Application of machine-learning methods in forest ecology: recent progress and future challenges

Abstract

Machine learning, an important branch of artificial intelligence, is increasingly being applied in sciences such as forest ecology. Here, we review and discuss three commonly used methods of machine learning (ML) including decision-tree learning, artificial neural network, and support vector machine and their applications in four different aspects of forest ecology over the last decade. These applications include: (i) species distribution models, (ii) carbon cycles, (iii) hazard assessment and prediction, and (iv) other applications in forest management. Although ML approaches are useful for classification, modeling, and prediction in forest ecology research, further expansion of ML technologies is limited by the lack of suitable data and the relatively “higher threshold” of applications. However, the combined use of multiple algorithms and improved communication and cooperation between ecological researchers and ML developers still present major challenges and tasks for the betterment of future ecological research. We suggest that future applications of ML in ecology will become an increasingly attractive tool for ecologists in the face of “big data” and that ecologists will gain access to more types of data such as sound and video in the near future, possibly opening new avenues of research in forest ecology.

Light detection and ranging explains diversity of plants, fungi, lichens, and bryophytes across multiple habitats and large geographic extent

“…Denser point clouds and terrestrial laser scanning combined with new machine learning techniques (Liu et al. ), spectral information (Lausch et al. ), and emerging environmental DNA technologies (Bush et al.…”

“…This renders LIDAR alone unsuitable for recording soil and leaf chemistry, but there could be differences in terrain and vegetation structure across habitat types mirroring these factors of which we are unaware. Denser point clouds and terrestrial laser scanning combined with new machine learning techniques (Liu et al 2018), spectral information (Lausch et al 2016), and emerging environmental DNA technologies (Bush et al 2017) might potentially remedy this situation in the future. Notably, advances within the field of imaging spectroscopy shows promising potential for mitigating the lack of information on for instance nutrient balance and soil pH, characterizing LIDAR measures.…”

Estimating Ecosystem Respiration in the Grasslands of Northern China Using Machine Learning: Model Evaluation and Comparison

“…Second, some large RE values may be caused by certain atmospheric turbulent events rather than real ecological processes [66]. ML models are well known for having the ability to automatically learn complex nonlinear relationships from input data [49,67]. Given that RE is difficult to constrain due to our limited understanding of the complex interactions among physical, chemical, and biological processes [5,14], ML models can help us accurately quantify the spatiotemporal variation in RE at regional or global scales.…”

LIDAR explains diversity of plants, fungi, lichens and bryophytes across multiple habitats and large geographic extent

“…This renders LIDAR alone unsuitable for recording soil and leaf chemistry, but there could be differences in terrain and vegetation structure across habitat types mirroring these factors of which we are unaware. Denser point clouds combined with new machine learning techniques (Liu et al 2018) and perhaps spectral information (Lausch et al 2016) could potentially remedy this situation in the future.…”