17 February 2011

Flood Disasters and Human-Caused Climate Change

[UPDATE: Gavin Schmidt at Real Climate has a post on this subject that  -- surprise, surprise -- is perfectly consonant with what I write below.]

[UPDATE 2: Andy Revkin has a great post on the representations of the precipitation paper discussed below by scientists and related coverage by the media.]  

Nature published two papers yesterday that discuss increasing precipitation trends and a 2000 flood in the UK.  I have been asked by many people whether these papers mean that we can now attribute some fraction of the global trend in disaster losses to greenhouse gas emissions, or even recent disasters such as in Pakistan and Australia.

I hate to pour cold water on a really good media frenzy, but the answer is "no."  Neither paper actually discusses global trends in disasters (one doesn't even discuss floods) or even individual events beyond a single flood event in the UK in 2000.  But still, can't we just connect the dots?  Isn't it just obvious?  And only deniers deny the obvious, right?

What seems obvious is sometime just wrong.  This of course is why we actually do research.  So why is it that we shouldn't make what seems to be an obvious connection between these papers and recent disasters, as so many have already done?

Here are some things to consider.

First, the Min et al. paper seeks to identify a GHG signal in global precipitation over the period 1950-1999.  They focus on one-day and five-day measures of precipitation.  They do not discuss streamflow or damage.  For many years, an upwards trend in precipitation has been documented, and attributed to GHGs, even back to the 1990s (I co-authored a paper on precipitation and floods in 1999 that assumed a human influence on precipitation, PDF), so I am unsure what is actually new in this paper's conclusions.

However, accepting that precipitation has increased and can be attributed in some part to GHG emissions, there have not been shown corresponding increases in streamflow (floods)  or damage. How can this be?  Think of it like this -- Precipitation is to flood damage as wind is to windstorm damage.  It is not enough to say that it has become windier to make a connection to increased windstorm damage -- you need to show a specific increase in those specific wind events that actually cause damage. There are a lot of days that could be windier with no increase in damage; the same goes for precipitation.

My understanding of the literature on streamflow is that there have not been shown increasing peak streamflow commensurate with increases in precipitation, and this is a robust finding across the literature.  For instance, one recent review concludes:
Floods are of great concern in many areas of the world, with the last decade seeing major fluvial events in, for example, Asia, Europe and North America. This has focused attention on whether or not these are a result of a changing climate. Rive flows calculated from outputs from global models often suggest that high river flows will increase in a warmer, future climate. However, the future projections are not necessarily in tune with the records collected so far – the observational evidence is more ambiguous. A recent study of trends in long time series of annual maximum river flows at 195 gauging stations worldwide suggests that the majority of these flow records (70%) do not exhibit any statistically significant trends. Trends in the remaining records are almost evenly split between having a positive and a negative direction.
Absent an increase in peak streamflows, it is impossible to connect the dots between increasing precipitation and increasing floods.  There are of course good reasons why a linkage between increasing precipitation and peak streamflow would be difficult to make, such as the seasonality of the increase in rain or snow, the large variability of flooding and the human influence on river systems.  Those difficulties of course translate directly to a difficulty in connecting the effects of increasing GHGs to flood disasters.

Second, the Pall et al. paper seeks to quantify the increased risk of a specific flood event in the UK in 2000 due to greenhouse gas emissions.  It applies a methodology that was previously used with respect to the 2003 European heatwave.

Taking the paper at face value, it clearly states that in England and Wales, there has not been an increasing trend in precipitation or floods.  Thus, floods in this region are not a contributor to the global increase in disaster costs.  Further, there has been no increase in Europe in normalized flood losses (PDF).  Thus, Pall et al. paper is focused attribution in the context of on a single event, and not trend detection in the region that it focuses on, much less any broader context.

More generally, the paper utilizes a seasonal forecast model to assess risk probabilities.  Given the performance of seasonal forecast models in actual prediction mode, I would expect many scientists to remain skeptical of this approach to attribution. Of course, if this group can show an improvement in the skill of actual seasonal forecasts by using greenhouse gas emissions as a predictor, they will have a very convincing case.  That is a high hurdle.

In short, the new studies are interesting and add to our knowledge.  But they do not change the state of knowledge related to trends in global disasters and how they might be related to greenhouse gases.  But even so, I expect that many will still want to connect the dots between greenhouse gas emissions and recent floods.  Connecting the dots is fun, but it is not science.


  1. Roger, the seasonal forecasting skill requirement indeed applies. But the Pall et al. paper is not focused on attribution of a single event. It tries to show what part of event probability (in this case a "UK 2000" type autumn flood event, with threshold 41 mm runoff or more) is attributable to climate change. While I agree that it does not prove that actual trends in losses are attributable to GHG emissions, it does indicate causes of shifting loss event probabilities.

  2. Just to say I've learned a lot from your posts since the Queensland floods, Roger. Thank you for putting these latest Nature papers into proper context - and for yesterday's very clear piece on the complex relationship between precipitation and flooding. The blogosphere is where the crazies are meant to dwell - but this seems an oasis of sanity, compared to our state broadcaster, for example.

    Richard Drake

  3. Received by email:

    "Hi Roger, thanks for the quick turn around on these papers, I bet you have been inundated (sorry) with requests to comment.

    Its been an important message from close observers of historical climate data (you being among my most often consulted) that we need to regionalize our studies, and minimize our use of global data sets to look for climate change signals in the noise. In this spirit, it seems most appropriate to ask about the relationship between rainfall, streamflow, and flooding trends in the _specific regions_ where these researchers have detected an increase in 1 day and 5 day rainfall storm intensity. I hope that is not viewed as a selective, biased approach. For years climate models have predicted increases in rainfall in certain relatively rainy areas (example, the US Northeast), and decreases in rainfall in many dry or seasonally dry areas (example, US southeast). If a global data analysis finds an increase in number or intensity of precipitation events over discrete but large regions of the globe, in parallel with increases in GHG's, we should not use global flood data to see if there's also an increase in flood and related harm (and by extension of Min et al, between GHGs and flood). Increases somewhere may be countered by declines elsewhere.

    So how about flood frequency data specifically for the quite large region that Min et al. point to, over the half-century of their study? You cite a report that tells us that 70% of the areas examined show no increase in flooding, and the 30% that do, half show an increase. So are those increasing flood areas coincident with these areas with increased rainfall, to a greater extent than expected by chance alone? There's the question to ask, hey?

    Also I think it will eventually strike many observers as, well, hairsplitting, to argue that a trend of increasing intensity of rainfall events cannot be connected to human well-being unless and until we can show that flood damage unquestionably rises even controlling for all other contributing factors. All analogies are flawed at some level, but its a bit like seeing sparks from a distant fire settle onto one's rooftop, and not worry about them because, watch as you may, they just keep going out. Rain events and major snowmelt events cause floods when they get intense enough, this is like saying a driver who has many close calls is likely to have a serious accident eventually. Even though other things (previously saturated soil for example) certainly help set the stage.

    And why just focus on flood? What about the other harms almost surely associated with intense precipitation (rain or snow) events, like traffic fatalities during the events themselves? And for intense rain specifically (but also large snow melt events), how about soil erosion on hill sides (even gently sloping), and sewage overflows into harbors and streams, and leaching of nutrients from lawns and farm fields, and deer kills in deep snow?

    Eric Olson, Brandeis"

  4. Note: Eric Olson emails to note that "southeast" should be "southwest" in the above . . .

  5. This article in Science, December 30, 2007

    How Much More Rain Will Global Warming Bring?

    Frank J. Wentz,* Lucrezia Ricciardulli, Kyle Hilburn, Carl Mears

    Climate models and satellite observations both indicate that the total amount of water in the atmosphere will increase at a rate of 7% per kelvin of surface warming. However, the climate models predict that global precipitation will increase at a much slower rate of 1 to 3% per kelvin. A recent analysis of satellite observations does not support this prediction of a muted response of precipitation to global warming. Rather, the observations suggest that precipitation and total atmospheric water have increased at about the same rate over the past two decades.

  6. Since no one has seemed to mention it.

    Taking into account land use change is necessary to determine frequency and reasons for flooding.

    I.E. In the municipality I live in, the planning commission changes from a maximum 20% impervious surface rule to a 60% impervious surface rule. Area's that only occasionally flooded in the past now flood multiple times a year.

    Two of my neighbors had their properties raised 18" as occasional flooding became permanently underwater.

  7. a. Interestingly, the Mendelsohn and Saher paper can no longer be found on the internet, at least not quickly. Your link is broken, as well as several other links found on the web and a search on the World Bank site won’t bring it up either.

    b. The thing about the nature articles is that Nature itself made the leap from the science findings to damages in the News piece by Q. Schiermeier through the decision to bring up the topic of insurance. (Not to mention that which is symbolically represented merely by the journal’s cover this week). With what I (maybe, naively) believe to be a particularly ballsy move, the article quoted Muir-Wood, an industry scientists. However, what he is quoted as saying is admirably clever. Initially it is stated that Dr. Muir-Wood backs the notion that one cannot put the blame of increased losses on climate change. Then, the article ends with a quote from him, “If there’s evidence that risk is changing, then this is something we need to incorporate in our models.”

    This is a very slippery slope and a brilliant double-dog dare. Without doing anything but sitting back and watching the headlines, one can form the argument that “science” supports the remodeling of the hazard risk above the climatological average and is more important then the risks stemming from socioeconomic factors. The reinsurance industry itself has published that socioeconomic factors far outweigh changes in the hazard in concern of losses. The point is (and that which has particularly gotten my knickers in a knot) is that Nature, et al. may wish to consider what it is that they want to accomplish. Is it greater involvement of federal governments in the insurance/reinsurance industry on the premise that climate change is too great a loss risk for private industry alone regardless of the financial burden it imposes? The move of insurance mechanisms into all corners of the earth under the auspices of climate change adaptation? Or simply a move to bolster prominence, regardless of whose back it breaks- including their own, if any of them are proud owners of a home mortgage? How much faith does one have in their own model when they are told that hundreds of millions of dollars in the global economy is being bet against the odds that their models produce?

    What Nature says matters to the world; what scientists say matters to the world- whether they care for the responsibility or not. That is after all, the game of fame and fortune (aka prestige).

  8. Different models predict different rates of increase of precipitation with temperature. There does appear to be a correlation of climate sensitivity with rate of precipitation increase. Models with higher rates of increase have lower climate sensitivities.


    Figure 9.18

  9. So we do, in fact, have a way of screening climate models to be used in ensembles. If Wentz, et.al. 2007 is valid, then any climate model with a low rate of increase of precipitation with temperature should be excluded. Of course, that leads to the problem that the ensemble climate sensitivity will then be at the low end of the IPCC range.

  10. Dr Pielke, just wondering if you could elucidate on a purported interview with Australian ABC. It appears your views on this study have been "under reported". See...http://www.abc.net.au/news/stories/2011/02/17/3141472.htm

    They say..."It is exciting to see the application of innovative approaches to connecting the dots between greenhouse gas emissions and damage from extreme events."

    You say above...:"But even so, I expect that many will still want to connect the dots between greenhouse gas emissions and recent floods. Connecting the dots is fun, but it is not science."

    Did they accurately represent your views? What did they leave out?


  11. The climate catastrophe promoters are still being given credibility they do not at all deserve in the public square.
    That is bad for the public, bad for scientists who tolerate this abuse by their peers,and bad for future policy.

  12. -11-ABC NEWS WATCH

    Sure, my comments were part of a "round-up" put together by the Science Media Centre (London) which you can see in full here:


    Here are my comments in full:

    "It is exciting to see the application of innovative approaches to connecting the dots between greenhouse gas emissions and damage from extreme events. Pall et al. seek to quantify the increased risk due to greenhouse gas emissions for a particular flood event in England and Wales in 2000. Their methodology extends an approach first applied in the context of the European heat wave of 2003. Wide acceptance of such a methodology will most likely have to await the ability to demonstrate skill in seasonal forecasting (of the future) that improves upon methods that do not consider the influence of greenhouse gases. This is particularly the case in situations such as flooding in Wales and England, where the authors observe that the region has seen no long-term trends in either flooding or precipitation. Such important research is in its infancy."

    Single soundbites rarely can capture the nuance involve with such issues, hence the importance of blogs o discuss in more depth.

  13. Thanks for the info. Surprisingly (to me) Australian ABC have still not found time to interview you directly!