No abstract

Data processing for forestry applications is challenged by the increasing availability of multi-source and multi-temporal data. The advancements of Deep Learning (DL) algorithms have made it a prominent family of methods for machine learning and artificial intelligence. This review determines the current state-of-the-art in using DL for solving forestry problems. Although DL has shown potential for various estimation tasks, the applications of DL to forestry are in their infancy. The main study line has related to comparing various Convolutional Neural Network (CNN) architectures between each other and against more shallow machine learning techniques. The main asset of DL is the possibility to internally learn multi-scale features without an explicit feature extraction step, which many people typically perceive as a black box approach. According to a comprehensive literature review, we identified challenges related to (1) acquiring sufficient amounts of representative and labelled training data, (2) difficulties to select suitable DL architecture and hyperparameterization among many methodological choices and (3) susceptibility to overlearn the training data and consequent risks related to the generalizability of the predictions, which can however be reduced by proper choices on the above. We recognized possibilities in building time-series prediction strategies upon Recurrent Neural Network architectures and, more generally, re-thinking forestry applications in terms of components inherent to DL. Nevertheless, DL applications remain data-driven, in contrast to being based on causal reasoning, and currently lack many best practices of conventional forestry modelling approaches. The benefits of DL depend on the application, and the practitioners are advised to ex ante subject their requirements to operational data availability, for example. By this review, we contribute to the technical discussion about the prospects of DL for forestry and shed light on properties that require attention from the practitioners.

No abstract

No abstract

Forest resource assessments based on multi-source and multi-temporal data have become more common. Therefore, enhancing the prediction capabilities of forestry dynamics by efficiently pooling and analyzing time-series and spatial sequential data is now more pivotal. Bayesian filtering and smoothing provide a well-defined formalism for the fusion or assimilation of various data. We ascertained how often the generic, standardized Bayesian framework is used in the scientific literature and whether such an approach is beneficial for forestry applications. A review of the literature showed that the use of Bayesian methods appears to be less common in forestry than in other disciplines, particularly remote sensing. Specifically, time-series analyses were found to favor <i>ad hoc</i> methods. Our review did not reveal strong numeric evidence for better performance by the various Bayesian approaches, but this result may be partly due to the challenge in comparing a variety of methods for different prediction tasks. We identified methodological challenges related to assimilating predictions of forest development; in particular, combining modeled growth with disturbances due to both forest operations and natural phenomena. Nevertheless, the Bayesian frameworks provide possibilities to efficiently combine and update prior and posterior predictive distributions and derive related uncertainty measures that appear under-utilized in forestry.