Research conducted in various parts of the world indicates a rise in the activity of mass movements, including an increase in the number of landslides, in recent decades (Innes, 1983; Winchester and Chaujar, 2002; Petley
In Duexi (China’s Sichuan province) a landslide with a volume of 8 million m3 was triggered, as a result of which 62 buildings and 1600 metres of road were destroyed, and 10 people were killed (Qui
The increasing number of landslides and related economic losses have resulted in the development of new methods of mapping landslide activity, e.g. aerial and satellite imagery (Murillo-García
Dendrochronology has also been used for developing maps of landslide activity (Catani
Dendrochronological techniques for landslide dating, which are used for the mapping of landslide activity and hazards, are based on the fact that the stems of trees growing on slopes are tilted during landslide episodes, which is reflected in the structure of their wood. Following tilting (after the landslide event), “reaction wood” is formed inside stems and tree rings start to become eccentric. These features of the wood anatomical structures, along with the macroscopic characteristics of the ring structure, allow landslide events to be dated with annual frequency (Demoulin and Chung, 2007; Gärtner and Heinrich, 2013; Lopez Saez
The study area is located on the massif of Sucha Mt (max 1040 m a.s.l.), in the Beskid Żywiecki Mts, in the Western Carpathians (Central Europe) (
The location of the study area in Central Europe, Western Carpathians (A) and the location of sampling sites for a dendrochronological study in the massif of Sucha Mt, in the Beskid Żywiecki Mts (B).Fig. 1
According to Hess (1965), the study area is situated in cold climatic zone with a precipitation of 1150–1350 mm. The average annual precipitation at the nearby gauging station (Żabnica, 550 m a.s.l.) is 1136 mm. The study area belongs to the lower montane vegetation belt, where common beech (
Landslides occurring in the study area pose a threat to both people and infrastructure. In 2010, landslides were triggered in the study area and on the massif of Prusów Mt in Milówka, in the neighbourhood of the study area. The most hazardous landslide destroyed local road and electricity networks and 9 buildings, and blocked a stream valley, causing flooding.
Samples for a dendrochronological study were taken from Norway spruce trees (
At each site, samples (cores) were taken from the stems of 3 trees growing at a distance of several to several dozen metres apart. Samples were taken by means of a Pressler borer from trees whose stems had not been damaged and presented with no defoliation. Trees which were tilted transversely to the slope’s inclination were also excluded from the study. This is because most trees affected by landslide movements have stems tilted in axes parallel to the slope. Furthermore, all tilted spruces have deformed trunk cross-sections parallel to the direction of the tilting. For sampling purposes, we also selected trees whose circumference measured at chest height was at least 50 cm. Sometimes, it was not possible to find 3 spruce trees meeting the above-mentioned requirements at one sampling site, in the case of which cores were taken from 1 or 2 spruce trees. In total, samples were collected from 131 trees.
Two cores were collected from each tree: one from the upslope side of the tree and the other from the downslope side. The samples were taken at chest height, according to the slope’s inclination. The location of each tree was marked using GPS. The cores collected were glued into wooden holders and sanded, following which the tree-ring widths were measured with an accuracy of 0.01 mm (LinTab measurement station). The eccentricity, the eccentricity index and its yearly variation were calculated based on a comparison of the tree-ring widths found on the opposite sides of a single tree stem (method following Wistuba
An example of dating landslide events from ring widths into (A) eccentricity, (B) eccentricity index [%] and (C) its yearly variation [%]: U — tree-ring widths in the upslope part of the stem [mm]; D — tree-ring widths in the downslope part of the stem [mm]; E — the eccentricity of the tree-ring [mm]; Ei — the eccentricity index of the tree-ring [%]; vEi — the yearly variation in the eccentricity index [%]; x — year (annual tree ring).Fig. 2
The development of the landslide hazard map, based on the calculated average frequencies of landslide events at the sampling sites, made it necessary to select the adequate data interpolation method. ArcGIS software (ArcGIS Desktop, 2017) was used to determine the method of interpolation, and then the method of recalculation and visualisation of dendrochronological data.
The method of data interpolation for landslide activity in the study should be global. This means that it applies a single mathematical function to all measurement points (1); should be precise,
The IDW method was rejected at the beginning because it created structures of the “bull’s eye” type (unreal accumulations of counter lines around extreme values, thus generating unlikely surfaces). The remaining 3 methods of interpolation available in the ArcGIS software were analysed (RBF, Spline, and Topo to Raster). These methods best reflected the possible spatial distribution of landslide episodes identified on the basis of dendrochronological data (
The results of interpolation for all points (A, C, E), and without 5 control points (B, D, F). Interpolation methods: the Radial-Basis Function (A, B), Spline (C, D), and Topo to Raster (E, F).Fig. 3
The RBF method is conceptually similar to fitting a rubber membrane through the measured sample values (the surface passes through the data values) while minimising the total curvature of the surface. The basic function selected determines how the rubber membrane fits between the values (ArcGIS Help, 2017). The next method, Spline, estimates values using a mathematical function which minimises the overall surface curvature. This results in a smooth surface which passes exactly through the input points (Childs, 2004). The last tool, Topo to Raster, uses an interpolation method specifically designed to create a surface which more closely represents a natural drainage surface, and better preserves both ridgelines and stream networks. This technique creates hydrologically correct DEM’s, and is based on the ANUDEM program developed by Hutchinson (1989, 2011).
A visual assessment of the quality of the interpolation methods was made by comparing the results on the interpolation maps (
The calculated errors of the interpolation methods. A Root Mean Square Error (RMSE) expresses the differences between original (real) values (2nd column) and the interpolated values (3rd, 5th and 7th columns).Points Radial Basis Function Spline Topo to Raster No. Real Value Value Difference Value Difference Value Difference 1.37 +0.72 1.06 +0.41 1.36 +0.71 0.95 –0.57 1.12 –0.40 0.91 –0.61 1.04 –0.17 0.98 –0.23 1.07 –0.14 1.48 +0.10 2.05 +0.67 1.37 –0.01 0.50 –0.71 0.30 –0.91 0.63 –0.58 1.07 –0.13 1.10 –0.09 1.07 –0.13 – 0.53 – 0.57 – 0.49
When presenting interpolated data we selected the intervals of the values of mean landslide activity per 10 years to be presented on the map of landslide activity: <0.50; 0.51–1.00; 1.01–1.50; 1.51–2.00; >2.00 (
The barriers applied in the process of dendrochronological-data interpolation in the massif of Sucha Mt, Western Carpathians, Poland: (A) watercourses, (B) the most-important ridgelines.Fig. 4
Average frequency of landslide events calculated for each sampling site in the massif of Sucha Mt, Western Carpathians, PolandFig. 5
A landslide activity map developed on the basis of the selected interpolation method – Spline with barriers (A) and the landslide hazard map for the massif of Sucha Mt, Western Carpathians, Poland (B). Maps of the distribution of landslides were created on the basis of the results obtained from the tree-ring eccentricity analysis.Fig. 6
Tree-ring chronologies in the massif of Sucha Mt reach back to 1883. The sampled trees were from 15 to 133 years old, but only a few of them were very young (11.5% trees were less than 30 years old). Almost 51.9% of all trees were 50 or more years old, and 6.9% were more than 100 years old. A total of 282 landslide events were identified in 131 of the spruce trees which were sampled, and the earliest landslide event observed in the tree-ring series dated back to 1919.
The average frequency of landslide activity noted for each sampling site varies from 0 events/10 years (
The landslide activity map (
On the landslide hazard map, it was found that 84 buildings (65.62 %) and 5.28 km of roads (65.3 %) are located in the medium and high landslide-hazard zone in the study area (
Dendrochronological data provide data on the temporal and spatial variability of landslide activity in the past, with yearly (or even seasonal) resolution from recent decades up to hundreds of years (depending on stand age) (Shroder, 1980; Butler, 1987; Schweingruber, 1996). In this study, we used each tree growing on a landslide as a separate living sensor of ground movement, recorded as macroscopic characteristics of the ring structure, such as tree-ring eccentricity.
One of the problems encountered in developing landslide hazard maps, based on the use of the dendrochronology, is the selection of wood anatomical structures. Usually, two features of wood anatomy,
Another problem is the selection of trees from which samples (cores) will be taken for analysis. The first limitation refers to the distribution of trees. Usually, some areas mapped on the basis of interpolation are not forested, and of course, landslide activity cannot be determined for these areas. The species composition of the stand from which samples are taken is also important. The wood anatomical structures which develop in trees affected by landsliding have been best recognised in conifers, spruce in particular (Stoffel
It is also indispensable to carefully select the sampling sites where the cores will be taken from trees. It is not enough to distribute the sampling sites at equal distances from one another. The distribution of the sampling sites should be dependent on the location of the landslide areas visible on the Digital Elevation Model developed from the LiDAR data (and, if possible, otherwise visible on topographic maps). Using the Digital Elevation Model, sampling sites should cover each separate slope and the surface of individual landslides visible on the map.
At the stage of editing the map, once the study of dendrochronological samples in the field and in the lab is completed, we should choose the method for interpolating the obtained results of landslide activity. The irregular distribution of sampling sites makes it difficult to choose the interpolation method. Caruso and Quarta (1998) used 4 methods of interpolation in their studies on a variety of environmental data, concluding that there was no single universal interpolation method. Depending on the problem being studied and the kind of data obtained, we should choose the most appropriate interpolation method. The interpolation of the spatial distribution of the values of environmental factors has not been given much attention so far (e.g. Gong
It is necessary to impose barriers during data interpolation to improve the automatic interpolation of data. In the events where the chosen interpolation method transfers the interpolated value (contour line) over a river flowing in the bottom of a valley to the opposite slope, a barrier should be placed on the course of the river. The value for the frequency of landslide activity on the opposite slope should come from the analysis of trees growing on the opposite slope. However, we should be careful when imposing barriers in small narrow valleys because it is possible that deep seated landslides can spread across several small river valleys (Xu
During the interpolation of the results of dendrochronological studies, the question arose of whether barriers at the landslide boundaries, visible on the Digital Elevation Model, should be established. The area of landslide activity can go beyond the visible boundaries of the existing landside bodies, given that landsliding can also occur on a slope without any landslide landforms visible on the model or in the field of a landslide (e.g. head scarps, landslide blocks and toes, and hummocky topography) (Papciak
Although there are numerous studies on landslide hazard assessment, there is no standard procedure for the preparation of the landslide hazard map (e.g. Guzzetti
Different types of data are used for preparing landslide maps, for example, geological and geomorphological data (Carrara
Landslide hazard maps developed on the basis of dendrochronological data show landslide activity, as opposed to hazard maps based on susceptibility to landslides occurring. Moreover, dendrochronological studies have allowed us to distinguish currently active landslides from relict landslides (Van Den Eeckhaut
Landslide hazard maps developed on the basis of dendrochronological data show ground activity in recent decades with seasonal precision (Schweingruber, 1996). Maps developed using other methods, such as e.g. remote sensing techniques, GPS measurements or inclinometric monitoring, only provide partial information on landslide history, as they are available only for a short period,
Landslide hazard maps based on dendrochronological data are relatively cheap because of the low cost of data collection and preparation. We only need a Pressler borer to extract a core from a tree, and a measuring system with software for tree-ring analyses. Other methods for collecting data on landslide activity, such as slope inclinometer monitoring or laser scanning, are relatively expensive (Chase
Trees record changes in wood anatomical structures (for example, the eccentric growth of the trees used in the study being discussed) under the influence of mechanical stress caused by various environmental factors, e.g. land-sliding, soil creep, snow avalanches or wind impact (Zielonka and Malcher, 2009). In order to minimise the risk of tree-ring eccentricity being caused by other processes and mistakenly be attributed to a landslide event, the record of the eccentricity index is compared to the record with the average level of tree-ring eccentricity occurring on a reference (control) slope. The reference, stable slope is located near the study landslides and is characterised by similar parameters,
Another limitation of a landslide map based on a tree-ring study is the nature of the data obtained. Dendrochronological data reflect the frequency of wood anatomical structures which occur periodically, and directly record tree stem tilting events. Landslide events, however, are dated only indirectly. Other methods, like inclinometer or geodetic monitoring, show direct changes in ground movements.
By using dendrochronological data, e.g. the eccentric growth of trees, it is possible to develop landslide hazard maps over a large area covering a whole mountain range or massif. These maps allow us to highlight areas which are potentially safe for existing buildings, roads, infrastructure and future development. Landslide hazard maps developed on the basis of dendrochronological data can be used in local spatial planning, providing grounds for preparing landslide risk assessments.
The most accurate results to be used in the development of maps are obtained when the trees growing in the mapped area are conifers, similar in age, and relatively old and homogenous in species composition. An important advantage of using dendrochronological data (and in particular the eccentric growth of trees) for preparing landslide hazard maps is the character of the data obtained. Numerical results include a temporal scale (several decades of landslide activity are recorded – and landslide event frequency is noted per decade), and a spatial scale (any selected area covered with trees). The use of dendrochronology to assess landslide activity is much cheaper than other methods, such as inclinometer monitoring. Modern methods of investigating landslide activity, such as interferometers or terrestrial laser scanning, allow very short data sequences to be obtained, while dendrochronology allows one to obtain 100 years of results of landslide activity, or even longer series of data.
The disadvantage of dendrochronology is the indirect nature of the results obtained. In fact, the results relate to the measurement of anatomical changes and a macroscopic characteristics of the ring structure occurring in the wood of tilted trees, whereas anatomical changes and a macroscopic characteristics of the ring structure occurring periodically in wood merely provide indirect information about landslide events. The second disadvantage of the method is the ability to record different environmental factors, such as wind, causing tree stems to tilt and wood anatomy disturbances to develop. The comparison of the dendrochronological results on the slope examined with those on a reference slope allows one to partially eliminate this error.