Long-Term Trends in the Chemical and Mechanical Denudation in a Small Carpathian Catchment

This study was carried out in the Bystrzanka catchment (49°37′51″N, 21°07′02″E) located in the Polish Flysch Carpathians (Fig. 1A, Starkel 1972). The study area has an average altitude of 487 m a.s.l. and an average slope of 9.6°. Slopes reach up to 35° in the western part of the catchment, primarily facing NE, while slopes in the eastern region only reach up to 15° and primarily face SW (Fig. 1B). The geology of the studied area is dominated by rocks from the Magura Nappe (Kozikowski 1956, Świdziński 1973). The occurrence of differentiated lithological complexes in the catchment is of particular importance. The western section of the catchment (600–753 m a.s.l.) is characterised by the occurrence of coarse- to medium-bedded sandstone and shale formations from the Magura series, which are characterised by their high resistance to denudation processes. The eastern section of the basin is composed of less resistant shale–sandstone inoceram layers (thin-, medium- and thick-bedded sandstones with shale and fucoid marls) from the Cretaceous–Paleogene period (Kopciowski et al. 1997, Fig. 1C). The presence of shale in the geological structure as well as the occurrence of weathered cover on the slopes favour denudation processes (Kozikowski 1956, Świdziński 1973). The bedrock in the catchment has undergone secondary folding and has been displaced along the Ropa axis fault – it is also cut by minor faults (Świdziński 1953, Sikora 1970). Around 30% of the catchment area consists of landslides.

Fig. 1.

A – Location of the study area and environmental characteristics of the Bystrzanka catchment area; B – slope; based on the deM with 10 m resolution; C – geological structure: 1– fault, 2 – minor thrust fault, 3 – major thrust fault, 4 – Inoceramus beds, 5 – Cięzkowickie sandstones, 6 – Magura sandstones, 7 – Sub-Magura sandstones, 8 – Krosno Bed, 9 – variegated shales; based on Wójcik et al. (2003).

The dominant soil types in the catchment are sandy clay loam and clay loam Inceptisols with a relatively low water capacity (Soil Survey Division Staff 2017). The thickness of the soil cover is 80–100 cm near the hilltops and increases to 200–400 cm along the slopes. Soils have formed on the slope cover, which – in areas built on Magurian sandstones – have a stony fraction of approximately 50–80%. In contrast, in the shale– sandstone series of the Inoceramian and Krosno beds, the proportion of skeletal fractions is 20– 40% (Adamczyk et al. 1973).

The main course of the Bystrzanka valley has a narrow bottom (rocky-alluvial) and steep slopes of up to 15 m; the valley is about 3–4 m wide in its middle course, and the riverbed is about 10 m wide at the mouth of the Bystrzanka valley where it leads into the Ropa (Niemirowska 1970, Kijowska-Strugała 2015).

Over time, the catchment has undergone significant changes in land use and land cover (LULC), primarily due to a major political transformation in 1989, which had a significant effect on southeast Poland. The eastern part of the catchment area has been used for agricultural purposes for many years. In 1969, cultivated land accounted for 48% of the area, while meadows made up only 3%. However, fundamental changes in the LULC have been observed in the past 60 years, with the area of cultivated land decreasing by up to 80% with a complementary eightfold increase in meadow area (Fig. 2). European Union (EU) policy supports the mowing of meadows, for example, for use as fodder. Two villages within the catchment area, Szymbark and Bystra, have a combined population of around 5000 people (Central Statistical Office of Poland 2022). The topographic database (BDOT10k – Polish acronym) of the catchment area indicates that the 448 buildings within the region are primarily concentrated along the valley axis. Over the past 50 years, an increase in the population and development of the two villages have led to a 19% rise in road density (Kijowska-Strugała 2019), highlighting how LULC and population density have contributed to changes in the area and spatial distribution of road surfaces in the region. The construction of new paved roads in the valley bottom has been accompanied by the abandonment of agricultural fields and cart roads on slopes. In the early 21st century, particularly in 2009 and 2010, the catchment saw a surge in engineering investments co-financed by EU funds. During this time, ditches along the main road were sealed, six landslides were stabilized, the largest bridge over the main stream was reconstructed and a 2-km section near the mouth of the Bystrzanka stream was reinforced with concrete (Kijowska-Strugała 2019, Gorlice Municipal Office 2024).

Fig. 2.

Land use an land cover (LULC) changes in Bystrzanka catchment (1 – cultivated land, 2 – forest (deciduous, mixed), 3 – grassland, 4 – orchards, 5 – agro-forestry transitional areas, 6 – urbanised area, industrial or commercial, sports or recreational areas, low-density buildings, roads; based on aerial photos Olędzki 2007, Zwoliński, Gudowicz 2011, Bochenek 2020).

Studies of meteorological and hydrological parameters have been carried out continuously from the meteorological station that is part of the Szymbark Research Station (SRS) of the Institute of Geography and Spatial Organization of the Polish Academy of Sciences since the 1970s. The SRS is located in the Bystrzanka catchment. The meteorological station is located at an altitude of 325 m a.s.l. This study used data from 1995 to 2023. Daily precipitation totals (wet precipitation and dry deposition) were determined using a Hellman rain gauge and an automatic VAISALA station. Due to the lack of other precipitation stations in the study area, the results of a single station located near the mouth of the Bystrzanka stream into the Ropa river were taken into account; this is despite the likelihood of a precipitation gradient in the study area as well as increased precipitation with altitude. Bochenek (2007) analysed the spatial differentiation between the total annual and seasonal precipitation in the Beskid Niski, Ciężkowice Foothills and Jasielsko-Sanockie Basin between 1970 and 1981 based on data from 27 precipitation stations. The calculated gradient of annual precipitation was 30 mm per 100 m of altitude. Based on this profile, precipitation at the highest point of the catchment (753 m a.s.l.) should be about 130 mm higher than the meteorological station, averaging around 995 mm.

The Olechnowicz-Bobrowska et al. (1970) classification was used to assess the frequency of days with a certain amount of precipitation. The ionic composition of precipitation water was determined from monthly samples of total deposition, derived from the daily precipitation totals.

Water level changes in the stream were recorded using a KB2 limnigraph and converted into flow rates. The flow curve was updated after each flood, especially when changes in bed morphology were observed. The Rc was calculated as the quotient of runoff (annual values) and precipitation (annual values). Floods and hydrological droughts were identified based on the magnitude of the discharge. The criterion proposed by Ozga-Zielińska and Brzeziński (1997) was used to identify flood events. Floods were also grouped according to their origin based on the terminology described by Lambor (1965). The lower discharge threshold for flood events was set at 0.689 m³ ∙ s⁻¹. Hydrological droughts were identified using the method proposed by Hisdal et al. (2004), with the threshold set at the 0.70 quantile. Over the study period, this upper threshold was 0.035 m³ ∙ s⁻¹.

Chemical and mechanical denudation studies have been carried out in the Bystrzanka catchment since the 1970s and continuously since 1995. Water temperature, pH and electrical conductivity (EC) relative to 25°C were measured daily using an Elmetron (CP-315, CC-311 and CX701) over the entire study period. Both pH and EC were calculated as monthly weighted averages with respect to precipitation or water flow in the channel. Water samples were collected at the mouth of the stream for chemical analysis. During floods, the frequency of water sampling was adjusted based on the rate of change in water level in the channel (from 15 min to about 1 h). The water samples were filtered using Whatman glass microfibre filters (GF/D) 2.7 μm and their ionic composition was determined using ion chromatography (Dionex ICS 1100) (Cl^–, S–SO₄²⁻, N–NO₃^–, P–PO₄³⁻, N–NH₄⁺) and atomic absorption spectrophotometry (VARIAN) (Ca²⁺, Mg²⁺, Na⁺, K⁺). All ions have been identified using ion chromatography since 2015. As other studies have shown, both spectrophotometric and ion chromatographic methods give very similar results (Thienpont et al. 1996). The concentration of HCO₃^– in the stream was measured by titration with 0.1M HCl. The chemical composition in the runoff (R) and precipitation (P) was calculated by multiplying the weighted ion concentration by the runoff or precipitation, respectively. The dissolved load (DL; kg ∙ a⁻¹) and yield (DY; kg ∙ ha⁻¹ ∙ a⁻¹) in the precipitation (DL_P; DY_P) and runoff (DL_R; DY_R) were also determined. The chemical denudation balance (kg ∙ ha⁻¹) was calculated as the difference between DL_R and atmospheric input (DL_P).

Suspended sediment concentration (SSC) was measured manually using the 1 L bottle method (Wren et al. 2000) in the hydrometric profile at 60% water depth as recommended by Newburn (1988). The frequency of sampling was approximately every 15–20 min during floods and decreased with decreasing water levels. During periods of low-intensity rainfall, which caused the water level in the stream to rise slowly, samples were taken every 1–2 h. Sampling frequencies were relatively consistent during the study period. Daily water samples were collected at 6 am (UTC) as part of the continuous monitoring of the stream. The presence of SSC was not observed during periods of average and low water levels in the Bystrzanka stream.

The water samples were filtered; any sediment accumulated on the filters was dried (105°C) and weighed to determine the SSC (g ∙ dm⁻³) in the laboratory at the SRS. The area-specific suspended sediment yield (SSY; Mg ∙ km⁻² ∙ a⁻¹) and absolute suspended sediment load (SSL; Mg ∙ a⁻¹) were calculated based on the SSC using the following equation: 1SSL=∑1nSSC ·Q ·Δt where: –

n – number of samples;

Statistical analyses were performed using Statistica v.12. The Shapiro–Wilk distribution test was used to assess the distribution of data. The non-parametric Mann–Whitney U test was used to compare the differences between the first (1995–2004) and last (2014–2023) decades of the research period. The Pettitt test was used to detect changes in the chemical denudation balance data. This method identifies a single point of change in the time series, that is, the point at which there is a significant change in the trend or mean of the distribution of the variable. The data were statistically analysed using a linear trend. A significance level of a = 0.05 was applied for trend analysis, as is standard in statistical research (e.g. Souvignet et al. 2012). For results with higher statistical significance and more pronounced trends, values were reported with a higher significance level (i.e. p < 0.01).

The mean annual air temperature during the study period was 8.8°C and exhibited a statistically significant increasing trend of 0.7°C per decade (α = 0.05; Table 1). A positive trend in mean monthly air temperature was observed in all months of the year except May (Table 1).

The mean annual precipitation during the study period was 864.5 mm, with annual totals ranging from 612.4 mm (2003) to 1180.0 mm (2014) (Table 1 and Fig. 3). The mean monthly precipitation ranged from 42.0 mm in December to 130.6 mm in July. Precipitation totals fluctuated each month between 1995 and 2022. The largest increase in precipitation occurred in August (12.5 mm per decade), while the largest decrease occurred in June (–18.1 mm per decade). The trend coefficient was influenced by the monthly precipitation totals and both annual and monthly rainfall totals were not statistically significant and did not show directions of change over the multi-year period. The average number of days with precipitation (Np ≥0.1 mm) was 178 days ∙ a⁻¹. In this context, a significant change was observed in 2012. Between 1995 and 2012, the average number of days with precipitation was 184 days ∙ a⁻¹ and exhibited a statistically significant positive trend of 12 days per decade. However, after 2012, the number of days with precipitation declined significantly, with a negative trend of 72 days per decade. This sharp decrease only stopped in 2020. The average number of days with precipitation between 2012 and 2023 was 170 days ∙ a⁻¹.

Fig. 3.

Annual totals of precipitation (P), runoff totals (R) and runoff coefficient (Rc) in 1995–2023.

The structure of daily precipitation was dominated by very light (≤1.0 mm) and light (1.0 < … 1.0 < P ≤ 5.0) precipitation, accounting for 32.2% and 41.5% of the days, respectively. The decrease in the number of days with precipitation between 2014 and 2023 was accompanied by a decrease in the number of days with very light precipitation. Between 1995 and 2022, there was an increase in the number of days with precipitation between 20 mm and 30 mm. Such precipitation, especially over periods of several days, resulted in significantly increased streamflow.

Differences in daily precipitation, annual snowfall (SF) and snowpack accumulation patterns observed in the first and last decades of the study have been attributed to climate change. The increase in air temperature, particularly during winter, has contributed to a negative trend in annual SF totals, despite the lack of any significant changes in precipitation during this hydrological half-year (HY). This warming has also led to a decrease in snow cover as well as the shortening of the snow cover retention period (Table 2).

This significant increase in temperature during the winter months has also affected the amount of SF and the duration of snow cover. The average annual SF was 117.6 mm (13.6% of total precipitation), while the average duration of snow cover was 67 days ∙ a⁻¹. The highest SF totals were recorded in the winters of 1999/2000 (222.8 mm) and 2012/2013 (214.0 mm), while the lowest totals were recorded in the winters of 2013/2014 (26.0 mm) and 2019/2020 (17.0 mm) (Fig. 4). Over the study period, there was a statistically significant negative trend in both annual SF (–29.1 mm per decade) and snow cover length (–16 days per decade).

Fig. 4.

Annual SF, share of snowfall in annual precipitation and duration of snow cover. SF, snowfall.

The mean annual runoff from the Bystrzanka catchment was 390.2 mm, representing 45.1% of the total precipitation (Table 1). The total annual runoff ranged between 85.0 mm in 2012 and 856.6 mm in 2010 (Fig. 3), with a coefficient of variation (Cv) of 36.8%. The mean runoff in the winter hydrological HY was 199.3 mm (51.1% of the annual total; Cv = 36.6%), while the mean runoff in the summer hydrological HY was 190.9 mm (48.9% of the annual total; Cv = 56.6%). The proportion of runoff in the winter hydrological HY compared with the annual total ranged from 22.8% in 2014 to 75.2% in 2022.

A total of 294 floods occurred in the main stream between 1995 and 2023. Floods resulting from continuous rainfall and from downpours were relatively similar – 95 and 94 events, respectively (Fig. 5A). The total flood discharge (i.e. the amount of water above the threshold discharge) was 2575 mm, which accounted for 22.8% of the total discharge. Floods after downpours discharged the most water (998.8 mm), while mixed floods (i.e. due to a combination of rainfall and snowmelt) contributed the least discard (229.2 mm; Fig. 5B). The mean annual flood runoff was 89.4 mm. The number of floods per year ranged from 2 (2012) to 22 (2010), while the magnitude of the flood runoff (water level above discharge threshold) ranged from 3.2 mm (2022) to 263.5 mm (2010) (Fig. 6). Between 2014 and 2023, there was an average of 9.7 floods, 1 less than between 1995 and 2004, and flood runoff was 29 mm lower. The duration of hydrological droughts ranged from 64 days (1998) to 303 days (2012). On average, hydrological droughts occurred on around 163 days ∙ a⁻¹, including 64 deep drought days ∙ a⁻¹. Over the period analysed, a statistically significant increase in the duration of hydrological droughts was observed (p = 0.033). In 1995–2004, they lasted 136 days on average, and 42 days longer in the last decade.

Fig. 5.

A – Number of floods distinguished by genetic classification (Lambor 1965) in 1995–2023; B – layer of water above discharge threshold (mm). Or – continuous rainfall, Ou – heavy rainfall, O(u + r) – mixed rainfall, r – snowmelt, M – mixed rainfall and snowmelt.

Fig. 6.

Flood runoff and number and hydrological drought duration from 1995 to 2023.

The mean annual pH of precipitation exhibited a statistically significant increasing trend of 0.3 units per 10 years (α = 0.05), with values ranging from 3.99 (2003) to 5.60 (2022) (Fig. 7; mean annual value = 4.90; Cv = 8.2%). However, there was no statistically significant trend observed in the mean annual EC (multi-year mean = 1.90 mS ∙ m⁻¹; range of variation 0.90–5.60 mS ∙ m⁻¹).

Fig. 7.

Mean annual pH and electrical conductivity (EC) in precipitation (P) and runoff (R).

The ionic composition of the precipitation was dominated by Ca²⁺ and S–SO₄²⁻ (Fig. 8A). The mean DL_P over the multi-year period was 42.39 kg ∙ ha⁻¹ (ranging between 29.2 kg ∙ ha⁻¹ and 83.7 kg ∙ ha⁻¹). The lowest DY_P recorded was for Mg²⁺ (1.61 kg ∙ ha⁻¹ ∙ a⁻¹), while the highest was for Ca²⁺ (10.16 kg ∙ ha⁻¹) (Fig. 9); this is related to the ‘washing out’ of the troposphere by precipitation after increasingly long periods without precipitation due to the accumulation of fine dust in the troposphere (Table 3). N–NO₃^- exhibited the lowest coefficient of variation in mean annual values (32%), while the highest value was associated with Cl^– (87.5%). There were statistically significant differences (p < 0.05) in Na⁺, K⁺, N–NH₄⁺ and pH between the first and the last decades of the study period. In addition, higher mean monthly concentrations of Cl^–, Mg²⁺, Na⁺, N–NO₃^– and S– SO₄²⁻ were observed in winter compared with the warmer periods.

Fig. 8.

Average ion concentration (%) of in A – precipitation and B – runoff.

Fig. 9.

Variation in the content of individual ions in dissolved yield precipitation (DY_P) and dissolved yield runoff (dY_r) between 1995 and 2023.

The mean annual pH in the runoff was 7.94 (ranging between 7.24 and 8.68; Cv = 5.3%) and exhibited a statistically significant decreasing trend of 0.3 per 10 years (α = 0.05). Similarly, the mean annual EC was 29.8 mS ∙ m⁻¹ (ranging from 17.5 mS ∙ m⁻¹ to 41.7 mS ∙ m⁻¹; Cv = 25.2%), with a decreasing trend of 5.1 mS ∙ m⁻¹ per decade. The mean annual DL_r was 859.9 kg ∙ ha⁻¹, ranging from 206.1 kg ∙ ha⁻¹ (2012) to 1729.3 kg ∙ ha⁻¹ (2010). The range of changes in the DL_R of individual ions over the multi-year study period is presented in Fig. 8B. The dominant ion in the Bystrzanka water was HCO₃^– (Fig. 9; average of 516.56 kg ∙ ha⁻¹ ∙ a⁻¹; Cv = 33.4%), though the ionic load of this species decreased by 17% in the last decade (2014–2023) compared with the first decade (1995–2004). A decreasing trend in the DY_R was observed for all ions and, in the case of S– SO₄²⁻, N–NO₃^–, K⁺, Mg²⁺, N–NH₄⁺, and P–PO₄³⁻ (Table 3), it was statistically significant. In addition, statistically significant differences (p < 0.01) were found in the DY_r between the first and last decades of the study period (except for Na⁺ and HCO₃^–). There was a statistically significant decrease in the mean monthly load of all species due to the decrease in water flow (except for N– NO₃^–, N–NH₄⁺). Furthermore, the loads of Cl^–, N– NO₃^– and S–SO₄²⁻ were higher in the colder half of the year compared with the warmer half of the year by 10%, 65% and 11%, respectively. This was influenced by the runoff regime (slightly higher but more stable runoff in winter) as well as the input of organic mineralisation products from the catchment.

The mean value of the total nutrient concentration (N–NH₄⁺, N–NO₃ and P–PO₄³⁻) for the studied period in the Bystrzanka stream was 1.94 mg ∙ dm⁻³. There was a statistically significant decrease in nutrient loads during the study period (Fig. 10). The average nutrient load in the first decade was 2.22 mg ∙ dm⁻¹, which decreased by 38% in the period between 2014 and 2023.

Fig. 10.

Changes in individual nutrients (P–PO₄³⁻, N–NO₃^–, N–NH₄⁺) and total nutrient loads in the Bystrzanka stream between 1995 and 2023.

The chemical denudation balance varied across the multi-year study period (Fig. 11), ranging from 177 kg ∙ ha⁻¹ (2012) to 1688 kg ∙ ha⁻¹ (2010) with a mean of 817 kg ∙ ha⁻¹. The year 2012 was an extremely dry year with a Rc of only 13.4%, while 2010 was characterised by record runoff (Rc = 73.1%). Statistical analyses (Pettitt test) revealed a statistically significant point of change in the chemical denudation balance in 2011. The average denudation value before 2011 was 939 kg ∙ ha⁻¹; in the years following, this value decreased to 668 kg ∙ ha⁻¹. The reason for the changes may have been lower precipitation since 2011 and lower average runoff by 32%. In addition, changes in LULC affected the amount of dissolved matter carried out from the catchment. A comparison of the two periods (i.e. the first and last decades) exhibited statistically significant differences (p = 0.045). The average value in the first decade was 927 kg ∙ ha⁻¹, 23% higher than in 2014–2023.

Fig. 11.

Chemical denudation balance in the Bystrzanka catchment between 1995 and 2023.

A multi-year decrease in DL_R was observed for all ions, while there were multidirectional changes observed in the DL_P: Na⁺, K⁺, Ca²⁺ and Mg²⁺ loads increased, while N–NO₃^–, S–SO₄²⁻ and N–NH₄⁺ loads decreased. The decrease in the chemical denudation balance was primarily influenced by changes in the ion composition of the stream.

SSC values varied across multiple flood events, reaching values of up to 45 g ∙ dm⁻³ during the extreme flood in 2010, when flows reached 57.7 m³ ∙ s⁻¹ (Fig. 12A). Such high SSC values were associated with natural factors (high daily precipitation and record flow) combined with anthropogenic factors, including engineering works that resulted in the accumulation of loose material both near and within the riverbed itself. Higher SSC values were recorded during summer floods over the study period. In addition, the SSC peaked both before (clockwise loop – supply of material mainly from the channel; Fig. 12A) and after (counter-clockwise loop) the peak discharge (Fig. 12B). Higher SSC values following peak discharge are a result of the delayed delivery of material from the upper part of the catchment, primarily as a product of the erosion of the stream channel and its tributaries, the erosion of roads and roadside ditches and from material accumulated in the vicinity of the channel due to anthropogenic activities.

Fig. 12.

Plot of SSC in the stream against discharge (Q) and precipitation (P) during the A – 3–4 June 2010 and B – 16–17 May 2014 flood events with a hysteresis loop and an example of concentration on an unpaved road in the Bystrzanka catchment. SSC, suspended sediment concentration.

The mean SSL between 1995 and 2023 was 3521 Mg ∙ a⁻¹ (ranging from 48 Mg ∙ a⁻¹ to 21,658 Mg ∙ a⁻¹) (Fig. 13). Statistical analysis (Pettitt test) revealed a statistically significant point of change in the SSL regime in 2016. The year 2010 experienced the highest SSY of 1666 Mg km⁻² ∙ a⁻¹ and the highest stream runoff of 896.7 mm: 95.8% of the annual SSL was removed during two consecutive flood events (May–June).

Fig. 13.

DL_R and SSL changes against runoff (R) between 1995 and 2023. DL_R, dissolved load runoff.

There was a statistically significant relationship between the runoff above the flood threshold and the SSY during the analysed period (Fig. 14). In addition, the SSY was found to be 317 Mg ∙ km⁻² ∙ a⁻¹ in the first decade (1995–2004); this was halved in the last decade of the study period (2014–2023; 125 Mg ∙ km⁻² ∙ a⁻¹).

Fig. 14.

Dependence of SSY and DY_R on the runoff above the flood threshold between 1995 and 2023. DY_R, dissolved yield runoff.