Evidence of karstification in chalk and limestone aquifers connected with stream systems and possible relation with the fish ecological quality ratio in Denmark

Although chalk aquifers are not too often associated with conduit flow, they are highly productive groundwater systems and, like limestone aquifers, they can be vulnerable to contamination when exposed to land use activities. The Danish carbonate rocks are generally recognized to be highly fractured and covered by thick Quaternary sediments. Fissure flow is pronounced, occurring in the upper 50–100 m due to Pleistocene glaciations. According to recently published maps of the distribution of karst in Europe, Denmark has no karst. However, this study concludes that karstified chalk and limestone aquifers are an important source of freshwater in Denmark. Four national datasets on karst features, groundwater flow, groundwater chemistry, and fish ecological quality ratio (EQR) data now indicate more heterogeneous structures and preferential flow pathways in the chalk and limestone aquifers than had been conceptualized and modelled with a national water resources groundwater/surface-water model in the recent past. This study provides new qualitative evidence that rapid and preferential flow of water and agrochemicals from the surface through thinner parts of the Quaternary cover layers, sinkholes and solution-enlarged fractures may likely impact the vulnerability of chalk and limestone aquifers. Additionally, due to the preferential flow system, some gaining streams discharged by karstified chalk and limestone aquifers show increased fish EQR values when using observed river daily discharge data instead of simulated daily discharge.


Introduction
Around the North Sea in North-West Europe, there is a chalk and limestone basin that occurs in England (UK), France, Belgium, the Netherlands, Germany, Sweden, and Denmark. The onshore margin of this basin is today used for water supply in most of these countries (Downing et al. 1993). Chalk and limestone aquifers supply about one-third of Denmark's potable water (Vangkilde-Pedersen et al. 2011). Around 15% of the ice-free continental land surface consists of carbonate rocks (Goldscheider et al. 2020), and more than 9% of the world's population depends on water from karst aquifers (Stevanovic 2019)-for example, groundwater from karst systems in Southern Europe and South-West China contributes 50% or more to regional freshwater supplies (Lu 2007;Hartmann et al. 2014;Chen et al. 2017;Kurkova et al. 2019). Clean and sufficient groundwater from these important aquifers in Denmark and elsewhere contribute to the maintenance of the ecological status in groundwater-dependent ecosystems associated with springs, river valleys, and the river network (Bonnaci et al. 2009) throughout wet and dry seasons in a cold climate (Graeber et al. 2017;Johansen et al. 2018) and warmer climate (Liu et al. 2009). A recent assessment of the quantitative status of Danish groundwater bodies in relation to the European Water Framework Directive River Basin Management Plans has revealed that about six of the 30 large chalk and limestone groundwater bodies are in poor quantitative status due to overexploitation of the aquifers (European Commission 2006;Henriksen et al. 2021a).
The development of karst drainage structures depends on the hydraulic gradient and the flux of carbon dioxide through the karst system (Bakalowicz 2005). Geomorphology and hydrology in karst terrains define the development of characteristic karst features like sinkholes, dolines, karst lakes, karst caves, and disappearing streams (White 1988). Karstified aquifers show some characteristics as a result of dissolution and physiographic features that may favor rapid, turbulent flow from the land surface through subsurface pipe-like voids and conduits that may reach the land surface forming visible springs or less visible wet surfaces near streams (MacDonald et al. 1998;Taylor and Greene 2008;Worthington and Soley 2017). This has implications for the vulnerability of the karstic chalk aquifer and the potential pollution risk of groundwater-dependent ecosystems due to interactions between aquifers, streams, and springs (Kresic and Mikszewski 2013;Goldscheider 2019;Hartmann et al. 2021;Kalvans et al. 2021;Maurice et al. 2021).
Karstic aquifers are often located in the shallow parts of the groundwater system and, hence, are connected to springs and gaining river reaches, thereby securing high quality and often stable ecological baseflow and high flow pulses important for the renewal of macrophytes and fish habitats in the stream network (Yang et al. 2008;Henriksen et al. 2021b;Graeber et al. 2014Graeber et al. , 2017. Comprehensive literature describes the relationship between stream flow statistics and the ecological flow regime of macroinvertebrates, macrophytes, and fish around the world Poff and Zimmerman 2010;Hoekstra et al. 2012;Westwood et al. 2017). Many EU member states have used ecological flow standards in implementing plans for the assessment of ecological flows in River Basin Management Plans (European Commission 2015; European Environment Agency 2021). Ecological flows were derived as part of the second river basin management plan (2014)(2015)(2016)(2017)(2018)(2019)(2020) and have been implemented again in the third river basin management plan for 2021-2027 for all river basin districts in Denmark (Henriksen et al. 2021a;Liu et al. 2020a, b).
A recent mapping of the distribution of exposed carbonate rock with karst development across Europe by Chen et al. (2017) indicates that Denmark has no karst aquifers (only 0.1% of carbonate rock outcrops). The European karst map by Chen et al. (2017) has used a pragmatic approach where only exposed carbonate rocks with karst development have been included in the hydrogeological map. Karst aquifers in Denmark are covered by thick Quaternary sediments; therefore, they have been overlooked and not recognized as karst aquifers by Danish hydrogeologists and in the international karst community. This is even though karst landscapes and other karst features in chalk and limestone aquifers have been highlighted in the Danish geological literature since the beginning of the 19th century (e.g. Ussing 1899; Jessen 1905). Baekgaard et al. (1982) published a detailed hydrogeological map of the European Union with karstified Upper Cretaceous chalk in central Jylland, which has been incorporated into a map of aquifers across Europe (BGR 2015).
The purpose of this study has been (1) to evaluate the degree of karstification in chalk and limestone deposits in Denmark; (2) to gather the available evidence on rapid (preferential) groundwater flow and karst behavior in groundwater and discharge to gaining streams connected to chalk and limestone aquifers in Denmark using national data on geology, hydrology, geochemistry and ecology, and (3) to discuss to which extent the possible karst development could affect the vulnerability of groundwater bodies and groundwater-dependent ecosystems. It is hypothesized here that, based on new understanding of karst behavior in confined chalk and limestone aquifers in Denmark, one can implement more heterogeneous structures and preferential flow pathways than had been conceptualized and modelled with a national water resources groundwater/surface-water model in the recent past (Henriksen et al. 2003). The need for a more accurate conceptualization is warranted in studies that support policy for protecting river discharge from karstified aquifers resulting in improved conditions for fish and more efficient groundwater source protection planning.

Research area
Denmark is located in Northern Europe and has an area of about 43,000 km 2 (Fig 1a). The average annual precipitation is about 750 mm. The average air temperature is 8-9 °C and is identical to the temperature of Danish groundwater. Several small to larger streams exist; about 85% of Denmark's streams are connected to aquifers with approximately equal portions in glacial derived sandy and clayey sediments (Sechu et al. 2021) and a small portion (about 100 km) to chalk and limestone sediments. The vast majority of the stream network represents gaining river conditions. Data on the stream network used in this study were created and are available at the Department of Ecoscience, Aarhus University, Denmark. The aquifers with which the stream network is in contact are based on the geological layers extracted from the Danish National Water Resources model (Henriksen et al. 2003) commonly known as the DK model.
Denmark is a low-lying land with hills generally less than 150 m high, and the Quaternary cover layers are dominated by glacial deposits from the latest glaciation (Weichselian) (Pedersen 2012). The Quaternary deposits in the research area are dominated by more than 25 m of cover layers and smaller areas with less than 20-25 m thickness, and in a few places chalk and limestone crops out at the ground surface (Fig. 2). Morphological features of the glacial deposits related to karstification of the upper parts of fractured chalk and limestone aquifers have been interpreted and reported in the Danish geological literature (Nilsson and Gravesen 2018). Based on this, a generic geographical information system (GIS) database with information on karst feature types, underlying carbonate lithology, thicknesses of glacial cover layers, and geographical position, was compiled and evaluated. In the research area, sinkholes are distributed over the entire area. In the western part are large karstic lakes (e.g. Vandet lake) located, and in the central part are gaining streams fed by the famous karstic springs in Rold Forest, forming a chalk aquifer-stream system (Berg 1951;Fig. 2).

Distribution of the chalk and limestone aquifers and stream system
The chalk and limestone aquifers were deposited during the geological periods of Upper Cretaceous (99.6-65.5 Ma) and Paleocene deposits from  Ma). The Upper Cretaceous chalk deposition in Denmark occurred in a part of the large North European carbonate sea where Denmark was covered by the sea and consisted of at least 450 m of chalk deposits (Surlyk et al. 2013). The chalk is a fine-grained deposit where more than 80% of the chalk mass has a grain size of less than 5 μm. The chalk in Denmark is a white and yellow-white micrite chalk that is generally soft except for thin yellow hard grounds, occasionally black chert layers, and thin marl layers. The main component is coccoliths but small clay content occurs (Håkansson et al. 1974). The Mid Danian carbonate consists of Bryozoan reef limestone interbedded with chert (Surlyk et al. 2006) and Danian chalk of muddy or micritic chalk mudstone (Blinkenberg et al. 2020). The Late Danian limestone deposits, named København limestone, is classified as a calcarenite and consist of a yellow-white limestone with some thick chert layers and are often a hard massive limestone (Stenestad 1976). The overlying Early Selandian b. c. Fig. 1 a Geological map of the carbonaceous rocks below the Quaternary cover layers in Denmark. The layer units from Upper Cretaceous to Selandian are all carbonaceous rock types. Geological layers in white color are deposited later than Paleocene. Locations with reported karst features are shown on the map. Dashed line box outlines the research area of this study. b Geological cross-section (X1-Y1, located in a) through Djursland where the chalk and limestone deposits gradually become younger from northeast to south-west (modified from Thomsen 1995). c Geological profile in TWT (TwoWayTraveltime, seconds) illustrating the structural elements and presence of thick records of sediments (up to 5 km thick) in the Danish Basin and slightly thinner in the Sorgenfrei-Tornquist Zone (line X2-Y2, a). The blue layer is the Upper Cretaceous chalk and Danian limestone (Chalk Group), the yellowish, greenish, and reddish layers are interbedded sandstones and mudstones, and the purple layer at the base is the Zechstein Salt (modified from Mathiesen et al. 2021) (Lower Paleocene) deposits were laid down after a period of erosion and often begin with glauconitic conglomerate and greensand, named Greensand limestone, followed by a massive light grey silty marl with strong bioturbation and without layering, named Kerteminde Marl (Heilmann-Clausen 1995). The pre-Quaternary unconformity thus truncates the Danian as well as Selandian units (Hallam 1984).
The geological and structural framework of the subsurface throughout the research area can be illustrated with two geological profiles that are constructed in vertical sections through the subsurface. The profiles give an insight into the geological and structural elements of the subsurface, including the position of faults and salt structures as well as the thickness, extent, and continuity of geological layers. Figure 1b shows a cross-section (line X1-Y1 in Fig. 1a) through Djursland with carbonate rocks from Upper Cretaceous to Danian where the chalk and limestone gradually become younger from northeast to southwest. The profile shows the stratigraphical position of the Upper Cretaceous chalk and Danian limestone and chalk layers in the subsurface. Figure 1c is a geological profile showing the largescale structures through the central and northern Jylland area (line X2-Y2 in Fig. 1a). The section is based on composite seismic profiles forming the background images (Mathiesen et al. 2021). In the section are highlighted layers from the Permian to Cretaceous. The Zechstein Group (Permian) contains thick salt deposits that locally have moved upwards to form salt diapirs or salt pillows in the Cretaceous and Danian sequence of chalk and limestone deposits. The chalk and limestone layers above and in the neighborhood are structural and chemically disturbed by the salt structure, with causes high salt concentrations in the surrounding aquifers. In the water management of the freshwater resource, the chalk and limestone aquifers above the salt structure have been omitted as a groundwater body in the research area (Troldborg 2020).

Databases and interpretation
This paper uses four national data sets to support hypotheses about the presence of overlooked areas with potentially karstified aquifers in Denmark. The datasets include: (1) karst features, (2) groundwater flow based on calculated hydraulic conductivities (K) derived from pumping tests in carbonate wells; (3) water quality based on nitrate concentrations over the drinking water quality standard (50 mg nitrate per liter) in monitoring wells in chalk/limestone and strontium isotopes in surface waters, and (4) systematic bias in fish-based EQR calculations in streams connected to chalk/limestone aquifers.

Pumping test data
A data set of transmissivity data from about 9,500 boreholes solely obtained from pumping tests in fractured chalk and limestone wells was collected from the national well database (JUPI-TER) of groundwater and water quality (GEUS 2022a). These data of varying quality were calculated from pumping tests of different lengths using water level measurements from both The distribution of boreholes with pumping tests performed in chalk and limestone aquifers is presented in Fig. 3. The data are not spread evenly in the chalk and limestone groundwater bodies, which consist of about 30 chalk and limestone aquifers larger than 15 km 2 in horizontal extension (Troldborg 2020). The geographical coverage of boreholes with pumping tests is most dense in the Metropolitan area of Copenhagen and Jylland in the center of the chalk and limestone aquifers. Much lower density occurs in the peripheral areas of the chalk and limestone groundwater bodies in most parts of the country (Fig. 3). Data are calculated from pumping tests undertaken in boreholes that were drilled for groundwater abstraction purposes (Kidmose et al. 2022). About 1.5 % of the estimated K values derived from chalk and limestone wells represent fracture/fissure-dominated groundwater flow with a median K value of 1 × 10 -4 m/s (<1 × 10 -2 and >1 × 10 -6 m/s). Approximately 3-4 % of all boreholes have a calculated K value lower than 1 × 10 -6 m/s (matrix-dominated flow).

Water quality in connected chalk and limestone aquiferstream systems
Diffuse nitrate contamination of groundwater from agriculture is causing severe environmental problems in Denmark (Carstensen et al. 2007). National regulations aimed at reducing the leaching of nitrate were successful, reducing the leaching of excess nitrate by approximately 33% on a national level (Kronvang et al. 2008). However, the Himmerland region in the northern part Fig. 3 The distribution of boreholes in Denmark with pumping tests in the chalk and limestone aquifers (in total 9,500 tests) as reported to the national borehole database. The boreholes are differentiated according to lithology. Latest delineation of the chalk and limestone aquifers (GWB2020) by Troldborg (2020)  (Fig 1a) shows a 31% reduction in nitrate leaching to groundwater, but only 8% of the monitored streams show a downward trend in nitrate concentration (Kronvang et al. 2008). In this area, many karstic springs discharge to the Lindenborg stream valley, where the spring water and some stream water sites were sampled in February 2015 for nitrate measurement; continuous measurements of natural tracers such as temperature and electrical conductivity (EC) in karstic springs have been successfully used to not only provide qualitative information about the karst drainage pattern (Liñán Baena et al. 2009;Luhmann et al. 2011) but also to estimate flow conduit volumes and mixing ratios by studying their time lag to precipitationinduced discharge peaks (Birk et al. 2004;Screaton et al. 2004).
In this field study, continuous measurements of EC at 30-min intervals from the Egebaek spring and Lindenborg stream, ~300 m upstream and downstream from the spring discharge to the stream, will be presented to show how the karstic spring and stream in a karstic area react to a precipitation event of 25 mm on 25-26 March 2015.
A national water quality dataset with time series of nitrate data in chalk/limestone aquifers is used for the assessment of the chemical status of the groundwater bodies as part of the third EU river basin management plan (Thorling et al. 2020). For each monitoring point, the mean value of the annual mean of nitrate concentrations for the period 2013-2018 was estimated by Thorling et al. (2020).
The source of water can be traced based on the samples' strontium (Sr) concentration and strontium isotopic ratio (Wang et al. 2006;Shand et al. 2009). Strontium isotope ratio is a conservative tracer with the 87 Sr/ 86 Sr ratio of waters not changing considerably from weathering of minerals until it reaches the surface waters through underground pathways. Anthropogenic Sr sources, for instance, fertilizers, can however change this isotope ratio (Hosono et al. 2007). The geographical distribution of strontium isotopes ( 87 Sr/ 86 Sr) in Danish surface water (streams, springs, and lakes) was studied by Frei (2011). Frei et al. (2020) found that Sr and strontium isotopes showed a close relationship, on a national scale, between groundwater in pre-Quaternary carbonate sediments and glacial-derived sediments and surface waters (stream, spring, and lakes). In this study strontium isotope compositions have been adopted from Frei and Frei (2011) at 25 surface-water sampling locations in streams, springs, and lakes. All the sampling sites can be related to Quaternary aquifers of glacial derived sediments or chalk and limestone aquifers.

Fish EQR ratio for chalk-and limestone-aquifer-related streams
The approach taken here is to build on the flow-biota relationship between flow regime statistics and fish ecological quality ratio developed by Graeber et al. (2014Graeber et al. ( , 2017, for identifying the ecological flow indicators of the gaining streams potentially connected to karstic aquifers by looking for areas with systematic bias in the calculated simulated fish EQR by comparing results of (Eq. 1) using observed and model-simulated discharge data: In the empirical Eq. (1), BFI signifies the baseflow index (baseflow volume divided by total flow volume); SIN is the class of stream sinuosity ranging between 1 and 4, where 1 is a straight channel, and 4 is a fully meandering stream calculated from stream length divided by linear distance; Fre25 is the annual frequency of events with flows above the 25th percentile (derived from flow duration curve), and Fre75 is the annual frequency of events with flows below the 75th percentile (derived from the flow duration curve).
It is hypothesized here that a reason for this directed bias (systematic underestimation when using modelled instead of observed daily discharge time series) is preferential flow processes due to immature karstic chalk and limestone aquifer conditions. To the authors' knowledge, no one has previously used evidence-based empirical equations for fish EQR (Eq. 1) for identifying potential karstic preferential flow dynamics.
For the ecological flow signature, the empirical formula (Eq. 1) for fish-based EQR has been derived for Danish streams (Graeber et al. 2014). The EQR ratio is expressed as a numerical value between 0 and 1, with high ecological status represented by values close to 1 and bad ecological status by values close to 0. The fish EQR is classified by the following thresholds: high (>0.94), good (0.72-0.94), moderate (0.40-0.71), poor (0.11-0.39) and bad (<0.11).
In the following, the simulated difference in fish EQR is used, where Q is flow rate or discharge. The main criteria used are: The justification for using a threshold of 0.06 is that it is a significant difference (or impact) when compared to changes that one sees for impacts from groundwater abstraction and/ or climate change which very rarely for Northern Jylland exceed a change of 0.06 when compared to the reference scenario, e.g. simulation compared to the simulation without abstraction (Liu et al. 2020b). Observed daily discharge time series and calculations of Fre25, Fre75, and BFI for the period 2004-2017 for potential immature karstified chalk and limestone, nonkarstic chalk and limestone, and Quaternary/pre-Quaternary dominated areas has revealed by trial-and-error that a set of supplementary criteria solely based on observed discharge could be made for providing a (2) ΔFish EQR Q obs , Q sim = fish EQR Q obs − fish EQR Q sim Criteria ∶ Δfish EQR Q obs , Q sim > 0.06 more robust and data-driven identification of karst areas for central and northern Jylland: Fre25-Fre75 (Q obs ) > 2.0, and BFI (Q obs ) > 0.80. The calculations were made by use of a cloud-based ICT tool (VandWeb) for comparing time series of observed groundwater abstractions, and simulated and observed daily flows developed for implementing and surveying ecological flows associated with licenses for groundwater abstraction (GEUS 2022b). VandWeb contains access to observed and simulated discharge data based on a national water resource model (Stisen et al. 2019;Henriksen et al. 2016). Flexible python tools allow calculations of fish, macrophyte, and macroinvertebrate EQRs in specified periods based on empirical equations developed by Graeber et al. (2014).

Geomorphological features
A survey of surface karst features by Nilsson and Gravesen (2018) has been supplemented with recent years of reports of surface karst features from citizens directly to the Geological Survey of Denmark and Greenland or via the media (in total 42 locations). The karst features can be separated into sinkholes that are funnel-shaped (including fluviokarst) or vertical-sided dolines, karst-lakes or sinkhole ponds with subsurface outlets, small caves, karst springs and losing stream sections to chalk and limestone aquifers (here named 'disappearing streams'). The locations of karst caves, karst landscapes, sinkholes/dolines, and disappearing streams occur across the entire country (Fig. 1a). Karst lakes and karst springs are so far only described for localities in the western part of Denmark.
Sinkholes/dolines have been mapped at three locations using high-resolution LiDAR elevation data in 1.6 and 0.4-m resolution with a vertical accuracy of a few centimeters (SDFE 2022). The regional frequency of these features is rather variable, with the highest density in the Thisted area ranging from 15 to 20 sinkholes/dolines per km 2 (Sørensen et al. 2017), and much lower density <1 sinkhole/doline per km 2 in the Kolindsund area (Jensen 2021) (Fig. 2). Sinkholes/dolines are primarily found in recharge areas. For comparison, a study in Chalk from southern England counts more than 150 dolines per km 2 in high-density areas (Mac-Donald et al. 1998).
It has been possible to make a characterization of a single sinkhole based on direct excavation of trenches across the sinkhole depression compared with the morphology of the high-resolution LiDAR data (Sørensen et al. 2017). It seems many circular depressions in the soil surface near Thisted and Kolindsund probably have an origin as cover-collapse sinkholes (Panno and Luman 2018).
Based on the present karst feature data, there is a relationship between locations with karst springs, lakes, and karst landscapes developed in Upper Cretaceous chalk, and disappearing streams above Bryozoan limestone, while sinkholes and caves develop in Upper Cretaceous chalk and Danian chalk and limestone (Fig. 1a). The majority of the karst features are located with less than 20-25-m thickness of Quaternary glacial derived sediments (Fig. 2). The observations (as part of this study) on findings of sinkholes in terrain at places with less than 20 m of glacial-derived sediments agree with North American studies in glacial terrains east of the Rocky Mountains where karst features (especially sinkholes) are predominantly observed in areas with a maximum of 15-20 m thickness of glacial derived sediments (Soller and Packard 1998;Weary and Doctor 2014).

High hydraulic conductivities in fault zones
The carbonate deposits contain tectonic faults and fractures. The faults are mainly vertical, while the fractures and conduits often follow bedding planes above or below chert and marl layers. The fractures and conduits are stained with yellow-brown Fe components from water flow (Rosenbom and Jakobsen 2005). Figure 4 shows Mesozoic fault zones that have been developed during the geologic periods Trias, Jura, and Cretaceous. It should be noted that the Mesozoic fault zones shown do not necessarily all cut the chalk layers in the Upper Cretaceous (See Fig. 1c).
The uppermost approximately 50-100 m of chalk and limestone have been disturbed by the Quaternary glaciers' large-scale glaciotectonic thrusting and folding. Isolated bodies of chalk or limestone are often included in the Quaternary layers. The erosion of the surface has also formed shallow depressions which have been further eroded by meltwater streams or depressions left in the land surface when ice, formerly covered by sediment, melts and kettle holes develop (Gry 1979). In an area south of Ålborg, the Skrivekridt is locally exposed and has a fractured aquifer that shows a combination of karst processes and fluvial erosion and deposition known as fluviokarst (Stenestad 2006).
Extraordinary high hydraulic conductivity (K) values represent K values above 1× 10 -2 m/s or more than 864 m/day, that are assumed to be reliable and represent hydraulic conditions in the chalk and limestone with pronounced fracture systems or fault zones with rapid non-Darcian (turbulent) groundwater flow (Marechal et al. 2008). The extreme high K values make up a ratio of 1.5% or a total of ~125 pumping tests. There appears to be a correlation between the boreholes in Upper Cretaceous chalk and the Mesozoic fault zones (Fig. 4). In the eastern part of the country, extreme high K values are primarily associated with the Greensand limestone and København limestone in the Metropolitan area of Copenhagen. These two carbonate rock types are characterized by a facies-controlled strong local-scale horizontal fissure development between hard and softer horizons in the two units (Jakobsen and Klitten 1999), that have probably been expanded further over time by chemical dissolution (karstification). High K values in Bryozoan limestone in the Copenhagen area are likely to be explained by local-scale glaciotectonic fractures. In the western part of the country, high K values occur evenly distributed between the carbonate rock types of København limestone, Bryozoan limestone, and Upper Cretaceous chalk. This kind of geological structure is prone to infiltration of surface water into the surface and discharge of groundwater from aquifers to gaining parts of surface waters.

Nitrate groundwater quality
High nitrate fertilizer concentrations (>50 mg nitrate/L) in chalk/limestone monitoring wells (red circles) are shown in Fig. 5 as average concentrations for the period of 2013-2018. There is a close relationship found between the location of monitoring wells with nitrate concentrations above 50 mg/L in chalk and limestone aquifers and cover thickness of up to 15-20 m.

Strontium in groundwater, streams, springs, and lakes
In accordance with Frei and Frei (2011) strontium isotope compositions of surface water connected to chalk and  (2022a)). The grey-shaded area is the latest delineation of chalk and limestone aquifers (chalk and limestone GWB2020, adapted from Troldborg 2020) Jylland Fyn limestone aquifers range from 0.7078 to 0.7082 in the central and northern Jylland. Surface water (streams, springs, lakes) in the research area connected to Quaternary aquifers (ks1-ks4) has strontium compositions larger than 0.7082. Strontium isotope composition data from Frei and Frei (2011) are shown in Fig. 5. Strontium isotope composition indicates that Sr isotopes are an excellent tracer to distinguish between Sr isotope samples collected in surface waters connected to chalk and limestone aquifers or Quaternary aquifers.

Water quality relations in the chalk aquifer-stream system
Nitrate concentrations measured in seven karstic springs in February 2015 reveal that the four springs discharging on the left bank of Lindenborg stream have elevated nitrate concentrations of 40-47 mg/L, while the three springs on the right bank show concentrations between 0.5 and 15 mg/L (Fig. 6). This distinction between the left and right stream valley can be attributed to land-use practices where the left bank of the stream valley is mainly used for agricultural production, while the right bank is a forested area. Nitrate concentrations in the stream also reflect the influence of spring water; nitrate concentration just upstream of the Egebaek spring is 16.8 mg/L, while a few hundred meters downstream this rises to 19.7 mg/L as a consequence of spring discharge with high nitrate concentrations to Lindenborg Stream. The reaction of the karstic Egebaek spring and Lidenborg Stream to a rain event on 25-26 March 2015 is presented by Strontium isotope ratio in surface waters indicates that the Sr source is groundwater originating from chalk/limestone aquifers (<0.7082) and glacial sediment aquifers (>0.7082) in accordance to Frei and Frei (2011). Streams connected to Quaternary (ks1-ks4) aquifers and chalk and limestone aquifers are also given. The locations with the discharge of groundwater originating from chalk/limestone discharge into lakes (location No. 1: Vandet lake), into springs (location No. 3: Trindborg spring), and into gaining stream sections of Dybvad Stream (location No. 2) and Villestrup Stream at two sampling locations (location No. 4) continuous EC measurements (Fig. 7). Before the rain event, both the upstream and downstream stream stations show a stable EC of 0.49 and 0.51 mS/cm, respectively, while EC in the Egebaek spring varies between 0.52 and 0.61 mS/cm. During the rain event, EC in Lindenborg Stream quickly decreases to 0.35 mS/cm both at the upstream and downstream stream station reaching a minimum value approximately at the end of the precipitation event. EC data from the Egebaek spring, however, does not reflect such a change and displays the previous variations in EC until 30 March 2015 with a time lag of 5 days after the start of the rain event. The decrease in EC is also much longer than for the stream. At both stream stations, EC values almost reach the original stable values 1 day after the event, while the karstic spring only returns to pre-event values 10 days after the onset of EC decrease.

Fish EQR index and flow dynamics in chalk streams
The stream network within the area with chalk and limestone aquifers is shown in Fig. 8. Those streams of the carbonate groundwater bodies where the carbonate The map also shows the location of the Egebaek spring and the upstream and downstream stations in the Lindenborg Stream where electrical conductivity (EC) was continuously monitored. The karst springs Kovadsbaek, Ravnkilde, and Lille Blaakilde were also investigated seven decades ago with measured discharge rates of 80-150 L/s by Berg (1951) Sjaelland Fyn Fig. 7 Electrical conductivity (EC) of spring water at the Egebaek spring and ~300 m upstream and downstream of the location where the spring discharges into the Lindenborg Stream (for the location refer to Fig. 6). The vertical black lines denote the duration of the rain event on 25-26 March 2015. EC values are recalculated to a reference temperature of 25 °C surface is less than 3m below the stream network in the national water resource model are shown with bold red lines (Troldborg 2020). It is assumed that there is potential hydraulic contact between the streams and underlying chalk and limestone aquifers on these stream stretches so that the stream can receive discharging groundwater through the stream bed. Catchments, where fish EQR calculated from observed discharge are more than 0.06 higher than fish EQR calculated from simulated discharge, are delineated in Fig. 8 (marked with 'Eflow').
Here the hypothesis is that observed discharge in karstic chalk areas has more pronounced peak flow fluctuation events (Fre25 above Q25) compared to nonkarstic chalk conditions, and at the same time without having an equally increased number of low flow fluctuation events (Fre75 below Q75). Karstic conditions at the time do not significantly impact modelled values compared to observed discharge calculated BFI.
Selected gauging stations for calculated Δ fish EQR (Q obs , Q sim ), Fre25-Fre75 (Q obs ), and BFI (Q obs ) with karstic and nonkarstic properties are shown in Table 1. Since the fish EQR Eq. (1) is developed for larger stream catchments (>10 km 2 ) the selected results are mainly from larger streams (and not small tributaries/headwaters). Figure 9 shows that the eco-hydrological status for the selected 19 streams in the research area can be classified into three grades (good, moderate and poor) based on the fish-based EQR index. Due to the recharge from the karst groundwater to the related streams, especially in the dry season, the specific low flow of karstic streams is generally larger than that of nonkarst streams. In Fig. 8, karstic streams are clustered towards the right side along the horizontal axis, and this would suggest that the ecological status of karstic streams is better.
In general the identified nine karstic stream catchments comply with all three criteria (Δ fish EQR > 0.06, Fre25-Fre75 > 2 and BFI > 0.8). Overall, these streams and catchments have a good-to-moderate ecological status when EQR is calculated based on Eq. (1) (the empirical relationship). For the other streams, one or more of the three criteria have not been met either because of BFI < 0.80, Fre25-Fre75 ≤ 2, or Δ fish EQR ≤ 0.06. If one only used the Δ fish EQR > 0.06 criteria then a total of four 'false positive' gauging stations would have been identified with karstic conditions out of a total of 19 gauging stations (i.e. 21% false positive). If one used only BFI and Fre25-Fre75 as criteria without using a comprehensive physically based groundwater/surfacewater model, only one false positive gauging station (i.e. 5% false positive) would have been identified. This means that the approach for central and northern Jylland is more robust. For the nonkarstic catchments, the calculated ecological status is typically moderate with one exception of 'Land Fjord Canal' (see Table 1), which gets the ecological status 'bad'. In general, the nine catchments with karstic streams are connected to chalk and limestone aquifers as expected, and two to lithostratigraphic deeper-lying Quaternary aquifers (ks3 and ks4). Fig. 8 The entire stream network is connected to chalk/ limestone (red line) and Quaternary aquifers (ks1 and ks4) in central and northern Jylland. Areas with systematic bias (underestimation) in fish ecological quality ratio (EQR) for simulated daily flow, compared to observed flow, are marked with 'Eflow' on the map where it is likely that the streams are connected to karstified aquifers or streams connected to Quaternary aquifers (ks1-ks4), where layer ks1 is most shallow and layer ks4 has a deeper lithostratigraphical location. Streams labelled 1-19 refer to stream numbers in Table 1 Eflow in streams connected to chalk and Quaternary aquifers (ks1 to ks4) Chalk  The 10 nonkarstic streams are predominantly 'without contact' or gaining streams in contact with the shallow Quaternary aquifer types ks1 and ks2; however, there are also examples of stream sections in contact with chalk and limestone aquifers for this group.

Discussion
In Paleocene time when the Danian and Selandian carbonates were deposited, significant and severe changes in sea level appeared. Thomsen (1995) interpreted up to four phases with very low sea level or even dry conditions with an outcropping of the chalk and limestone surface due to uplift and erosion. The paleo-geographical and climatic conditions indicate that there may have been the right conditions for the formation of karst in the surface of the chalk and limestone deposits in Northern Europe before the end of the Paleocene time (Downing et al. 1993). Northern Europe was located ~900 km further south than now so the Danish area was at a latitude of about 45 °N (Ziegler 1990). The climate was constantly warmer (around 20 °C) with arid climate conditions (Hancock 1975). For parts of the chalk and limestone surfacing in the Danish land area at that time, the conditions for the development of karst at the top of the limestone deposits were therefore ideal. Seismic surveys of the chalk and limestone surface (Top Chalk surface) under the North Sea suggest that karst topography was subaerially exposed in the North Sea in mid-Paleocene time (Jenyon 1987;Huuse 1999). The present karstification of the on-shore fractured chalk and limestone aquifers with fracture systems to a depth of 50-100 m has most likely been developed after the end of the most recent ice age 11,700 years before the present, allowing ample time for subsequent expansion of fissure openings by chemical dissolving processes of chalk and limestone materials along the fissures with the help of circulating groundwater. However, it cannot be ruled out that the karst formation already started during earlier interglacial/interstadial periods in the Holocene period. In addition, it cannot be excluded that some of the larger fault systems were formed in the late Mesozoic (Upper Cretaceous). However, regardless of the time of formation, these fault zones will also likely be exposed to circulating groundwater since the end of the last ice age, with karstification as a result. Thus far, limited observations have been made of fieldverified karst features on the land surface like sinkholes, springs, and disappearing streams in Denmark. In the subsurface, only very few caves are known. There may be doubts as to whether they are carbonate rock mines established 400 years ago when Denmark was a major exporter of flint material to neighboring countries. There is, as far as the authors are aware, no knowledge of underground drainage conduits dominated by rapid groundwater flow. However, it should be mentioned that many geophysical video logs in open carbonate rock abstraction boreholes show turbulent groundwater flow in fracture zones. Facies and structural studies using geophysical methods compared with cored borehole material are valuable to explore karst system conduits and fracture systems (Chalikakis et al. 2011;Chatelier et al. 2011).
Contamination of chalk aquifer groundwater with fertilizers (Foster 2000) and pesticides (Malaguerra et al. 2012) has been known since the 1960s. The hydraulic connectivity between the fractured chalk aquifer and the land surface is closely related to the thickness of the confining layer and geological heterogeneity. Chalk aquifers are most vulnerable to groundwater pollution where the cover layers are thin or absent. In areas where the thickness of the glacialderived sediments over the chalk in Denmark is less than 15-20 m, the chalk aquifers are categorized as very vulnerable. A similar assessment of chalk aquifer vulnerability is reported from the US where chalk or potentially karstified chalk aquifers superimposed with thinner glacial deposits (less than about 15 m) are categorized as very vulnerable (Weary and Doctor 2014). The geological and geochemical heterogeneity in glacial-derived sediments can be described by dividing the cover layer into conceptual submodels, each with an indexed vulnerability to pollution transport using the depth development of fractures and sand lenses, and analysis of complex redox structures, as a proxy for the spread of pollution vertically Hansen et al. 2021). Even minor heterogeneities with high hydraulic conductivity may form preferential flow paths and dominate flow and transport through the otherwise impermeable clayey till cover layers (Sidle et al. 1998;Harrington et al. 2007;Harrar et al. 2007;Kessler et al. 2013). Adams et al. (2016) point out that the absence of karst features on the land surface does not imply the absence of karst processes in the subsurface. It has been speculated that small and more simple sinkholes might indicate relatively small conduits in underlying carbonate bedrocks and larger structures might be situated close to underlying cave systems. Intensive forestry and agricultural operations within karst landscapes can more or less erase these features (Sørensen et al. 2017;Panno and Luman 2018;Hofierka et al. 2018).
Karstic preferential flow catchments show a tendency for more significant systematic bias in calculated fish EQR based on empirical Eq. (1), with an underestimated fish EQR value for simulated discharge compared to observed discharge mainly due to a more narrow Q75-Q25 band and underestimation of Fre25 for karstic streams compared to nonkarstic. At the same time, identified karstic catchments for observed discharge in central and northern Jylland have Fre25-Fre75 > 2.0 and BFI > 0.80 and good to moderate ecological status. Fish EQR, as pointed out by Liu et al. (2020a), is positively related to BFI and negatively related to Fre75, underlining the importance of a stable discharge regime with a generally rare occurrence of low flow events. The positive relation to Fre25 with more frequent slight disturbances due to weak peak flows (exceeding Q25) reflects an improved quality of the fish habitats and community. Since karstic preferential flow conditions support both dynamics, e.g. higher frequency of Fre25 compared to other aquifers and increased baseflow through conduit voids, such stream conditions are in general favorable for complying with a good status of such streams. Since Fre25-Fre75 is also part of the macrophyte indicator, which in addition has a duration of three times Q25 flow events, good status for aquatic plants is often also obtained for karstic streams. Finally, since also Q90/Q50 typically are above 0.5 for such streams, karstic streams also score good status if the physical/morphological index is in good status.
Other studies on the base flow index in karst areas in Indonesia, Bulgaria, and the UK are in good agreement with the chosen BFI > 0.8 in this study. An Indonesian study has estimated BFI values for karst springs up to 0.9 (Fatchurohman et al. 2018). Vasileva et al. (2017) used a BFI of 0.8 for Bulgarian karst systems, while the UK baseflow index is set at 0.9 for chalk (UK_CEH 2022)-for example, in the Pang and Lambourn river catchments in southern England, the BFI indexes were calculated to be 0.87 and 0.96, respectively (Maurice et al. 2006;Maurice 2009). However, the research reported here shows that the criteria limits that distinguish karst streams from nonkarst streams are not universal and can likely not be transferred to other parts of Denmark or abroad. Chalk and limestone streams for Sjaelland in the eastern part of Denmark do not fulfill these supplementary criteria, so either the chalk and limestone on Sjaelland are to a less degree karstic than in central and northern Jylland, or other stressors (urbanization and groundwater abstraction) impact the low flows (BFI and Fre75), requiring specific analysis of usable thresholds for Sjaelland and the eastern islands.
The authors believe that it is therefore the overall Δ fish EQR > 0.06 criteria, to a high degree defined by the difference in Fre25 and Fre75 between observed and simulated flow (and driven by pulses of rapid precipitation impacts), that may be transferable to Sjaelland and Fyn, even though BFI and low flows are almost zero in dry summers in this part of Denmark. However, a validation based on proxy basin tests of the developed methodology for Sjaelland and Fyn, and other areas in Northern Europe and abroad, is needed, and as part of that, more universal supplementary criteria eventually need to be identified.

Conclusion and perspectives
Although the chalk aquifers are not too often associated with conduit flow, they are highly productive groundwater systems, like limestone aquifers, and can be highly vulnerable to contamination when exposed to land use activities. This study shows that karstified chalk and limestone aquifers are important freshwater resources in Denmark. The occurrence of karst aquifers underneath confining cover layers of glacial-derived sediments is more widespread than hitherto recognized and new to the international karst community. In an updated version of the world karst map, Denmark and probably other parts of the world must be included in the map as a new category of karst aquifers. Evidence of rapid groundwater flow due to the presence of karstified chalk and limestone aquifers in Denmark has been evaluated using four selected indicators based on national data on geology, hydrogeology, geochemistry, and ecological flow indicators. Surface karst features indicate more heterogeneous structures and preferential pathways to the chalk and limestone aquifers than conceptualized at present in Denmark. Overall, the chalk and limestone appear to be gently karstic with local-scale karst development. Identification and implementation of appropriate measures to protect the aquifers require a proper conceptualization and modelling of the freshwater cycle in the chalk and limestone aquifers. A better understanding of the morphology, distribution, and evolution of the various karst features in the land surface is required to provide insights into the character of the underlying carbonate geology and karst aquifers in Denmark. The presented ecological flow ratio for fish has so far only been investigated for karst conditions in central and northern Jylland, Denmark. However, it must be considered quite probable that in areas elsewhere, a fish EQR equation can be established based on regression, and that the methodology described here can inspire criteria for ecological flow for other areas. It is concluded that, to the authors' knowledge, this is the first time that fish EQR empirical formulas have been used to evaluate possible karstic conditions based on an analysis of runoff time series in streams. permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.