Nobel laureate Richard Feynman said it best: “It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If it doesn’t agree with experiment, it’s wrong.”
Testing theories with data is how theories are validated. Yet the need to “‘kick the tires’ on a theory is often overlooked when the media assesses extreme weather.
A recent CBC News story headline declared: “Yes, we’re getting more extreme rainfall, and it’s due to climate change, study confirms.”
According to the CBC, which carried the report on its flagship news show, “Warmer temperatures due to climate change lead to wetter air, and we’ve seen more extreme rainfall and flooding across North America. But is there really evidence that the two are related? Yes, there is.”
Well, no there isn’t. The story was based on an Environment Canada study, Human influence has intensified extreme precipitation in North America. Here’s what the CBC report missed:
• The Environment Canada study presented theoretical projections from models that the data did not consistently validate. In some regions, the models predicted changes and trends quite contrary to those reported and also contrary to measured data.
• It missed the big picture. Watersheds are complex and precipitation is only one-factor affecting flood risk — warmer temperatures mean mitigating factors like less runoff from snowmelt. The study referenced that fact, but not the CBC story.
The CBC’s French service, Radio-Canada (R-C) coverage also linked the Environment Canada study to “sudden intense downpours,” flash flooding, and sewer geysers.
It also neglected data contradicting model results, and limitations of the models in projecting such extreme events.
CBC and R-C coverage has been a problem in the past. The R-C Ombudsman found that a reported increase in severe “100-year” storms was not supported by data and required correction.
The CBC Ombudsman encouraged journalists to be “clearer with their choice of tenses” when separating past, present, and future weather phenomena.
Yet CBC continues to report on urban flooding with statistics saying rainfall is now more frequent when those statistics are not observed in measured data at all but are rather from a numerical simulation modelprojecting the future.
There is an ongoing tendency to confuse models and actual data.
Nobel laureate Daniel Kahneman, the author of Thinking, Fast and Slow, suggested we fall victim to cognitive biases by Thinking Fast, thus moving us away from the rational thought of Thinking Slow.
For example, a “substitution bias” causes us to reject a computationally complex judgment for an easily understood one.
It may be easier to point to neat model results as “proof,” particularly when these results confirm our expectation than to wade through noisy, conflicting observational data. We cannot argue that all floods are caused by climate change.
Fast reporting also missed key study details in the rainfall story, particularly key limitations.
The study itself says, “Many of the physical processes that produce extreme rainfall occur at spatial scales smaller than those that can be reliably simulated by available models. Local-scale events are not well captured.”
It also stated model results may diverge from observations in the west and central west — observations show decreases in extreme precipitation while all three models show increases.
In some regions, it turns out all models are “wrong”!
Are we merely being difficult here? If one- to five-day precipitation is projected to increase, does that not imply that there will be more runoff and flooding in large rural watersheds? Not necessarily.
Even if precipitation increases, warmer weather might mean less snow, and thus less snowmelt.
Such mitigating factors are among the reasons the official attribution study for Alberta’s 2013 flood stated categorically that “no anthropogenic influence can be detected for one-day and three-day surface runoff.”
A broader analysis of major floods across North America and Europe using observational data found “the number of significant trends in major-flood occurrence across North America and Europe was approximately the number expected due to chance alone.”
Checking theories with data shows that there is yet no change in significant floods.
What if we actually look more directly at local data?
For example, how about using rain data to check theoretical models? Data for 651 weather stations in Environment Canada’s Engineering Climate Datasets show that a mere 4.9 percent of annual maximum one-day rainfalls have to date shown a statistically significant increase.
And what about those shorter (two hours or less), sudden intense downpours? For this large data set, that rainfall increased in only just over four percent of the cases.
These small increases, and their flip side (statistically significant decreases), are relatively few and can be explained by chance. So the evidence is lacking on changes in the particular rain events linked to urban flooding.
All of us, including the media, must embrace the reality that weather and infrastructure systems are complex with no simple answers.
Reporting accurately and fairly on complex phenomena means not glossing over model uncertainties or omitting conflicting data.
In promoting action on flooding, certainly a crucial priority, one CBC interviewee stated that “time is not a luxury.” Given the gaps in reporting, let’s hope there’s time for some slow thinking about floods before we rush into fast decisions.
Read more at Financial Post