Understand risk. Review evidence. Take informed action.

Getting started

1. Select a community from the dashboard selector.
2. Review its current risk score and warning status.
3. Inspect the evidence behind the assessment.
4. Review expected impact and prioritized actions.
5. Then inspect observations, sources, and history.

The dashboard follows a decision-first hierarchy: Warning → Evidence → Actions → Observations.

Risk and warnings

Scores and warning states are calculated by deterministic rules. AI may explain or recommend actions, but it cannot change the authoritative score.

Evidence and recommended actions

Triggered evidence lists every deterministic rule that contributed to the score. Expected impact describes exposure, disruption, and urgency. Priority actions state what should be reviewed or coordinated next.

SentinelAI supports human decisions. It does not replace emergency authorities, official forecasts, or local response protocols.

Live dashboard updates

SentinelAI checks for persisted operational changes every 30 seconds while the page is visible. A changed risk, warning, observation, or recommendation updates in place without reloading the complete page.

Live sync active

The browser can retrieve the latest SentinelAI state.

Live sync delayed

The last valid view remains visible while connectivity recovers.

Source freshness

Source freshness is different from live sync. It reports when an IoT station or online provider last delivered evidence.

Fresh

Evidence arrived within the expected provider interval.

Delayed

The expected delivery interval has been exceeded.

Stale

No evidence has arrived for twice the expected interval.

Reference

A replay dataset is historical and does not become stale.

History, trends, and export

• Select a variable and source before comparing readings.
• Choose the latest 30 or 100 readings, a preset period, or a maximum 31-day custom range.
• The table shows community time, UTC, value, change, source, and quality.
• Use Export CSV for a bounded copy of the selected evidence.

Trend lines, deltas, and anomaly markers describe persisted history. They are not forecasts and cannot change risk.

Know the data mode

Live

Current hardware or online provider observations.

Estimated

Historical reanalysis values with retained provenance.

Replay

A verified historical sequence presented frame by frame.

Simulated

Reproducible demonstration evidence, never a live emergency.

Forecast evidence

Forecast evidence contains provider model values for future valid times. SentinelAI preserves provider, model selection, retrieval time, horizon, and location provenance separately from measured observations.

Select Open Early Warning Outlook to inspect the deterministic 1-, 6-, 24-, and 72-hour horizons. Each horizon shows evidence, confidence, uncertainty, and an Event → Exposure → Possible Impact → Action translation. It cannot change the authoritative current assessment.

The Anticipatory Intelligence Brief explains the persisted outlook, presents conditional scenarios, and organizes preparedness actions by horizon and audience. Every safety action links to an approved IFRC or WMO preparedness-guidance reference. These organizations do not supply or endorse SentinelAI's monitoring, forecast, assessment, or Outlook. If AI is unavailable, the same section displays deterministic fallback guidance.

Using the Garsen replay

• Restart returns the deterministic sequence to its beginning.
• Next persists the next historical frame and updates risk.
• Play story animates only the contextual presentation.
• Story arrows never change evidence, alerts, or risk.