An efficient approach for constructing a 3D model of soil samples using the transparent soil technique has been developed. The workflow of this 3D modelling method encompasses experimental configuration, image acquisition, data processing, and model generation. The core step involves surface generation of the samples, which is achieved through a surface-fitting algorithm based on Non-Uniform Rational B-Splines (NURBS). The effectiveness of this 3D reconstruction and visualization methodology was validated via an application in a flow-induced soil erosion test.

This study examined long-term (2008–2021) water flux data in three different age (83-, 48- and 20-year-old) plantation forests in the Great Lakes region in Canada. The mean annual evapotranspiration (ET) values were 465 ± 41, 466 ± 32 and 403 ± 21 mm year⁻¹ in the 83-, 48- and 20-year-old stands, respectively, with the highest ET observed in the 83- and 48-year-old stands. The 20-year-old forest had the highest mean WUE of 4 g C kg H₂O⁻¹ over the study period as compared to a mean WUE of 3.6 g C kg H₂O⁻¹ in the two older stands, suggesting that the youngest forest was more efficient in terms of carbon sequestration and water use. The younger stand also exhibited greater sensitivity to water stress and higher drought coupling.

Water stable isotopes from the Willamette River, Oregon USA indicate a decrease in water originating from the high-elevation snow zone (blue bars) in late summer low streamflow over 13 years. Changes in peak snow water equivalent explain some of the decline, but a temporal declining trend is still present.

This figure illustrates the relationship between urban vegetation (leaf area index, LAI) and impervious surface fraction (f_imp) across Asian cities in different climate zones. Higher LAI is observed in tropical rainforest zones (Kuala Lumpur, Singapore), while urbanisation intensifies in humid subtropical zones (Shanghai). Variations in fimp and LAI highlight the impact of urbanisation on vegetation and hydrological dynamics, emphasising distinct urban climate challenges.

The past two decades have been dominated by negative anomalies in annual rainfall, causing a general lowering of water tables and persistent storage deficits. Spatio-temporal patterns of recharge are also strongly influenced by vegetation cover, with coniferous forests having high evapotranspiration losses that inhibit groundwater recharge. Most stream flow is generated by shallow groundwater in the catchment headwaters, which is relatively young.

Recent studies have successfully applied machine learning models to predict stream solute concentrations from high-frequency sensor data. However, when applied to watersheds outside the training set, model performance ranges from biased to entirely inaccurate, even with seemingly similar watersheds. Our results demonstrate that machine learning models reliably predict solute concentrations only where trained, as differences in solute concentration patterns and sensor-solute relationships limit their broader applicability.

Future projections for the Doam Watershed show significant worsening of drought severity, particularly, under the SSP585 scenario, highlighting the urgent need for proactive water management. Statistical analysis with the Mann–Kendall test confirms increasing drought frequency using SPI, SPEI and PDSI indices. Critical drought periods are expected around the 2070s and 2080s, as demonstrated by heat maps, while copula models suggest a strong relationship between drought duration and severity, providing a robust framework for drought risk management.

Regular droughts occurred most in rainfed agricultural areas (central/southeastern CA) for over 50 pentads, whilst flash droughts dominated irrigated areas (central, southeastern, northern CA) for 2–5 pentads. NDVIz and SSMI showed a significant 9–12 month lag correlation, both declining with drought, especially flash droughts. Rainfed agriculture was highly sensitive to soil moisture loss, whilst irrigation mitigated this effect, improving water use efficiency and shifting the correlation towards positive.

This study develops a novel analytical model for leaky aquifers, addressing depth-dependent decreases in aquitard specific storage (S_s) and hydraulic conductivity (K), along with variable pumping rates. Results reveal that S_s and K decay dominate drawdown, depletion, and leakage fraction patterns, while pumping variability creates local drawdown and depletion maxima. The findings emphasise prioritising depth-dependent parameters in groundwater management, particularly in glacial regions with significant vertical heterogeneity, to optimise resource use and mitigate hydrological risks.

Year-long monitoring of seawater and cave groundwater reveals strong tidal influence, with a 20–30 min lag indicating high hydraulic connectivity in this karst aquifer. Seasonal variations, largest in autumn, affected seawater mixing, as seen in dissolved oxygen levels, while stable pH highlighted the system's buffering capacity. These findings enhance understanding of groundwater-seawater interactions in hydrological processes in coastal karst systems under changing environmental conditions and confirm cave's significance for sea-level reconstructions.

We assessed the extension of the soil column in gridded catchment rainfall–runoff modelling as an approach to sufficiently capture the shallow groundwater system and its interaction with the rest of the catchment hydrological processes. We implemented and assessed different vertical profiles of saturated hydraulic conductivity that exceed the typical soil dataset and measurement maximum depth (2 m). Results demonstrate the influence of the involved soil parameters on discharge and groundwater dynamics prediction performance and catchment water balance partitioning.

Model results showed that a stormflow response was more likely in the high snow zone (HSZ) than in the low snow zone (SZ)L, likely due to the higher soil moisture content. Rainfall intensity was a good predictor of the magnitude of the stormflow response, whereas burn category influenced the lag to peak times. The effect that year post-fire had on the stormflow response was ambiguous. Results indicate that widespread overland flow only occurred at the severely burned LSZ site.

Based on meteorological observations and precipitation isotopes collected between January 2019 and December 2022 in Poyang Lake, the largest freshwater lake in China, we confirmed significant sub-cloud evaporation during the precipitation processes in humid climate regions, revealed the consistent temporal variations of precipitation isotopes monitored at the surface and estimated at the cloud base by modifying the sub-cloud evaporation effect, and identified various meteorological factors affecting the precipitation isotopic changes.

Historical trends from 1991 through 2020 were analysed from high elevation snow stations along a 2500 km corridor of the North American Rocky Mountains. This revealed declining maximum and April 1 snowpacks, and earlier snowpack peaks. A latitudinal pattern was observed, with slight changes in northern regions (blue) and increasing decline rates southward (pink and red). This latitudinal pattern contrasts with some climate warming patterns, which increase northwards (MapChart graphic).

The quantification of water quality in rivers in time and space is the basis for effective and impactful management. To investigate the patterns of spatial and temporal variability of water quality, we analyse stream water concentrations of anthropogenic, geogenic and biogenic constituents in 1215 German catchments. Each constituent showed a distinct pattern of spatial and temporal variability that was explainable by landscape characteristics and underlying processes.

This study introduces a spectral-based method to estimate hydraulic flux in the vadose zone by coupling thermal diffusivity with observed soil temperature spectra. Unlike traditional models that rely on fixed flux or diffusivity, this approach uses time-frequency analysis to capture subsurface variability. Two boundary conditions are tested, with results showing high agreement under constrained outlet heat flux scenarios. The method provides a practical framework for characterising subsurface heat–water interactions and enhancing understanding of unsaturated zone hydrodynamics.

The subsurface lateral flow on the downslope (near the Benggang wall) is associated with the collapse of the Benggang wall, which can be an indicator of the collapse events.

Effects of the aquatic flexible vegetation with varying morphologies on wave-sediment motion were studied in the field. The presence of vegetation caused a dual reduction of the near-bed wave orbital velocity, which increased the threshold of sediment motion. Wave velocity determined the sediment incipient motion, and the sediment content was controlled by wave-current velocities once suspended.

Identification of the spatial distribution of three sources of streamflow, total phosphorus (TP) and nitrate-nitrite nitrogen (NNN) in New Zealand. TP primarily comes from near-surface event flow and is correlated with the number of days with high rainfall, catchment elevation and dairy cow density. Most NNN originates from shallow groundwater discharge and is correlated with climate, geomorphology and geology factors, land cover and livestock density.

CNN model obtained higher performances for streamflow forecasting than LSTM and XGBoost models at lead times of 4–6 days. The interpretability of the optimal models with the SHAP method showed that the antecedent streamflow consistently dominated the streamflow forecasting performances of machine learning methods. With the increase of lead times, the contributions of soil moisture at the depth of 28–100 cm and leaf area index for low vegetation gradually increased, and they even surpassed the contributions of antecedent streamflow.

This study developed a water-salt transport model incorporating irrigation and drainage regulations and validated it using field data from the Hetao Irrigation District. Sensitivity analysis identified ditch spacing, saturated soil water content and specific yield as critical parameters. The model achieved a Nash–Sutcliffe efficiency of 0.63 and provides a robust tool for managing soil salinisation and optimising water use.