Can remote sensing map forests ? In principle, this is nothing new. Scandinavian countries are pioneers in this field and have already been using LiDAR technology  for several years. What about in France ? There has not been much research into this subject, and questions remain : can these methods be transferred to France ? Scandinavian forests are fairly flat and homogeneous ; French forests contain more species and mixtures. Too many parameters ? Not so.
Currently, remote sensing technologies such as LiDAR are evolving, delivering increasingly precise spatialized data. How can the data gathered by the scanners be linked to the information collected on the ground by forest rangers ? The FORESEE project  uses remote sensing to develop evaluation tools to implement a balanced forest biomass network, using reliable estimates of availability on an operational scale.
The FORESEE project
In order to test the methods on areas with contrasting stands and topography, 4 study sites were selected as a representation of French forests :
- The Alps : mountain forests
- The Landes Forest : softwood plantations
- The Lorraine Region : leafy plains forests
- The Vosges Mountains (Haut-Rhin department) : an area of 1,360km2 which includes a variety of forest types, the best type of validation site.
The project brought together researchers, managers and private stakeholders. It was an opportunity to deal with the problems they each face. "Managers don't need to understand how the laser works, points out Jean-Matthieu Monnet, researcher at Irstea's Grenoble center specializing in the development of forest inventory methods based on LiDAR data. They are interested in the maps produced by the black box of modeling. Are they precise enough ? Can they be used ?"
Improving method reliability
In order to meet the expectations of the wood industry, scientists started by reproducing methods tested on the various study sites. Faced with reassuring results, comparable to those achieved in forests with simple structures and species mixtures, researchers started working to improve methods and evaluate their reliability.
In partnership with teams from the French National Forest Office (ONF), researchers from Irstea's Montpellier center developed models based on a physical interpretation of LiDAR point clouds, with particular focus on understanding laser-plant interactions. "Tree foliage, whether young or old, oak or fir, does not react in the same way when it is lit up by a laser flash. When this foliage is all mixed up in a forest, it can cloud the relationship between data collected and the view of the actual forest. We have managed to obtain much clearer results than we expected."
Work was also undertaken to detect trees using LiDAR data: height, surface, volume, etc. Using an algorithm based on geometric characterization, it is now possible to extract the individual characteristics of each tree, whether it is hidden by another tree or not. Will it be possible to differentiate individual trees by their species ?
"For a long time, we observed the forest by taking aerial photos, infrared in particular, although this results in a qualitative overview. Using LiDAR, it is now possible to create a spatialized view with volume distribution rather than a general overview of a region. We are clearly targeting quantitative data."
At Irstea's Grenoble center, work has focused on the influence of the quality of regional reference points and LiDAR data on the precision of the maps obtained. More accurate maps can be obtained by increasing the number of data points used as references and using higher density LiDAR readings. This, however, leads to increased costs. Results will allow managers to optimize resources linked to LiDAR inventories according to their expectations of accuracy.
According to national statistics, forest surface areas are increasing, along with the quantity of available wood. "However, statistics and public authorities do not specify which key areas should be exploited," notes Monnet. Using LiDAR means that mapping can become more refined, down to forest unit or landowner scale. For example, on the Vosges site (1,360km2), researchers have been able to estimate the volume of wood for forest areas of 25m x 25m with an average deviation of 30%! For larger plots, precision is close to that achieved using traditional ground methods.
Issues limiting the transfer of these new methods to managers include : awareness of margins of error present in every measurement taken on the ground. Researchers are working to qualify this error ; however, according to Monnet, "significant support work is needed." A further limitation is : the cost of data (gathering raw data - aircraft rental, LiDAR)! “This requires the implementation of large-scale data gathering policies to lower costs.” Meanwhile, project specifications have been drawn up, describing all the steps needed to create a forest inventory using LiDAR, from raw data to finished maps. These are valuable tools for managers.
A real French research community has grown up from the initial 4 years of research. “Before, everyone worked on their own. The project clearly created a drive, bringing all the ongoing work across the country together.” This French expertise now needs to be consolidated.
Estimating forest productivity from archive photos
Towards the end of the project, a new method was developed, based on old photographs. By selecting sequences of old photographs (from the Second World War onwards), researchers from Irstea and Inra were, for example, able to recreate the height of the trees. “We virtually rebuilt trees to the height they were at the time, which enabled us to estimate previous rates of forest growth. In turn, this gave us an idea of how forests might grow in the future, telling us what volume would be available in different places.”
Digital techniques make this easy. But what about poorly preserved silver prints ? Scientists faced real problems to fix these images to neutralize any distortions and estimate the height of the trees correctly. “The method provides a concept of time, where LiDAR is focused on a given moment.” Researchers hope to develop this process further.
For more information
- Consult the websites of the Mountain Ecosystems (EMGR) research unit in Grenoble and the Land, Environment, Remote Sensing and Spatial Information (TETIS) research unit in Montpellier.
- Using remote sensing in forest ecology.
LiDAR, an aerial laser scanner, collects data on topography and plant structure by sending laser flashes that are reflected by any objects in their path. The raw data are then recorded as a 3D point cloud.
 FORESEE (2010-2014): Forest Resource Estimation For Energy, ANR Bioénergies Project 2010. Partners: FCBA (coordinator), Inra, ONF (R&D network), Irstea Montpellier center (TETIS JRU) and Irstea Grenoble center, IGN, S
yntégra (surveying company), Union de la coopération forestière française (French Forest Cooperatives Union).