Project title : Uncovering the role of fisheries management in reducing poverty and malnutrition
Description of the project : The aim of this project is to develop a multidisciplinary understanding of which key aspects (governance, social, cultural, and economic) make a community successful in alleviating poverty and malnutrition in the context of small-scale fisheries. The project considers coastal sub-Saharan Africa. It combines satellite imagery, ecological, demographic, dietary and human health information from the past 20 years to assess the temporal dynamic of poverty and malnutrition.
Main mission : Artificial Intelligence (AI) is generating new horizons to tackle big societal challenges. Recent improvements in IA applied to satellite images have enabled accurate estimates of local poverty conditions. Although these methods have rapidly advanced, they have not performed so well in predicting indicators of malnutrition, probably because there is a lack of direct causality between simple satellite images and key nutrition variables of interest. A solution to overcome this limitation is to complement satellite images with auxiliary data providing additional high-resolution spatial information on relevant variables of interest to better predict local malnutrition. Although deep learning models fed with auxiliary variables have been used in various fields and applications such as land cover classification, they have never been used to predict malnutrition. Since auxiliary variables are often high-resolution gridded data, their incorporation can produce maps with a higher level of detail compared to classical approaches but can also promote a landscape perspective on malnutrition prediction.
The candidate will implement a deep learning model with several branches integrating both satellite images (Landsat and/or Sentinel) and auxiliary information to predict malnutrition at the scale of villages in sub-Saharan Africa. Since the drivers of malnutrition are certainly complex and interconnected - war and conflict, climate change, natural disasters, and poverty - a wide variety of auxiliary variables reflecting these conditions will be extracted and tested to improve the prediction of malnutrition indicators.
Activities : Three main tasks should be achieved within 18 months (+ potential internship):
Build upon deep learning models previously implemented to predict poverty and provide, for the first time, high-resolution malnutrition predictions for coastal sub-Saharan Africa.
Extract and test spatial auxiliary variables able to improve the prediction of malnutrition indicators.
Statistically assess whether proximity to managed marine areas results in positive human health and nutrition outcomes.
The net salary ranges between €2,200 and €2,380 per month.
Contract duration: 18 months
