Disaster early warning systems: What works and who gets left behind?
From alert to agency.
One in five adults has been impacted by a disaster in the past five years, and impact follows vulnerability.
Within low-income countries the burden falls further on the most vulnerable: 37% of adults with primary education or less report being impacted, against 20% with higher education.
Early warnings.
Four in five of those impacted were warned. The warning gap follows the income gap.
Good progress for the UN Early Warnings for All initiative, which commits the world to universal coverage by the end of 2027.
Digital and broadcast channels reach furthest.
SMS reaches more people than WhatsApp; radio and television remain important broadcast channels. The key question is not only how many warnings people receive, but which channels extend coverage.
Much of the technology needed to close the warning gap already exists.
Expanding SMS-based warnings is the single largest opportunity to close the coverage gap. Phone ownership is not the same as reliable network coverage, but it is a clear place to start.
Being warned and being able to act are different things.
The ability to act splits more sharply by income than coverage does. High coverage does not guarantee response: more than nine in ten adults impacted by a sandstorm were warned, yet only around half of them could act.
The second and third channels materially increase the likelihood of action.
Beyond three, each extra channel adds four to eight points to the share able to act, and resilience barely moves, from 58 to 60. Having multiple channels raises the chance a message is received; repetition across channels raises the chance it is believed.
Warnings work best alongside preparedness. Warning infrastructure alone does not produce action.
Agency is the belief that you can protect yourself and your family in a disaster. A household plan is one known to everyone in the home over the age of 10.
The two compound. Together they triple the share able to act compared with having neither.
What the findings ask of policymakers.
They face the highest rate of disaster impact, the lowest warning coverage, and fewer than half of warned adults are able to act (48%).
Prioritise the countries where low coverage and high exposure overlap, particularly across Africa and parts of Asia.
One warning is better than none, but the jump in both action and individual resilience comes with the second and third channel.
Of adults who received no warning, 77% own a mobile phone, more than 170 million people. SMS-based warnings are the largest opportunity to close the gap.
The difference between 27% and 83% acting on a warning is closely linked to whether people have a household plan and believe they can protect their family.
A coverage figure records that a warning was sent. The share of warned people able to act records whether it worked.
From insight to action.
A pre-registered review of the evidence, synthesised narratively and thematically.
Protocol developed and registered on the Open Science Framework before commencing the review.
Academic databases and grey literature, including a call for evidence. Studies published from 1990 onwards were eligible.
Quantitative, qualitative, mixed-methods and economic evaluations were included.
Quality assessment using the appropriate tool for each study design.
Narrative synthesis using SWiM guidelines for quantitative studies, and thematic analysis of qualitative studies.
A consultation group was consulted three times during the project: at protocol development, interim findings and final reporting.
111 studies across 54 countries, published between 1995 and 2026.
Most studies explored experiences or associations rather than establishing causal effects. Most appeared from 2020 onwards, and quality varied by design.
Robust effectiveness evidence concentrated in high-income settings, while implementation evidence more often came from lower-middle- and low-income settings.
National systems and hydrometeorological hazards dominate the evidence base.
Hydrometeorological hazards were the focus of most studies: especially floods, heavy rain, cyclones, storms, drought and extreme heat.
Community-based systems were mainly flood-focused and implemented in lower- and middle-income settings. All five global-system studies concerned earthquake warning systems.
One tsunami study found fewer deaths. Other findings on deaths, illness and injury were mixed.
| Hazard | Mortality | Morbidity |
|---|---|---|
| Tsunami | Fewer deaths (one study) | No evidence found |
| Heat | Mixed or unclear | Fewer cardiac-related ambulance call-outs |
| Floods | Mixed or unclear | No evidence found |
| Typhoons | Mixed or unclear | Fewer typhoon-related injuries |
Some studies reported fewer typhoon injuries, or fewer cardiac-related ambulance call-outs during heat warnings, but these benefits were not consistent across the evidence base.
The strongest evidence is for reduced asset losses, particularly from floods.
reduced damage when people knew what protective action to take.
Reduced asset losses, particularly from floods.
Evidence of reduced losses was more mixed.
Only one descriptive study examined livelihood outcomes, assessing perceived effects on crops and livestock without statistical testing.
Effective warning systems have five interconnected components.
Multiple complementary communication channels improve reach.
Trusted local intermediaries and institutions support credibility.
Local languages and accessible, understandable formats improve comprehension.
Clear, location-specific instructions on what to do support action.
Preparedness training, evacuation planning and response support make action feasible.
These components work together to reduce the impact of extreme weather events. Technology alone cannot deliver impact.
Who gets left behind, and what is missing from the evidence?
- Digital or text-only systems can exclude people with limited connectivity, literacy or device access.
- Language, disability, age, education, occupation and institutional trust shape whether a warning is understood and acted on. Gender shapes access and action.
- Sparse subgroup analysis hides differential effectiveness.
Evidence on differential effectiveness for underserved groups remains limited. Where reported, disabled people, those with low literacy or no education, and those facing a language barrier were left behind.
Communities value credible, timely and locally relevant warnings that give them time and clear options to act.
- Community participation and local ownership.
- Trusted actors and clear communication through multiple channels.
- Preparedness, evacuation planning and coordinated institutions.
- Low trust and delayed, confusing or inaccessible warnings.
- Weak infrastructure, limited funding and poor maintenance.
- Fragmented responsibilities and insufficient local capacity.
Organisations and governments recognised the value of systems but reported recurring constraints in infrastructure, coordination, funding, responsibilities and local capacity.
The Poll and the review converge on what makes warnings useful.
| Shared priority | World Risk Poll | Systematic review |
|---|---|---|
| Multiple channels | Second and third channels raise action | Multi-channel dissemination is a core component |
| Preparedness | Plans and agency lift action | Training, evacuation planning and response support enable impact |
| Trust and relevance | Credible, repeated warnings support belief | Trusted intermediaries, local languages and clear instructions |
| Inclusion | Income shapes reach and action | Accessibility and subgroup evidence remain weak |
| Outcome focus | Measure ability to act, not reach alone | Link warnings to practical response capacity |
tools.gsec.org.uk/wrp-explorer
Slides
Controls
Move the cursor over any chart to isolate a bar or line while you talk about it. Swipe left and right on touch screens. The address bar keeps the slide number, so a link to #23 opens at slide 23.