James Lovelock: Humans are too stupid to prevent climate change.

Here is a pretty good video interview.

Right here.

 DaveInVA wrote: 
Their main problem seems to be the hundred foot rule.  This is frankly unfeasible in the modern world.  However I can see that 10% of the instruments could generate enough data to produce a data set that you can match the rest to.  This is, in fact, what happens: in this article, they say

"But climate scientists who analyze the data say that they are able to account and adjust for the faulty locations by comparing warming trends they spot at bad sites to trends they see at good ones.

"If you use only the sites that currently have good siting versus those that have not-so-good siting, when you look at the adjusted data basically you get the same trend," said Jay Lawrimore, chief of the climate monitoring branch at NCDC.

Lawrimore admitted that Watts' volunteers had discovered real problems with sensor siting, but he said that even when those sites' heat readings were adjusted down, they still showed a steady overall rise in temperatures.

"The ultimate conclusion, the bottom line is that there really isn't evidence that the trends have a bias based on the current siting," he said. 

 callum wrote: 
The only problem is they claim a very large number of them are compromised for various reasons so that much data fudged out of the equation still will not give accurate results...Also with that many of them out of whack you run the risk of using them as the good data and fudging the ones that were right which of course would make the trend calculated useless. I am not saying there isn't global warming but I am saying it needs to be based on accurate data and not guess by people with an agenda. Especially when many of these peoples livelihoods depend on Government grants and funding they are more likely to tweak the data to give them what they want to hear.

 DaveInVA wrote: 
There are two ways of dealing with that sort of thing.  Firstly, you can spend a lot of time and effort re-placing your sensors.  Then you can argue and lobby government to restrict urban crawl so that the set of instruments you very carefully installed on the outskirts of a city doesn't get a restaurant next door.  It is, however, unlikely that you can ever solve this problem by simply 'fixing' the problems (possible to do in theory, in practice it just gets silly).  The other way is to recognise that this is likely to happen and apply some statistical process to identify outliers, instruments with an offset from their local data-set etc.  Those instruments that are just way out of line you discard/investigate - presumably there has been some malfunction or something.  Those that have a similar trend with an offset you simply adjust for the offset.  It happens all the time in the physical sciences.  A good example is one of my first year labs where we had to bounce an ultrasound beam off a steel 'mirror' and measure the angle produced.  however the system for setting the angle the beam made with the mirror was hopelessly inaccurate.  We ended up with a smooth curve based on the sin of the angle centred around about 75 degrees, instead of 90 degrees as we had expected*. This was just due to the offset and we allowed for that in the write-up.  Had we been doing the experiment many times over with slightly different set-ups each time we would have simply focussed on the trends and changes in each data-set and used a set of carefully measured calibration readings to scale the readings off again.

Lazy8 - I agree on all points.  However some of the stuff people are talking about in the media seems to show that not everyone is familiar with the process.  I've certainly never seen anyone explain it in any depth in the media.  Its important that scientists talk about methodology as well as results; if nobody else understands how we get there, it might as well be religion**.  Using data constructively is an great skill and something I'm not very good at doing; I have great respect for the people that do put all this stuff together.



*In fact, due to my tutor popping down into the labs, unscrewing our equipment as we shouted Don't touch that! Please...Noooo...ohbugger... we ended up with a smooth curve up to about 60 degrees, an instantaneous drop and a second section of curve.  But thats professors for you.

**thank you Asimov

 Lazy8 wrote: 
One of the problems that they touched on and that I've also seen touched up by other networks is that many of the Data stations are to close to things that affect their readings. For instance they found one station had a vent from a nearby restaurant blowing on it. Until they correct the compromised data stations any data used from them is misleading and wrong.

 callum wrote:

Well, it's Fox News, so damned sure it ain't neutral.

I think everybody gets the complexity of the problem, that's really not the issue. The issue that's come up is that given that complexity—that difficulty in gaining insight from the data—it's important to not overstate its accuracy. It comes with error bars that are large compared to the trends they're trying to find.

When you have noisy data that you can't recreate (and unless somebody's got a time machine we're stuck with what we have) the way to deal with it is to assign error bars to it and weight it accordingly: the noisiest data gets the least influence on the reconstruction.

The article claims that all the data sets that the various agencies use overlap, so they really aren't independent. True enough; that's what happens when you try to use as large a dataset as possible to nail down trends in stochastic phenomena. There are within those datasets however independent subsets that can be gleaned for insight. Some of the best are ice core data, which by their nature filter out day-to-day/month-to-month spikes and troughs in the data. The tops of glaciers don't generally have urban heat-island effects to correct for either. They don't describe the whole world's climate because there aren't mile-thick glaciers very close to the equator, but we have that data from both the far northern and southern hemispheres. It tells us that the earth's climate has had large fluctuations in the past and that we are in the midst of a warming trend that has been going on a long time.

There are error bars on this data too, but the broad agreement between northern and southern hemispheres makes it pretty compelling. What it doesn't do for us is draw conclusions. That's our job.

Unfortunately the problem of noisy data (which we know how to handle) has been overwhelmed by the problem of noisy activists. In this case the most noise gets the most attention, and it is drowning out the actual science.

 fidget wrote: 
Buy controlling interest in your local water company... soon.

 DaveInVA wrote: 
A confusing article; it couldn't decide where it stood on this, but still didn't seem very neutral.  The economist article further down the page goes a long way to explaining where the differences come about.  Its nigh on impossible to construct a data-set and associated model that is accurate; some degree of compromise is always involved.  

To take an analogy - imagine you were running an event like a sports meet where you were giving out free water to even attendees.  Each person was allowed to drink all the bottles of water they liked, and special bins were provided to recycle the bottles; several water stations were set up.  Now imagine you wanted to know how many bottles of water were actually drunk by the different contestants in each event (to have an idea the next time round).  You need to find out whether the 100m sprint uses more water than the 1500m endurance etc - where you should priorities the water at the next meet at a different ground.  You could count the numbers of crates of water given out at each station, and distribute them evenly amongst the nearby events.  You could count the numbers of empty bottles at the bin and compare distances to each event.  You could do a survey at each event and multiply up.  You could head count people who go to the water stations over time.  You could do all sorts of imaginative things with accounting systems, but you could end up with several data-sets:

Averages of consumption (bottles given out at each station)

Averages of waste (bottles in each bin, visits to toilets?)

Averages of visit numbers (people who visit each water station)

Survey results (etc)

Supply - how many bottles of water did you have to start off?  During the day it might be rather difficult to keep track of where it is all going, but you should know the total number of bottles used over the whole day.

Each of these methods have different advantages and disadvantages, but what produces the best (most accurate*) data?  And how do you relate that to the number of bottles of water used by each different contestant at each different event?  Probably, the best data is generated by some sort of average** of all the different data sets.  But here we come across another problem.  One set of data gives you information about water used at one set of locations, one about bottles thrown away at another set of locations.  How do you relate all the data together to answer the original problem (how much water does each contestant use at each event?).  

To relate each set of data to the actual information you want you need a model.  So you could assume that the water given out at a certain water station is evenly spread between the events around it and split the ground into 'cells' based on which station was nearest.  Or you could do some more complicated analysis just to work out where people go during the day.  Obviously this kind of analysis can be combined - if lots of water is given out in one part of the ground, and lots is thrown away in another then you can assume the people are picking up water and taking it across the ground.  Even the simple survey is complicated: how do you know you got all the right people?  You have to adjust for the fact that a certain sort of person will answer a survey.

With all your data sets added up you can start checking if they are consistent.  You compare summary statistics between them, starting simple and getting more complex.  An obvious place to start is with the total amount of water drunk - do any of your data sets agree with the grand total (almost certainly not!)? Do any of the methods give anomalies (impossibilities like the 100m hurdle participants each drinking 200 bottles of water each or data that stands out, like the water station with the huge amount of consumption because a box of bottles got dropped)?  Once you have fiddled with your data sets to get them all to agree on summary conclusions you can start doing some analysis of the results.  Here you have to be careful again.  Some data might give a really accurate picture of the actual usage in one area, but be awful taken as a whole.  Some data might be great given particular local conditions (are there taps?).  Some data might be plain wrong.
Of course, if someone comes to you with microchips to put in the bottom of all your water bottles, then you could track every bottle used.  That way you could work it all out exactly for one event, and you would have a 'perfect' data set for one day of one event.   Then you can compare your gathered data from lots of events to the 'perfect' data set (or sets), and use this as some pre-determined stationary points for your model to fit to.  That is a very powerful technique and allows you to make a lot of decisions about the data you gathered.  Comparing the data you gathered on the days you also got a 'perfect' model can also tell you how to offset your models given those particular local conditions (weather, number of attendants etc).
So there we go - analysis of a data set is tortuously complex, and I've only been talking about the amount of water drunk at a large sports fair.  Climatologists deal with a many-variable system with a huge number of data points, methods of gathering data, methods of correcting and collating data, correction factors, proposed correction factors and unknown variables.  This is why the data can be better and worse than other data, and also why climatologists never announce anything with certainty.  This is also part of the problem many people have with the climate-gate emails.  When academics say things like 'hide a bump' or to minimise high temperatures, the general meaning is that they have some historical data set that goes back a long way that they wish to adjust to fit a more accurate data set that doesn't cover a long period.  Understanding the science, maths and 'art' of modelling with a data set is difficult - but I hope I have given some idea of how it is done.

*this is a problematic statement in itself - how do you measure accuracy? Do you want to match the actual data or match the trends, or match closely at a certain point or what?
**you can take the mean, or a weighted mean (giving more importance to one set of data or some other thing) 

 samiyam wrote: 
That's really scary - but I'm not surprised, somehow

 miamizsun wrote: 
 It's the first step in shunting the control of the US space program over to the private sector so Chase Manhattan and the Bank of America can end up owning all American access to space.  This will lead to complete monopolization of new techs and all power generation from space.

 DaveInVA wrote: 

NASA Data Worse Than Climate-Gate Data, Space Agency Admits

25 March 10

Monckton in Utah: Hitlerian hyperbole and deviations from the truth

 marko86 wrote: 
South Talpatti Island as it was known in Bangladesh or New Moore Island or Purbasha as it was known in India was a small uninhabited offshore island that emerged in the Bay of Bengal in the aftermath of the Bhola cyclone in 1970 and disappeared at some point in the 2000s.  link

Now you see it, now you don't.  Oops, there it is again.

 sirdroseph wrote: 
Yep, I bet he even promotes that off of Southpark...

 Inamorato wrote: 

I think it was this guy:

 marko86 wrote: 
World peace through global warming! I suppose the deniers will attribute this to volcanic activity, sunspots, or trickery by Al Gore.

Disputed isle in Bay of Bengal disappears into sea

NEW DELHI - For nearly 30 years, India and Bangladesh have argued over control of a tiny rock island in the Bay of Bengal. Now rising sea levels have resolved the dispute for them: the island's gone.

New Moore Island in the Sunderbans has been completely submerged, said oceanographer Sugata Hazra, a professor at Jadavpur University in Calcutta. Its disappearance has been confirmed by satellite imagery and sea patrols, he said.

"What these two countries could not achieve from years of talking, has been resolved by global warming," said Hazra.

The clouds of unknowing

There are lots of uncertainties in climate science. But that does not mean it is fundamentally wrong

The Economist

 jadewahoo wrote: 
I have to say, It is actually more then I like to read. It's not exciting prose, by any means. It is truth though. If you are going to be nay-sayer, you should at least read it,, ALL OF IT, but alas,, I dream.