Natural forest regeneration on non-forest land in Lithuania is only recognized as a forest after it reaches the age of twenty years. If these areas are not properly inventoried and registered, they may be treated as abandoned agricultural land. Consequently, this can lead to legal sanctions or a decision of landowners to clear large areas leading to the removal of young trees. This study aims to investigate the potential of timely inventory of afforestation on agricultural land by employing a combination of advanced spatial analysis, artificial intelligence, and remote sensing techniques. We hypothesized that the improvement of abandoned land identification in Lithuania depends on mapping and specifying trees outside currently registered forest lands. Primarily, we mapped and specified areas covered by woody vegetation using remotely sensed data based on airborne laser scanning (ALS) and aerial imaging. A combination of individual tree detection and area-based approaches in processing the ALS data resulted in most promising solution for mapping the objects of interest. Spectral information from optical images increased the accuracy of estimations. Then, we improved the characterization of mapped objects by reconstructing their historical development using simulation models based on machine learning and deep learning approaches. All available auxiliary historical geographic information was used to facilitate the identification of afforestation on non-forest lands and the abandonment of agricultural lands. Alongside mapping areas grown with woody vegetation, we managed to achieve better understanding of factors behind abandonment of agricultural lands and clearing out potential new forests. We suggest our approaches to be used also for wall-to-wall land-use change mapping in the past or future in accordance with forthcoming EU legal acts for the LULUCF sector within the frames of greenhouse gas management efforts.