#4 Extreme Event Attribution (EEA)

This week I spoke with Dr Chris Brierley, climatologist and lecturer at UCL. He said that if I was ever going to answer the question I asked two posts ago, “if we are changing the climate, are we similarly changing the weather?” I would have to explore the science of Extreme Event Attribution (EEA). So, here we go:

In 2004, Dr. Stott and his colleagues at the UK Met Office published a paper in Nature proving that climate change had doubled the chances of the record-breaking 2003 European heat wave, killing many tens of thousands of lives.

Stott et al., 2004: Human contribution to the European heatwave of 2003

This marked the birth of EEA - developed to assess the level to which we as humans are influencing extreme events. The Bulletin of the American Meteorological Society (BAMS) was set up, annually dedicating a special issue to the extreme event attribution studies that were carried out the previous year. It was sifting through these publications, that I found a key question that I am desperate to explore and I will use it to frame my future research and blogs:

“How important is human-induced climate change in increasing the probability and/or intensity of extreme events?”  

“How does it work?!” – I hear you ask

EEA investigates the many different possible factors influencing extreme events and attempts to quantify their relative importance.These are carried out either using observational data, for example precipitation data collected from a fixed station, or by using climate model simulations of the possible weather in the current climate. More often than not, both these two methods are actually used in conjunction in a ‘multi-method’ approach which can sometimes be the only way to assess confidence. It must be noted, however, that results from these different methods, in fact even between different model simulations in the same study, may not match statistically, making them harder to compare.

Where do we stand?

1. There is varying attribution confidence for each 'type' of extreme weather

There are 3 KEY elements needed for conducting a reliable attribution study, summarised in the table below, and, well, quite simply if you don't have all three then your attribution won't be so "top-notch let's publish it asap": you need good modelling, a high quality observational record and a strong background in the science behind the event (...oh and not to mention lots of money and resources to actually run the models which can run over 1000 times just to show the effects of the changes of one variable like CO2. 

Source : NOAA 2016 - Assessment of state of event attribution science for different extreme event types

Extreme heat events, for example, are one of the easiest events to attribute as they fit all the three criteria and as a result are the most published.

Sun et al., 2014: shows the relationship between anomalies of the number of heat wave days and of summer mean temperatures in Eastern China

However, other extreme events like extra/tropical cyclones and severe convective storms (e.g. tornadoes) are influenced by a far more complex combination of variables so their physical mechanisms, in context to climate change, are not so well-understood. Additionally, the sporadic nature of their occurrence means they have a far lower quality/length of observational record, making them even more difficult to model. I wanted to find out more about this, so I asked Tornado-Titans, my favourite Instagram account (who happen to be doing my dream job chasing and documenting tornadoes) whether they were experiencing these issues:

Click here for IG Account: Tornado Titans

2. Involvement of other forcings

When studying the relative role of anthropogenic forcings, the researchers also need to understand the many other factors at play in producing the extreme event. This includes both natural variations that would be present even without climate change, and other human-caused factors.

For example, wildfires are linked to climate change, however, the risk of fire is also linked to past forest management, natural climate variability, human activity within the forest etc.

3. Framing the research

One of the biggest and most recent controversies over research framing was brought to my attention by Dr Brierley; papers studying the 2011-2015 droughts in California had different phrasing of research questions:

• The paper in which the research question focussed only on the role of precipitation deficit in causing the drought found little to no anthropogenic influence.
• However, the papers that broadened their research questions to studying the effects of both precipitation deficit and high temperature did.

In terms of statistical reliability, all were significantly confident in answering the questions they had set, but how can papers proving very different answers all be right?

Even within the same paper, attribution questions have to be framed differently for each methodology used. As the field is relatively new, there are still differing opinions on how to conduct and interpret attribution investigations. In fact, even defining what makes "EXTREME" weather is debated, and Jack Wharton does an excellent blog post on this if you fancy a follow-up!

4. Selection bias 

This does not affect the validity of an attribution case individually BUT is relevant when conducting a meta-analysis across a range of studies. These include things like studying events with an expected strong anthropogenic influence (choice bias), shown by the distorted proportion of BAMS papers of 2015 on 'heat' related topics; or choosing events of interest to the analyst such as those funded by stakeholders in wealthier countries, exemplified by the bias number of studies on cyclones hitting the coastal cities of the US (type bias). 

It is clear that at the moment it is not possible to definitively answer how climate change is impacting all weather events...but challenges aside, the progress that has been made in such a short amount of time from the first EEA paper is pretty impressive.

See you next week!