Climate models yield confidence question
Grand statements about climate change impacts are all very well for scientists - a global average temperature rise of so many degrees Celsius, a global change in precipitation of such-and-such percent.
But no-one lives on the global average. We all have a home - and what might be very useful, be you a farmer or a city-dweller, would be some precise indications of what the future holds for your farm, your street, your village.
It's precisely what many people here at the UN climate talks are worrying about.
A couple of years ago, the UK government, using science from the Met Office and elsewhere, published a detailed study of likely climate impacts across the UK itself.
The idea was to tell authorities, businesses and communities what they could expect in decades to come in terms of rainfall changes and other parameters important when planning the future.
The project came in for a kicking from some climate modellers who said it was simply impossible to make localised projections with any kind of confidence, given the current state of modelling science.
Now, on the fringes of the UN talks, the Met Office - at the government's request - has published a new study plotting likely climate impacts on 24 countries around the world.
Twenty-one computer models of climate were quizzed for answers on issues such as vulnerability to floods, rainfall changes and suitability for growing crops.
And you can interpret at least some of the findings, again, as an exercise in the unfeasible.
The UK is actually one of the best-studied countries in the world owing to a tradition of weather measurements that dates back centuries.
So the findings for the UK are among the most definite in the report.
As the Daily Telegraph put it, "good news for farmers" - virtually all of the UK's farmland is set to become more productive.
As the Guardian reported it - "millions more at flood risk".
When you look at the figures a little more, however, you see distinct differences in the confidence associated with each of those conclusions.
In calculating the proportion of UK farmland likely to become more fertile, the models' answers ranged from 60% to 99% - pretty firm stuff - and only one projected any losses in any parts of the country.
The flooding picture, however, is different, with estimates ranging from a 56% reduction in flood risk to a 180% increase.
Looking into other countries, even bigger discrepancies materialise.
The change in flood risk to Bangladesh - surely one of the most flood-prone countries in the world even without climate impacts - ranged from -59% to 557%
Dry Egypt could be better off by 100%, or worse off by 206%.
And an eagle-eyed colleague spotted that the proportion of Peruvians likely to be under more serious water shortages was calculated to be a round 0%.
The Met Office team explained that the impacts of melting glaciers were not included in their modelling - and that's set to be a serious issue not only in Peru but the much more populous nations around the Himalayas.
When quizzed about these figures, one of the Met Office scientists said that many other projections were based on single computer models.
Putting the range of uncertainty in the public domain from this large suite of models was, she said, "intellectually honest".
Fair enough. But the exercise also surely gives you an insight into the limits of current modelling when the various models, each of them supposed to be "state-of-the-art", reach such divergent conclusions.
As a policymaker, as a business leader, as a citizen, would you make decisions on the basis of these models?