Discussion on 'Large floods in South East Queensland: is it valid to assume they occur randomly?' by GM McMahon and AS Kiem

Abstract

No doubt the authors’ paper was received with many mutterings of ‘You have to be out of your cottonpickin’ minds, Piglets. I’ve been doing flood frequency analyses since I was in three-cornered pants and the idea that floods come in cycles has to be total bovine excrement!’ It is understandable that longterm compliance with Australian Rainfall and Runoff (Pattison 1977; Pilgrim 1987; Ball et al. 2016) could result in practitioners truly believing that annual floods are perfectly random entities and that they always occur according to the Log-Pearson Type III (LP3) distribution. That statistical distribution, first devised in 1888 by Pearson to describe skewed data, has come to us through five U.S. studies from 1966 to 1982, after which Stedinger and Griffis (2008) commented: ‘Bobée and Ashkar (1991, 76) observe that since the official adoption of the LP3 distribution in the United States and Australia, “its application to the study of floods has been both extensive and widespread.” Still a concern is whether the adopted LP3 distribution with log-space moments is a good choice . . . the true distribution will never be known.’ Bypassing that philosophical profundity in favour of practicality, Australian Rainfall and Runoff (Ball et al. 2016) continues to follow U.S. practice in its Book 3 Chapter 2 and encapsulates it in TUFLOWFlike flood frequency software. Since then, the U.S. has produced Bulletin 17C (England et al. 2018) to strengthen a number of identified areas of weakness and has resulted in USGS PeakFQ version 7.1 and USACE HEC-SSP 2.1 software. But with all its tweaking, the LP3 distribution is not omnipotent. Under the sub-heading Decadal Trends in Annual Peak Streamflow, Mastin et al. (2016, 12) declare: ‘In the Pacific Northwest region, decadal shifts in precipitation are linked to atmospheric circulation and sea surface temperatures (Cayan et al. 1998). As result, decadal trends in annual peak flows are evident at many sites’ as shown by their Figure 7. It appears that the semicyclicity of flooding is the new reality for flood hydrologists. An examination of LP3-advocating Bulletin 17C flood records at 01134500 Moose River at Victory VT suggests the biggest floods may occur on a 20-year semi-cycle on the U.S. East Coast (England et al. 2018, Table 10.3). And it is not that Australians are ignorant of the lumpiness of flooding. Australian Rainfall and Runoff (Ball et al. 2016) Book 3 Chapter 2.2.1 states: ‘Climate may experience pseudo-periodic shifts that persist over periods lasting from several years to several decades. There is growing evidence that parts of Australia are subject to such forcing and that this significantly affects flood risk . . . practitioners are therefore advised to keep abreast of new developments.’ Questions for the authors are: