**3. Materials and Methods**

Changes in groundwater levels in Warsaw have been studied by means of the following (Figure 3):


**Figure 3.** Factors and methods controlling groundwater infiltration and recharge in an urban area ([27], modified).

Measurements of the water table depth are made automatically with continuous control. The research station monitors three aquifers: two Quaternary aquifers (shallow and deeper) and the Oligocene aquifer (Figure 2), using 1 piezometer and 2 wells. The analysis was performed on a

27-year measurement series in hydrogeological years 1993–2019; therefore, the total number of daily observations was 9861 for each aquifer. In the frame of the groundwater monitoring system in Poland, the observation site of water levels is located in Warsaw (II-22-1, the western part of the city; Figure 1) in the deeper Quaternary aquifer. Measurements of groundwater levels have been performed at this site since 1994 [28].

Analysis of the changes of groundwater levels was performed using statistical methods most commonly applied in assessment studies of monitoring data [29–34]. The performed analysis allows the determination and characterization of statistical data and prediction of groundwater level changes.

Seasonal variability of the water level results from cyclic climate changes. It was characterized by annual amplitude, median value and analysis of charts with water table fluctuations. The annual amplitude was calculated as the difference between the highest and lowest multiannual mean water level, or as the difference between the absolute highest and the absolute lowest registered water level. The coefficient of variation, i.e., ratio of the standard deviation to the mean value, was also calculated. The measure of multiannual change of groundwater levels is the value of standard deviation of mean annual groundwater levels [35].

The magnitudes of estimated changes in the trend of meteorological variables in this study were estimated by the application of the Sen slope method. This technique calculates the gradient as a change in the measurements correlated with units of temporal change. The advantages of this method include an allowance for missing data, avoidance of assumptions about the distribution of tested data, and averting the effects of gross data errors and outliers [36]. Therefore, the method reduces the consequences of missing data and/or anomalous trends therein using the median values of the time series of various slopes that were detected as an evaluation tool [36,37]:

$$\beta = \operatorname{median} \frac{(Xj - Xi)}{(j - i)} \text{ for all } j > i \tag{1}$$

In this equation, *Xj* and *Xi* denote values dated at times *j* and *i*, and time *j* is after time *i*. The estimator β is the median overall combination of the recorded pairs for the entire dataset where trend analysis is performed, and a positive β indicates an increasing trend, while a negative value indicates a decreasing trend [37].

Assessment of the trend of changes in the water table level in the studied aquifers using the Sen slope method was based on the average monthly positions of the water table.

The magnitude of infiltration was determined based on measurements of groundwater levels. The methodology of assessing the magnitude of recharge at a regional scale is the focus of many hydrogeological investigations, and their synthesis has been presented e.g., by de Vries, Simmers [38], Scanlon et al. [39], and Assouline [40].

Observations of the groundwater levels of the shallow Quaternary aquifer (first aquifer) allowed the determination of the recharge using the water table fluctuation (WTF) method (e.g., [41,42]). The method is based on an assumption that the increase in groundwater level is caused by infiltration. In intervals with no recharge, the water level decreases. Recharge is calculated as follows:

$$R = \frac{S\_y \, dh}{dt} = \frac{S\_y \, \Delta h}{\Delta t} \tag{2}$$

where

*R*—recharge [L/T] (mm/year) *Sy*—specific yield [-] *h*—hydraulic head [l] (mm) *t*—time [T] (year)

The method is mainly used for short observation periods in areas with a shallow groundwater level, which causes quite large changes in its position over time [39].

The mean annual sum of precipitation from the multiannual interval 1993–2019 in the Warsaw area at 570 mm was used in the calculations of the infiltration coefficient [43]. The meteorological data came from the Warszawa-Ok ˛ecie Station, located in the south of Warsaw. A similar average amount of rainfall (571 mm) was recorded in the years 2001–2010 in Warsaw [44]

#### **4. Results**

The depth of the water table of the shallow Quaternary aquifer was at intervals from 3.62 to 7.76 m and a mean depth of 5.91 m over the period of 1993–2019. The deeper Quaternary aquifer had a depth of 6.40 to 10.13 m and a mean depth of 8.21 m. In the Oligocene aquifer, the extreme values of the depth to the water table were from 21.75 to 32.28 m, at a mean depth of 25.44 m (Table 1). The characteristic groundwater levels in three aquifers were determined in order to recognize the dynamics of changes indicating similarities and differences in the hydrodynamic conditions.


**Table 1.** Basic statistical parameters of the analyzed aquifers.

\* based on daily measurements.

The value of the 1st quartile (5.56) and 3rd quartile (6.28) was exceeded over 2440 times in the case of depth to groundwater of shallow Quaternary aquifer (Table 1). Exceedance of the 1st quartile was found in two periods: March 1993 to October 1996 and June 2010 to October 2012. Exceedance of the 3rd quartile was found in two periods: September 2005–March 2007 and June 2016–October 2019. The value of the 1st quartile (7.78) and the 3rd quartile (8.70) was exceeded also over 2440 times in the case of depth of groundwater of the deeper Quaternary aquifer (Table 1). Exceedance of the 1st quartile occurred in two periods: March 2006 to October 2007 and June 2009 to August 2012. Exceedance of the 3rd quartile occurred in two periods: June 1996 to July 2000 and October 2017 to October 2019. The Oligocene aquifer is different compared to Quaternary aquifers in the case of the 1st and 3rd quartiles. The 1st quartile was exceeded from October 2013 to October 2019, while the 3rd quartile was exceeded from January 1993 to July 1999.

Taking into consideration the coefficient of variation, it can be seen that variability of depth of the groundwater table is statistically significant in the case of shallow Quaternary and Oligocene aquifers (Table 1).

A similarity can be observed between the Quaternary aquifers with regard to low groundwater depths. The amplitudes (WNG-NNG; highest low groundwater depth to lowermost low groundwater depth) were similar in the first and second aquifers and reached a maximum of 3 m (Table 2). In both aquifers, there were similar differences between the mean depth to the water table in the multiannual interval and the mean depth at low levels (SNG; average low groundwater depth). In the case of the Oligocene aquifer, the amplitude (WNG-NNG) exceeded 10.5 m, whereas the difference between the mean depth to the water table in the multiannual interval and at low levels was 1 m. This indicator points to the similarity of hydrodynamic conditions in the Quaternary aquifers and the distinctiveness of the Oligocene aquifer.


**Table 2.** Characteristic groundwater levels.

The ratio of annual mean to multiannual mean is 0.17 in the first aquifer, 0.20 in the second aquifer and 0.13 in the third aquifer, which indicates significant multiannual changes at small seasonal amplitudes.

A comparison of daily measurements from both Quaternary aquifers shows that larger amplitudes occur in the shallow aquifer (4.14 m) than in the deeper aquifer (3.73 m). Analysis of the water table position on an annual basis did not reveal cyclic changes related to seasonal fluctuations of the water table, with typical high levels in spring–summer and low levels in autumn–winter. The water table position was not affected by seasonal infiltration changes. There was no direct relationship between precipitation and groundwater levels of the Quaternary aquifers (Figure 4). The analysis of trends of long-term changes carried out using the Sen's slope method indicates a decreasing trend of the water table in a shallow Quaternary aquifer (Table 3).

**Figure 4.** Monthly average groundwater levels of Quaternary aquifers.


**Table 3.** Sen's slope trend water table position.

In the deeper Quaternary aquifer, the trend of changes of the water table position was opposite in the multiannual interval. However, detailed analysis of the groundwater levels in both aquifers shows that only in the first part of the multiannual interval (when the Quaternary aquifer was intensively exploited for industrial requirements), the water table of the deeper aquifer was at a lower level, indicating larger differences between both aquifers. After reducing the exploitation of groundwater due to the economic transformation in Poland in the early 1990s, since 1997 the water table of the deeper Quaternary aquifer began to increase systematically, and since 2001, its position has been analogous to the fluctuations in the shallow Quaternary aquifer. Similar dynamics of fluctuations in the multiannual interval 2001–2019 may indicate that after reducing exploitation in the deeper Quaternary aquifer, they are influenced by the same factors (recharge, drainage) that decide on the analogous water table position. The similarity between the water table positions of both Quaternary aquifers suggests the existence of a hydraulic connection between them in zones beyond the area of the research station (Figure 5).

**Figure 5.** Monthly average potentiometric surface of the Quaternary aquifer in well no. II-22-1 [20].

Assessment of the reliability of accumulated data allowing for spatial interpretations in the Warsaw city was achieved by analysing changes of water table positions of the deeper Quaternary aquifer in well no. II-22-1 [28], located in the westernmost part of the city (Figure 5). Comparison of the water table fluctuations in both Quaternary wells indicates their clear similarity, which allows the local observations in the research station to be transferred to spatial assessments (Figures 4 and 5, Table 3).

The potentiometric surface of the Oligocene aquifer displays an opposite trend compared to both Quaternary aquifers. The trend of changes in the potentiometric surface of the Oligocene aquifer in the multiannual interval 1993–2019 points to the gradual restoration of hydrostatic pressure. In 1993, the potentiometric surface became stabilized at 80.83 m a.s.l., whereas it presently stabilizes at 90.72 m a.s.l., which shows a rise of the potentiometric surface in the multiannual interval by about 10 m (Figure 6).

**Figure 6.** Monthly average potentiometric surface of Oligocene aquifer.

Trend analysis of the water table levels using Sen's slope method confirms the rising trend of the Oligocene potentiometric surface (Table 3).

The gradual restoration of pressure in the Oligocene aquifer results from reduced exploitation. The largest increase in the potentiometric surface was observed in 1997–2003 (Figures 6 and 7). After 2003, the potentiometric surface position changed in a much narrower range. There was excessive exploitation of groundwater from the Oligocene aquifer in Warsaw in the 1970s, reaching 50,000–60,000 m3/d, resulting in the formation of a vast depression cone. Following the protection of these waters, in 1977, the construction of new water intakes for industrial purposes was banned, and the intake by existing industrial plants was strongly limited [45,46]. At present, 53 public sources rely on groundwater from Oligocene aquifer in the Warsaw city (as of 13.12.2019) [47]. Water from this aquifer is also used by hospitals and industries requiring water of the highest quality. The introduced restrictions have brought expected effects. A similar case took place in Barcelona, in which the intensive extraction of water from deltaic aquifers caused large drawdowns, leading to seawater intrusion [48]. Since the 1970s, many industries have migrated from the city, which reduced water demand, resulting in a progressive recovery of heads.

Accepting the urban area as the watershed-scale balance area and undertaking the attempt to identify the causes of groundwater level changes on a short-term and long-term basis, the basin water budget was evaluated for the shallow Quaternary aquifer. Recharge was assessed based on the analysis of changes of the water table position monitored in the first aquifer using the water table fluctuation (WTF) method. The specific yield taken for the calculations was at 0.10 [41,49–51]. This value is typical for sediments with low hydraulic conductivity (the aeration zone comprises sands with the addition of the silt fraction and clay).

Due to the lack of seasonal water table fluctuations, annual recharge was not discussed; instead, the method was modified by assessing recharge in intervals similar with regard to the dynamics of the water table. Based on the analysis of the water table position, four intervals were distinguished, differing in water table fluctuations as reflected in the magnitude of recharge (Figure 8).

**Figure 7.** Annual amplitudes of the water table in Quaternary and Oligocene aquifers.

**Figure 8.** Water table in the shallow Quaternary aquifer in the selected research periods.

In the first interval, in the hydrological years 1993–1994, and in the third interval, in 2010–2011, infiltration was a long-term process (from 2 to 9 months), which caused a significant rise of the water table (1.2–2.9 m). The recharge coefficient was calculated as the percentage contribution of recharge (R) in the mean precipitation (P), assuming 570 mm/year of the precipitation for Warsaw. In these intervals, the annual recharge was from 130 to 160 mm, at a recharge coefficient from 23 to 28% (Table 4). In the second interval, in 1995–2009, the infiltration time was much shorter (1 to 2 months) and caused smaller seasonal fluctuations of the water table. Annual recharge was about 80 mm, and the recharge coefficient was 14%. For 16 years, the water table fell by almost 2 m. The fourth interval, in 2012–2019, was characterized by slightly lower values of parameters; infiltration was short-term (5–20 days), and despite a few such episodes during the year, systematic fall of the water table was observed. In 2019, the water table was at 7.71 m, i.e., 1.8 m below the mean value for the entire multiannual interval 1993–2019.

**Table 4.** Assessment of recharge of the shallow Quaternary aquifer using the WTF method.


In the first aquifer, only long-term recharge lasting several months can result in changes in the water table position. The observed differences in the water table position and its gradual decrease from 2012 are not caused entirely by climate change as reflected in precipitation recharge. The distribution of annual totals of precipitation did not differ significantly from the remaining intervals during this period (Figure 4). This factor is not the only reason for long-term and non-cyclic changes. The lack of typical seasonality is the cumulative effect of many elements, which are interdependent and cannot be analyzed separately.

#### **5. Discussion**

Recharge in urban areas depends on a number of factors, including climate conditions and the dynamics of their changes [52,53], lithology of subsurface sediments [54,55], topography [56], land use and land cover [52,57], and thickness of the aeration zone [55,58].

The change of climate conditions in Warsaw was recorded as anomalously high precipitation in 2017 (821 mm) and anomalously low precipitation in 2015 (297 mm) (Figure 4). Measurements of groundwater levels in the research station have indicated that this change has no direct impact on the change of groundwater levels, even in the case of the shallowest Quaternary aquifer. In all analyzed aquifers, the influence of precipitation on water table fluctuations was insignificant due to the rather unfavourable filtration parameters in the aeration zone and the impact of urban infrastructure, especially in the subsurface zone. The reaction of the water table, even after intense precipitation, was greatly delayed. Multiannual trends of changes in the position of the water table in Quaternary aquifers and a stable trend of increasing potentiometric surface in the Oligocene aquifer are notable.

In urban areas, the magnitude of recharge depends on land use. The research station is located in Warsaw's centre and was developed at a time when there was hardly any infrastructure around it. However, in the 27-year interval, numerous changes in its surroundings have taken place, particularly related to the development of the University of Warsaw, Ochota Campus infrastructure, which may have had a significant impact on groundwater recharge and flow in the shallowest zone. The recharge zone also includes more distant regions, covered by typical urban housing. New multi-family buildings, parking lots and streets were constructed, modifying direct recharge from the surface, which is typical of urban agglomerations. The construction of multi-story buildings requiring deep foundations permanently modified the structure of groundwater flow [59,60]. Some buildings required drainage during construction, which periodically changed the hydrodynamic system. This factor is difficult to estimate because of construction activities in the urban area, but has a significant influence on the local position of the water table.

In urban areas with high water demands, the hydrodynamic conditions are modified due to significant, focused groundwater sources. There are only a few groundwater sources near the research station and in all of Warsaw; therefore, restoration of piezometric surfaces was documented in the Oligocene aquifer. Leakage of the water supply system and infiltration of water from the leaking network to the aquifer, documented even in the vicinity of the research station (Figure 9), can be observed in Warsaw.

**Figure 9.** Electrical resistivity tomography profile in the research station area (January 2020).

The content of water leaking from the water supply system due to leakage or failure is roughly estimated in Warsaw at about 18% (mean value of water loss for Poland according to the Central Statistical Office for 2006), which gives the absolute value of 61,200 m3 water per day. This value

should be increased by another 18% resulting from leakage from the sewage system, which is 4160.7 km long in the Warsaw agglomeration [24]. Taking into account that these are estimates in many cases are understated, it can be concluded that a total of about 125,000 m<sup>3</sup> of water per day recharges the shallow Quaternary aquifer within this urban area. In Barcelona, the water supply network losses provide 22% of the total recharge of aquifers [48].

Water use does not increase proportionally with the increasing population of Warsaw. This may result from more economical water use due to rising costs and ecological lifestyles. This lower water source results in smaller amounts of water potentially recharging the subsurface zone due to leakage of the water system. The cumulative impact of many elements causing periodical rise or fall of recharge in urban areas allows for an assessment of their impact on the groundwater environment. In the case of the three analyzed aquifers, water level fluctuations are slow and occur over longer periods. Quaternary aquifers are characterized by multiannual intervals with higher recharge and much longer episodes of water table fall, which, compared to the entire 27-year interval, results in a falling trend towards decreasing water table levels. Such a case is very distinct in the shallow Quaternary aquifer since 2012. The Oligocene aquifer displays a trend of gradual but uneven restoration of hydrostatic pressure over the analyzed time span. The highest increase in the potentiometric surface was found in 1997–2003, after which the increase was much slower. The restoration of pressure in the Oligocene aquifer in the Warsaw area is confirmed by monitoring observations performed by the Polish Hydrogeological Survey [61].

#### **6. Conclusions**

Analysis of the changes in groundwater levels in several aquifers—two Quaternary ones and an Oligocene one—points to diverse factors controlling their location. In Quaternary aquifers, recharge conditions and dynamics of changes depend on local factors: infiltration conditions in the region of the monitoring system and development of the area, including elements characteristic of urbanized regions. In the case of the deep Oligocene aquifer, the position of the potentiometric surface depends on water circulation in the regional hydrogeological system, which covers a much larger area than the Warsaw agglomeration. Statistical analysis is a useful tool for interpreting multiannual measurements, although numerous natural and anthropogenic aspects need to be considered for its correct application. Groundwater monitoring systems usually exist in urban areas for several years, whereas statistical analysis requires long-term measurement series to analyse the trend of changes. The position of the water tables of the Quaternary aquifers across a long-term interval showed their fall, whereas the deep Oligocene aquifer displayed an opposite trend, with the restoration of hydrostatic pressure resulting in a rise in the potentiometric pressure by about 10 m.

Analysis of the changes in groundwater levels in time and factors influencing them in Warsaw does not show the impact of climate change (e.g., through extreme precipitation) on groundwater. Seasonal trends and long-term changes in levels have not been observed in the studied aquifers in the urban area.

The studies have shown that groundwater recharge and drainage in urban areas are components of natural processes related to hydrogeological and climate conditions and anthropogenic factors resulting from the groundwater source, water distribution, and changes in spatial management. When analysing the cumulative impact of these factors, often with opposite effects with regard to groundwater recharge, a relatively rapid decrease in the water table position may be observed in the shallowest groundwater aquifer. Therefore, it is critical to correctly manage water resources and increase water retention conditions through increased area covered by green zones, increased application of precipitation for urban green management and application of innovative solutions to enhance the infiltration of precipitation in high-density urban zones. Analysis of data sets containing long-term measurement series allows for a reliable assessment of the existing conditions (water efficiency), aids in planning sustainable water management and, to some degree, allows for the prediction of the effects of irrational water management or inappropriate development of urban infrastructure.

**Author Contributions:** Conceptualization, E.K. and J.J.M.; methodology, E.K., J.J.M., D.P. and A.W.; software, D.P. and A.W.; validation, E.K., J.J.M., D.P. and A.W.; resources, J.J.M. and A.W.; data curation, J.J.M. and A.W.; writing—original draft preparation E.K., J.J.M., D.P. and A.W.; visualization, D.P. and A.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received funding from the University of Warsaw.

**Conflicts of Interest:** The authors declare no conflict of interest.
