Computer Models
Sunday, November 21, 2010 4:08:22 AM
On two occasions I have been asked, 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
- Charles Babbage
We are just past the one year anniversary of Climategate when emails from University of East Anglia's Climate Research Unit were released. This is where the HADCRUT dataset for historical temperatures is put together. Phil Jones works there and was at the center of climategate. In the emails, there were a lot of things that raised eyebrows. Jones telling people to delete emails in response to FOI requests. Refusal to hand over datasets. Or alternatively, saying the dataset is out there, but not answering where they were, which ones, and where the meta data was located. The HARRY_README file is also something all programmers should read. It just showed how the temperature dataset is really in shambles.
Before getting into the topic at hand, computer models, I want to say that I'm all for saving the environment and doing what we can because it's the right thing to do. Even back in the 80's, when people got called tree huggers and ridiculed in many other ways, I was always for pushing for recycling and all the rest. Those views have not changed. But up until climategate, I was fooled into believing that catastrophic man made global warming was real. After climategate, having seen the data myself, I can only say that if you're still one of the people who believes in catastrophic man made global warming, please research the data.
There is a long list of items that are not debated by either side.
1. CO2 is a greenhouse gas.
2. A specific amount of warming per doubling (logarithmic scale) of CO2.
3. CO2 by itself will not cause runaway global warming.
4. We are coming out of the Little Ice Age.
5. Except for Mann and a few holdouts, the Medieval Warm Period (MWP) does exist as did the Roman Warm Period. Contoversy still exists as to how warm it was when the Vikings settled on Greenland.
6. Forcings and feedbacks with respect to CO2 are the only way to achieve runaway global warming.
7. The Urban Heat Island effect is real. This is another one that AGW proponents took a long time to accept.
8. Clouds and water vapour are the primary greenhouse gas that affects the climate of the planet.
9. It has gotten warmer over the past 100 years. See #4.
10. The uncertainties about the forcings and feedbacks are real.
Items #5 and #7 are recent developments. At least when it comes to the public. And those were points being pushed by skeptics. Mann actually tried to erase the MWP all together.
But skeptics and proponents of AGW basically agree on all these points.
A great many AGW proponent are incredulous when they first hear of this. The reason is that questions like "Do you believe in global warming?" is a loaded one. All skeptics believe that the Earth has warmed over the past 100 years. In fact, they will push that the Little Ice Age existed and that the MWP existed before that. So skeptics are very much PRO climate change. They are the ones that brought back the MWP from Mann's graph.

See the near static temperatures for a thousand years and then the sharp uptick in recent years. This is all bogus. This is why many people turn skeptics. This graph has been debunked more times than all UFO events, 9/11 conspiracies and JFK conspiracies put together. Yet, there are still scientists who back it. Why? I really don't know.
Now we can get into the topic of computer models.
A big question that arises is how can a non climate scientist know if the scientists aren't being honest about their claims? Should we not take the scientist's claim over anyone else's that isn't a scientist? And who am I to say that climate scientists are wrong? No, I'm not a climate scientist, but I don't need to be one because there are a few points that are universal when it comes to computer models. Everyone can use them. And you don't need to know squat about the climate.
It all starts with Babbage's famous quote I posted at the top. Here it is again.
On two occasions I have been asked, 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
- Charles Babbage
This question was asked by government officials if I'm not mistaken. Perhaps senators or congresspeople. I think they were asking if the machine would be able to catch common mistakes that humans do. In any case, Mr. Babbage is quite correct that it's foolish to think that if you put in wrong data that the correct results will come out. You can't balance your budget if you use the wrong figures. That sort of thing.
So right here, everyone and anyone has a very good tool to know if a computation has zero chance of success. It doesn't mean you'll know if the output is correct. But for certain cases, it will tell you with 100% certainty that the output will be wrong. As said earlier, if you see someone use the wrong ledger to input budget information, you don't need to be an accountant to know that the output will be wrong.
RULE #1: Garbage in, garbage out.
Remember it.
We can now go back to Mann's infamous hockey stick graph. Everyone on the planet who knew a little bit of history knew this graph was garbage. There is ample evidence that the MWP existed. This is when the Vikings settled on Greenland. They left as the Little Ice Age was beginning. There was evidence of the MWP all over the globe. Climate scientists were still trying to deny it. But geologists were well aware of this for a long time.
So we knew with 100% certainty that the MWP existed. The above graph was created using a proxy. Instrumental temperature measurements only go back to the 1880's or thereabouts. But historical records and geological evidence was clear that the MWP existed 1000 years ago. So when Mann produced this graph, it was obvious that the input data was wrong. This means that all the research papers that used Mann's graph and data were also wrong.
Rule #1 cannot be avoided. But that did not stop Mann.
He knew his proxies were bad. In fact, the data after 1960's for the proxies showed a decline. What did Mann do? He hid the decline. Spliced it or hid it behind another dataset on the graph. If the proxy isn't valid for certain years, then there is nothing indicating that it would be valid for past years either. It should have been thrown out entirely. More than anything, this is what came out the most from climategate when it was revealed exactly what Mann had done. It wasn't a clever trick. It was sheer dishonest behaviour.
Not only was the MWP proof that his data was wrong, but now we knew that unreliable data was also present. Anything that used this data would now have to play by rule #1. Garbage in, garbage out. So all any skeptic had to do was look if they used Mann's data anywhere and they knew the results of that paper were bogus. So obvious was it that people who knew nothing about climate science were laughing at Mann, Jones, Briffa and all those who supported their data and results.
Climategate wasn't so much a realization that scientists weren't being honest, but rather a revelation as to the extent of what they thought they could get away with. Perhaps they weren't aware of rule #1. I don't know. But it certainly looks that way as most predictions are now done with computer models.
That brings us to the next area I want to talk about. Actual computer models as used by climate scientists. These models have a mindboggling amount of variables. Most of them are very uncertain. What's worse is that you can make the models do anything you wish because the sensitivity to certain variables is so great. With any great uncertainty comes the fact that this data is essentially no data at all. Normally, one would hope that you'd have data that was in the ballpark. But the forcings and feedbacks are a very contentious issue where pro AGW scientists argue amongst themselves. So is it possible that the models are correct? Well, perhaps a few of them if they closely resemble themselves. But there's a good chance that they're all wrong just because of the uncertain input data.
What I find most fascinating is how climate scientists expect people to believe their models because it came from a computer. This isn't the 90's anymore. People use computers. The success and failures are theirs, not the machine's.
Someone reading this might say... "Hold on now. Just because something is uncertain doesn't mean you can't get something relatively close and update it as you gain more information." I completely agree. In fact, this is the way science works. You make a hypothesis. You can make a theory to explain the hypothesis with models and whatever data you have on hand. And over time, you either refine your hypothesis (and models) or you reject the hypothesis outright if it can't be updated.
That gives scientists a LOT of room to wiggle. Again, how can I say their models are bogus? Because the Earth is the only true model. Yes, you can predict certain short term patterns. And yes, you can predict certain long term ones. But one has to understand the view of climate scientists. They don't believe in AGW because they are certain of it. If they were, we could look at the data and go "HOLY SHIT!!!". But even climate scientists don't have the holy grail that would prove them right.
What they have is a lack of alternative explanation.
Read that sentence again. Once more. Do you understand what that means? Read it again. Yes, it means exactly what it says. The ignorance of climate scientists is proof that they are correct. I'm not making that up. In fact, if someone suggests an alternative explanation with any degree of credibility, they will have to research it to prove it wrong. They wouldn't have to do that if they let skeptics also have funding for research, but that's another topic I don't want to get into.
What would it take to simulate the long term behaviour of Earth's climate? How many variables do you think it would take? How much information would you need? At what point would you be able to tell what percentage is due to human activity and what percentage is due to natural variations? The fact that they cannot tell us this at ANY degree of confidence tells us that their models are bogus.
The point I'm making here is not about the input and it's not that they can't tell us a valid result. It's about the model itself. Babbage said he couldn't understand how you could expect to have the correct results with the wrong input. Well, here's another thing that Babbage would not understand either. How can you expect to have the correct result if the algorithm is wrong?
RULE #2: Garbage algorithm, garbage out.
There is just no way you can tell me that they have the correct algorithm. It's impossible. Even if you use the correct data, you will get incorrect results. With the climate, there is much uncertainty not only about the input data, but also about the algorithms used. Heck, we can't even get a decent temperature dataset that doesn't include stations that have moved, are on top of asphalt, have been omitted, vast regions like the polar regions that have virtually no recording stations at all, urban heat island effect, etc. And now we're supposed to believe that they have solved how climate of the Earth works based on this data?
RULE #3: If you can't predict the past, you can't predict the future.
We have past data. As unreliable as most of it is, there is a way to test if some of it is usable. You test your models against two things. The null hypothesis and past results. So if you input random data, the model using actual data should give better and different results. You'd be amazed as to how many models and proxies don't pass the null hypothesis. Take Mann's graph again. It doesn't pass the null hypothesis.
The other test is to take older data and see if you can predict data that is still in the past, but a little in the future from the input data. If you get a good correlation, then you may be onto something. However, you can't cherry pick the data either. For example, most models that actually do this kind of test avoid the 40's to 70's like the plague. This was a cooling period when climate scientists were proclaiming that another ice age was upon us. But predicting this cooling period is a thorn in the side of climate scientists as far as predictive abilities go.
So there you go. When someone says that you're not a climate scientist or <insert title here>, you know that there are certain things that you can tell without knowing a thing about that field. Bogus data, bogus algorithms and not being able to use a model on past data and doesn't give better results than random data are all things that anyone can see right away.
RULE #1: Garbage in, garbage out.
RULE #2: Garbage algorithm, garbage out.
RULE #3: If you can't predict the past, you can't predict the future.
There is nothing here that is specific to any field. These are just facts of causality. If you see any of those three rules happening, then you know that causality has been broken and you need not look any further.
This post is dedicated to all those that aren't affected by causality.


