As many of the rest of you, I’ve been following the extreme weather in the south and south-eastern United States with some interest over the last few weeks. A nice discussion of the links between the record tornado outbreaks, synoptic situation and very warm conditions in the Gulf of Mexico can be found in Jeff Master’s blogs from April. One major consequence of the enhanced rainfall associated with the extreme weather (combined with seasonal snowmelt) has been the significant flooding along the length of the Mississippi River (documented in some detail with extensive links at Wikipedia.
One of the most interesting aspects of the flooding has been the decision making process surrounding the management of the floodwaters at various points during their path down the river. In the USA, flood risk management is split between several agencies at many levels of government, providing a significant challenege to co-ordinate efforts. One early example of this in the 2011 flood was the legal conflict between the US Army Corps of Engineers and the state of Missouri over the intentional destruction of the Bird’s Point levee in order to protect towns further downstream from inundation but resulting in significant flooding of surrounding areas as shown in this NASA image of the day
Making decision like this in a complex geophysical system is obviously very challenging, particularly when people’s property is at stake. As waters surged further downstream, and threatened Baton Rouge and New Orleans, it was very interesting that the Army Corps decided to release maps of potential inundation from various scenarios was made available to the press and public. A key part of the decision making here, opening the Morganza Floodway and flooding a 3,000 square mile area of Cajun county, was jusitified by its near inevitability in all the scenarios studied.
Providing the public access to the detailed uncertainty analysis which underlies key decisions is an important topic for many different geosciences and has been discussed before on this blog and it reminded me of a couple of recent papers I’ve been reading. The key message of both papers is that simply providing uncertainty information to the public doesn’t provide any guarantee that this information will be correctly understood and interpreted for decision making.
The first study (Gigerenzer et al., 2005) looks at how people in different countries use and interpret probabilistic forecasts of rain. In particular, the study asked people in five different cities what the statement “There is a 30% chance of rain tomorrow” indicates to them. Interestingly only in New York did the majority of those surveyed correctly identify that this would indicate that the forecast indicated it would rain “on 30% of days like tomorrow”. In Amsterdam, Berlin, Milan and Athens those surveyed were more likely to think that the forecast indicated that it would rain for 30% of the time tomorrow, with a small proportion who thought the statement indicated rain over 30% of the forecast region. This study highlights one problem with communicating probabilistic information, that it requires a very clear definition of the intended forecast variable (or reference class) along with the forecast. Gigerenzer et al. also tentatively argue that prologned exposure to probabilistic forecasts makes it more likely for the general public to correctly interpret them.
A further problem with probabilistic information is identified by the study of Budescu et al. (2009). This study examines the way in which members of the public interpret statements in the fourth IPCC report associated with a given likelihood of occurence (for example, very likely is defined as having a likelihood of occurence of more than 40%). Even when a translation table for the verbal to numerical values was included in the survey, there was still a low consistency between the IPCC authors intended expression of likelihood and that assumed by the respondents. As noted by Budescu et al., these differences in the interpretation of language by different people are a common feature of many studies, and although they propose several ways to improve on the IPCC AR4 approach, this will nonetheless remain a problem for any attempt to communicate uncertainty verbally.
It seems clear therefore, that there remains a great deal of work to be done in Meteorology, as in other geo-sciences, to understand exactly how best to use and communicate the vast amounts of data about uncertainty which are now available to us.