We describe in this paper our response to the EGC Challenge 2017. Exploratory data analysis has first lead to understand the distribution of variables and detect strong correlations. We then defined two new variables combining dataset variables. Several classification algorithms have been experimented for the first task of the challenge. Performances have been evaluated by 10-fold cross validation. It has resulted in selecting the best unilabel and multilabel classifiers. On both unilabel and multilabel levels, the best classifier outperforms the reference scores by approximately 2%. We also explored the second task of the challenge. On one hand, association rules have been searched. On the other hand, the initial dataset has been enriched with domain knowledge such as climate data (rainfall, temperature, wind) or taxonomic data in the field of botany. Furthermore, geographical and cartographic data have been used in a visualisation tool for representing trees.