Predict diet diversity from space: socio-ecologically informed use of satellite imagery with machine learning

As an essential part of food security, diet diversity often requires frequent household food security surveys to monitor. Diet diversity is typically measured by the number and range of food groups consumed by households over a given time period, and is often considered an indicator of nutrition intake. To have a good estimation of the diet diversity of rural households, particularly in less-developed regions, asks for high-resolution and longitudinal household survey data with high financial and personal costs. In this topic, we will investigate the role that satellite data could have in monitoring and estimating diet diversity of rural households. To date, remote sensing data and methods have been used to predict indicators that traditionally require surveys to estimate, such as poverty. The use of remote sensing data for food security however is mostly on the availability dimension through crop extent and yield mapping, yet the role of RS in predicting the utilization dimension of food security has not been explored.