MAS-DEWS: A multi-agent system for predicting Africa's drought

摘要

Droughts continue to be the number one hydro- and meteorological natural disaster inflicting Africa. In the last decade alone, the Continent contributed 55% of the world's droughts and 64% of deaths emanating from these disasters. With over 24% of her population affected, Africa accounted for 47% of people affected by droughts in the same period. Given that rain-fed agriculture is a major source of food in most African countries, the droughts have contributed to the acute food-insecurity; this may explain why 35 of the 45 poorest countries are in the Continent. There is consensus that drought early warning systems that contextualize culture and risk, have higher chance of succeeding in reducing vulnerabilities and building resilience. A homegrown drought early warning system designed around cheaper wireless sensor weatherboards, mobile phones and weather lore has the ability to address Africa's unique challenges and opportunities. However, integrating these diverse components requires some level of intelligence that only found in multi-agent systems. We present the design, implementation and evaluation of a comprehensive system made up of seven sub-systems that are linked up together by intelligent agents that were implemented using the Java Agent Development. The system's evaluation, carried out in three locations in Africa, revealed the potential of the system.