Influence of the Precipitation Pattern on the Water Runoff Dynamics From the Loess Gully in the Lublin Upland (Eastern Poland) in the Years 2003–2022

Atmospheric precipitation was recorded using a heated TPG-023 digital pluviograph from A-STER, with a precipitation step of 0.1 mm (resolution) and a time step of 1 s. Such recording made it possible to determine the instantaneous rainfall intensity, important for statistical calculations of rainfall-runoff relationships. Measured data for calculating water runoff was obtained from a hydrological station installed at the mouth of the gully, consisting of a Thomson triangular weir, a water gauge patch and a digital recorder ‘Thalimedes’ developed by OTT company (Fig. 2).

Recording of data done in HYDRAS software (OTT HydroMet), occurred every 10 min with an accuracy of 1 mm. The water flow is calculated from the water levels according the Thomson’s formula (Byczkowski 1999): 1Q=1.4h5/2Q = 1.4{h^{5/2}}

where: –

Q – water flow [m³/s],

Due to the number of data, they were usually compiled on a monthly basis, according to hydrological years, covering consecutive months from November to October of the following calendar year. The warm half of the year covers the period from May to October (V–X) and the cold half-year from November to April (XI–IV). Effective precipitation includes all precipitation events which triggered water runoff from gully catchment. As a rule, water runoff was recorded with rainfall above 20 mm and sometimes above 10 mm, therefore these rainfall thresholds were used to distinguish days with precipitation.

Maximal rain intensity is obtained directly from the pluviography and rain efficiency is calculated as quotient of rainfall sum and rain maximal intensity. Runoff surface index and runoff coefficient calculated from the relationship of total runoff to catchment area and total rainfall.

The basic statistical analyses were performed in MS Excel. It was also used to fit linear models (rectilinear and curvilinear) to the distributions of the analysed parameters: rainfall and water runoff, and to determine the determination coefficient (R²), allowing to estimate the quality of the model. On the contrary, the Statistica program (TIBCO Software Inc.) was used to perform linear correlation analysis, Statistica program adjust the linear regression models and their quality parameters and generate non-linear models using the segment regression equation: 2y1=a1x+K(x<=b) and y2=a2x+K(x>b){y_1} = {{\rm{a}}_1}{\rm{x}} + {\rm{K}}(x < \; = {\rm{b}}){\rm{ and }}{y_2} = {{\rm{a}}_2}x + {\rm{K}}({\rm{x}} > {\rm{b}})

where: –

a₁, a₂ – regression coefficients,

In the segmented regression analysis, the quasi-Newton estimation method and the number of iterations = 50 were applied. The breakpoint (b) was determined automatically and according to the user, so that the explanation of the variance of the variable was the largest. A standard level of statistical significance was also adopted for the generated models, i.e. a = 0.05. Due to the amount of data, regression analysis was performed for water runoff on an annual, seasonal and monthly basis.

The several-year period preceding the 20-year registration series was characterised by high annual precipitation totals and extreme events, such as intense snowmelt in the spring of 1996 and two heavy rainstorms, in September 1997 and June 1999. Numerous active erosion and piping landforms remained, including the dissection of the bottom of the gully (Rodzik, Zgłobicki 2000, Zgłobicki 2002). Weather conditions varied throughout the study period. There were large fluctuations and contrasts in the annual and seasonal distribution of precipitation. These are distinctive features of the ‘capricious’ climate of the Lublin Upland (Siwek 2006, Kaszewski 2008, Janicki et al. 2014, Janicki 2016).

In terms of annual precipitation total sums, according to Kaczorowska’s (1962) classification, normal years (90–110% of the average) prevailed in the study catchment between 2003 and 2022. Precipitation total sums, close to the average, occurred in as many as 12 cases, while 4 years were classified as wet (111–125% of average), 2 as extremely dry (below 50% of average), and 1 each as dry (75–89% of average) and extremely wet (above 150% of average). There were no years classifed as very dry (50–74% of average) and very wet (126–150% of average). In recent years, the spread of precipitation sums has increased, as evidenced by the occurrence of both extremes of annual precipitation total sums in the last 3 years of the 20-year measurements period (Fig. 3).

Fig. 3.

Distribution of annual sums of precipitation and annual volumes of water runoff from the gully catchment for the years 2003–2022; precipitation classifcation according to Kaczorowska (1962): Vd – extremely dry year, D – dry year, N – average, W – wet, Vw – extremely wet and blue dotted line means the average precipitation for total period.

Precipitation of the warm half-year, from May to October, was predominant (average 382 mm), with totals ranging from 175 mm to 535 mm (Fig. 4). The precipitation of the cold half-year (from November to April) reached values from 115 mm to 250 mm (average 190 mm). Monthly precipitation maxima occurred only in the warm half-year and the month with the highest precipitation most often was July (average 82 mm) with a maximum of 213.6 mm in 2011.

Fig. 4.

Distribution of half-year precipitation and water runoff from studied gully catchment in Kolonia Celejów in 2003–2022; c-p – cold half-year, w-p – warm half-year.

In August 2006, precipitation total sums were as high as 234 mm, a record for the entire 20-year series of surveys. Probably this is an absolute record, as this month saw a then-record monthly precipitation sum at many meteorological stations in the Lublin region. In the study catchment, the sum of monthly precipitation repeatedly exceeded 100 mm such as 138.8 mm in September 2010, 140.1 mm in June 2013, 130.2 mm in May 2014 or 116.2 mm in October 2020 (Fig. 5). Such a spread of maximum precipitation total sums characterises the transient nature of the Lublin Upland climate (Siwek 2006, Kaszewski 2008, Janicki et al. 2011). The lowest monthly precipitation total sums were recorded mostly in February (mean 29 mm) and sometimes also in November, October or January.

Fig. 5.

Monthly precipitation total sum and runoff between 2003 and 2022 in the studied gully catchment.

Effective daily precipitation, resulting in surface runoff, began with totals of 7–10 mm, depending on temporary environmental conditions (Table 1).

Daily precipitation >10 mm occurred from 6 to 25 times a year, with an average of 15 cases (Table 1), typical of the Lublin region (Kaszewski 2008). Precipitation >20 mm per day and intensity >0.5 mm per min in the nature of torrential downpours was recorded several times a year (an average of 4 per year). During the 20-year study period, there were no extreme heavy precipitation events with high totals and high intensity at the same time. Also, long periods (seasons) of precipitation, with above-average rainfall totals, shaping the dynamics of runoff were not observed. Meanwhile, the occurrence of several precipitation events in direct succession was recorded, while between 2003 and 2006, thaws and melts were superimposed with rainfall, accelerating the thawing of snow (Table 2).

In this context, the greatest hydrological impact was caused by frontal torrential rainfall, which occurred in 2010, 2014, 2016 and 2021. The largest temporary water runoff was generated by single rainfall events with a yield of >30 mm (Table 2).

In the first half of the monitored 20-year period, annual values of water runoff ranged from 43 m³ in 2008 to almost 40,000 m³ in 2003 (Table 3). The extremely dry year of 2003 maintained the base runoff of water stored in the loess cover in previous years (Fig. 6). In 2003–2006, cold halfyear runoff (4000–35,000 m³, with an average of 9200 m³), associated with snowmelt, predominated. The highest snowmelt runoff usually occurred in March (Fig. 6).

Fig. 6.

Monthly water runoff sums distribution from the loess gully catchment Kolonia Celejów in the 20-year period 2003–2022.

In 2003, mid-winter January melt (in the middle and end of the month) was also productive, with an effect enhanced by rainfall. Warm-season half-year runoff totals in those years (13003000 m³) were only comparable to some monthly totals: in May 2005 (1700 m³) and august 2006 (1200 m³). In the following years, 2007–2009, runoff was extremely low, with no runoff at all in many months (Fig. 6), while in 2008 it occurred only in January and May. after this calm period, there was a surge in the volume of water runoff in the following years, boosted in part by the stabilisation of the baseline runoff. The average annual water runoff in 2010–2018 was 40,000 m³, and in 2017 a maximum runoff about 76,000 m³ was recorded. The relatively long, wet period was predominated by warm half-year outflows, ranging from about 7400 m³ to about 4200 m³ (average about 2300 m³), mainly in May (2014), July (2011, 2017) and August (2017). In the cold half-year, the highest runoffs were recorded in January and February (2014), despite the periodic disappearance due to freezing of the base runoff stream. In the last 20-year measurement period of 2019–2022, the average annual water runoff slightly decreased to a value close to the average for the entire measurement period (almost 2500 m³). However, its fluctuations in different scales: annual, seasonal and monthly increased significantly (Fig. 6A) The highest runoff in this period occurred in 2021 and amounted to about 35,000 m³ (Table 2). The runoffs of the warm and cold half-years were similar, ranging from about 5000 m³ to 21,400 m³ with an average of 14,400 m³. Cold half-year runoffs were relatively even, with predominating snowmelt runoffs in February and March. Warm half-year runoffs reached similar values, but their monthly magnitudes varied more (Fig. 6B). Runoffs predominated in August (2021), October (2020) or May 2019 and 2022 even in September.

Periodicity in water runoff from the catchment, which occurred throughout the 2003–2022 measurement period, became apparent in the form of four periods with different precipitation regimes and consequently, with different volumes and patterns of water runoff (Figs 7A–C). The 2003–2006 period was dominated by runoffs of the cold half-year (4–35,000 m³, with an average of 9200 m³), associated with snowmelt.

Fig. 7.

Average monthly precipitation total sums and runoff in the gully catchment: A – nival period 2003–2006, B – steady period 2007–2009, C – pluvial period 2010–2018, D – variable period 2019–2022.

The highest snowmelt runoff usually occurred in March (Fig. 7B). In 2003, mid-winter January snowmelt (in the middle and end of the month) was also productive, with the effect enhanced by rainfall. In 2010–2018, annual precipitation total sums were high, ranging from 541 mm to 711 mm (Table 4). The number of days with precipitation >10 mm was highly variable (from 6 to 25), but daily total sums did not exceed 41 mm. Against this background, the very wet year 2010 stood out, with abundant precipitation and high annual total sums. In general, an upward trend in water runoff from the catchment prevailed (Fig. 7C). Winters were not very snowy; an average of 64 days with snow cover occurred. The lack of snowmelt runoff was compensated by numerous summer rainstorms, generating rapid surface runoff (Kociuba et al. 2014, 2015), conditioned by earlier ground saturation. Such a relationship is well illustrated by the case of the precipitation/runoff relationship in May/June 2013 (Fig. 7B). In the following years, 2019–2022, annual precipitation total sums were the most variable, ranging from 342 mm to 728 mm, making the averaged precipitation conditions close to the multi-year average (Table 4). The exceptional year in this regard was 2020, with a maximum number of days with rainfall of >10 mm and a maximum annual rainfall total of 728 mm. Winters were generally low in snow, with cover lingering from 5 days to 59 days. Despite this, the proportion of pronivial to propluvial runoff was equalised in the annual water runoff (Table 4). The year 2022 was exceptionally dry, and despite the occurrence of snowmelt, a decrease in water runoff was observed.

Water runoff for averaged months of the entire 20-year measurement period does not differ significantly from the hydrogram for rivers of the Lublin Upland, with a complex ground-snowrain regime (Michalczyk, Wilgat 1998, Kociuba, Stępniewska 2002). The annual course distinguishes a snowmelt maximum and a secondary bipartite precipitation maximum, with climaxes in July and May. The low-flow period lasts from September to December, with a slight increase in October. This pattern of annual gully runoff shows increasing instability. This may be a response to changes in the distribution of major atmospheric circulation types observed in this part of Europe (Wibig 2001, Lorenc et al. 2012, araźny et al. 2021). The tendency, characteristic of the second half of the 20th century, to lengthen the periods between productive melts continues (Rodzik et al. 2021). Similar trends for snowmelt in the Central Volga region (Russia) are documented by Yermolaev et al. (2022). However, the lack of a consistent trend in the amount of annual water runoff is confirmed, as it was for annual precipitation totals in Central Europe during the 20th century (Niedźwiedź et al. 2009).

An increase in annual precipitation total sums (Fig. 8), especially summer precipitation, is mainly responsible for the increase in water runoff from the gully during the pluvial period.

Fig. 8.

Multi-year distribution of the precipitation total sums and annual runoff between 2003 and 2022 from the gully catchment. Position of X axis determines the average annual water runoff from the catchment.

A decrease in annual precipitation total sums and summer precipitation in 2012, 2013 and 2015, as well as a decrease in winter precipitation in 2012 and 2015 determined the trend reversal. At the same time, there were numerous rainstorms, especially in 2015 and 2017, which determined the maximum seasonal and annual runoffs. The overall decreasing trend of annual water runoff in the recent period is interrupted by the outstandingly wet year of 2020, with peak annual precipitation and high warm half-year precipitation (Fig. 8). The continuation of the increase in runoff in the following year, is the result of high summer precipitation total sums and the occurrence of downpours (Table 1). The years 2019 and 2021 are also marked by an increase in winter precipitation (Fig. 5). The exception, however, becomes the extremely dry hydrological year of 2022, but with a runoff close to the multi-year average. The precipitation of 2020 and 2021 renewed the water resources of the catchment so that a relatively high baseflow was recorded. A 2-year delay in the response to precipitation in the Lublin region was noted when analysing groundwater table fluctuations and the occurrence of low-flow period in river (Janicki et al. 2011).

Linear correlation between precipitation total sums and runoff from the gully catchment, calculated for annual, seasonal, monthly actual and averaged data, showed no statistically significant relationship between the variables (r < 0.5). also, linear regression analyses showed no statistically significant relationships. As shown, in the case of the occurrence of a series of high precipitation after a dry period and a wet period (Fig. 5), runoff is determined by the instantaneous retention capacity of the catchment. Thus, the direct effect of precipitation pattern on runoff dynamics seems clear for the 2010–2018 pluvial period, for monthly precipitation of the warm half-year.

Therefore, a non-linear regression analysis was performed for the same data (Table 5). The results of the segmented regression method used showed significant (R² > 0.5) and strong relationships (R² > 0.7) between precipitation total sums and gully water runoff. The best model fit was obtained for the hydrological year and the warm half-year (Table 5). Weaker model fit and lower coefficient of determination were found for the cold half-year and monthly data. These relationships are complex and difficult to fit into a mathematical model. An attempt to determine precipitation/runoff relationships for three gully catchments on the Nałęczów Plateau was made by Józefaciuk and Karczewski (1991), by using data from water gauge readings daily. They explained the radically different results obtained from neighbouring catchments by the variation in such parameters as the degree of afforestation of steep slopes, soil erosivity, distribution of fields and thickness of loess cover, while they did not take into account the different shape and elongation of catchments.

The maximum daily water runoffs from the gully were generated by intense snowmelt and rainstorms, especially torrential rainfall (Table 1). The total water runoff reached a maximum of 10,000 m³, while the flow rate was more than 800 dm³ · s⁻¹. The specific runoff rate reached 700 dm3 · s⁻¹ · km⁻² and was similar to the rates of the river drainage catchments of low hydrological rank in the Lublin Upland (Maruszczak et al. 1984, Janicki et al. 2014, Janicki 2016). In the case of torrential rainfall, the high runoff rates were determined by the earlier saturation of the ground by the preceding rainfall (Figs 9A, B).

Fig. 9.

Response of a loess gully catchment water runoff to a downpours series depending on the preceding conditions: A – during the dry period, B – during the wet period.

However, in no case did the unit surface runoff index rates exceed 9 mm, indicating that local retention: interception, detention and infltration absorb most of the water in the water balance of the catchment. The variability of local conditions affects the variation of the runoff coeffcient (from about 1% to 28%, average >7%). This testifes to the high water capacity of loess and the considerable retention capacity of the catchment, with a system of gullies overgrown with oak-hornbeam forest. This is especially evident in the dry season, as infltration rates of loess in the dry state are high (Demczuk et al. 2022).

The catchment’s retentivity became apparent, especially during the record daily and monthly precipitation for the entire 20-year period, which occurred during the frst years of measurements. Despite the low precipitation, the basal runoff of water from the loess cover functioned in 2003–2004. The highest daily precipitation (64.5 mm) was recorded on 4 May 2005, as an all-day precipitation with signifcant runoff (Table 1). In contrast, the record monthly precipitation (234 mm) was gauged in August 2006 and consisted of 4 precipitation events >20 mm, most of which were retentive, as the summer had been very dry earlier. Disproportionately little surface runoff occurred only in the second half of August (Fig. 9A).

The role of gullies in the water balance for branching and forested gully systems in loess areas is ambiguous and depends on local conditions and precipitation patterns. It is fairly unanimously accepted in the literature that gullies are lines of periodic water runoff and drain headwater catchments (Poesen et al. 2003). Loess catchments over agricultural use with young gullies in the Moldavian Highlands (Jonita 2006) and in the Loess Plateau of China (Chen et al. 2024) reduce water storage and facilitate rapid runoff into river channels. Also, individual loess gullies in the Lublin Upland, especially those of road genesis, channel and accelerate surface runoff (Ziemnicki, Naklicki 1971, Józefaciuk, Karczewski 1991, Mazur 2008). Froehlich (1982) attributed a similar role of gullies in the riparian zone of the Carpathian catchment. On the other hand, it seems that branching loess gullies with permanent vegetation cover inhibit surface runoff and increase the retention capacity of the catchment (Mazur, Pałys 1991) as is evident from observations on gully runoff on the Nałęczów Plateau, among others (Józefaciuk, Karczewski 1991.

The Kolonia Celejów gully system, with its large number of drainless, retains rainwater from low- and moderate-efficiency precipitation. Once the rainfall total exceeds about 10 mm, rapid surface runoff (quick flow) occurs when it overlaps with the base runoff. In the study catchment, the unit runoff coefficient is low, ranging from 1% in the stable period, 2–6% in the variable period up to about 8% of annual precipitation in the nival and pluvial periods, which confirms the significant role of the degree of water saturation of the loess cover in shaping runoff. The highest values were recorded in June 2013 as a result of the superimposition of two rainstorms (clustering) under conditions of high moisture content of the loess cover in the preceding period (Fig. 7B). Studies conducted in dry denudation valleys (Ziemnicki, Orlik 1971, Janicki 2014, 2016) have shown that natural relief adapted to episodic runoff drains more easily than in the case of gullies with an undeveloped channel and uneven bottom with absorbent litter. During the 2003–2022 measurement period, the annual runoff rate in the Kolonia Celejów gully catchment did not exceed 9% of annual precipitation (Table 3). In specific years, the runoff coefficient ranged from 0% to 8.6% (mean < 4%). The highest values of this coefficient were obtained for the wet year 2017 with numerous rainstorms and the dry year 2003, but with extreme snowmelt. On the other hand, in the extremely wet year 2020, the coefficient was quite low after a dry year, in the absence of torrential rainfall, suggesting a high absorptive capacity of the loess cover.