    Birkeland, K.W. 2001. Spatial patterns of snow stability throughout a small mountain range. Journal of Glaciology 47(157), 176-186.