The task of tree species recognition is to recognize the tree species using photos of their leaves and barks. In this paper, we propose an evidential k-nearest neighbors (k-NN) combination rule. The proposed rule is adapted to classication problems where we have a large number of classes with an intra-class variability and an inter-class similarity like the problem of tree species recognition. Finally, we compare the performance of the proposed solution to the evidential k-NN.