There but for the grace of God, or the small boy who cried `Wolf’

I am writing this piece as a follow-up to my article on flooding in North Carolina. The news over the July 4^th weekend has been dominated by the tragedy caused by the flash flood which occurred in Texas. For background you should read the Substack article by Roger Pielke. Readers in the UK may remember the flash flood which struck the village of Boscastle in Cornwall on 16^th August 2004. By sheer good fortune there was no loss of life but the damage to lives and businesses was large. A similar flood occurred in Cockermouth in Cumbria on 19^th November 2009, again with no loss of life but causing enormous economic damage to the town of Cockermouth.

Flash floods are a very particular kind of flood which occurs when very heavy rain falls over a short period in a river basin that is narrow and steeply sloping in its upper reaches. This concentrates the run-off so that the water runs into the river and then downstream very quickly. Once the river flattens out the water continues to move rapidly but the water level rises and overwhelms everything in its path until it reaches a flood plain where the water can spread out over a much larger area.

While flash floods can be devastating and cause significant loss of life, their shocking nature arises from the speed at which everything happens. In contrast, the floods which cause the worst loss of life and economic damage – see the Wikipedia page listing the floods ranked highest in terms of loss of life – fall into two very different categories.

The first are due to heavy and prolonged rainfall or snow melt over extended periods in the upper reaches of large river basins. The water accumulates and runs (sometimes quite slowly) downstream to affect more heavily populated areas where people have few options for getting out of harm’s way. In the past, poor communications and infrastructure meant that there was little or no warning of imminent flooding. Today, such events cause great economic damage but smaller loss of life if affected communities heed the warnings that are given. The worst floods have occurred – and still occur - in China (Yangtze, Yellow, and Pearl Rivers) and the Indian sub-continent (Indus, Ganges and Brahmaputra Rivers).

The second are due to storm surges when a combination of strong winds (in the wrong direction) and high tides overwhelm coastal defences. The Netherlands has many recorded floods that caused significant loss of life due to storm surges, but there have been many storm surges caused by tropical cyclones in China and other parts of East Asia. The conditions which give rise to storm surges are reasonably well-understood and there is usually adequate warning to implement evacuation procedures if people heed the warnings.

Tsunamis are like storm surges, but they are caused by major earthquakes, volcanic eruptions, and landslides. Vulnerability to tsunamis is very concentrated with Alaska, Chile, Indonesia, and Japan worst affected. Some Pacific Islands such as Samoa and the Solomon Islands are unlucky enough to be exposed to both tsunamis and storm surges. Their occurrence is effectively unpredictable, though there may be sufficient time to issue warnings and implement emergency evacuations.

These distinctions are important because following tragedies like that in Texas, certain questions always raised: Why wasn’t this prevented? Or: Why weren’t warnings issued? Invariably, the answers run along the lines of “well it is complicated (and/or expensive)”.

As Pielke points out, the region of Texas (Hill Country) where the disaster occurred is referred to as “flash flood alley”.[1] It has both (a) an escarpment that makes it prone to heavy rainfall fuelled by warm moist air from the Gulf of Mexico, and (b) a topography and geology that causes the Guadalupe River to become a raging torrent very quickly. It is known that similar floods have occurred at least 4 times in the last 50 years. Throughout the region, flash floods are a routine occurrence. For example, less than a month earlier on June 12^th 13 people died because a flash flood near to San Antonio swept away their vehicles.

Pielke expresses incredulity that there appears to be no warning system in place either for the Guadalupe River or the wider region covering flash flood alley. Against this, various sources repeat a point made by the county judge that they deal with floods on a regular basis and they have arrangements for sending out weather and flood warnings as well as advice for evacuation. What happened, according to this view, was entirely unexpected and it was made worse by the fact that it occurred overnight on a public holiday. Various weather warnings of heavy rain and flash floods were issued between midnight and 04.15 am on July 4^th. The key flood gauge rose from minor flooding (10 ft) at 03.00 am to extreme flooding (30 ft) at 04.30 am. The water level rose even faster as the water moved downstream.

At the heart of this question is the classic issue of false negatives and false positives that besets all efforts to design medical tests and other warning indicators. Weather forecasters in the UK are haunted by the well-known episode nearly 40 years ago when a TV weather forecaster was asked whether a storm expected to affect the south of England overnight would be accompanied by hurricane-force winds. He said quite explicitly that would not happen, only for the storm to cause the greatest amount of wind damage to trees and property in living memory. Technically, the storm wasn’t a hurricane, but the episode was seen as indicative of the unreliability of weather forecasts. This was a very harsh judgement. There was little practical that anyone could have done had a different answer have been given.

The lesson drawn by the Met Office was that false negatives have a much higher cost (there but for the grace of God) than false positives (the boy who cried ‘Wolf’). Over recent decades it has started to issue many weather warnings for routines weather events – rain, wind, snow, thunderstorms, lightning, heat and fog – for relatively large areas. As the number of warnings has increased, their accuracy and informational value has declined to the point where they have become little more than background noise. To counteract this effect, the Met Office and broadcasters have become increasingly strident in reporting such warnings.

Focusing on the issue of floods, being told that heavy rain is expected is usually of little value. I live in the upper section of the Tweed Valley in the Scottish Borders. Heavy rain in the surrounding hills causes the river very quickly, though these are not flash floods. When the ground is saturated, the increase in water volume causes the river to overflow its banks, which have been built up, and to flood surrounding neighbouring fields. That happens several times in a typical. The floods cause little damage, apart from to a few unwary sheep. That is the point of flood plains.

However, roughly 1 in 5 years the amount of water is sufficient to rise above the next set of embankments, which protect the road. As a result, the road is flooded to a depth of 1+ metres in multiple places and becomes impassable for most vehicles. In these conditions houses on the edge of the road can be at risk of flooding and many properties further downstream are at risk.

Warning of such events would be useful to those at risk from flooding. In truth, it is very difficult for any weather forecaster to identify the circumstances that give rise to such floods because these include the water saturation of hills and fields, the level of several large reservoirs, and precisely where rain falls. Issuing one or more warnings each year would rapidly cause most people to regard them as not being sufficiently reliable to act on.

There is another, more political, aspect to weather warnings. The Met Office and other official weather agencies have become obsessed with linking every weather extreme to climate change. Almost daily announcements of this and that weather record are accompanied by claims that this is what would be expected due to climate change, even though this often takes the form of an unprovable assertion that all aspects of weather are going to be more variable in future. Combining an explicit commitment to lobbying with a desire to attract media attention by issuing all-embracing weather warnings puts the Met Office in the position of the little boy who cried ‘Wolf’ too often. They are not believed by many in the small number of cases when warnings are both useful and likely to be correct.

People who live in Florida – and in cyclone-prone regions of Asia - report that responses to hurricane warnings are tempered by the considerable uncertainty about where such storms will come ashore. Even evacuation warnings may be heeded reluctantly because they tend to cover much greater areas than are likely to be affected by most storms. Such uncertainty is not the fault of weather forecasters. It reflects the random nature of most storms. Equally, it is difficult for forecasters to take account of the costs of false positives to those who are expected to act on warnings.

Rather than increasing the frequency and stridency of weather warnings, there is something important that weather forecasters could do. That is to improve our collective understanding of the variability of weather outcomes making use of good quality weather data. If they wish to claim that certain events are a consequence of climate change, first they should demonstrate that they have a real understanding of weather variability over not just 50 years but 100 or even better150 years. Such periods are a necessary requirement to make claims about changes in the distribution of weather outcomes based on real evidence rather than garbage-in garbage-out models.

Weather disasters like that in Texas prompt, perhaps understandably, a rush to judgement and a much less reasonable desire to score political points. The same happened after the state’s electricity grid experienced a catastrophic failure in February 2021 due to severe winter storms (also over a public holiday), when even more people died. There is considerable merit in Pielke’s argument for a (Natural) Disaster Review Board that functions in a manner akin to the National Transportation Safety Board (NTSB). On the other hand, the NTSB relies upon political and technical procedures that have been forged painfully over nearly 60 years. It is often slow in issuing its findings and recommendations. The latter may be implemented reluctantly or not at all.

It is also naïve to believe that a Disaster Review Board would lessen the pressures that arise out highly polarised views of the role of climate change in increasing the probability of extreme weather events. To be effective any such board would have to be barred from commenting on the role of climate change in natural disasters. If its assessments were used in the way that the Met Office, NOOA and other weather agencies treat extreme weather events as opportunities for advocacy about action on climate change, then their acceptance would be politicized and their impact greatly reduced. On the other side, of course, imposing such a restriction would be vehemently resisted by those who believe that climate change is the most important issue of our age.

As an optimist, I would hope that the first step in formulating policies to address the impact of extreme weather events – from flash floods to winter storms – is understand both their frequency and how their occurrence translates to threats to both human life and the infrastructure on which we rely. Texas is a state of extremes, weather and many other things. For anyone to propose solutions the first things required are both modesty and open-mindedness.

Unfortunately, the realist in me says that none of this will happen. The Met Office and similar weather agencies will go on being shills for a particular view of climate change. The rest of us will simply conclude that they are either incompetent or unwilling to discharge their core functions. Institutional change is very hard under the most favourable conditions. At best what happens is that dysfunctional and unreformable organisations decline into irrelevance, which for a period makes them worse in attempt to counteract their decline.

For those who are not tied into official weather forecasts, I suggest that you consider switching to one of the increasing set of commercial or open-source forecasts. In both the UK and Italy, I have found Accuweather to be more useful than most of the alternatives. For data we use the Open-Meteo API which allows the user to choose from weather data sources with a variety of resolutions and coverage including historical data from the European Centre for Medium-Term Weather Forecasting. There are many alternatives to the Met Office and their ilk. Please use them!

[1] It is important to note that some people familiar with the area who have commented on Pielke’s article point out that this term is not commonly used in the region. There is, however, a widespread awareness among locals that the area is extremely vulnerable to flash floods.