This study investigates the impacts of climate change on the hydrology and soil thermal regime of ten sub-arctic watersheds (northern Manitoba, Canada) using the Variable Infiltration Capacity (VIC) model. We utilize statistically downscaled and bias-corrected forcing datasets based on 17 general circulation model (GCM) - representative concentration pathways (RCP) scenarios from phase 5 of the Coupled Model Intercomparison Project (CMIP5) to run the VIC model for three 30-year periods: a historical baseline (1981–2010), and future projections (2021–2050: 2030s and 2041–2070: 2050s), under representative concentration pathways (RCPs) 4.5 and 8.5. The CMIP5 Multi-Model Ensemble (MME) mean-based VIC simulations indicate a 15–20% increase and 10% decrease in the projected annual precipitation and snowfall, respectively over the southern portion of the basin and >20% rainfall increase over the higher latitudes of the domain by the 2050s. Snow accumulation is projected to decline across all sub-basins, particularly in the lower latitudes. Projected uncertainties in major water balance components (i.e., evapotranspiration, surface runoff, and streamflow) are more substantial in the wetland and lake-dominated Grass and Gunisao watersheds than their eight counterparts. Future warming increases soil temperatures >2.5°C by the 2050s, resulting in 40–50% more baseflow. Further analyses of soil temperature trends at three different depths show the most pronounced warming in the top soil layer (1.6°C 30-year-1 in the 2050s), whereas baseflow increases substantially by 19.7% and 46.3% during the 2030s and 2050s, respectively. These results provide crucial information on the potential future impacts of warming soil temperatures on the hydrology of sub-arctic watersheds in north-central Canada and similar hydro-climatic regimes.

Soil–rock mixtures are widely encountered in geotechnical engineering projects. The instability and failure mechanism of grap-graded soil–rock mixtures under rainfall conditions has always been the focus of geological disaster research. To deeply explore the mechanism of seepage deformation of soil–rock mixtures, an indoor physical permeability test that considers soil–rock mixtures with different fine contents was conducted, and a particle-scale numerical simulation test of the permeability evolution was carried out using the coupling model of PFC3D and ABAQUS. The test results showed that the spatial distribution of fine particle loss along the height direction could be divided into three areas: top loss, middle uniform, and bottom loss area. The “island” effect of coarse particles, which is caused by excessive fine content and makes the fine particles bear more load, was eliminated with the loss of fine particles. In this preset working condition of coarse and fine particle diameters, setting FC to 35% may be the best way to fill the voids between the coarse particles. Particle migration leads to a change in the load-bearing skeleton structure, thereby causing seepage deformation. Therefore, the particle-scale numerical test method can better reproduce the seepage deformation process of grap-graded soil–rock mixtures.

Groundwater plays a crucial role in sustaining global food security but is being over-exploited in many basins of the world. Despite its importance and finite availability, local-scale monitoring of groundwater withdrawals required for sustainable water management practices is not carried out in most countries, including the United States. In this study, we combine publicly available datasets into a machine learning framework for estimating groundwater withdrawals over the state of Arizona. Here we include evapotranspiration, precipitation, crop coefficients, land use, well density, and watershed stress metrics for our predictions. We employ random forests to predict groundwater withdrawals from 2002-2020 at a 2 km spatial resolution using in-situ groundwater withdrawal data available for Arizona Active Management Areas (AMA) and Irrigation Non-Expansion Areas (INA) from 2002-2009 for training and 2010-2020 for validating the model respectively. The results show high training (R2≈ 0.86) and good testing (R2≈ 0.69) scores with normalized mean absolute error (NMAE) ≈ 0.64 and normalized root mean square error (NRMSE) ≈ 2.36 for the AMA/INA region. Using this method, we spatially extrapolate the existing groundwater withdrawal estimates to the entire state and observe the co-occurrence of both groundwater withdrawals and land subsidence in South-Central and Southern Arizona. Our model predicts groundwater withdrawals in regions where production wells are present on agricultural lands and subsidence is observed from Interferometric Synthetic Aperture Radar (InSAR), but withdrawals are not monitored. By performing a comparative analysis over these regions using the predicted groundwater withdrawals and InSAR-based land subsidence estimates, we observe a varying degree of subsidence for similar volumes of withdrawals in different basins. The performance of our model on validation datasets and its favorable comparison with independent water use proxies such as InSAR demonstrate the effectiveness and extensibility of our combined remote sensing and machine learning-based approach.

The climate of the Eurasia inland basin (EIB) is characterized by limited precipitation and high potential evapotranspiration; as such, water storage in the EIB is vulnerable to global warming and human activities. There is increasing evidence pointing to varying trends in water storage across different regions; however, a consistent conclusion on the main attributes of these trends is lacking. Based on the hydrological budget in a closed inland basin, the main attributes of changes in actual evapotranspiration (AET) and terrestrial water storage (TWS) were identified for the EIB and each closed basin. In the EIB and most of its closed basins, the TWS and AET showed significantly decreasing and non-significantly increasing trends, respectively. The primary cause underpinning the significantly decreasing TWS in the EIB was increasing AET. Approximately 70% of the increase in AET has been supplied by increased irrigation diversions and glacial melt runoff. At the basin scale, similar to the EIB, changes in AET were the predominant factor driving changes in TWS in most basins; the exception to this was the Balkhash Lake basin (BLB), Iran inland river basin (IIRB), Qaidam basin (QB), and Turgay River basin (TuRB). In these basins, changes in precipitation largely contributed to the TWS changes. The AET consumption of other water resources was the main factor contributing to AET changes in seven of 16 basins, including the Aral Sea, Caspian Sea, Junggar, Monglia Plateau, Qiangtang Plateau, and Tarim River basins. The increase in precipitation contributed more than 60% of increasing AET in four of 16 basins, particularly in the Helmand River basin and QB (>90%). Changes in precipitation and consumption by other water supply sources contributed to approximately half of the AET changes in the other five basins, including the Inner Mongolia Plateau, Issyk-Kul Sarysu, BLB, IIRB, and TuRB basins.

Quantifying the uncertainty linked to the degree to which the spatio-temporal variability of the catchment descriptors (CDs), and consequently calibration parameters (CPs), represented in the distributed hydrology models and its impacts on the simulation of flooding events is the main objective of this paper. Here, we introduce a methodology based on ensemble approach principles to characterize the uncertainties of spatio-temporal variations. We use two distributed hydrological models (WaSiM and Hydrotel) and six catchments with different sizes and characteristics, located in southern Quebec, to address this objective. We calibrate the models across four spatial (100, 250, 500, 1000 $m^2$) and two temporal (3 hours and 24 hours) resolutions. Afterwards, all combinations of CDs-CPs pairs are fed to the hydrological models to create an ensemble of simulations for characterizing the uncertainty related to the spatial resolution of the modeling, for each catchment. The catchments are further grouped into large ($>1000 km^2$), medium (between 500 and 1000 $km^2$) and small ($<500km^2$) to examine multiple hypotheses. The ensemble approach shows a significant degree of uncertainty (over $100\%$ error for estimation of extreme streamflow) linked to the spatial discretization of the modeling. Regarding the role of catchment descriptors, results show that first, there is no meaningful link between the uncertainty of the spatial discretization and catchment size, as spatio-temporal discretization uncertainty can be seen across different catchment sizes. Second, the temporal scale plays only a minor role in determining the uncertainty related to spatial discretization. Third, the more physically representative a model is, the more sensitive it is to changes in spatial resolution. Finally, the uncertainty related to model parameters is dominant larger than that of catchment descriptors for most of the catchments. Yet, there are exceptions for which a change in spatio-temporal resolution can alter the distribution of state and flux variables, change the hydrologic response of the catchments, and cause large uncertainties.

Flow regimes are critical for determining physical and biological processes in rivers, and their classification and regionalization traditionally seeks to link patterns of flow to physiographic, climate and other information. There are many approaches to, and rationales for, catchment classification, with those focused on streamflow often seeking to relate a particular response characteristic to a physical property or climatic driver. Rationales include such topics as Prediction in Ungauged Basins (PUB), helping with experimental approaches, and providing guidance for model selection in poorly understood hydrological systems. While scale and time are important considerations for classification, the Annual Daily Hydrograph (ADH) is a first-order easily visualized integrated expression of catchment function, and over many years is a distinct hydrological signature. In this study, we use t-SNE, a state-of-the-art technique of dimensionality reduction, to classify 17110 ADHs for 304 reference catchments in mountainous Western North America. t-SNE is chosen over other conventional methods of dimensionality reduction (e.g. PCA) as it presents greater separability of ADHs, which are projected on a 2D map where the similarities are evaluated according to their map distance. We then utilize a Deep Learning encoder to upgrade the non-parametric t-SNE to a parametric approach, enhancing its capability to address ‘unseen’ samples. Results showed that t-SNE was an effective classifier as it successfully clustered ADHs of similar flow regimes on the 2D map. In addition, many compact clusters on the 2D map in the coastal Pacific Northwest suggest information redundancy in the local hydrometric network. The t-SNE map provides an intuitive way to visualize the similarity of high-dimensional data of ADHs, groups catchments with like characteristics, and avoids the reliance on subjective hydrometric indicators.

Snow density is one of the essential properties to describe snowpack characteristics. To obtain the spatial variability of snow density and estimate it accurately in different periods of snow season still remain as challenges, particularly in the mountains. This study analyzed the spatial variability of snow density with in-situ measurements in three different periods (i.e. accumulation, stable, melt period) of snow seasons 2017/2018 and 2018/2019 in the middle Tianshan Mountains, China. The performance of multiple linear regression model (MLR) and three machine learning models (i.e. Random Forest (RF), Extreme Gradient Boosting (XGB) and Light Gradient Boosting Machine (LGBM)) to simulate snow density has been evaluated. It was found that the snow density in melt period (0.27 g cm-3) was generally greater than that in stable (0.20 g cm-3) and accumulation period (0.18 g cm-3), and the spatial variability of snow density in melt period was slightly smaller than that in other two periods. The snow density in mountainous areas was generally higher than that in plain or valley areas, and snow density increased significantly (p < 0.05) with elevation in the accumulation and stable periods. Besides elevation, latitude and ground surface temperature also had critical impacts on the spatial variability of snow density in the middle Tianshan Mountains, China. In this work, the machine learning model, especially RF model, performed better than MLR on snow density simulation in three periods. Compared with MLR, the determination coefficients of RF promoted to 0.61, 0.51 and 0.58 from 0.50, 0.1 and 0.52 in accumulation period, stable period and melt period respectively. This study provide a more accurate snow density simulation method for estimating regional snow mass and snow water equivalent, which allows us to achieve a better understanding of regional snow resources.

The hydroclimatology of Northern South America responds to strongly-coupled dynamics of oceanic and terrestrial surface-atmosphere exchange, as moisture evaporated from these sources interact to produce continental rainfall. However, the relative contributions of these two source types through the annual cycle have been described only in modeling studies, with no observational tools used to corroborate these predictions. The use of isotopic techniques to study moisture sources has been common in assessing changes in the water cycle and in climate dynamics, as isotopes allow tracking the connection between evaporation, transpiration, and precipitation, as well as the influence of large scale hydroclimatic phenomena, such as the seasonal Inter Tropical Convergence Zone migration. We characterize the isotopic composition of moisture sources becoming precipitation in the Andes and Caribbean regions of Colombia, using stable isotopes data (δ18O, δ2H) from the Global Network of Isotopes in Precipitation (1971-2016) and contrasting it with moisture trajectory tracking from the FLEXPART model, using input from ERA-Interim reanalysis to compute the relative contribution of oceanic and terrestrial sources through the annual cycle. Our results indicate that most precipitation in the region comes from terrestrial sources including recycling (>30 % for all months), Orinoco (up to 28 % monthly for April), and the northern Amazon (up to 17 % monthly for June, July, and August); followed by oceanic sources including the Tropical South Pacific (up to 30 % monthly in October, November, December) and Tropical North Atlantic (up to 30 % monthly for January). These outcomes highlight the utility of combining stable isotopes in precipitation and modeling techniques to discriminate terrestrial and oceanic sources of precipitation. Further, our results highlight the need to assess the hydrological consequences of land cover change in South America, particularly in a country like Colombia where water, food and energy security all depend directly on precipitation. .

Aaron Smith1, Doerthe Tetzlaff1,2,3, Marco Maneta4,Chris Soulsby3,21IGB Leibniz Institute of Freshwater Ecology and Inland Fisheries Berlin, Berlin, Germany2Humboldt University Berlin, Berlin, Germany3Northern Rivers Institute, School of Geosciences, University of Aberdeen, UK4Department of Geosciences, University of Montana, Missoula, Montana, USACorrespondence to: Aaron Smith ([email protected])

Water quality dynamics depend strongly on hydrologic flow paths and transit time within catchments. In this paper we use a travel time tracking method to simulate stream salinity (as measured by electrical conductivity) in the Duck River catchment, NW Tasmania, Australia. The approach couples the StorAge transit time modelling approach with two different approaches to model electrical conductivity. The first assumes the catchment has a cyclic salt balance (rainfall source, stream flow sink) that is in dynamic equilibrium and evapoconcentration of salt is the only process changing concentration. The second assumes that the salinity of water in catchment storages is a function of water age in those stores, without explicitly simulating salt mass balance processes. The paper compares these alternate approaches in terms of salinity simulation, simulated stream water age distributions, and simulated storage age distributions. Both salinity simulation approaches reproduce stream salinity with high fidelity under calibration and perform well under validation. The simulations using the age-related solute concentration approach produce less biased results and thus high model efficiencies for validation periods. This approach also produces more consistent model parameter estimates between periods. There are systematic differences in the resultant age distributions between models, particularly for the solute balance based simulations where parameters (catchment storage size) changed more between calibration periods. The effect of time varying versus static storage selection functions are compared, with clear evidence that time varying storage selection functions with parameters linked to catchment conditions (flow) are essential for adequate simulation of event concentration dynamics.

We combine physics-based groundwater reactive transport modeling with machine learning techniques to quantify hydrogeologic model and solute transport predictive uncertainties. We train an artificial neural network (ANN) on a dataset of groundwater hydraulic heads and 3H concentrations generated using a high-fidelity groundwater reactive transport model. Using the trained ANN as a surrogate model to reproduce the input-output response of the high-fidelity reactive transport model, we quantify the posterior distributions of hydrogeologic parameters and hydraulic forcing conditions using Markov-chain Monte Carlo (MCMC) calibration against field observations of groundwater hydraulic heads and 3H concentrations. We demonstrate the methodology with a model application that predicts Chlorofluorocarbon-12 (CFC-12) solute transport at a contaminated site in Wyoming, USA. Our results show that including 3H observations in the calibration dataset reduced the uncertainty in the estimated permeability field and infiltration rates, compared to calibration against hydraulic heads alone. However, predictive uncertainty quantification shows that CFC-12 transport predictions conditioned to the parameter posterior distributions cannot reproduce the field measurements. We found that calibrating the model to hydraulic head and 3H observations results in groundwater mean ages that are too large to explain the observed CFC-12 concentrations. The coupling of the physics-based reactive transport model with the machine learning surrogate model allows us to efficiently quantify model parameter and predictive uncertainties, which is typically computationally intractable using reactive transport models alone.

Midwestern cities require forecasts of surface nitrate loads to bring additional treatment processes online or activate alternative water supplies. Concurrently, networks of nitrate monitoring stations are being deployed in river basins, co-locating water quality observations with established stream gauges. Here, we construct a synthetic data set of stream discharge and nitrate for the Wabash River Basin - one of the U.S.’s most nutrient polluted basins - using the established Agro-IBIS model. While real-world observations are limited in space and time, particularly for nitrate, the synthetic data set allows for sufficiently long periods to train machine learning models and assess their performance. Using the synthetic data, we established baseline 1-day forecasts for surface water nitrate at 12 cities in the basin using support vector machine regression (SVMR; RMSE 0.48-3.3 mg/L). Next, we used the SVMRs to evaluate the improvement in forecast performance associated with deployment of additional sensors. Synthetic data enable us to quantitatively assess the expected value of an additional nitrate sensor being deployed, which is, of course, not possible if we are limited to the present observational network. We identified the optimal sensor placement to improve forecasts at each city, and the relative value of sensors at all possible locations. Finally, we assessed the co-benefit realized by other cities when a sensor is deployed to optimize a forecast at one city, finding significant positive externalities in all cases. Ultimately, our study explores the potential for AI to make short-term predictions and provide an unbiased assessment of the marginal benefit and co-benefits to an expanded sensor network. While we use water quantity in the Wabash River Basin as a case study, this approach could be readily applied to any problem where the future value of sensors and network design are being evaluated.

With global warming and rapid urbanization, urban agglomerations over the Loess Plateau (LP) are suffering from various urban disasters. Urbanization has aggravated the decreasing trends of extreme precipitation in Taiyuan and Xi’an urban agglomerations (UAs) and enhanced the increasing trends of extreme precipitation in Luoyang, Hohhot and Xining UAs during 1979–2018. Meanwhile, the number of light rain days decreases in almost all the cities, indicating the sensitivity of light rain days to urbanization. The climate change is a primary contributor to the change of urban precipitation during 1980–2000. However, the urbanization contribution has been increasing gradually since 2000, and the urbanization further amplifies the trend of extreme precipitation caused by the climate change. In terms of the physical mechanisms, the rapid increasing surface temperature and aerosol particles are closely related to the urban precipitation. Our findings provide a systematic understanding of the urbanization effects on the extreme precipitation over the LP and may play an important role in the mitigation of urban disasters.

The hydraulic properties of coastal aquifer systems are relevant to various hydrogeological, hydro-ecological and engineering problems. This study presents an analytical solution for predicting groundwater head fluctuations induced by dual-tide in multilayered island aquifer systems, consisting of an unconfined aquifer on the top and any number of leaky aquifers below. The solution was derived via the methods of matrix differential calculus and separation of variables. It is more general than any existing analytical solutions for the tidal pressure propagation since the new solution can consider multilayered aquifer systems along with the effects of leakage and aquifer length. Using this solution, we illustrated potential errors that may occur due to neglecting one or more vital factors affecting groundwater fluctuations. Besides, we articulated the groundwater response to the dual-tide in complex coastal aquifers. Considering that some thin semipermeable layers may be ignored in practical field investigation, we also demonstrated the effects due to simplification of aquifer layers. The results showed that with the increase in the number of overlapped leaky layers, the tidal propagation in the bottom part of multilayered aquifer system approaches that in a single confined aquifer with the same transmissivity and storage.

This paper presents a taxonomy (hierarchical organization) of hydrological processes; specifically, runoff generation processes in natural watersheds. Over 120 process names were extracted from a literature review of papers describing experimental watersheds, perceptual models, and runoff processes in a range of hydro-climatic environments. Processes were arranged into a hierarchical structure, and presented as a spreadsheet and interactive diagram. For each process, additional information was provided: a list of alternative names for the same process, a classification into hydrological function (e.g. partitioning, flux, storage, release) and a unique identifier similar to a hashtag. The taxonomy provides a method to label and search hydrological knowledge, thereby facilitating synthesis and comparison of processes across watersheds.

Groundwater recharge in highly-fractured volcanic aquifers remains poorly understood in the humid tropics, whereby rapid demographic growth and unregulated land use change are resulting in extensive surface water pollution and a large dependency on groundwater extraction. Here we present a multi-tracer approach including δ18O-δ2H, 3H/3He, and noble gases within the most prominent multi-aquifer system of central Costa Rica, with the objective to assess dominant groundwater recharge characteristics and age distributions. We sampled wells and large springs across an elevation gradient from 868 to 2,421 m asl. Our results suggest relatively young apparent ages ranging from 0.0±3.2 up to 76.6±9.9 years. Helium isotopes R/RA (0.99 to 5.4) indicate a dominant signal from the upper mantle across the aquifer. Potential recharge elevations ranged from ~1,400 to 2,650 m asl, with recharge temperatures varying from ~11°C to 19°C with a mean value of 14.5±1.9°C. Recharge estimates ranged from 129±78 to 1,605±196 mm/yr with a mean value of 642±117 mm/yr, representing 20.1±4.0% of the total mean annual rainfall as effective recharge. The shallow unconfined aquifer is characterised by young and rapidly infiltrating waters, whereas the deeper aquifer units have relatively older waters. These results are intended to guide the delineation and mapping of critical recharge areas in mountain headwaters to enhance water security and sustainability in the most important headwater dependent systems of Costa Rica.

Typical applications of process- or physically-based models aim to gain a better process understanding or provide the basis for a decision-making process. To adequately represent the physical system, models should include all essential processes. However, model errors can still occur. Other than large systematic observation errors, simplified, misrepresented, inadequately parametrized or missing processes are potential sources of errors. This study presents a set of methods and a proposed workflow for analyzing errors of process-based models as a basis for relating them to process representations. The evaluated approach consists of three steps: (i) training a machine learning (ml) error-model using the input data of the process-based model and other available variables, (ii) estimation of local explanations (i.e., contributions of each variable to a individual prediction) for each predicted model error using SHapley Additive exPlanations (SHAP) in combination with principal component analysis, (iii) clustering of SHAP values of all predicted errors to derive groups with similar error generation characteristics. By analyzing these groups of different error-variable association, hypotheses on error generation and corresponding processes can be formulated. That can ultimately lead to improvements in process understanding and prediction. The approach is applied to a process-based stream water temperature model HFLUX in a case study for modelling an alpine stream in the Canadian Rocky Mountains. By using available meteorological and hydrological variables as inputs, the applied ml model is able to predict model residuals. Clustering of SHAP values results in three distinct error groups that are mainly related to shading and vegetation emitted longwave radiation. Model errors are rarely random and often contain valuable information. Assessing model error associations is ultimately a way of enhancing trust in implemented processes and of providing information on potential areas of improvement to the model.

The magnitude and spatial heterogeneity of snow deposition are difficult to model in mountainous terrain. Here, we investigated how snow patterns from a 32-year (1985 – 2016) snow reanalysis in the Tuolumne, Kings, and Sagehen Creek, California Sierra Nevada watersheds could be used to improve simulations of winter snow deposition. Remotely-sensed fractional snow-covered area (fSCA) from dates following peak-snowpack timing were used to identify dates from different years with similar fSCA, which indicated similar snow accumulation and depletion patterns. Historic snow accumulation patterns were then used to 1) relate snow accumulation observed by snow pillows to watershed-scale estimates of mean snowfall, and 2) estimate 90 m snow deposition. Finally, snow deposition fields were used to force snow simulations, the accuracy of which were evaluated versus airborne lidar snow depth observations. Except for water-year 2015, which had the shallowest snow estimated in the Sierra Nevada, normalized snow accumulation and depletion patterns identified from historic dates with spatially correlated fractional snow-covered area agreed on average, with absolute differences of less than 10%. Watershed-scale mean winter snowfall inferred from the relationship between historic snow accumulation patterns and snow pillow observations had a ±13% interquartile range of biases between 1985 and 2016. Finally, simulations using 1) historic snow accumulation patterns, and 2) snow accumulation observed from snow pillows, had snow depth coefficients of correlations and mean absolute errors that improved by 70% and 27%, respectively, as compared to simulations using a more common forcing dataset and downscaling technique. This work demonstrates the real-time benefits of satellite-era snow reanalyses in mountainous regions with uncertain snowfall magnitude and spatial heterogeneity.