Remoteness as a proxy for social vulnerability in Malawian Traditional Authorities An open data and open-source approach

Abstract

Within the humanitarian aid sector there is a need to assess the social vulnerability to natural hazards within developing countries on a granular geo-spatial level. This is commonly done through the use of qualitative and paper-based field surveying methods, which are costly and time consuming. These field assessments are often project-based, linked to a certain area and consequently results in data filled with gaps. The current and most common methodology used within the humanitarian community, including the International Federation of the Red Cross and Red Crescent Societies (IFRC), is a so-called Vulnerability and Capacity Assessment (VCA). Through this methodology, insights into the vulnerability and coping capacity of a community with regard to natural hazards and climate change are gathered. In this way, humanitarian organisations are better informed on the susceptibility to natural hazards pre-disaster and the aid neediness of communities’ post-disaster. More often than not, a VCA is not conducted or the outcomes on an affected community are not readily accessible by information managers, causing uninformed aid provisioning pre-disaster and in the immediate disaster response phase. The fact that there is no information readily available for these areas, creates a need for alternative evaluation methods based on secondary and open data sources. In this research, through the use of open data; OpenStreetMap (OSM), The Humanitarian Data Portal (HDX), Malawi Spatial Data Platform (MASDAP), and CEISIN/Facebook data, several remoteness indicators are created using Postgis, PgRouting, Osmosis, and other open-source geospatial analysis tools to generate a proxy for social vulnerability. The created remoteness indicators are field tested in Malawi using the digital surveying tools OpenDataKit (ODK) and OpenMapKit(OMK) to validate the initial outcomes of the geo-spatial analysis with the aim of calibrating the geospatial parameters to real-life values. The proxy is then created based on the remoteness indicators, which are trained through machine learning (ML) techniques to an existing social vulnerability index (SoVI) for Malawi. The results are visualized through an interactive map using a web application where the proxy result and the existing vulnerability index can be compared. To make this process generally applicable in other disaster-prone countries, the development of the proxy is done with the use of open data and open-source software and the results are interactively visualised through an online dashboard to make the process more transparent and replicable. The developed proxy predicts the SoVI scores for Malawian Traditional authorities with an accuracy of 64% and can be run on this granular level within hours, compared to months or even years using traditional vulnerability assessments. This potentially creates a fast and accurate assessment alternative for decision-makers who decide on the project areas of humanitarian organisations and it may provide a fast and evidence-based insight into vulnerability in the immediate response phase after a natural hazard.