Why considering ecosystem data for disaster risks?
Historically, disaster risk reduction measures focused massively on grey infrastructure, such as dykes, sea walls or gabions. The 2004 Indian Ocean tsunami raised awareness about the role of mangroves in protecting entire communities and their livelihoods from the waves. At UN Environment and UNEP/GRID-Geneva (Global Resources Information Database), we believe in innovative solutions to disaster risks, and we have been promoting the role of ecosystems in reducing hazard, exposure and vulnerability since 2008.
Moreover, we strongly believe that understanding risks means providing tools for decision-makers to build cost-effective solutions to those threats.
Mangroves, sea grasses and corals have been proven to reduce wave energy and related impacts from storm surges; forests and other vegetation can reduce landslide susceptibility by removing excessive soil water content and stabilizing the soil through root networks; forests can also buffer drought by regulating humidity and increasing precipitations, through albedo, roughness, shadow, heat absorption and evapotranspiration processes.
Whilst promoting the role of ecosystems through projects and advocacy, we realized that visual assets such as maps can play a crucial role in convincing decision-makers to invest in such approaches through policies, projects and programmes. The availability of global datasets provides an opportunity to visually compare the restoration and conservation potential of various ecosystems to population exposure to hazards. Ultimately, we show suitable areas where ecosystem management can be used as opportunities to protect the greatest number of people globally.
UN Environment and UNEP/GRID-Geneva collaborate on this since several years. But it is only recently that this materialized in the production of some of the datasets that we show here. UNEP/GRID-Geneva prepared the data and processed it, by overlaying hazard population exposure and ecosystem coverage at the global scale, for several combinations. All the input data are open-access and documentation is soon to be shared on how to use the output layers.
Forest management to reduce floods in Africa: presenting data used and approach
For the VizRisk challenge, we decided to show the potential of forest management in reducing flood risk for the African continent. Hence two main datasets were used as input layers: forest coverage data from the World Resources Institute (itself resulting from a merge of data on closed, opened forest and woodlands) and flood population exposure.
Why this data?
Because flood is one of the most common and pervasive hazards in the African continent, exposing over 13 million people yearly. African forest ecosystems (including the savannah woodlands) are one of the world’s set of lungs: the tropical rainforest from West Africa and the Congo basin stores 171 gigatons of carbon. But deforestation rates are high – about 2-3 % per year – and the loss of these ecosystems could prove dramatical for carbon storage and ecosystem services. Forest management is essential to ultimately reduce flood risk, especially when climate change is increasing the variability of rainfall patterns, leading to more intense and frequent flash flood phenomena.
The flood exposure dataset includes an estimate of the annual physical exposition to river floods based on a statistical model for 25 to 1,000 years return period events. A population grid for the year 2015, provided by the Joint Research Centre Data Catalogue (European Commission, Joint Research Centre (JRC)) has also been used. Unit is expected average annual population (2015 as the year of reference) exposed (inhabitants). This product was designed by International Centre for Geohazards /NGI for the Global Assessment Report on Risk Reduction (GAR). It was modeled using global data.
|Layer Theme||Original Dataset Title||Provider||Reference period||Spatial resolution||Link|
|River Floods||Flood hazard # years||GAR 2015, PREVIEW / Global Risk Data Platform||25 to 1000 years RP model||~ 0.01 dd||http://preview.grid.unep.ch|
|Population density||GHS Population grid||Joint Research Centre Data Catalogue||2015||1 Km||https://data.jrc.ec.europa.eu/dataset|
|Forest||WRI Current forest coverage||World Resources Institute (WRI) Atlas of Forest Landscape Restoration Opportunities||2000 – 2009||0.008333 dd||http://www.wri.org/applications/maps/flr-atlas/|
Table 1. Summary table of input data layers used
These datasets are open-access and available at the links cited above. They were resampled at a 10×10 km resolution grid, for computational purposes.
Each cell thus stores average values at the indicated spatial resolution. The overlaying of the two heavy datasets was done with GRASS GIS software based on Python scripts, but similar processes could be done through other tools, such as QGIS and have been tested with higher resolution datasets for national scale mapping. The output layer is a geopackage, with an attribute table, storing values for each cell, such as the proportion of closed, opened or woodland forests, as well as flood frequency and population exposed to floods.
In addition, it includes a categorization in 2 main groups: areas where forest restoration would be most beneficial for reducing flood exposure, and areas were protection would be better indicated. This differentiation was done by using quantiles for the forest proportion of each cells: the lower 50% of values were attributed to ‘restoration’, the rest to ‘protection’. The median is situated at approximately 65% of forest coverage for the 10×10 km cell for the African continent.
|**Summary of statistics for African continent|
|*Dataset #1 – flood population exposure|
|# people annually exposed to floods||13,224,644|
|*Dataset #2 – estimated forest coverage|
Table 2. Summary of statistics for the African continent. This information was retrieved through QGIS, by selecting relevant features through ‘Select Features by Expression’ and then requesting through the Processing Toolbox ‘Basic statistics for fields’
Dot representation for small scale mapping
Each cell was originally a square of approximately 10×10 km. Several limitations came to mind for the designing of the map: since each cell stores average values, area-based choropleth map could lead to a misleading impression that the whole area is equally affected by flood risk and suitable for measures such as forest protection or restoration. Because of its coarse spatial resolution, the data does not integrate well the presence of dense urban areas which are unsuitable for large scale forest restoration.
In other words, the data was designed for raising awareness and communicating about the enormous potential of forest management practices to reduce flood; not for project design. It is thus preferable to visualize it at the global, regional and country scale.
Thus, instead of visualizing the data with 10×10 km areas, centroids of polygons were created, for better visualization at the small scale – for the whole African continent. The scatter effect of dots enables the user to quickly grasp the areas suitable for forest management to reduce floods at the continental scale. At a country scale, the dot representation – although not indicating with accuracy where forest protection and restoration should happen – was deemed less misleading than having a whole area represented: it gives an indication that forest management practices could be beneficial in the surrounding area, but not exactly where it should happen.
Uploading the data on MapX platform
UN Environment and UNEP/GRID-Geneva have been building MapX, an online, open source cloud solution for mapping and monitoring the sustainable use of natural resources. The MapX platform is optimized for lowInternet bandwidth, to ensure access and use from any environment. Emphasis is put on data security, integrity and access and counts over 1300 data layers and engagement with 400 stakeholders. MapX uses Mapbox GL JS to render interactive maps from vector tiles and Mapbox styles.
In MapX, we created a new specific space (called ‘project’ in the MapX terminology) for storing and visualizing data for the VizRisk challenge: “Viz Risk Challenge 2020”.
The project was made public (which means that anyone can visualize the content within it) and can be freely accessed at: https://bit.ly/30tLsML
Each user can explore the data by zooming in, toggling layers, clicking on features to see attribute information.
The VizRisk project includes the input layers for the African continent, with some summary information on each dataset such as the data sources. It also includes an animation, with legend, data sources and some contextual information that explains the data.
The data on forest management for flood reduction was uploaded as GeoJSON, as this is one of the standards for open source data, and it is supported by MapX at upload and download. For the project, a basemap style from Mapbox was customized, although allowed changes in parameters through MapX are more limited than through the Mapbox platform.
For creating a project and adding layers, the general workflow is the following:
- Login into MapX website
- Creating through the toolbox a new project, specifying users’ restrictions (in our case, ‘public’)
- Adding data sources through the toolbox of the project
- Creating new views based on data sources and ‘styling’ of these views for each dataset to be displayed
- Adjusting zoom-scale styling, adding summary information about the layer (metadata), configuring users’ access (in our case, ‘public’)
For reproducing this workflow, we refer to the knowledge page of MapX: https://www.mapx.org/knowledge_base/
In this knowledge base some technical guides are available on how to get started with MapX, to manage a project workspace, uploading spatial data, building story maps and other functions.
Results from the data analysis
The analysis highlights that in Africa, of over 13.2 million people yearly exposed to floods, 8.4 million of them could directly benefit from forest management practice to reduce their exposure to flood hazard.
Forest conservation measures, for instance, would be most beneficial in countries like DRC, Gabon, the coastal area of Nigeria. In areas where forest coverage is lower, restoration measures could be more suitable.
Louise Schreyers is a GIS Analyst at UN Environment, Disaster Risk Reduction Unit.