SolutionFire, flood and forest-loss intelligence for Thailand
An AI assistant for the people who protect forests and villages. It watches satellite detections, weather and maps, ranks what threatens people first, explains why in plain words, and drafts the response for a duty officer to approve before anything goes out.
Real NASA FIRMS satellite fire detections for Chiang Mai and Chiang Rai on 28 Mar 2026. Incidents, people and teams are synthetic. Not an official warning.
The context
The signals that matter sit in different places and update at different speeds. The AI opportunity is to join them up, show where risk is building and why, and help people act on it sooner.
Much of Thailand's flood risk builds during the rainy season, broadly from around May to October, when heavy rain and rising rivers can coincide.
Flash floods can arrive within hours in hilly areas, while river floods in low-lying plains build over days. Each needs a different lead time.
Weather, water, disaster-response and local agencies each hold part of the picture.
Public sources such as NASA satellite fire detections, Open-Meteo weather and air quality, and OpenStreetMap villages and protected areas can be combined on one map.
General background only. Specific agencies, data, thresholds and procedures will come from stakeholders during discovery.
What is the current issue?
The issues under each area are working hypotheses to test with stakeholders, not findings.
Rainfall, water levels, spatial layers and field reports in one place
AI opportunity. One ingest step that maps every source to a common schema and flags gaps and outliers before scoring.
See the smoke and weather screenA risk score per area from combined signals
AI opportunity. A transparent score per area from weighted layers, with a plain explanation of what drives it.
See a fire's score breakdownThreshold-based alerts and coordinated actions
AI opportunity. Rules that fire on the score, a drafted message a duty officer approves, and a checklist that follows the severity.
See fires near villages and the drafted actionsOne shared picture for decision makers
AI opportunity. A shared map, response board and drafted situation report, from the same data the alerts use.
See the briefingHow can AI help?
Each workflow starts from work response teams already do. Rules and models do the watching and drafting; a named person decides what happens next.
New readings are mapped, timed and checked before they reach a risk score, so people trust what they see.
Rain, river, spatial and population data arrive on their own schedules.
Fields, units and time zones are aligned, and each reading is placed in an area.
Out-of-range values, gaps, duplicates and dated tables are flagged with a reason.
Decides on each flagged record, with the reason in front of them.
A checked batch with its age and flags, logged for audit.
Tools the system calls
Illustrative function names for a pilot design.
Today
Each source is checked in its own tool, and odd readings are noticed late, if at all.
With AI
One checked dataset with every gap and outlier flagged before it can move a score.
The intelligence layer
The four areas share one data layer and one set of AI services. Data flows up; approvals and the audit log wrap every step.
Public sources, combined on one map
Rules and models score, the agent drafts
Each role sees its own view
What would the user see?
The opening screen is a real map, not an empty one: satellite fire detections for Chiang Mai and Chiang Rai on 28 March 2026, ranked by risk score. Click a fire to see why it scored what it did, or ask a suggested question.
Disaster AlertChiang Mai · Chiang RaiThailandIndiaDuty officerDecision support · not an official warning · fire data NASA FIRMS · incidents synthetic
See which fires threaten villages first and why, then approve, edit or hold each drafted advisory.
See the nearest fires with distance and wind direction, and report back from the field.
One shared map and a drafted situation report for each meeting.
Manage sources, quality flags, layers and scoring weights.
Key screens
Each tab opens the same app on a saved archive day: the briefing, a fire's score breakdown, fire danger, smoke and forest loss. The satellite detections are real; incidents and responses are illustrative.
Click a fire cluster on the map, or a card in the panel, to open it.
Disaster AlertChiang Mai · Chiang RaiThailandIndiaDuty officerDecision support · not an official warning · fire data NASA FIRMS · incidents synthetic
The morning briefing: one ranked list instead of a thousand hot spots. The scores come from a fixed rule over real detections, wind, villages, protected areas and slope.
What comes next?
This page is the concept and working demo. Pilot scope, data, budget and timeline are agreed only after stakeholders have seen it.
The scope areas this solution covers.
This page, on a real archive day of fire detections.
Adjust as more information arrives.
Validate the problem, data and needs.
Scope, data, integrations, criteria, budget, timeline.
Show the idea working, so stakeholders react to something concrete.
Output: a stakeholder-ready demo.
One agreed area, real data where access allows, run alongside today's process.
Output: a scoped pilot with agreed measures.
More areas and sources on the same platform, then prediction and a digital twin.
Output: a supported service.
Built for public safety
Anyone acting on a flood alert will ask where it came from and who signed it off. The design answers that from the start. The list on the right is what we need from stakeholders before a pilot.
Answers are pre-written from a saved archive day, not live. Risk scores are a fixed, transparent formula.
Fire detections are real NASA FIRMS data for 28 Mar 2026. Incidents, people and teams are synthetic. Nothing here is an official warning.
No public alert leaves the system until a duty officer has read and approved it.
Each area shows the layers, weights and readings behind its score, so it can be challenged.
Review the demo with your team.Tell us what a pilot must prove.
The fire detections in this demo are real satellite data for 28 March 2026; the incidents, scores and responses are illustrative, not a forecast or an official warning. This page is a demonstration, not an official system of any agency. For real warnings, follow official channels.