Forest ecosystems play a fundamental role in natural balances and climate mechanisms through their contribution to global carbon storage. They also represent important reservoirs of biodiversity. The sustainable management and conservation of forest resources is, therefore, crucial in the current context of global warming and biodiversity loss. To tackle such challenges, earth observation data has been identified as a valuable source of information capable to provide stakeholders with informative indicators to support the decision making process related to forest ecosystems management. In particular, LiDAR remote sensing has proven to be a powerful tool for the characterization of structural properties of forest ecosystems, which in turn are valuable information for their monitoring. To deal with this particular issue in the context of simultaneous forest variables estimation in a multi-output regression setting, we propose a deep learning based strategy to combine together information coming from metrics derived from high density 3D-point clouds acquired by airborne laser scanning (ALS) and topography descriptors extracted from the same source. The performance of our framework is evaluated on two stand-level forest variables of interest: Total Volume (referring to total wood volume) and Basal Area. It is compared to KNN and Random Forest (RF) algorithms, and also to a single-output version of our framework. As a result, the proposed multi-output framework performed better than KNN and RF and achieved a R2 value of 0.72 and 0.73 for Total Volume and Basal Area, respectively. The obtained results are similar to those obtained when running two single-output deep learning frameworks and underline that the availability of additional topography descriptors to enrich the information provided by standard LiDAR-derived metrics brings interesting performance improvements in the estimation of both forest variables of interest.