Summary
Claims made in so-called climate change event attribution studies suffer from gross over-certainties and cannot be trusted.
The techniques used in these studies are in their infancy and do not warrant the trust put into them.
These studies assume either
(a) perfect forecasting models, or
(b) known, uncertainty-free causes of climate change.
Neither condition holds.
Because of this, attribution claims are far too certain or are wrong.
They should not be used in any policy decisions.
Introduction
Time for a change
... There is therefore no ‘natural’ state of the climate, defined as one operating without man’s influence.
We can, however, guess what the climate would look like without man’s influence, but we’d never be able to independently check whether our guess is true.
... Attributions
Certain current weather events are said to be attributable to‘climate change’. These events, some say, would not have appeared or would have been markedly different if the climate was in its ‘natural’ state.
Curiously, events attributed to climate change are always ‘extreme’ or harmful; they are never beneficial.
Nobody bothers to check whether in changed climates there will be an increase in pleasant summer afternoons, or better crop-growing weather.
Re-searchers look only for the bad; it is therefore only the bad that will be reported.
This demonstrates an irreparable confirmation bias in attribution studies. ...
No consistency
... the growing practice of attributing every bad weather event to ‘climate change’ has become a concern to other scientists.
Warnings about going too far (the sky is always falling) and diluting the message are already appearing. ...
Model-based claims
Recall that the output of models of a changed climate and the natural climate are compared to compute probability ratios for a particular event.
... all claims are conditional on the quality of these models.
... model-based climate-attribution claims assume perfect models – which is absurd.
... The models can’t be ‘good enough’ – they have to be faultless for the attribution to have a definite meaning.
Since models are imperfect, this is never the case.
... Note also that the climate models used must demonstrate skill in predicting the kinds of extremes studied.
This is no simple task.
... skill at predicting extremes is low or absent – models tend to exaggerate them
... The global models have to also predict local events well, which they do not.
... It is a well-known adage that any model can be made to fit old data perfectly.
This is why only skill at predicting data never before seen or used in any way must be the only true judge of model performance.
... Conclusion
The desire to say that current notable, harmful or extreme events are caused by man’s activities is strong.
Strangely, this is accompanied by a lack of desire to claim man’s activities produce any beneficial effects.
All events investigated are ‘bad’ events, so these are all that will be reported.
This introduces a strong bias in attribution reports, one that is likely tied to a desire to blame every untoward weather event on global warming.
The journal Climate Change even boasted as much in a call for papers on attributions.
They said pushing attributions in the press can produce ‘teachable moments within a short time after an event’, and ‘can bring clarity to a complex question’.
It is true enough that claims of attribution are clear, but they are also wrong or misleading, as we have seen.
... Models have too much mean prediction error, and unknown but presumably large prediction error of extreme events.
The are thus not trustworthy.
... Observation-based attribution studies assume that man is the sole or most important cause of the changes in observations from before and after an ad hoc date.
Claims that this is so are unproven because the actual or changed climate models used to make them are imperfect.
Also, the uncertainty in measurements of past events, which can be substantial, is never accounted for, rendering these studies meaningless.
It is not that attribution studies are impossible; it is just that they are poor, or worse.
They should therefore not be used for decision making in any public way."