Positive Plant-Plant interactions and spatial Patterns in Pyrenean Post-mine tailings – SixP

Bringing together researchers and partners involved in airborne data acquisition, the management of former mining sites, the geochemistry of contaminated soils, plant ecology in constrained ecosystems, and the development of artificial intelligence methods, the SIXP project was organized into three major work packages.

The first WP was dedicated to the characterization of the environmental gradients present on the different sites studied, whether in terms of contamination, contextualization of this pollution (by characterizing local topography, directions of rainfall runoff), or other variables necessary to understand vegetation dynamics (such as soil fertility). All the data acquired was integrated into a GIS accessible to all the project consortium, enabling them to precisely contextualize the areas of observation and experimentation used in the second and third WP.

The second WP was dedicated to the study of interactions between plants along contamination gradients. The variation in these interactions was studied along gradients located at different elevations, using both observational (by measuring the distribution of species in relation to each other) and experimental (by transplanting species close to potentially “nurse” plants or without neighboring plants) approaches.

The third WP was dedicated to automatic mapping of plant species using AI methods applied to very high-resolution aerial imagery. We have evaluated several reference models in semantic segmentation and object detection, characterized images coming from the different study sites, countered the spatial error intrinsic to acquisition campaigns, regressed biophysical and biochemical variables, and generated digital terrain models from 3D point clouds.

On-site measurements have highlighted the importance of creating detailed contamination maps and assessing metal environmental availability while considering the geological, mineralogical, and geomorphological context of the sites. Measurements of metal-environmental availability by acetic acid has proven to be accurate across all studied soils. Finally, the results have demonstrated a very high ecotoxicological risk for all studied soils concerning Cd, Zn, and Pb. To complement this environmental characterization, UAV acquisition has enabled the production of LIDAR point clouds or aerial imageries with millimeter- or centimeter-scale resolutions that are rarely obtained in mountainous contexts.

In community ecology, the results have confirmed the importance of facilitation in highly contaminated areas. Facilitation is even more significant when interacting species or individuals are functionally differentiated. In particular, plants highly adapted to metal presence seem to promote the presence of less-adapted species. Therefore, facilitation contributes to maintaining plant diversity in contaminated areas. Regarding the underlying effects, microclimatic improvements play a particularly important role. In corresponding sparsely vegetated areas, hot and dry summer days can be particularly stressful for vegetation, even at high altitudes. These microclimatic improvements are linked to shading and transpiration by metal-tolerant species. Additionally, mining residues in mountainous and sloped areas create unstable and rocky soils. The possibility that positive effects are linked to soil stabilization is still under analysis. Finally, negative effects (opposite to facilitation) have also been observed but in the less contaminated parts of the gradients. There, some plants can absorb and accumulate important amount of metals in their leaves. After leaf fall and decomposition, local metal concentration can increase, thereby affecting neighboring plants sensitive to metals in these less contaminated areas. Noticeably, all these effects disappeared during the heatwave of summer 2022, which had a severe impact on vegetation. Considering the significant role of facilitation in contaminated areas, this raises questions about the long-term impact of climate change on the studied vegetation.

In the field of artificial intelligence, we have obtained several significant results. We demonstrated that plant classification should be addressed as an object detection task and that the spatial imprecision of annotations requires special consideration. We have also proposed a semi-supervised approach for biophysical variable regression, as well as a solution to learn how to generate a digital terrain model from a 3D point cloud.

The studied sites still show high concentrations of metals even several decades after mining activities ceased, and the ecotoxicological risks are very high for all the soils studied with respect to Cd, Zn, and Pb. Even on less impacted soil, we found several species able of hyperaccumulate metals in their leaves. Following SIXP, several perspectives are drawn concerning metal-accumulating plants in other contexts (other regions, other types of metal-rich soils), in order to gain a general understanding of this unique adaptation in the plant kingdom. The perspectives also concern the long-term evolution of plant communities on metal-rich soils in the context of climate change. For more applied perspectives, the knowledge acquired about vegetation on mining soils, and in particular about the pivotal role of facilitation, now enables to test the use of mixtures of species between which facilitation is expected. These mixtures could help the reclamation and revegetation of mining sites, both in the study area and in other regions, in order to limit the dispersion of contaminated particles in the environment. These applications would complement the already well advanced phytoremediation technologies. Locally in the region of the study, regular meetings with stakeholders enabled us to share the knowledge acquired. Frequent contacts with pastoral associations led to the planning of joint visits to the sites studied, in order to identify areas problematic for sheep grazing, where both vegetation and dust particles are contaminated.

In mountainous environments, airborne data acquisition by drone generally does not allow work at centimetric resolutions. The project made it possible to improve acquisition methods at these scales (flight paths, targets, data post-processing), thereby enhancing their performance for future applications. The efforts made on acquiring original very high-resolution data, correcting and annotating them, provide us with a high-quality dataset from which it is possible to train deep models. Existing solutions have shown their limitations in this specific context, and more advanced research studies should be conducted to manage to exploit such available images to map plant species and their interactions. Coupling 2D and 3D data is also promising.

Mine tailings are witnesses of exploitation of ore bodies which took place several decades ago. These tailings are an important part of the 100 000 heavy-metal polluted sites which require urgent rehabilitation in Europe. These tailings represent a sizeable source of contaminated material spreadable in the environment. Despite their toxicity for non-adapted species, rare heritage plant communities, metallicolous grasslands, established on them gradually over many years. Six-P project aims to assess the role of plant-plant interactions in these specific plant communities, both as a key system to understand variation in plant-plant interactions along stress gradients and as a possible restoration tool. In mountainous areas, tailings form particularly harsh environments for plant growth (metal toxicity, climatic constraints). In such conditions, positive plant-plant interactions are expected according to a dominant ecological theory: The Stress Gradient Hypothesis (SGH). However, this hypothesis has been poorly investigated along (metal) pollution gradients. In addition, and regardless of pollution gradients, there is a need to better define its conditions of application. SixP aims to contribute to this active field of research by focusing on four directions: i) to characterize the variation of plant-plant interactions along gradients of metal phyto-availability, while explaining the specific role of metallicolous species in these interactions; ii) to better identify the effects of multiple stress factors on these interactions; iii) to specify the plant functional strategies at stake; and iv) to assess the effect of plant-plant interactions at the community scale. The project will be implemented in several mine tailings in the Pyrénées at different altitudes (in the montane zone, and at the subalpine-alpine zone). At each site, several areas will be specified from peripheral low-contaminated areas towards tailings centers corresponding to a gradient of metal phyto-availability. The first three research directions will then be addressed by experimentations manipulating species in interaction. As for the last direction, the combination of very high resolution airborne data (lidar, multispectral images) covering the studied areas with in situ observations in a deep learning framework will be used to map species distribution and their geomorphological position. Spatial patterns of the different interacting species (aggregation vs repulsion) will exhibit the effects of plant-plant interactions on the long-term. Six-P relies on a multidisciplinary consortium with expertise in ecology, metals biogeochemistry, airborne data acquisition, computer vision, machine learning and management of post-mining sites. In addition to the valuable and general knowledge acquired on plant-plant interactions, benefits are expected in the phyto-management domain, by proposing the use of several species associations as viable alternative to the already available techniques. Management of rare metallicolous grasslands of heritage value could also be improved thanks’ to project results. Finally, Six-P will increase the interest of computer vision and artificial intelligence groups on ecological issues. The deep neural network models designed in SixP could also be applied to other problems in ecology, since transferability of deep networks is improving regularly.