A random forest model for Norway spruce and Scots pine tree rings in Europe

Authors:
  • KBKamil Bielak
    Kamil Bielak
  • GBGediminas Brazaitis
    Gediminas Brazaitis
  • AGAksel Granhus
    Aksel Granhus
  • SHStig-Olof Holm
    Stig-Olof Holm
  • MLMagnus Löf
    Magnus Löf
  • AJAris Jansons
    Aris Jansons
  • MMMarek Metslaid
    Marek Metslaid
  • TNThomas Nord-Larsen
    Thomas Nord-Larsen
  • ANArne Nothdurft
    Arne Nothdurft
  • HPHans Pretzsch
    Hans Pretzsch
  • RRRicardo Ruiz-Peinado
    Ricardo Ruiz-Peinado
Open Access
Publisher website
Google Scholar
Abstract
Norway spruce, which is sensitive to drought, and Scots pine, which is drought-resistant, are two of the most significant conifer species in Europe. In mixed stands, they can utilize resources more efficiently than in pure stands, leading to higher yields and reduced risk. Tree ring research is often used to study their growth in response to complex environmental factors. Machine learning, though rarely applied to tree ring analysis, might be well suited for modelling these complex relations. Data from 22 triplets (1 mixed and two pure plots of Norway spruce and Scots pine) covering a temperature and precipitation gradient of 3.2-9.2°C and 613 to 1075 mm respectively, were used in this study. On each plot, trees were mapped and measured for dbh, height and height to the crown base. 4490 increment cores were collected and synchronized in the lab. A random forest model with relative DBH, age, competition, mixture and climate variables explained 76.4% of the variation and proved effective in describing ecological relationships.