Tropical cyclones moving into boreal forests: Relationships between disturbance areas and environmental drivers

Sci Total Environ. 2022 Oct 20:844:156931. doi: 10.1016/j.scitotenv.2022.156931. Epub 2022 Jun 27.

Abstract

Tropical cyclones (TCs) are common disturbance agents in tropical and subtropical latitudes. With global warming, TCs began to move to northern latitudes, with devastating effects on boreal forests. However, it remains unclear where and when these extraordinary events occur and how they affect forest structure and ecosystem functioning. Hence knowing which geomorphological features, landforms, and forest types are most susceptible to severe wind disturbance is vital to better predict the future impacts of intensifying tropical cyclones on boreal forests. In October 2015, catastrophic TC Dujuan hit the island of Sakhalin in the Russian Far East. With a wind speed of 63 m·s-1, it became the strongest wind recorded in Sakhalin, damaging >42,000 ha of native forests with different levels of severity. We used high-resolution RGB satellite images, DEM-derived geomorphological patterns, and the U-Net-like convolutional neural network to quantify the damaged area in specific landform, forest type, and windthrow patch size categories. We found that large gaps (>1 ha) represent >40 % of the damaged area while small gaps (<0.1 ha) only 20 %. The recorded canopy gaps are very large for the southern boreal forest. We found that the aspect (slope exposure) is the most important in explaining the damaged area, followed by canopy closure and landform type. Closed-canopy coniferous forests on steep, west-facing slopes (typical of convex reliefs such as ridges, spurs, and peaks) are at a much higher risk of being disturbed by TCs than open-canopy mountain birch forests or coniferous forests and broadleaved riparian forests in concave reliefs such as valley bottoms. We suggest that the projected ongoing poleward migration of TCs will lead to an unprecedentedly large area of disturbed forest, which results in complex changes in forest dynamics and ecosystem functioning. Our findings are crucial for the development of mitigation and adaptation strategies under future changes in TC activity.

Keywords: Convolutional neural network; Forest disturbance risk; Landform analysis; Satellite data.

MeSH terms

  • Cyclonic Storms*
  • Ecosystem
  • Forests
  • Taiga*
  • Trees