by Climatologists figuring out which data makes their models better
[Via Ars Technica]
In order to improve long-term predictions of global climate change, we need more information about the current and changing environment. Unfortunately, in the current era of government budget problems, expensive satellite climate studies are being cut, so it is important to identify the measurements we need the most, choosing among things like air temperature, pressure, humidity, radiance at various wavelengths, radiation transfer to and from the surface, etc.
One possible way of prioritizing is to figure out which of those measures would help us the most when it comes to projecting future climate change, and focus research funds there. A paper that recently appeared in the Proceedings of the National Academies of Science presents a statistical method for doing this and shows that surface temperature measurements may not be the most useful data to improve surface temperature predictions.
[More]
Reading the comments reveals a tremendous misunderstanding of what the report was about. It is not about cherry-picking data to get the best model.
Data for things as complex as climate models are collected by a wide variety of methods, each with their own error – some measures are quite precise while others have large error bars.
The researchers took a model which incorporated a lot of data from a wide variety of sources and asked the question – Could a smaller number of more precise measurements provide a model that was just as robust as one where all the data could be included?
They wanted to know that if only some methods for collecting data could be carried out, which would be the most informative? It might be that using 3 highly precise forms of measurements would produce a climate model virtually identical to one which included all the available data.
If that were the case, they we should make sure we continue to accumulate data from those 3 forms of measurement.
And that is what they found. Several measurements from satellites for instance, as well as some terrestrial measurement, can reproduce a climate model as well as including all the data used to create the model. Just a handful of approaches can create a model just as robust as including all the different approaches.
In a time when we seem to be limiting ourselves, to preventing some spending for research, we now have an idea of the relative importance of several data collection methods. we can make sure that he get the data using the approaches that will be most informative.
A Man With A Ph.D. 