Modeling groundwater redox conditions at national scale through integration of sediment color and water chemistry in a machine learning framework.

Redox conditions play a crucial role in determining the fate of many contaminants in groundwater, impacting ecosystem services vital for both the aquatic environment and human water supply. Geospatial machine learning has previously successfully modelled large-scale redox conditions. This study is the first to consolidate the complementary information provided by sediment color and water chemistry to enhance our understanding of redox conditions in Denmark. In the first step, the depth to the first redox interface is modelled using sediment color from 27,042 boreholes. In the second step, the depth of the first redox interface is compared against water chemistry data at 22,198 wells to classify redox complexity. The absence of nitrate containing water below the first redox interface is referred to as continuous redox conditions. In contrast, discontinuous redox conditions are identified by the presence of nitrate below the first redox interface. Both models are built using 20 covariate maps, encompassing diverse hydrologically relevant information. The first redox interface is modelled with a mean error of 0.0 m and a root-mean-squared error of 8.0 m. The redox complexity model attains an accuracy of 69.8 %. Results indicate a mean depth to the first redox interface of 8.6 m and a standard deviation of 6.5 m. 60 % of Denmark is classified as discontinuous, indicating complex redox conditions, predominantly collocated in clay rich glacial landscapes. Both maps, i.e., first redox interface and redox complexity are largely driven by the water table and hydrogeology. The developed maps contribute to our understanding of subsurface redox processes, supporting national-scale land-use and water management.

Evaluating the impact of muon-induced cosmogenic 39Ar and 37Ar underground production on groundwater dating with field observations and numerical modeling.

Groundwater dating by radioactive cosmogenic tracers such as 39Ar relies on the decay rate from a known initial atmospheric activity (100%modern). Thereby, it is assumed that cosmogenic 39Ar production in the subsurface is negligible at depths below the water table and that contributions from natural rock radioactivity are minor or missing. Here we present 39Ar data from aquifers located in quaternary glacial sediments and tertiary limestones in Denmark, which unequivocally demonstrate that cosmogenic production can induce considerable age biases. 39Ar values larger than 100%modern are observed at relatively shallow groundwater depths in non-radiogenic rocks. These activities are compared to calculations based on previously assessed depth-dependent production rates in rocks and realistic estimates of the emanated fractions to the water phase. The water residence time distribution with depth, which was determined by numerical flow modeling and particle tracking, underpinned the significance of muon-induced 39Ar production. The short-lived isotope 37Ar is produced by similar processes as 39Ar and demonstrated its usefulness as an indicator of local underground production in an aquifer. The significance of cosmogenic underground production in other possible recharge scenarios was then assessed by explicitly simulating the radioargon accumulation and decay in a 2D synthetical numerical model. These simulations demonstrated that underground production is negligible when the water infiltrates freely in a porous aquifer. However, in the presence of a confining layer impeding the infiltration at shallow depths (<30 m), as is the case in our study site in Denmark for instance, over-modern 39Ar activities (>100%modern) may occur. The age concluded from the dissolved activities is then possibly biased towards young values. Special attention should thus be paid to the recharge rates when using 39Ar for dating groundwater. 37Ar activities provide complementary information about the strength and mechanisms of underground production.

Comparison and evaluation of model structures for the simulation of pollution fluxes in a tile-drained river basin.

The European Union Water Framework Directive requires an integrated pollution prevention plan at the river basin level. Hydrological river basin modeling tools are therefore promising tools to support the quantification of pollution originating from different sources. A limited number of studies have reported on the use of these models to predict pollution fluxes in tile-drained basins. This study focused on evaluating different modeling tools and modeling concepts to quantify the flow and nitrate fluxes in the Odense River basin using DAISY-MIKE SHE (DMS) and the Soil and Water Assessment Tool (SWAT). The results show that SWAT accurately predicted flow for daily and monthly time steps, whereas simulation of nitrate fluxes were more accurate at a monthly time step. In comparison to the DMS model, which takes into account the uncertainty of soil hydraulic and slurry parameters, SWAT results for flow and nitrate fit well within the range of DMS simulated values in high-flow periods but were slightly lower in low-flow periods. Despite the similarities of simulated flow and nitrate fluxes at the basin outlet, the two models predicted very different separations into flow components (overland flow, tile drainage, and groundwater flow) as well as nitrate fluxes from flow components. It was concluded that the assessment on which the model provides a better representation of the reality in terms of flow paths should not only be based on standard statistical metrics for the entire river basin but also needs to consider additional data, field experiments, and opinions of field experts.

A distributed modelling system for simulation of monthly runoff and nitrogen sources, loads and sinks for ungauged catchments in Denmark.

We validated an existing physically based 3D MIKE SHE groundwater resource model (DK-model) at 175 Danish gauging stations covering different catchment sizes in order to calculate monthly water runoff in the 50% ungauged part of Denmark. Model performance was in most cases good (61% of gauging stations had a Nash-Sutcliffe (NS) coefficient >0.60) but nevertheless showed a large seasonal and georegion specific bias. Therefore, bias correction factors had to be developed before applying the DK-model simulations of runoff in the ungauged areas. Simulated monthly runoff from ungauged areas and the measured monthly runoff from 178 gauging stations were distributed to 2663 smaller Hydrological Units (ca. 15 km(2)) and linked with a new empirical model for flow-weighted monthly total nitrogen (TN) concentrations (R(2) = 0.43; P < 0.0001) developed based on 20 years of observations (1990-2009) in 83 small catchments for calculation of monthly gross diffuse TN-loads from HU's. Nitrogen retention was calculated in streams, lakes and wetlands utilising both lake specific models and rate coefficients to calculate N retention in surface water bodies. The whole model complex was linked in the DK-QN concept for simulation of monthly TN losses from point sources and diffuse sources, TN retention and resulting loadings to Danish coastal waters. The DK-QN model was validated in 118 gauged catchments and the model simulations had for >25% of the observations of monthly discharge weighted TN concentrations a NS larger than 0.26. Catchment specific monthly TN-loadings were modelled with a higher performance as 50% of the catchments had a NS greater than 0.75. The model concept allows calculation of N retention in streams, lakes and wetlands and the average annual model calculated N retention amounted to 21% of the modelled gross riverine TN loadings. The average annual gross TN loading to surface freshwater in Denmark derived from diffuse sources amounted to 97 000 tonnes N (91% of gross TN loadings) which is 54% of the total estimated N-leaching from the root zone on the Danish land area (212 000 tonnes N) during the period 1990-2009.