Auditing the Climate Data

Auditing the Climate Data, by Dr David Evans, from a forthcoming book.

We need to maintain the highest level of confidence in climate data, because the understanding and stewardship of our planet is at stake. We need to apply the lessons and higher standards of fields other than climate science.

I would have little confidence in a public company whose accounts are not audited. I would not invest in such a company, and of course could not invest in such a company on any respectable stock exchange — it is a condition of being listed on the exchange that a company has its accounts audited by an outside auditor every six months.

Twenty years ago, I would have had little confidence in the benefits claimed for a drug in a medical journal. The pressure on researchers and companies to exaggerate were tremendous, and auditing of the processing of results from drug trials was not possible because that information was not disclosed. Since then the medical industry has greatly improved its procedures. It is now standard practice to make the raw data from the drug trials, the methods used to process that data, and the software used to process that data all publicly available — so that interested parties can replicate the final results from the raw data. Government approval for a drug now requires this. Confidence was mostly restored.

I have little confidence in certain climate datasets because (a) they are not audited, (b) they are often revised (sometimes substantially) years or decades after the fact, (c) the adjustments made for imperfect collection instruments are not fully transparent, and (d) the raw data and the pathways to the final datasets are not made fully public with understandable documentation and all relevant software. Beyond that, there is an amazingly consistent tendency for revised datasets to better fit the theory — by sheer coincidence the initial data was nearly always wrong, and after revision it agrees more with the theory.

Replicability is at the heart of science. The final dataset must be replicable from the raw data, meaning someone else can start with a copy of the raw data, run the processing software (full source code supplied), and produce exactly the same final dataset. If human judgement enters the processing, then it needs to be documented and all consequent actions made part of, or accessible to, the processing software.

Climate data is like drug trial data back in the bad old days. The current “trust me” approach is not good enough. Too often the trend in the final version of a dataset appears to differ substantially from the trend in the raw data. There might be good reasons for this, or there might not, but no one apart from a few insiders knows because the data processing is kept secret.

The history of science is replete with examples where raw data has been misinterpreted or incorrectly processed, and wrong conclusions drawn. The process of science is simple: do not trust anything unless it can be replicated. Again, the motto of the Royal Society is relevant: “nullius in verba” or “take nobody’s word for it.” Reputations, motives, money, popularity, political status — they are all irrelevant. What distinguishes science from other fields of knowledge acquisition is that observations and replication are the ultimate authorities on what is accepted as true.

Secret climate business is no longer adequate on a matter as important as the environment and temperature trends. The examples in the financial and medical arenas should be followed, and a regime of full disclosure implemented. Without full disclosure, confidence in climate matters will remain low and dissatisfaction will remain widespread.

The solution is audits that can replicate the final climate datasets from the raw data. This is a narrow technical exercise. The climate scientists should welcome it, just as a responsible public company welcomes an audit of its finances and an honest medical researcher welcomes making all their processing methods and software public. If the climate scientists’ methods and conclusions are basically sound, this will be revealed by the audits. But, and this is not a reflection on either the character nor expertise of climate scientists, I am not going to take their word for it.

Feedback welcome.