Post-doctoral research position Predicting Environmental Changes Using Deep Learning, Satellite Remote Sensing and Future Climate Projections. Case studies in Southern Kenya and the Brazilian Amazon W/M

27 Mar 2024
Job Information

Organisation/Company

IRD Occitanie

Research Field

Environmental science

Researcher Profile

First Stage Researcher (R1)

Country

France

Application Deadline

27 Apr 2024 - 00:00 (Europe/Paris)

Type of Contract

Temporary

Job Status

Full-time

Offer Starting Date

1 Jul 2024

Is the job funded through the EU Research Framework Programme?

Not funded by an EU programme

Is the Job related to staff position within a Research Infrastructure?

No

Offer Description

Context

MOSAIC (Multi-site Application of Open Science in the Creation of Healthy Environments Involving Local Communities) is a European project focusing on Planetary Health. It aims to integrate data from diverse sources to elucidate the relationships between environmental changes and human health in two distinct study sites in Kenya and Brazil. In Kenya, the project focuses on southern Kenya and northern Tanzania, addressing the repercussions of drought, land degradation, and biodiversity loss on the Maasai herders' way of life, including aspects of their health and environmental sustainability. Meanwhile, in the Amazon, the study sites delve into the health and environmental impacts of deforestation and mining on local and indigenous populations. To achieve its objectives, MOSAIC will harness a wide range of data sources, ensuring a comprehensive analysis of the environmental and health landscapes at its study sites. This includes remote sensing and Earth observation data for tracking land use changes and climate variables; health surveillance system data; and local population data through community-based participatory approaches for insights on environmental observations and socio-economic indicators. Additionally, environmental and ecological data on biodiversity and geographical characteristics will be collected on-site, and open data repositories will be used to enrich the contextual understanding of each site. These diverse data streams will collectively inform the project’s effort to model health outcomes from environmental changes, fostering a holistic approach to addressing planetary health challenges.

Research Objectives

Within the scope of the MOSAIC project, comprehensive studies are set to be undertaken to collect essential data and develop models to uncover the main drivers behind human health issues and biodiversity loss at each study site. This post-doctoral research will be dedicated to constructing sophisticated models designed for the prediction of pivotal future events that may impact these areas. Such efforts aim to facilitate the adoption of anticipatory mitigation tactics. For example, if reductions in forest coverage and droughts are identified as significant contributors to mental health challenges in Kenya, we aim to develop predictive models that can accurately forecast these occurrences. Through the integration of these forecasts, we aim to anticipate future crises, thereby enabling the implementation of timely and impactful mitigation measures.

Proposed Methodology

This research stands out for its use of both Earth observation data and future climate projections to forecast environmental changes and extreme weather events. To do this, we will leverage Deep Neural Networks (DNNs), which have shown promise across various fields like drought detection [1] and deforestation monitoring [2]. These successes are largely due to Foundation Models designed for analyzing satellite images [3][4]. Such models are pre-trained on extensive datasets of satellite images and can be easily adjusted to new tasks, showcasing remarkable flexibility even with limited amounts of task-specific data. Hence, an important first step in this post-doctoral research will be to adapt these models to our study areas in Southern Kenya and the Brazilian Amazon. By addressing the unique challenges of these regions through a unified approach, we aim to establish a clear methodology for fine-tuning neural networks to predict environmental variables from satellite images. This effort is expected to make a substantial contribution to the field of climate change mitigation through artificial intelligence. Moreover, this project will pioneer the integration of climate change prediction variables, specifically from the CMIP6 dataset [5], into our models. This integration seeks to evolve these systems beyond mere detection, enabling them to forecast future environmental threats with better precision. By incorporating predictions of future climate variables, we expect to enhance our models' ability to preemptively identify potential environmental crises, empowering authorities to implement effective mitigation strategies in advance. Additionally, our research will explore the synergistic prediction of related environmental events within a single modeling framework. Evidence from the field of Machine Learning suggests that joint prediction of related variables can significantly improve model accuracy by introducing a form of regularization [6]. For example, predicting deforestation alongside potential flood occurrences could yield more accurate forecasts, offering novel insights into the interconnected nature of environmental phenomena. This innovative approach has the potential to redefine predictive modeling in environmental science. Given environmental and climate data's complexity and variability, incorporating diverse datasets from different regions will be crucial to enhance the robustness of our models. Therefore, the researcher will also be expected to integrate publicly available databases to enrich model development's pretraining phase and bolster our predictive frameworks' overall effectiveness. [

1] An, J., Li, W., Li, M., Cui, S., & Yue, H. (2019). Identification and Classification of Maize Drought Stress Using Deep Convolutional Neural Network. Symmetry, 11, 256. https://doi.org/10.3390/sym11020256 . [2] Ortega, M. X., Bermudez, J. D., Happ, P. N., Gomes, A., & Feitosa, R. Q. (2019). Evaluation of deep learning techniques for deforestation detection in the amazon forest. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 4, 121-128. [3] Fuller, A., Millard, K., & Green, J. (2024). CROMA: Remote Sensing Representations with Contrastive Radar-Optical Masked Autoencoders. Advances in Neural Information Processing Systems, 36. [4] Jakubik, J., Roy, S., Phillips, C. E., Fraccaro, P., Godwin, D., Zadrozny, B., ... & Ramachandran, R. (2023). Foundation models for generalist geospatial artificial intelligence. arXiv preprint arXiv:2310.18660. [5] Kim, Y. H., Min, S. K., Zhang, X., Sillmann, J., & Sandstad, M. (2020). Evaluation of the CMIP6 multi-model ensemble for climate extreme indices. Weather and Climate Extremes, 29, 100269. [6] Zhang, Y., & Yang, Q. (2018). An overview of multi-task learning. National Science Review, 5(1), 30-43.

Requirements

Research Field

Environmental science » Ecology

Education Level

PhD or equivalent

Skills/Qualifications

Expected candidate

We seek a dynamic candidate with a proven track record in developing deep learning models, especially within the realm of computer vision, and an interest in addressing environmental and health challenges through innovative technological solutions. The ideal candidate will immerse themselves in the MOSAIC project, contributing to data collection and description, co-supervising Ph.D. students, and engaging in critical discussions on data infrastructure to ensure the project's longevity and impact. A commitment to participatory science and the ability to render complex models interpretable to local communities is essential, as these elements are vital for fostering trust and ensuring the practical application of our research findings.

Working environment

The French Research Institute for Sustainable Development (IRD) IRD is a French public scientific and technological institution (EPST) that has been operating in the Global South for over 60 years. It is under the dual supervision of the ministries responsible for Research and Foreign Affairs. Its research, expertise, valorization, and training activities aim to contribute to the economic, social, and cultural development of the Global South. Nearly 40% of the Institute's staff are assigned overseas and to French overseas territories. Video presentation:The IRD in 230 seconds.

UMR Espace-Dev

The Mixed Research Unit ESPACE-DEV bases its scientific activities on defining integrated approaches to societal and environmental sustainability. These approaches are developed across various themes and methodologies, drawing on the unit's multidisciplinary skills, with a scientific approach known as sustainability science. This guides the research towards approaches that prioritize the search for solutions, co-construction with stakeholders, and the development of support and data-sharing infrastructures. The scientific orientations aim to address and deal with complex questions and problems on the following fundamental themes: 1) socio-ecological and energy transitions, 2) health and well-being of societies while preserving resources and ecosystems, and 3) vulnerability, adaptation, and viability of territories including insular and coastal systems. By definition, ESPACE-DEV is multidisciplinary and multi-institutional.
 

Languages

FRENCH

Level

Basic

Languages

ENGLISH

Level

Excellent

Additional Information
Benefits

Working conditions

Throughout your professional journey, the IRD supports you in developing your skills. The institute provides a range of tools such as a digital integration path, access to continuous training, promotion, and mobility. Depending on the activities, the IRD offers the possibility to work from home 1 to 3 days a week. By joining the IRD, you will benefit from:

● 32 days of leave + 13 RTT (for a full-time position at 38 hours and 30 minutes per week)

● Collective catering

● Optional annual subscription to the Association of Social Works: holiday-leisure and sports-cultural benefits

● A contribution of 15€/month towards social protection

Selection process

IRD - Chercheur Prédictions des changements environnementaux H/F

Website for additional job details

https://emploi-recrutement.ird.fr/offre-de-emploi/emploi-chercheur-predictions-…

Work Location(s)

Number of offers available

1

Company/Institute

IRD

Country

France

Geofield

Where to apply

Website

https://emploi-recrutement.ird.fr/offre-de-emploi/emploi-chercheur-predictions-…

Contact

State/Province

Occitanie

City

Montpellier

Website

https://emploi-recrutement.ird.fr/offre-de-emploi/emploi-chercheur-predictions-des-changements-environnementaux-h-f_230.aspx

Street

911 Avenue Agropolis, 34394 Montpellier

Postal Code

34000

STATUS: EXPIRED