Remote sensing of forests with airborne laser scanning has been used increasingly in the last decades. Due to its capability to provide very detailed three-dimensional information on the structure of forests and individual trees it is very suitable to detect changes in forests. However, current change detection methods in this field often cannot exploit the full potential of the information. The increasing gap between the ability of these methods and the information available in increasingly denser airborne laser scanning data has led to this research. Therefore a three step approach is developed that aims to use the full three-dimensional information in the data to detect harvested and fallen trees in forests. First, differences in the repeated data are obtained through an adapted cloud-to-cloud comparison, such that the data can be reduced to the changed points only. Secondly, from the changed points the individual changed objects are extracted and finally classified as either removed trees or other objects. The findings show that harvested and fallen trees can be detected with an overall accuracy of up to 94% and an omission error of 0%. The results further indicate that smaller trees and changes below the canopy can be identified, although with lower success rates. All in all, it can be seen that the potential of the three-dimensional nature can be better exploited by considering each individual point of the data.