Precipitation Amounts Triggering Landslide Processes in the Western Part of the Nałęczów Plateau (Lublin Upland, Poland)

The studied sites are located in the vicinity of Zbędowice and Kazimierz Dolny, in the western part of the Nałęczów Plateau, Poland. The Nałęczów Plateau forms part of the Lublin Upland, stretching between the Vistula and Bystrzyca valleys, in a WNW–ESE direction. It forms a broad, flat bed rising 200–220 m a.s.l. (with a length of around 45 km and a width of 10–13 km). To the north and south, the region is a morphological edge. The bedrock of the western part of the Nałęczów Plateau is formed from a diverse range of fractured rocks of Maastrichtian and Palaeogene ages (Pożaryski 1948, Pożaryska 1967, Henkiel, Nitychoruk 1983, Abdel-Gawad 1986, Żarski et al. 1998, Machalski 2005). The topography of the sub-Quaternary basement in the Vistula region exceeds 150 m, decreasing to around 70 m towards the east (in the vicinity of Wąwolnica) (Harasimiuk, Henkiel 1976). This deep bedrock topography was formed at the end of the Early Pleistocene in connection with the development of the Vistula Valley and its erosion. These features determine the contemporary structure of the valley network in the Nałęczów Plateau (Harasimiuk 1980, Henkiel, Nitychoruk 1983, Pożaryski et al. 1994) (Fig. 1).

Fig. 1

In the Nałęczów Plateau, Pleistocene glacial, fluvioglacial and limnoglacial deposits from the Odra glaciation directly overlie sub-Quaternary rocks. North of the Bystra valley, the thickness of the glacial deposits exceeds 20 m. On the other side of the Bystra valley, boulder clay deposits form a thin, discontinuous cover. The formation of these sediments results from the location of this part of the Lublin Upland in the marginal zone of the Odra continental glacier in the post-maximal stadial (Fig. 2A) (Harasimiuk, Henkiel 1976, 1978, Żarski 1988, Pożaryski et al. 1994).

Fig. 2

The youngest Pleistocene sediments are represented by a continuous loess cover from the Weichselian glacial period. These sediments originate from a uniform period of loess formation, which was preceded by several interstadial phases during which poorly developed, initial fossil soils were formed (Harasimiuk, Henkiel 1978, Maruszczak 1991). In the western part of the Nałęczów Plateau, the average depth of the loess cover is 15–20 m, while in the northern and marginal areas of this region and along the valleys of the Bystra and Grodarz rivers, the loess depth reaches 30 m. This significant loess cover thickness results from the infill of depressions, valleys and pre-Quaternary valleys (Fig. 2B) (Harasimiuk, Jezierski 1998).

The relief of the Nałęczów Plateau is dominated by elements relating to loess cover thickness and the formation of loess subsoils. In the upper parts of the loess uplands, a network of valleys and undulating relief is observed in fine-grained deposits. In areas where the loess cover thickness exceeds 10 m, the relief is of a denudational subtype with low altitude differences (up to 30 m), shallow denudation basins and depressions without outflows. The denudation–erosion sculpture subtype appears in areas where the loess thickness reaches 15–20 m. These areas are characterised by altitude differences of up to 60 m and a dense network of erosion–denudation valleys. In contrast, in areas where the loess thickness reaches 30 m, an erosional relief subtype with a contemporary system of extensive systems of erosion–denudation valleys will develop. Altitude differences in these areas may reach up to 100 m and the network of valleys and top parts is cut by ravines whose density reaches up to 17.6 km ∙ km⁻² (mean = 8.6 km ∙ km⁻²; own research based on LiDAR data). According to the data of Kęsik (1961), the density of gullies network ranges from 0.5 km ∙ km⁻² in the southern part of the Plateau to 5.0 km ∙ km⁻² between Kazimierz Dolny and Puławy in the north-western part of this region. Such strong incision of the western part of the Nałęczów Plateau is connected with the deep erosional base of the Vistula River (Fig. 2C) (Maruszczak 1973).

The presence of loess rocks as aeolian deposits, which are poorly cemented and porous (and thus susceptible to thawing and suffosion), in contact with low- and very low-permeability glacial sediments (boulder clays, proglacial clays) and steep ravine slopes of up to 50°, favours landslide formation processes.

Atmospheric precipitation is often the key factor driving slope instability processes. With time, precipitation is transformed into surface runoff and evaporation, with a part retained in the slope cover. The important features of precipitation, ultimately determining the occurrence or intensity of the associated processes, are its height, duration and intensity.

Close to the Nałęczów Plateau, there is one meteorological station from the Institute of Meteorology and Water Management in Puławy (IMGW) that has conducted meteorological observations since 1951. Until 2000, there was a precipitation station of the Institute of Meteorology and Water Management situated in Kazimierz Dolny; however, over 50 years, the data are only 55% complete. In addition, there are single observation stations located in the Plateau, but their observation periods do not exceed 10 years. Precipitation in the marginal zone of the Lublin Upland is characterised by large spatial and seasonal variability. In this area, precipitation is associated with diversified hypsometry, slope exposure and the presence of valleys of various sizes, as well as deep denudation and erosional valleys.

The meteorological station in Puławy is located close to the Nałęczów Plateau at an altitude of 142 m a.s.l. The average total precipitation for the years 1951–2020 is 576.4 mm, with the highest annual precipitation level recorded in 1974 (797.3 mm) and the lowest recorded in the first observation year (1951) with 403.6 mm. According to Kaczorowska’s classification (1962), extremely dry years (i.e. below 50% of the long-term mean precipitation level, 289.2 mm) were recorded in the 70-year observation period. According to this classification, years with average precipitation (90–110% of the average precipitation) dominated, of which there were around 30. There were only 5 very wet years (126–150% of the average precipitation), namely, 1966, 1974, 1980, 2014 and 2017. Notably, these years are none of those in which catastrophic landslide events occurred in the Carpathians and Carpathian Foothills (Fig. 3).

Fig. 3

Precipitation in the summer half-year (May–October) averaged 64% of the annual precipitation levels for 1951–2020. In this period, only two precipitation seasons differed in character, i.e. precipitation in the winter half-year (November–April) exceeded the summer (1967 and 1979). In 14 cases, the precipitation prevalence in the summer half-year exceeded 70%, with a maximum value of 80.5% recorded in 1974. Similarly, as in the case of precipitation in the summer and winter seasons, the precipitation amount recorded in climate seasons also fluctuated considerably. The summer season, from June to July, exhibited the highest seasonal sums of precipitation (37.8% on average). The lowest precipitation levels were observed in the winter season (December–February), accounting for 17% of the annual precipitation sum (Fig. 4).

Fig. 4

The distribution of monthly mean precipitation levels at the Puławy station is marked by a distinct peak in July and August, with the mean sum exceeding 65 mm. The highest average precipitation sums were recorded in July (80.7 mm). The highest monthly precipitation sums were recorded in July 2006 and August 2011, with values of 228.4 mm and 222.5 mm, respectively (Fig. 4).

As shown in Figure 5, days without rainfall dominated in the period 1951–2020, during which time they accounted for more than 56% of the days of the year. Days with very weak (0.1–1.0 mm) and weak (1.0–5.0 mm) precipitation constituted 13.9% and 16.6% of the year, respectively. Moderate and heavier precipitation (>5.0 mm) accounted for 9.3% of the days of the year. During the study period, precipitation exceeding 20 mm ∙ day⁻¹ was recorded an average of 3.6 times ∙ year⁻¹. However, 1981 was an exception, with nine occasions when rainfall exceeded 20 mm. The maximum daily precipitation values were recorded in July 1962 (73.3 mm), May 2014 (68 mm) and May 2001 (63.5 mm).

Fig. 5

The structure of atmospheric precipitation in the Nałęczów Plateau area was determined by analysing precipitation sequences for the period between 1951 and 2020. The combination of at least 2 days with precipitation and a precipitation sum of ≥0.1 mm were defined as a precipitation sequence. Above-average precipitation sequences were determined based on the analysis of the duration and total precipitation level that occurred within the sequence. Threshold values defining above-average precipitation sequences and precipitation sum were defined as values whose probability exceeded 10% (i.e. 90^th percentile). Based on this criterion, four classes were identified:

Class 1. Above-average precipitation with a relatively short duration.

During the study period (1951–2020), there were 2626 precipitation events lasting more than 1 day, of which 57% occurred in the period from April to October (1489 cases). The highest precipitation sum was 129.5 mm over 11 days (from 1 August to 11 August 2006). The longest precipitation sequence lasted 21 days (from 15 January to 4 February 1983), amounting to a total of 35 mm. In the warm period, the longest precipitation sequence lasted 16 days, with a sum of 63 mm. The highest number of recorded rainfall sequences fell in June and July, with average values of 18–20 mm.

From an analysis of all precipitation sequences taking place during the vegetation period (i.e. from April to October), the 90^th percentile in precipitation duration was ascertained at 6 days, together with a precipitation sum of 38.8 mm. Class 2 precipitation, responsible for intensified surface runoff, flushing and initiation of landslide processes, occurred 54 times during the study period (3.8% of all precipitation sequences lasting longer than 1 day). The mean precipitation and duration values of this class were 63.2 mm and 9.4 days (0.29 mm ∙ h⁻¹), respectively (Fig. 6).

Fig. 6

The high precipitation period coincides with the vegetation season and the associated agrotechnical cycle and lasts from April to September/October. Apart from mountainous areas and the coastal zone, the Lublin Upland is characterised by numerous abrupt and heavy precipitation events, which may result from the climate being more continental in this part of Poland (Chomicz 1951, Maruszczak, Trembaczowski 1958, Parczewski 1960, Maruszczak 1986, Starkel 1986, 1996, 1997, 2008, Rodzik et al. 1998, 2008, Kaszewski, Siwek 2005, Siwek 2006, Kaszewski 2008, Lorenc et al. 2012). This part of the upland belt is also characterised upland by significant areas of high thermal and moisture contrasts, such as loess marginal zones, areas adjacent to zones of humid depressions or wide river valleys (the Vistula Valley), which generate downpours (Parczewski 1960, Rodzik et al. 1998, 2008, Rodzik, Janicki 2003).

During geomorphological mapping of the western part of the Nałęczów Plateau, we mapped around 297 landslides (area of about 30 km²). However, these are preliminary results that may change with subsequent research. The average landslide density in the area is about 10 landslides ∙ km⁻². The maximum intensity of landslides was identified in the Zbędowice area and exceeds the study area’s average value by more than three times. The most common landslides are shallow turf landslides, where the thickness of the gravitationally displaced rock package does not exceed 50 cm, and whose area reaches up to 1 acre. In terms of surfaces, the largest landslides occur in areas with the thickest loess cover (typically over 10–15 m thickness), lying on poorly permeable glacial till, where the depth of ravine slits is greatest. The intensity of landslide processes decreases with the upward movement of these valley forms – most landslides were observed on slopes of up to 35° (76% of the recorded landslides), with the highest intensity occurring on 25–35° slopes, accounting for 37% of the recorded landslides.

Six slopes in the vicinity of Kazimierz Dolny: Doły Podmularskie (K02, K03), Norowy Dół (K01)] and Zbędowice [Łachowy Dół (Z01a), on the road from Parchatka to Zbędowice (Z01), Szkutny Dół (Z03) were selected for analysis (Table 1, Fig. 9). The selection of these study objects was based on literature analysis, environmental interviews with local inhabitants and geomorphological mapping of landslide formations in the western part of the Nałęczów Plateau. The selected slopes exhibited visible signs of landslide activity. They are located close to other landslide niches of various sizes, which are distinctly discernible from the surrounding region, and indicate that contemporary landslide processes have been in operation. The selected sites represent erosional ravine valleys with varying exposures and slopes, now shaped by the episodic flow of precipitation and snowmelt water. All the slopes are built of loess rocks with thicknesses varying from 15 m to 30 m (Vistula glaciation); in the lower parts of the profiles, clay sediments are identified, originating from the Odra glaciation period, or sandy/gravelly units underlie the loess (in the northern part of the research area). At the contact between the sedimentary rocks and glacial deposits, water effusions or zones of increased slope cover moisture were observed.

Fig. 9

The results of our fieldwork show that the studied area experiences intense slope processes. These forms vary in size, with a significant proportion of the landslides being 20–40 m long and tending to occur in the lower part of the slopes, with visible zones of wetter soil in their vicinity. However, the boreholes did not reveal the presence of a shallow groundwater table.

Since the upper parts of the slopes were located in the vicinity of the landslides, the location of sliding surface (its entry and exit) was interpreted to coincide with the location of cracks, and colluvium thickness was ascertained based on the shape of intact slopes. Additionally, the Define Slip Surface Grid and Radius method from GeoStudio was used to determine the slip surface. This method is used to define the rotation centres and radii for circular and composite slip surfaces. To localise the position and inclination of the impermeable layer (e.g. clay layer), we identified its outcrops at the toe of the slope and in the beds of watercourses.

Table 2 contains the basic geotechnical parameters of the loess slope cover for each slope. The loess deposits from different sites exhibit very similar shear strength parameters and permeability coefficients, which is probably an effect of the highly consistent grain size distribution. These soils can be classified as silt and are characterised by medium permeability, relatively high internal friction angle values and very low cohesion, which are consistent with the results of loess studies from other regions (Xing et al. 2016, Xu et al. 2018).

Rainfall data from the vegetation seasons of 1951–2020 were used to determine the precipitation sum values, including different probabilities of occurrence (i.e. return periods) (Fig. 10). The depth of daily rainfall with a 100-year return period (1% occurrence probability) is slightly lower than 100 mm, similar to the maximum 24-hour rainfall values recorded in meteorological stations in the centre of Poland in the 1951–2015 period (Pińskwar et al. 2019). However, the precipitation sum values for longer periods are quite low relative to those estimated for the Nowy Wiśnicz Foothills in the northern part of the Carpathians (Demczuk et al. 2019) or even for the Lisbon Region (Vaz et al. 2018).

Fig. 10

The next step of the analysis was the development of a slope model for each analysed landslide; these models were then further subjected to hydrological and geotechnical calculations, assuming the meteorological values for the years 1973–2020 as boundary conditions. As a result of these calculations, we determined the general trend of slope equilibrium condition changes, in addition to gaining insights into precipitation episodes threatening slope stability.

The calculation results (Figs 11 and 12) indicate that the average FS value for the studied objects in the years 1973–2020 reaches an increased probability of landslide occurrence (1 < FS < 1.3) according to Wysokiński’s (1980) criteria. The lowest safety coefficient values were identified in the autumn–winter season. Evaporation from the ground and plants is inhibited during this period. The lowest FS value occurs at the turn of March and April, associated with relatively low precipitation, largely accumulated in the form of snow cover. This results in high landslide inertia and delayed slope response to precipitation. However, at the start of the vegetation season (April), the FS increases, which is related to intensified evaporation that exceeds precipitation. In mid-May, a slight fluctuation/decrease in the FS value was observed, which is associated with an increase in precipitation sum and duration, in addition to high, stabilised infiltration of precipitation into slope cover. At the end of August and the beginning of September, the value of FS decreases rapidly due to decreasing evaporation from plants and the ground.

Fig. 11

Fig. 12

Analysis of slope stability during the 1973–2020 period, including the mean monthly SF values (Figs 12 and 13), indicates that slope stability is significantly correlated with seasons. The analysis shows that there is a pattern of SF fluctuation. Generally, the months at the beginning of the year are the most unfavourable from the perspective of slope stability. Also visible is a slight trend of increasing slope stability in summer months within the analysed years, which is a probable consequence of global warming.

Fig. 13

During our analysis, we implemented 138 computational scenarios for six ravine slopes in the western part of the Nałęczów Plateau. The computational scenarios considered the simulated variations in the precipitation amount and its intensity (i.e. uniform, increasing or decreasing).

The critical precipitation level for landslide triggering is mainly related to the condition of the soil on the initial day of analysis (initial soil moisture content condition). The average critical precipitation level for the analysed slopes was 93.8 mm, with a minimum value of 35.2 mm, and an average precipitation duration of about 27 days. The critical precipitation level for the wet soil is 23% lower than that for dry slopes and 19% lower than the critical precipitation level for humid soils. Increasing precipitation intensity is the most conducive factor to landslides, irrespective of initial soil moisture. This type of precipitation series allows rapid, easy infiltration into slope cover due to the precipitation intensity being smaller than or similar to the infiltration parameters of the rocks forming the slopes. The infiltration process increases the moisture content in the slope soils and creates a saturated zone characterised by low shear strength. The table below summarises the results for the extreme and mean values of the critical (threshold) precipitation levels (Table 3).

Figure 13 shows the distribution of the obtained critical precipitation results for all the studied slopes. Notably, the (threshold) critical precipitation values shift towards the left-hand side of the plot with increasing initial soil moisture, and the spectrum of precipitation intensity data increases. The critical precipitation for landslide triggering in wet slopes is more intense, on average, than that triggering landslides in cases of dry or humid slopes. The distribution of the critical precipitation levels mainly depends on the size of the slope cover involved in the landslide process. The larger the volume (i.e. the greater the length of the slope), the greater the amount of precipitation needed over a longer time for the initiation of landslide processes.

In general, our results indicate that the precipitation levels initiating failure of the analysed slopes are low compared to threshold precipitation values reported in Polish (Gorczyca 2004, Gil, Długosz 2006) and world literature (Lumb 1975, Guzetti et al. 2007, Peruccaci et al. 2017, Marjanovic et al. 2018, Vaz et al. 2018, Teja et al. 2019, He et al. 2020). For example, for Hong Kong, Lumb (1975) reports that landslides are initiated with rainfall totals exceeding 50 mm in the 15 days preceding the landslide events, with the principal rainfall initiating landslides (on the day of occurrence) in excess of 50 mm. Gil and Długosz (2006) report that the precipitation initiating shallow mass movements for the Carpathian area corresponds to a precipitation total of 200–250 mm for 20–40 days, whereas 100 mm of high-intensity precipitation is conducive to the formation of mudflows and combined mud–stone slides. Gorczyca’s (2004) results from observations of Carpathian landslides show that such events can be initiated by smaller precipitation depths. The relationships between the values of cumulative precipitation and their duration for the analysed slopes, presented in Figure 14, demonstrate that, for the most part, the values from this study fall within the threshold ranges given for various regions. However, it should be stressed that the relationships from previous literature are empirical and usually take into account short observational periods of several days; thus, they may not be necessarily representative of all possible cases of rainfall distribution over time.

Fig. 14

The analysis of pore pressure distribution showed that slope failure is usually related to water accumulation in the soil in the lower part of slopes, while the middle part of the slope remains unsaturated. This means that the slope instability occurs as a result of reduced suction pressure; this appears reasonable considering that the slope angle in the lower part of the slope exceeds 40°, while the internal friction angle values in loess are 30.4°–33.9° and the cohesion is close to 0 kPa. Such a relationship deviates from previous observational results and analyses of the causes of mass movements in the Carpathians (Thiel 1989, Demczuk et al. 2019), where slope failure is mostly accompanied by saturation of the whole slope with water. However, these results from the slope stability analysis for loess slopes or steep slopes indicate that, in many cases, slope failure results from a reduction in suction pressure and partial saturation of slope cover. Therefore, the precipitation levels initiating the failure of loess slopes are lower than those for the Carpathian slopes, whose inclinations are usually gentler.

A comparison of the calculation results with data on the levels of precipitation sequences for warm periods of the years 1951–2020 highlighted the sequences corresponding to Class 2, as determined by the 90^th percentile for the precipitation sum and duration. However, a 2-day extension of the minimum precipitation duration (up to 8 days) is observed. Out of the 54 precipitation events classified as extreme (Class 2), only 7 coincided with the calculation results. Comparing the results obtained with those for the slopes of the Wiśnickie Foothills, calculated using the same method, the ravine slopes require, on average, half the precipitation, with a 16% shorter duration for landslide initiation.