Geomorphology 253 (2016) 189198

Contents lists available at ScienceDirect

Geomorphology
journal homepage: www.elsevier.com/locate/geomorph

How different are the results acquired from mathematical and subjective
methods in dendrogeomorphology? Insights from landslide movements
Karel ilhn
Department of Physical Geography and Geoecology, Faculty of Science, University of Ostrava, Chittussiho 10, 710 00 Ostrava, Czech Republic

a r t i c l e

i n f o

Article history:
Received 9 June 2015
Received in revised form 19 October 2015
Accepted 20 October 2015
Available online 21 October 2015
Keywords:
Dendrogeomorphology
Landslide
Tree sensitivity
The Orlick hory foreeld

a b s t r a c t
Knowledge of past landslide activity is crucial for understanding landslide behaviour and for modelling potential
future landslide occurrence. Dendrogeomorphic approaches represent the most precise methods of landslide
dating (where trees annually create tree-rings in the timescale of up to several hundred years). Despite the advantages of these methods, many open questions remain. One of the less researched uncertainties, and the
focus of this study, is the impact of two common methods of geomorphic signal extraction on the spatial and temporal results of landslide reconstruction. In total, 93 Norway spruce (Picea abies (L.) Karst.) trees were sampled at
one landslide location dominated by block-type movements in the foreeld of the Orlick hory Mts., Bohemian
Massif. Landslide signals were examined by the classical subjective method based on reaction (compression)
wood analysis and by a numerical method based on eccentric growth analysis. The chronology of landslide movements obtained by the mathematical method resulted in twice the number of events detected compared to the
subjective method. This nding indicates that eccentric growth is a more accurate indicator for landslide movements than the classical analysis of reaction wood. The reconstructed spatial activity of landslide movements
shows a similar distribution of recurrence intervals (Ri) for both methods. The differences (maximally 30% of
the total Ri ranges) in results obtained by both methods may be caused by differences in the ability of trees to
react to tilting of their stems by a specic growth response (reaction wood formation or eccentric growth). Finally, the ability of trees to record tilting events (by both growth responses) in their tree-ring series was analysed for
different decades of tree life. The highest sensitivity to external tilting events occurred at tree ages from 70 to
80 years for reaction wood formation and from 80 to 90 years for eccentric growth response. This means that
the ability of P. abies to record geomorphic signals varies with not only eccentric growth responses but also
with age.
2015 Elsevier B.V. All rights reserved.

1. Introduction
Landslides are very broadly extended geomorphic processes that
often occur in populated areas (Gutirrez et al., 2010; Van Den
Eeckhaut and Hervs, 2012), where they can present a signicant natural hazard. Therefore, knowledge of the chronology of past landslide activity is important for modelling their future potential occurrence
(Borgatti and Soldati, 2010). Moreover, data regarding the spatial distribution of landslide activity are necessary for safe land-use planning. Unfortunately, direct information regarding past landslide activity is often
scarce, particularly in remote areas. Landslides inventory and mapping
is required to better understand where and when they may occur. Nevertheless, as the dates of triggering of past events are often uncertain,
the application of absolute dating methods becomes necessary (Lang
et al., 1999).
Dendrogeomorphic methods represent the most precise way of dating past landslide activity in areas with forest cover located in

E-mail address: karel.silhan@osu.cz.

http://dx.doi.org/10.1016/j.geomorph.2015.10.012
0169-555X/ 2015 Elsevier B.V. All rights reserved.

temperate zones in the timescale up to several hundred years (Corona

et al., 2014), although some limitations exist. For example, a limitation
of dendrogeomorphic landslide dating is the age of the trees studied.
Since the introduction of dendrogeomorphic analysis, coniferous trees
have been favoured (Alestalo, 1971; Shroder, 1978; Bgin and Filion,
1988; Lopez-Saez et al., 2012a) probably due to their good visibility of
tree-rings, easy sampling, or good visible growth responses. Nevertheless, many landslide areas are populated by broad-leaved trees, and
studies investigating broad-leaved trees have become more frequent
in recent decades (Fantucci and Sorriso-Valvo, 1999; Guida et al.,
2008; Stefanini, 2004; ilhn et al., 2014). According to Shroder
(1978), the most frequent landslide-associated event affecting trees is
the tilting of their stems; nevertheless, damage to the roots or burying
of the stem base can occur as well (Stoffel and Corona, 2014). For coniferous trees, the dominant inuence on growth following tree tilting is
the formation of reaction (compression) wood on the lower side of
the stem. The cells of compression wood have thicker and more rounded walls of the tracheids, which is the reason why tree rings containing
compression wood are macroscopically identiable (Timell, 1986). Nevertheless, coniferous trees do not create reaction wood in all cases of

190

K. ilhn / Geomorphology 253 (2016) 189198

stem tilting, and a potential minimum tilting value for its creation had
not been dened until recently (ilhn and Stoffel, 2015). Moreover, potential delay in the compression wood formation after the tilting event
can occur (Shroder, 1978). However, compression wood identication
in tree ring series is a subjective approach, and results can be inuenced
by a researcher error. Additionally, broad-leaved trees create reaction
(tension) wood on the upper side of their tilted stems. Unfortunately,
this type of reaction wood is practically unidentiable at a macroscopic
scale (Westing, 1965). In spite of potential limitations regarding compression wood analysis, many recent studies of landslide dating have
been based on this subjective approach (Lopez-Saez et al., 2012a,b;
ilhn et al., 2012, 2013). Another growth reaction to stem tilting is eccentric growth. Tilted coniferous trees start to produce wide tree-rings
on the lower side of the stem, whereas the upper side of the stem is
dominated by suppressed growth (Braam et al., 1987). The position of
wide tree-rings and suppressed growth is usually opposite in the case
of broad-leaved trees. Identication of a tilting event based on dating
of this reaction has a mathematical base, as the eccentricity is based
on the calculation from measured tree-ring widths, and subjective errors can be excluded. Unfortunately, some limitations even for the eccentric growth analysis exist. The eccentric growth can occur even as a
response to non-geomorphic events (e.g. strong wind). Due to the
aforementioned difculties with identication of tension wood, the
analysis of tree-ring series of broad-leaved trees is dominantly focused
on eccentric growth as a reaction to stem tilting (Van Den Eeckhaut

et al., 2009; ilhn et al., 2014). To date, the resulting effect of these
two different approaches (reaction wood vs. eccentric growth) on reconstructed chronologies or spatial reconstruction is not known. However, answering this question is crucial for increasing the accuracy of
past events identication. Therefore, as noted above,
dendrogeomorphology is characterized by some biases (while it is
very accurate), so that a comparison of different methods is highly
important.
Another open question is the changing sensitivity of trees to geomorphic events with increasing age. For example, ilhn et al. (2013)
found the highest sensitivity of Crimean pine (Pinus nigra ssp.
pallasiana) to rockfall events at ages of 80 to 90 years. The highest sensitivity of P. nigra to landslides was found to be present in two phases,
between the ages of 40 to 60 years and 120 to 130 years (ilhn and
Stoffel, 2015) in the dependence on the dominant growth response (reaction wood vs. eccentric growth). However, for Norway spruce (Picea
abies (L.) Karst.), a very common tree species occupying landslides in
the northern hemisphere, the most suitable ages for recording tilting
events have not been determined yet.
Based on the aforementioned open questions, the objective of this
study is to show the effects of using subjective and mathematical approaches (reaction wood and eccentric growth, respectively) on different aspects of dendrogeomorphic reconstruction of landslide
movements. This objective will be achieved using the following strategies: (i) evaluating landslide event chronologies obtained using both

Fig. 1. Location, geomorphology, and geology of the studied area. A location of the study area in the Czech Republic, B geomorphic map of the studied landslide area (1 position of the
topographic prole in Fig. 2; 2 sampled trees; 3 border of studied domains; 4 gully; 5 debris talus; 6 partial landslide block; 7 alluvial fan; 8 main scarp; 9 bedrock outcrop; 10 accumulation lobe), C geology of the studied landslide surroundings (1 uvial sands and gravel, 2 marlstone and sandstone, 3 marlstone, 4 limestone claystones, 5
loess, 6 fault), and D LiDAR image of hillslope affected by the studied landslide.

K. ilhn / Geomorphology 253 (2016) 189198

191

Fig. 2. Topographic proles, geomorphic features, and disturbed trees on the landslide area. A main scarp with bedrock outcrop; B surface of subsided landslide block; C upslope
tilted trees on slightly rotated landslide block; D disrupted marl layers at the foot of the landslide area; E deformed trees on the steep parts of the landslide area.

approaches, (ii) comparing the effect of each approach on the reconstructed spatial activity of landslides, and (iii) assessing the changes in
sensitivity to tilting events of P. abies with increasing age for both
approaches.
2. Study area
For the purpose of this study, a landslide area dominated by blocktype movements was selected. The studied landslide area is located in
the foreland of the Orlick hory Mountains (5006 N., 1612 E.;
~ 300 m asl; Fig. 1A). The probable pre-conditioning factors include
the climatic conditions and geological structure of the hillslope. The climate of the region can be characterized as moderately humid with a
mean annual temperature of 78 C and a mean annual precipitation
of 550700 mm (Quitt, 1971; Tolasz et al., 2007). The landslide area developed in the sub-horizontal layers of limestone marls (ech, 2001) on
a hillslope located on the left bank of the Divok Orlice River (Fig. 1C, D).

The existence of faults parallel with the slope direction and the past undercutting of the hillslope by the Divok Orlice river can serve as potential preparatory factors. The age of undercutting, which is evidenced by
the presence of an old river branch at the foot of the hillslope in some
places, is not known up to now.
The landslide area is characterized by a frontal shape with a width of
~1.2 km and a NE orientation. The maximum vertical distance is approximately 75 m. The landslide area can be dened as block-type movements (according to Cruden and Varnes, 1996) with several levels of
mildly inclined, subsided blocks separated by steep steps (Fig. 2E). The
northern part of the landslide area is represented by one subsided
block with several steps decreasing in height to the north (Figs. 1C
and 2B). The middle part of the landslide area has up to three levels of
subsided blocks (Figs. 1C and 2C) and, including its horizontal movements, shows the greatest elongation in this section. The southern
part of the landslide area is dissected by individual blocks separated
by a network of gullies. At the mouth of the gullies, two long alluvial

K. ilhn / Geomorphology 253 (2016) 189198

192

Fig. 3. Methods used for determining landslide movement. A onset of reaction (compression) wood in coniferous trees; B extraction of the geomorphic signal based on abrupt changes
in the eccentricity of tree rings.

fans (up to 100 m long) were created (Fig. 1B). Several man-made structures such as forest roads have been built on parts of the slope. The
fronts of some of the subsided blocks are compressively deformed
(Fig. 2D), resulting in the formation of accumulation lobes. The main
scarp is very steep (more than 35) and has bedrock outcrops in several
locations (Fig. 2A). Debris tali have formed below some of the outcrops.
There are several linear subsidence features that likely follow fault lines
on the plateau above the main scarp. The landslide area is covered by a
forest dominated by Littleleaf Linden (Tilia cordata Mill.), European
Beech (Fagus sylvatica L.), Norway Spruce (P. abies), European Larch
(Larix decidua Mill.), and Scots Pine (Pinus sylvestris L.).
Because landslide reactivation often impacts only a limited part
of the whole landslide area (Floris and Bozzano, 2008; Lopez-Saez
et al., 2013; ilhn et al., 2013), the studied landslide area was divided
into ve quasi-homogenous domains (Fig. 1B). This approach is
particularly suitable for dendrogeomorphic research (ilhn, 2012).
Dendrogeomorphic research focused on the following parts of landslide
area (not described from upper to lower parts, but focusing on the
analysed areas). Domain I is the northern part with one subsided long
block; domain II is the middle part with three vertical levels of blocks;
domain III consists of a connective level of blocks bordering domain II
from the south; domain IV consists of two vertical levels of blocks
Table 1
The number and age of P. abies in individual landslide domains.
Domain

Trees (number)

Maximal age

Minimal age

Mean age (stdev)

I
II
III
IV
V
Total

33
10
12
27
11
93

142
108
125
133
136
142

16
32
35
68
66
16

108.8 (31.9)
44.3 (28.1)
91.3 (33.5)
108.4 (19.3)
115.4 (23.5)
101.5 (32.9)

with a large debris talus in the middle of the domain; and domain V is
the southern part of the landslide area with blocks separated by a network of gullies.
3. Methods
A total of 372 increment cores were extracted from 93 disturbed
P. abies trees in an area of block-type movements. Tree-ring series
were analysed using two different methods: mathematical-based analysis of eccentric growth and the subjective method of reaction wood
identication. Results from both methods were analysed separately to
assess their inuence on the dendrogeomorphic analysis of a landslide
area.
First, detailed eld geomorphic mapping (at 1:500 scale) was performed. The mapping was produced using eld investigation and a digital elevation model (DEM) created from LiDAR data. Along with other
map features such as landslide blocks, rocky scarps, gullies, and debris
tali, the positions of the disturbed (tilted) trees selected for sampling
were recorded.
For the purpose of this study, P. abies were sampled. Trees were selected by two criteria: their position on the geomorphic map with an effort to evenly distribute the sampled trees within the landslide area
(Corona et al., 2014) and visible external disturbance caused by landslide movements (tilted or bent stems). Trees were sampled preferably
on the surface of blocks with a variety of topographical surface slope angles. The steepest parts of the main scarp were ignored. Trees were sampled using a Pressler increment borer (max. length: 40 cm; diameter:
0.5 cm) at the height of maximum stem bending. Four increment
cores were extracted from each tree perpendicular to each other (one
from the lower side of the stem, one from the upper side of the stem,
and two parallel to the slope). Additionally, 30 individual specimens
of P. abies were sampled (two cores per tree) from a stable position

K. ilhn / Geomorphology 253 (2016) 189198

193

Fig. 4. Chronology of landslide movements in individual domains and for the total landslide area as reconstructed from reaction wood analysis.

outside the landslide area (stable plateau above the main scarp) to create a reference chronology reecting only the effects of climatic conditions or occasional insect outbreaks on tree growth (Cook and
Kairiukstis, 1990). The reference chronology was created using a
double-detrending procedure in the Arstan software (Cook, 1983).
The negative exponential curve (or linear regression) was used in the
rst step, and the cubic smoothing spline function was applied in the
second step (Holmes, 1994).
All samples were processed following a standard procedure
(e.g., Stoffel and Bollschweiler, 2008, 2009). The individual steps of
this procedure included sample drying, gluing samples into the woody
supports, sanding, tree ring counting, and measuring ring widths
using dendrochronological TimeTable and PAST4 software (VIAS,
2005). The increment curves were cross-dated with the reference chronology to identify and correct possible false or missing tree rings.
Two dating approaches of tilting events were applied. First, a subjective analysis of reaction (compression) wood presence (Fig. 3A) was
carried out on all four increment cores from each tree using a binocular
microscope. Reaction wood can be found in two forms (pronounced and
mild). Tracheids with compression wood anatomy (thicker and

rounded walls) can be visible in the latewood part of a ring in the case
of mild reaction wood, whereas pronounced reaction wood contains
such tracheids throughout the entire width of the ring (Lopez-Saez
et al., 2012a). The onset of both reaction wood types was considered
as a tilting signal. Second, a mathematical approach, based on tree ring
eccentricity calculation and analysis was applied. The determination of
a tree tilting event was based on eccentric growth using the approach
published by ilhn et al. (2014) and ilhn and Stoffel (2015). Eccentricity was analysed by comparison of two tree ring series growing perpendicular to each other (one from the lower side of the stem and the
second perpendicular to the rst). This strategy reduces the risk of missing a ring while still providing a chance at calculating tree-ring eccentricity (Shroder, 1978). In the rst step, tree-ring eccentricity (e) was
calculated according to Braam et al. (1987):
e

ac
ac

where a is the ring width from the upper side of the stem and c is the
ring width perpendicular to a. Eccentricity values were separated into

Fig. 5. Chronology of landslide movements in individual domains and for the total landslide area as reconstructed from eccentric growth analysis.

K. ilhn / Geomorphology 253 (2016) 189198

194

Fig. 6. Proportion of dated landslide events based on data from reaction wood analysis and eccentric growth analysis in individual landslide domains.

three intervals (marked as 0, 1, and 2), according to Fig. 3B. The occurrence of specic interval patterns (e.g., 0022; Fig. 3) allows determination of the intensity of the tilting signal. For the purpose of this study,
this pattern was used only to determine whether a tilting signal occurred or not.
Landslide activity in individual domains was expressed using the
standard event-response index (It) according to Shroder (1978):
It

X
Rt =
Nt  100%

where Rt is the number of trees with a tilting signal in year t and Nt is the
number of all sampled trees living in year t. The It values were calculated
individually for each domain and for each tree group. To dene a landslide event, at least two trees had to show a landslide signal and the It
value had to be at least 7.5%.
The frequency of landslide events recorded by a single tree was
expressed by the recurrence interval (Ri), calculated as:
X
Ri At =
LSt

Fig. 7. Spatial distribution of mean recurrence interval between two landslide movements, reconstructed based on data from reaction wood analysis and eccentric growth analysis.

K. ilhn / Geomorphology 253 (2016) 189198

195

Fig. 8. Difference (absolute value) of the recurrence interval maps, reconstructed based on data from reaction wood analysis and eccentric growth analysis.

where LSt is the number of tilting signals in the tree rings of a tree and At
is the total age of the tree. The Ri values from each tree were spatially interpolated using the ordinary kriging model due to irregular spatial pattern of sampled trees locations, with the grid cell size of 1 m due to
minimal distance between trees, using the Surfer 8 software (Golden
Software, 2002). Resulting grids based on reaction wood data and eccentric growth data were subsequently subtracted from each other to
determine places on the landslide area with the highest differences between results obtained by the two different approaches.
Finally, based on the assumption of Stoffel and Corona (2014), an attempt was made to express changes in a tree ability to record tilting
events with increasing age. The number of tilting events recorded by
each tree was summarized for individual decades of tree live and
expressed as the number of recorded events per tree in each decade.
This evaluation was based on data from both reaction wood and eccentric growth analysis.
4. Results
4.1. Number and age of sampled trees and growth disturbances
From the entire landslide area, 93 individual P. abies trees were sampled. In total, 372 increment cores were extracted. The highest number
of trees was sampled in domain I (33 trees) and the fewest number of
trees was sampled in domain II (10 trees). The mean age of the sampled
trees was 101.5 years (stdev = 32.9 years); the oldest and youngest
trees were 142 and 16 years old, respectively; the oldest trees were
sampled in domain V (mean = 115.4 years), whereas the youngest
were sampled in domain II (mean = 44.3 years). Details about the number and age of the sampled trees in the individual domains are listed in
Table 1.
Among the analysed tree-ring series of P. abies, a total of 118 indications of reaction (compression wood) were identied. On the other
hand, based on the analysis of eccentric growth, a total of 149 tiltingevent signals were identied. Very strong event signals were identied

in 24 cases, strong signals in 28 cases, weak signals in 34 cases, and very

weak signals in 63 cases.

4.2. Chronology of landslide events

The chronology of landslide events based on reaction wood analysis
obtained from the tree-ring series of P. abies contained 20 total events
across all domains (Fig. 4). Twelve years are identied as years of
landsliding, as some events occurred in the same year. The oldest
event occurred in domain I in 1909 and the youngest event occurred
in the same domain in 2002. The highest number of recorded events occurred in domain V (six events), and only three events were recorded in
domains II and IV. The highest mean It value reached events in domain II
(26.7%), while the lowest It value came from events in domain I (9.5%).
The rst period of frequent landslide activity occurred between 1909
and 1932. This period was followed by 36 years without dated activity.
Between 1968 and 2002, nine event years were recorded, while during
the last 12 years no events were found.
Eccentric growth analysis discovered 39 total landslide events across
all domains (Fig. 5). These events occurred in 24 landslide years. Nine of
the 12 event years reconstructed from reaction wood analysis were also
recorded by eccentric growth analysis (1932, 1968 and 1977 events
were not recorded). Moreover, eccentric growth analysis enabled the
dating of fteen additional landslide years (1943, 1965, 1972, 1974,
1980, 1981, 1985, 1988, 1989, 2001, 2005, 2006, 2009, 2011, and
2012). The highest number of events recorded occurred in domain V
(10 events), whereas only ve events were dated in domain IV. As in
the reaction-wood-based chronology, the oldest event was found in domain I and dated to 1909, while the youngest events occurred in domains I and II in 2012. The highest mean It value was found in domain
III (25.7%) and the lowest It value occurred in domain IV (10.7%). The
chronology obtained from using reaction wood data demonstrates
two periods of higher landslide activity very similar to the chronology
based on eccentric growth data (19091943 and 19652012).

Fig. 9. Sensitivity of P. abies to tilting events based on reaction wood analysis expressed as: A number of dated tilting events in individual decades of tree life; B number of dated tilting
events per one tree in individual decades of tree life.

196

K. ilhn / Geomorphology 253 (2016) 189198

Fig. 10. Sensitivity of P. abies to tilting events based on eccentric growth analysis expressed as: A number of dated tilting events in individual decades of tree life; B number of dated
tilting events per one tree in individual decades of tree life, and proportion of signal intensities.

Thirty ve per cent of all dated landslide years were discovered by

both methods simultaneously (not only by one method). However,
54% of event years were reconstructed using eccentric growth analysis
only. Using reaction wood analysis alone enabled reconstruction of
12% extra event years. The benets of using both methods for
reconstructing landslide events in individual domains are shown in
Fig. 6.

recorded tilting events is shifted by about one decade in comparison

with data based on reaction wood analysis (i.e., to ages of 80 to 90
years). After this period, the number of events decreases. In general,
an above-average number of recorded tilting events based on eccentric
growth analysis was observed in trees aged 60 to 120 years (Fig. 10A).
Even the proportion of very strong and strong tilting signals is above average during this period with a peak at tree ages of 90 to 100 years
(Fig. 10B).

4.3. Spatial landslide activity

5. Discussion
The spatial distribution of landslide movements was expressed by
the interpolated distribution of the trees Ri values. From data based on
reaction wood analysis (Fig. 7), it is clearly visible which parts of the
landslide area are most active. Generally, the most active part (with
the shortest periods between two events) occurs in the centre of the
landslide area in domains II and III with three vertical levels of subsided
blocks. This statement is supported by the results of the eccentric
growth analysis (Fig. 7). These domains contain the highest number of
vertical block levels and the greatest landslide area length. Moreover,
surface subsidence along the fault line above the main scarp is greatest
in this area (up to 40 m). The highest reconstructed landslide activity reveals dynamic landslide area development in this location. Frequent
landslide movements broke the edge of the marl plateau into several
levels of blocks and moved these blocks the greatest distance from
their original locations (up to 100 m in comparison with surrounding
stable slope without landslide movements). The area of the next highest
activity is located at the southern border of the landslide area in domain
V. Generally, low activity was recorded in domain IV, while domain I is
dominated by closely spaced changes in activity frequency.
Recurrence maps obtained from both approaches show very similar
results (coefcient of correlation = 0.89). Nevertheless, subtraction of
the two grids revealed locations with the highest differences between
the spatial distributions of both datasets. According to Fig. 8, the maximum difference between the two grids is no higher than ~ 40 years,
which is ~30% of the maximum Ri in either grid. This observation conrms the expected strong agreement between the grids. The highest differences are located at the southern border of the studied landslide area
in domain V, in the central part of domain IV on the debris talus, and in
several locations within domain I.
4.4. The ability of P. abies to record tilting events in tree rings
Based on reaction wood analysis, the number of tilting events recorded during individual decades of tree life initially increases with
age. The highest number of identied tilting events was recorded at
tree ages of 60 to 80 years (Fig. 9A). After the age of 80 years, the number of reaction-wood-based tilting events abruptly decreases. The same
results are clearly visible even when the number of recorded tilting
events is weighed by the number of trees living in a particular decade
(Fig. 9B).
The number of dated tilting events based on eccentric growth analysis initially increases with age as well. However, the highest number of

5.1. Chronological consequences of mathematical and subjective method

use
As the landslides dominated by marly rocks are sensitive to
precipitation totals (Van Asch et al., 1999) the verication of
dendrogeomorphic dating can be done by comparison with extreme
rainfall events. The dating accuracy can be veried by the occurrence
of heavy precipitation events in most of the dated landslide years
(e.g., 1909, 1977, 1985, 1997, 1998, 2000; tekl et al., 2001) in the
study region. The precipitation total in these years exceeded at least
100 mm during one day.
Chronologies reconstructed from each method provide clear evidence of signicant differences resulting from the use of the two
methods. The chronology obtained from data based on eccentric growth
analysis indicated twice the number of landslide events as data based on
reaction wood analysis. Although some of the events were identied by
both methods, use of eccentric growth analysis alone identied 54% of
all events. ilhn and Stoffel (2015) had already demonstrated a higher
sensitivity of eccentric growth analysis compared to using macroscopic
observations of reaction (compression) wood for P. nigra ssp. pallasiana.
Although a different tree species was used in this study, the results presented here fully support the conclusions of these authors. Differences
in the ability of trees to form compression wood vs. eccentric growth
are potentially caused by different tilting thresholds of tree stems required to initiate growth reaction. Based on the results of this study,
P. abies can form eccentric growth at lower values of stem inclination
than necessary for the formation of reaction wood. In addition, the sensitivity of eccentric growth analysis is higher, even though the reaction
wood analysis was carried out on four radii of the stem, as a tree can be
tilted several times in different directions (although it is common to analyse only two radii; Corominas and Moya, 1999; Carrara and O'Neill,
2003; Stefanini, 2004). Accordingly, further research should focus on
the relationship between the magnitude of stem deformation and the
intensity of eccentric growth/reaction wood formation.
5.2. Consequences of mathematical and subjective method use on spatial
reconstruction
Both methods of geomorphic signal extraction provided very similar
spatial distributions of recurrence intervals (Ri). Identical maximum
ranges of reconstructed Ri depend on the maximum age of the trees.

K. ilhn / Geomorphology 253 (2016) 189198

Tree age limits do not allow for the reconstruction of higher Ri values;
this means that the real period between two landslide events in the
less active parts of the landslide area could be higher than what was
found from the reconstructed chronology.
Based on the assumption that different thresholds of stem tilting
produce different growth reactions (reaction wood vs. eccentric
growth), it can be supposed that the greatest differences between results obtained by the different methods can be found in the trees with
low stem inclination. Data based on eccentric growth in these trees
should show higher frequency of tilting events in these cases. Accordingly, areas with signicant differences between both grids probably
correspond to locations characterized by low stem inclination. This assumption should be further investigated by research focusing on
dendrometric evidence.
5.3. Ability of P. abies to record tilting signals
One of the aims of this study was to analyse the changing ability of
P. abies trees to record tilting signals in their tree-ring series. This question is quite new for landslide research, as the inuencing geomorphic
factors have not been intensively studied (ilhn and Stoffel, 2015).
However, solving this open question is crucial for establishing a generally recommended sampling strategy for a wide range of tree ages. Initial results are known only for impacts of debris ows on P. nigra
(ilhn et al., 2015), rockfalls on P. nigra (ilhn et al., 2013), and for
landslides on P. nigra and F. sylvatica (ilhn and Stoffel, 2015). In this
study, the highest sensitivity of P. abies to landslide movements was recorded at tree ages of 60 to 80 years for reaction wood formation and
ages 80 to 90 years for the formation of eccentric growth. These results
are quite different from the results of ilhn and Stoffel (2015). P. abies
trees seem to be better able to allocate their wood material along stem
circumferences during ages of limited ability to create reaction wood,
nevertheless the next research of this difference reason should be
realised in the future. The result is a shift of the highest sensitivity to
stem tilting from the formation of reaction wood toward eccentric
growth. An open question remains the effect of changes in tree size
(stem diameter) on the sensitivity of a tree to landslide movements.
6. Conclusion
This study highlights how the choice of a dendrogeomorphic method can imply some biases on the results. For this research, 93 trees of
P. abies were sampled from one landslide area. The onsets of reaction
(compression) wood formation and abrupt changes in tree-ring eccentricity were used as landslide signals.
The chronology of landslide movements based on eccentric growth
analysis indicated nearly twice the number of events compared to the
chronology based on reaction wood analysis. This means that the use
of eccentric growth is a much more sensitive approach to landslide dating than the classical analysis of reaction (compression) wood in the
tree-ring series of P. abies. Nevertheless, each method can identify
only a part of the landsliding events, and not necessarily the same
ones. Therefore, the combination of both methods for landslide research
can be recommended. The assumption that trees react in different ways
to various values of stem inclination requires additional study, as it can
have consequences on the spatial reconstruction of landslide activity.
The sensitivity of P. abies to landslide movements via its ability to
create reaction wood or eccentric growth as response to stem tilting
changes with increasing tree age. The greatest ability to record tilting
events via reaction wood formation and eccentric growth occurs at
tree ages of 70 to 80 years and 80 to 90 years, respectively. This indicates
differences in a tree ability to allocate its woody material in different decades of tree life. Future research should focus on changes in tree sensitivity to landslide activity during various age periods (not only decades)
or various intervals of stem diameter.

197

Acknowledgement
This study was funded by the Czech Science Foundation project no.
15-02067S. The English language was reviewed by Elsevier Language
Editing Services. I offer my sincere gratitude to Mauro Soldati and an
anonymous reviewer for their insightful comments and to Takashi
Oguchi for the thorough editing.

References
Alestalo, J., 1971. Dendrochronological interpretation of geomorphic processes. Fennia
105, 1139.
Bgin, C., Filion, L., 1988. Age of landslides along the Grande Rivire de la Baleine estuary,
eastern coast of Hudson Bay, Qubec (Canada). Boreas 17, 289299.
Borgatti, L., Soldati, M., 2010. Landslides as a geomorphological proxy for climate change:
a record from the Dolomites (northern Italy). Geomorphology 120, 5664.
Braam, R.R., Weiss, E.E.J., Burrough, P.A., 1987. Spatial and temporal analysis of mass
movement using dendrochronology. Catena 14, 573584.
Carrara, P.E., O'Neill, J.M., 2003. Tree-ring dated landslide movements and their relationship to seismic events in southwestern Montana, USA. Quat. Res. 59, 2535.
ech, S., 2001. Geological map 1:50 000 Rychnov nad Knnou. Czech Geological Survey.
Cook, E.R., 1983. A Time Series Analysis Approach to Tree-ring Standardization (PhD Dissertation) University of Arizona, Tucson, USA.
Cook, E.R., Kairiukstis, L.A., 1990. Methods of Dendrochronology. Kluwer Academic Publisher Dordrecht (408 pp.).
Corominas, J., Moya, J., 1999. Reconstructing recent landslide activity in relation to rainfall
in the Llobregat River basin, Eastern Pyrenees, Spain. Geomorphology 30, 7993.
Corona, C., Lopez-Saez, J., Stoffel, M., 2014. Dening optimal sample size, sampling design
and thresholds for dendrogeomophic landslide sampling. Quat. Geochronol. 22,
7284.
Cruden, D.M., Varnes, D.J., 1996. Landslide types and processes. In: Turner, A.K., Schuster,
R.L. (Eds.), Landslides Investigation and MitigationSpecial Report 247. Transportation
Research Board, US National Research Council, Washington, DC, pp. 3675.
Fantucci, R., Sorriso-Valvo, M., 1999. Dendrogeomorphological analysis of a slope near
Lago, Calabria (Italy). Geomorphology 30, 165174.
Floris, M., Bozzano, F., 2008. Evaluation of landslide reactivation: a modied rainfall
threshold model based on historical records of rainfall and landslides. Geomorphology 94, 4057.
Software, G., 2002. Surfer Version 8 Reference Manual. Golden Software Inc., CA.
Guida, D., Pelni, M., Santilli, M., 2008. Geomorphological and dendrochronological analyses of a complex landslide in the Southern Apennines. Geogr. Ann. A 90, 211226.
Gutirrez, F., Soldati, M., Audemard, F., Blteanu, D., 2010. Recent advances in landslide investigation: issues and perspectives. Geomorphology 124, 95101.
Holmes, R., 1994. Dendrochronology Program Library User Manual. Laboratory of TreeRing Research, Tucson, Arizona, U.S.A.
Lang, A., Moya, J., Corominas, J., Schrott, L., Dikau, R., 1999. Classic and new dating
methods for assessing the temporal occurrence of mass movements. Geomorphology
30, 3352.
Lopez-Saez, J., Corona, C., Stoffel, M., Astrade, L., Berger, F., Malet, J.P., 2012a.
Dendrogeomorphic reconstruction of past landslide reactivation with seasonal precision: Bois Noir landslide, southern French Alps. Landslides 9, 189203.
Lopez-Saez, J., Corona, C., Stoffel, M., Schoeneich, P., Berger, F., 2012b. Probability maps of
landslide reactivation derived from tree-ring records: Pra Bellon landslide, southern
French Alps. Geomorphology 138, 189202.
Lopez-Saez, J., Corona, C., Stoffel, M., Berger, F., 2013. High-resolution ngerprints of past
landsliding and spatially explicit, probabilistic assessment of future activations:
Aiguettes landslide, Southeastern French Alps. Tectonophysics 602, 355369.
Quitt, E., 1971. Klimatick oblasti eskoslovenska. Academia, Praha 73 pp. (in Czech).
Shroder, J.F., 1978. Dendrogeomorphological analysis of mass movement on Table Cliffs
Plateau, Utah. Quat. Res. 9, 168185.
ilhn, K., 2012. Dendrogeomorphological analysis of evolution of slope processes on
ysch rocks (the Vsetnsk vrchy Mts; Czech Republic). Carpathian J. Earth Environ.
Sci. 7, 3949.
ilhn, K., Stoffel, M., 2015. Impacts of age-dependent tree sensitivity and dating approaches on dendrogeomorphic time series of landslides. Geomorphology 236,
3443.
ilhn, K., Pnek, T., Hradeck, J., 2012. Tree-ring analysis in the reconstruction of slope
instabilities associated with earthquakes and precipitation (the Crimean Mountains,
Ukraine). Geomorphology 173174, 174184.
ilhn, K., Pnek, T., Brzdil, R., Havl, D., Duek, R., Hradeck, J., 2013. Dating of bedrock
landslide reactivations using dendrogeomorphic techniques: the Mazk landslide,
Outer Western Carpathians (Czech Republic). Catena 104, 113.
ilhn, K., Pnek, T., Tursk, O., Brzdil, R., Klime, J., Kaikov, L., 2014. Spatio-temporal
patterns of recurrent slope instabilities affecting undercut slopes in ysch: a
dendrogeomorphic approach using broad-leaved trees. Geomorphology 213,
240254.
ilhn, K., Pnek, T., Hradeck, J., Stoffel, M., 2015. Tree-age control on reconstructed
debris-ow frequencies: examples from a regional dendrogeomorphic reconstruction
in the Crimean Mountains. Earth Surf. Process. Landf. 40, 243251.
Stefanini, M.C., 2004. Spatio-temporal analysis of a complex landslide in the Northern Apennines (Italy) by means of dendrochronology. Geomorphology 63,
191202.

198

K. ilhn / Geomorphology 253 (2016) 189198

tekl, J., Brzdil, R., Kakos, V., Je, J., Tolasz, R., Sokol, Z., 2001. Extrmn denn srkov
hrny na zem R v obdob 18792000 a jejich synoptick piny. Nrodn
klimatick program R, Praha (140 pp.).
Stoffel, M., Bollschweiler, M., 2008. Tree-ring analysis in natural hazards research an
overview. Nat. Hazards Earth Syst. Sci. 8, 187202.
Stoffel, M., Bollschweiler, M., 2009. What tree rings can tell about earth-surface processes:
teaching the principles of dendrogeomorphology. Geogr. Compass 3, 10131037.
Stoffel, M., Corona, C., 2014. Dendroecological dating of geomorphic disturbance in trees.
Tree-Ring Res. 70, 320.
Timell, T.E., 1986. Compression Wood in Gymnosperms. Springer, Berlin (2150 pp.).
Tolasz, R., Mkov, T., Valerinov, A., Voenlek, V., 2007. Climate atlas of Czechia. esk
hydrometeorologick stav. Palackho univerzita, Praha, Olomouc (255 pp.).
Van Asch, T.W.J., Buma, J., Van Beek, L.P.H., 1999. A view on some hydrological triggering
systems in landslides. Geomorphology 30, 2532.

Van Den Eeckhaut, M., Muys, B., Van Loy, K., Poesen, J., Beeckman, H., 2009. Evidence for
repeated re-activation of old landslides under forest. Earth Surf. Process. Landf. 34,
352365.
Van Den Eeckhaut, M., Hervs, J., 2012. State of the art of national landslide databases in
Europe and their potential for assessing landslide susceptibility, hazard and risk. Geomorphology 139 (140), 545558.
VIAS (Vienna Institute of Archaeological Science), 2005. Time Table. Installation and Instruction Manual. Ver. 2.1, Vienna.
Westing, A.H., 1965. Formation and function of compression wood in gymnosperms II.
Bot. Rev. 34, 5178.