Seeking two Master-level internships in Montpellier about cocoa trees and lidar

rvezy · November 5, 2025, 3:39pm

Analysis of the relationships between cocoa tree architecture and pod production in agroforests using terrestrial lidar data
Detailed characterization of leaf surface distribution and light environment in cocoa agroforests using terrestrial lidar point clouds

Leave a comment below or send me a message to candidate.

First one: MSc Internship – Analysis of the relationships between cocoa tree architecture and pod production in Cameroonian agroforests using terrestrial LiDAR data

Context

As part of the agroecological transition, agroforestry systems are emerging as sustainable, multifunctional production models. They maintain economically viable yields while providing essential ecological functions such as microclimate regulation, biodiversity conservation, and carbon storage.

In Cameroon, cocoa (Theobroma cacao) is often cultivated in agroforestry systems alongside trees that provide timber, fuelwood, or non-timber forest products (fruits, bark), or that serve primarily as shade. These systems display a high degree of structural and productive variability, which remains poorly understood.

Yields in cocoa agroforests vary greatly between plots. This variability may be linked to structural and architectural differences among trees and between plantations (age, planting density, species diversity, etc.). Such properties, which directly affect a tree’s ability to acquire and distribute resources, are however difficult to quantify systematically because of the three-dimensional complexity of trees and the limited accessibility of their crowns.

The rise of terrestrial LiDAR (TLS) technologies opens new opportunities, as they allow precise 3D characterization of tree structure. The challenge now lies in extracting relevant architectural indicators from these large datasets in order to describe phenotypic variability and relate it to production performance.

Internship objective

The general objective of this internship is to analyze the relationship between the architectural properties of cocoa trees and their pod yield, using data from terrestrial LiDAR surveys conducted in Cameroonian agroforests.

The intern will take part in the full LiDAR data analysis pipeline and perform statistical analyses on structure–production relationships, including:

Spatial matching between field inventory data (tree positions, yield monitoring, diameter at breast height, etc.) and LiDAR data.
Isolation of individual trees within the LiDAR point cloud.
Segmentation of the point cloud into woody and leafy components.
Reconstruction of tree models from the woody point cloud.
Statistical analysis of the effects of architectural traits and contextual variables (e.g., plot planting density, shade trees) on yield.

The work will build upon existing 3D point cloud processing algorithms, which will be adapted to the specific characteristics of cocoa agroforests.

Candidate profile

Level: Master’s 2 or final-year engineering student
Background: Agronomy, ecology, forestry, modeling, applied mathematics, or computer science
Skills and interests:
- Interest in processing large 3D datasets (computational algorithms)
- Programming skills (Python, R, etc.)
- Statistical analysis (R)
- Interest in agroecological research
- Strong writing skills
- Desired qualities: autonomy, scientific rigor, initiative, and enthusiasm for interdisciplinary work

Supervision

The internship will be supervised by Olivier Martin (CIRAD UMR AMAP), Rémi Vezy (CIRAD UMR AMAP), and Ivan Cornut (CIRAD, Yaoundé, Cameroon).

Duration: 6 months

Location: Montpellier, France

References

Wibaux, T., Lauri, P.-É., M’Bo Kacou, A. A., Kouakou, O. P., & Vezy, R. (2025). A spatial perspective on flowering in cauliflorous cacao: architecture defines flower cushion location, not its early activity. Annals of Botany, mcaf107. https://doi.org/10.1093/aob/mcaf107
Peynaud, E., & Momo Takoudjou, S. (2024). Terrestrial LiDAR point cloud dataset of cocoa trees grown in agroforestry systems in Cameroon. Data in Brief, 53, 110108.
Mbile, J. G. N., Saj, S., Enock, S., Mala, W. A., & Harmand, J.-M. (2025). Cacao stand rehabilitation practices affect long-term cocoa production in agroforestry systems in Cameroon. Agroforestry Systems, 99, 199.
Raumonen, P. et al. (2013). Fast Automatic Precision Tree Models from Terrestrial Laser Scanner Data. Remote Sensing, 5, 491–520.

Second one: MSc Internship: Detailed characterization of leaf surface distribution and light environment in Cameroonian cocoa agroforests using terrestrial LiDAR point clouds

Context

As part of the agroecological transition, agroforestry systems offer a promising response to sustainability challenges by combining agricultural productivity, resilience to climatic hazards, and diversity of resources produced. These multifunctional systems provide both goods (timber, fruits, bark) and ecosystem services (carbon storage, microclimate regulation, soil protection).

In Cameroonian cocoa agroforests, the cocoa trees are often grown under a canopy of tall associated trees that create a favorable microclimate. These shade trees alter the light environment, which can significantly influence the health, yield, and lifespan of cocoa trees. Cocoa is classically considered a shade-tolerant species, with optimal yield typically achieved under around 30% shade.

Despite their importance, the relationships between the light environment of cocoa trees, their leaf surface distribution, and their agronomic performance (yield, plant health) remain poorly understood. This is due to the high heterogeneity of agroforestry systems (species diversity, complex plant architecture), the multiple interactions among species (light, nutrients, root competition), and the impact of microclimatic conditions on pest dynamics.

Traditional methods for estimating leaf area and measuring intercepted light (such as hemispherical photography or light sensors) are useful but have major limitations: they are point-based, labor-intensive, and do not provide continuous, three-dimensional characterization over an entire plot.

Terrestrial LiDAR (TLS) sensors, on the other hand, enable detailed 3D descriptions of vegetation structure through point clouds. AMAPvox, a model developed within the UMR AMAP research unit, transforms these point clouds into voxel spaces (3D cubes) that include estimates of leaf surface density. These voxelized representations can then be used to realistically model light irradiance at any point within a cocoa plantation and over different time periods. However, species-specific parameterization and evaluation of the model are necessary to obtain realistic estimates of leaf surface density.

Internship objectives

The internship will have two main goals:

To parameterize the AMAPvox model for cocoa agroforests in order to estimate the distribution of leaf area for cocoa trees and associated species from TLS point clouds.
To develop methods to derive, from the voxelized outputs, the amount of light intercepted by cocoa trees.

Tasks

The intern will:

Preprocess TLS point clouds acquired in cocoa agroforests representing a gradient of structural complexity:
- Develop or benchmark existing algorithms for separating woody and leafy components
- Generate a digital terrain model
Estimate the parameters required for AMAPvox modeling, including:
- Estimation of leaf angle distribution
- Determination of optimal voxel size
Develop an approach to quantify the amount of light intercepted by cocoa trees over a given period using AMAPvox outputs.
Evaluate the estimated leaf surface areas and modeled light distribution by comparison with in situ measurements (hemispherical photographs, light sensors, LAI-2200) or in silico reference measurements (manual measurements on the point cloud).
Analyze the relationships between received light, yield, and plant health across the structural complexity gradient.