4.1. Hydrochemical Analysis of the Study Area
The minimum, average, and maximum values of ion concentration, pH, and electrical conductivity (EC) are shown in
Table 1.
The average values of the major ions were used to construct a Piper diagram (
Figure 4). At Opačac, Prud, and Klokun springs, the distribution of major ions is characteristic for limestone-dominant aquifers, with Ca
2+ as the most abundant cation, followed by Mg
2+, Na
+, and K
+. Mandina mlinica spring site is abundant in Na
+ and Cl
- due to sea proximity; Mg
2+ and K
+ are elevated for the same reason, whereas Mandina mlinica abstraction site shows typical chemistry for karst groundwaters. The observed values of pH ranged from 7.28 to 8.14, and they were within the standard range of the freshwater (i.e., from 6.5 to 8.5). Values of electrical conductivity (EC) at Prud, Klokun, and MM
A were elevated but well within the drinking water threshold (<2500 µS/cm) [
56], as a result of marine influence due to their position close to the coast. At MM
S, even the minimum values of EC and Cl
− evidenced total mixing of spring water with the seawater. Contrary values observed at the two adjacent springs at MM
S and MM
A demonstrated the heterogeneity of karst systems in terms of preferential flows and differences in freshwater/seawater interface. Furthermore, the dissolution of sulfate-rich minerals (i.e., gypsum, anhydrite) from evaporite deposits could also be reflected in the elevated concentration of sulfates (SO
42−) and subsequently, EC. Evaporite deposits are common in the central and southern Adriatic basins, and Bosnia and Herzegovina [
57]. As the majority of Prud’s catchment is located in the neighboring country, evaporite deposits in the underground were the most likely cause of increased sulfate concentration. Additionally, the anthropogenic activity (industry, agriculture, domestic wastewater) could also add up to the high concentration of SO
42−. Sulfates are abundant in seawater at 2.712 g/L in 35 Practical salinity units (PSU) [
58], therefore, a marine origin is also plausible. Perry et al. [
59] have given a strong argument for using the SO
42−/Cl
− ration as a seawater mixing indicator. The average value for seawater is 10.3. Adhering to that, the results confirm the authors’ assumption regarding seawater mixing in Prud and Klokun (
Table 1).
Relatively high bicarbonate (HCO
3−) concentrations confirmed that the groundwater flow occurs in a carbonate aquifer. Nitrate concentrations were generally low, i.e., up to 5.35 mg/L at all analyzed springs, and they did not exceed the standard drinking water threshold (50 mg/L) [
56]. Although there are many possible sources of nitrate in the groundwater, the majority of them are linked to human activities. However, intensive agriculture and husbandry in the study area were not reflected in high nitrate concentrations in the groundwater. The most probable causes of a low concentration of NO
3− were: (i) high discharges at Opačac, Prud, and Klokun (i.e., on average, 4.6, 5.7, and 1.4 m
3/s) resulting in a high dilution of nitrates; (ii) preferential and piston groundwater flow paths through a highly fractured and karstified underground, and (iii) complex groundwater recharge and mixing patterns in the overlying catchments. The reliable interpretation is constrained by the lack of data on nitrate concentrations from surface waters and soil. In particular, rivers in karst
poljes are the main recipients and receive a surplus of nitrogen that is not taken up by plants, and a high concentration of nitrates is expected, particularly during several spraying seasons throughout the year. This was supported by low values of potassium at all analyzed locations except in MM
S, where high values of K
− originated from the seawater. A similar study was conducted in China [
60] where the authors confirmed karst hydrogeochemical parameters such as pH, conductivity, HCO
3−, Ca
2+, SO
42−, and Mg
2+ as good indicators of various land use and other environmental changes.
The Piper diagram describes the dominant hydrochemical facies of the investigated springs (
Figure 4). Opačac, Klokun, Prud, and MM
A belong to the calcium type, i.e., Ca-HCO
3 and mixed Ca-HCO
3 and Ca-Mg-Cl facies. MM
S belongs to the Na-Cl type due to its coastal location. Since the study area consists of carbonate deposits, i.e., limestone and dolomite, they add a significant amount of Ca
2+ and Mg
2+ to the water.
The bacteria, total organic carbon, and inorganic nitrogen content are shown in
Table 2.
Assessing the given microbiological data, what immediately stands to attention is the high aerobic bacteria content (6000 units/mL) at Prud. Total coliforms follow suit with 1800 units in 100 mL, as well as fecal bacteria. This reflects the poor purification capacity of karst aquifers due to the thin or non-existent overlying soil and, generally, the absence of fine-grained materials in the aquifer (i.e., clay, sand) [
61]. Total organic carbon values and inorganic nitrogen content show a slight correspondence with bacterial colonies, especially for Opačac and Prud springs. Pronounced fractions of fecal bacteria are derived from animal manure coupled with higher values of inorganic nitrogen building a connection with agricultural runoff (fertilizers) [
62]. Groundwater pollution generated by intense nitrogen fertilization is a common issue in numerous agricultural communities [
62,
63,
64].
In general, the water is of good quality, with slightly elevated sulfates most likely on account of evaporite deposits from the neighboring Bosnia and Herzegovina. The microbiological data is high, but the water is being chlorinated prior to distribution to end-users and thus properly managed. Even though the study supposed that the continual monthly hydrochemical monitoring over a period of hydrological year is sufficient, the obtained results showed otherwise. The limitations of the hydrochemical study were the sparse data set and lack of continued daily monitoring, therefore, a more intricate groundwater monitoring network is needed for a conclusive study, over a longer period of time [
64].
4.2. Assessment of Climate Change Impacts
Hydrological modeling and assessment of climate change impacts on water resources were performed for Imotski polje and the catchments of Prud, Klokun, and Mandina, respectively. The measured data on mean annual air temperature and precipitation from 1961 to 2012, as well as the data on the spatial distribution of the runoff, and their generated values for the period until 2050, were used in the analyses. Statistics of the mean annual air temperature and the values generated by Aladin, RegCM3, and Promes are shown in
Table 3.
In comparison to the 30-year reference period (from 1961 to 1990), all models evidenced an increase in the average values of mean annual air temperature until 2050. In Imotski polje, Promes evidenced the highest increase in the average value of mean annual air temperature, up to 1.4 °C, followed by Aladin with 0.5 °C, and RegCM3 evidenced a negligible increase of 0.1 °C. Slightly lower values were observed in the catchments of Prud, Klokun, and Mandina mlinica, where increases of 0.9 °C, 0.6 °C, and 0.1 °C were indicated by Promes, Aladin, and RegCM3, respectively. In both parts of the study area, the maximum values of the mean annual air temperature tend to increase more significantly until 2050. For Imotski polje, Promes showed the highest increase, up to 2 °C, followed by Aladin with 1.3 °C, and RegCM3 with 0.5 °C. In the catchments of Prud, Klokun, and Mandina mlinica, slightly lower increases of 1.4 °C, 1.3 °C, and 0.8 °C were evidenced by the models. Regarding the minimum values of the mean annual air temperature, mixed results were obtained, whereas RegCM3 evidenced a decrease of −0.4 °C in both parts of the study area, while Promes and Aladin showed a slight increase.
The linear regressions evidenced increasing trends of 2.3 °C/100 years by Promes, followed by Aladin with 0.9 °C/100 years, for Imotski polje. Similar results were obtained for Prud, Klokun, and Mandina mlinica catchments, where Promes evidenced an increase of 1.4 °C/100 years, and Aladin 1 °C/100 years. In both parts of the study area, RegCM3 provided a negligible increase in the mean annual air temperature until 2050. Due to the long tradition of using RegCM3 in Croatia for climate predictions, the models often display conservative projections, small warm biases in air temperature, and general overestimation of precipitation [
9]. Furthermore, the increasing trends in the mean annual air temperature obtained by Promes and Aladin fit well into regional studies and attest to the hypothesis of faster warming trends in the Mediterranean region. In particular, Branković et al. [
9] described warming rates in the Adriatic region and showed an increase in the mean annual air temperature of 0.07–0.22 °C/decade for the period from 1951 to 2010, and 0.29–0.71 °C/decade for the period from 1981 to 2010. Furthermore, they projected the most significant increase for the summer and autumn periods, with an increase of up to 5.5 °C until 2100. Similarly, a higher intensity of warming during the warmer parts of the year (i.e., spring and summer) was evidenced by Bonacci et al. [
5]. Moreover, the observed results are in accordance with Berkeley Earth [
11], grid-based datasets of global air temperature. They described the current mean rate of air temperature change of 2.94 ± 0.26, 3.54 ± 0.33, and 3.35 ± 0.40 °C/100 years for Europe, Croatia, and Split, respectively.
Statistics of the mean annual precipitation and the values generated by Aladin, RegCM3, and Promes are shown in
Table 4.
All models pointed out the stable or slightly decreasing trends in precipitation until 2050. In particular, Aladin evidenced a decrease of approximately 57 mm/100 years, RegCM3 of 61 mm/100 years, and Promes by 129 mm/100 years in Imotski polje. Furthermore, a more significant decrease in the maximum values of the mean annual precipitation was evidenced by all models, i.e., by approximately 100–250 mm (
Table 4). Promes indicated the most significant decrease, while Aladin and RegCM3 showed similar predictions. Conversely, the minimum values of the mean annual precipitation are expected to increase by 200–250 mm until 2050. A nearly identical situation was observed in the catchments of Prud, Klokun, and Mandina mlinica, where models indicated slightly decreasing trends in precipitation of 57 mm/100 years (Aladin), 62 mm/100 years (RegCM3), and 131 mm/100 years (Promes). Additionally, the behavior of the minimum and maximum values of the mean annual precipitation is similar to the one observed in Imotski polje. These results are in accordance with most climate models for the Mediterranean region which predict a gradual decrease in precipitation and an increase in variability. However, these model-based changes in precipitation are usually characterized by high uncertainty [
5,
9,
12].
Based on the available data on the mean annual air temperature and precipitation, and their spatial distribution in the period from 1961 to 1990, the mean annual runoff for the same 30-year reference periods were calculated by Turc and Langbein empirical expressions.
Table 5 shows the statistics (minimum, average, maximum) of the time series of the calculated mean annual runoff in the period from 1961 to 1990.
Generated time series of the mean annual runoff in the periods from 1991 to 2020, and from 2021 to 2050 according to Aladin, RegCM3, and Promes are shown in
Table 6.
Furthermore, the time series of the historical and generated values of the mean annual runoff in Imotski polje and catchments of Prud, Klokun, and Mandina mlinica are shown in
Figure 5.
The results evidenced that the average values of the mean annual runoff in Imotski polje will decrease until 2050. In particular, Aladin and RegCM3 evidenced a decrease of 10% and 8%, respectively, while Promes indicated a 15% decrease when compared to historical runoff (from 1961 to 1990,
Table 4 and
Table 5). The maximum values of the mean annual runoff showed a more significant decrease, namely by 10%, 8%, and 22% according to Aladin, RegCM3, and Promes, respectively. All three models indicated an extreme increase of the minimum mean annual runoff, namely by 44%, 100%, and 88%. In the catchments of Prud, Klokun, and Mandina mlinica, all three models evidenced an almost identical decrease in the average values of the mean annual runoff as in Imotski polje. Projected changes in the maximum values of the mean annual runoff are similar to those projected for Imotski polje, with a foreseen decrease up to 10%, 12%, and 22%. Regarding the minimum values of the mean annual runoff, the same increase as in Imotski polje was evidenced by all three models.
Furthermore, from 1971 to 2000, E-OBS datasets evidenced 34 to 40 CDD in Imotski polje and catchments of Prud, Klokun, and Mandina mlinica. EURO-CORDEX RCMs predicted an increase in CDD by 3 to 8 days from 2021 to 2050, for the 75th percentile under RCP4.5 and RCP8.5 scenarios. In the period from 2071 to 2100, CDD is expected to increase by 8 to 13 days in coastal parts of the study area, and by 3 to 8 days in the hinterlands, where the continental effect of the climate is more pronounced. This increase in CDD could be reflected in the prolonged droughts, especially during the warmer season (i.e., spring and summer) when the hydrological minimum occurs, posing a threat to agricultural production, which is characterized by a low percentage of irrigated areas and a relatively high percentage of production that requires irrigation (e.g., fruits). Regarding the CWD, which is far greater in mountainous regions, E-OBS indicated 9 to 11 CWD in the period from 1971 to 2000 in the study area. Models showed a mixed-signal for the period from 2021 to 2050, i.e., for the 25th percentile, both scenarios predicted a decrease by 1 to 4 days, while stable trends were provided for the 50th and 75th percentile. The same result was projected for the period from 2071 to 2100. Furthermore, the median value (50th percentile) under RCP8.5 predicted reductions in CWD mostly across Italy and Balkan countries, which could have a significant impact on the duration and intensity of the flood waves in the study area. Concerning Rx1D, climate models indicated a clear increasing signal on a short time horizon (2021–2050), and according to the median value, the maximum annual daily rainfall could increase from 5 to 15 mm. The same result was obtained for the 75th percentile. On a long time horizon (2071 to 2100), the most significant increases in Rx1D were detected for the 75th percentile, from 15 to 25 mm in lowland parts of the study area, and 25 to 35 mm in mountainous areas.
Since the pressure on water resources will continue to grow, owing to adverse climate change effects (prolonged drought, consequently, greater seawater intrusion), observation and extended analyses of the endangered water resources is an obligation.