The French convergence institute CLAND (Climate change and land use) and its partner research unit ECOSYS are looking for a motivated candidate for a 12-month postdoctoral position on the mapping of agricultural organic nitrogen sources across France, at the interface between remote sensing, deep learning and agri-environmental sciences.
Scientific background
Organic fertilisation from livestock manure is one of the main sources of diffuse nitrogen in France, with roughly 300 million tonnes of manure produced each year. It is also one of the practices promoted to increase carbon storage in agricultural soils. Any spatially explicit mapping of organic fertilisation pressure first requires an accurate localisation of the sources - livestock buildings and facilities - combined with a quantitative estimate of the manure produced at each site.
The French national livestock identification database (BDNI), updated quarterly and coupled with the Land Parcel Identification System (RPG), already provides the location of farms and their animal numbers, but says nothing about the physical structures housing the animals and storing the manure. Recent work has demonstrated that agricultural facilities can be detected from satellite imagery using deep learning (de Senneville et al., 2025), an approach that can be transferred to livestock buildings. The BDNIÅ~RPG coupling can therefore serve as a georeferenced training set.
The project further aims to add a temporal dimension: the cattle production cycle - alternating between winter housing and summer grazing - strongly changes where and when nitrogen is excreted. The ultimate goal is to identify spatial concentrations (hotspots) and periods of peak production (hotmoments) of reactive nitrogen, and to assess how these match areas of potential NH₃ emission.
This spatially explicit quantification of the available organic residual products (ORP) also opens a second line of enquiry: the carbon storage potential of agricultural soils. The localised manure volumes determine where, when and in what quantities organic inputs can be applied to fields. The resulting maps will therefore make it possible to weigh the expected benefit for soil carbon storage against the associated nitrogen losses, and to inform tradeoffs between these two objectives at the territorial scale.
Tasks
- Build a georeferenced training set from the BDNIÅ~RPG coupling, focusing on the cattle sector (the best covered by the BDNI), and use it to train a deep-learning object detector identifying cattle buildings and their associated structures;
- Develop and validate a model linking the characteristics of the detected structures (built area, number of buildings, annexes) to the likely herd size expressed in livestock units (LU);
- Reconstruct annual herd dynamics per site and approximate the grazing/housing cycle using a satellite proxy;
- Produce a spatio-temporal map of organic manure production and identify hotspots and hotmoments;
- Disseminate results through publications in peer-reviewed journals and presentations at national and international conferences.
Profile
- PhD in remote sensing, geomatics, data science, agronomy, environmental science or a related field (obtained, or to be obtained by the starting date);
- Solid practice of geospatial data processing and GIS;
- Familiarity with livestock systems, nitrogen biogeochemical cycles or French agricultural databases (BDNI, RPG) is an appreciated asset;
- Documented experience in scientific writing and publication in peer-reviewed journals;
- Autonomy, a taste for interdisciplinary work and the ambition to pursue a scientific career;
- Good communication skills in English; knowledge of French is an asset but is not required.
How to apply
- Applications should be sent by email to Raia Silvia Massad (raia-silvia.massad@inrae.fr), Bertrand Guenet (guenet@geologie.ens.fr) and Marco Carozzi (marco.carozzi@inrae.fr) before 15 September 2026.
- Your application should include a motivation letter, a detailed academic CV (with publication list) and the contact details of two referees.
- The selection committee will review applications as soon as the call closes; shortlisted candidates will be invited for an interview.
- You will be based at the UMR ECOSYS, on the AgroParisTech campus (22 place de l'Agronomie, 91120 Palaiseau, France).
Reference: de Senneville A. et al. (2025). Towards Large Scale Geostatistical Methane Monitoring with Part-based Object Detection. arXiv:2507.18513.
The French convergence institute CLAND (Climate change and land use) and its partner research unit ECOSYS are looking for a motivated candidate for a 12-month postdoctoral position on the mapping of agricultural organic nitrogen sources across France, at the interface between remote sensing, deep learning and agri-environmental sciences.
Scientific background
Organic fertilisation from livestock manure is one of the main sources of diffuse nitrogen in France, with roughly 300 million tonnes of manure produced each year. It is also one of the practices promoted to increase carbon storage in agricultural soils. Any spatially explicit mapping of organic fertilisation pressure first requires an accurate localisation of the sources - livestock buildings and facilities - combined with a quantitative estimate of the manure produced at each site.
The French national livestock identification database (BDNI), updated quarterly and coupled with the Land Parcel Identification System (RPG), already provides the location of farms and their animal numbers, but says nothing about the physical structures housing the animals and storing the manure. Recent work has demonstrated that agricultural facilities can be detected from satellite imagery using deep learning (de Senneville et al., 2025), an approach that can be transferred to livestock buildings. The BDNIÅ~RPG coupling can therefore serve as a georeferenced training set.
The project further aims to add a temporal dimension: the cattle production cycle - alternating between winter housing and summer grazing - strongly changes where and when nitrogen is excreted. The ultimate goal is to identify spatial concentrations (hotspots) and periods of peak production (hotmoments) of reactive nitrogen, and to assess how these match areas of potential NH₃ emission.
This spatially explicit quantification of the available organic residual products (ORP) also opens a second line of enquiry: the carbon storage potential of agricultural soils. The localised manure volumes determine where, when and in what quantities organic inputs can be applied to fields. The resulting maps will therefore make it possible to weigh the expected benefit for soil carbon storage against the associated nitrogen losses, and to inform tradeoffs between these two objectives at the territorial scale.
Tasks
- Build a georeferenced training set from the BDNIÅ~RPG coupling, focusing on the cattle sector (the best covered by the BDNI), and use it to train a deep-learning object detector identifying cattle buildings and their associated structures;
- Develop and validate a model linking the characteristics of the detected structures (built area, number of buildings, annexes) to the likely herd size expressed in livestock units (LU);
- Reconstruct annual herd dynamics per site and approximate the grazing/housing cycle using a satellite proxy;
- Produce a spatio-temporal map of organic manure production and identify hotspots and hotmoments;
- Disseminate results through publications in peer-reviewed journals and presentations at national and international conferences.
Profile
- PhD in remote sensing, geomatics, data science, agronomy, environmental science or a related field (obtained, or to be obtained by the starting date);
- Solid practice of geospatial data processing and GIS;
- Familiarity with livestock systems, nitrogen biogeochemical cycles or French agricultural databases (BDNI, RPG) is an appreciated asset;
- Documented experience in scientific writing and publication in peer-reviewed journals;
- Autonomy, a taste for interdisciplinary work and the ambition to pursue a scientific career;
- Good communication skills in English; knowledge of French is an asset but is not required.
How to apply
- Applications should be sent by email to Raia Silvia Massad (raia-silvia.massad@inrae.fr), Bertrand Guenet (guenet@geologie.ens.fr) and Marco Carozzi (marco.carozzi@inrae.fr) before 15 September 2026.
- Your application should include a motivation letter, a detailed academic CV (with publication list) and the contact details of two referees.
- The selection committee will review applications as soon as the call closes; shortlisted candidates will be invited for an interview.
- You will be based at the UMR ECOSYS, on the AgroParisTech campus (22 place de l'Agronomie, 91120 Palaiseau, France).
Reference: de Senneville A. et al. (2025). Towards Large Scale Geostatistical Methane Monitoring with Part-based Object Detection. arXiv:2507.18513.
A propos des cookies
Qu’est-ce qu’un « cookie » ?
Un "cookie" est une suite d'informations, généralement de petite taille et identifié par un nom, qui peut être transmis à votre navigateur par un site web sur lequel vous vous connectez. Votre navigateur web le conservera pendant une certaine durée, et le renverra au serveur web chaque fois que vous vous y re-connecterez.
Différents types de cookies sont déposés sur les sites :
- Cookies strictement nécessaires au bon fonctionnement du site
- Cookies déposés par des sites tiers pour améliorer l’interactivité du site, pour collecter des statistiques
> En savoir plus sur les cookies et leur fonctionnement
Les différents types de cookies déposés sur ce site
Cookies strictement nécessaires au site pour fonctionner
Ces cookies permettent aux services principaux du site de fonctionner de manière optimale. Vous pouvez techniquement les bloquer en utilisant les paramètres de votre navigateur mais votre expérience sur le site risque d’être dégradée.
Par ailleurs, vous avez la possibilité de vous opposer à l’utilisation des traceurs de mesure d’audience strictement nécessaires au fonctionnement et aux opérations d’administration courante du site web dans la fenêtre de gestion des cookies accessible via le lien situé dans le pied de page du site.
Cookies techniques
Nom du cookie
Finalité
Durée de conservation
Cookies de sessions CAS et PHP
Identifiants de connexion, sécurisation de session
Session
Tarteaucitron
Sauvegarde vos choix en matière de consentement des cookies
12 mois
Cookies de mesure d’audience (AT Internet)
Nom du cookie
Finalité
Durée de conservation
atid
Tracer le parcours du visiteur afin d’établir les statistiques de visites.
13 mois
atuserid
Stocker l'ID anonyme du visiteur qui se lance dès la première visite du site
13 mois
atidvisitor
Recenser les numsites (identifiants unique d'un site) vus par le visiteur et stockage des identifiants du visiteur.
13 mois
À propos de l’outil de mesure d’audience AT Internet :
L’outil de mesure d’audience Analytics d’AT Internet est déployé sur ce site afin d’obtenir des informations sur la navigation des visiteurs et d’en améliorer l’usage.
L‘autorité française de protection des données (CNIL) a accordé une exemption au cookie Web Analytics d’AT Internet. Cet outil est ainsi dispensé du recueil du consentement de l’internaute en ce qui concerne le dépôt des cookies analytics. Cependant vous pouvez refuser le dépôt de ces cookies via le panneau de gestion des cookies.
À Savoir
- Les données collectées ne sont pas recoupées avec d’autres traitements
- Le cookie déposé sert uniquement à la production de statistiques anonymes
- Le cookie ne permet pas de suivre la navigation de l’internaute sur d’autres sites.
Cookies tiers destinés à améliorer l’interactivité du site
Ce site s’appuie sur certains services fournis par des tiers qui permettent :
- de proposer des contenus interactifs ;
- d’améliorer la convivialité et de faciliter le partage de contenu sur les réseaux sociaux ;
- de visionner directement sur notre site des vidéos et présentations animées ;
- de protéger les entrées des formulaires contre les robots ;
- de surveiller les performances du site.
Ces tiers collecteront et utiliseront vos données de navigation pour des finalités qui leur sont propres.
Accepter ou refuser les cookies : comment faire ?
Lorsque vous débutez votre navigation sur un site eZpublish, l’apparition du bandeau « cookies » vous permet d’accepter ou de refuser tous les cookies que nous utilisons. Ce bandeau s’affichera tant que vous n’aurez pas effectué de choix même si vous naviguez sur une autre page du site.
Vous pouvez modifier vos choix à tout moment en cliquant sur le lien « Gestion des cookies ».
Vous pouvez gérer ces cookies au niveau de votre navigateur. Voici les procédures à suivre :
Firefox ; Chrome ; Explorer ; Safari ; Opera
Pour Obtenir Plus D’informations Concernant Les Cookies Que Nous Utilisons, Vous Pouvez Vous Adresser Au Déléguée Informatique Et Libertés De INRAE Par Email à Cil-dpo@inrae.fr Ou Par Courrier à
INRAE
24, chemin de Borde Rouge –Auzeville – CS52627
31326 Castanet Tolosan cedex - France
Dernière mise à jour : Mai 2021
