In this study, we evaluate the use of texture descriptors obtained from very high-resolution satellite imagery to improve tree species classification in a tropical forest. Gray level co-occurrence matrix (GLCM), gabor and wavelet features were combined with reflectance bands of the WorldView-2 (WV-2) and QuickBird-2 (QB-2) satellite sensors to perform species classification. The use of QB-2 and WV-2 reflectance data yielded 26.6% and 44.8% of average accuracy, respectively. Texture features combined with QB-2 bands did not improve significantly the results. However, the combination of GLCM features with WV-2 reflectance bands increased the average accuracy up to 5.7%, exceeding 15% for some species. Pioneer species such as Cecropia hololeuca and climax species such as Cariniana legalis were classified with approximately 90% accuracy. The results highlight the potential use of WV-2 imagery to monitor tropical forest environments.