One of the challenges posed by climate change is the increased risk of catastrophic natural events. These are likely to disproportionately affect most vulnerable areas across the world, many located in resource constrained, low income countries. The international development effort may be significantly undermined by such events, because of the foregone social expenses that could have been utilised for expanding healthcare, education and infrastructure investments.
Signatories to the United Nations’ Sustainable Development Agenda committed in 2015 to achieving the 17 ambitious Sustainable Development Goals by 2030. Therefore, risk sensitive investment in developing countries is crucial for ensuring the long-term resilience of projects implemented in pursuit of this vision.
Experienced data science practitioners who offer services in the private sector might see this as a relatively mundane problem. At Appsilon we have observed the transformative power of data, which our clients tap into when they implement decision support systems in their businesses to mitigate risk. One might assume that a similar approach could be taken to enhance the decision making capacity of governments and international aid organizations to incorporate natural disaster risk factors into allocating scarce resources.
Visualise – a simple, but comprehensive decision support tool for policymakers in natural disaster risk managementHowever, it is a recurring theme in international development where solutions created in the developed world fail when implemented in developing countries as they lack appropriate understanding of the unique context and challenges the regions face.
Despite significant efforts to gather, analyze and disseminate data on natural disaster risk, it remains wanting both in quality and quantity in these countries. Furthermore, the technical capacity to interpret the data on the ground may be lacking. Even when such expertise is available, the political will and financial capacity may not be sufficient to take full advantage of the data and invest in risk reduction activities.
All of these challenges require a measured and holistic approach combined with significant on-the-ground experience and relationships. Therefore, if data science companies wish to contribute to tackling this problem, there exists a clear need for a deeper collaboration with development practitioners.
We were approached by Dr Junko Mochizuki, who is a member of the Risk and Resilience program’s team at the International Institute for Applied Systems Analysis, to provide support in her project aimed at improving disaster risk management policy in Madagascar.
Madagascar, an island country located in the Indian Ocean, is particularly exposed to natural disaster risk with frequent cyclones ravaging its weak infrastructure. For example, storms, torrential rain and strong winds unleashed by the Ava cyclone in 2018 caused USD$130 million damage and USD$156 million in losses, totaling 2.9% of its GDP. To put this in context, Madagascar averages a GDP growth of 2.5% per year.
Furthermore, the majority of the population dwells in rural areas and depends on agriculture for their livelihood. These individuals are disproportionately affected by such disasters, a detail not necessarily encapsulated in aggregate income figures. The government support is highly restricted as Madagascar is one of the poorest countries in the world with a GDP per capita at USD$460.
Dr Mochizuki and her team set out to understand how to support Madagascar in developing appropriate policies to mitigate this risk. Two questions were posed during the Training Program on Disaster Risk Assessment and Optimization of Public Investments in Reducing Economic Losses she participated in Madagascar in 2015:
“How can we strengthen contingency funding and the mainstreaming of disaster risk reduction at the same time?”
“What can a cash-strapped government do when donors themselves do not seem to allocate funding based on the tangible needs of a country’s natural disaster risks?”
The Visualise project led by Dr Mochizuki will answer these questions. Its key deliverable is a user-friendly analytical and decision-support tool, which should inform macroeconomic, budgetary and development planning and empower officials to better communicate the country’s needs to international aid organizations.
Similar previous interventions in Madagascar did not prove successful because of two major factors. First, there was no consistent deep-level data. Disaster risk assessment and census type information was being collected by various international organizations and these data were neither interoperable or exhaustive. Second, the tools that had been deployed were complicated and inaccessible to government officials and policymakers.
Visualise addresses both of these problems. It takes advantage of country-wide household surveys and combines them with a variety of already available and curated multidimensional poverty indicators. These granular data are then loaded into the back-end of a comprehensive visualization tool, which allows users to understand the impact of natural disasters at an unparalleled level of detail. Instead of lists of numbers, complex charts and statistics, the users are presented with an easy-to-use, self-explanatory resource.
The project fully embraces the complex contextual reality of the country and needs of the end-users. The tool will be deployed in close collaboration with Madagascar’s Ministry of Finance. The next step is to enhance its functionalities with macroeconomic modelling for simulating economy recovery depending on allocation of resources.
I believe that this is a good example of the exploratory approach we described at the launch of our AI for Good initiative. While we can think of many ways in which decision support systems we develop for our commercial clients could be utilized in a developing country context, partnering with experts from IIASA who have on-the-ground knowledge ensures the positive impact of our contribution.
Follow me on Twitter at @tbaraslupski
Sign up for the AI for Good newsletter