The East Africa famine calls into question the wisdom of investing in early warning systems without improving take-up, writes Linda Nordling.
When the UN declared a state of famine in the Horn of Africa in July, one group of scientists was not surprised. In August 2010, the USAID-funded Famine Early Warning Systems Network (FEWS NET) had issued a prediction of poor March–May rains this year.
The tens of thousands of people that have died since then in East Africa, and the millions that remain hungry, are a harsh indictment of the ability of science-based early warnings of disaster to make a difference to the continent's most vulnerable people.
As more and better African science data make predictions easier, more effort will be needed to understand and improve the uptake of early warnings at local, national and international levels.
Why warnings fail
So why did the rainfall warning fail to avert the disaster in the Horn of Africa? The problem is multifaceted, and not easy to address.
Writing in Nature, FEWS NET scientist Chris Funk lists a number of reasons, from conflicting climate research predicting wetter conditions in East Africa, to political obstacles (especially conflict in Somalia) and the region's rapid population growth outpacing agricultural yields. 
Funk believes the answer is more science and better agricultural practices to improve yields in areas at risk of drought and food shortages. Working towards a more sustainable long-term strategy for feeding the region's population will limit the need for emergency response, he says.
While most media reports have focused on the international community's failure to respond, Funk focuses on what must be done in Africa to improve uptake.
He is right to do so. The failure of warnings to filter down into African policy at all levels is a graver problem than that of mobilising emergency aid, criticised by many observers for creating a dependency culture in formerly self-sufficient regions.
Data poor no more
The past decade's strong science push in Africa has boosted the understanding of its climate, diseases, and political and economic systems.
This data flood has produced a wealth of mathematical models, and systems are now in place — or being set up — for predicting food price variability, malaria outbreaks, floods and even armed conflicts.
They are all being billed as tools that can revolutionise the continent's response to emergencies.
But it is not just FEWS NET that is struggling to live up to expectations. Last month, Nigeria's emergency management agency accused state governments of ignoring early warnings of floods that killed more than an estimated 140 people and left tens of thousands homeless.
In Nigeria, the problem was a failure of state governments to act upon warnings by clearing drains and evacuating people living in flood plains.
The weakest link
Similar barriers held back the famine warning in the Horn of Africa. "There is a weak link between early warning and response," says Gideon Galu, a scientist with FEWS NET based in Nairobi, Kenya.
Although FEWS NET works with decision-makers in the national government to highlight its findings, the link-up to local authorities and rural communities relies on national government structures, he says.
Kenya, for example, has a food security steering group that passes scientific information on to decision-makers. Nongovernmental organisations and other partners can attend meetings and get a consensus on the scale and magnitude of a problem.
But the final link in the flow of information is problematic in Kenya, and in Ethiopia; and it's particularly weak in countries like Somalia where governance is fragile. There are scant resources and information dissemination networks to make sure that messages reach the local level.
These failures illustrate a key obstacle that many African countries must overcome. The impact of science-based early warning systems hinges on governments' capability (and perhaps willingness) to turn warnings into readiness on the ground.
Avoiding warning fatigue
In time, the early warning community will learn from its mistakes as it establishes strong networks that reach down to the community level — an activity that FEWS NET is working on.
The organisation is also working to address the underlying vulnerability to food crises via the US Feed the Future programme, a US$3.5 billion investment in strengthening agriculture in Ethiopia, Kenya and other countries.
Better links to local government and rural communities will also improve FEWS NET's capability to collect data on variables such as rainfall and temperature, making prediction more of a two-way street, adds Gulu. "They are now able to provide us with additional data about what is happening in the field," he says.
But there is a downside: the explosion in early warning systems means that individual warnings could grow into a cacophony of calls to emergency action that compete for policymakers' attention and scarce resources.
Such 'warning fatigue' would make it harder to discern credible alerts and act on them.
In that future, a warning system well-connected to decision-makers with functioning lines of communication with local officials is likely to be more effective than one where all the money has gone to improving the science while neglecting the capacity to put the evidence to good use.
Journalist Linda Nordling, based in Cape Town, South Africa, specialises in African science policy, education and development. She was the founding editor of Research Africa and writes for SciDev.Net, Nature and others.