People must both understand and trust the climate information they receive if they are to respond in an adequate manner.
In 2000, the Limpopo river basin in southern Africa experienced a very substantial rainfall for many days as a result of unusual cyclone activity. Experts knew that it would result in serious flooding - of a magnitude never experienced before by rural communities in Mozambique. Yet very few villages were informed about it.
Most communities had no electricity or radio, yet people were usually able to successfully predict floods by observing ants. Ants build their homes underground; when groundwater rises, they leave their nests - and people know that the water is rising.
On this occasion, the flood came so rapidly there was no time for the groundwater to rise, or for ants to react before the river overflowed.
When someone who had heard the experts' prediction drove to a certain village to tell them to evacuate, the local chief asked him, 'Who are you and why should I do what you say? Since the times of my ancestors, floods have only occurred after ants leave their homes. Now the ants are not moving and you come and ask me to leave?'.
As in most of the Limpopo Valley, many people did not evacuate. About 700 people drowned.
With climate-change risks, traditional knowledge is increasingly not enough because our past experience does not necessarily apply to present and future risks.
There is no point in generating advanced climate models without also preparing to communicate this new knowledge in ways that are sensitive to the context and that can be understood and trusted - especially amongst marginalised communities where trust of the local government may be low.
This article is an excerpt from International Alert's new report on 'Climate Change, Conflict and Fragility'