Rapid mapping

Why this matters

The command post must decide within the hour which villages are surrounded by water, which roads are still usable and where to send boats. Waiting for a complete 3D model takes too long. Rapid mapping answers urgent questions with sufficient data, but must state clearly how far it covers, when the images were taken and what is not yet confirmed. Otherwise gaps on the map may be read as safe areas.

Rapid mapping versus full mapping

A rapid map may be a quick mosaic without GCPs, or single images with numbered points: enough to show “roughly what is where”, but not for precise measurement. Work that needs areas or volumes must follow the UAT 361 workflow. Images without complete coordinate data should not be placed on a map by guessing their position.

At national and international level there are satellite mapping mechanisms. The Copernicus Emergency Management Service (Rapid Mapping) provides free maps of disaster extent and damage within hours to days to authorised users, and the International Charter Space and Major Disasters provides free satellite data when a registered national disaster authority requests it. Satellites cover wide areas, while drones are more detailed and can fly under cloud, so they complement each other.

Analysing flooded area

CELL_M = 25
flood = ["............", "...~~.......", "..~~~~......", ".~~~~~~.....",
         "..~~~~~~~...", "...~~~~~~~..", "....~~~~~~..", "......~~~..."]
houses = [(1, 2), (2, 6), (3, 3), (3, 8), (4, 5), (5, 1), (5, 9), (6, 6), (7, 10), (2, 10)]

wet = sum(row.count("~") for row in flood)
area_ha = wet * CELL_M ** 2 / 10_000
hit = [h for h in houses if flood[h[0]][h[1]] == "~"]
print(f"flooded cells {wet} of {len(flood) * len(flood[0])} = {area_ha:.2f} ha")
print(f"houses in flooded cells: {len(hit)} of {len(houses)} -> {hit}")

flooded cells 35 of 96 = 2.19 ha
houses in flooded cells: 4 of 10 -> [(3, 3), (4, 5), (5, 9), (6, 6)]

Coverage, age and completeness are different questions

The knowledge unit on drills separates three indicators. Coverage is the share of the area with usable data. Data age is the time from capture to use. Completeness is the share of required items that have usable data. A map may cover almost the whole area yet be several hours old in a fast-rising flood.

from datetime import datetime

aoi_cells, no_data_cells = 96, 6
captured = datetime(2026, 9, 20, 7, 40)
used = datetime(2026, 9, 20, 11, 10)
required_items = ["bridge B1", "school shelter", "road R12", "clinic", "pump station"]
valid_items = ["bridge B1", "school shelter", "clinic"]

coverage = (aoi_cells - no_data_cells) / aoi_cells
age_h = (used - captured).total_seconds() / 3600
completeness = len(valid_items) / len(required_items)
missing = [i for i in required_items if i not in valid_items]
print(f"coverage {coverage:.1%}, data age {age_h:.1f} h, completeness {completeness:.0%}")
print("still missing:", missing)

coverage 93.8%, data age 3.5 h, completeness 60%
still missing: ['road R12', 'pump station']

Training datasets and AI

Training AI to classify floods and damage uses labelled drone image datasets such as FloodNet (Rahnemoonfar and colleagues, 2021), post-flood imagery, and RescueNet (2023), 4,494 images after Hurricane Michael with 10 damage classes. Read the data licence conditions first, and AI results must be checked by people, as in the WFP example where AI counted damaged houses and specialists reviewed the results.

Module lab

Common mistakes

Summary

• Rapid maps answer urgent questions with sufficient data, but must state extent, time and status
• Satellites (Copernicus EMS, International Charter) and drones complement each other
• Areas and buildings from classification still need confirmation
• Coverage, data age and completeness are different questions and must be reported separately

Check your understanding

1. 40 flooded cells of 20 × 20 m each are how many hectares?
2. An area of interest of 200 cells has 10 with no data. What is the coverage?
3. Images captured at 06:30 are used at 09:00. How many hours old is the data?
4. Why are gaps on a rapid map dangerous?
5. What status should AI counts of damaged houses have before review?