Remote Sensing Labs
Browse all docs

Disaster Response

Disaster response is the domain where the value of an image is measured in hours, and where the analyst is rarely the person who acts on the result. A flood or a wildfire produces a decision backlog — evacuations, road closures, crew tasking, aid requests, insurance reserves — long before it produces a clean map, and the people holding those decisions work from partial information under time pressure. The mechanics live elsewhere on this site — Flood Mapping is the workflow for where water is sitting that should not be, and Burn Severity Mapping is the workflow for how hard a fire changed the ground. This page is about who needs those maps, on what clock, and what makes the answer usable when it arrives. It uses floods and wildfires as its examples; earthquakes, landslides, and storm damage follow the same response logic but rely on methods this site does not yet cover, such as damage grading from very-high-resolution imagery and radar coherence-change damage proxy maps.

The decisions these maps are actually feeding

The same flood or burn map serves several audiences with conflicting needs. Incident command and emergency management want a common operating picture now — where the hazard is, what it threatens, and where crews and evacuation orders should go in the next operational period. Humanitarian and civil-protection agencies want population exposure: how many people, which settlements, which routes, so that shelter, water, and access can be planned. In a major disaster, national authorities can also request satellite imagery and rapid maps in the first place, through mechanisms such as the International Charter “Space and Major Disasters”, which authorized users activate, or the Copernicus Emergency Management Service’s rapid mapping. Insurers and recovery funders want a defensible footprint and a damage estimate to set reserves, triage claims, and release payouts. Infrastructure and utility operators want to know which specific assets — substations, pumping stations, road segments, rail — fall inside the affected area so they can dispatch inspection and restoration; the exposure overlay in Infrastructure Monitoring produces exactly that list. A map built for one of these audiences, on its clock, is worth more than a faithful extent that arrives after the decision it was meant to inform.

The response clock, and why latency beats accuracy early

The value of a disaster map decays with time, and steeply. In the first hours of a flood or the active phase of a fire, a rough extent delivered while crews are still being positioned is worth more than a meticulous one delivered after the operational period it belonged to has closed — the same trade the agriculture domain frames around phenology, except here the window is measured in hours rather than weeks. This inverts the usual instinct to refine. A response product runs on a clock set by the event and by the incident-action-planning cycle, not by the analyst’s sense of completeness. It also means the sensor that can see the event at all often beats the one that would see it best on a clear day, which is why radar carries so much of a flood’s acute phase, for the reasons Flood Mapping sets out. For an active wildfire, the acute-phase input is not a severity map at all but active-fire detections — thermal hotspots from MODIS and VIIRS, distributed through NASA’s FIRMS globally within about three hours of overpass, and from geostationary satellites at far shorter intervals — together with fire-perimeter mapping. Latency here is not just processing time; it is revisit, tasking, downlink, and delivery combined, and it is the number most worth optimizing.

Activation-time mapping versus post-event assessment

Two jobs hide under the word “response”, and conflating them produces bad products. Activation-time mapping is the acute phase: fast, coarse, good-enough extent to direct crews, evacuations, and access decisions while the event is still unfolding. Post-event damage assessment is the recovery phase: slower, more careful work that quantifies what was lost, feeds insurance and reconstruction funding, and establishes a baseline for monitoring recovery. Burn severity belongs entirely to the second job. Even an initial severity assessment is made around containment — typically within days of it, and emergency stabilization teams often start before the fire is fully contained — and the extended assessment waits for the next growing season. The two jobs want different things from the imagery: the first prizes speed and coverage and tolerates a rough edge; the second prizes a defensible, well-validated footprint because money and legal exposure hang on it. A product designed for one is usually wrong for the other, and stating which phase you are serving is the first honest thing a disaster map can do.

What makes a product usable in a response

Four things separate a map a response can act on from one that merely describes the event. The first is the reporting unit: responders act on communities, road segments, parcels, service areas, and administrative wards, so an extent that is not already summarized into exposure by those units pushes the hardest work onto the person with the least time. The second is a plain confidence statement — built the way the flood workflow’s outputs section describes — because a responder who cannot tell a firm result from a guess will either over-trust it into a bad evacuation call or discard it entirely. The third is delivery format and channel: a result that cannot reach the incident’s mapping system, the field tablet, or the situation report in a form it can ingest is, operationally, not delivered. The fourth, and the one most specific to this domain, is the asymmetry in the cost of being wrong. A missed inundated neighborhood or a missed high-severity slope can cost lives or leave people unwarned, while a false positive wastes a crew and erodes trust in the next map — so responders often err toward flagging possible impact in the acute phase and tighten the footprint for the recovery assessment.

Where the domain stays hard

Validation is scarce exactly when it matters most: ground truth during an active disaster is fragmentary, dangerous to collect, and often arrives after the decisions have already been made, so acute-phase maps are trusted with far less field confirmation than any analyst would like. Cloud, smoke, and the timing of an overpass can mean the imagery simply does not exist for the hours that matter, and no amount of processing conjures an observation that was never captured — the honest response to a gap is to say so, not to fill it. Urban and vegetated settings complicate the signal: water under a canopy or among buildings needs different radar reasoning than open water, and damage beneath an intact roof is hard to see from above, so a clean extent can quietly understate the human impact. And the last mile is organizational, not technical: a correct map that does not reach the right desk, in the right format, inside the operational period, changes nothing.

Where to go next

The workflow mechanics live in Flood Mapping, which walks the extent-and-change question from framing to a reproducible result, and Burn Severity Mapping, which does the same for fire effect. Both lean on the multi-temporal reasoning the Change Detection concept develops, on the Cloud Masking discipline that keeps a shadow from reading as impact, and — in the cloud-covered acute phase — on the all-weather sensing the SAR Basics concept explains.

Sources