by Judith Curry
In Part I, the Congressional testimony of John Christy and Francis Zwiers on extreme events was discussed. In this post, I focus in on issues related to floods. This topic was also discussed in a previous post on Attribution of Extreme Events Part II.
Both Francis Zwiers and John Christy discussed the two recent papers appearing in Nature, by Pall et al. and Min et al. The relevant text from their testimony is excerpted below:
Francis Zwiers’ testimony
While the research required to answer this question specifically in the context of recent events is yet to be completed, two new papers in Nature (Min et al., 2011; Pall et al., 2011) have presented evidence that changes in the intensity of extreme precipitation since the middle of the 20th century may be linked to human induced global warming, and that in at least in one instance, that human influence on climate had likely substantially increased the risk of flooding.
Further, a warmer atmosphere can, and does, hold more water vapour, which has been detected in data (Santer et al., 2007; Willett et al., 2007; Arndt et al., 2010), and which implies that more moisture is available to form precipitation in extreme events and to provide additional energy to further intensify such eventsv. Many of these expected changes have been observed, and in some instances, are beginning to be linked to human induced warming of the climate system.
Heavy and extreme precipitation events have also received a considerable amount of study. Heavy precipitation has contributed an increasing fraction of total precipitation over the regions for which good instrumental records are available (Groisman et al., 2005; Alexander et al, 2006), and particularly over the US (Karl and Knight, 1998; Kunkel et al., 2007; Peterson et al., 2008; Gleason et al., 2008), indicating an intensification of precipitation extremes. Direct examination of precipitation extremes, such as the largest annual 1-day accumulation, or the largest annual 5-day accumulation, also shows that extreme precipitation has been intensifying over most parts of the world for which suitable records are available (Alexander et al., 2006; Min et al., 2011, Figures 3, 4), with an increase in the likelihood of a typical 2-year event of about 7 percent over the 49 year period from 1951 to 1999.
Climate scientists have long argued that an intensification of extreme precipitation is an expected consequence of human influence on the climate system (e.g., see Allen and Ingram, 2002; Trenberth et al., 2003). Indeed, models do intensify extreme precipitation in response to increasing greenhouse gas concentrations and Min et al. (2011) recently showed, using an ensemble of models, that the observed large-scale increase in heavy precipitation cannot be explained by climate variability, and is most likely due to human influence on climate. However, as with extreme temperature events, place and event based research is required to determine whether increasing greenhouse gas concentrations have altered the odds of a given type of event. Pall et al. (2011) demonstrate a suitable approach, and show that human influence from increased greenhouse gas contributions had substantially increased the odds of flooding in England and Wales in the autumn of year 2000 as compared to the world that would have been if greenhouse gas concentrations had remained at pre- industrial levels (Figure 5).
John Christy’s testimony
Svensson et al. 2006 discuss the possibility of detecting trends in river floods, noting that much of the findings relate to “changes in atmospheric circulation patterns” such as the North Atlantic Oscillation (i.e. natural, unforced variability) which affects England. For the Thames River, there has been no trend in floods since records began in 1880 (their Fig. 5), though multi-decadal variability indicates a lull in flooding events from 1965 to 1990. The authors caution that analyzing flooding events that start during this lull will create a false positive trend with respect to the full climate record.
Flooding events on the Thames since 1990 are similar to, but generally slightly less than those experienced prior to 1940. One wonders that if there are no long-term increases in flood events in England, how could a single event (Fall 2000) be pinned on human causation as in Pall et al. 2011, while previous, similar events obviously could not? Indeed, on a remarkable point of fact, Pall et al. 2011 did not even examine the actual history of flood data in England to understand where the 2000 event might have fit. As best I can tell, this study compared models with models. Indeed, studies that use climate models to make claims about precipitation events might benefit from the study by Stephens et al. 2010 whose title sums up the issue, “The dreary state of precipitation in global models.”
In mainland Europe as well, there is a similar lack of increased flooding (Barredo 2009). Looking at a large, global sample, Svensson et al. found the following.
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.
The dreary state of precipitation in global models
Stephens, G. L., T. L’Ecuyer, R. Forbes, A. Gettlemen, J.‐C. Golaz, A. Bodas‐Salcedo, K. Suzuki, P. Gabriel, and J. Haynes (2010), Dreary state of precipitation in global models, J. Geophys. Res., 115, D24211, doi:10.1029/2010JD014532.
“New, definitive measures of precipitation frequency provided by CloudSat are used to assess the realism of global model precipitation. The character of liquid precipitation (defined as a combination of accumulation, frequency, and intensity) over the global oceans is significantly different from the character of liquid precipitation produced by global weather and climate models. Five different models are used in this comparison representing state‐of‐the‐art weather prediction models, state‐of‐the‐art climate models, and the emerging high‐resolution global cloud “resolving” models. The differences between observed and modeled precipitation are larger than can be explained by observational retrieval errors or by the inherent sampling differences between observations and models. We show that the time integrated accumulations of precipitation produced by models closely match observations when globally composited. However, these models produce precipitation approximately twice as often as that observed and make rainfall far too lightly. This finding reinforces similar findings from other studies based on surface accumulated rainfall measurements. The implications of this dreary state of model depiction of the real world are discussed.”
So the issue is this. Even if we assume the following:
- Sea surface temperatures are warming
- More evaporation occurs from warmer sea surface temperatures (provided that surface wind speeds don’t decrease)
- warmer atmosphere can accommodate more water vapor
- Given the constraints of the Clausius- Clapeyron equation, more water vapor in the atmosphere implies more global precipitation
Stephens et al. implies that these premises does not necessarily imply an increase in the intensity or frequency of floods, nor does it necessarily “load the dice” in favor of such events. Floods depend critically on the spatiotemporal distribution of rainfall, which the climate models simulate poorly. If the hypothesized increase in global rainfall is associated with a larger number of weak rainfall events, or a more even seasonal distribution of rainfall, then we wouldn’t necessarily expect to see more frequent floods.
So even if you accept the above 5 premises, which scenario is more likely: the scenario with an increase in floods, or a scenario with no increase in floods? Well, as per Stephens et al., climate models simply don’t provide any useful information on the subject of localized heavy precipitation events. Climate models may have more potential utility in providing statistics on drought, which are tied to larger scale circulation regimes, are regional in scope and have longer duration (e.g. seasonal or longer).
Learning from modeling floods on weather time scales
The challenge of modeling floods is illustrated in this paper by Hopson and Webster. Even using what is arguably the best weather forecast model in the world (ECMWF), there are substantial biases in the rainfall distributions (see their Fig 3 and associated discussion), which can be adjusted. The model has skill in predicting floods for rivers where there is a large catchment area (this particular application is for the Ganges and Brahmaputra Rivers.)
In terms of attribution studies or future projections, it seems that there might potentially be some skill in terms of slow rise floods in Asian monsoon region for the rivers with large catchment areas, provided that the climate models can capture the Asian monsoon rainfall and the statistics of the intraseasonal oscillation.
Floods are notoriously difficult to predict, particularly flash floods. Coarse resolution global climate models do a “dreary” job of simulating the local distribution of rainfall events. Apart from the actual physics of precipitation and the complex interactions between the cloud microphysics and small scale turbulent and convective dynamics, getting the regional distribution of rainfall events depends on accurate simulation of the statistics of weather events, ENSO, blocking patterns, etc. Coarse resolution climate models do not do a particularly good job in simulating the dynamics that are relevant for formation of extreme precipitation events.
The argument might be made that if there are biases in the climate model precipitation, you could still discern a change in the extreme events, either with time or in attribution experiments (e.g. with natural forcing only, and natural plus anthropogenic forcing). However, there is no reason at all to think that the bias statistics would remain constant for these different climate states. Hence, I conclude that climate models have nothing to say on the attribution or prediction of floods, with the potential exception of a few regions such as the Asian monsoon where the floods are driven by relatively predictable rains over large catchment basins.
Further, prediction of actual river flooding depends on more than a rainfall simulation; it also requires a river routing model. Flash floods are unpredictable on timescales more than a day or two, and are not simulated at all in coarse resolution climate models. Hence I personally have no confidence in the studies of Pall et al. and Min et al. in terms of being able to draw any conclusions about the attribution of floods.
It is possible that the higher resolution time slice experiments conducted for CMIP5 may produce better regional distributions of precipitation events, but that remains to be demonstrated. On the broader subject of extreme events, droughts in principle should be much more predictable by coarse resolution climate models.
In summary, I agree with Christy’s statement that “extreme events are poor metrics for global change.”
Moderation note: this is a technical thread that will be moderated for relevance.