Impact Assessment of Natural Hazards
General Module
This tutorial showcases how to combine hazard extent shapefiles with other datasets in QGIS in order to assess the impact of a natural hazard.
Natural hazard maps obtained from already-available hazard mapping services or produced from optical or radar satellite data in most cases are insufficient on their own to provide useful information that may be used to direct humanitarian assistance and mitigation actions. Natural hazard extent maps must be combined with other geographical information and further analyzed to generate localized statistics and quantities that may be applied in resource mobilization. The following publicly available datasets are combined with an example hazard map (a flood map of Nigeria obtained from the GloFAS Global Flood Monitoring (GFM) service) in this guide:
- Administrative boundaries - Humanitarian Data exchange service (HDX)
- Road network - Open Street Map
- Human settlement footprint - DLR Geoportal
Pre-processing
Disaster maps in most instances are in raster format since they are derived from satellite imagery. To load a raster layer in QGIS, go to Layer -> Add layer -> Add raster layer and navigate to the raster file’s directory. Depending on the information contained in the raster the raster file will open as a black and white image (Figure 1) where white pixels represent affected areas while the black pixels represent unaffected regions.
It is important to transform the hazard map from raster to vector format for additional analysis and interaction with other datasets. It is feasible to vectorize pixels that only have the value of 1 in situations where the hazard map is binary, meaning that pixels with a value of 1 represent affected areas and pixels with a value of 0 represent non-affected areas. Pixels with a specific value can be vectorized in QGIS using the “Vectorizing Raster Classes “ tool that is implemented in SAGA toolbox in QGIS. The parameters required that need to be specified for this operation include Grid which is the hazard raster map, the class selection which is the band containing information to be vectorized, “Class Identifier” where the user has to specify the pixel value to be polygonised and finally the vectorizing mode and in this case the “one single multi-polygon object” is chosen (Figure 2).
Alternatively, the whole image can be vectorized based on the pixel values. To do this go to Raster -> Conversion -> Polygonize (Raster to Vector) (Figure 3).
The resulting vector will cover the whole extent of the image but will have two kinds of attributes based on the initial pixel values. At this point, the non-affected polygons can be removed by opening the attribute table and selecting the attribute value that represents non-affected regions. The procedure as illustrated in Figure 4 is as follows; in the attribute table first click toggle editing button to enable editing of the layer, type 0 (value representing non-affected areas) in the expression bar that appears at the top of the table, then click update all. When the zeros are highlighted, click on the delete button.
The resulting polygons will have geometry errors such as overlaps and slivers. To correct geometry inconsistencies in a vector layer go to Vector->Fix Geometries and select the vectorized layer(Figure 5). This will detect all geometry inconsistencies and correct them.
The geometry-corrected layer can then be exported as a shape file and saved by right-clicking on the layer -> export -> ”Save feature as” and assigning an appropriate coordinate reference system (CRS) (Figure 6).
Intersection with Administration boundaries
The amount of land area affected by hazards within a given administrative region can be derived from the intersection between hazard maps and administrative boundaries. Administrative boundaries can be derived from various sources depending on the specific region, for example for Europe, boundary data can be sourced from https://ec.europa.eu/eurostat while for Africa, administrative boundaries can be sourced from https://gadm.org/.
To quantify the amount of land area affected by the hazard in every admin region, the first step is to load the administrative regions and the vectorized hazard map, this can be done by selecting Layer -> Add Layer -> Add Vector layer and navigating to the specific directories containing the specific vector layers.
As shown in Figure 7 the affected administrative regions would first need to be extracted from the whole administration layer, this way the processing time for subsequent calculations is reduced since operations will be on a smaller dataset. To extract the admin regions that are affected by floods go to Vector -> Data Management Tools -> Join attributes by location. In the subsequent window in the “join to features in” tab, select the administrative boundaries layer, under “Geometric predicate” select the type of spatial relation to use as a filter, in this case, intersect, overlap or contain would be appropriate and lastly in the “By comparing to” tab select the hazard vector layer.
The results of the join-by-attributes operation (Figure 8) is a subset of the administrative polygons that contain the affected regions which is shown in Figure 9.
To determine the amount of flooded area that falls under each administrative region, a spatial union between the two layers is computed. To perform a union in QGIS go to Vector -> Geoprocessing -> Union and add the input layer as the flood polygons and the overlay layer is the admin boundaries (Figure 10).
The last step is computing the areas that overlap, this can be done by going to Vector -> Geometry Tools -> Add Geometry Attributes and selecting the union layer created in the previous step. The resulting will result in a new layer with an attribute column containing the overlapping areas.
Intersection with road network
The length and classification of roads affected by floods can be extracted by analyzing the intersection between a road network and the hazard polygons.
The road network can be sourced from Open Street Map or third-party distributors free of charge. Once the road network is loaded, it can be visualized as a categorized layer to see the different road classes as shown in Figure 11. The goal is to derive the length of affected roads for each class.
The first step is clipping the road network to the extents of the hazard polygons to remain with affected roads only. This can be done via Vector -> Geoprocessing -> Clip in the input layer the input is the whole road network while the overlay layer is the polygonised hazard areas (Figure 12). This process clips a vector layer using the features of an additional polygon layer. Only the parts of the features in the input layer that fall within the polygons of the overlay layer will be added to the resulting layer.
The attributes of the features are not modified, although properties such as area or length of the features will be modified by the clipping operation. If such properties are stored as attributes, those attributes will have to be manually updated.
In the final step, the cumulative length of each affected road class is measured. To do this a virtual layer is generated from an SQL query of the clipped road network. To create a virtual layer, in the main menu go to Layer -> Add Layer -> Create/edit virtual layer. Assign the virtual layer a name in the Layer name window and in the query-window input the SQL query as illustrated in Figure 13.
To view the results of the query, right click on the newly created virtual layer and open the attribute table. A three-column table will appear with the ID in the first column, road class in the second column and the cumulative length of each affected road class in the third column (Figure 14).
Intersection with Human Settlement data
An estimate of affected population can be derived from the number of settlements affected by a hazard event. Human settlement footprint provides a binary raster dataset derived from satellite imagery showing where humans live worldwide. The human settlement footprint is provided as a geoportal service that can be accessed via the link https://geoservice.dlr.de/web/maps/eoc:wsf2019 and navigating to the Downloads menu and selecting World Settlement footprint 2019 which is the latest dataset. This opens a tiled map of the world, to download the settlements of a specific area, click on the tiles covering the area, if the area of interest is covered by more than one tile, click on the multiple tiles and download them as a list on the left download panel as shown in Figure 15.
The settlement footprint is downloaded in geotiff format and can be loaded in QGIS by selecting from the main menu Layer -> Add Layer -> Add Raster Layer then selecting the specific folder where the raster data is stored (Figure 16).
To extract pixels that are flooded from the Human Settlement footprint raster, the “Clip Raster by mask layer” is used. To mask the raster using the flood polygons, go to Raster -> Extraction -> Clip Raster by mask layer. The input layer is the human settlements footprint raster and the mask layer is the hazard polygons (Figure 17).
The last operation is quantifying how many settlements are affected by the flood event. This can be done by generating a report of the count of pixel values in the masked HSF raster. This can be achieved using the “Raster layer unique values report” tool (Figure 18).
Since the HSF raster is a binary raster, there will be two values only; the number of pixels having the value of 1 and those having the value 0. Basing on an assumption of how many persons live in one settlement, an estimate of the affected population can therefore be inferred from the pixel count (Figure 19).