4 November 2010: More Paradise climax avalanches.

How would you forecast wind loading?

Looks like forecasting the directions of daily windloading in the images you've presented isn't easy, but finding the daily aspects that aren't windloaded looks fairly straightforward in the absence of crossloading.

Statement 1
Statement 2
Yes, as I have already stated several times, aspect is in general a poor predictor of avalanches.

Statement 3
Which is it? Statement 1, Statement 2, or Statement 3? You agree that the relationship between aspect and instability may change rapidly-which means aspect is a poor predictor of avalanches. In the same post you go on to say that it's unsafe to suggest that aspect is a poor predictor of avalanches. You can't have it both ways.

I think the problem is that we are using different meanings of the word "predictor" and we haven't agreed what we are trying to predict.

Are we trying to predict the conditional probability of avalanche GIVEN I decide ski that aspect?  That I believe is what trumpetsailor is referring to.  Furthermore, we have lots of additional information besides aspect like snowpack history, results of stability tests, weather history, current weather, slope angle.  All of that, of course, goes into our mental model when we try to calculate P(Avalanche | skied, info) (the probability of avalanche given that I ski the slope, given all the other info I have).

The Grimsdottir paper observes a rather event: the joint probability of avalanche AND slope skied.  It sounds like they then try to derive the conditional probability by dividing through an estimate of the probability of the aspect being skied, in symbols P(Avalanche AND skied)/ P(Skied) = P(Avalanche | skied).  Furthermore, they assert that they have integrated out the "info" term there is in the above expression by not conditioning their observations on "info", so are just finding P(avalanche | skied) for each aspect and comparing these probabilities.  I'm not sure if I think that the integration assumption is a good one or not, but would lean towards not.

Good thread!

This means that, after removing certain assumptions and biases from the analysis, it was clear that aspect is a poor predictor of avalanches.

It doesn't mean aspect is useless, but on the other hand all of the following are true:

Instability lingers on cold aspects.
Cold aspects are less prone to thaw instability.
South facing slopes are prone to crust formation.
Radiation recrystallisation is problematic on south aspects.
Surface hoar and depth hoar form independent of aspect.
Wind loading patterns are chaotic and cannot be resolved simply by evaluating the prevailing wind direction.

OK I get it... It think.
So in stupideze what you are saying is. " Over a season you cannot rely upon aspect alone to avoid skiing a dangerous slope, you must take other data into consideration on any given day"

Looks like forecasting the directions of daily windloading in the images you've presented isn't easy, but finding the daily aspects that aren't windloaded looks fairly straightforward in the absence of crossloading.

Are you serious? While it would be very nice to just "pretend that cross loading is absent", the fact is that cross-loading can be predominant in many types of mountain terrain. But I get what you're saying: "if we just ignore the facts, we can predict wind loading".

I believe aspect is an excellent predictor of instability in combination with other factors that, in general, vary with time. I do agree that aspect alone is not enough.

Yes, but we were discussing aspect as a predictor of avalanches. We weren't discussing aspect + weather + stratigraphy related to a specific situation.

It's misleading to new folks reading this forum to dismiss the importance of aspect entirely because its predictive power isn't simple.

Excuse me?

TrumpetSailor, I don't want to be sharp, but you're not in any position whatsoever to lecture me about responsible communications. I go out of my way to provide good, safe information here and anywhere else that I participate in such discussions.

To this point, I've actually cited research in support of my statements here, I've provided some visualisations, and I've also cited well-accepted facts about the very complicated relationship between aspect and specific types of instability.

You've cited no research and provided no facts whatsoever.

It's misleading to new folks reading this forum to dismiss the importance of aspect entirely because its predictive power isn't simple.

Well, given the research and facts that I cited above, why don't you tell us about the predictive powers of aspect.

OK I get it... It think.
So in stupideze what you are saying is. " Over a season you cannot rely upon aspect alone to avoid skiing a dangerous slope, you must take other data into consideration on any given day"

Yes, exactly. I suppose we can borrow from another famous saying "the avalanche doesn't know you're an aspect".

I think the problem is that we are using different meanings of the word "predictor" and we haven't agreed what we are trying to predict.

Are we trying to predict the conditional probability of avalanche GIVEN I decide ski that aspect? That I believe is what trumpetsailor is referring to. Furthermore, we have lots of additional information besides aspect like snowpack history, results of stability tests, weather history, current weather, slope angle. All of that, of course, goes into our mental model when we try to calculate P(Avalanche | skied, info) (the probability of avalanche given that I ski the slope, given all the other info I have).

The Grimsdottir paper observes a rather event: the joint probability of avalanche AND slope skied. It sounds like they then try to derive the conditional probability by dividing through an estimate of the probability of the aspect being skied, in symbols P(Avalanche AND skied)/ P(Skied) = P(Avalanche | skied). Furthermore, they assert that they have integrated out the "info" term there is in the above expression by not conditioning their observations on "info", so are just finding P(avalanche | skied) for each aspect and comparing these probabilities. I'm not sure if I think that the integration assumption is a good one or not, but would lean towards not.

You'll have to take this up with Dave McClung and Harpa Grimsdottir. I didn't participate in the research.

Good thread!

Agreed!

Are you serious? While it would be very nice to just "pretend that cross loading is absent", the fact is that cross-loading can be predominant in many types of mountain terrain.

Yes, I'm serious. Adding a qualifer to the assessment only assesses its scope of validity. It means that, if I were looking for slopes that weren't windloaded immediately following those days, I would've needed to find some that weren't magnets for crossloading in addition to being predominantly windward.

Well, given the research and facts that I cited above, why don't you tell us about the predictive powers of aspect.

Here are a few thoughts. Please keep in mind that, when looking for links on the web with an eye toward proving a point, one can usually find a favorable source. I did quickly comb the academic literature as well, but the necessarily careful terminology of the modern avalanche researcher does not lend itself well to a quick search. I'm not an avalanche expert, but I do spend a lot of time thinking about and observing snow. Strong absolute statements about anything related to snow science, where, in my experience, the authoritative answer to any specific question is "maybe", get my goat.

An incomplete list of thoughts:

• Windloaded aspects are frequently prone to slides. Backcountry Skiing, Volken et. al (p. 103)
• Different aspects tend to be different. (see Human Factors in the link).
• Weak layers tend to be aspect-dependent. (See Depth Hoar Distribution for only one example.)
• Similar aspects on nearby peaks tend to exhibit similar behavior. (look for the word "east") See also The Avalanche Handbook (p. 175).
• Solar effects occur mostly on sunny aspects.

• Instability lingers on cold aspects.
  Cold aspects are less prone to thaw instability.
  South facing slopes are prone to crust formation.

• The Avalanche Handbook (p. 246) regarding aspect, two sentences down from the last one quoted in an earlier post:

  "Singular data about the present situation, for example, from the Avalanche Bulletin, are normally much more accurate than some fixed rule about aspect, which would many "if," "then," "but" qualifiers."

We differ, in part, in the meaning of the word "aspect". To me, it means the orientation of a slope, usually with respect to something. I think more about sunny aspects, shady aspects, windloaded aspects, etc. than I do cardinal directions, i.e. NE aspects. I will say that in the winter, I'm partial to NW aspects (mmm. pow.), and in the summer, I'm partial to SW aspects (mmm. corn.).

Winter's almost here - time for pow.
[tt]
.SUNDAY NIGHT...SHOWERS LIKELY. SNOW LEVEL 3500 FEET.
.MONDAY...SHOWERS LIKELY. SNOW LEVEL 3000 FEET.
.MONDAY NIGHT...SNOW AND RAIN LIKELY. SNOW LEVEL 2500 FEET.
.TUESDAY...SNOW AND RAIN LIKELY. SNOW LEVEL 3500 FEET.
.TUESDAY NIGHT...SNOW AND RAIN LIKELY. SNOW LEVEL 3000 FEET.
.WEDNESDAY...MOSTLY CLOUDY WITH A CHANCE OF SNOW AND RAIN. SNOW LEVEL 2500 FEET.
.WEDNESDAY NIGHT...CLOUDY WITH A CHANCE OF SNOW AND RAIN. SNOW LEVEL 3000 FEET.
.VETERANS DAY...CLOUDY WITH A CHANCE OF SNOW AND RAIN. SNOW LEVEL 3500 FEET.
.THURSDAY NIGHT...SNOW AND RAIN LIKELY. SNOW LEVEL 3000 FEET.
.FRIDAY...MOSTLY CLOUDY WITH A CHANCE OF SNOW AND RAIN. SNOW LEVEL 3000 FEET.
[/tt]