This wiki page is intended at presenting the rationale for the incorporation of natural hazard data in OSM or in a fork project of OSM. We present here how natural hazard data can be collected and mashed up with "traditional" OSM data to create a synthetic view of the exposure of populations and infrastructure towards natural disasters.
Natural hazards pose a threat to the life and economic activities of many populations all over the world, especially in developing countries, where the vulnerability is higher and the Early Warning Systems less sophisticated (if existing at all). Since OSM is more and more frequently used for humanitarian activities, it becomes relevant take advantage of the data already contained in OSM to evaluate the risk of populations towards natural disasters. Many use cases can be thought of:
- After an earthquake, the search and rescue teams want to know where one can expect to have avalanches or landslides triggered by the seismic shaking.
- When starting a new humanitarian / development project, one wants to know if the chosen location (for a building, a water source, a refugee camp...) is at high risk of natural disaster
- Anyone wanting to start a business or to buy a piece of land should be allowed to do so knowing the risk of natural disaster at this location.
理論的背景 (Theoretical background)
- 災害 (Hazard): 災害とは、自然のあるいは人的に引き起こされた出来事で、破壊を引き起こす可能性をもたらすものです。
- 災害区域 (Hazard zone): 災害が発生 (過去において既に発生しているか、もしくは地形学的な証拠に拠った場合にも) したことが当然のように論議される区域のこと。
- 脆弱性 (Vulnerability): 住民の脆弱性は、災害に襲われた際に損失や罹災の可能性に関連しています。これは建物の品質や、教育水準、社会経済情勢などによって定義されます。
- 曝露 (Exposure): 曝露とは直接的に災害の脅威にさらされる施設と生命に相当する。GIS (地理情報システム) の用語としては、曝露は施設と災害区域の交点です。
- リスク (Risk): リスクとは、曝露と脆弱性の結合です。
- 非常に脆弱なコミュニティ (脆弱な建物、経済的に厳しく、教育水準が低いなど) は、災害区域の中にさえ位置しなければ、リスクには直面しないだろう。
このプロジェクトでは、我々はシンプルに災害区域を既にOSMの中に存在している村や道路や建物などに重ね合わせることを提案する。この意味において、リスク分析ではありませんが、曝露分析が可能となります。後者は追加の脆弱性に関する情報を必要としており、OSMと互換性のあるマナー (一方は膨大なリソースだけでなくフレキシブルなプロセスまでも求めます) では容易に集めることはできません。
As some terms are often mixed up, it is important to draw a quick reminder:
- Hazard: a hazard is a natural or man-made event that carries a potential for destruction
- Hazard zone: a zone where it can be reasonably argued that a hazard could take place (either because it has already taken place in the past or because of geo-morphological evidences)
- Vulnerability: the vulnerability of the population is related to the likelihood of loss and suffering if a hazard was to strike. It is defined by the quality of the buildings, the level of education, the socio-economic situation...
- Exposure: the exposure corresponds to the facilities and lives that are directly threatened by a hazard. In GIS terms, the exposure is the intersection between the facilities and the hazard zones
- Risk: the risk is the conjunction of an exposure and a vulnerability
- A 9.5 earthquake striking in the middle of a non-inhabited desert creates a null risk
- A seasonal landslide of a given intensity will create a higher risk for poor communities living in weak buildings that for communities living in well-engineered buildings
- A very vulnerable community (weak buildings, no economic mean, low education...) will face a null risk as long as it is not located within a hazard zone
In this project we simply propose to overlay the hazard zones on what is already existing in OSM: villages, roads, buildings... In this sense, an exposure analysis is made possible, but not a risk analysis. The latter would necessitate additional vulnerability information, that can not easily be collected in an OSM-compatible manner (one would require a much less flexible process as well as huge resources).
As of today (22.04.2012), tags exist for flood-prone areas only. Why not extend it to all natural (and man-made?) hazards?
Types of hazards
Before anything, it is important to define what hazards are relevant and whether / how they can be identified:
|Seismic hazard||The seismic hazard is usually defined as a seismic intensity value (often expressed in PGA - peak ground acceleration - in m/s2) that is likely to be reached in the coming 50 years, with a probability of 95%.
In other words, a value of 2.5 at a certain location means that, at this particular location, there is more than 95% chance to have an earthquake in the next 50 years that will generate shaking of at least 2.5 m/s2.
|N/A||Not easily collectable by the crowd. Necessity to use existing dataset (GSHAP). If no further data is collected, would it be more relevant to create a map template simply overlaying the OSM features on the seismic map?|
|Landslides||The tag hazard_type=landslide comprises all the hazards that involve massive displacement of earth, mud, stones... Although those hazards are quite different from each other, it is not easy for non-specialists to make the distinction. As this distinction is not essential for general humanitarian purposes, we thus propose to simplify the process the ontology.||Landslide, rockfall, debris flow||Past events (often well known in rural communities), hazard zones identified by relevant NGOs and national agencies|
|Snow avalanches||Only snow avalanches that are relatively predictable and that directly affect populations are taken into account. We thus consider the big ones (winter powder avalanches and spring wet avalanches), but not the slab avalanches, well known by skiers and mountaineers, but usually not affecting villages.||N/A||Past events (often well known in rural communities), hazard zones identified by relevant NGOs and national agencies|
|Floods||A flood is a significant overflow of the water level that submerges a piece of land.||N/A||Past events (often well known in rural communities), hazard zones identified by relevant NGOs and national agencies|
Any other hazard?
Qualification of hazards
It is not enough to define a hazard zone only with a polygon. Some attributes are required to store (at least) the intensity (is it a powerful landslide, or a relatively weak one?) and the return period (does it come back every year or once per generation?). The challenge here is to keep the model simple enough for non-specialists to be able to input data, but complex enough to keep some significance. A very simplified and intuitive approach is proposed here:
- Low intensity: a low intensity hazard is powerful enough to damage (but not to destroy) a house that would hypothetically be located on its path
- High intensity: a high intensity hazard is powerful enough to destroy a house that would hypothetically be located on its path
NB: by "house" we mean the type of habitat construction that is most commonly found in the region where the hazard is considered. The intensity is thus a value that is relative to the local environment.
Hazard return period
The return period is given as the approximate return time period, in years. A seasonal flood coming every year will thus be given a return period of 1. A landslide that has been observed once in a generation will be given a return period of 30.
As the return period are usually not known precisely (and as they anyway fluctuate), a better approach is to categorize it in 3 very easily understandable categories:
- < 3 years: any hazard that occurs almost every year (seasonal flood, spring avalanche...). Those are very well-known from communities, as they impact there life and activities every year)
- 3 - 50 years: any hazard that occurs more than once in a lifetime
- > 50 years: any hazard that happens once (or less in a lifetime)
The significance of a hazard is given as a function of the intensity and the return period. A tentative example is shown below:
|Return period < 3 years||Return period 3-50 years||Return period > 50 years|
|Low intensity||High significance||Medium significance||Low significance|
|High intensity||High significance||High significance||Medium significance|
NB: the significance is a concept that will be used for the rendering. E.g.:
- High significance => Red
- Medium significance => Orange
- Low significance => Yellow
Some flood zones have already been mapped in OSM. If this multi-hazard approach gains momentum and proves to be useful and universal enough, the existing flood-related information might be translated into the new tag-system.
The following tags are aimed at carrying enough information so that relevant analysis can be carried out, while keeping the model simple enough.
- Should the hazard data be stored along with the traditional OSM data? If not, what should be the most appropriate structure to ensure compatibility and ease of use?
- What is the risk of vandalism and how to fight against it? As real estate prices highly depend on the safety of the location (who would like to buy a flat in what is known as landslide-prone area?), wouldn't some people be tempted to create nonexistent hazard zones or to suppress actual zones?