Need to Know

Science is understanding data, not data.

Science requires data, and it requires models to understand the data. The best models will be based on well established principles of physics, chemistry and biology. In the context of climate science a brief discussion about the nature of the data and the models used to understand the data will help, but, of course, not be sufficient.

Ten years ago, Thomas Knutson pointed out that “if we had observations of the future we would obviously trust them more than models”. In climate science, we could add that if we had data from well designed measurements in the past, we would also use them, but, unfortunately it is only is the past few decades that a few such measurements are available, driven in major part by improved theoretical models and advances in instrumentation. It is a blessing that many well qualified scientists have spent their time working on the data and the models.

A simple example of the difficulties needed to turn numbers into meaningful data is the time of observation issue for temperature measurement. If measurements had been made at say 3 PM at one location, and then a new station keeper starts taking measurements at 7 AM, without adjusting for the change in time, just looking at the numbers would make it difficult to draw any conclusions. There are people who develop techniques for dealing with this situation. One solution is to compare the measurements with the jump to measurements at nearby stations where the time of observation stayed the same.

Global measurements of temperature only go back to 1880 with a few exceptions and those longer series are bedevilled by serious calibration issues not that the early part of the instrumental record is a golden relic.  More like a rusty old pot that needs a lot of restoration. Paleoclimate scientists have identified proxys for climate variables that can be used, but again, each of these requires careful construction.

Twitter is full of foolish people tweeting about only trusting the unevaluated data. This is about the same as demanding printouts of random number sequences.  To even begin to use data one has to understand how it was measured.  How the data was acquired.  How the data was calibrated.  What the relationship of what was measured to the parameters of interest.  These are just some of the questions that must be answered before the extracted information can be evaluated using theory and associated models.

Similar questions apply to models. As George Box said, all models are wrong, some are useful.  Useful for answering what questions.  Comparable to which data sets.  An interesting point not much commented on is that neither models or data sets do not have to be complete to be useful, just that they isolate the question under study and that there are no significant interferences, or even if there are known intererences what their effects will (approximately) be.

This, of course, raises the question of what do we know, and how can that knowledge be used to guide policy.  We know a lot, and we know more every year. Just go read the IPCC reports.  Even better read the same chapter from WGI in 1990, 1995, 2001 , 2007 and  2014.  As Ms. Liverlara said in high school.  Compare and contrast.

We also know that a human response to a challenge is to deny that it exists, and that when that challenge threatens economic or political interests the response is vociferous. Ask yourself, are 97% of the world’s climate scientists contriving an environmental crisis and being exposed by a plucky band of billionaires and oil companies?