Been around for a while with cpdn and i got to thinking: Is this doing any good? I've seen the results page but how to we know they are accurate? So my idea is running a climate model say from the start of year 2000 with all the important starting points and having it run through say 2014. Then, when these are finished, look at the model and compare it to what actually happened between 2000-2014. At that point we would know how accurate these models could be/are. Am i along the right path or am i not looking at it from the right point of view?

Comparing theory with real data IS what is done.
Two examples of this were the Mid-Holocene project, and the Millennium project.

But the models are being run by and for researchers external to Oxford, and publishing the results of their research is largely up to them.

And it doesn't really apply to the "attribution" research, which is mostly what is being done at the moment. These models only run over a short period of time, in a limited area, for whatever event is being studied.

That's not the fundamental purpose of this project. Have you read pages linked in "Main page" at top of blue section, left side of this page? There are pages with explanations and others with scientific papers based on CPDN results.

In the pre-boinc part of CPDN, three 15-year segments were run on numerous tasks for direct comparison of results.

Then a few of us ran about sixty 200-year "Spinups" to provide a set of static oceans for 160-year tasks. Then came coupled atmosphere with dynamic oceans, eventually broken into four 40-year parts because of participant complaints about task length.

Now, global climate models run with regional hi-resolution windows for specific research scientists under parameters designed. Like all projects we run, the scientists have their methods of verification.

[EDIT: Oops, you beat me to it again, Les!]

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

Les and astroWX thank you very much for this information.

Hi jayw900,

This is a good point, but there's slightly more to it even than Les and astroWX have correctly explained.

The way our climate modelling experiments work is to run tens of thousands of models to form what are called ensembles of models. Each model is a plausible but very slightly different version of the world. The climate, as you probably know, is a highly complex, chaotic system, and even models starting with essentially the same starting conditions can end up with different climates.

It's the comparison of thousands of models within an ensemble that lets us make statistical statements, for example about the probability of an extreme weather event occurring.

However, it also means that each individual model may not, on its own, match exactly with reality, but that's ok because we're not looking at one model on its own, we are looking at the average over thousands of models.

So, whilst your model should look plausibly like the real world (unless it's a model from a "world that might have been within climate change" ensemble, in which case it wont!), it may not match exactly what happened with the weather in the real world.

I hope that helps,

And thanks for crunching!

Cheers,
Hannah

---

Hannah Rowlands
--
No longer Communications Officer for climateprediction.net, as of October 2015