This is a post that has been at the top of our to-do list for years…but with so many other to-dos it hasn’t made it on to our website. However, with water on the ground across the country, it seems like a good time to get this draft version up.
Statistics are tough
Statistics were not my favourite subject at school (and most definitely not at University as the concepts got more and more challenging). I expect the same can be said for many. However, statistics have their place – especially when you want to understand flood likelihoods, severities and maps (have a read of this blog post to get a primer on these terms if they are unfamiliar).
Will my home flood?
As a business owner, home owner, or planner what you really want to know when the water is rising, or better yet when you are making decisions or planning for future floods, is:
what is the probability that my home/business will flood?
And, not just if it will flood today, or tomorrow, or this year – but if it will flood while I pay off my mortgage, while I live here, or while my children live here.
Using statistics to understand if my home will flood?
As a starting point, most flood maps, decisions and reports are based on a design likelihood (we’ve got another blog post brewing on why this is not best practice). This is usually referred to as an X-year flood (e.g. 100-year flood, 1000-year flood). What this actually means is that there is a 1 in X chance of a flood of a given severity occurring this year or in any year. So, for example, there is a 1% chance that a 100-year flood will occur in any year. The percentage representation of this is called the Annual Exceedance Probability or AEP – the annual chance that a flood if a given size will be equaled or exceeded. This is much cleaner way of describing flood sizes and likelihoods than using return periods. Return periods can mislead – for example, it is commonly believed that if a 100-year event has occurred, it will not re-occur for another 99 years, which is incorrect.
Using encounter probabilities to better understand and make better decisions.
Another way to think about hazard likelihood is through the use of encounter probabilities, where it is possible to calculate the likelihood of encountering an flood of a given size over a defined time period—for example, the length of an average mortgage (25 years) or the average lifespan of a human (75 years). Table 1 shows that for a 1% AEP event there is a 22% chance that an event of this size or greater will occur over a 25-year period. As a homeowner (or potential home buyer looking to buy a home) in a 1% AEP floodplain – the knowledge that there is more than a 1 in 5 chance that your house will flood while you pay off your mortgage could be a game changer.
Table 1: Encounter probabilities for various likelihoods.
|Annual Exceedance Probability (AEP)||Indicative Return Period||Encounter Probability of Occurrence in 25 years||Encounter Probability of Occurrence in 50 years||Encounter Probability of Occurrence in 75 years||Encounter Probability of Occurrence in 100 years|
|30%||Once every three years||100%||100%||100%||100%|
|10%||Once every 10 years||93%||99%||100%||100%|
|3%||Once every 33 years||53%||78%||90%||95%|
|1%||Once every 100 years||22%||39%||53%||63%|
|0.1%||Once every 1000 years||2%||5%||7%||10%|
And what about climate change?
Unfortunately, all the stats presented above assume a stationary climate (i.e. an assumption that history is representative of the future). This is not the case. As climate changes, the stats of flood and other climate-driven natural hazards are becoming more and more challenging. And, therefore we need a new way to think about and mitigate damages that considers multiple futures. We’ll muse on this in a future post. In the meantime, there’s lots of other flood musings on our pet projects page, or you can check out our approach to dynamic and changing probabilities to flood using Probability of Inundation (POI) curves in our work for the City of Vancouver.
In summary, understanding flood likelihoods is key to good decision-making. Return-periods are a poor means of communicating probabilities. And, the use of encounter probabilities is a much better tool to help people understand the statistics of what seem like extreme events.