Skip to main content
Dryad logo

Tree ring evidence of rapid development of drunken forest induced by permafrost warming


Fujii, Kazumichi (2022), Tree ring evidence of rapid development of drunken forest induced by permafrost warming, Dryad, Dataset,


Climate warming risks permafrost degradation causing “drunken” trees toppling along with thaw, but drunken forest also occurs with hummocky ground on continuous permafrost. However, no evidence has previously clarified whether tree leaning is activated by climate warming or is part of a natural hummock formation process. We test hypothesis that climate warming accelerates development of drunken forest. Results showed that trees’ leaning synchronizes with development of soil hummocks and that tree leaning events are recorded in darkened lignin-rich tree rings. Cell morphology in tree rings shows that tree leaning is caused not by loosening of soil foundation in warming summer but by winter frost heaving. Activities of tree leaning and hummock formation increase with increasing air temperature. Reconstruction of drunken forest development using tree ring record suggests that hummock formation has shifted from periodic events until 1960 to continuous mound rising. Recent warming increases hummock development and soil potentials for carbon storage on continuous permafrost, but further warming risks permafrost degradation and hummock collapse with carbon loss.


Field site. We investigated 50 mounds in the Mackenzie upland area between Inuvik and Tsiigehtchic, which are located in the Northwest Territories, Canada (N68°03’, W133°30’; Tables S1, S2). The soil is underlain by continuous permafrost that has developed during ice ages with no coverage by Laurentide ice sheet. To reconstruct the development of soil hummocks and drunken trees, we analyzed the tree-ring records of black spruce (Picea mariana L.) and soil microrelief. Inuvik has a subarctic climate; the mean annual air temperature is –8.8ºC. The soil surface is covered by lichen (Cladonia mitis L. and Cladonia stellaris L.)-covered mounds and moss (Pleurozium schreberii L., Hylocomium splendens L., and Sphagnum fuscum L.)-covered depressions.

Experimental design. We collected a total of 50 tree disks from 50 hummocks at two plots with contrasting permafrost table depths and mound sizes and with clayey and sandy soils (Table S2). Clayey soils are derived from a mixture of fluvial clayey sediments and fine-grained glacial till (27–37% clay and 22–30% sand), while sandy soils are derived from glaciofluvial sands (8–12% clay and 73–75% sand). To analyze the changes in the activities of mound growth and tree leaning before and after warming, we compared young trees born after 1910 and mature trees born before 1860 (>50 and <100 years old in 1961, respectively).

Soil and microrelief survey. We measured the hummocky microrelief of the soil surface (relative elevation) and the maximum depths from the thawed soil to the permafrost table in August (active layer thickness) by excavating soil pits at the mound and depression sides surrounding the tree stumps (at a distance of 20 cm from the trunk). In the well-developed hummocky microrelief, the depths from the active soil layer to the permafrost table varied widely between the mound and depression sides surrounding the tree stumps. The ratios of the maximum and minimum active layer depths to the permafrost table were calculated and defined as an indicator of hummock size (Table S2).

Tree-disk sampling and observation. Tree-disk samples were collected by cutting the tree stems perpendicular to the stem axis at a height of 30 cm above the ground, where the reaction wood formation is at its maximum. The wood tissues were observed using optical microscopy after staining lignin and non-lignin tissues (protein and cellulose) using safranin and fast green dyes, respectively.

Dendrochronological methods. We applied dendrochronological analysis to estimate disturbance events (soil hummock formation)[12]. The annual widths of tree rings were measured using a stereomicroscope (MZ6; Leica, Solms, Germany) with a system for tree-ring measurement from Velmex Inc. (Broomfield, NY, USA) in two directions on the major axis, including the maximum wood radius and two perpendicular directions on the minor axis of the wood disk (a total of four radiuses). The tree-ring data (Fig. S3) were visually cross-dated and statistically verified using the COFECHA program. Due to high density of tree rings (10 tree rings in 1 mm), 28 tree disks of 50 mounds (38 mounds in 10 clayey soil sites, 12 mounds in 6 sandy soil sites) satisfied dendrochronological quality check (cross-dating). 28 mounds including 22 clayey soil mounds on fine-grained glacial till and 6 sandy mounds on glaciofluvial sands were used for reconstruction of hummock formation.

Detection of tree leaning and hummock formation events. The annual intensity of tree leaning was defined as the ratios of maximum annual tree-ring width to average annual tree-ring width (Data S1) and calculated by dividing the maximum annual tree-ring width by the average annual ring width of the other three directions (±90º, 180º relative to maximum tree-ring width). The peaks when the annual intensity of tree leaning exceeded 2 were defined and counted as active tree-leaning events, during which the trees produce round cells (reaction wood) (percentage of active tree-leaning events in Fig. S4b). After this filtering step, the peaks when the annual intensity of tree leaning exceeded that of the previous year were counted as new events of mound rising (percentage of active mounds in Fig. 4b).

Reconstruction of soil hummock formation. Among fifty tree samples, trees leaning on the mound edges (n = 28) were used to reconstruct soil hummock development, and tree-ring widths were summed up to calculate the wood increment curves (Data S2). Then, the extent of reaction wood formation was calculated from the ratios of the maximum wood radius to the average wood radius to reconstruct soil hummock development (Fig. 1e). The cases in which the extent of reaction wood formation exceeded 1 indicated that trees were leaning towards the orientation of the mound slope.

Statistics. The mean values of the extent of reaction wood formation and the annual intensity of tree leaning were used for statistics. The meteorological data (mean annual air temperature) were obtained from the nearest meteorological stations (Inuvik; 1958–2014), and the deficit data (1911–1958, 1971) were complemented by the values estimated using the simple linear regression with the data of the nearest station (Fort Good Hope; 1911–2014). Arctic-wide summer air temperature anomalies (1820–1974) were inferred from the existing tree-ring record. To analyze the extent of reaction wood formation or annual intensity of tree leaning or mound growth before and after warming, the differences in the slopes of the linear regression equations were compared using t-tests between the young and mature trees, and the correlations between the annual intensity of tree leaning and air temperature were tested using simple linear regression. Statistics were performed using the Sigma Plot software (14.0 SPSS Inc., Japan).