Building damage assessment
Background
This methodology was developed and implemented as part of the crisis response following the passage of Cyclone Chido through Mayotte on December 14, 2024. It does not repeat the methodology used for cyclone Haiyan, but is based on the BAR methodology, which it partially reproduces in terms of the 4 categories of visible wind disaster damage based on images. However, it does not include an assessment of the type of structure, as this is difficult to determine unambiguously from vertical imagery, as it can usually only be determined from the roofing material. This feature is provided instead. This methodology is suitable for wind disaster damage, but not for damage caused by other types of disaster (e.g. earthquakes), for which a vertical view cannot determine the extent of damage to buildings.
The following have been created and put online on the JOSM server:
- a model with presets, specific to Cyclone Chido, but which can easily be used for other disasters.
- a cartographic style (also known as a coloring model), presented in this video.
A test project was created in the zone of Mamoudzou, the most populated city in Mayotte. It was used to gather feedback from OSM cartographers and to improve the default models and instructions. These detailed instructions are available in English, French, Spanish and Portuguese and are accompanied by several explanatory videos in English or French only (see links below). This test project also made it possible to compare the evaluations made by the OSM community with those made by the Copernicus project (see section below).
Imagery
Damage mapping using imagery can only be done if post-disaster imagery is shared. In the case of the mapping done after Cyclone Chido hit Mayotte, CNES / Airbus Pleiades satellite images were provided as part of the Copernicus program and made available by IGN through this site. Use of these images has been granted to OSM contributors.
Mapping
This involves applying labels to buildings mapped using pre-disaster imagery. Two attributes are added to each building:
- on pre-disaster imagery, the type of roof material using the key roof:material=*. The presets template features a drop-down list with the roof material types presented in the reference wiki page. The coloring template adds a specific figure for concrete and another for metal (small dashed lines), while all other materials have the same figure (large dashed lines).
- on post-disaster imagery, the level of damage, according to 4 different categories: none (green), minimal (yellow), significant (orange) and complete (red). The presets detail each of these categories, which come from existing methodologies already tested on this type of context, such as the Harvard Institute's BAR methodology. The label is damage:DisasterName=* (damage:Chido=* in the case of cyclone Chido) so that other disasters can be mapped in the future without disrupting existing assessments. It is possible to use damage=*, but remember to modify the key at the end of the assessment or before any new damage mapping.
This document provides examples from Mayotte for roof material types and damage levels.
Two mapping methods for adding these labels were tested, depending on the habits of each OSM mapper:
- an approach that essentially uses the mouse, as shown in this video
- an alternative that uses the Task List plugin and only the keyboard, as shown in this other video. Once you've got the hang of it, you'll normally be able to map faster.
Additional metadata tags are also available, which can be added to each object:
- damage:DisasterName:event=* where the value is the GLIDE number of the disaster, to be found on this reference site. For Chido, for example, this number is TC-2024-000225-COM
- damage:DisasterName:assessment=* where the value is the date of the imagery used, in yyyy-mm-dd format
- source:damage:DisasterName=* where the value is that of the organization that created the imagery.
It is advisable to add these metadata tags at the end of the assessment session, before sending the data, as adding them manually for each building would be tedious. It's easy to select all the objects that don't yet carry these tags with a search like
"damage:DisasterName" (-"source:damage:DisasterName" OR -"damage:DisasterName:assessment")
Once selected, the metadata tags for these buildings can be filled in all at once using the preset template.
Tips
Avoid creating jobs that are too large - e.g. 150 m square is a good option that can be completed in about 15 minutes.
It is preferable to complete and correct pre-disaster buildings in advance (non-rectangular shapes when they are clearly rectangular, heterogeneous offsets, etc.) so as not to discourage contributors working on damage mapping.
Correcting buildings is still necessary when they are obsolete, need to be subdivided, or have such poor geometry that they need to be improved for damage information to be meaningful. This video shows an example of such a correction.
Cross-referencing with Copernicus EMSR data
Mayotte was the subject of a post-Chido assessment as part of the Copernicus Emergency Management Service - Rapid Mapping 780 (EMSR780), on the northeastern part of Mayotte. The resulting data, in the form of points for buildings, is open data and intersects the vast majority of OSM buildings. This cross-referencing was carried out via PostGIS and enabled the recovery of Copernicus information from OSM buildings. This MapStore context includes this intersection layer, which can be viewed in the “ESMR info in OSM buildings” map view and downloaded from the layer list.