Introduction
Civilizations have typically obtained water from natural and constructed surface-water resources throughout most of human history. Only during the last 50–70 years has a significant quantity of water for humans been obtained through pumping from wells (Alley et al.
2002). During this short time, alarming levels of groundwater depletion have been observed in many regions, especially in semi-arid and arid areas that rely heavily on groundwater pumping from clastic sedimentary basins (Konikow
2013; Taylor et al.
2013; Wada et al.
2010,
2012). Groundwater has commonly been a source of high-quality freshwater and an important safeguard against uncertain inter-annual and inter-decadal shortfalls in precipitation and surface-water supplies (Hanson et al.
2012). However, overdraft of this important resource has only accelerated during the twenty-first century (Wada et al.
2011) and is further threatened by future climate uncertainty (Milly et al.
2008; Mirchi et al.
2013). Despite continued unsustainable groundwater abstraction in many areas, water policy efforts continue to respond to near-term crises and fail to anticipate long-term future conditions (Karl et al.
2009).
Large inter-annual variability of precipitation and streamflow (Dettinger et al.
2011) and heavy reliance on groundwater pumping for agricultural irrigation (Scanlon et al.
2012) has created an especially precarious scenario for water management in California’s Central Valley (CV) aquifer system, which has had overdraft conditions in its southern portion for decades (Brush et al.
2013; Faunt et al.
2009; Konikow
2013; Scanlon et al.
2012). Historically abundant snowmelt runoff from the Sierra Nevada provides an estimated 54% of water for CV crops, on average (Faunt et al.
2009), with the remainder provided by direct precipitation and groundwater. Average temperatures in California (CA) are expected to increase by 1.5–4.5 °C by the end of the twenty-first century (Cayan et al.
2008), which will decrease the proportion of precipitation as snow and initiate earlier spring snowmelt and runoff (Cayan et al.
2008,
2010; Vicuña and Dracup
2007), increase evapotranspiration (ET) and decrease late-season baseflow (Hayhoe et al.
2007; Huntington and Niswonger
2012), and likely increase the likelihood of co-occurring flooding and water-shortages in the same water year (Knowles et al.
2006; Swain et al.
2018). Future population growth and land-use change in CA, USA, and will likely increase drought risk (Barnett et al.
2008; Cayan et al.
2010), and elevate competition for existing water resources (Gleick
2000). These stressors are not unique to CA, but are symptomatic of increasing vulnerability of water resources worldwide (Stewart et al.
2005; Vicuña et al.
2011), especially in snowmelt-fed semi-arid and arid regions (Konikow
2013).
To reverse the negative effects of overexploitation of groundwater resources, groundwater must transition from being treated mainly as an extractive resource to one in which recharge and subsurface storage are pursued more aggressively. This remains a challenge because unlike surface-water reservoirs that are typically replenished on annual timescales, the clastic sedimentary aquifer systems are replenished on much longer time scales (Taylor et al.
2013), especially if no particular effort is devoted to augmenting groundwater recharge. Managed aquifer recharge (MAR) has been used for decades to supplement natural recharge and to strategically store surface water in groundwater aquifer systems for future water supply (Bouwer
2002), often as part of a conjunctive use (Bredehoeft and Young
1983) and/or water market framework (Israel and Lund
1995), including in select groundwater basins in CA (e.g., Asano
2016; Kletzing
1987), elsewhere in the Southwestern US (e.g., Jacobs and Holway
2004), and globally (e.g., Dillon et al.
2019). There is interest in expanding use of MAR in CA, both to offset overdraft and to hedge against future decreases in snowpack-water storage and changes in the timing and volume of surface-water availability. Established MAR projects commonly use dedicated infiltration basins located over locally coarse-texture geological deposits to increase recharge, but increasingly, MAR on agricultural fields during nongrowing seasons (Ag-MAR) has been proposed as an alternative to infiltration basins (Dahlke et al.
2018; Harter and Dahlke
2014; Niswonger et al.
2017). Studies have noted that even during periods of water scarcity, wet-season high-magnitude streamflows (HMF) can often provide ample unmanaged surface water for MAR during nongrowing seasons in CA (Beganskas and Fisher
2017; Kocis and Dahlke
2017) and elsewhere (Chinnasamy et al.
2018). The largest quantity of winter HMF in CA are in the CV, where most of the statewide groundwater overdraft occurs. HMF in the CV typically occur during episodic 5–7 day windows (Kocis and Dahlke
2017), and can be quite large (3.2 km
3 annual average during years with HMF). However, augmenting groundwater recharge with ephemeral HMF remains a challenge in the CV because the sedimentary aquifer system is composed of mostly silt and clay sediments (Faunt et al.
2009) that form nearly ubiquitous, multiple confining layers that create semiconfined conditions and limit infiltration rates over most of the landscape.
Geologic heterogeneity is ubiquitous across scales and strongly affects movement of water and solutes through the subsurface; however, seldom are enough subsurface data available to represent heterogeneous features explicitly in models (De Marsily et al.
2005; Koltermann and Gorelick
1996). As a result, development of groundwater models often involves simplifying or up-scaling heterogeneity to enable adequate representation of regional groundwater flows for purposes of regional water resources management. For example, in models of sedimentary aquifer systems (e.g., Phillips and Belitz
1991; Fogg
1986) where the aquifer sediments amount to 20–50% of the aquifer systems (the remainder being aquitard sediments), effects of the ubiquitous aquitard beds are approximated by regionally reducing the vertical hydraulic conductivity (K
v) relative to the horizontal hydraulic conductivity (K
h) by several orders of magnitude in order to match both horizontal and vertical hydraulic gradients. This approach is capable of producing good approximations of regional flows and groundwater budget components, but tends to smooth any local variations in recharge and vertical flow in the sedimentary connected network (e.g., Fogg et al.
2000; Fleckenstein et al.
2006). Stochastic methods like transition-probability-based-indicator geostatistics provide opportunity for realistically representing subsurface heterogeneity of the major aquifer and aquitard facies while honoring measured data (Carle and Fogg
1996; Weissmann and Fogg
1999; Weissmann et al.
1999). Results from studies implementing these methods show strong influence of subsurface heterogeneity on groundwater/surface-water interactions (Engdahl et al.
2010; Fleckenstein et al.
2006; Liu
2014).
A number of studies have measured infiltration and recharge processes in porous media in field settings (e.g., Batlle-Aguilar and Cook
2012; Bresciani et al.
2018) and laboratory settings (e.g., Fichtner et al.
2019). Other work has demonstrated the importance of geologic heterogeneity on infiltration processes and stream–aquifer interactions, including the often-significant contribution of focused stream leakage in arid and semi-arid areas (e.g., de Vries and Simmers
2002; Bresciani et al.
2018; Irvine et al.
2012), including the CV (Fleckenstein et al.
2006). Geologic heterogeneity is also important for natural recharge processes and MAR in karst systems (Hartmann et al.
2017; Xanke et al.
2017). In heterogeneous clastic sedimentary systems, even a small fraction of permeable hydrofacies in correlated random media tend to be interconnected in three dimensions (Fogg et al.
2000; Harter
2005), especially in the absence of spatially persistent geologic unconformities, providing potential recharge pathways to semi-confined aquifer systems. Interconnected, highly permeable sand and gravel deposits have been shown to occur in select locations in the southern CV that are potentially conducive to considerably higher rates of regional recharge than would be possible over the rest of the landscape (Weissmann et al.
2004). Studies have suggested the presence of these features in the northern CV (Meirovitz
2010; Shlemon
1967), compelling further study of MAR dynamics in this system. Several studies have focused on synthesizing MAR suitability characteristics from a combination of spatial data, including remote-sensed imagery, geologic maps, and soil surveys to identify favorable surface site characteristics for MAR in CA (O’Geen et al.
2015), and elsewhere (Adham et al.
2010; Ghayoumian et al.
2007). These studies provide a valuable initial survey of site suitability, but do not account for deeper subsurface geologic heterogeneity that has been shown to be important for recharge (e.g., Weissmann et al.
2004).
This research aims to explicitly simulate variably saturated water-flow dynamics in a highly resolved representation of complex subsurface geologic heterogeneity of the CV that includes both interconnected, highly permeable sand and gravel deposits and more typical silt- and clay-dominated sediments. Additionally, the goal of this research is to (1) gain insight into infiltration and recharge phenomena that are challenging to observe and have not been included in regional-scale groundwater models, and (2) guide MAR strategies for regions reliant on diminishing snowpack water storage and in overdrafted groundwater basins.
Discussion
Increasing water scarcity has accelerated overdraft of groundwater resources in CA (Scanlon et al.
2012) and globally (Wada et al.
2011) and has created newfound interest in replenishing overdrafted aquifer systems with MAR. Studies have shown that thoughtful implementation of MAR practices can increase sustainability of groundwater resources (Niswonger et al.
2017). However, widespread implementation of MAR is often impeded by institutional barriers like transfer of water rights and water accounting uncertainty (Asano
2016), along with infrastructure limitations, including water conveyance and land acquisition costs (Gailey
2018), and water quality concerns (Hartog and Stuyfzand
2017). Past research has shown the importance of heterogeneity on recharge processes in clastic sedimentary aquifer systems (e.g., Fleckenstein et al.
2006; Irvine et al.
2012). Results presented here further demonstrate that subsurface sedimentary geologic architecture is an important consideration for infiltration and recharge processes, and especially so when considering MAR effectiveness. Results show a highly transient recharge response that is consistent with field experiments (Batlle-Aguilar and Cook
2012) and a wide range of recharge rates that are possible across the landscape, including rapid, high-volume recharge in select areas where sand and gravel IVF outcrop at land surface. Siting MAR over IVF deposits has the potential for outsized recharge rates compared to the rest of the landscape.
For example, these results suggest that the hypothetical cumulative recharge volume for site 1 during the 180-day scenario (~2.0 × 10
8 m
3) is >1/10th of the estimated annual groundwater overdraft for the CV, CA, during 2003–2010 (Famiglietti et al.
2011). Interannual variability of precipitation in CA is large, and is typically derived from just a few storm events (Dettinger et al.
2011), and excess surface water available for recharge in CA is flashy and typically falls within a short (<10 day) window (Kocis and Dahlke
2017), so identifying sites that can accommodate large volumes of recharge during a short period of time is paramount. Conversely, results show that laterally continuous fine-grained facies can impede MAR rates considerably, and may limit MAR feasibility over much of the landscape. Results show that vertical interconnection of coarse-texture facies is important for MAR, rather than just the upper-most facies designation. This suggests that GIS-derived surface metrics of recharge potential, while valuable, are not fully diagnostic of MAR potential. Rather, these data products should be considered as an important component of a thorough site evaluation that includes investigation of deeper subsurface geologic architecture.
Importantly, IVF deposits have been identified in several CA river fans, including the American, Tuolumne, and Kings rivers, and likely occur in other major river fans that drain high-elevation, glacially influenced catchments on the west side of the Sierra Nevada (Meirovitz
2010; Weissman et al.
2005; Weissmann and Fogg
1999; Weissmann et al.
2004). In addition, similar coarse-texture, glacially influenced IVF have been identified in other fans in the Western USA (e.g., Pierce and Scott
1983). The authors believe that results presented here can foster renewed effort on the part of the hydrologic sciences community to identify and catalog locations with IVF for MAR. These results demonstrate that the recharge potential for these features is sufficiently strong that they could be considered for special land use prioritization such as a
recharge preserve.
One may conclude that the relationship between proportion of coarse-texture facies and recharge potential illustrated here can be deduced from first principles, and is thus trivial. Certainly, in a general sense, this finding is obvious; however, to the authors’ knowledge, no study has used a 3D variably saturated water flow code to explicitly simulate MAR dynamics through a highly heterogeneous domain. This approach couples a detailed representation of subsurface geology with physically realistic water flow physics to elucidate important processes that can (1) help improve representation of recharge processes in coarse-resolution, management-focused groundwater models, (2) help prioritize site investigation and data collection for proposed MAR projects, and (3) inform management entities to the potential impacts of MAR.
Results illustrate an important dichotomy between change in storage and pressure response in the aquifer system. Indeed, results show that a pressure response can be registered in wells screened in the semi-confined aquifer system several kilometers from the origin of the recharge stress. Of course, the increase in pressure is not related to physical water from the recharge site entering that well, and these results illustrate this important concept, which can be a challenging to convey to laypersons, and has important implications for water rights and water management. For example, recently-passed groundwater management legislation in CA requires the creation of local groundwater sustainability agencies (GSAs) to limit both the “chronic lowering of groundwater levels” and “significant and unreasonable reductions in groundwater storage” (Kiparsky et al.
2016). Results presented here demonstrate that MAR can help mitigate both of these impacts. While the benefit of physical change in storage occurs locally, the increase in groundwater heads can be regionally beneficial, potentially benefitting adjacent jurisdictions outside of the immediate GSA.
These results also suggest that while networks of interconnected coarse-texture facies provide a conduit for rapid infiltration and widespread pressure response, the fine-texture facies accommodate a substantial fraction of the total recharge volume. This finding is consistent with other work showing the importance of fine-texture facies storage in aquifer systems (e.g., Konikow and Neuzil
2007), and challenges common aquifer-system conceptual frameworks, wherein fine-texture facies are often considered a non or minimally contributing component of the aquifer system. These findings suggest that fine-texture facies may in fact be the largest reservoir in this alluvial aquifer system. Preliminary sensitivity analyses indicate this response is fairly robust to parameter uncertainty (see section ‘
Parameter sensitivity on storage accommodated by fine-texture facies’). Importantly, these findings support conceptual models of groundwater flow and storage in alluvial aquifer systems that include fine-texture facies as an important storage reservoir (e.g., Konikow and Neuzil
2007). In essence, the connected network of coarse-texture facies provide for relatively fast flow and recharge phenomena, while the fine-texture facies end up accommodating most of the changes in storage but on longer time scales. The storage depletion and replenishment can be viewed as a two-stage process, in which rapid declines in storage occur in the coarse-texture aquifer network followed by slow drainage (leakage) from the fines. Conversely, during storage augmentation, the immediate increases occur in the aquifer network on time scales of days to months, followed by much slower but pervasive increases in storage in the fine-texture facies on time scales of months to years. From a whole-watershed perspective, one can deduce fine-texture facies to be the largest (but least accessible) reservoir within this system, followed by coarse-texture facies, and finally surface-water reservoirs, which are the most readily accessible and replenishable. In general, these results are somewhat reminiscent of the leaky aquifer analytical model development by Neuman and Witherspoon (
1972), who pointed out an investigation bias toward the hydrology of aquifers and suggested that future work should focus on the aquifer-aquitard complexes that compose aquifer systems.
The authors acknowledge some limitations to the approach—for example, the TPROGS technique for developing the geologic domain is informed by ample conditioning data; however, the approach is inherently stochastic, which can limit the robustness of facies prediction in areas of the domain with sparse conditioning data. In addition, only a single TPROGS realization was used for these simulations, and the authors acknowledge that a more rigorous ensemble approach could provide greater insight into potential facies distributions within the domain. In addition, the model spin-up included a domain-wide 1 mm day
−1 recharge flux during the final year to facilitate model convergence for subsequent MAR simulations. This boundary condition may be unrealistic with respect to recharge rates reported for this semi-arid area (25–275 mm year
−1; Fleckenstein et al.
2006) and may contribute to some overestimation of antecedent soil-moisture conditions in the uppermost model cells and influence recharge rates for subsequent MAR simulations. Importantly, this study should not be treated as a thorough site investigation for MAR in this region. Rather, the authors present these findings as a proof-of-concept to demonstrate the influence of geologic heterogeneity on MAR dynamics in a hypothetical but physically realistic domain. Moreover, the simulations presented here do not consider several surface processes that influence real-world MAR feasibility and dynamics, including topographic site limitations, evaporative losses, and clogging effects (Bouwer
2002). In addition, the authors acknowledge that a more detailed investigation of the role of geologic heterogeneity on
K upscaling and a rigorous uncertainty or sensitivity analyses of hydraulic properties would permit broader interpretation of these findings, and thus warrants further study. Despite these limitations, these findings have implications for understanding MAR dynamics and for assessing MAR feasibility in clastic alluvial groundwater basins in CA and globally.
Conclusions
This research explores variably-saturated water flow dynamics in a highly heterogeneous geologic domain that reflects the complex alluvial geologic architecture on the east side of the Northern Central Valley, CA. The research objectives are to inform MAR implementation in CA and elsewhere by (1) highlighting the role of subsurface geology for recharge dynamics and (2) identifying important recharge phenomena that are not easily observed or simulated by typically coarse-resolution regional groundwater models. The approach uses the variably saturated water-flow code, ParFlow, to simulate recharge over a range of configurations of unconsolidated alluvial geology, including over sand and gravel IVF deposits that interconnect from land surface to the deeper semi-confined aquifer system. Two sets of recharge scenarios were simulated at five sites to evaluate system response to both prolonged, and shorter successive recharge stress.
Results show a large (nearly 2 order-of-magnitude) range of cumulative recharge volumes between sites that is dependent primarily on the configuration of subsurface geologic facies. Recharge rates were highly variable in time, with all sites showing relatively high initial rates (e.g., >100 and >50 cm day−1 during the first 10 days of simulation time for sites 1 and 2, respectively) followed by rapid decay to a quasi-constant recharge rate. Results demonstrate that the overall subsurface geologic architecture, rather than just the upper-most soils or facies designation, is important for recharge. All sites showed >4× reduction in average recharge rates for the last 30 days compared to the first 30 days of the 180-day simulation, and a ~2× reduction in cumulative recharge for year 3 compared to year 1 for the multi-year simulations. This behavior reflects the effect of groundwater mounding that limits rapid filling of coarse-texture UZ facies.
Results suggest that the majority of recharge volume is accommodated by filling unsaturated-zone facies, but where there is sufficient hydraulic communication between land surface and the deeper aquifer system, the majority of the pressure response is propagated through the saturated aquifer system once the recharge wetting front intersects the water table. Results show that if there is sufficient hydraulic connection between the recharge site and the semi-confined aquifer system, the recharge pressure response can be widespread and rapid, propagating over several kilometers over a period of days or weeks. These results provide a valuable illustration of two physically distinct benefits of recharge: (1) local increases in groundwater storage and (2) the possibility of a more widespread re-pressurization effect in the regional aquifer system. The distinction between these responses has important implications for water rights, groundwater management regulations, and other water-policy issues.
Results also suggest that while the majority of water volume and pressure response is transmitted through coarse-texture facies, the majority of the recharge volume is eventually stored in fine-texture facies, even for sites that have disproportionally large fractions of coarse-texture facies. This result suggests that fine-texture facies are the largest, albeit least accessible, reservoir for recharge in this system. This finding has important implications for aquifer conceptualization, because fine-texture facies are often considered as aquitards (or aquicludes) that do not appreciably participate as part of the overall aquifer system.