Global analysis of groundwater pumping from increased river capture

In many regions globally, groundwater overuse exceeds natural replenishment, leading to immediate consequences such as reduced river flows and devastating impacts on freshwater ecosystems. In alluvial aquifers in particular, groundwater pumping contributes to river flow reduction in two significant ways: first, by intercepting water that would naturally discharge into the river, and second, by lowering groundwater levels below the riverbed, causing river water to infiltrate. Despite these critical interactions, large-scale water resources assessments often overlook the relationship between groundwater and surface water, hindering a comprehensive understanding of the consequences of groundwater pumping on both the groundwater and surface water systems. Our study, utilizing a coupled global-scale groundwater–surface water model, reveals that approximately 20% of globally pumped groundwater stems from diminished streamflow, while 16% results from reduced storage. Projections for the end of the century, accounting for climate change, suggest potential increases to 30% from reduced streamflow and a decrease to 12% from reduced storage. Notably, our results highlight that the impact on streamflow is more widespread and linked to smaller pumping rates, contrasting with impacts on storage associated with higher pumping rates. This study shows the crucial need to include groundwater–surface water interactions in large-scale water resources assessments, not only for accurate estimates of freshwater availability but also for a comprehensive understanding of the far-reaching impacts of groundwater overuse related to increasing water demands and climate change.


Introduction
Over the past 50 years, global water demands have more than tripled, driven by growing populations, economic development, expansion of irrigated cropland, and climate change [e.g . 1].Only less than 1% of all water on our planet is accessible for human water use and supporting freshwater ecosystems.Most of this accessible water is stored underground in aquifers.When water from rivers, streams, and lakes is insufficient to satisfy water demands, groundwater is often used as an additional water source.Today, more than two billion people (representing 30% of the world population) [2], and nearly half of our freshwater ecosystems [3], depend, at least partly, on groundwater.Irrigation is by far the largest sector of freshwater use worldwide, and accounts for 17% of global cropland area and 40% of global food production [4].Approximately 40% of the current irrigated agricultural area globally depends on groundwater [5,6], highlighting the central role of groundwater in maintaining and enhancing global food security.Especially irrigated regions in drier climates depend heavily on groundwater [e.g . 7].
In many regions around the world, pumping rates exceed rates of recharge from percolating rain, melting snow, and river infiltration.Consequently, when pumping rates exceed recharge rates for long periods, groundwater levels will decline.Declining groundwater levels can have various negative impacts, such as drying of wells and increasing pumping costs, groundwater depletion, land subsidence, saltwater contamination of aquifers, a reduction of groundwater-driven evapotranspiration fluxes, and a reduction of river flows and water levels in lakes and wetlands [8][9][10][11][12].Most of these impacts are related to ecological losses and damages [e.g.8,13,14].To better understand the various and widespread impacts of groundwater pumping worldwide and to be able to adapt to, mitigate, or avoid impacts, a detailed global understanding of groundwater systems and dynamics is essential.
Regional-scale studies have long emphasized the importance of including groundwater-surface water dynamics and interactions for accurate model-based estimates of groundwater availability in aquifers or watersheds, and for detailed assessment of groundwater use impacts on groundwater and surface water resources [14,15].At the global scale, however, groundwater dynamics and head-dependent fluxes are still often ignored [14][15][16].The largest challenges in including groundwater dynamics at the global scale are computation efforts and data scarcity limiting regional accurate parameterization and the possibility for detailed data and model-driven evaluations [15].By including groundwater dynamics and groundwater-surface water interactions on a global scale, it is possible to assess how a drop in groundwater levels caused by groundwater pumping affects groundwater and surface water availability and associated impacts [8,17].Two recent studies estimated that in almost half of the watersheds where groundwater is pumped, it is notably pumped at the expense of river flow [18], already putting ecosystems at risk for one-third of these watersheds [8].Currently, however, there is not a comprehensive worldwide assessment that quantifies the extent to which pumped groundwater originates from increased river capture-meaning the groundwater that would otherwise naturally flow into streams or infiltrate into the groundwater system (as illustrated in figure 1).Such an assessment would enhance our understanding of how groundwater pumping affects surface water resources and provide insight into the extent to which the adverse effects of groundwater pumping are buffered by groundwater-surface water interactions.
At the scale of the USA, an estimate of increased capture does exist.It was estimated that about 85% of groundwater being pumped was derived from the capture of river water or evaporation, with the remaining 15% derived from groundwater storage [19].These continental-scale estimates underscore the significance of accounting for increased river capture in global-scale water resources assessments.Moreover, they emphasize that evaluations of groundwater pumping should not only include its impacts on groundwater storage (which is currently often the focus of large-scale groundwater use assessments), but also should consider its impacts on streamflow, the connected ecosystems, and evaporation.
Furthermore, the spread between current global estimates of groundwater depletion is large [14], varying from 5000 km 3 to 25 000 km 3 for 1960-2010.The lower estimate does account for groundwater lateral flows and groundwater-surface water interactions, while the higher estimate excludes these flows and interactions.The first estimate aligns with estimates derived from calibrated regional studies [8,19] and, as such, indicated that groundwater lateral flows and interactions are relevant to include in large-scale water resources assessments [20].
In this study, we provide the first global-scale quantification of increased river capture and assess its proportional contribution to the total groundwater pumping within sub-basins.The quantification of increased river capture will deepen our understanding of how groundwater pumping affects both groundwater and surface water resources.It will also provide new insights into the degree to which groundwater-surface water interactions mitigate the adverse consequences of groundwater pumping, particularly buffering decreasing groundwater levels.This is useful new knowledge that will, for example, help to better understand where and when groundwater-dependent ecosystems are at risk, and to move towards more sustainable groundwater management worldwide.
In our analysis, we focus on two timeframes: the recent past  and the end of the century (2050-2100).We use modeled outputs of groundwater dynamics, flows, and interactions obtained from a physically based global-scale groundwater-surface water model (GSGM) [8].We used historical data on climate and water demand as well as projections for future climate and water demand.First, we assessed the main drivers of increased river capture.Second, we evaluated the direct and indirect consequences of increased capture on groundwater's contribution to river flow, with a particular focus on low-flow conditions.In doing so, this research offers a perspective on the interdependence of groundwater use and groundwater-surface water interactions.Additionally, we provide a first insight into the direct and indirect impacts of global groundwater pumping.The analysis highlights regions where groundwater pumping has already led to the depletion of groundwater storages and a reduction in streamflow, and it identifies regions that could face substantial environmental consequences unless current groundwater pumping practices transition towards more sustainable approaches.

Data input
In this study, model results of a previously performed and published study, using a physically based globalscale GSGM [8] were used.Here, a general overview of the model is provided, and important aspects of the model for this study's application and analysis are highlighted.A more detailed model description is provided as supplementary material.
The model consists of the global hydrology and water-resources model PCR-GLOBWB2.0[21] and is dynamically coupled, via groundwater recharge/capillary rise and groundwater discharge/river infiltration, to a two-layered groundwater flow model based on MODFLOW [20,22].In the groundwater model, groundwater heads, dynamics, and interactions are simulated.The coupled model was run transiently at a daily-to-monthly timestep and high spatial resolution (5-arcminutes globally).Detailed descriptions of the model setup, model structure, coupled modules, evaluation of model results, and discussion on model sensitivity to parameter settings and boundary conditions are given in previous publications.These publications cover the model and its applications [21][22][23][24][25][26][27].
In this study, we particularly focused on the impacts of groundwater pumping on groundwatersurface water interactions and estimated groundwater depletion.To this end, the results of two model setups are compared: (1) a naturalized run assuming no human water use and (2) a human-impacted run including groundwater and surface water uses.The model was run over the period 1960-2100.For the period 1960-2010, the WATCH forcing data [28] was used for the meteorological input.Sectoral water demands were estimated for industries, households, livestock, and irrigation using datasets provided by [26], and were allocated to groundwater and surface water resources using a dynamic allocation scheme including sector-dependent return flows [8,22].Projected future water withdrawals were included by assuming a 'business-as-usual' scenario, in which industrial and domestic demands, as well as the extent of irrigated areas, stay unchanged after 2010, and where irrigation demands only increase or decrease as a result of climate change [8].For the future period, CMIP5 RCP8.5 was used as an emission scenario (the worst-case scenario) [8].Model sensitivity to climate input was tested by running the GSGM using the result of three Global Climate Models (GCMs), as provided by the Inter-Sectoral Impact Model Intercomparison Project ISIMIP (www.isimip.org/) and bias-corrected using the WATCH forcing dataset [28].These three GCMs represent the globally wettest, average, and driest GCMs (GFDL-ESM2M, HadGem2-HS, and MIROC-ESM-CHEH respectively).In this study, we used the model results for the future period to estimate increased river capture under different climate forcings.The previous studies [8,20] showed that groundwater heads and dynamics are most sensitive to parameter settings of conductivity and second to drainage level.In this study, we used the best-performing parameter settings for these parameters, concluded after evaluations against observed groundwater heads and dynamics [8].

Estimation of increased river capture and groundwater depletion
In the natural situation (in general), groundwater and surface water are connected, and groundwater is drained by rivers and streams (figure 1(a)).When groundwater pumping starts, groundwater levels begin to drop.As a result, groundwater discharge will decrease (figure 1(b)), or even reverse into infiltration (figure 1(c)), and evapotranspiration supported by capillary rise from groundwater will decrease as well.The groundwater and surface water systems are still connected and the volume change in groundwater drainage/river infiltration is head-dependent [8,17] (figures 1(b) and (c)).When pumping continues, and the groundwater and surface water systems get disconnected, permanent losses of groundwater from its storage will occur (groundwater depletion) (figure 1(d)).The interaction between groundwater and surface water becomes one-way and infiltration occurs at a constant rate [8,17] (figure 1(d)).In this study, we quantified how much of the pumped groundwater stems from river discharge (increased river capture) and how much stems from groundwater storage (groundwater depletion).We did not explicitly estimate the water pumped that leads to reduced evaporation, which [19] estimates as part of the total increased capture.In our current water demand and water use model, we cannot disregard the influence of irrigation on evapotranspiration, making it impossible to analyze the effects of a declining groundwater table.
Increased river capture, cQ GW [l 3 T −1 ], was calculated at a monthly timestep by comparing simulated groundwater discharges (see SI), Q GW , [l 3 T −1 ], under natural and human-impacted conditions: where Q GW_hum and Q GW_nat are human and natural Q GW respectively.In case Q GW_hum is larger than Q GW_nat no increased capture was estimated.Groundwater depletion, GWD [l 3 ], was estimated at the cell level using head declines caused by groundwater pumping (estimated over the periods 1960-2010 and 2050-2100) and aquifer storativity or specific yields (for confined and unconfined aquifers, respectively): where dH is groundwater head decline [l], S is storativity or specific yield [-], and A is the grid-cell area [l 2 ].

Estimation of pumping from increased capture of river discharge and groundwater storage
For the analyses in this study, we used averaged model outcomes over a historical period (1960-2010) and future period (2050-2100), and show results and the sub-basin level (i.e. level 6 HydroSHEDS [29]).We focused our analysis on sub-basins with notable groundwater abstractions (larger than 0.01 m y −1 ).The fraction of pumped groundwater from river discharge, F GW [-], was calculated over the sub-basin level as: where cQ GW_sb is the cQ GW summed over the subbasin level and GWP_ sb is the groundwater pumped summed over the sub-basin level.
The fraction of pumped groundwater stemming from groundwater storage, F GWS [-], was calculated over the sub-basin levels as: where GWD _sb is GWD summed over the sub-basin.

Impact on river low flows
Next, we studied the impact of groundwater pumping on river flows in more detail.We analyzed the impact of groundwater pumping on the contribution of groundwater discharge to river flows.First, we analyzed the impact of pumping on groundwater discharge to the streams.For that, we estimated monthly groundwater discharges (i.e.Q GW_hum and Q GW_nat respectively) at the grid-cell level and accumulated this along the drainage network.Changes caused by groundwater pumping were compared to the natural conditions, resulting in F Q [-]: Next, we studied the impact on river low flows.We focused specifically on river low flows as the relative contribution of groundwater to streamflow, and the dependency of healthy ecosystems on groundwater is largest during times of low rainfall and low flows.Low flows are calculated as the 90th percentile of average monthly streamflow (Q90, routed along the drainage network), meaning 10% of the months have streamflows below the Q90 value.Changes in Q90 caused by groundwater pumping are estimated as, F Q90 [-]: Q 90_hum and Q 90_nat are estimated low flows under human and naturalized conditions respectively.
Results of both F Q and F Q90 are averaged and presented at the sub-basin level.

Increased river capture and depletion fractions
It is estimated that globally, 20% of pumped groundwater comes from increased river capture (averaged over the model period 1960-2010) and 16% comes from groundwater storage, resulting in groundwater depletion.The global distribution of fractions shows a global average fraction of 0.25 and 0.15, of pumped groundwater stemming from increased river capture and groundwater from storage, respectively (figures 2(a) and (c)).
It is estimated that at the end of the century, 30% (20%-35%) of pumped groundwater globally will come from increased capture and 12% (6%-40%) from storage (numbers between brackets represent the wettest and driest GCMs).The future global distribution of fractions is estimated at 0.22 (0.2-0.45) and 0.1 (0.1-0.2) for increased capture and groundwater storage, respectively (figures 2(b) and (d)).
Globally, differences in fractions exist (figure 2).In general, higher fractions of increased river capture can be found in regions where groundwater pumping ranges from low to medium (i.e.10-100 * 10 6 m 3 m −2 ), while smaller fractions are found in regions where demands are higher (i.e.>100-1000 * 10 6 m 3 m −2 ).In general, higherend ranges of fractions are found for parts of the northeastern USA, west and central Europe, South America, and Russia (figure 2(a)).Higher fractions of groundwater from storage (figure 2(c)) are found in regions where demand is high, for example, in the Central Valley and High Plains aquifer systems in the USA, Indus and Upper Ganges river basins, the Middle East, and parts of Asia.
Regional differences in fractions exist as well, as presented in figure 2. For example, the fraction of pumped groundwater from increased river capture (figure 2(a)) ranges from 0.2-0.8 for the Central Valley (USA); 0.4-0.8 for the High Plains aquifer (USA); 0.3-0.7 for the Indus basin; 0-0.1 for the Upper Ganges basin; 0-0.1 for the North China Plains; and between 0.3 and 0.7 for southern Europe.In general, similar global and regional patterns are projected for the future period (figures 2(b) and (d)).
Figure 3 shows histograms illustrating the worldwide distribution of fractions found in the maps.
When considered alongside the maps in figure 2, it is evident that increased river capture contributes significantly more to global pumping activities than groundwater storage, as indicated by the higher counts (meaning more sub-basins) in the histograms.Also, higher fractions of increased river capture are estimated than those of groundwater depletion.Comparing the past to the future period (considering the average climate forcing) shows that there will be a small rise in water extraction from increased river capture in the future, particularly within the range of 0.2-0.4 and 0.8-1, contributing to a 10% overall increase increased river capture globally.Conversely, smaller fractions of groundwater pumping will occur from storage, particularly in the ranges 0.4-1, contributing to a 4% decline in pumping from storage globally.

Impact on the contribution of groundwater to river flows
On the global scale, the contribution of groundwater to streamflow decreases by 15% when groundwater is pumped.It is estimated that by the end of the century, this decrease will shift to a global increase of 6% under the average climate forcing (and decrease by 1%-2% for the wettest and the driest forcings, respectively).The impact on groundwater contributions varies significantly across the globe.Over the historical period, for example, the most substantial decreases have been observed in regions such as the Midwest of the USA, Mexico, the Mediterranean region, and the Indus River basin.Looking ahead to the future, it is evident that some of these regions will experience a worsening of the negative impacts, notably in the Midwest of the USA and the Indus/India region.On the other hand, certain areas will experience an increase in groundwater discharge contributions, particularly in parts of the Central Valley aquifer, the Mediterranean region, and some regions in North China.The future prediction for these regions is that surface water irrigation will increase, which will have a positive impact on groundwater discharges.
Globally, the decrease in river low flows due to groundwater pumping is less than 5%.These changes in river low flows, however, show large regional differences (figure 4).Substantial decreases, ranging from −30% to more than −50%, have been simulated in, for example, the Indus River basin, the Middle East, southern Europe, Mexico, and Central Valley regions.Conversely, certain regions are estimated to have notable increases of 30% to more than 50% in river low flows, including parts of the lower Ganges basin and the North China Plain.The estimates for the future period show even more pronounced impacts and regional differences.It is estimated that by the end of the century, river low flows will decrease by 11% (7% both for the wettest and driest climate forcings) globally.New regions with substantial decreases in river  low flows (−30% to more than −50%) are expected to emerge, including the eastern part of the USA, Southern Europe, Middle Europe, large parts of the Middle East, and India.On the other hand, regions showing significant increases in river low flows (30% to more than 50%) are expected to develop in parts of South America, parts of the northwest Sub-Sahara region, and parts of Asia, including a large portion of the Ganges basin.

Drivers of increased river capture and contribution to groundwater availability
On the global scale, it is evident that increased river capture is more widespread and extensive than groundwater depletion, as indicated in both maps and histograms (figures 2 and 3).Increased river capture occurs in nearly all regions, whereas depletion tends to be concentrated in areas with high groundwater demands, particularly in drier climates (figure 2).This difference in spatial patterns shows that for the wetter regions, where groundwater and surface water are still connected, increased river capture has a significant positive contribution to the available groundwater that can be pumped from the system without depleting the groundwater.In the drier regions, however, where groundwater pumping rates are large, increased river capture and groundwater pumping are far from equilibrium, and groundwater and surface water get disconnected.The lower relative contribution of increased river capture to the total pumped groundwater in these drier regions does not mean that the estimated volume of increased river capture is smaller than the estimated volume for wetter regions.On the contrary, the volume of increased river capture can be predicted to some extent from the volume of groundwater withdrawals.At the sub-watershed level, there is a moderate correlation between the two variables (R 2 = 0.6).Comparatively, the correlation  between groundwater abstractions and the volume pumped from storage shows a slightly smaller correlation (R 2 = 0.5), suggesting increased capture happens before groundwater depletion is experienced.When we aggregate these results at the larger river basin level by combining several sub-basins, the correlation between increased river capture and the volume of pumped groundwater strengthens significantly (R 2 = 0.8, figure 5).
On the contrary, the fraction of increased river capture cannot be explained by groundwater pumping at all (R 2 = 0.1).When examining the data more closely, however, it is apparent that the greatest proportions of pumping resulting from increased river capture are observed in cases with moderate to high groundwater pumping volumes.Specifically, for pumping volumes ranging from 10 to 1000 m 3 d −1 , the average fractions range from 0.26 to 0.18 (figure 6(a)).
Similarly, the relationship between pumping from storage and groundwater pumping does not exhibit a direct one-to-one correlation (R 2 = 0.1).Nevertheless, when we analyze the distribution of fractions across different groundwater abstraction classes, a distinct pattern emerges.When groundwater abstractions reach high levels, specifically 10 000 m 3 m −2 and above, the average fraction of pumping from storage is 0.21.This value is notably higher than what is observed for the other abstraction classes.
In general, groundwater uses are strongly driven by climate conditions.In drier regions around the world, surface water often falls short of meeting the substantial water demands, leading to the utilization of groundwater as an additional water source.Figure 6(b) illustrates the extent to which the proportions of pumping from increased capture or storage can be attributed to groundwater recharge.The bar plot reveals that the proportions of capture are relatively consistent across all recharge classes, with values ranging between 0.2 and 0.25.In the highest recharge class, however, the fraction is notably lower, averaging around 0.1.Conversely, when analyzing the fraction of depletion, the results are strikingly distinct.They indicate that the highest fractions are associated with the lowest recharge class, with an average of 0.23.In contrast, significantly smaller fractions are observed for the higher recharge classes.
The results depicted in figure 6 provide valuable insights.They demonstrate that predicting the fraction of pumping from increased capture, and thus the dependence of groundwater pumping on groundwater-surface water interactions, is on average considerably more challenging than predicting the fraction of pumping from storage.Regarding the latter, we can deduce that it is most significant in regions characterized by high water demands and low rates of recharge.These regions are typically where groundwater and surface water become disconnected, except for regional drainage connections.It is worth noting, however, that a lower fraction in this context does not necessarily imply a lower volume of capture, as observed in the scatterplot (figure 5).Furthermore, as evidenced by the maps in figure 3, there is a significant regional variation in the estimated fractions, highlighting the complexity and diversity of groundwater and surface water dynamics across different geographical areas.Also, looking towards the future, it is apparent that the rate of increase in capture surpasses that of depletion, and it becomes even more widespread (figure 2).This observation aligns with findings from previous studies, highlighting the growing significance of increased river capture as a key factor influencing groundwater dynamics, and its impact on surface water resources and ecosystems [e.g. 8, 13, 14].

Better understanding of the impacts of groundwater pumping
When considering the impacts of increased river capture and groundwater storage losses, the most substantial effects are observed in the driest regions of the world.In these areas, groundwater is intricately linked to regional drainage systems, and the water demands are particularly high.As a result, these regions experience significant impacts on groundwater baseflow and the streamflow low flows, primarily due to extensive groundwater abstractions and associated depletion (figure 4).
Conversely, regions characterized by strong connectivity between the groundwater and surface water systems, typically found in wetter climates, experience smaller impacts.In these areas, groundwater pumping rates tend to be lower, and depletion is not as pronounced.Although capture from river flow remains relatively high in these regions, it stays in balance with the recharge rates, and the groundwater and surface water systems maintain their interconnectedness.Impacts on groundwater baseflow and streamflow low flows are less pronounced for these regions (figure 4).
The predicted future trends suggest that even a slight increase in groundwater demand and the fraction of increased river capture can lead to larger impacts on groundwater baseflow and streamflow low flows.This underscores the sensitivity of groundwater-surface water interactions to relatively minor changes in water demand and climate conditions.Additionally, these impacts are more widespread compared to changes in groundwater storage, which predominantly occur when the groundwater and surface water systems become disconnected and groundwater pumping reaches its maximum effect.
This conclusion aligns with previous estimations at regional and global scales regarding the environmental impact of groundwater pumping, emphasizing the critical importance of understanding, and managing, groundwater-surface water interactions in the face of evolving water demands and changing climate patterns.

Uncertainties
The model uncertainties that exist in this study have to be acknowledged, which involve various aspects such as input data and parameterization.While the primary objective of this research was not to dive into a detailed examination of these uncertainties (as this had previously been done), ranges have been provided based on the outcomes of a prior sensitivity analysis considering different climate models.In a previously performed sensitivity analysis, where parameter settings of key parameters were systematically varied [8], it became evident that estimates of groundwater heads and groundwater discharge show a higher sensitivity to variations in saturated conductivity compared to drainage levels.It was concluded that the model consistently produced robust trends across diverse parameter settings.In the sensitivity analysis using various climate data sets, it is observed that the climate model representing the driest condition predicts the most substantial global impacts, while the wettest condition results in comparatively smaller impact estimates.Another assumption that influences the results is the assumption of constant water demands (fixed at 2010 levels) except for irrigation water demands, which adjust based on climate impacts.In reality, water demands will rise, following population growth and socio-economic development, and irrigated areas are likely to expand.Adding larger complexity in future water demands and water use scenarios, and more detailed regional data is likely to contribute valuable additional information and regional relevancy beyond the first-order estimates presented in this study.

Conclusion
In conclusion, our study underscores the critical importance of incorporating increased river capture in large-scale water resources assessments, particularly in the context of groundwater pumping.We have observed that when abstractions are at their highest, capture reaches its maximum levels, leading to significant impacts on streamflow and notably affecting metrics such as streamflow low flow and groundwater baseflow.Conversely, in regions characterized by lower abstraction rates, we have observed a relatively high contribution of capture, ranging between 20% to 80%, highlighting the dependency of groundwater pumping on the complex interactions between groundwater and surface water.In such regions, the impacts on streamflow appear to be less pronounced.
Looking ahead to one future climate scenario (RCP8.5), it is evident that the consequences of increased capture will become more pronounced as water availability and demand continue to rise, especially in the face of climate change.Our findings emphasize the sensitivity of groundwater-surface water interactions in this context.
To mitigate and prevent adverse impacts, it is imperative to maintain a balance between groundwater pumping and the replenishment of aquifers through river recharge and infiltrating rainwater.Sustainable groundwater management strategies should prioritize the enhancement of recharge and replenishment, while concurrently reducing water demands.
It is worth noting that our study represents a pioneering effort in estimating the role of increased capture, offering a first-order approximation of its implications.We acknowledge the presence of major assumptions, however, particularly in the realm of water demand projections.Future research should focus on refining these estimates, thereby enabling a comprehensive evaluation of model uncertainties and sensitivities.This will ultimately lead to a better understanding of the regional relevance of global estimates, and enable more effective groundwater management practices in the face of changing hydrological conditions.

Figure 1 .
Figure 1.Schematization of the effect of groundwater pumping on groundwater-surface water interactions and groundwater storage.(a) A natural gaining stream.(b) b1: Pumping starts (q1) and groundwater levels begin to drop, reducing groundwater discharge (GD), but still draining.b2: At first groundwater is taken out of storage.Eventually, a new equilibrium is reached where all pumped water comes from reduced groundwater discharge and evaporation.(c) c1: pumping continues (q2) and groundwater levels drop further and pumping results in river infiltration.c2: More groundwater is taken out of storage, but again, a new equilibrium is reached.(d) d1: Pumping continues further (q3) and groundwater levels drop below the river bed.Surface water infiltration reaches a maximum, independent of groundwater depth.d2: groundwater is persistently taken out of storage leading to a continuous lowering of the water table at a faster rate if pumping rates are higher than surface water infiltration and diffuse recharge over the depression cone.The volume of water that is captured by groundwater pumping in (b), (c), (d) increases river capture.Evaporation (E) is also impacted by pumping, as indicated in the figure.Adapted from[8], with permission.© 2019, The Author(s), under exclusive licence to Springer Nature Limited.

Figure 2 .
Figure 2. Fractions of groundwater pumped from increased river capture (a), (b) and storage (c), (d) at the sub-watershed scale for the past (a), (c) and future (b), (d) periods.

Figure 3 .
Figure 3. Histograms of the worldwide distribution of fractions of pumping from increased river capture (a) and from groundwater storage (b).Comparing the historical (blue) and future periods (red).

Figure 4 .
Figure 4. Impacts of groundwater pumping on groundwater's contribution to river flow (a), (b) and river low flows (c), (d), for the past (a), (c) and future periods (b), (d) presented as the fraction of change compared to the naturalized conditions at the sub-watershed level.The data in (a) and (b) are masked for watersheds where groundwater pumping is significant (white), and the data in (c) and (d) show missing data (mv) for regions where no river discharge is simulated.

Figure 5 .
Figure 5. Scatter plot of simulated increased river capture against groundwater abstractions averaged over larger river basins (200 total globally), showing R 2 of 0.84.

Figure 6 .
Figure 6.Global distribution of fractions of increased river discharge and pumping from groundwater storage per groundwater abstraction class (a) and recharge class (b).Results are from the past period and shown at the sub-watershed level.