Niche Modeling

13
Feb
Lognormal Snowfall

Here is the distribution of annual snowfall in Law Dome Antarctica over the last 750 years (blue), compared to a normal (dashed red) and a lognormal (solid red) distribution.

Remember that in the finest Popperian tradition we are trying to disprove that the snowfall in the last few decades at Law Dome has been unusual. To do this, I have used a robust approach of aggregation (splitting the series into equal sized section), estimating the parameters of the lognormal distribution, then plotting the actual mean snowfall in the final aggregate against the calculated confidence limits.

The previous plots using the van Ommen approach, and an assumed normal distribution, have shown the recent snowfall at LD is significantly high at aggregation from about 25 years on.

While a qualitatively similar result is found when tested with a lognormal distribution, the ‘unnaturalness’ is further reduced. Here the snowfall could be regarded as quite unusual when aggregated between 25 and 40 years, capturing the length of the snowfall period since 1970, but beyond that the snowfall is marginally significant.

However, we don’t really know that this data has a lognormal distribution. There might be arguments for a related distribution with a ‘fatter tail’ that would suggest that extreme events are more probable than the lobnormal distribution would suggest.

Remember the ABC transcript regarding Antarctic snowfall that started this was:

Well, it is a real smoking gun I guess. It could be that we have just happened to find something that is really one in actual fact, thousands flukey event to get such a large snowfall. The more natural interpretation is that there is something been going on in the last 30 to 40 years and we know what that something is. It is the human impact on the atmosphere.

In my view there is some evidence that the event is unusual, and certainly I haven’t falsified the claim here with robust statistics, I don’t think it rises to a 1:38,000 year event as claimed. I suspect that ‘data peeking’ was going on. That is, seeing a large snowfall event, a novel methodology was sought that profiled that event with the highest possible significance.

Tagged as:
Published by david stockwell in: All Climate Statistics
If you like this blog please take a second from your precious time and subscribe to my rss feed!

DavidLHagen

“The more natural interpretation is that there is something been going on in the last 30 to 40 years and we know what that something is. It is the human impact on the atmosphere.”

Besides the statistical test for “something been going on”, this is also the scientific equivalent of a either/or logical fallacy by ignoring other possible causes. e.g. the Pacific Decadal Oscillation, combinations of ocean/atmospheric oscillations, CFC's affecting the ozone hole, variations in solar cycles, other anthropogenic effects, and combinations of such effects. Thus, even if there is a statistical significant variation in lognormal snowfall, the cause of “anthropogenic warming” would not be statistically demonstrated as not having been compared to all other known causes, let alone yet unknown ones.
Anonymous

“The more natural interpretation is that there is something been going on in the last 30 to 40 years and we know what that something is. It is the human impact on the atmosphere.”

Besides the statistical test for “something been going on”, this is also the scientific equivalent of a either/or logical fallacy by ignoring other possible causes. e.g. the Pacific Decadal Oscillation, combinations of ocean/atmospheric oscillations, CFC’s affecting the ozone hole, variations in solar cycles, other anthropogenic effects, and combinations of such effects.

Thus, even if there is a statistical significant variation in lognormal snowfall, the cause of “anthropogenic warming” would not be statistically demonstrated as not having been compared to all other known causes, let alone yet unknown ones.
Charlie A

“That is, seeing a large snowfall event, a novel methodology was sought that profiled that event with the highest possible significance.”That portion of the blog post is a bit over-the-top. It seems to assume an evil intent. There are many other possible explanations for how the claims made in the paper came to be written, and were passed through the peer review process.
tasvanommen

No evil intent here, and for the record, a “desired” result in no way drove my methodology.
davids99us

I didn't mean to imply evil intent. It's human nature to look at the data and try to design analysis is a certain way that supports your view, me too. I have said many times, everyone is biased, but doing a science should involve using processes that correct for out personal biases. I am trying to wrap up this with the most recent post.
Charlie A

“That is, seeing a large snowfall event, a novel methodology was sought that profiled that event with the highest possible significance.”

That portion of the blog post is a bit over-the-top. It seems to assume an evil intent. There are many other possible explanations for how the claims made in the paper came to be written, and were passed through the peer review process.

Anonymous

No evil intent here, and for the record, a “desired” result in no way drove my methodology.

Anonymous

I didn’t mean to imply evil intent. It’s human nature to look at the data and try to design analysis in a certain way that supports your view, me too.

For example, it so happens that the smoothing parameter used by Tas seems to optimize the significance. Was a 10 year Gaussian smoother chosen prior to running the analysis, or as a result of running the analysis at a range of windows, and then choosing 10 years? Its not evil either way, but I think the latter way diminishes confidence in the result.

I am always asking myself if the way I am analyzing things is truly objective. And that is why I run analysis at a range of parameters and scales, not just at a specific scale, because it could be that the scale you chose optimizes the result, and every other scale doesn’t show the influence. I have said many times, everyone is biased, but doing a science should involve using processes that correct for out personal biases.

I am trying to wrap up this with the most recent post.

Charlie A

“That is, seeing a large snowfall event, a novel methodology was sought that profiled that event with the highest possible significance.”That portion of the blog post is a bit over-the-top. It seems to assume an evil intent. There are many other possible explanations for how the claims made in the paper came to be written, and were passed through the peer review process.
tasvanommen

No evil intent here, and for the record, a “desired” result in no way drove my methodology.
davids99us

I didn't mean to imply evil intent. It's human nature to look at the data and try to design analysis in a certain way that supports your view, me too. For example, it so happens that the smoothing parameter used by Tas seems to optimize the significance. Was a 10 year Gaussian smoother chosen prior to running the analysis, or as a result of running the analysis at a range of windows, and then choosing 10 years? Its not evil either way, but I think the latter way diminishes confidence in the result. I am always asking myself if the way I am analyzing things is truly objective. And that is why I run analysis at a range of parameters and scales, not just at a specific scale, because it could be that the scale you chose optimizes the result, and every other scale doesn't show the influence. I have said many times, everyone is biased, but doing a science should involve using processes that correct for out personal biases. I am trying to wrap up this with the most recent post.

Lognormal Snowfall

Categories

Blogroll

Advertising

Comments

Archives

tag cloud

Most commented

Copyright © 2008 - 2012 Niche Modeling