Hydrogeology of a complex karst catchment in Southern Dalmatia (Croatia) and Western Herzegovina (Bosnia and Herzegovina)

ABSTRACT Our study focuses on a sizeable transboundary karst catchment in Croatia and Bosnia and Herzegovina, extending over 2000 km2. A complex underground conduit system and extreme karst forms heterogeneity are the main characteristics of the area in question. Since determining the boundary of such a large and complex catchment is difficult, we used different kinds of data sets, of which the most relevant are the available geological data, hydrochemical data, hydrological data, and tracing tests data, to divide the regional catchment into six subcatchments. We also examined past archived reports and carried out new hydrological investigations of several major and minor springs. Our research results in a hydrogeological map that can be used as a base for establishing site-specific groundwater protection zones, for water balance calculations and the planning of new research in this area, especially the ones regulating combined cross-border efforts to prevent groundwater contamination and ensure sufficient drinking water.

The objective of this study was to produce a hydrogeological map of a large regional karstic catchment, situated at the border of Southern Dalmatia (Croatia) and Western Herzegovina (Bosnia and Herzegovina), with delineation into several subcatchments or influential areas.Determining the catchment boundaries and the catchment area in karst is a highly complex task (Bonacci et al., 2013).Because surface water and groundwater from these subcatchments flow into neighbouring subcatchments, and consequently, the boundaries between them cannot be considered as groundwater or surface water divides, we have used the term influential area (Main map).
Our map was created mainly by using information from unpublished expert reports at the Croatian Geological Survey Archive combined with the data collected during our field surveys conducted from 25 th September 2013 to 29 th July 2015.During this period, our hydrochemical analysis comprised monthly sampling of 15 springs and two surface lake water in Croatia and three springs in Bosnia and Herzegovina, each sampled once.As a part of this study, we also carried out hydrogeological mapping and tracer experiment.Delineation of the hydrogeological system was achieved using structural geological information from previous research, combined with water balance calculation, tracing experiments and springs' hydrogeochemical properties.

Study area
The catchment area is situated in two countries, covering a part of Western Herzegovina (in Bosnia and Herzegovina) and a region of Southern Dalmatia (in Croatia) (Figure 1).The whole research area extends to over 2000 km 2 .The recharge zone of this large catchment is predominantly in Bosnia and Herzegovina, whereas the discharge zone is mainly along the Croatian coast.The elementary determinants of the area comprise heterogeneous relief, expressed elevation differences over a small area and a geological basis with typical karstic relief formed in the zones known as the deep karst (Matić et al., 2012).Characterisation of the area is an interchange of huge mountains and karst poljes.Vrgoračko polje is the lowest 'hydrogeological cascade' to which groundwater flows from hypsometrically higher karst poljes − the Duvanjsko, Imotsko and Rastok poljes (Main map).The rapid circulation of groundwater, often under pressure, allows the formation of powerful springs at lower elevations (Zerem, 2016).The area is abundant in good-quality spring water, of which only a small amount has been used for public water supply (Ivičić & Pavičić, 1997).The highest peak in Western Herzegovina is Pločno on the Čvrsnica mountain (2228 m a.s.l.), while the highest coastal Dinaric mountain in Southern Dalmatia is Biokovo with its highest peak St Jure (1762 m a.s.l.).The area includes several important rivers, Ričina, Suvaja, Sija, Vrljika, Tihaljina, Sita, Mlade and Trebižat, which represent one river with receding, sinking and underground stream sections, with a longitude of about 106 km from spring to mouth in the Neretva River (Bonacci et al., 2013) (Main map).Most surface water flows towards the Neretva River, but part of the groundwater flows southwards.According to Köppen's climate classification (2011), on the coast Hot-summer Mediterranean climate is prevalent (Csa), while further inland, a Warm-summer Mediterranean climate (Csb) prevails.

Hydrogeology
Hydrogeological relations of the area are very complex due to morphology, anthropogenic influences (the construction of tunnels and hydroelectric power plants, the development of irrigation systems, massive karst groundwater pumping, urbanisation, etc.), and the climate.We conducted our fieldwork in the Tihaljina-Mlade-Trebižat valley, from the Vrgorac area up to the final stretch of karst springs in the vicinity of Baćina lakes and Prud spring (Main map).More than 100 temporary and permanent karst springs emerge here.Annual discharges of the larger springs range from 1.5 to maximum 10 m 3 /s (Klokun, Modro oko Vrelo Lištice, Vrelo Tihaljine, Prud spring, Opačac, Klokun BiH).
The study area is composed of sedimentary rocks of the Permo-Triassic, Triassic, Upper Jurassic, Lower and Upper Cretaceous, Palaeogene, Neogene and Quaternary ages (Bonacci et al., 2013;Slišković, 1994).Older Permo-Triassic rocks have been tectonically lifted to the surface over younger rocks (reverse faulting) and consist of evaporite rocks with carbonate and clastic rocks (sandstones and marls).The Triassic strata (mainly dolomites) form partial barriers to water circulation within the study area.Jurassic rocks, dolomites and limestones, normally lie beneath limestones of the Lower Cretaceous.The Cretaceous deposits, composed of stratified to massive limestone with rare dolomites prevail in this area.Palaeogene and Neogene consists of a wide range of sediments, but predominantly clastic rocks (sandstones and marls).Here the Dinaric Eocene flysch sediments (Marinčić, 1981) can have a diverse impact on groundwater movement.The Quaternary deposits extend throughout karst poljes, as well as along the open streamflows.The most permeable rocks of the area are limestones and dolomites, while the least permeable rocks belong to flysch series and Quaternary deposits.In this area, karst poljes behave as 'hanging barriers' (Figure 2) with its Quaternary sediments, and springs usually appear in the north and northwest edges of poljes.In the central zones of poljes, there are estavelles and in the east and southeast parts, ponors are developed.The lowest poljes, such as Vrgoračko, represent a total groundwater barrier due to low elevation and morphology.One of the reasons why upper karst polje barriers were defined as 'hanging' is because groundwater constantly streams underneath them (Figure 2).Thin flysch series, which are present sporadically, behave similarly.
Underground seawater intrusions are also common in the discharge zone of the investigated area.Coastal springs and Baćina lakes salinisation is less frequent due to the significant flysch barrier that separates the karst background from the sea.Flysch barrier extends from the seaside at Makarska city and continues along the lower part of Biokovo mountain to the vicinity of Baćina lakes, which are dominated by the less permeable dolomites, dolomitic limestones, dolomitic and limestone breccias.

Materials and methods
This study combined pre-existing geological and hydrogeological data with new findings collected during our fieldwork campaigns between September 2013 and July 2015.The pre-existing archived unpublished data was mainly focused on the sanitary protection zones, which included delineations of smaller catchment areas of springs.Therefore, for the first time, a uniform hydrogeological map of the whole catchment has been made.The data for the map production included: i) four pre-existing geological maps on the scale of 1:100.000,which we combined and interpreted into a hydrogeological map, ii) springs locations and their characteristics which we found in unpublished archival reports, iii) collection of discharge data used to calculate water balance, iv) tracing experiment data from historical reports, and v) new simultaneous tracing experiment from 2014.

Cartographic data
We created a hydrogeological map by combining four geological maps and attributing hydrogeological characteristics to geological units.These geological maps are: Ploče (Marinčić et al., 1979), Imotski (Raić et al., 1978), Metković (Raić et al., 1977) and Mostar (Mojičević & Laušević, 1966).We encountered a significant challenge when combining the geological units at the edges of different sheets of these four maps.Furthermore, a part of the area surrounding Baćina lakes was mapped in 2015 in more detail which reduced any potential discrepancies about merging sheets.As a result, we grouped stratigraphic units into nine hydrogeological categories based on their permeability, given that hydrogeological categories belong to geological units with similar capacity of storing and transmitting water according to their lithological properties (Gragnanini et al., 2015).We finally digitised all of the data in the GIS environment, shown within the Main map.
using digital titrator (HACH, USA).In Western Herzegovina, we undertook one sampling campaign in February 2014.Concentrations of Na + , K + , Mg 2+ and Ca 2+ were obtained using the method of atomic absorption with AAS Analyst 700 (Perkin Elmer, USA), whereas concentrations of NO 3 -, PO 4 3--P, SO 4 2and Cl -were obtained with ion chromatography apparatus (LabAlliance, USA).Our measurements were all conducted in the Croatian Geological Survey laboratory.We used Piper diagram (Figure 3) to show relative ratios of major ions median values in the water of this area.

Hydrological data
Given the complexity of the hydrological conditions of the catchment and the lack of appropriate measurement data, the water balance assessment was made by analysing available climatological and hydrological data available at the Croatian Hydrometeorological Survey and the Agency for the Adriatic Sea Water Area, Mostar.This was made during hydrogeological investigations of Baćina lakes (Terzić et al., 2015).Estimates for water balance were made based on thirty years dataset, 1961−1990.
In the report by Turc (1955) and Langbein (1938), empirical models were used to calculate water balance in a modified and developed GIS application (Horvat & Rubinic, 2006).These approaches use the average annual rainfall and air temperature (often the only available climatological data in hydrological runoff analysis) as spatially variable input parameters.
By examining the spatial assessment of annual precipitation and average annual air temperatures in our catchment area, the spatial distribution of effective annual precipitation was calculated.The process started with the spatial estimation of input meteorological data (temperature and precipitation average annual data) in the delineated catchment.Obtained results were then verified in two steps: i) with calibration of input meteorological data by establishing corrective correlations between measured meteorological data in the field (selected climatological station) and values obtained based on their spatial distribution; ii) calibration of evaluated annual runnof values and effective precipitation (Turc and Langbein method) with measured discharge data data in a wider regional area, which was extended to a referenced 30 years period based on a regression relationships data from nearby hydrogeological stations.

Tracing experiment
To characterise groundwater flow direction, we conducted a new simultaneous tracing experiment during our fieldwork in Vrgoračko polje on 30 th September and 2 nd October 2014 (Main map).Vrgoračko polje is located in Southern Dalmatia, northeast of Biokovo mountain, where the Matica River is the only permanent stream that winds from its springs in the northwest before splitting to the south and the east and then sinking.The difference in altitude between the springs, estavelles and ponors is approximately eight metres.In Vrgoračko polje, seasonal runoff goes through estavelles and ponors into the Neretva River (east of Vrgoračko polje).It drains to Baćina lakes through an artificial tunnel (monitoring pointtunnel T2), leaving the Vrgoračko polje dry throughout the summer.This seasonal alternation of flooding and dry periods is repeated annually (Posilović et al., 2018).The excess water is further drained from the Baćina lakes through artificial tunnel (monitoring pointtunnel T3) into the sea.
We analysed the new tracing experiment made in 2014 against the older tracings collected from archived reports (Bojanić, 1980;Ivičić & Pavičić, 1997;Slišković, 1994Slišković, , 2000Slišković, and 2014) ) to interpret flow pathways and groundwater recharge processes (Flury & Wai, 2003;Weiler & Flühler, 2004).For this experiment, we used sodium fluorescein and sodium naphthionate as tracers.As sodium fluorescein has slightly better properties (minimal sorption, higher detection level, higher solubility in water), it was used in the northeast Vrgoračko polje, where tracing has not been done previously.By comparison, we used naphtionate in the central part of the polje because of a previously established but not certain connection (Ivičić & Pavičić, 1996).The simultaneous tracing experiment was carried out during a flood wave retreat when the underground was saturated with groundwater but the natural runoff gradients, generally towards the sea, were re-established.Despite the low gradients, the flow conditions were changing rapidly due to the high karstification of the terrain, and choosing a suitable place to insert the tracer was challenging.Ponors are located mainly along the side of the Matica River, so their activity mainly relates to the levels of the river.
Consequently, the main problem during our tracing experiment was that ponor water does not submerge during high-water levels.At the same time, ponors dry up in a short period during the lower water levels.As a result, we decided to insert tracers two days apart.The first tracer of fluorescein (10 kg) was injected in the northwest polje on 30.9.2014, as the drainage direction of Vrgoračko polje is from northwest to southeast.Hence, the water levels fall in northwest polje first.In the ponor about 10 l/s of water plunged at the time of the experiment.The second tracer, naphthionate (60 kg dissolved in 250 l of water), was injected on 2.10.2014,50 metres from ponor Crpala at the bank of river Matica, between stone blocks, where plunging of water was heard (Figure 6).The mean daily water level of the Matica river was 181 cm on the day of fluorescein injection, measured in Staševica.On the day of the naphthionate injection, it dropped to 153 cm.Water samples were analyzed on a Perkin-Elmer (USA) spectrofluorometer LS55 and at Klokun and Mandina mlinica with field fluorometer GGUN-FL (Albillia, Switzerland).

Results & discussion
As the Dinaric karst groundwater flows through highly intricate channels and pathways that cannot be directly observed, we used various methods presented in this paper to categorise the regional catchment into six influential areas, which we have termed subcatchments (Main map).It was necessary to classify these subcatchments because their groundwater fluctuates depending on the water table levels affected by changes in precipitation.Given the hydrogeological conditions during late winter and spring, for instance, when snow melts and precipitation increases, water often discharges across subcatchments.Furthermore, we examined high to low permeability rocks and structural geological elements, whose main characteristics affect subcatchments boundaries.This investigation is of great importance in the future management of this transboundary catchment because two countries are beneficiaries of its water resources, and ensuring good quality of the groundwater is crucial for sustainable development and protecting the integrity of its ecosystems (Ly et al., 2019).

Cartographic data -Main map representation
The primary outcome of our research was a production of the Main map, in which we delineated this regional catchment into six subcatchments: Mostarsko blato (a-250 km 2 and b-359 km 2 ), Tihaljina-Mlade-Trebižat (a-688 km 2 and b-179 km 2 ), Opačac (176 km 2 ), Prud catchment (94 km 2 ), Klokun and Modro oko catchment (449 km 2 ), Baćina lakes and Mandina Mlinica catchment (100 km 2 ).In addition, Mostarsko blato and Tihaljina-Mlade-Trebižat, two very big subcatchments, are further divided into two subdivisions.This was important because surface flow often passes across these subdivisions when the water level is higher.Still, when the water level is lower, groundwater flow dominates, and a more significant percentage of water flows within the boundaries of these subdivisions.We conceptualised the recharge area for this wide drainage zone which extends from Prud spring on the border with Bosnia and Herzegovina, along the edge of the Neretva River valley and to Ploče, and along the Adriatic coast to Gradac, where a complete flysch barrier reaches the coast.The whole catchment area is characterised by short-lived water transit, from a few days to a few months, suggesting low-storage capacity, high variability of spring discharge and fast transport of potential pollutants to the spring.The Main map shows that topographic divides do not correspond to the underground catchment areas.In particular, these divides represent subcatchment boundaries that interact with neighbouring ones.This is also true for the Tihaljina-Mlade-Trebižat, spreading over two catchments.In particular, these boundaries cannot be treated as strict divides because the flow frequently crosses over them, subject to seasonal changes.Therefore, we argue that influental areas are a more appropriate term when descibing similar systems, given the challenges of independently describing the Prud spring and Klokun spring catchments, as they share a large area in which groundwater drains toward them.For the same reason, dividing catchments, such as Klokun and Mandina mlinica, or Modro oko and Klokun who share similar dynamic and geochemical characteristics is even more complicated.Consequently, dividing this complex karst system by delineating subcatchments based on how the water drains towards different zones in different weather conditions is more constructive.In other words, all the groundwater divides are zonal and contingent upon hydrological conditions, which means they often change position.Only in places where the flysch and carbonate rocks are in contact (e.g.south of Gradac, near Mandina Mlinica spring) groundwater divide is not zonal.Instead, it follows the reverse fault plane to the town of Makarska and spreads north to the city of Imotski, where the aquifer is built of carbonate rocks with considerably lower flysch units.

Hydrogeochemical facies
When defining our subcatchments, we also considered groundwater hydrochemical characteristics.Major ion analysis revealed three hydrogeochemical facies of springs, indicating different groundwater origin (Figure 3).These are i) calcium -hydrogen carbonate (Ca 2+ -HCO 3 − ) ii) calcium -sulfate / hydrogen carbonate (Ca 2+ -SO 4 2− /HCO 3 − ), and iii) sodium -chloride (Na + -Cl − ).This suggests that dissolution of carbonates, evaporites and seawater intrusion are the principal drivers of groundwater chemistry.As most of the investigated area is built of carbonates the Ca 2 + -HCO 3 − facies is dominant.Previous investigations results indicated that a significant number of the sampled water from the study area have higher concentrations of sulfate ions (Slišković et al., 1998;Štambuk-Giljanović, 1994) compared to the remaining Croatian karst groundwater.In this instance, the water of these karst aquifers differs because of the underground presence of evaporites, gypsum and anhydrite (Plummer & Back, 1980;Šiftar, 1986;Šušnjara et al., 1992;Tišljar, 1992).Namely, outcrops of evaporites are present in Western Herzegovina.Given the width of the catchment, there are raised levels of sulfate ions in Klokun BiH spring (in Bosnia and Herzegovina), marked within Piper diagram (Figure 3).Furhermore, water of Southern Dalmatia springs: Prud, Lukavac and Stinjevac belong to mixed Ca 2+ -SO 4 2− /HCO 3 − facies.On the other side, Na + -Cl − facies is present in the coastal area and can indicate fresh-seawater interaction and how deep into the continent the sea has an impact.The increased content of Na + (K + ) and Cl − is detected at spring Mandina mlinica and at the Mindel spring in Baćina lakes.These springs are located closer to the edge of the flysch barrier, in contact with carbonates that are more permeable.Further to the east, the sea's effect decreases.This is influenced by the Neretva River, which flows into the sea and acts as a barrier to seawater intrusions.

Water balance
Maps (Figure 4) showing the spatial distribution of specific discharge for the period 1961-1990, which were obtained by using the Langbein and Turc methods, are made within the unpublished report 'Hydrogeolocial investigations of Baćina lakes' (Terzić et al., 2015), are one of the essential layers in catchment delineation.Differences between Langbein and Turc methods can be seen in Figure 5.
Within the report, subcatchments mean annual runoff was estimated.Tables 1 and 2 give the mean annual temperature, mean annual precipitation and total mean annual discharge for each subcatchment.Langbein's method provided lower values of specific discharge in all the analysed areas, compared to those obtained with Turc's method.For example, in the northeastern parts of the southern Dalmatia runoff and precipitation is larger, and in the areas with higher altitude it even exceeds 50 ls −1 km 2 .Here the differences between the two methods are  (Terzić et al., 2015).Subcatchments numbers are explained in Table 1.(Terzić et al., 2015).Subcatchments numbers are explained in Table 1.around 10 -20%.In coastal areas, surface runoff is about 10 ls −1 km 2 , and the results of Turc runoff are higher by 20 30% compared to the Langbein method.However, this area falls outside the boundaries of the analysed catchment, so the actual differences in the results of assessments obtained according to the two mentioned methods are significantly lower, ranging from 10 -20%.

Tracing experiment
Our new tracing results from an experiment carried out in September, and October 2014 with maximum velocities are shown in Figure 6.In Table 3 fluorescein whereas in Table 4 naphthionate appearance locations are indicated.
The previously confirmed connection with Oćuša spring (Ivičić & Pavičić, 1997) was not confirmed during the more recent experiment.However, we did confirm the connection of Vrgoračko polje ponors with spring Klokun, where the tracer arrived in high concentration and with a noticeably high velocity of 3.23 cm/s.This discovery is of critical importance for the design of sanitary protection zones in the area because, thus far, researchers have recorded significantly lower speeds (0.73 cm/s) (Ivičić & Pavičić, 1997).According to this result, the traced ponor within Vrgoračko polje, around the river Matica, must be included in a rigorous protection regime.
In the Main map, all previous tracing results ( Bojanić, 1980;Ivičić & Pavičić, 1997;Slišković, 1994Slišković, , 2000Slišković, and 2014) ) were analysed during catchment   vić, 2014), the groundwater divide spreads to the Rakitno polje, the outermost section of the Prud catchment.East from Rakitno, the groundwater divide is morphological as it goes along the mountain Čabulja; somewhere in that zone, transverse groundwater divide goes through Mostarsko blato.Water from this area flows from Mostarsko blato and Trebižat River to the Neretva River.Part of the water flows underground following surface flow, but with dye tracing, also a diagonal and transversal water flow is confirmed.

Conclusion
The study has contributed to our understanding of the hydrogeological aspects of this large karst catchment area.It covers an area of around 2000 km 2 and stretches across two neighbouring countries.The area is characterised by deep and intense karstification, which results in numerous surface karst forms and a complex subsurface fracture-conduit system that generates unpredictable groundwater flow.The presented comprehensive hydrogeological investigation has shown that this system has a zone of underground and surface runoff (sometimes intertwined) between flysch hydrogeological barrier on the Adriatic coast and the Neretva River.Furthermore, research indicated that most of the investigated major springs have a similar regime, suggesting that they recharge from the same aquifer in their hinterland.Hence, higher concentrations of sulfate ions in several springs in Southern Dalmatia are also an indication of recharging from more distant areas -mainly west and north of Tihaljina-Mlade-Trebižat River, where evaporite rocks are present and where the domination of sulfate ions in water is also apparent.The primary outcome of our work was a delineation of a big karst catchment into six subcatchments areas from which surface and groundwater flow in different directions.Their boundaries are positioned according to geological setting and hydrogeological relations.The subcatchment boundaries should not be considered as strict divides because, in different hydrological conditions, groundwater or surface water intersects them.Yet, they are fundamental for water balance calculations and estimating their water's influence on (sometimes distant) karst springs.Furthermore, new terminology which defines subcatchments as influential areas is used.The new terminology can be used when describing similar karst catchments of such proportions where subcatchments interaction dominates.Also, a basis for future research, that can further expand the knowledge of catchment behaviour, was given.In addition, this should improve the quality of establishing sanitary protection zones, and possible use of additional quantities of spring water in the water supply during summer seasons as well as better management of transboundary catchments, which can often be challenging due to countries' development differences and policies.

Figure 1 .
Figure 1.Study area position in: (a) sketch of Mediterranean karst area (Bakalowicz, 2015) and (b) cross-border countries Croatia and Bosnia and Herzegovina.

Figure 3 .
Figure 3. Piper diagram -median ion values (*one sampling campaign ion values of springs in Bosnia and Herzegovina).

Figure 5 .
Figure 5.The difference in the results obtained by the methods of Turc and Langbein for the period 1961-1990(Terzić et al., 2015).Subcatchments numbers are explained in Table1.

Figure 6 .
Figure 6.Monitoring points and tracing experiment results.

Table 3 .
Basic tracing data from the northwestern part of the Vrgoračko polje with fluorescein.

Table 4 .
Basic tracing data from the northwestern part of the Vrgoračko polje with naphthionate.