Biomass estimation is a crucial component in the areas of carbon storage and forest management. Biomass can be analyzed rather at forest or at Single Tree Level (STL) and studies in both directions can be found in literature. Nevertheless, working at STL provides more accurate estimations. Such a level of information can be obtained by means of LiDAR data, but this type of data is usually expensive and not available for all Countries. Thus, few studies at STL can be found in developing countries and tropical forests. This study presents one of the first works for biomass estimation in tropical forest located in Gorgona Island, Colombia. Adaptations have been made to state-of-the-art methods for them to properly work in a complex forest, where trees sizes vary a lot from one area to another. Preliminary results are shown for some areas, allowing to see the capabilities of the adaptation.