Welcome to the webpage of the research group on the foundations for statistical methods in the environmental sciences (FORMES). The group is led by Prof. Yves Atchade. Our work focuses on statistical methods and theory with relevance in environmental sciences.

We are always looking for interested students (undergrad, master, phd), and postdocs. Please see here for an ad that we are currently running for a postdoc position. Please reach out (atchade at bu dot edu) if interested.

Here are some problems that we are currently working on.

Large scale Bayesian inference. We aim to develop computationally tractable Bayesian framework for large scale modeling problems. Theoretically, we are keenly interested in the behavior of posterior distributions in the high-dimensional limit, and how to best sample from them using Markov Chain Monte Carlo or related methods.

Rainfall prediction in West Africa and deep learning. Reliable rainfall predictions have the potential to improve the livelihood and the resilience to weather extremes of millions of people in Africa and other developing countries. However our ability to do these predictions remains stubbornly poor in many parts of Africa. In this project we are exploring various approaches for building physics-aware statistical models (from simple to deep learning models) to improve rainfall prediction in West Africa.

Learning to invert remote sensing data at local scales. Satellite remote sensing data lead to challenging inverse problems driven by the radiative transfer equation. Due to the constraints of processing remote sensing data at the scale of the entire planet, the methodologies currently employed to solve these radiative transfer inverse problems remain adhoc, and requires extensive prior information. We are working to develop modeling approaches (including re-purposing existing foundational models) for solving radiative transfer inverse problems at local scales.

Field project: a network of environmental sensors in West Africa. We are working to deploy and maintain a network envinronmental sensorin Cote d'Ivoire. The purpose is to determine to what extend data collected from inexpensive environmental sensors can be coupled with remote sensing data to improve estimates of environmental states. This is a joint project with ENSEA, ABIDJAN, CI and the tahmo foundation.