Donna Laframboise and Richard Tol have some pretty devastating criticisms of the IPCC working group 3 report.

Phil Jones is now on the record saying that he doesn’t consider it normal scientific practice to share data and results with other scientists who wish to replicate his findings.  And, it is pretty clear that Hughs tended to get a big fat pass from all his reviewers of published works, stating that no reviewer ever asked for his data, methodoloby, or computer code.

Warwick Hughes makes a pretty good case that in fact Jones was quite open with his data and working papers, as long as he thought the requestor was on his side.  Once he found out certain people were working to replicated and find errors in his work, those people were locked out.   The impressin one gets from his article is that there is now a pretty easy answer to “how can all these climate scientists be so wrong?”  The answer is that they have never had any scrutiny whatsoever on their results, and anyone who attempted such scrutiny were marginalized and vilified by the inner core community.

Stephen Mosher writes:

When it comes to deciding whether to share data or not, standards have nothing to do with the decisions Jones made and he knows that. He knows he shared confidential data with Rutherford while he denied it to McIntyre and Hughes. He knows he regarded the confidentiality of those agreements quixotically. Violating them or hiding behind them on a whim. This was scientific malpractice. Lying about that now is beyond excuse.

Glenn Reynolds brings an interesting example of post-modernist science, where getting the right answer is more important than being factually correct:

Bellesiles, for those who don’t remember, was a historian at Emory who wrote a book making some, er, counterintuitive claims about guns in early America — in short, that they were much rarer than generally thought, and frequently owned and controlled by the government. Constitutional law scholars who expressed doubts about this were told to shut up by historians, who cited the importance of “peer review” as a guarantor of accuracy, and who wrapped themselves in claims of professional expertise.

Unfortunately, it turned out that Bellesiles had made it up. His work was based on probate records, and when people tried to find them, it turned out that many didn’t exist (one data set he claimed to have used turned out, on review, to have been destroyed in the 1906 San Francisco earthquake). It also turned out that Bellesiles hadn’t even visited some of the archives he claimed to have researched. When challenged to produce his data, he was unable to do so, and offered unpersuasive stories regarding why.

Bellesiles eventually lost his job at Emory (and his Bancroft Prize) over the fraud, but not until his critics had been called political hacks, McCarthyites, and worse. But what’s amazing, especially in retrospect, is how slow his defenders — and the media — were to engage the critics, or to look at the flaws in the data. Instead, they wrapped themselves in claims of authority, and attacked the critics as anti-intellectual hacks interested only in politics. Are we seeing something similar with regard to ClimateGate? It sure looks that way to me.

One of my pet peeves in the climate debate is how some folks will immediately describe differences in opinion or interpretation to the fact that someone is lying.  I wanted to show an example of how reasonable people can disagree from the same data set.  This is from a paper written by Vincent Gray (spsl3) in response to an analysis of South Seas sea levels in a series of SEAFRAME reports here.  Mr. Gray believes the authors of the reports have exaggerated sea level rise, and I am sympathetic to his analysis, but I really wanted to show how multiple people can draw different conclusions from the same data.

To begin, lets take the sea level data for Tuvalu from here.  We will graph the raw data, and use Excel to plot a least squares linear fit (the scale on the left is in meters)

The trend we get is about 5.2mm per year of sea level rise  — the actual study Gray is commenting on shows 6mm per year, but its data only went through 2008.

The most noticeable feature on this chart is the depression in 1998, which Gray attributes to the super strong el Nino of that year.  So, I first took this anomalous data out by pasting in data for that period from a previous period (with the months synchronized)

OK, this cut the sea level trend in half, to 2.7mm a year.  Of course, this kind of data fill-in leaves much to be desired.  It was simply an experiment on my part.   I think a better test is to look at the trend since this anomalous event

The trend since the 1998 el Nino has been 0.6mm a year.

So, from the same data, we can reach trends that are an order of magnitude different, from 0.6mm to 5.4mm.  I think the original authors of the study were remiss in not doing more sensitivity analysis, and it would be an interesting test to see if presented with such an anomaly that reduced rather than increased the trend, whether they would have handled it the same way.

Never-the-less, I hope you can see why even reasonable people can draw different conclusions from the same data set.  Thanks to a reader for sending me the original link.

Weather is a chaotic stochastic system.  Outcomes that we typically like to measure – severe storms, tornadoes, hurricanes, temperatures, snowfall — all have mean or average behavior with a large bell-curve or normal distribution around that mean.

With all the talk of record snow in Washington or light snowfall in certain Olympic venues, I feel that a reminder is in order:  There is very little one can deduce about changes or drift in the mean from one or two isolated events in the tail ends of the distribution.   If a kid in your high school gets a perfect score on her SAT, does this mean that the average kid is getting higher SAT scores, that this kid’s score is a symptom of “global smartening?”  Or is this kid’s performance just an isolated event in the tail of the test score distribution?   Katrina and the Washington blizzard seem to occasion a lot of climate conclusions, when in fact I think those conclusions are virtually impossible from such events.

The only really useful role I can see that these extreme events play in the scientific debate is to weed out the credible climate commentators from the charlatans.    If an alarmist says, for example, that the heavy snows in Washington are not necessarily inconsistent with global warming, then he or she is probably relatively safe.  But run away quickly from anyone who says manmade CO2 caused Katrina or, even more incredibly, the Washington snowstorms — they are just nuts.

Of course, the argument typically morphs into folks arguing that extreme events themselves are more prevalent, in other words somehow the standard deviation of the distribution has expanded.  This, in my mind, is one of the weakest arguments in the alarmist arsenal.  The evidence for this is extremely weak (example), and a number of metrics (such as for hurricane activity and large tornadoes) have actually declines over the last decade.  What tends to happen is that the reporting frequency of such events increases, which increases the general perception of having more extreme events — but scientists are supposed to be able to see past such observation biases.

A corollary to this is that extremes in one part of the world do not necessarily mean that the world average is moving in that direction.  Those of us in the US would have sworn January was a cold month, but globally it turns out January was actually a pretty warm month, at least on the historic scale of the last 30 years.  I remember when agricultural futures were first popularized, farmers often went bankrupt forgetting just this corollary.  They would see weather in their area terrible, with terrible crop yields ahead, and they would go long on these crops in the futures markets, only to find the weather in other areas was quite good and they lost a fortune on their futures.