The recent paper, published in the journal Science, on the 2005 and 2010 Amazon droughts, by Lewis and co-authors, raises the obvious question - are the predictions of 'Amazon die-back' becoming reality?

It is now more than a decade since the Met Office Hadley Centre's earlier climate model was found to project long-term drought and forest loss in the Amazon due to future climate change. Other models, including more recent Met Office models, give less extreme results. However, instead of rushing to say "Look - it's happening!" we need to look very carefully at whether these could be just natural events instead.

Indeed, this is just one example of a whole new aspect of climate science - detection and attribution of climate change impacts. We are now confident that global temperatures are increasing, and that this is most likely due to human greenhouse gas emissions, but as for saying whether this is the cause of events such as the Amazon droughts and Australian floods there's much more work to do there.

Identifying trends in our changing ecosystems

It is very easy to perceive trends or changes in data when, in fact, it is just random noise.

How to identify true trends

• We need a long enough dataset to reliably cover the timescales of variation and change
• When the data show large fluctuations there is a need to employ rigorous statistical techniques to extract any signal of change from this noise.
• Even if a trend is real, it requires a systematic, scientific approach to determine its cause - just because two things are changing at the same time, it does not necessarily mean that one causes the other.

Processes

Understanding of the processes, including controlled experiments as far as possible, are key to attributing changes to their causes.

• Mapping ecological impacts - it is now possible to manipulate environmental conditions over a few hectares of forest, for example by shielding the forest floor from rainfall in order to simulate drought.
• Virtual reality - It's not possible to manipulate for regional climate changes - we only have one Earth - but we can use the 'virtual reality' of computer models, where different factors such as greenhouse gas or aerosol changes can be switched on and off in order to see which combination best explains observed changes in reality.

Meeting the scientific challenge

This will be one of the most challenging tasks of writing the IPCC Fifth Assessment Report. There is already a huge appetite for news on the apparent impacts of climate change, and we can expect to see papers or 'grey literature' reports emerging which attempt to attribute impacts such as ecological changes, food shortages and humanitarian disasters.

The science community's task in IPCC AR5 will be to objectively assess this evidence and establish the true, scientifically well-founded picture. After the controversy over statements on the Amazon in the IPCC Fourth Assessment, establishing the facts over Amazon drought will be top priority.