Abstract
This work aimed to evaluate the impacts caused by extreme frost events in an ecological restoration area. We grouped the species in three ways: (1) type of trichome coverage; (2) shape of the seedling crown; and (3) functional groups according to the degree of damage caused by frost. The variables of the restored area and species characteristics were selected to be subjected to linear generalization analysis models (GLMs). A total of 104 individuals from seven species were sampled. The most affected species were Guazuma ulmifolia Lam. (98% of leaves affected), followed by Cecropia pachystachia Trécul and Hymenea courbaril L. (both 97%), Inga vera Willd. (84%), and Senegalia polyphylla (DC.) Britton & Rose with 75%. Tapirira guianensis Aubl. was considered an intermediate species, with 62% of the crown affected. Only Solanum granulosoleprosum Dunal was classified as slightly affected, with only 1.5% of leaves affected. With the GLM analysis, it was verified that the interaction between the variables of leaf thickness (Χ² = 37.1, df = 1, p < 0.001), trichome coverage (Χ² = 650.5, df = 2, p < 0.001), and leaf structure culture (Χ² = 54.0, df = 2, p < 0.001) resulted in a model with high predictive power (AIC = 927,244, BIC = 940,735, Χ² = 6947, R² = 0.74, p < 0.001). Frost-affected crown cover was best explained by the interaction between the three functional attributes (74%). We found that there is a tendency for thicker leaves completely covered in trichomes to be less affected by the impact of frost and that the coverage of the affected crown was greatly influenced by the coverage of trichomes. Seedlings with leaves completely covered in trichomes, thicker leaves, and a funneled or more open crown structure are those that are most likely to resist frost events. The success of ecological restoration in areas susceptible to extreme events such as frost can be predicted based on the functional attributes of the chosen species. This can contribute to a better selection of species to be used to restore degraded areas.