A common water quality issue is an excess of nutrients which can lead to problems such as eutrophication. Stream restoration is one method by which improvements in water quality may be attempted. One strategy is increasing hyporheic zone flow at baseflow by addition of instream structures. The hyporheic zone can be an area of increased biogeochemical activity, with potential enhancement of reactions such as denitrification. However, the comparative effects of various instream restoration techniques, as well as the role of watershed setting and corresponding environmental characteristics in which restoration occurs (e.g., hydraulic conductivity, stream slope), are still poorly understood. In this study we numerically modeled groundwater and surface water interaction in a 200 m second order stream reach in southwestern Virginia using MIKE SHE. We calibrated the model to hydrologic and tracer data available during field tests of restoration techniques. We then simulated different types of instream restoration techniques (e.g., fully and partially channel-spanning weirs and buried structures), and varied hydrologic and biogeochemical controlling factors driven by watershed setting. The measured effects for this sensitivity analysis were direction and magnitude of surface water-groundwater exchange and amount of denitrification. We found that factors related to watershed setting had the greatest effect on surface water-groundwater exchange and on denitrification, including streambed hydraulic conductivity, natural or background stream topography and slope, and groundwater levels. Type and number of instream structures also influenced surface water-groundwater exchange and denitrification, but to a lesser degree, and the effect of structures was in turn controlled by watershed setting. Watershed setting was thus the largest control, both on exchange overall, and the effectiveness of structures. Human effects on watersheds such as agriculture and urbanization therefore likely play a role in whether reach-scale restoration practices succeed in achieving water quality goals. More broadly, restoration efforts at the watershed scale itself, such as reducing fertilizer use or improving stormwater management, may be necessary to achieve ambitious water quality goals. Nevertheless, reach-scale restoration efforts such as in-stream structures may play a useful role in certain watershed settings. Furthermore, other reach-scale restoration techniques that affect streambed topography, such as addition of pool-riffle sequences, may be more effective, and bear investigation. / Master of Science / A common water quality issue is an excess of nutrients which can lead to problems such as algal blooms. Stream restoration is one method by which improvements in water quality may be attempted. One strategy is increasing hyporheic zone flow by addition of instream structures. The hyporheic zone is an area of the stream bed and banks where there is increased biogeochemical activity, with potential enhancement of reactions that may remove nutrients such as denitrification. However, the comparative effects of various instream restoration techniques, as well as the role of watershed setting and corresponding environmental characteristics in which restoration occurs (e.g., hydraulic conductivity, stream slope), are still poorly understood. In this study we numerically modeled groundwater and surface water interaction in a 200 m headwater stream reach in southwestern Virginia using MIKE SHE. We calibrated the model to hydrologic and tracer data available during field tests of restoration techniques. We then simulated different types of instream restoration techniques (e.g., fully and partially channel-spanning weirs and buried structures), and varied hydrologic and biogeochemical controlling factors driven by watershed setting. The measured effects for this sensitivity analysis were direction and magnitude of surface water-groundwater exchange and amount of denitrification. We found that factors related to watershed setting had the greatest effect on surface water-groundwater exchange and on denitrification, including streambed hydraulic conductivity, natural stream topography and slope, and groundwater levels. Type and number of instream structures also influenced surface water-groundwater exchange and denitrification, but to a lesser degree, and the effect of structures was in turn controlled by watershed setting. Watershed setting was thus the largest control, both on exchange overall, and the effectiveness of structures. Human effects on watersheds such as agriculture and urbanization therefore likely play a role in whether reach-scale restoration practices succeed in achieving water quality goals. More broadly, restoration efforts at the watershed scale itself, such as reducing fertilizer use or improving stormwater management, may be necessary to achieve ambitious water quality goals. Nevertheless, reach-scale restoration efforts such as instream structures may play a useful role in certain watershed settings. Furthermore, other reach-scale restoration techniques that affect streambed topography, such as addition of pool-riffle sequences, may be more effective, and bear investigation.

This work focuses on the use of Artificial Intelligence (AI), and in particular Machine Learning (ML) to tackle quality and quantity aspects of both surface water and groundwater. Traditionally, river water quality modelling and contaminant transport in groundwater studies resort to the solution of physical-based (PB) equations, which aim to define a conceptual model of reality. The complexity of the processes involved, in some cases undisclosed or indiscernible, calls for a sensitive parameterization by the modeler. For such reason, the PB models can be limited by the complexity of the system, the availability of data, and the consequent need for simplifying assumptions. On the other hand, ML models are data-driven and rely on algorithms to identify patterns in data. These techniques aim to extract a surrogate representation of the reality by learning existing correlations in data. They can handle complex and non-linear relationships between variables and can be more flexible and adaptable to new environments. However, they are directly affected by the quality and quantity of available data, requiring larger datasets than PB models. To explore the potential of these methods in addressing surface and groundwater challenges,we experimented with different algorithms in three distinct applications. First, we compared two ML techniques for a water quality catchment-scale model and the most performing was then employed to fill the gaps in environmental time series and to enhance the prediction of a PB model in the groundwater context. Therefore, in the first part of this work, a water quality model of the Adige River Basin is presented and discussed. For this purpose, Random Forest (RF) and Dense feed-forward Neural Network (DNN) were applied and compared to a standard linear regression (LR) approach and an Importance Features Assessment (IFA) of the drivers was performed. DNN showed to be more flexible and effective in detecting non-linear relationships than RF. LR performed at a satisfactory level, similar to RF and DNN, only when drivers linearly correlated to the observational variable were used, and a limited fraction of variability was explained. However, important drivers, non-linearly related to the water quality variables of interest introduced a significant gain when DNN was used. Regarding the variables investigated, water temperature and dissolved oxygen were modeled accurately, using RF or DNN, and sufficient accuracy was obtained by using the minimum information available, represented here by the Julian day of the measurements embodying the seasonality. The other variables showed instead a more balanced influence by the complete set of drivers, appreciable in the IFA procedure for DNN and RF, and a geogenic origin and anthropogenic disturbances were confirmed for chemical contaminants. The proposed analysis, by means of ML algorithms and through the IFA of the drivers, can be applied to predict spatial and temporal variability of contaminant concentrations and physical parameters and to identify the external forcing exerting the most relevant impacts on the dynamics of water quality variables. The second part of the thesis investigated the use of the DNN algorithm to gap-fill time series measurements, for daily flow rate and daily water temperatures from different sites downstream of the Careser glacier, in Pejo valley (northeastern Italy). Thus, an in-depth analysis of the streamflow response to the hydrological regime alterations of the glacier was carried out, through the reconstruction of the time series of the flow rate measured at a gauging station downstream the glacier, in the period 1976-2019. The water temperature time series, instead, were correlated to the macro-invertebrate population’s statistics in the same period at four sites along the Careser stream from the glacier to the reservoir immediately downstream the Careser Baia gauging station. In the first step, the water temperature was modelled just through the Julian day and air temperature information and, subsequently, precipitation, reconstructed flow rate, and evapotranspiration were introduced for sensitivity analysis of the features. With air temperature projections, the DNN model of the water temperature was also applied to simulate future scenarios up to 2050, considering different emission pathways. In this case, DNN proved to be a reliable tool for gap-filling the observational time series, even for time series with many gaps. The reconstructions of the water temperature allowed us to estimate the delay between the warming in air and water temperature and the effect on the biological invertebrate species in the glacier streams. The sensitivity analysis of the features was again key in underlining the contributions of the forcing available, unveiling the combined effects of the warming in air temperature and the decline of flow rate on the water temperature increase. The in-depth analysis of the flow rate revealed, besides the dramatic reduction of streamflow, the anticipation of the summer peak and the negligible influence of the precipitation in these alterations. Lastly, the framework for an ML-PB hybrid model in the context of contaminant transport by groundwater was presented. In this procedure, the contaminant concentration at several sampling locations was associated with physical parameters characterizing the aquifer. Through a synthetic case, a DNN model was employed to predict the physical parameters and a simplified PB equation was used to project the concentration into the future. The analysis demonstrated the capability of DNN to predict physical parameters by capitalizing on the information contained in the available concentration measurements. The thesis is articulated through 7 chapters. In Chapter 1, a broad overview of Machine Learning is presented, with its specific applications in Water sciences and the consequent motivations and objectives of this research. In Chapter 2 the main Machine Learning basic concepts are clarified and presented, in order to set the floor for the successive developments in which ML is applied to surface and subsurface hydrology. Chapter 3 covers the Machine Learning and statistical algorithms employed for modeling in the current research. In Chapter 4, Adige water catchment case study is presented and discussed. In Chapter 5, the gap-filling time series procedure for Careser case study is presented for both the variables investigated. In Chapter 6 the results of the hybrid Machine-Learning Physics-Based application of a groundwater model on synthetic data are presented. Finally, remarks and conclusions are summarized in Chapter 7, which provides also perspective work for these applications.

>Magister Scientiae - MSc / In the Kruger National Park (KNP) which is the flagship conservation area in South Africa, the impact on groundwater should be kept to a minimum as groundwater plays a vital role in sustaining ecosystem functioning and sustaining baseflow to streams and rivers. For this reason groundwater has been recognized as one of the environmental indicators that need to be monitored. The KNP has adopted a Strategic Adaptive Management (SAM) approach with clear ecosystem management goals. The achievement of these goals is evaluated by using environmental indicators. These indicators are evaluated against thresholds of potential concern (TPC). TPCs are a set of boundaries that together define the spatiotemporal conditions for which the KNP ecosystem is managed. TPCs are essentially upper and lower limits along a continuum of change in selected environmental indicators. Historically, groundwater recharge and surface water interaction with rivers has tended to be overlooked in the KNP. This study proposes a conceptual model of groundwater recharge processes in the KNP, defining when and how groundwater recharge occurs. Two methods were used, the Cumulative Rainfall Departure (CRD) and stable isotopes of ²H and ¹⁸O. An adapted version of the CRD which incorporates a long and short term memory of the system was used to identify possible recharge processes. Further, using the CRD method a reliable reconstruction of the long term groundwater level trends are simulated using monthly rainfall totals with reference to the average rainfall over the entire time series 1936-2009. The stable isotope of ²H and ¹⁸O samples from cumulative rainfall samplers, surfacewater (streams and rivers) and groundwater from boreholes were collected monthly for approximately one year (May 2010 to July 2011). The isotope composition of the groundwater was used to establish whether recharge was immediate or delayed. Additionally, the isotopic composition of surface-water from rivers and streams were compared to that of groundwater to identify surface-water interactions. Groundwater recharge in KNP occurs during the rainy summer months (December to March) and very little to none during the dry winter season (April to September). Recharge takes place during rainfall sequences 100mm or more. The stable isotope records collected from cumulative rainfall, groundwater and surface water (streams and rivers) indicate that groundwater experiences evaporation prior to infiltration. As the KNP experiences high evaporation rates, insignificant rainfall sequences contribute little or zero to recharge. The CRD analysis of groundwater level fluctuations shows that recharge to the aquifers respond to dry and wet cycles that last for 6 to 14 years. The KNP experienced several periods of below-average rainfall and hence no significant recharge took place to the basement aquifers. During a normal rainy season the water levels rise somewhat then starts receding again. It is only during major rainfall events that may occur every 100yrs to 200yrs causing the aquifers to fully recharge. This was perfectly illustrated by the high groundwater levels after the 2000 major rainfall event that recharged the aquifers fully. During below average rainfall years the overall water level trend is drastically declining. The system experiences higher natural losses than gains due to outflow of groundwater to streams and rivers. The KNP is divided down the center by two geological formations, granites along the west and basalts along the east. The combination of the CRD model and the stable isotopic analysis suggest that the dominant recharge processes that occur in the southern region of the KNP are direct recharge via piston flow and indirect recharge via preferred pathways particularly streams and rivers. Along the eastern half of the KNP on the Basalts and Rhyolite direct recharge via piston flow are dominant. Groundwater is not recharged via small streams and rivers (Sweni and Mnondozi Rivers) as it was found that at these particular sites these rivers are detached and do not interact with groundwater. Along the western granitic areas the dominant recharge process are indirect recharge. Recharge takes place via preferred pathways particularly streams and rivers. It was found that ephemeral rivers (Nwatsisonto River) act as sinks for groundwater recharge and influent-effluent conditions are experienced along seasonal rivers (Mbyamiti River). The large perennial Sabie and its tributary the Sand River are consistently fed by groundwater, above all maintaining base flow during the dry season. These rivers act as basin sinks receiving groundwater discharge all year round. Using the stable isotope composition of rainfall, surface-water and groundwater to act as a natural tracer, in combination with the CRD method proved invaluable to confirm the plausible recharge processes. The study provided a conceptual understanding of the groundwater system in the KNP forming the foundation to developing acceptable limits (TPCs) of the groundwater levels in the KNP. The model will serve as a guide for the recharge processes and for deciding on the location and time frames for data collection to ultimately set TPCs for groundwater in the KNP to sustainably manage the resource.

No description available.

Hydroelectric releases on the Deerfield River in northwestern Massachusetts affect surface water-groundwater interactions there by daily reversing the head gradient between river and groundwater. Artificially elevated stage drives river water into the riparian aquifer. Water budget analysis indicates that roughly 10% of this bank-stored water is permanently lost from the river system in a 19.5 km reach, likely as a result of transpiration by bank vegetation. Field observations as well as two-dimensional modeling results show that water losses are not uniform throughout the study reach. Riparian aquifer transmissivity in river sub-reaches largely determines the magnitude of surface water-groundwater exchange as well as net water loss from the river. These newly documented dam-induced losses from river systems inform decisions by river managers and hydroelectric operators of additional tradeoffs of oscillatory dam-release river management.

No description available.

Increases in reactive nitrogen from human actions have led to negative impacts on surface water (SW) and groundwater (GW) quality, and it is important to better understand denitrification processes in aquatic systems. The hyporheic zone has unique biogeochemical conditions, and is known to attenuate contaminants originating from SW and traveling through the hyporheic zone, together with necessary reactants. However, the ability of the hyporheic zone to attenuate contaminants from deeper upwelling GW plumes as they exit to SW is less understood. I used MODFLOW and SEAM3D to simulate hyporheic flow cells induced by riverbed dunes and upwelling GW together with mixing dependent denitrification of an upwelling nitrate (NO3-) plume. My basecase model scenario entailed dissolved organic carbon (DOC) and dissolved oxygen (DO) advecting from SW and DO and NO3- advecting from GW, which is typical of water in agricultural land uses. I conducted a sensitivity analysis to determine controls on mixing dependent denitrification. Mixing dependent denitrification increased with increasing hydraulic conductivity, decreasing lower bottom flux, as well as increasing DOC in SW and NO3- in GW. Non-mixing dependent denitrification also occurred when there was SW NO3-, and I found its magnitude was much greater than mixing dependent denitrification. Nevertheless, potential for hyporheic zones to attenuate upwelling NO3- plumes seems to be substantial, though highly variable depending on biogeochemical reaction rates as well as geomorphic, hydraulic and biogeochemical conditions. Stream and river restoration efforts may be able to increase both mixing and non-mixing dependent reactions by increasing hyporheic zone residence times. / Master of Science

Restoring hydrologic connectivity between the channel and floodplain is a common practice in stream and river restoration. Floodplain hydrology and hydrogeology impact biogeochemical processing and potential nutrient removal, yet rigorous field evaluations of surface and groundwater flows during overbank floods are rare. We conducted five sets of experimental floods to mimic floodplain reconnection. Experimental floods entailed pumping stream water onto an existing floodplain swale, and were conducted throughout the year to capture seasonal variation. Each set of experimental floods entailed two replicate floods occurring on successive days to test the effect of varying antecedent moisture. Water levels and specific conductivity were measured in surface water, shallow soils, and deep soils, along with surface flow into and out of the floodplain. Total flood water storage increased as vegetation density increased and or antecedent moisture decreased. Hydrologic flow mechanisms were spatially and temporally heterogeneous in surface water, in groundwater, as well as in exchange between the two and appeared to coexist in small areas. Immediate propagation of hydrostatic pressure into deep soils was suggested at some locations. Preferential groundwater flow was suggested in locations where the pressure and electrical conductivity signals propagated too fast for bulk Darcy flow through porous media. Preferential flow was particularly obvious where the pressure signal bypassed an intermediate depth but was observed at a deeper depth. Bulk Darcy flow in combination with preferential flow was suggested at locations where the flood pressure and electrical conductivity signal propagated more slowly yet arrived too quickly to be described using Darcy's Law. Finally, other areas exhibited no transmission of pressure or conductivity signals, indicating a complete lack of groundwater flow. Antecedent moisture affected the flood pulse arrival time and in some cases vertical connectivity with deeper sediments while vegetation density altered surface water storage volume. Understanding the variety of exchange mechanisms and their spatial variability will help understand the observed variability of floodplain impacts on water quality, and ultimately improve the effectiveness of floodplain restoration in reducing excess nutrient in river basins. / Master of Science

The pollution of streams and rivers is a growing concern, and environmental guidance increasingly suggests stream restoration to improve water quality. �Solute retention in off channel storage zones such as hyporheic zones and floodplains is typically necessary for significant reaction to occur. �Yet the effects of two common restoration techniques, in stream structures and inset floodplains, on solute retention have not been rigorously compared. �We used MIKE SHE to model hydraulics and solute transport in the channel, inset floodplain, and hyporheic zone of a 2nd order stream. �We varied hydraulic conditions (winter baseflow, summer baseflow, and storm flow), geology (hydraulic conductivity), and stream restoration design parameters (inset floodplain length, and presence of in stream structures). �In stream structures induced hyporheic exchange during summer baseflow with a low groundwater table (~20% of the year), while floodplains only retained solutes during storm flow conditions (~1% of the year). �Flow through the hyporheic zone increased linearly with hydraulic conductivity, while residence times decreased linearly. �Flow through inset floodplains and residence times in both the channel and floodplains increased non linearly with the fraction of bank with floodplains installed. �The fraction of stream flow that entered inset floodplains was one to three orders of magnitude higher than that through the hyporheic zone, while the residence time and mass storage in the hyporheic zone was one to five orders of magnitude larger than that in floodplain segments. �Our model results suggest that in stream structures and inset floodplains are complementary practices. / Master of Science

Increased reactive nitrogen from human activities negatively affects surface water (SW) quality. The hyporheic zone, where SW and groundwater interact, possesses unique biogeochemical conditions that can attenuate contaminants (e.g., denitrification), including mixing-dependent reactions that require components from both water sources. Previous research has explored mixing-dependent denitrification in the hyporheic zone but did not address the effects of varying SW depth as would occur from storms, tides, dam operation, and varying seasons. We simulated steady and unsteady hyporheic flow and transport through a riverbed dune using MODFLOW and SEAM3D, and varied SW depth, degree of sediment heterogeneity, amplitude and frequency of sinusoidal fluctuations, among others to determine these effects. We found that increasing steady state surface water depth from 0.1 to 1.0 m increased non-mixing dependent aerobic respiration by 270% and mixing-dependent denitrification by 78% in homogeneous sediment. Heterogeneous hydraulic conductivity fields yielded similar results, with increases in consumption due to variation in correlation length and variance of less than 5%. Daily SW fluctuation, including variation of amplitude, period, and sinusoidal versus instantaneous changes had significantly less impact than longer-term trends in SW depth. There is potential for the hyporheic zone to attenuate NO3- in upwelling groundwater plumes. Restoration efforts may be able to maximize the potential for mixing-dependent reactions in the hyporheic zone by increasing residence times. / Master of Science / Increased nitrogen in runoff from human activities negatively affects surface water quality. The hyporheic zone is where surface water and groundwater interact, and the mixing between the waters can help to this nitrogen to undergo reaction (denitrification), potentially stopping the contaminant from spreading. Previous research has explored this idea, but has not addressed the impact of varying surface water depth, as would realistically occur due to storms, tides, dam operation, and varying seasons. We simulated both constant and fluctuating surface water conditions on a riverbed dune to see the effects on hyporheic flow and denitrification. Test variables included the surface water depth, the degree of sediment heterogeneity, the amplitude and frequency of surface water fluctuations. We found that increasing the steady-state surface water depth had the most dramatic increase on the amount of reaction undergone. This trend was also seen in heterogeneous sediment. Any daily-scale surface water fluctuations, including runs that varied the amplitude, period, and sinusoidal vs instantaneous changes in surface water depth, had significantly less impact than longer-term trends in surface water depth.