Willow traits outperform taxonomy in predicting phytoremediation services.

Phytomanagement of contaminated sites can mitigate exposure risks for surrounding populations while providing numerous ecological services. To meet these goals, trait-based models are proposed to guide plant selection. However, this relies on the assumption that plant traits can effectively predict key services and that traits of individual species can be used to predict mean, or community-level traits of a given species assemblage. To critically evaluate these assumptions, we conducted a mesocosm study where three willow species were planted in all possible combinations (1 to 3 species) in contaminated soil under a controlled environment, for 110 days. At the community-level (for every mesocosm), we measured ten functional traits and three phytoremediation services (i.e. phytoextraction, phytostabilization, and translocation factor). We evaluated the differences between observed community-level traits and expectations from traits of the corresponding species grown in monocultures. Then, we compared the predictability of phytoremediation services through species composition and community-level traits. Our results indicate that, despite the short phylogenetic gradient, willow species exhibit distinct and predictable traits within assemblages. Moreover, trait values measured here are comparable to the values retrieved on the TRY database, confirming the potential of global databases to guide trait-based efforts in phytoremediation. Phytoremediation services were not predicted by species composition (mean R = 0.05), but rather well explained by community-level traits (mean R = 0.52). This suggests that models incorporating functional information are better suited to predict and understand phytoremediation services. Phytoextraction was generally correlated to fast-to-intermediate aboveground growth strategies with fast-growing belowground tissues, while translocation factors were associated with slower root growth. Phytostabilization was associated with faster-growing root systems and stress-tolerant capacities from the aboveground tissues. This experiment represent a strong test for trait-based models, given the short phylogenetic and contamination gradients tested. This reinforces the potential of trait-based models in phytoremediation and phytotechnologies more broadly.