This is about understanding the aims of your communication, because you can’t achieve your aims unless you are clear about them in your own mind. A warning aims to change people’s behaviour, through clear advice and messaging – a forecast aims only to give people information which they can use in their own decision-making: there is no advice or message. The two require different communication techniques and would be evaluated very differently. 

This will help you understand the needs of your audiences so that you can provide them with the information that they want and need, and not just the information that you want to give them or assume that they want. Different audiences might want different things, so you may have to prepare different outputs for them. You may also want to work with audiences to help them know how to respond to the forecast – what actions they may consider. This is particularly advised in schools, but also with others, such as infrastructure managers. When communicating to the public, use the channels that they are already used to using for similar information rather than trying to invent your own: work with journalists in all media, weather forecasters etc. They will also likely know their audiences well and be able to help. Regular meetings with those who are going to try to interpret and disseminate forecasts can also help ensure that any new people are familiar with the format and ready to help their audiences.

Just as you as a communicator had to decide what the aim of your communication is, so your audience also needs to know this aim. If the audience are expecting a warning with a clear behaviour message, they will be confused and unhappy with a forecast that doesn’t give them that. Forecast information can help people make all sorts of decisions, depending on their own circumstances – it is useful even when absolute probabilities are low and there is no official ‘advice’. For example, people might choose to practice an evacuation drill, test shut down procedures for power stations, identify diversion routes that avoid tunnels or bridges – all depending on their responsibilities. You’re not advising people to do any of those things as part of the forecast, but you can give people a sense of decisions that they or others might make. Making it clear why the forecast is being made public and how it might be of use to some people will help the audience know how to respond to it.

It is tempting to avoid making decisions on behalf of your audiences, and instead to allow every variable to be customised (e.g., the geographical area and time frame of the forecast, the threshold of event size being considered, metrics such as intensity or magnitude). However, if communicating with the public, this is not helpful for most and instead makes the forecast more complex and difficult to understand. Add as little customisation as possible to the main interface: if possible, allow the audience to vary only one thing - the location for which the forecast applies. For more experienced audiences you can allow more customisation via a ‘settings’ option but work with these audiences to identify the variables they need to change and keep the customisation via settings only to those.

People find it increasingly easy to interpret formats as they become familiar with them, so it is useful to expose people to a format frequently. Additionally, familiarity with what ‘normal’ looks like in terms of the likelihood of a seismic event will help people interpret the likelihoods displayed in times where the risk level is higher and so make sense of it (which it is important that they are able to do).

Geographical maps are formats that are very familiar to most audiences, and so they frequently express a liking for information presented in that way. A geographical map can be useful to help them know the area over which a forecast is calculated and valid, but for individuals only interested in the forecast at one particular location, reading absolute risks off a map (illustrated as isolines) seems to be harder than simply displaying the absolute risk probability to them. It may also confuse people into thinking that the isolines on the map represent something other than the likelihood of an event of a set magnitude (e.g. the geographical distribution of intensity of, or damage caused by, a forecasted earthquake, since such maps are often used after an event to illustrate exactly that).

It’s easy to concentrate on communicating the difficult aspect of the numbers involved in the likelihood, but forget that the numbers involved in the impact (e.g. magnitude or intensity) also need to be given a context for people to know what this might mean for them. Although an earthquake of the same magnitude or even intensity can have very different effects depending on the types of buildings in an area, just giving people a sense of what sort of effects different levels of earthquake might have, and in a way that references their experience (e.g. giving examples of earthquakes that they might remember reports from) rather than theoretical (e.g. not through a description like ‘light damage likely’, which people found unhelpful in our interviews).

People shown a probability as a more solid and imaginable expected frequency perceive that probability to present a higher chance of happening than when shown a percentage. However, such expected frequencies also help people discriminate between low probabilities, which would have several decimal points if shown as a percentage (and where changes of whole orders of magnitude are difficult for people to discriminate). A percentage, though, even with decimal places, is seen as clearer and easier to read, so presenting the absolute risk as a percentage in a big, bold font as the main output seems sensible, whilst also showing the interpretation underneath as an expected frequency. Graphics such as bar charts or icons are not helpful for the low probabilities most often applicable to seismic events.

Giving context to the absolute risk of a seismic event is important to help people interpret the (otherwise fairly meaningless) probability of such an event. If they have become used to seeing the forecast regularly (see point 5 above) then they may already have an idea of what ‘normal’ (baseline) is, but the format should not rely on that knowledge. Experiments where people were given the baseline risk in an area as a piece of context to a forecast gave inconsistent results. Where people were given the relative risk between the current and baseline probabilities (e.g. ‘twice as high as average’) people had a higher perception of the risk (if it was elevated above baseline), but it also seemed to inhibit discrimination between different probabilities.  We therefore don’t recommend using either a baseline or relative risk as a way to add context to an absolute forecast probability.

What does seem to give people useful context to interpret the current probability of a seismic event however, is a graphical comparison of this probability against the average probability in other cities with which they are likely already to have some sense of the likelihood of an earthquake. These cities are likely to be within their country, but may be international in the case of some famously high hazard locations (e.g. Tokyo or Los Angeles). The scale on the risk ladder should be linear, not logarithmic (it can be cut off just above the highest baseline risk). We acknowledge that calculating these comparator risks is complex. Some members of the audience are likely to find the risk ladder ‘too much information’ – especially alongside both the percentage and expected frequency formats described in point 7 - so it is probably better as optional additional information.

Even if you don’t communicate this information as part of your forecast service, work with others who can provide this information to the relevant audiences. People want to know what to expect if an event occurs – how long an aftershock sequence might last, what sort of effects an earthquake might have, how long before things are likely to return to normal, and who has responsibility for different actions.

Although an automated forecast service might be cost-effective, in order to really support your audiences you should also provide alongside it a short (1-2 sentences) written explanation for the forecast, and a continually-available personal service for those who need to ring to have an explanation of the forecast. It may be worried members of the public, journalists, or key infrastructure managers or political decision-makers – all will need to be supported by a one-to-one service, staffed by people who understand the technicalities of the forecast but are able to communicate about it at any level required. This is no small undertaking, but an important one.

It is helpful to be able to pre-empt potential misunderstandings (such as the difference between a forecast and a prediction, or that large seismic events can happen without warning at any time even in ‘low hazard’ areas). However, it is important to approach this task with humility. Recognise that there is a great deal not known within the geosciences, and that a lot of information is not ‘true’ or ‘false’ but a matter of degree, interpretation or debate. Also recognise that many different cultures have different views on the role of aspects such as fate and that insensitivity to this can easily create barriers between groups and foster mistrust – which can undermine all attempts at communication.

More information on how to fight earthquake misinformation

Example of the guidelines being tested. See http://earthquake-forecast.wintoncentre.uk  (source code freely available for use at https://github.com/WintonCentre/rise-dashboard).

1. Becker, J.S., Potter, S.H., McBride, S.K., Wein, A., Doyle, E.E.H., Paton, D. (2019) "When the earth doesn’t stop shaking: How experiences over time influenced information needs, communication, and interpretation of aftershock information during the Canterbury Earthquake Sequence, New Zealand." Int. J. Disaster Risk Reduct. 34, 397–411. doi:10.1016/j.ijdrr.2018.12.009
2. Becker, J.S., Potter, S.H., McBride, S.K., H. Doyle, E.E., Gerstenberger, M.C., Christophersen, A. (2020) "Forecasting for a Fractured Land: A Case Study of the Communication and Use of Aftershock Forecasts from the 2016 Mw 7.8 Kaikōura Earthquake in Aotearoa New Zealand." Seismol. Res. Lett. doi:10.1785/0220190354
3. Wein, A., Potter, S., Johal, S., Doyle, E., Becker, J. (2016) "Communicating with the public during an earthquake sequence: Improving communication of geoscience." Seismol. Res. Lett. 87, 112–118. doi:10.1785/0220150113
4. McBride, S.K., et al. (2018) "Developing earthquake forecast templates for fast and effective communication." 11^th Natl. Conf. Earthq. Eng. 2018, NCEE 2018 Integr. Sci. Eng. Policy 12, 7310–7316.
5. Michael, A.J., et al. (2019) "Statistical seismology and communication of the USGS operational aftershock forecasts for the 30 November 2018 Mw 7.1 Anchorage, Alaska, earthquake". Seismol. Res. Lett. 91, 153–173. doi:10.1785/0220190196
6. Dallo, I., Stauffacher, M., Marti, M. (2020) "What defines the success of maps and additional information on a multi-hazard platform?" Int. J. Disaster Risk Reduct. 49, 101761. doi:10.1016/j.ijdrr.2020.101761
7. Dallo, I., Marti, M. (2021) "Why should I use a multi-hazard app? Assessing the public’s information needs and app feature preferences in a participatory process" Int. J. Disaster Risk Reduct. 57, 102197. doi:10.1016/j.ijdrr.2021.102197