Fri Mar 4

People could've prepared for the floods better if the impacts of weather forecasts were clearly communicated

Written by Kate R Saunders, Lecturer, Queensland University of Technology

The extreme rain and flooding bombarding Australia’s east coast has inflicted a heavy toll on lives and livelihoods. This, however, could’ve been minimised if weather warning systems had been clearly translated into on-the-ground, local impacts for communities.

Early and detailed impact warnings can give people sufficient time to act. This should include continually updated predictions showing the spatial extent of flooding, timing, and the levels of uncertainty. All this should be easily accessible via a central hub of information.

Access to this sort of information depends on the council. Residents in Brisbane, where half a year’s worth of rain fell in three days ^[1], had only a PDF of a map showing the flood extent. It didn’t show the water’s depth, just whether there was predicted water – and a house submerged from floor to ceiling is very different from having some water on your front lawn.

Improved warnings are unlikely to protect houses, as there’s an inevitable loss of infrastructure that goes along with large catastrophic floods. However, it means people can prepare for floods. Possessions can be moved, vulnerable loved ones can be reached, and people can be evacuated in a timely manner - not in the middle of the night ^[2].

The huge clean up after the floods is underway in many towns across the east coast. AAP Image/Jason O'Brien

Weather warning vs impact warning

The Bureau of Meteorology issues warnings about extreme rainfall, river heights and possible floods. These are usually communicated on social media and in the news.

Individual councils translate these weather warnings into possible impacts for their communities, such as how high the floodwaters may rise and what infrastructure may be damaged.

Early weather warnings are often broad in location and time due to large uncertainties in the predictions, even with the best weather models and mathematical and statistical methods behind them. As the event draws nearer, forecasters can make better predictions with more certainty, with greater detail about the location and timing of the event.

Impact warnings must evolve in the same way: broad at first, and becoming more targeted and specific with time. But the available warnings didn’t make the best use of information about the probability of flood water at different locations and times.

The map provided by the Brisbane City Council, for example, showed only one possible predicted scenario ^[4] for the flood extent at the maximum river height. This doesn’t show people when their place will go under, and many homes were inundated before maximum river height was reached.

The map also didn’t show the predicted water depth at given locations, and didn’t show any uncertainties in its predicted extent. This makes it harder for people to make decisions about how and when to act.

While there are many ways this map could be improved upon, many other councils in flood-affected areas didn’t even have this.

Making good decisions

Broad flood estimates for catchments aren’t always useful for individual action. To make effective decisions for your family and home:

Location matters: to act, it’s important to know if you’ll be cut off by flood waters
time matters: knowing your property is at risk of flooding in three days time offers different mitigation options than “likely to flood” in seven hours
uncertainty matters: people will act differently based on a one in 100 risk than a one in ten risk, and interpret descriptions ^[5] (such as “likely”) in different ways.

Indeed, research shows communicating forecast uncertainties and probabilities helps people make better decisions, increases trust in the forecast ^[6], and even improves compliance with warnings ^[7].

There’s also a much lower cost in taking action early. Of course, early warnings need to be balanced with false-alarms. Acting early is often done in the face of uncertainty, but waiting until the event is about to occur doesn’t give people enough time to act.

When events are low probability and high impact, it’s often better to err on the side of caution.

What needs to improve?

In Australia, different councils have different resources and funding for translating complex weather data into usable flood impact information. For most councils, this isn’t part of their day-to-day business, and they likely don’t have necessary expertise on-staff given how infrequent flood events of this magnitude are.

As climate change brings more frequent and intense natural disasters, improving weather impact warnings and the resources of councils is more urgent than ever.

As climate change brings more frequent and intense natural disasters, improving weather impact warnings and the resources of councils is more urgent than ever. AAP Image/Jono Searle

Drawing from the warning systems in place for bushfires is a good place to start. Australia has a highly localised warning system for bushfires ^[9], which includes: Monitor Conditions, Evacuate Now and Shelter in Place. Information about which roads are cut off are also widely communicated.

A similar tiered warning system linking weather and actions would have greatly helped residents make decisions. Low-lying areas with a moderate to high risk of flooding could then better understand how their flood or isolation risks were growing over time.

What’s more, information about where a flood can occur, its depth, when it will happen, and how likely at a given point in time and space needs to be visualised in an accessible way. This could include interactive, local maps showing the predictions and uncertainty in how the event will evolve.

Australia can also look to the United Kingdom Met Office, which focuses on impacts in their weather warnings ^[11]. It issues colour code warnings based on the impact and the likelihood of the event – for example, a code orange can be issued when an event either has a very high likelihood and a moderate impact, or a very high impact and a moderate likelihood.

Issuing timely, actionable warnings based on impacts isn’t easy. Spatial estimates of floods require advanced modelling techniques, especially when predicting in real time. Uncertainty is also a difficult concept to visualise and communicate, but there is ongoing research ^[12] to address this problem.

These challenges can’t be tackled in an ad hoc manner in the midst of a natural disaster. Data science, better data, and technology all offer solutions for this challenge, and must be an ongoing focus in the lead-up to future, inevitable floods.