Groundwater temperature anomalies in central Europe

As groundwater is competitively used for drinking, irrigation, industrial and geothermal applications, the focus on elevated groundwater temperature (GWT) affecting the sustainable use of this resource increases. Hence, in this study GWT anomalies and their heat sources are identified. The anthropogenic heat intensity (AHI), defined as the difference between GWT at the well location and the median of surrounding rural background GWTs, is evaluated in over 10 000 wells in ten European countries. Wells within the upper three percentiles of the AHI are investigated for each of the three major land cover classes (natural, agricultural and artificial). Extreme GWTs ranging between 25 °C and 47 °C are attributed to natural hot springs. In contrast, AHIs from 3 to 10 K for both natural and agricultural surfaces are due to anthropogenic sources such as landfills, wastewater treatment plants or mining. Two-thirds of all anomalies beneath artificial surfaces have an AHI > 6 K and are related to underground car parks, heated basements and district heating systems. In some wells, the GWT exceeds current threshold values for open geothermal systems. Consequently, a holistic management of groundwater, addressing a multitude of different heat sources, is required to balance the conflict between groundwater quality for drinking and groundwater as an energy source or storage media for geothermal systems.


Introduction
Groundwater is an important resource for society and industry. Within the European Union (EU), it is the main source of drinking water, supplying about 50% of the total demand [1]. However, it is equally important for agriculture. Depending on the country and type of agricultural production, up to 90% of the water for irrigation originate from groundwater [2]. In the industrial, commercial and residential sectors the use of groundwater as a resource for heating and cooling purposes is increasing worldwide [3]. Additionally, the surrounding ecosystem strongly depends on the groundwater quality and temperature [4][5][6][7][8][9][10][11]. Multiple uses of groundwater lead to high competition between different interest groups. Consequently, a holistic groundwater management in terms of quantitative, qualitative and thermal issues, as well as sensible regulations of this highly demanded source are essential [12,13].
The EU water framework directive (WFD) [14] defines the status of groundwater in terms of quantity and chemical quality. Groundwater quality and dependent ecosystems strongly rely on physical and chemical properties, which are in turn influenced by the groundwater temperatures (GWTs) [15,16]. The temperature determines natural bacterial and fauna community composition as well as biogeochemical processes [7,17]. An increase in GWTs enhances the propagation of pathogen microorganisms, which in turn endanger the hygienic state of groundwater and therefore its use as a drinking water resource [8]. Thus, the WFD classifies heat input into the aquifer as pollution. However, a study by Hähnlein et al [18] on the legal status of shallow geothermal energy use reveals great differences between European countries: regulations are based on national or regional water management and/or ground-water protection authorities, different ministries or technical guidelines with the main purpose of the protection of groundwater as drinking water resource [19]. Furthermore, these regulations mostly concentrate on the temperature of reinjected water from industrial cooling processes and/or open geothermal systems. Until now, little attention has been paid to other anthropogenic heat sources, which may have an even larger and more widespread impact on GWTs [20][21][22][23].
Shallow GWTs are subject to seasonal variations down to a depth of 10-15m [24]. Comparable to air temperatures, GWTs also depend on altitude and latitude [25]. For instance, mean GWT fluctuates between 2°C and 20°C between northern and southern Europe [26]. However, the natural state of GWT is altered by human activities. While groundwater is globally affected by increasing temperatures due to climate change [27][28][29][30][31][32][33], there are regional, anthropogenic impacts elevating GWT above its average and natural state. Changes in land use and advancing urbanisation in particular, directly influence groundwater recharge, level and temperature [34,35]. Increased surface temperatures due to artificial, sealed surfaces and underground structures raise the GWT beneath cities leading to so-called subsurface urban heat islands (SUHI) [36][37][38][39]. These SUHIs are often quantified by measuring the urban heat island intensity, which is defined as the difference between GWT in the urban area and in the rural background. In Germany, Menberg et al [23] determined average SUHI intensities of about 3-7K, but also detected local hot spots with GWT up to 20K warmer than the rural background temperature. Further GWT anomalies induced by underground car parks, construction sites, wastewater treatment plants, mine, landfills or power stations are also observed [25,36,40,41]. In their study on GWTs in Germany, Benz et al [42] introduced the anthropogenic heat intensity (AHI), which relates average rural background temperatures to local temperature measurements. They found GWTs to be much more impacted by human activity than by atmospheric and surface temperatures. However, they did not comprehensively discuss the encountered GWT anomalies. Hence, there is still a lack of understanding of these temperature extremes, and many questions remain unanswered in regard to the locations, frequencies, implications and associated point sources of such small scale and local temperature anomalies.
This study therefore aims to map, track and discuss the occurrence of temperature anomalies in shallow aquifers in central Europe. Based on (multi-) annual mean GWT data from ten European countries (table S1 is available online at stacks.iop.org/ERL/14/ 104012/mmedia), we determine the corresponding anthropogenic heat intensities (AHIs) to identify extreme, positive GWT anomalies. The AHI max , defined as the upper 3% percentile of all AHIs, are selected for each of the three major land cover classes (natural, agricultural and artificial) and linked to the detailed CORINE land cover types. We chose the upper 3% to assure AHIs, which are significantly above the measurement accuracy. Wells located under artificial surfaces, often in vulnerable aquifers, are examined in more detail in order to identify potential heat sources. Finally, we briefly discuss these GWT anomalies in the context of national regulations and assess the current and potential impact on our society.

Materials and methods
Groundwater temperatures Shallow GWT data from 44 205 wells in ten countries in central Europe are the basis for this study. GWT data originate from monitoring networks and are provided by local authorities, environmental agencies or hydrogeological services (table S1). While 11% of the wells are equipped with GWT data loggers, most wells were monitored manually as part of chemical analyses. The highest well densities can be found in France, south-west Germany and Belgium, whereas only few sampled wells are located in Denmark and Slovakia ( figure 1(a)). To standardise the data set and to eliminate seasonal GWT variations, data from all wells are averaged over the time span from 2003 to 2017 following the procedure given in Benz et al (2017a). In their approach, each temperature measurement is represented by a vector of a unit length of 1 and directed towards the month of measurement for a clocklike segmentation of the months. The output is the mean of all measurement vectors for one location, known as seasonal radius r, which is equal to zero for uniformly distributed measured data, and equal to one if they were collected in the same month. Following the recommendation by Benz et al (2017a), all wells with a depth 60m and r0.25, which indicates a bias-free annual mean, are considered for the further analysis (figure S1).

Anthropogenic heat intensity
For each well the AHI is defined as the difference between GWT at the well location and the median of surrounding rural background GWTs (GWT r ) [42] (equation (1)). Based on the definition by Benz et al [42], AHI is a measure of the anthropogenic influence on GWTs. Yet, in this study AHI also detects thermal disturbances caused by natural sources, as we apply it to wells in urban as well as rural areas The input parameters to determine the rural background temperature are the bias-free GWT, geographical elevation and night-time light intensity.
Elevation data are extracted from the Global 30 Arc-Second Elevation (GTOPO30) model and downloaded with Google Earth Engine [43]. Night-time lights from Version 4 of the DMSP-OLS Night-time Lights Time Series, processed by NOAA, were also extracted with Google Earth Engine. Since the night light data are only available up to January 2014, a 10 year average (01/2004 to 12/2013) was chosen. Night-time light intensity is expressed as a digital number (DN) running from 0 to 63 indicating an increasing urban activity [44]. All wells with a nighttime light of DN<15, an elevation ±90m and within a distance of 47km to the analysed location are considered for the calculation of rural background temperature [42]. To ensure meaningful statistics and to avoid an impact by outliers AHI is only determined, if at least five wells fulfil these criteria.

Land cover classification
The CORINE Land Cover (CLC) [45] classification scheme consists of three hierarchical levels with 44 land cover classes at the third and most detailed level (figure S2). Based on Level1, we define three main land cover classes: (1) natural, (2) agricultural and (3) artificial. The natural class is a combination of CLC's classes 'forest and semi natural areas' and 'wetlands'. The agricultural class contains CLC's 'agricultural areas' and the artificial class includes all 'artificial surfaces'. The calculated AHIs are categorised into and separately analysed for these three main classes ( figure S3).

GWT anomalies
The wells within the upper 3% percentile of each class are specified as temperature anomalies AHI max . All AHI max wells within the artificial land cover class are closely inspected via satellite images (Google Earth). Based on observed common characteristics, such as land use, economic activity and settlement structures, we defined specific land utilisation classes (LUCs) with detailed subclasses and identified possible heat sources of these hot spots.

Results and discussion
Statistics of GWT anomalies (AHI max ) Based on the bias-free annual mean GWT (12 151 wells) an AHI could be evaluated for 10 656 wells ( figure 1(b)). AHI is uniformly distributed over all known measurement depths, proving its independence of depth ( figure S4). Its distribution is given in figures S5-S7. Figure 1(c) displays the wells within the top three percentiles, which represent 318 GWT anomalies (AHI max ) in total. 97% of these hot spots are located in Austria, France and Germany, which have the highest AHI well density overall. In Belgium, hot spots exist only in agricultural areas. Slovakia, Switzerland and Luxembourg have only one hot spot in the class artificial and natural, respectively. Czech Republic, Denmark and Netherlands do not show any (figure S8). The hot springs in Austria and Southwest Germany, as well as accumulations of hot spots in the Upper Rhine Graben (URG) and Eastern Germany clearly stand out ( figure 1(c)). The URG is a densely urbanised region with multiple industrial areas, while East Germany is widely known for its former coal and ore mining. The minimum values of AHI max of the classes natural, agriculture and artificial are 2.3K, 1.7K and 3.9K respectively (figure S9).
To illustrate the link between land cover and temperature anomalies, the Level3 CLC classes for wells with an AHI are compared with the CLC classes of the AHI max wells (figure 2). A shift in the percentages of wells in each land cover class between these two sets is evident. Hence, it becomes apparent for which land cover temperature anomalies appear more frequently. For wells located on natural land cover, the percentage of wells in coniferous and mixed forests decreases from AHI to AHI max , whereas the percentage of wells associated with transitional woodland-shrub and natural grasslands triples. The latter are therefore more likely to contain GWT anomalies. One explanation is that soil temperatures and/or GWT beneath grass or farming land are typically higher than those beneath a forest, due to differences in incident solar radiation and evapotranspiration [46,47].
In contrast, the shift from non-irrigated arable land to pastures in the agricultural class cannot be exclusively explained by physical effects due to vegetation or shielding foliage. According to Herb et al [48], ground surface temperatures (GSTs) beneath grass and land with different plant canopies are similar. A possible explanation for the anomalies is deforestation, which is known to cause subsurface temperature anomalies [49][50][51][52], that are detectable at depths of 20-100m [53]. Regarding the temporal and horizontal extent of such temperature anomalies, a lateral spread of several hundred metres over 100years can occur [54]. Nevertheless, one has to notice that AHIs>3K under both natural and agricultural surfaces result from hot springs or local anthropogenic sources, such as contamination caused by landfills, mining or waste water treatment plants.
In the artificial class, the share of discontinuous urban fabric shifts towards industrial areas and continuous urban fabric. Multiple previous studies on SUHIs indicated local hot spots within dense urban areas and industrial sites, which is also evident in our current findings here. Epting et al [41], Menberg et al [23] and Ferguson and Woodbury [55] noticed a strong correlation between the highest underground temperatures and the density of buildings, in particular buildings with heated basements. For the city centre of Cologne and Winnipeg, Zhu et al [56] found an increase in GWT of up to 5K, which compares closely with the median of the AHI max of artificial surfaces in this study (figure S9). Epting et al [57] observed an increase of GWT up to 6 and 8K in dense industrial and commercial areas of Basel. Single point heat sources in industrial areas were also mentioned by Ferguson and Woodbury [58], Bucci et al [40] and Menberg et al [23].
GWT anomalies (AHI max ) beneath artificial surfaces The outcome of the detailed visual inspection and examination of the surroundings of the 45 artificial AHI max wells are six LUCs with 20 detailed subclasses (figures 3 and S10). With a mean AHI of 7K, the LUC 'factory' has by far the largest impact on GWT, whereas the mean AHI of 'industry parks' is the smallest with on average 5K. With regard to the share of each utilisation class, most of the hot spots are within 'city' (33%), followed by 'factory' and 'industry park' with 27% and 24%, respectively. In the following, possible heat sources within specific LUCs are discussed.
In the LUC 'industry park', different industrial branches such as plastic, paper, electronic, chemical or  machinery construction companies are mixed with office buildings and supermarkets. Here, high GWTs can originate from multiple heat sources such as basements with heating installations, sealed surfaces or injection of cooling water. These interfere with each other and can add up so that the distinct heat source of the groundwater anomaly is difficult to identify. Bucci et al [40] also referred to heat fluxes from buildings into the ground originating from industrial exothermic processes inside the buildings as cause for high GWT above 17°C in an industrial district close to Turin city.
In the LUC 'waste', one well is close to a landfill with an enclosed waste recycling plant, while the other one is on the premises of a waste disposal facility with detention basins and compensating reservoir. Benz et al [36] also identified a wastewater treatment plant in Osaka, Japan, as a local heat source for increased GWT.
Despite a high thematic accuracy of over 85%, wrong classifications of CLC classes can also occur [59]. Here, two wells in the artificial class are located on farmland and a fruit plantation, and thus actually fall into the agricultural class and the LUC 'farming'.
The LUC 'automotive' refers to wells located at a car workshop, a car race track and car dealer. The common characteristic of the automotive class are sealed surfaces and possible contamination with petroleum hydrocarbons [60].
The high mean AHI and standard deviation of the LUC 'city' stand out and reflect the significant, yet variable impact of the different subclasses and of the corresponding heat sources. High GWT in city centres are due to the interference and superposition of heat input by sealed surfaces and underground structures, as already described in several SUHI studies [23,36,38,40,41,56,[61][62][63][64]. A conspicuous cluster of wells showing increased GWT were observed close to underground car parks and therefore, classified as separate subclasses. The fringe subclass contains less dense urban areas. A hot spring in Austria, having the highest AHI (27.0 K) of all artificial wells, falls within this subclass and causes the overall high AHI and standard deviation of LUC 'city'. Disregarding this natural temperature anomaly leads to a mean 'city' AHI of 5.0±1.7K.
The LUC 'factory' comprises wells situated on the property of a detached, single factory that is not part of an industrial park. All seven wells in the subclass power plant are at the same location in France, whereas the remaining subclasses are only represented by one well location each. GWT anomalies with temperatures over 30°C in the vicinity of power plants were also reported by Menberg et al [23].
Heat sources of AHI max For 16 out of the 45 hot spots of the class artificial, we were able to identify potential heat sources summarised into seven classes (table 1). It is important to note that other underground heat sources such as industrial cooling, geothermal applications or sewage pipes are likely [22,39,40,61], but could not be detected with the here proposed method relying on satellite imagery and local knowledge. The highest temperature anomaly is associated with a hot spring in Austria. All remaining temperature anomalies and heat source classes refer to anthropogenic activities. Based on their spatial extent and impact magnitude they can be divided into two groups. The first group consists of heat sources that are scare, but have a large extent, such as contaminations and mining operations. Basements, district heating (DH) networks, swimming pools and underground car parks are the second group. They are rather local sources, but are more frequent in urban environments and therefore also have an extensive impact on GWTs.
The first group, containing the heat sources contamination and mining, exhibits the highest GWT and AHI of all identified anthropogenic heat sources with temperatures of up to 8K warmer than the rural surrounding. The three wells in the class contamination Table 1. Individual values, means and standard deviations (std) of the groundwater temperature (GWT) and anthropogenic heat intensity (AHI) for the 16 identified heat sources and seven heat source classes of the hot spots (AHI max ) within artificial areas. refer to two wells in LUC 'waste', and one well is at a car race track (LUC 'automotive'). Exothermic chemical and biological degradation processes in landfills or contaminated sites can result in higher GWTs [23,40]. Krümpelbeck [65] reported temperatures up to 60°C in a landfill. Similar to landfills, exothermic biogeochemical weathering processes, called acid mine drainage (AMD), cause high temperatures in mines and their remote surroundings [66]. Reports by Felix et al [67] and LfULG [68] confirm AMD as heat source of one particular well in the LUC subclass fringe, situated in a hard coal mining district in eastern Germany. Furthermore, they described increased GWT in remote observation wells due to coal seam fires reaching temperatures up to 90°C within the pithead stocks. The high GWT and AHI of the well in the subclass 'farmland', located in an area in eastern Germany famous for ore mining, could also be associated with AMD. The second group includes the small scale and local heat sources basements, DH networks, swimming pool and underground car parks. The well linked to warming from basements, is 2m away from a shopping mall in Karlsruhe, Germany. While the AHI of this well is lower than the ones associated with contamination and mining, almost every building in a city has a basement, which typically also hosts the heating installation of the building. Epting and Huggenberger [21], Benz et al [61] and Epting et al [69] also emphasised the large impact of basements on GWT and due to their high heat flux and dominant area, named them as the dominant drivers of SUHIs.
Correlating local DH network plans with well positions, we could classify the heat source of three wells of the subclass city centre as DH. In DH networks, water with temperatures up to 160°C circulates under high pressure through pipes under many urban areas [70]. Depending on season and type of insulation, heat losses up to 20% occur [71]. Benz et al [61] pointed out that DH pipes are a prominent source of anthropogenic heat fluxes. The time series in figure S11 also clearly demonstrate the impact of DH heat fluxes on a groundwater observation well 3.5m away from the pipe. Regarding the mean GWT at 6m depth, representing the middle of the aquifer, AHI is as high as 8K. Consequently, the heat input by DH pipes, especially in case of a local leakage is not negligible and should be considered more carefully.
Water with lower temperatures than in DH pipes is also released into aquifers by leaking swimming pools. Cracks in the pool or loose tiles can cause leakage rates of 70 m 3 d −1 [72]. Another case study about a municipal swimming pool in Montreal reports a leakage rate of 350-700m 3 per day into the underlying aquifer [73]. Even if the swimming pool is watertight, the basin releases heat to the subsurface. One of the wells in LUC 'city' is located 4m away from a municipal swimming pool in Germany and the GWT of 16°C is likely to be influenced by the heat release of the pool. Menberg et al [23] even noticed a GWT of 20°C for an observation well next to a swimming pool in Frankfurt, Germany. At another municipal swimming pool in Germany, temperatures of 25°C beneath the swimming pool and increased GWT of 1-3K in the downgradient were measured [74].
In previous SUHI studies, underground car parks were intensively discussed as sources for GWT anomalies [21,22,56]. This is in accordance with our findings that reveal underground car parks as the most frequent heat source of temperature anomalies in the class artificial (table 1). Warm, exhausted fumes and a poor ventilation lead to heat accumulation, so air temperature strongly increases in underground car parks. Iskander et al [75] recorded temperatures above 25°C in summer at the lowest level of an underground car park. We also recorded air temperatures of up to 30°C in an underground car park and correspondingly high GWT of almost 20°C in an observation well within this car park (figure S12). The correlation between these two temperatures is obvious and therefore the heat input of underground car parks into the aquifer is evident.

Regulations
Despite the multitude of underground heat sources, only open geothermal systems are currently regulated by legally binding temperature thresholds in Austria (20°C), Denmark (25°C) and the Netherlands (25°C) [76,77]. Four wells out of all 318 hotspots exceed the 25°C threshold value, though they are natural hot springs in Germany and Austria. A maximum temperature (T max ) of modified groundwater of 20°C and a relative change (ΔT) in GWT of ±6 K is given in the geothermal installation guidelines in Austria (legally binding) and Germany (recommended) [76,77]. For all hot spots, we detected 13 wells that exceed T max and 38 with an AHI exceeding ΔT of 6K. While four of these temperature anomalies are associated with natural hot springs, the remaining nine temperature infringements, or rather 34 for AHI exceeding ΔT, are associated with anthropogenic heat sources. The majority of wells with a higher AHI than the 6K temperature difference (ΔT) are in the artificial land cover class and located in Austria, France, Germany and Switzerland. When comparing our results with the accepted ΔT and T max , we found that the mean AHI of the LUCs 'automotive', 'city' and 'factory' are slightly above the ΔT limit, while the mean AHI linked to the heat source classes 'contamination' and 'mining' are 1K or even more than 2K above the criteria respectively. Since GWT is averaged, the information of seasonal positive or negative extreme values of the time series is not accounted for in this analysis. Individual GWT measurements might exceed the maximum GWT T max more frequently. From the GWT time series in figures S11-S13, it becomes apparent that GWT peaks caused by basements, contamination, mining and DH surpass the T max -limit several times while annual mean values remain below the threshold. In case of aquifer thermal energy storage systems, seasonal variation of GWTs also cannot be detected by AHI since the mean GWT is equal or close to the GWT r . Accordingly, the number of wells momentarily exceeding 20°C is expected to be significantly higher than those found based on annual mean GWTs.

Conclusion
This study detects GWT anomalies in central Europe and identifies large-and small-scale anthropogenic heat sources such as mining and underground car parks. These extreme and until now unregulated heat sources seriously impact our groundwater. When GWTs continue to increase, groundwater cooling systems are no longer efficient [55,69]. Furthermore, high GWT might also affect groundwater quality and ecology (e.g. [5,6,[78][79][80]). In some urban areas, where aquifers are already contaminated with heavy metals and organic compounds, an increase of GWT by only 5K might also entail a decrease of dissolved oxygen and may lead to a mobilisation of other contaminants such as arsenic [81][82][83][84]. Nevertheless, elevated GWTs provide the opportunity to harness more energy from the aquifer using shallow geothermal systems or make the operation of such systems more efficient [56,[85][86][87][88]. Overall, increased GWTs have multiple, long-term consequences and therefore, the complex interaction between heat sources and heat sinks in consideration of the aquifer characteristics should be further studied and also regulated. All these influencing factors have to be incorporated into future urban subsurface planning. Regulations should be more flexible, so that depending on the specific aims of the policy of cities and communities, the focus of groundwater management can be on groundwater as a resource for drinking water and/or as an energy resource. The use of numerical heat transport models could maximise the positive effects of increased GWT in order to meet the needs of various interest groups and to preserve the natural state of our groundwater ecosystems.

Acknowledgments
CT is grateful to the funding received through GRACE, the Graduate School for PhD students of the KIT-Center Climate and Environment at the Karlsruhe Institute of Technology (KIT). The work was also supported by the German Research Foundation (project no. B2850/3-1). The authors thank Alistair Fronhoffs (Vlaamse Milieumaatschappij), Arlette Liétar Stapleton (Bundesamt für Umwelt) for the provision of data and additional information. Special recognition is also given to Nicolas Weidenthaler for the field work and collection of GWT data in Mannheim, Germany. We also thank the two reviewers for their comments. We acknowledge support by the KIT-Publication Fund of the Karlsruhe Institute of Technology.

Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available for legal and/or ethical reasons.