I thought I would briefly comment a little more about the claims made in John McLean’s thesis. Something to bear in mind is why we do research. Essentially, research is very simply about trying to understand something; to answer some question, or test a hypothesis. If you’re lucky, you can undertake a carefully designed, controlled experiment that produces data that can be easily use to answer the question that was posed. In many cases, however, this isn’t possible, and you need to undertake some kind of complex data analysis in order to get an answer.
Consider global surface temperatures. We’d like to understand if we’ve warmed over the last 100 years or so and, if we have, by how much. The problem is that we didn’t set up monitoring stations in the mid-1800s with this in mind. We do, however, have temperature measurements that go back to the 1800s, so we can work with these. However, instruments have changed, measuring stations have moved, the number of measurements has changed, the environment in which the measurements are made can have been artificially altered, the time at which measurements were made can have changed, and – in some cases – even the way in which measurements were made has changed. Therefore, if you want to construct some kind of global surface temperature record, you need to try and correct for these various non-climatic factors.
However, doing so requires developing some kind of data analysis technique and also using some judgement as to how to process this data. It can’t be perfect, but you can test to see how various methodological choices influence the results. People might even disagree with some of these choices; this doesn’t make them wrong. Some of the data might even be processed in a way that is clearly wrong, but if there is a lot of data, you might not be able to check how every data point is influenced by the analysis method. Again, you can check how this kind of thing would influence the results.
So, what about John McLean’s thesis? There’s nothing wrong with checking the data used to generate global temperature datasets. However, if this is to be a serious research project, then a key aspect is to understand how potential errors might influence the results. Simply pointing out possible errors tells us little if we don’t also understand the significance of these errors. By itself, it doesn’t really advance our understanding at all. If you really want to advance our understanding, you need to do more than simply highlight possible data errors (I’m ignoring for now that Berkeley Earth appears to have already flagged most of the issues highlighted in McLean’s thesis).
All I’m really trying to point out is that research is fundamentally about improving our understanding. Auditing a data set so as to point out possible errors can certainly play a role, but by itself does little. This is especially true if another analysis has already identified most of these issues and shown that the impact on the result is negligible.