215

214

ACTA CARSOLOGICA 45/3 – 2016

ACTA CARSOLOGICA 45/3 – 2016

COBISS: 1.01

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)
Analiza prenikle vode v urbani kraški jami pod madžarsko prestolnico (budimpešta)
Katalin Fehér1, József Kovács2, László Márkus3, Edit Borbás4, Péter Tanos5 & 
István Gábor Hatvani6*

Abstract	UDC  551.444:504.5(439.151)
Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani: Analysis of drip water in an urban karst cave beneath the Hungarian capital (Bu­dapest)
Our geological heritage is increasingly threatened by anthro­pogenic activity. This is especially true of the Pál-völgyi Cave System beneath Budapest. It is among the 150 longest and at the same time most endangered cave systems in the world. The aims of the study were (i) to set up a monitoring system in the cave, (ii) to track the daily changes in the quality and quan­tity of drip water, and (iii) to determine the exposure of the cave. Monitoring was conducted at two locations in a shallow area next to a fracture zone (site name: TG) and one lying in a tectonically less disturbed, geologically more homogeneous location 20 m deeper (site YC). The data obtained in respect of 13 variables were assessed using descriptive statistics, prin­cipal component- and periodicity analyses. At first glance, it was apparent that the eight water quality parameters differed in quantity between the two sites. Furthermore, using principal component analysis it was shown that in the fractured-shallow setting, anthropogenic activity (external urban pollution, e.g. de-icing, decrease of land cover etc.) is the driving process de­termining water quality. At the tectonically less fractured site (YC) external influences originating above ground may be added to the natural karst-forming processes. The assessment of drip intensity and electric conductivity again highlighted the differences between the sites in terms of their reaction to pre­cipitation. With regard to diurnal periodicity, although pH and Eh indicated a mature periodic behavior at both sites (covering 56-65 % of the total observed time), at site TG electric conduc­tivity displayed diurnal periodicity over only 21 % of the total time, compared to 56 % at YC. All results pointed towards a conclusion that at site YC daily periodicity and water quality are much more connected to natural processes, while at site TG anthropogenic external influences suppress these.
Key words: cave drip water, hydrochemistry, karst, Pál-völgyi Cave System, time series analysis, urban pollution.
Izvleček	UDK  551.444:504.5(439.151)
Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Pé­ter Tanos & István Gábor Hatvani: Analiza prenikle vode v urbani kraški jami pod madžarsko prestolnico (Budimpešta)
Naša geološka dediščina je vse bolj ogrožena zaradi človekovih dejavnosti. To še posebej velja za jamski sistem Pál-völgyi pod Budimpešto. Sodi med 150 najdaljših in hkrati najbolj ogroženih jamskih sistemov na svetu. Cilji raziskave so bili (i) vzpostaviti sistem spremljanja izbranih parametrov v jami, (ii) spremljati dnevne spremembe kakovosti in količine pre­nikle vode, in (iii) določiti izpostavljenost jame onesnaženju. Raziskava je bila izvedena na dveh lokacijah v plitvem območju blizu razpoklinske cone (ime vzorčnega mesta: TG) in na tek­tonsko manj pretrti in geološko bolj homogeni lokaciji 20 m globlje (ime vzorčnega mesta: YC). Pridobljeni podatki 13 spremenljivk so bili obdelani s pomočjo opisne statistike, analize glavnih komponent in periodičostno analizo. Na prvi pogled je bilo očitno, da se je osem parametrov kakovosti vode kvantitativno razlikovalo med obema mestoma. Poleg tega se je s pomočjo analize glavnih komponent pokazalo, da je mesto v plitvem pretrtem območju pod velikim vplivom antropogenih aktivnosti (urbano onesnaženje, npr. sredstva proti zmrzo­vanju, zmanjšanje pokrovnih plasti, itd.), ki vplivajo na kako­vost vode. Na tektonsko manj preoblikovanem mestu (YC) se zunanji vplivis površja mešajo z naravnimi kraškimi procesi. Ocena intenzivnosti kapljanja in električne prevodnosti je pon­ovno izpostavila razlike med vzorčnimi mesti v smislu njunega odziva na padavine. V zvezi z dnevno periodičnostjo je vzorčno mesto TG kazalo dnevno periodičnost več kot 21 % celot­nega časa, mesto YC 56 %, čeprav sta pH in Eh izkazali zrelo periodično obnašanje (56-65 % celotnega opazovanega časa) na obeh mestih. Vsi rezultati kažejo na sklep, da je na mestu YC dnevna periodičnost in kakovost vode veliko bolj povezana z naravnimi procesi, medtem ko jih na mestu TG antropogeni zunanji vplivi zadušijo.
Ključne besede: jamska prenikla voda, hidrokemija, kras, jamski sistem Pál-völgyi, analiza časovne vrste, urbano onesnaževanje.

1 Department of Environmental and Landscape Geography, Eötvös Loránd University, Pázmány P. sétány 1/C, Budapest, 
H-1117 Hungary, e-mail: feher.katoke@gmail.com
2 Department of Physical and Applied Geology, Eötvös Loránd University, Pázmány P. sétány 1/C, Budapest, H-1117 Hungary, e-mail: kevesolt@geology.elte.hu
3 Department of Probability Theory and Statistics, Eötvös Loránd University, Pázmány P. sétány 1/C, Budapest, H-1117 Hungary, e-mail: markus@cs.elte.hu
4 Department of Physical and Applied Geology, Eötvös Loránd University, Pázmány P. sétány 1/C, Budapest, H-1117 Hungary, e-mail: borbas.edit12@gmail.com
5 Institute of Mathematics and Informatics, Faculty of Mechanical Engineering, Szent István University, Páter K. u. 1, Gödöllő, H-2100 Hungary, e-mail: tanospeter@gmail.com
6 Institute for Geological and Geochemical Research, Research Center for Astronomy and Earth Sciences, Hungarian Academy of Sciences, Budaörsi út 45., Budapest, H-1112 Hungary. Tel.: +36 70317 97 58; fax: +36 1 31 91738; e-mail: hatvaniig@gmail.com
*Corresponding Author.
Received/Prejeto: 01.09.2016

ACTA CARSOLOGICA 45/3, 213–231, POSTOJNA 2016

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Introduction

In the last 150 years, due to extensive urbanization and the appearance of vast metropolises, the landscape of their setting has changed drastically. With construction and mining activity, even the original relief conditions have been altered, and the indigenous flora has complete­ly disappeared. Budapest, the capital of Hungary, has been no exception: land-use and the nature of the surface cover has changed from natural to e.g. garden-suburbs or blocks of flats, and together with this building a wide va­riety of services such as the laying of pavements and the construction of roads has been provided (Mari & Fehér 1999). Trends such as these on the surface should alert scientists to the need for an investigation of sub-surface water systems, since, as a direct result the amount of con­taminants seeping into the ground has increased, eventu­ally reaching the karst- and thermal waters and causing the deterioration of their water quality (Veni 1999). 
With urbanization and its effects, geological riches (e.g. karst systems) are ever more under threat (Bolner et al. 1989; Brocx & Semeniuk 2007). This is especially true of those metropolises that are located upon remark­able geological sites. The Pál-völgyi Cave System, unique in the world, lies on the right bank of the Danube under Budapest. The cave system is 30.1 km length (Hungarian Cave Database 2016), the longest in Hungary and among the 150 longest in the world (Gulden 2016). More im­portantly, it is the recharge zone for the world heritage-designated thermal springs of Budapest. The problems discussed below are particularly important in a karst en­vironment, especially in an urban environment such as the Pál-völgyi Cave system, the subject of the research (Fig. 1A, 1B).
If the cover soil is intact above the karst, it is able to filter pollutants arriving with the infiltrating waters. If, however, this upper surface is damaged by e.g. silvi­culture, construction, mining or agricultural activity, these waters may reach the lower layers without facing any obstacles (Bolner 1995), guided mainly by the com­plex karst fracture system (Williams 2008). These waters will eventually manifest their anthropogenic origin in the sub-surface environment in drip water (Kern et al. 2009; Baldini et al. 2012; Hartland et al. 2012), or in spe­leothems (Kogovšek 2011; Siklósy et al. 2011; Baker & Fairchild 2012) in karst caves (Fairchild & Baker 2012). These characteristics typical of anthropogenic activity can be traced mainly in the elevated concentrations of ions (e.g. Na+, SO42-, Cl-, NO3-) present in the water and high electric conductivity values (Hem 1985) as “con­taminants” with respect to the natural karst environ­ment. The electric conductivity of drip water is one of the best indicators of pollution occasioned by anthropo­genic activity in the hydrological system of such caves. Its analysis in terms of intensity and chemical compo­sition already has a long history (e.g. Genty & Deflan­dre 1998; Baker & Brunsdon 2003). In an undisturbed setting, conductivity values should change together with bicarbonate concentrations. However, when the system is disturbed1, e.g. by ions originating from sewage waters (Kogovšek 2011), conductivity does not change accord­ingly. If a whole cave is located in a geologically similar setting and large differences are observed in the drip wa­ter’s conductivity at different monitoring sites, it may be reasonably suspected that this is a result of external pol­lution and different sources of drip water. Generally, the least likely explanation is that these phenomena result from the geological setting.
Following the natural processes and tracking the changes induced by the previously discussed phenom­ena is quite straightforward. Local monitoring stations (See section 2.2) need to be set up to analyze the drip water (Kogovšek 2011; Baldini et al. 2012; Hartland et al. 2012; Kogovšek & Petric 2013; Liu & Brancelj 2014). In the area, systematic measurements of drip water quality have been carried out since the mid-1980s (Bolner et al. 1989; Mádl-Szőnyi et al. 2007; Fehér 2009). The results of these analyses indicate a continuous increase in the concentrations of various polluting components, again providing motivation for further, deeper inspection.
In the light of the preceding considerations the au­thors felt compelled to formulate the following goals: (i) the setting up of a monitoring system in the Pál-völgyi Cave System – subjacent to anthropogenic activity on the surface, (ii) the tracking of daily changes in the qual­ity and quantity of drip water, and (iii) the determination of the cave’s exposure, with the aim of providing an ex­ample for the further examination of other caves in such settings.

1 A system is disturbed when its natural regime is changed by anthropogenic influences.

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)


Fig. 1: Location of the Buda thermal karst and the study area in the Rózsa Hill in Budapest, Hungary A), the two sampling sites marked within the Pál-völgyi Cave System (red lines) B) and the conceptual model of the Rózsa Hill discharge area C), where 1: Permian-Lower Triassic evaporitic-carbonate strata; 2: Triassic carbonates; 3: 119 Szépvölgy limestone; 4: Buda marl; 5:Tard clay; 6: Kiscell clay; 7: Miocene formations; 8: local/intermediate flow systems; 9: regional flow systems; 10: regional karst + basi­nal fluids; 11: basinal fluids; 12: complex water-rock interaction; 13: structural elements (based on Erőss et al. 2012).

Materials and methods

Site description
The study area, The Buda thermal karst –and as a part of it the Pál-völgyi Cave System - was chosen to illus­trate the exposure of caves to urban anthropogenic activ­ity. It is one of Europe’s largest functioning thermal karts systems, and a part of the main karst water reservoir of the Transdanubian Central Range (Fig. 1A). The Meso­zoic carbonate sequence in certain parts of the range can reach a thickness of a couple kilometers, ensuring petro­logical continuity, and thus providing the hydrodynamic connection between the thermal karst and the other sub-areas (Mádl-Szőnyi & Tóth 2015). One of the regional discharge areas of the range’s aquifers is the Buda thermal karst, and as part of this, the Rózsa Hill itself (Fig. 1A), the location of the study site. This is a hypogene karst sys­tem on the border of an emerged carbonate block and a sedimentary basin (Erőss 2010; Erőss et al. 2012; Mádl-Szőnyi & Erőss 2013; Fig. 1C). The Buda thermal karst has developed in a regional discharge area of the aquifer system of the Transdanubian Range. However, the range of hills of which the Rózsa Hill area is a part serves as local recharge area, and this is where the Pál-völgyi cave is found. 
On the Rózsa Hill, because of periglacial processes, the eroded rocks produce a vast amount of clastic rock. However, as a result of downhill creeping, the sedi­mented loess mixed with clastic rock, giving a diverse (fine/ coarse) cover on the karst strata. Thus, it can be concluded that the friable carbonate bedrock and the previously mentioned cover in certain settings func­tions as an epikarst and moderates infiltration/pollu­tion (Virág et al. 2011). These Rózsa Hill coarse strata, only partially function as aquiclude layers, as supposed earlier by Mádl-Szőnyi et al. (2007). In the present case, beneath the coarse strata a Buda Marl formation can be found (estimated macroporosity: 1.5 %; Kleb et al. 1993). Its hydrological characteristics are determined by faults (Gáspár et al. 2015) and beneath the marl a Szépvölgyi Limestone Formation (matrix porosity: 5-10 %; Albert 2010) can be found. 
The Pál-völgyi Cave is located mainly in the previ­ously mentioned limestone, but extends into the marl as well. Its system of passages follows the 25-30° dip of the limestone and marl. The thickness of the marl above the cave is 20-70 m in general (Gáspár et al. 2015) and 40-60 m above the sampling sites.
As discussed above, the land use –as is true of Buda­pest in general- has changed on the Rózsa Hill as well. In the 18th Century, approx. 40 % of its area was covered by forests and bushes; by the mid-1980s this had decreased to 8 %. In the meanwhile, the proportion of built-up area increased from 2 % to 85 %, reaching 90 % by the turn of the 21st Century (Mari and Fehér 1999; Fig. 1B).
Thus, caves beneath such areas function as unique natural laboratories to gather further information about the current hydrological characteristics (water retention/conductance) of the coarse cover strata.
Sampling sites and measurements
One possible way of determining the threat of pollution in the Rózsa Hill area is the chemical analysis of drip wa­ter. The physico-chemical composition of drip water in a sense functions as tracer, reflecting the characteristics of infiltrating waters of natural and/or anthropogenic origin, e.g. sewage waters coming from damaged drain pipes or other washed-down pollutants. The cave galler­ies are located halfway between the surface and the karst water table (average depth 50 m). This setting provides ideal conditions for the detection of possible pollution from the surface before it reaches the water table.
Two monitoring sites were chosen for the study. One is located in the Térképész Gallery (TG) 40 m be­neath the surface, next to the tectonically fractured zone of Szép Valley, while the other is the Y Corridor (YC), 60 m deep, lying beneath a declivous pediment in a geo­logically more homogeneous location than TG (Fig. 1B; cross section: Fig. A1).
Based on research experience gained in the course of previous monitoring campaigns in the cave, a decision was made to commence the systematic measurement of a much wider set of parameters instead of the previous occasional sampling (Bolner et al. 1989; Fehér 2009). In this way, it should be possible to separate natural geo­chemical processes from the anthropogenic ones. Thus, the following monitoring strategy was developed.
The measurements can be categorized into two groups: (i) continuous high frequency (hourly) and (ii) bi-weekly. In the first, pH, Eh (mV) and electric conduc­tivity (EC; µS cm-1) were measured, while in the course of the bi-weekly measurements water quality (Na+; K+; Ca2+; Mg2+; Cl-; HCO3-; NO3-;SO42- (mg l-1) and drip inten­sity (ml h-1) measurements were performed on samples taken from the two sites. The hourly-measured data were retrieved during the collection of the bi-weekly samples. This was carried out for both sites from 15.07.2013 to 13.04.2014.
An ODEON Range data collector (Fig. A2) and the sampling instrument were set up so the drip water in the cave would fall into its funnel and, passing along a plas­tic tube, reach the bottom of the sampling cell. When the cell is full the water goes into the sampling bottle (Fig. A2). This way water circulation and the constant water coverage of the electrodes is ensured.
In addition to the collected data, the daily sum pre­cipitation data (mm) of the Ferenchegy meteorological station (N47°31’01.22”; E19°00’56.06”; ELEV: 230 me­ters above Baltic Sea level; NLCO 1975), were acquired from the Hungarian Meteorological Service for the time interval 15.07.2013 to 13.04.2014. It should be noted that snow – only 1 cm – fell on only one day (01.02.2014), so the delay effect due to snow melt (i.e. a change in aggre­gate conditions) could be disregarded.
Methods used
The different sampling frequencies of the measured pa­rameters clearly determined the set of methods that could be employed to assess their time series. On the one hand, in the case of the bi-weekly water quality sam­ples it was sufficient to use principal component analysis (PCA) to find the driving processes of drip water qual­ity (for details see the following section). On the other hand, the high frequency (hourly) time series complied with the requirements of the analysis of periodic/spectral components to investigate the different sites’ exposure to anthropogenic effects.
Principal component analysis
Although the water quality indicators correspond to tem­porally changing processes, the less frequent bi-weekly measurements do not show any detectable temporal structure: the consecutive samples can be considered as being independent of each other; thus, the conditions for the application of PCA are satisfied (Hatvani et al. 2015; Kovács et al. 2015). The choice of PCA over factor analy­sis is explained by the fact that the latter decomposes the covariance structure, while PCA breaks the full variance down into principal components without loss of infor­mation (Abdi & Williams 2010). Since the amount of contaminants and other water quality indicators is highly relevant, it would be meaningless to scale them to unit variance. Hence a covariance matrix is preferred in PCA over the correlation matrix. It should be noted that PCA does not automatically decide how many principal com­ponents should be considered satisfactory in the aim of representing variability (Kaiser 1960).
Periodicity analysis
The high frequency (hourly) sampling provided an op­portunity for periodicity analysis. A multitude of meth­ods is suitable for the determination of a periodic or spec­tral component of a time series. Stationary time series can be relatively easily analyzed using a standard Fourier transformation; this method is, however, incapable of telling us when those frequencies occurred. The simple Fourier analysis presupposes the permanent presence of oscillating components with constant amplitude. How­ever, in a propagating process the amplitude of a spectral component is not necessarily invariable over time. This fact calls for time-frequency analysis of the power spec­tra, in order to study the temporally localized behavior of a time series in the frequency domain. Some of the most frequently used procedures are the Short Term Fou­rier Transform (STFT), in which the traditional Fourier Transform is multiplied by a fix-sized sliding window, or multi-resolution analysis or wavelet analysis, which may also be regarded as the enhancement of the STFT (Gröchenig 2001). 
As discussed by Kaiser (1994), the STFT, as with other windowed Fourier transformation techniques, represents an inaccurate and inefficient method of time–frequency localization, as it imposes a window length or “response interval” on the analysis. For analyses where a predetermined scaling may not be appropriate, because of a wide range of dominant frequencies, a method of time–frequency localization that is scale-independent, such as wavelet analysis, should be employed. Wavelet transformation was established in order to further the quest for balance in time and frequency resolution (Tor­rence & Compo 1998). The wavelet approach is localized in time and scale (frequency), resulting in a time-scale (time-frequency) decomposition of the signal using scaled localized oscillating functions, the wavelets. By fragmenting the data into short dampening “wavelets” instead of infinitely oscillating long sine waves, the wave­let transform can be used to analyze time series that contain nonstationary power at different frequencies (Daubechies 1990).
In the first attempt, the applied STFT signaled clearly that the presence of the diurnal component is not permanent (for details, please see Appendix 3). In order to refine the results and utilize significance levels for the more exact detection of the periodic components, wavelet spectrum analysis (WSA) was used. One of the frequently used basic wavelets also employed in the computations is the Morlet wavelet (Morlet et al. 1982), obtained by localizing a complex sine wave with a Gauss­ian envelope (Fig. 2). The wavelet transformation Wn(s) may be defined as the convolution of the data and the wavelet function (1):

	(Eq. 1)
in which the asterisk (*) represents the complex con­jugate, ‘Xn’ the original data stream, ‘s’ the scale, ‘.’ the basic (mother) wavelet function and ‘.’ the degree of the resolution. “The mother wavelet provides a source function to generate daughter wavelets, by scaling and translating it. This way it creates a self-similar structure, and as a result the data’s periodic component over time can be examined” (Kovács et al. 2010). This enables the computation of the proportion of the presence/detecta­bility of the daily periodic component in the observation days. Note here that the output graphs of WSA have edge artifacts, since the wavelet is not completely localized in time. Therefore, as is the general practice in wavelet analysis (Torrence & Compo 1998) a cone of influence (COI; i.e. the region where edge effects cannot be ignored meaning that wavelet power estimation is reliable) is in­troduced.
Before applying wavelet spectrum analysis, it was necessary to get rid of any trend contamination by re­moving it using a locally estimated scatterplot smooth­ing (LOESS) (Cleveland 1979; Cleveland & Devlin 1988). This removed the large, long-term fluctuations, while preserving the daily oscillations. 
All mathematical and statistical computations were performed using R 3.2.3 (R Core Team 2013), IBM SPSS 20 and the saturation indices were calculated using Phre­eqc for Windows. For the visualizations of the results, CorelDRAW Graphics Suite X7 and MS Office 2013 were used.

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)


Fig. 2: Graph of the Morlet wavelet.

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Results

Water quality analysis on the data between July 2013 and April 2014
Overview on groundwater chemistry 
As a first step, clear differences were brought out regard­ing their hydrogeochemical characteristics by plotting the main ion composition of the drip- water at the two sites on Piper diagrams (Piper 1944). It turned out that the samples of drip water taken at site TG belong to the Na-Ca and Cl-SO4 facies (Fig. 3A), while those from site YC belong to the Ca-Na and Cl-SO4-HCO3 facies (Fig. 3B) according to the nomenclature of Back (1966).
In other words, at site TG Na and Cl, while at site YC Ca and Mg, and SO4 and HCO3 are dominant. There­fore, in the case of site YC the concentration of ions re­sembles that of a karst system better. The most explicit difference between the sites is that the concentration of most of the anions and cations was higher at TG (Tab. 1A) than at YC (Tab. 1B) even if considered in mmol l-1 (Tab. 1), e.g. the chloride and sodium content was more than one magnitude higher at TG (Fig. A4). 
If the saturation indices are calculated, the differ­ence between the two sites is again explicit. Site TG is close to the equilibrium state regarding calcite with slight over-saturation (SIcc: 0.05-0.08). At YC, however, again, slight, but higher over-saturation is seen in the case of calcite (SIcc: 0.14-0.48), dolomite (SIdol: 0.2-0.9) and aragonite (SIar: 0.3), than at TG.
As a result of the nine month long sampling campaign and the following laboratory analysis, it was found that conductivity at site TG varied be­tween 5,300-6,000 µS cm-1, while at site YC between 900-1,100 µS cm-1(Fig. 4B). The difference between the two sites could also be observed in the case of drip intensity ranges (TG: 30-80 ml h-1, YC: 20-60 ml h-1; Fig. 4B respectively). It may also be inferred from the graphs (monotone decreasing trend from the begin­ning of the measurements), that EC was higher in summer, although measurements were not available for it during that time.
At both sites, the changes in EC and drip intensity in general followed each other. When drip intensity de­creased, so did EC and vice-versa. In the time period as­sessed, two events can be brought up as examples of the parallel increase of both parameters. The first (Event 1) occurred in November-December 2013 and the second (Event 2) in January-February 2014. The volume of change significantly differs at the two sites (Fig. 4). In the case of EC, at site TG the change was only 200 µS cm-1 during the first and 600 µS cm-1,during the second event, while at site YC it was only 10 µS cm-1 and 30 µS cm-1 respectively. A similar case was observed in drip inten­sity as well, where again bigger changes were observed at site TG (15 & 25 ml h-1) than at site YC (5 ml h-1) during both events. 
As a rough estimate, it can be seen that in the case of Event 1 it took about ~1 day for TG and ~2 days for YC to respond to the effect of the precipitation of the previous days, while in the course of Event 2 such a dis­tinction cannot be made because of the lack of outlying precipitation peaks. It should be noted that in the case of the other precipitation events no explicit change was seen in either the EC or the drip intensity parameters. However, it is suspected that in the course of both Event 1 & 2, the previously seeped-down water had not yet ex­ited the system, so the prolonged and/or smaller precipi­tation events – especially in the case of Event 2 – reached an already water-saturated system, causing a response in drip intensity and EC. The only explicit exception may be January 2014 (Fig. 4), when at site TG there was an increase in drip intensity, but without the expected de­crease afterwards, probably because of the effect of the precipitation in Event 2.
The driving processes of water quality
To explore the determining background parameters of water quality in the drip water datasets of the two sites PCA was used. The first principal component (PC) in the case of site TG explained 77 %, the second PC, however, only 21 % of the total variance. In the case of site YC these figures were 83 % and 13 % respectively, meaning that the remaining variance explained by the further PCs was less than 2 % and 4 % for TG and YC respectively, indicating that these subsequent factors may be ignored (Fig. 5).
At site TG, chloride and sodium were much more determining than all the other components. Site YC is, contrastingly, clearly bicarbonate dominated, with chlo­ride and calcium also being important in both PCs (Fig. 5B). In addition, sulfate is only important in the second PC at site YC. Thus, the most notable difference between the two sites is to be found in the parameters related to natural karst processes, since at site YC bicarbonate has large factor score in the first PC, while at site TG neither bicarbonate nor calcium has any significance in the de­termination of the first two PCs.
Periodicity analysis of the data between October 2013 and January 2014
Prior to periodicity analysis, the long-term variability was removed from the pH, Eh and EC time series of the two sites using LOESS. The characteristic shorter fluc­tuations were thus preserved intact, while the “masking” trend was omitted (e.g. see Fig. A5). As the next step, a time-frequency analysis concentrating on diurnal perio­dicity was conducted for the time series of pH, Eh and EC measured at both observation sites using WSA (e.g. Fig. A6). For each series, time intervals were sought in which the 24 hour period was detectable. Then the proportion of the number of days with daily periodicity to the overall number of observation days was calculated for each time series, and the two observation sites characterized and compared using these figures. It may be conjectured that the obtained characterization is indicative of the transfer of contaminating substances.
The proportion of time with a daily period present vs. the full time period was calculated (as in Kovács et al. 2010), and found to be of similar magnitude at both sites for pH and Eh. EC, however, was peculiar in this sense, since at site TG it indicated daily periodicity over a much smaller portion of time than at YC (Tab. 2).
As a final step, the daily mean course was deter­mined by averaging out the fluctuations for each hour of the day (Fig. 6). In the case of all three parameters, the first and most explicit difference is that at site TG the amplitude of the daily period is much larger than at YC, approx. 5.5, 3, and 3 times larger for pH, Eh, and EC, re­spectively. In addition, in the case of pH and Eh there is a clear shift between the minima and maxima. At YC the peaks of the curves occur 6h (pH) and 2h (Eh) later than at TG. As for EC, a clear and mature daily period is only present at YG. Such comparisons between the time lags would thus be inappropriate. As an additional observa­tion, pH and EC show clearly opposing patterns. When pH decreases, EC increases, and vice-versa at both sites (Fig. 6).


Fig. 3: Facies categorization of the drip water at sites TG red dots and YC blue triangles.

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)

Tab. 1: Descriptive statistics of cations and anions at TG (number of samples: 17) A) and YC (number of samples: 15) B) (15.07.2013-13.04.2014). M: mean, MED: median, SD: standard deviation, CV: coefficient of variation, MIN: minimum, MAX: maxi­mum.
A)
 
TG 
Statistic / Paremeter
 Measurement unit
 HCO3-
 SO42-
 Cl-
 K+
 Mg2+
 Ca2+
 Na+
 NO3-
 
M
 (mg l–1)
 117.8
 208.5
 1739.1
 6.6
 50.7
 308.3
 766.2
 66.4
 
MED
 116.7
 213.4
 1719.3
 6.4
 48.6
 305.4
 754.5
 66.7
 
SD
 6.6
 12.5
 79.7
 0,4
 6.4
 10.6
 55.4
 1.2
 
CV
 6 %
 6 %
 5 %
 6 %
 13 %
 3 %
 7 %
 2 %
 
MIN
 105.1
 176.0
 1648.4
 6.2
 41.7
 297.7
 709.4
 64.0
 
MAX
 128.4
 220.6
 1878.9
 7.3
 66.0
 328.3
 954.5
 68.4
 
M
 (mmol l–1)
 1.9
 2.2
 49.1
 0.2
 2.1
 7.7
 33.3
 1.1
 
YC
 B)
 
M
 (mg l–1)
 162.7
 171.2
 112.0
 3.7
 49.5
 75.7
 32.0
 77.9
 
MED
 157.6
 168.8
 109.9
 3.4
 48.6
 74.4
 32.0
 78.6
 
SD
 20.7
 13.6
 6.0
 0.4
 2.6
 5.1
 1.6
 2.0
 
CV
 13 %
 8 %
 5 %
 12 %
 5 %
 7 %
 5 %
 3 %
 
MIN
 140.1
 155.3
 106.4
 3.3
 45.2
 68.7
 29.2
 72.8
 
MAX
 198.5
 215.8
 124.1
 4.3
 53.3
 84.0
 34.8
 80.1
 
M
 (mmol l–1)
 2.7
 1.8
 3.2
 0.1
 2.0
 1.9
 1.4
 1.3
 



Tab. 2: The presence of a daily period in percentages compared to the total assessed time. 
 
 pH
 Eh (mV)
 EC (µScm-1)
 
TG
 61.8 %
 56 %
 21 %
 
YC
 65 %
 59.9 %
 56 %
 



Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani


Fig. 4:Precipitation events (mm) A) along with the Z-scores of EC and Drip intensity at sites TG (blue triangle) and YC (red square) B) for 15.07.2013 to 13.04.2014. The light blue vertical bands indicate the precipitation events that started filling up the karst and the dark blue ones the precipitation events that triggered the change in EC and drip intensity.


Fig. 5: Results of PCA at site TG A) and YC B).

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)


Fig. 6: pH A), Eh B), and EC C) average daily fluctuations of the de-trended time series, repeated four times for better visualiza­tion, at the two sites.

Discussion

On the basis of the presented analyses significant differ­ences may already be observed between the two moni­toring sites at the level of the descriptive statistics. This is particularly interesting since both are located in the same geological formation, albeit in different tectonic settings: TG is situated near the fracture zone of the Szép Valley 40 m below ground, with water reaching it quickly via seepage from the surface, while YC is located in a much less fractured rock, under a declivous pediment at a depth of 60m (Fig. A1). Therefore, water seeps down slower and in smaller quantities. In general, besides the local differ­ences, the observed values were clearly way above what might reasonably be considered as natural for both the cations and anions. The values in our study area, e.g. for (i) chloride were 2 to 30 times more, and (ii) sulfate ~3.5 to 4 times more (at sites YC and TG respectively) than the maxima in the Postojna cave (Slovenia), which is known to be affected by the residues (chlorides, nitrates, sulfates and phosphates) of the nearby military facilities. This is noteworthy even if we consider that the Pál-völgyi Cave System has a marl cover (Fig. 1C). Interestingly, the nitrate values were higher in the Postojna Cave than in the Pál-völgyi Cave System, indicating the fecal origin of the pollution (cesspools) in the previous case (Kogovšek 2011).
Regarding saturation, the system is slightly over-saturated in general. The natural processes (e.g. car­bonate-equilibrium reactions) dominating at YC (Fig. A4) are able to manifest themselves. For example in the process when the drip water exits the rock and gets in touch with the cave air – in the water collector - and its equilibrium CO2 exits the water and the becomes slightly more over-saturated. In the case of TG, however, because of the higher anthropogenic influence the carbonate-equilibrium reactions are suppressed, as seen from the lower bicarbonate values as well.
The bi-weekly drip-intensity drove EC values, re­flecting the well-known phenomenon that fallen pre­cipitation seeps in through the epikarst, bringing with it substances from the surface. This, in turn, tends to lead at first to a swift increase EC, followed by a dilut­ing of the water (Ford & Williams 2007; Liu & Brancelj 2014). The speed of the decay was different for the two sites during/after two precipitation events (Events 1 & 2 in Fig. 4), but due to the lack of continuous drip inten­sity measurements, only estimations were made. Unfor­tunately, measurements as dense as those found in Liu & Brancelj (2014) were not available for the examined period to trace the impact of precipitation on drip in­tensity on a finer scale. However, an additional special set of data was available with continuous drip intensity measurements (20.11.2014 to 17.12.2014) from the area examined (Fig. 7).
Therefore, the lag between the precipitation events and the increase in drip intensity and EC could be de­termined exactly. In the course of Event 3, drip intensity started to increase at TG after 34 h and at YC after 44 h, while the change in EC followed it one hour later at both sites (at TG after 35 h and at YC after 45 h). In the course of Event 4, the previously seeped-down water (start­ing with Event 3) had not yet exited the system, so the prolonged and smaller precipitation events reached an already water-saturated system, causing an even slower response in drip intensity (TG: 68h; YC:81h). This was somewhat similar to the phenomena shown in Fig. 4. This is the same phenomenon that presumably took place in the course of Event 2 (Fig. 4). As for EC in Event 4, at TG it took the system 290 h to respond, probably due to the fact that the high amount of substances that reached the karst during Event 3 (6,700 µS cm-1) had not yet exited the system, and at the time of Event 4 it only decreased a 100 µS cm-1, while at YC it started to increase parallel to drip intensity, due to the already known differences between the two sites (Fig.7).
Although, measurements did not cover summer, the starting points of the EC graphs at both sites are in line with the documented phenomenon that conductiv­ity is higher in summer (Batiot et al. 2003).
As seen, the EC values were different for the sites, indicating a greater degree of anthropogenic influence at TG. Compared, however, with a cave in a pristine alpine site (Austria) 60 m below ground (as is site YC), the EC values recorded even at site YC were a minimum 3.5 times higher, and in some cases more than 25 times higher (Kern et al. 2011). This indicates the in general higher degree of disturbance and exposure of the ex­plored section of the Pál-völgyi Cave System, although latter has a marl cover (Fig. 1C). At the same time, EC was found to be on the same scale as in another dis­turbed cave system in Slovenia (Kogovšek 2011). As for drip intensity, it was more variable at site TG, this origi­nating from the difference in the settings of the sites, i.e. TG is closer to the fracture zone and located closer to the surface (.h=20 m).
stochastic realtionships
With the stochastic analyses, the parameters explaining most of the variance were sought, and consequently con­nected to the main processes governing the quality of the drip waters. In general, the presence of sodium and chloride might be the result of natural processes, such as weathering, but then again, large differences in con­centrations within a cave located in a uniform geological setting alert us to the possible presence of external pol­lution. These may include, for example NaCl, used for de-icing for decades (Granato 1996), or the residues of detergents, disinfectants etc. On the other hand, elevated sulfate concentrations in drip water may be connected to natural processes of the soil (Hem 1985), or to the de­composition of organic matter of anthropogenic origin. Nitrate, however, is always an indicator of human activ­ity in the area, originating as it does in fertilizer usage or sewage waters, for example (Hem 1985; Lerner et al. 1999; Motyka et al. 2005).
In the present situation and study, these parameters were indeed related to anthropogenic and/or natural karst processes, as described in the study of Daoxian and Cheng (2009), though to a different degree at the two sites. In the case of site TG, it became evident that the parameters combined in the first two PCs are in close relationship with external urban pollution. The most pronounced were sodium and chloride originating from de-icing materials and/or domestic sources, reaching the subsurface flow (Panno et al. 2006). Natural processes do not seem to have a relevant role here at all.
On the contrary, site YC’s first PC is bicarbonate dominated - representing the natural karst processes (Ford & Williams 2007) - but chloride is also present, again indicating an external influence from the surface, such as sulfate in the second PC (Kogovšek 2011). The reason for this lies in the setting of the sites, meaning that at YC there is a much greater chance for rock-water interaction to take place due to the slow seepage caused by the more consolidated environment, while in the case of TG the water does not have time to dilute the sub­stances from the rock (Fig. A1). 
Periodic behavior
The periodicity analysis of the pH and Eh time series indicated high variability and a diurnal periodic behav­ior detectable over a substantial part of the observation period at both sites. The slight difference in the degree of detectability as well as the lag between the peaks of the periods at the two monitoring sites may be due to their geological settings, as previously described. After examining the results more closely, it became apparent that pH and EC were in anti-phase. The pH rose and EC dropped during daytime at both sites, as has been ob­served elsewhere in a karst underground river (Daoxian & Cheng 2009). These opposing patterns in the cases of pH and EC were found to hold in the bi-weekly data as well (Fig. 8A). Higher CO2 content (i) decreased pH, and (ii) increased the solution capability, total dissolved sol­ids (TDS, Fig. 8B, calculated from the measured cation and anion concentrations), and as a result, EC as well (Fig. 8 inset table).
The similarity of pH and Eh and the difference of EC in terms of daily periodic behavior at the two moni­toring sites closely reflects the situation described in the Introduction in detail, thus providing evidence of the presence of anthropogenic influence on the area. At the 

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Fig. 7: Precipitation events (mm) A) along with the Z-score trans­formed EC and Drip intensity at sites TG B) and YC C) for 20.11.2014 to 17.12.2014. The light blue vertical bands indi­cate the precipitation events that started filling up the karst and the dark blue ones the precipitation events that triggered the change in EC and drip intensity.


Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)

Conclusions

Exploring urban caves is equally important from the per­spective of tourism, as well as therapeutic and scientific uses, and is a highly important task in protecting our geological heritage. For one thing, urban caves are much more affected than ones in areas remote from human activity by factors such as increases in the proportion of built-up -areas, cracks in the drainage system, the lack of natural topsoil etc. in changing the behavior of natural karst processes. With the presented study, the explicit dif­ferences between the anthropogenically much more af­fected site (TG, near a fracture zone) and the tectonically un-disturbed site (YC) were highlighted. The processes at site TG were found to be different from the natural karst ones in operation at YC in the following ways. At TG, (i) Its water quality parameters reflected urban activity from both a descriptive and stochastic point of view to high­er extent in other anthropogenically effected sited de­scribed in literature; (ii) EC and drip intensity responded more swiftly and dynamically to precipitation; and (iii) as for periodicity, the external pollution inputs originating the urban environment, and as a minor factor, the lack, for example, of natural topsoil coverage disturbed and masked the periodic behavior which might be expected from such a karst environment. This was most explicit in the case of EC. The Pál-völgyi Cave System is of par­ticular importance since it is a key recharge zone of the thermal springs of Budapest. This way continuous moni­toring of its environment if of exceptional importance. Because the infiltrated polluted waters may change (i) the natural karst environment of the cave, (ii) its forma­tions and minerals and (iii) the aerosols in it as well and consequently have a direct impact on its exploitation for therapeutic purposes and endanger the cave as a geologi­cal heritage.

Acknowledgements

The authors would first like to thank the Directorate of the Duna-Ipoly National Park for authorizing the re­search in the cave, Paul Thatcher for his work on our English version, and say thanks for the stimulating dis­cussions with our colleague Zoltán Kern. We would like to thank the Lithosphere Fluid Research Group and Prof László Bozó, chief advisor of the Hungarian Meteoro­logical Service, OMSZ, for providing the meteorological data. In addition, we would like to thank the help of the Bekey Imre Gábor and the Pagony Cave Research Groups for their help. We are also grateful for the support of the “Lendület” program of the Hungarian Academy of Sci­ences (LP2012-27/2012) and the János Bolyai Research Scholarship of the Hungarian Academy of Sciences. This is contribution No. 34. of the 2ka Palaoclimatology Re­search Group.

References

Abdi, H. & L.J. Williams, 2010: Principal component analysis.- Wiley Interdiscip. Rev. Comput. Stat., 2, 433–459.
Albert, G., 2010: Volumetric modelling of cavities and pores in the Pál-völgy Cave, Budapest (in Hungar­ian).- Földtani Közlöny, 140, 3, 263-280.
Back, W., 1966: Hydrochemical facies and ground-water flow patterns in northern part of Atlantic Coastal Plain.- U.S. Geol. Surv. Prof., Paper 498-A, pp. 42.
Baker, A. & C. Brunsdon, 2003: Non-linearities in drip water hydrology: An example from Stump Cross Caverns, Yorkshire.- Journal of Hydrology, 277, 151-163.
Baker, A. & I.J. Fairchild, 2012: Drip Water Hydrology and Speleothems.- Nature Education Knowledge, 3,10, paper 16.
Baldini, J.U.L., McDermott, F., Baldini, L.M., Ottley, C.J., Linge, K.L.,Clipson, N. & K.E. Jarvis, 2012: Identi­fying short term and seasonal trends in cave drip water trace element concentrations based on a dai­ly-scale automatically collected drip water dataset.- Chemical Geology, 330-331, 1-16.
Batiot, C., Linan, C. & A. Andreo, 2003: Use of total or­ganic carbon (TOC) as tracer of diffuse infiltration in a dolomitic karst system: the Nerja Cave (An­dalusia, southern Spain).- Geophys Res Lett, 30, 22, 2179–2183.
Bolner, T.K., 1995: Karst water protection problems in­dicated by dripping water analyses in Buda thermal karst area.- Acta Carsologica, 24, 525–534.
Bolner, T.K., Tardy, J. & L. Némedi, 1989: Evaluation of the environmental impacts in Budapest’s caves on the basis of the study of the dripping waters.- In: 10th International Congress of Speleology, UIS, pp. 634-639, Budapest.
Brocx, M., & V. Semeniuk, 2007: Geoheritage and geo­conservation: history, definition, scope and scale.- J R Soc West Aust, 90, 53–87.
Cleveland, W.S., 1979: Robust Locally Weighted Regres­sion and Smoothing Scatterplots.- Journal of the American Statistical Association, 74, 829–836.
Cleveland, W.S. & S.J. Devlin, 1988: Locally Weighted Regression: An Approach to Regression Analysis by Local Fitting.- Journal of the American Statistical Association, 83, 596–610.
Czuppon, Gy., Kármán, K., Németh, S., Kiss, K., Demé­ny, A., Szilárd, J., Kern, Z., Bíborka, M., Kohán, B., Haszpra, L. & Z. Siklósy, 2014: Hydrogen and oxy­gen isotopic variation of cave drip waters: implica­tions for recent climate and paleclimate signal in stalagmite.- In: Climate Change: The Karst Record (VII) KR7: 7th International Conference. Abstract book, pp. 62-63, Melbourne.
Czuppon, Gy., Kern, Z., Kármán, K., Németh, S., John, Sz., Haszpra, L., Kohán, B., Kiss, K., Siklósy, Z. & Zs. Polacsek, 2013: Spatial and Temporal Variations of .D and .18O Values of Cave Drip Waters: Impli­cations for Paleoclimate Signal in Stalagmite.- Cen­tral European Geology, 56, 2-3, 274-276.
Daoxian, Y. & Z. Cheng, 2009: Annual Report of the IGCP513 “Karst Aquifers and Water Resources” China Working Group in 2009.- [Online] Avail­able from: http://www.karst.edu.cn/public/upload/files/20150116-IGCP513%20report%20in%202009.pdf [Accessed 17th May 2016].
Daubechies, I., 1990: The wavelet transform time-frequency localization and signal analysis.- IEEE Trans. Inform. Theory, 36, 961–1004.
Erőss, A., 2010: Characterization of fluids and evaluation of their effects on karst development at the Rózsa­domb and Gellért Hill, Buda Thermal Karst, Hun­gary.- PhD thesis. Eötvös Loránd University, Buda­pest, pp. 171.
Erőss, A., Mádl-Szőnyi, J. & A. Csoma, 2012: Hypogenic karst development in a hydrogeological context, Buda Thermal Karst, Budapest, Hungary.- In: Malo­szewski, P., Witczak, S. & G. Malina (eds.) Ground­water Quality Sustainability: IAH Selected Papers on Hydrogeology 17. CRC Press - Taylor and Francis Group, pp. 119-133, London.
Fairchild, I.J. & A. Baker, 2012: Speleothem Science: From Process to Past Environments, John Wiley & Sons Ltd, pp. 450, Chichester.
Fehér, K., 2009: A Rózsadombi-termálkarszt szennyeződés-veszélyeztetettségi vizsgálata (in English: The pol­lution-risk assessment of the Rózsa Hill thermal karst).- M.Sc thesis. Eötvös Loránd University, Bu­dapest, pp. 79.
Ford, D. & P. Williams, 2007: Karst Hydrogeology and Geomorphology.- John Wiley & Sons, pp. 562, Chichester.
Gáspár, A., Virág, M. & A. Erőss, 2015: Karst porosity estimations from archive cave surveys - studies in the Buda Thermal Karst System (Hungary).- Inter­national Journal of Speleology, 44, 151-165.
Genty, D. & G. Deflandre, 1998: Drip flow variations un­der a stalactite of the Pere Noel cave (Belgium). Evi­dence of seasonal variations and air pressure con­straints.- Journal of Hydrology 211, 208–232.
Granato, G.E., 1996: Deicing chemicals as a source of constituents in highway runoff.- Transportation Re­search Record: Journal of Transportation Research Board, 1533, 50-58. doi: 10.3141/1533-08
Gröchenig, K., 2001: Foundations of Time-Frequency Analysis.- Birkhäuser, pp. 360, Basel.
Gulden, B., 2016: [Online] Available from: http://www.caverbob.com/wlong.htm [Accessed 17th May 2016]. 
Hartland, A., Fairchild, I.J., Lead, J.R., Borsato, A., Baker, A., Frisia, S. & M. Baalousha, 2012: From soil to cave: Transport of trace metals by natural organic matter in karst dripwaters.- Chemical Geology, 304-305, 68-82.
Hatvani, I.G., Kovács, J., Márkus, L., Clement, A., Hoff­mann, R. & J. Korponai, 2015: Assessing the rela­tionship of background factors governing the water quality of an agricultural watershed with changes in catchment property (W-Hungary).- Journal of Hy­drology, 521, 460-469.
Hem, J., 1985: Study and interpretation of the chemical characteristics of natural water.- U.S. Geological Survey, Water-Supply Paper 2254, pp. 263.
Hungarian Cave Database, 2016: [Online] Available from: http://www.termeszetvedelem.hu/index.php?pg=caves [Accessed 3rd June 2016].
Jakucs, L., 1971: A karsztok morfogenetikája (in English: Morphogenetics of Karsts).- Akadémiai Kiadó, pp. 310, Budapest.
Kaiser, G., 1994: A Friendly Guide to Wavelets, Birkhäus­er, pp. 300, Cambridge.
Kaiser, H.F., 1960: The application of electronic comput­ers to factor analysis.- Educational and Psychologi­cal Measurement, 20, 141-151.
Kern, Z., Fórizs, I., Pavuza, R., Molnár, M. & B. Nagy, 2011: Isotope hydrological studies of the perennial ice deposit of Saarhalle, Mammuthöhle, Dachstein Mts, Austria.- The Cryosphere, 5, 291-298. doi:10.5194/tc-5-291-2011
Kern, Z., Fórizs, I., Perşoiu, A. & B. Nagy, 2009: Stable isotope study of water sources and of an ice core from Bortig ice cave, Romania.- Data of Glaciologi­cal Studies (Materialy Glyatsiologicheskikh Issledo­vaniy), 107, 175-182.
Kirkwood, B.R. & J.A.C. Sterne, 2003: Essential medi­cal statistics (2nd ed). Blackwell Science, pp. 513, Malden, Massachusetts.
Kleb, B., Benkovics, L., Dudko, A., Gálos, M., Juhász, E., Kertész, P., Korpás, L., Marek, I., Nádor, A. & Á. Török, 1993: Complex geological investigations and drillings in the surroundings of Rózsadomb. Geologi­cal, petrophysical, tectonic and palaeokarst analysis and evaluation.- Department of Engineering Geol­ogy, Budapest Technical University, Project Report Phare 134/2.
Kogovšek, J., 2011: Impact of chlorides, nitrates, sulfates and phosphates on increased limestone dissolution in the karst vadose zone (Postojna Cave, Slovenia).- Acta Carsologica, 40, 2, 319-327.
Kogovšek, J. & M. Petrič, 2013: Increase of vulnerability of karst aquifers due to leakage from landfills.- En­viromental Earth Sciences, 70, 2, 901-912.
Kovács, J., Hatvani, I.G., Korponai, J. & I.S. Kovács, 2010: Morlet wavelet and autocorrelation analysis of long-term data series of the Kis-Balaton water pro­tection system (KBWPS).- Ecological Engineering, 36, 1469-1477.
Kovács, J., Márkus, L., Szalai, J. & I.S. Kovács, 2015: De­tection and evaluation of changes induced by the diversion of River Danube in the territorial appear­ance of latent effects governing shallow-ground­water fluctuations.- Journal of Hydrology, 520, 314-325.
Lerner, D.N., Yang, Y., Barrett, M.H. & J.H. Tellam, 1999: Loadings of non-agricultural nitrogen in urban groundwater.- In: Ellis, B. (ed.) Impacts of urban growth on surface and groundwater quality. IAHS Publication No. 259, pp. 117–124, Birmingham.
Liu, W. & A. Brancelj, 2014: Hydrochemical response of cave drip water to snowmelt water, a case study from Velika Pasica Cave, Central Slovenia.- Acta Carsologica, 43, 1, 65-74.
Mádl-Szőnyi, J. & Á. Tóth, 2015: Basin-scale conceptual groundwater flow model for an unconfined and confined thick carbonate region.- Hydrogeology Journal, 23, 7, 1359-1380.
Mádl-Szőnyi, J. & A. Erőss, 2013: Effects of regional groundwater flow on deep carbonate systems fo­cusing on discharge zones.- In: Proceedings of the International Symposium on Regional Groundwater Flow: Theory, Applications and Future development, 21st-23rd June 2013, Xi’an, China. China Geologi­cal Survey, Commission of Regional Groundwater Flow, IAH, 71-75, China.
Mádl-Szőnyi, J., Virág, M. & A. Erőss, 2007: A Szemlő-hegyi-barlang csepegővizeinek vizsgálata a budai márga törmeléktakarón át történő beszivárgás ér­tékelése céljából (in English: Investigation of drip­ping water in the Szemlő Hill cave in order to assess infiltration through the Buda marl debris mantle).- Földrajzi Közlemények 131, 55, 371-388.
Mari, L., & K. Fehér, 1999: The impacts of land use change on the Buda termal karst: a study of Szemlő-hegy cave.- In: Bárány-Kevei I. & J. Gunn (eds.) Es­says in the ecology and conservation of karst. Acta Geographica Szegediensis, 36, pp. 104–111, Szeged.
Meyer, K.W., Feng, W., Breecker, D.O., Banner, J.L. & A. Guilfoyle, 2014: Interpretation of speleothem cal­cite d13C variations: Evidence from monitoring soil CO2, drip water, and modern speleothem calcite in central Texas.- Geochimica et Cosmochimica Acta, 142, 281-298.
Morlet, J., Arens, G., Fourgeau, E. & D. Giard, 1982: Wave propagation and sampling theory; Part I, Complex signal and scattering in multilayered media.- Geo­physics, 47, 203-221.
Motyka, J., Gradzinski, M., Bella, P. & P. Holubek, 2005: Chemistry of waters from selected caves in Slovakia – a reconnaissance study.- Environmental Geology, 48,6, 682-692.
NLCO: National Land and Cartography Office, 1975: Projection Guideline for the application of the Uni­form National Projection System, Budapest.
Panno, S.V., Hackley, K.C., Hwang, H.H., Greenberg, S.E., Krapac, I.G., Landsberger, S. & D.J. O'Kelly, 2006: Characterization and Identification of Na-Cl Sources in Ground Water.- Ground Water 44, 2, 176-187. doi: 10.1111/j.1745-6584.2005.00127.x
Piper, A.M., 1944: A graphic procedure in the geochemi­cal interpretation of water-analyses.- Trans Am Geophys Union, 25,914-923.
R Core Team, 2013: R: A language and environment for statistical computing. R Foundation for Sta­tistical Computing, Vienna.- [Online] Available from: http://www.R-project.org [Accessed 17th May 2016].
Santos, C.A.G., Galvao, C.O. & R.M. Trigo, 2003: Rain­fall data analysis using wavelet transform. – In: Ser­vat, E. et al. (eds.) Hydrology of Mediterranean and semiarid regions: papers selected for the international conference on Hydrology of the Mediterranean and Semi-Arid Regions, 1st-4th April 2003, Montpellier, France. International Association of Hydrological Sciences, Unesco, 195-201, Wallingford.
Siklósy, Z., Kern, Z., Demény, A., Pilet, S., Leel-Ossy, Sz., Lin, K., Shen, C.C. & É. Széles, 2011: Speleothem and pine tree as sensitive indicators of environmen­tal pollution – a case study of the effect of uranium -ore mining in Hungary.- Applied Geochemistry, 26, 666-678.
Stieber, J. & Sz. Leél-Őssy, 2015: Changed or back into the climate of Peace Cave?.- Karsztfejlődés, XX, 263-282.
Szalai, Z., 2008: Spatial and temporal pattern of soil pH and Eh and their impact on solute iron content in a wetland (Transdanubia, Hungary).- Acta Geo­graphica Debrecina Landscape and Environment, 2,1, 34-45.
Torrence, C. & G.P. Compo, 1998: A Practical Guide to Wavelet Analysis.- Bulletin of the American Meteo­rological Society, 79, 61-78.
Végh, D., Kovács, J., Maucha, L. & L. Márkus, 2005: Mathematical analysis of runoff depletion time series of karst springs around Jósvafő. Karsztfejlö­dés, 10, 91-105. [Online] Available from: http://www.karsztfejlodes.hu/kotetek/2005/A%20J%C3%-93SVAF%C5%90%20K%C3%96RNY%C3%89-KI%20KARSZTFORR%C3%81SOK%20V%C3%8-DZHOZAM.pdf [Accessed 17th May 2016].
Veni, G., 1999: A geomorphological strategy for con­ducting environmental impact assessments in karst areas.- Geomorphology, 31,1-4, 151-180.
Virág, M., Mádl-Szőnyi, J. & A. Mindszenty, 2011: Natu­ral and anthropogenic effects in the drip water of the Szemlőhegy Cave (in Hungarian).- Bányászat, 81, 387-400. [Online] Available from: http://www.matarka.hu/koz/ISSN_1417-5398/81k_2011/ISSN_1417-5398_81k_2011_387-400.pdf [Accessed 17th May 2016].
Williams, P.W., 2008: The role of the epikarst in karst and cave hydrogeology: a review.- International Journal of Speleology, 37, 1-10.
Zámbó, L. & T. Telbisz, 1999: The influence of the soil zone on karst corrosion and karren development.- In: Bárány-Kevei I. & J. Gunn (eds.) Essays in the ecology and conservation of karst. Acta Geographica Szegediensis, 36, pp. 187-192, Szeged.
Zámbó, L.,Horváth, G. & T. Telbisz, 2001: Investigations of microbial origin of karst corrosion of soils de­pending on different temperatures.- Chinese Sci­ence Bulletin, 46, 1, 28-32.

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)

Appendices

Appendix 1


Fig. A1: Cross section of the studied area of the Pál-völgyi Cave System with the location of the two sampling sites, the T galley (TG) and the Y corridor (YC).

Appendix 2


Fig. A2: Build-up of the ODEON data collector and sampling on site (photograph on the right).

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani

Appendix 3. STFT results


Fig. A3: STFT results of pH at site YC.

STFT was applied first to the series, using Bartlett window and window width 200 data points (Fig. A3).
High amplitudes can be observed along the 24 hourly periods (0.0417 frequency, broken line) and its close neighborhood in the majority of the time-domain. It is very difficult however to distinguish, whether it is an effect of the choice of the window width, or a result of a genuine diurnal period. It would be straightforward to narrow the window width but then due to the poor sepa­ration of frequencies the 24 hourly period would disap­pear in the whole last third of the time domain. To the contrary, by widening the window, the 24 hourly period would completely fill in the full time domain as if it were detectable any time. Further, to our best knowledge no statistical test is available for assessing the significance of the estimated amplitude at a given time, therefore the re­quired proportion of days with periodic behavior could not be found with prescribed reliability. This necessitates a further analysis, when time and frequency resolutions do not change at the expense of each other.

Appendix 4


Fig. A4: Cation concentrations at TG A) and YC B) and anion concentrations at TG C) and YC D).

Analysis of drip water in an urban karst cave beneath the Hungarian capital (Budapest)

Appendix 5


Fig. A5: Example on LOESS on pH at site TG with the original data the smoothing curve and the residuals.

Appendix 6


Fig. A6: Z-score standardized pH values (upper graph) and daily periodicity (lower graph) at site YC. The domain of the 5 % significance level against red noise are colored by red and marked with a thick black contour. The dashed areas outside the thin black line mark the COI. The black horizontal line indicates the approximated place of the daily period.



224

ACTA CARSOLOGICA 45/3 – 2016un-fractured YC site, the karst processes drive periodic­ity, while in the shallower TG, located next to the frac­ture zone, it is likely to be the topsoil and vegetation. However, it is suspected that the non-periodic patterns and the decreased daily periodicity seen in the case of EC at TG (wavelet results, Tab. 2) are the results of (i) an­thropogenic contamination, and (ii) the fact that in the fractured zone next to TG (Fig. A1), water seeps down quite fast, (iii) as a result gathering the substances from a larger area/watershed and in a more concentrated way than at YC. 

Another reason for the periodic patterns of the continuously measured parameters may be attributed to biological activity (root/rhizosphere respiration and soil organic matter decomposition; Zámbó et al. 2001; Meyer et al. 2014) in the soil lying over the area. Jakucs (1971), observing the atmosphere in the soil, found that CO2 has a daily periodicity in different bio- and climate-specific karst micro areas (Szalai 2008), and so does carbonate in covered karst during continuous and prolonged seepage (Zámbó & Telbisz 1999). Both considered the solution of bicarbonate and the CO2 concentration of the seep-water as a main factor of “soil effect”, clearly determining the chemistry of the seeping water, consequently determin­ing the pH of the water seeping into the karst. These pro­cesses and the anthropogenic effects are superimposed on the periodic patterns at sites TG and YC.
Although measurements have been conducted for drip water in the Pál-völgyi cave by Bolner et al. (1989), Bolner (1995) and Fehér (2009), and in other caves in Budapest by Mádl-Szőnyi et al. (2007), these only in­volved periodic/intermittent sampling campaigns. Thus, comparing the results of the unique continuous cam­paign presented here with their findings would not give a meaningful picture.
It should be noted that similar, continuous drip wa­ter collection has been performed in other Hungarian caves outside the capital, mainly for stable isotope com­position and water chemistry in e.g. the Csodabogyós Cave (Czuppon et al. 2013, 2014), but this took place in areas located in an anthropogenically undisturbed set­ting. In addition, continuous climate- (Stieber & Leél-Őssy 2015), and spring runoff measurements (Végh et al. 2005) are also available from natural karstic areas.

Katalin Fehér, József Kovács, László Márkus, Edit Borbás, Péter Tanos & István Gábor Hatvani


Fig. 8: Bi-weekly measurements for pH, EC A) and TDS B) at YC and their correlation matrix (inset table).