Trends in groundwater storage
Figure 5 shows the computed trends in GWS for the Euro-Mediterranean region from 2003–2020. The trend analysis of GWS provides an indication of an increase or decrease of GWS in a certain area over the period under consideration. The trend in GWS is given here as an annual mean over the entire time span from 2003–2020 and can therefore vary over shorter periods. A negative trend over the specified time period is equivalent to a decrease in groundwater resources, while positive trends indicate an increase. It should be noted that GRACE-derived data always refer to the entire vertical groundwater column and thus represent the sum of GWS changes in multilayered aquifer systems, where present, and observed groundwater level records of individual aquifers may therefore differ. Negative trends in GWS may either be the result of a decrease in groundwater recharge or indicate that groundwater abstraction exceeds groundwater recharge in the long-term. Positive trends in turn indicate either an increase in groundwater recharge and/or reduction in groundwater use and hence a recovery of groundwater resources.
A very significant trend in GWS (p ≤ 0.01) can be observed in about 82% of the study area, while 5% is still significant (p ≤ 0.05). No significant trends (p > 0.05) are found in 13% of the study area, e.g. in larger parts of northern Germany and south-western France, or scattered in smaller areas of North Africa, Spain, Italy or Turkey (Fig. 5). About 80% of the area with significant trends reveal negative values and about 20% positive values. The actual values of the not-significant trends (most of which are very small with average absolute values at around –0.6 mm/year for negative trends and 0.5 mm/year for positive trends) were still included in the computation of spatially averaged trends (for countries, climate zones, etc.).
Trends in GWS in different climate zones
Based on the distribution of trends in GWS over the study area, it can be seen that positive and negative trends occur in all major climate zones (Figs. 2a and 5; Table 1). However, the average of each major climate zone is negative and ranges from –1.5 to –4.4 mm/year. The strongest negative trends of subclimate zones are found in the arid hot steppe of the MENA region, with an average value of –7.2 mm/year and in the cold region with hot and dry summers in Eastern Europe with an average trend of –5.7 mm/year. However, these zones represent less than 3% of the total investigated area. Similarly, the few polar regions with high altitudes and tundra-dominated landscapes are also subject to strong negative trends with values ranging from –2.8 to –5.9 mm/year, though representing less than 1% of the area. This can be observed, for example, in the Alps, the Balkans, and the Carpathians. The cold regions with warm summers and without a dry season, which cover about 19% of the study area, show a mean annual trend of –2.9 mm/year. The hot arid desert-like regions, which account for almost 51% of the study area and are present mainly in the MENA region, show an average negative trend of –1.7 mm/year, and most temperate regions of central and western Europe (about 19% of the study area) show average negative trends between –0.7 and –2.5 mm/year. It appears that negative trends on average extend across all climate zones, with a slight tendency towards more negative trends in polar (–4.4 mm/year), arid (–3.8 mm/year) and cold regions (–3.4 mm/year) compared to temperate regions (–1.5 mm/year), though also great differences in distribution within the regions can be observed.
Trends in GWS for individual countries
The average annual trends in GWS in mm/year for the major regions are given in Table 2. Those for each country, including the lower and upper confidence intervals, are given in Table 3 and are also presented in Fig. 6. The results of the GWS trend analysis show that an average negative trend is observed in 36 of the 47 countries, while 11 countries show a positive average trend (Fig. 6; Table 3). In 33 countries, also the upper confidence level is in the negative range, which means that there is a 95% certainty of a decrease in GWS. Only for five countries (Yemen, Portugal, Estonia, the United Arab Emirates and Denmark) are also the lower confidence levels positive (Table 3), indicating a 95% certainty of positive trend in GWS. The overall mean of the trends across the Euro-Mediterranean region is –2.1 mm/year, which corresponds to a loss of water volume of about 29,500 million m3/year, with the Arabian Peninsula showing the largest annual loss of about 12,000 million m3, Northern Europe experiencing a small annual gain in groundwater of about 440 million m3 (Table 3).
The strongest negative average trends can be observed for Iraq (–8.8 mm/year) and Syria (–6.0 mm/year), but also other countries of the Arabian Peninsula show clear negative trends, as for example Kuwait (–5.7 mm/year), Jordan (–4.3 mm/year) and Saudi-Arabia (–3.9 mm/year). However, countries in Western, Central and Eastern Europe are also affected from negative average trends in GWS, with Switzerland being the most affected (–5.7 mm/year), followed by Hungary (–5.7 mm/year) and Austria (–5.1 mm/year). The countries of the Balkan region also show average negative trends, ranging from very strong trends in Slovenia (–4.7 mm/year) to less strong trends in North Macedonia (–1.9 mm/year). In Southern Europe, the situation is ambivalent, with strong negative trends in Italy (–2,0 mm/year) and weak negative trends in Spain (–0.3 mm/year) and Greece (–0.1 mm/year), while Portugal reveals a positive trend (2.1 mm/year). Average positive trends of GWS per country are mostly found in Northern Europe such as in Latvia, Lithuania, Estonia and Denmark with values between 0.6 and 6.0 mm/year. North African countries reveal weak to strong negative trends for Libya, Algeria, Tunisia, and Morocco (–0.9 to –2.9 mm/year), while Egypt shows a slightly positive trend (0.2 mm/year; Table 3).
It should be noted, however, that the mean value is often not representative for the entire area of a country, as the trend in many countries varies greatly from region to region. This is particularly noticeable in the countries of Northern Africa, for example, as well as in Spain and Turkey.
Comparison of trend in GWS with recharge and abstractions on country level
The sustainable use of groundwater resources is usually derived from the comparison of the amount of natural groundwater recharge and groundwater abstractions (Fig. 7; Table 3). This can give hints for possible causes of negative trends, which can be attributed to regions where abstractions exceed natural recharge. For a comparison, the GRACE derived trends in GWS were converted to an annual increase or decrease in GWS volume per country (calculated from the mean GWS trend in mm/year by cell-wise multiplication with the cell-area and summed up over each country).
At the country level, the recharge values from the FAO, the WaterGAP model, and computed based on ERA5-Land data, reveal partly large differences (Fig. 7), which is either caused by the different calculation methods and/or by different periods under consideration. The FAO groundwater recharge data are long-term annual averages of the reference period of 1961–1990 (FAO 2021), generated either by estimating annual infiltration rate of precipitation (in arid countries) or by computing river base flow (in humid countries, FAO 2021). The WaterGap data include diffuse groundwater recharge and groundwater recharge from surface-water bodies, which are in turn the results of an elaborate modelling procedure (Müller Schmied et al. 2021), but refer to the period of 2003–2016 only. The computations based on ERA5-Land data are the result of a simpler water balance approach, but therefore refer to the exact study period of 2003–2020.
The groundwater abstraction data of the FAO comprise the annual gross amount of water extracted from aquifers, including withdrawal of renewable groundwater, as well as water from overabstraction of renewable groundwater or withdrawal from fossil groundwater (FAO 2021). Values are taken from 2017 (newest available values) to make them comparable across countries, since for many countries, earlier values are patchy, which leads to errors especially in countries where abstraction rates have strongly increased or decreased over the study period. The WaterGap net abstraction from groundwater data refers to 2003–2016, but, contrary to the FAO data, includes return flows caused by human water use (mostly due to irrigation). Moreover, the GRACE, WaterGAP and ERA5-data all come at different spatial solutions, which can lead to inaccuracies especially when calculating volumes for smaller countries. Overall, for the mentioned differences, a direct comparison of GWS trends with average values of groundwater abstraction and groundwater recharge data on country level is only meaningful to a certain degree, i.e. in a semiquantitative way (Fig. 7). Table 3 shows the differences of recharge and abstraction for the FAO data, the WaterGap data, and the recharge from WaterGap and ERA5-Land data minus abstraction from FAO. These data already exhibit a strong variability.
The differences of annual recharge and net abstraction from the WaterGap data are positive for all countries except Yemen, which seems somehow unrealistic compared to other studies. It can only be assumed that the net abstraction values are probably underestimated, which might be a result of the overestimation of return flows by human water use or overestimation of the annual recharge rates. Strongly declining groundwater levels, e.g. on the Arabian Peninsula, could otherwise not be explained.
The differences of groundwater recharge and abstractions from FAO data show a more diverse picture. Unfortunately, Iraq and Syria, the two countries with the strongest negative trends in GRACE GWS, lack FAO abstraction data. The data exhibit nonsustainable abstractions for most countries on the Arabian Peninsula and also for Northern Africa (Algeria, Libya, Tunisia, Egypt). In Europe, negative net balances from FAO data are only given for Portugal. The results are widely consistent with losses in groundwater volume from GRACE-derived GWS data for the Arabian Peninsula and Northern Africa, apart from Qatar and United Arab Emirates, which exhibit even a gain in GWS. On the other hand, countries with high losses in GWS from GRACE data in Europe, like France, Poland, Germany and Hungary, do not exhibit unsustainable use of groundwater from FAO data. Since the FAO abstraction data seem rather up to date and reliable (apart from possibly unrecorded illegal withdrawals), the recharge data refer to the period of 1961–1990. A possible explanation is that the observed decline in groundwater resources in these regions is the result of a climatic decline in recharge in the period of 2003–2020 (considered for the GWS trend) compared to the earlier reference period, as stated, e.g. by Fliß et al. (2021) for southern Germany.
In order to investigate this possible influence, WaterGap recharge values as well as recharge values for 2003–2020, calculated based on a simple water balance approach from ERA5-Land data, were used instead of FAO recharge for computation of sustainable use along with the FAO abstractions. The results from the WaterGap recharge and FAO abstraction data are still somehow unrealistic for most countries, with loss values only for five countries on the Arabian Peninsula, leading to the assumption that also the WaterGap recharge values could be overestimated. Though there are still some contradictions, the computed results based on ERA5-Land recharge and FAO abstractions show the most correspondences with changes in groundwater volumes from GRACE GWS data. This is the case for most countries of the Arabian Peninsula and Northern Africa, which show partly high losses in GWS volumes and at the same time a nonsustainable groundwater use. Most countries with gains in GWS (except United Arab Emirates, Egypt, Lebanon, and Yemen) now show a recharge surplus over abstraction. However, still there are many countries, especially in Europe (e.g. Romania, France or Germany) that show partly high losses in groundwater volume from GRACE-derived GWS, but have a high recharge surplus over abstraction at the same time. Again, a possible explanation could be the different time periods considered: while GRACE GWS and ERA5-Land recharge data include low recharge values in the extremely dry summers in central Europe in 2018–2020, the FAO abstraction data from 2017 do not include presumably increased abstractions in the same period, leading to a distorted balance. Though it must be admitted that the differences are partly that large (e.g. for Germany and France), this probably can not be the only reason.
As mentioned, the comparisons are only possible in a semiquantitative way and a direct comparison of GWS gain or loss volumes with volume differences in recharge and abstractions are not meaningful due to different calculation methods and reference periods. On the other hand, the comparisons here show that the traditional approach of using recharge and abstraction data, which are subject to great uncertainties, are often not suitable for the determination of a sustainable use of groundwater resources.
Comparison of trends in GWS with trends in natural groundwater recharge
To analyse possible climatic influences on the GWS, a trend analysis (seasonal Mann–Kendall Test and Sen’s Slope) for the monthly recharge values computed from ERA5-Land data from 2003–2020 was carried out. Apart from smaller regions on the Arabian Peninsula and Northern Africa, the trends were not significant over most of the study region, especially in Europe (results not shown). However, several studies found that groundwater recharge decreased in the past two decades, compared to a three-decade reference period from 1971 to 2000 such as the study by Fliß et al. (2021) for southern Germany, which could play a role in some regions, even if the (monotonic) trends within the study period are mainly not significant. In any case, it can be assumed that groundwater withdrawals (which can be partly also influenced by climate variability especially regarding abstractions for irrigation) play a dominant role, especially when downward trends in GWS are large.
Regional GWS trends, related land use and population density
The regional distribution of the GWS trend within and across countries can look different depending on many factors, including population density and land use—for example, the European region with some densely populated urban centres is much more homogeneously used for agriculture and settlement, while the population density in the MENA region is mainly limited to coastal regions or isolated urban or agricultural areas inland (Fig. 2c). It can be seen throughout the study region that areas with high groundwater abstractions are often more affected by negative trends than other areas (Fig. 2d). Moreover, these regions widely coincide spatially with those in which there are either agricultural areas irrigated with groundwater (e.g. in central Saudi-Arabia, parts of Italy and France) or higher population densities (e.g. eastern Iraq, the Po valley in Italy and the Île-de-France region around Paris; Fig. 2c). This suggests that the negative trends in GWS can at least partly be attributed to nonsustainable groundwater abstractions, though there are also exceptions, for example in parts of south-western Spain, where not significant or even positive trends prevail in spite of high groundwater abstractions. This may result either from the fact that the spatially distributed data of net groundwater abstraction include irrigation return flow, which might mitigate negative trends in GWS due to abstraction for agriculture, or, in some arid zones, abstractions may not always be directly spatially connected to land use or population density, as groundwater is sometimes abstracted from more remote areas (e.g. in southern Libya).
Trends in GWS for zones of different aquifer productivities
The extent to which groundwater can be explored generally depends on the productivity of the aquifers and their accessibility. Highly productive, shallow aquifers are generally easier to explore than deep and less productive ones; however, their surface exposure also plays an important role in groundwater recharge, as they generally respond more strongly to precipitation events than deep and covered aquifers. Much less productive and deep aquifers are tapped more frequently when water demand is high and water from other sources is not available. From the results of the trend analysis in GWS, there is no clear indication that certain types of aquifers are particularly affected by positive or negative trends when considering the entire region; however, a clearer picture emerges when specific regions or single geological provinces are considered. The complete analysis of all geological provinces can be found in Table S2 of the ESM.
A negative trend in GWS can be observed for all types of aquifers in Europe and Turkey, with average trends ranging between –1.9 and –3.5 mm/year. However, the highly productive intergranular, fractured and karst aquifers show somewhat stronger negative trends, on average with –3.5 mm/year, than the moderately productive aquifers (with –2.5 mm/year) and the high to low productivity unconsolidated porous aquifers (with –2.0 and –2.5 mm/year, respectively). Low-productivity intergranular and fractured aquifers as well as practically nonaquifer rocks show also slightly less negative trends of –2.2 and –1.9 mm/year on average, respectively. Since they are usually hardly used for water supply, this could be an indication of climatic causes.
The aquifers in North Africa show a slightly lower negative trend than in Europe with average values between –0.3 and –2.0 mm/year, whereas the low-productivity unconsolidated porous aquifers even show positive trends of 0.6 mm/year. However, the latter are only found in about 2% of the region. The strongest negative trend is shown by the low-productivity unconsolidated porous aquifers, with –2.0 mm/year.
In the Arabian Peninsula, generally many more negative trends can be observed, as already mentioned previously. The highly productive unconsolidated porous aquifers stand out in particular, with an average trend of –10.1 mm/year, which is certainly due to their generally easier accessibility. The moderately productive intergranular, fractured and karst aquifers are also heavily stressed, with a mean value of –6.4 mm/year. The other aquifer types, including the low-productivity ones, also show negative trends between –2.5 and –3.3 mm/year (Table 4).
Selected regional examples of GWS trends
In the following, selected regions or countries in western and southern Europe, the Arabian Peninsula and Northern Africa (Fig. 8) are discussed in more detail. With an annual groundwater consumption of about 4,800 million m3 and an annual groundwater recharge of between 4,000 and 37,000 million m3, Portugal shows sustainable use of groundwater, which is also reflected in a slight gain in GWS of 189 million m3/year. Spain also seems to run a sustainable groundwater management strategy with an annual groundwater consumption of about 6,400 million m3 and an estimated amount of renewable groundwater of about 34,000 million m3. However, it recorded a slight mean annual loss in GWS of about 143 million m3/year. Positive trends in GWS are especially observed in the southern Iberian Massif, which consists of intergranular and fractured sediments and magmatic and metamorphic rocks of low productivity. A moderately negative trend in GWS prevails in the Iberian Cordillera, where a complex sequence of intergranular, fractured, and karstified formations of moderate to high productivity is exposed, alongside unconsolidated sedimentary formations of low to high productivity (Fig. 8a,b).
France experiences a mean annual groundwater loss of about 1,270 million m3, corresponding to 23% of the total groundwater abstraction of about 5,500 million m3. With an estimated amount of renewable groundwater recharge between around 83,000 and 120,000 million m3/year, the GRACE-derived country-wide loss in GWS cannot be explained from averaged nonsustainable extractions. On the other hand, it is limited to some regions in central and eastern France; thus, the Anglo-Paris basin reveals a mean annual loss of 576 million m3, the Trans Graben basin of 127 million m3, the Massif Central of 159 million m3, and the Lion-Camargue basin of about 208 million m3 (Fig. 8a and b), together accounting to already 72% of total groundwater loss. These areas are already known for declining groundwater levels, as described by Maréchal and Rouillard (2020).
For Italy, a mean annual groundwater loss of about 561 million m3/year was calculated, while the mean annual renewable groundwater volume is reported to be between about 40,000 and 85,000 million m3. Data on groundwater use are not available for Italy. A negative trend is observed in many coastal areas, but is mainly attributed to the northern and central parts of Italy, especially the central mountainous regions of the Apennines and the Po Valley (Fig. 8b). The latter consists of a highly productive unconsolidated porous aquifer and reveals a mean annual loss of 249 million m3, while the Apennines consists of nonwater-bearing rocks and highly productive fractured and karstified aquifers (Fig. 8a). A decrease in precipitation in Italy due to climate change was already observed before the 2000s (Polemio and Casarano 2008; Ducci and Tranfaglia 2008) and also groundwater overuse (Lancia et al. 2020), which both can be the causes for negative trends in GWS. The results correspond to some regional observations of groundwater depletion, e.g. in coastal aquifers (Sappa and Vitale 2001) and large parts of the mountainous regions in central Italy, as evidenced by declining groundwater levels and spring discharges (Fiorillo et al. 2015; Lancia et al. 2020). However, other studies have not identified declining spring discharges in the northern Apennines or the piedmont (Cervi et al. 2018; Bastiancich et al.2021).
Most north African countries show slightly negative trends in GWS, with values ranging from –0.9 to –2.9 mm/year, and only Egypt shows a slightly positive trend of 0.2 mm/year, mostly visible along the Nile and the northern part of the country (Fig. 8d). The positive trends are also visible in northern Libya, with the exception of a narrow coastal strip where negative trends occur. Other areas with positive trends are found in northern Tunisia and Algeria as well as along the north-western Atlantic coast of Morocco. Large areas with negative trends are present in the inland areas of Egypt, Libya and in large parts of the Atlas Mountains, which stretches from Morocco over Algeria to Tunisia. The different aquifer types in the north African countries are subject to similar variations and trends, with the low productivity intergranular and fractured aquifers even showing slightly positive trends with 0.6 mm/year, but accounting for only about 2% of the area. The moderate and highly productive intergranular, fractured and karstified aquifers reveal weak to moderate negative trends of –0.3 and –1.1 mm/year, respectively, which outcrop in about 62% of the region (Fig. 8c). The trends in north-western Algeria correspond with the depletion of groundwater resources observed by Berhail (2019) and the gradual decline of groundwater levels in the coastal zone in south-western Morocco identified by Malki et al. (2017) and Ouhamdouch et al. (2018).
The region most affected by negative trends in GWS is the Arabian Peninsula. It is characterized by the south-eastern Arabian shield basement complex of low-permeability igneous and metamorphic rocks and the north-eastern stable and mobile shelf of mainly karstified carbonate rocks and consolidated sediments. The latter merges in the north into the collision zone of the Arabian and Eurasian Rift, the Zagros mountain belt (Fig. 8e). In all Arab countries, groundwater plays a crucial role in water supply, but due to natural water scarcity, many aquifers are not sustainably managed, resulting in declining groundwater levels. All countries show a strong negative trend in GWS, with Iraq at –8.8 mm/year and Syria at –6.0 mm/year being the most affected (Fig. 8f). Groundwater abstraction data are not available for these countries, but the region is known for severe groundwater abstraction, as for example the city of Erbil in northern Iraq which is located in the centre of a large depression cone affecting the Khleisha uplift and the Zagros fold belt (Stevanovic and Iurkiewicz 2009; Nanekely et al. 2017; Awadh et al. 2020).
In Jordan, the recorded annual groundwater abstraction is about 615 million m3, thus already officially exceeding the officially recognized average annual recharge rate of FAO (2021) of 540 million m3. The rather strong mean annual decline in GWS of –4.3 mm/year corresponds to an average annual storage loss of 388 million m3, which exceeds the official annual groundwater deficit of about 235 million m3 in 2017 (MWI 2017). This can probably at least be partly attributed to illegal groundwater abstractions that are not officially counted. Moreover, the FAO estimations of the long-term recharge or renewable groundwater resources could also be overestimated, as indicated by computations of recharge from 2003–2020 based on ERA5-Land data. Also, the overuse of transboundary aquifers, e.g. by Saudi Arabia, certainly contributes to the computed losses. Saudi Arabia officially withdraws over 21,000 million m3 groundwater per year, exceeding by far even the most optimistic estimations of natural recharge. This causes an average annual storage loss within the countries’ borders of about 7,500 million m3, which is reflected in a strong cone of depression, affecting the Hail Ga’Ara arc and the Jafr Tabuk basin (Fig. 8f). Other countries such as Kuwait and Israel also show negative trends on average with –5.7 and –3.2 mm/year, respectively. Qatar, Libanon, Oman, Yemen and the United Arab Emirates, however, show even positive mean trends between 0.2 and 4.2 mm/year, which is limited to the central, sparsely populated desert and mountainous areas, with the more densely populated coastal areas and the western and eastern parts showing slightly negative trends. For the eastern part of the United Arab Emirates, this coincides with the regional observation of Yilmaz et al. (2020). It has to be mentioned though, that the results of the GRACE–derived data for the smaller countries of the Arabian Peninsula has to be interpreted with some care.
Regarding aquifer types in the Arabian Peninsula, most are strongly affected by the negative trends in GWS, which could be due to the natural water scarcity in this region and the additional use of low-productivity groundwater systems that contribute less to water supply in water-rich countries. The most affected aquifers are the highly productive unconsolidated porous aquifers with an average trend of –10.1 mm/year, which are generally shallow and easy to develop, followed by moderately productive intergranular and fractured aquifers (including karstified rocks) and practically nonaquifer rocks (porous or fissured) with –6.4 and –3.3 mm/year, respectively. Both aquifer types occur in about 26% of the area (Fig. 8e).