In this paper, the main objective is to predict total bearing capacity (TBC) of pretensioned spun concrete piles (PSCP) using Machine Learning (ML) methods namely Reduced Error Pruning Tree (REPT), Gaussian Process (GP), Artificial Neural Networks (ANN) and two novel hybrid models including: Cascade Generalization based Gaussian Processes (CG-GP) and Cascade Generalization based Artificial Neural Networks (CG-ANN) based on data from 95 PSCP piles installed at the Hoa Binh 5 wind power plant project in Vietnam. For model development, field-estimated TBC values obtained from Pile Driving Analyzer (PDA) tests were used as the output parameter. The predictive capability of the models was validated using common statistical indicators, namely Mean Absolute Error (MAE), Coefficient of Determination (R2) and Root Mean Square Error (RMSE) with 70% of the data used for training and 30% for testing. The results indicated that the proposed hybrid CG-ANN model (R² = 0.935, RMSE = 44.691 ton, MAE = 30.215 ton) outperformed all other models including CG-GP (R2 = 0.929, RMSE = 50.738 ton, MAE = 37.812 ton), Artificial Neural Networks - ANN (R2 = 0.926, RMSE = 47.963 ton, MAE = 32.167 ton), REPT (R2 = 0.776, RMSE = 75.350 ton, MAE = 53.115 ton) and GP (R2 = 0.916, RMSE = 52.785 ton, MAE = 39.967 ton) in the correct prediction of the TBC of PSCP. The results demonstrate that the hybrid CG-ANN model can serve as an efficient and reliable tool for rapid, accurate estimation of PSCP bearing capacity, thereby helping reduce the time and cost associated with elaborate field testing.

This experimental study evaluates the influence of soil consistency on the ultimate bearing capacity and elastic modulus of single geogrid-reinforced and unreinforced stone columns installed in soft-to-very soft clay. Laboratory model tests were conducted in the Ho Thuong Tin area of Hanoi, Vietnam, under three controlled soil conditions with liquidity indices (IL) of 0.78, 1.0, and 1.5. A total of six displacement-controlled load tests were performed in a unit-cell setup, with and without a geogrid reinforcement layer at the column head. Results show that the inclusion of geogrid significantly enhanced both the ultimate bearing capacity and stiffness of the stone column: the bearing capacity increased by approximately 12–25%, and the elastic modulus (E₅₀) increased by 19–27% relative to unreinforced columns. The study indicates that geogrid reinforcement at the column head provides a technically material-saving and straightforward alternative to complete encasement, with potential economic advantages particularly for soft clays where lateral confinement is limited. The findings provide new experimental data to inform the design of geogrid-reinforced stone columns under varying soil conditions.

Accurate reservoir inflow forecasting is critical for real-time water management in monsoon-dominated basins. This study develops a weighted Voting Ensemble model to predict daily inflow to the A Luoi hydropower reservoir in central Vietnam using multi-station rainfall and lagged inflow data. Five machine learning models MLP, RF, KNN, XGB, and Ridge Regression, were trained on a unified feature set containing current and lagged rainfall inputs and three runoff memory terms, and subsequently combined using performance-based weights derived from time-series cross-validation errors. Evaluation using MSE, RMSE, and NSE shows that the ensemble outperforms all standalone learners, reducing RMSE by 12–25% and improving NSE from 0.70–0.91 (best individual models) to 0.92 on the test set. SHAP analysis is also used to explain model predictions and highlight the most influential features. During an independent verification period, the ensemble maintained strong performance (NSE ≈ 0.98), accurately capturing rising and recession limbs and minimizing peak-flow underestimation. These results demonstrate the robustness and operational feasibility of weighted ensemble learning for short-term inflow forecasting, offering valuable support for reservoir operation, flood mitigation, and water allocation in data-rich reservoir systems.

This study investigated wave-induced suspended sediment dynamics and seabed morphological changes around the Truong Sa Island in the East Vietnam Sea. The analysis was conducted using the MIKE 21/3 Coupled Model FM, which integrates wave, hydrodynamic, sediment transport, and bed evolution modules to simulate seabed responses to both seasonal and interannual wave forcing during 2013–2015. The model was forced with ERA5 reanalysis data and validated using observed tidal and bathymetric datasets from the Vietnam Academy of Science and Technology (VAST). Results revealed that seabed changes were most pronounced in the northeastern nearshore region of the Truong Sa Island. It was driven by the prevailing northeast-directed wave conditions, which also contributed to elevated suspended sediment concentrations (SSCs). Seasonally, winter exhibited more dynamic seabed changes and greater suspended sediment variability than summer. Under the most severe wave conditions in 2015, the simulated domain experienced a net seabed incision with an average depth of -0.54 m. Spatially, the nearshore areas were characterized by significant seabed incision and high SSCs, whereas offshore regions showed minimal seabed changes and low SSCs. Moreover, persistent sediment deposition was observed in the harbor southwest of the Truong Sa Island, implying the need for regular dredging to maintain operational capacity. The findings underscore the need for integrated management strategies that balance coastal development, marine ecosystem protection, and navigational safety.

Accurate climate projections are crucial for water resource management in the Lake Toba region of Indonesia, where rainfall supports agriculture, tourism, and hydropower. However, rainfall projections from CMIP6 global climate models (GCMs) often exhibit systematic biases due to coarse spatial resolution and model assumptions. This study applies the statistical bias-correction method, Quantile Delta Mapping (QDM), to adjust CMIP6 rainfall outputs using observed data from seven weather stations around Lake Toba. The study employs ten probability distributions and compares two distribution identification approaches: monthly and full-period. Results show that monthly distribution identification (QDM1) improves model performance more consistently, reducing the Kolmogorov-Smirnov (KS) error by the raw CMIP6 data by 47.26% compared to 20.39% from the full-period approach (QDM2). Post-correction projections using a Multi-Model Ensemble Mean (MMEM) for 2015–2050 under scenarios SSP2-4.5 and SSP5-8.5 indicate moderated rainfall trends in 2050 after correction, notably reduced overestimation in wet seasons. These more conservative projections suggest slight decreases during MAM, contrasting with the uncorrected CMIP6 tendency toward wetter futures. It also suggests the future use of regional climate models (RCMs) and nonparametric or machine-learning methods to capture more complex rainfall distribution patterns in the region.

In the face of accelerating land degradation from open-pit mining, ecological rehabilitation has become a cornerstone strategy for restoring ecosystem functions and sustainability. This study evaluated the early outcomes of a pilot multi-strata agroforestry model designed for post-bauxite mine spoil at the Tan Rai site, Vietnam. The system combined Pinus caribaea Morelet, Bixa orellana L., and Sphagneticola trilobata (L.) Pruski, integrated with soil amendment products such as controlled-release NPK fertilizers, superabsorbent polymers, and polyacrylamide. Over a six-year observation period, comprising two years of active management followed by four years of natural succession, plant growth performance, nine soil physicochemical properties, and eight heavy metal concentrations were assessed through six sampling campaigns. Results indicated significant improvements in vegetation structure and soil quality. P. caribaea showed strong height growth, while B. orellana exhibited rapid diameter expansion and reached reproductive maturity by year six. Vegetation cover stabilized above 75% after 12 months, accompanied by increasing species diversity. Within the first 24 months, soil organic carbon, available nutrients, cation exchange capacity, and pH improved markedly, while heavy metal levels remained within the safe thresholds specified by QCVN 03:2023/BTNMT. These findings highlight the potential of multi-strata agroforestry, when combined with targeted soil amendments, to accelerate ecological recovery on post-bauxite mine spoils. However, continued monitoring is required to ensure long-term sustainability.

Estuarine environments are highly productive and biodiverse ecosystems that are particularly vulnerable to environmental pollution. This study assessed the presence and ecological risks of microplastics (MPs) and ten heavy metals in surface sediments from two major estuaries of the Red-Thai Binh River system in northern Vietnam: Ba Lat and Bach Dang. Sediment samples were collected during the rainy season (July 2024) under low tide conditions. The average concentrations of heavy metals followed the order: Hg < Cd < As < Ni < Cu < Cr < Pb < Zn < Mn < Fe. While most metal concentrations, except for Fe, were below Vietnamese regulatory limits (QCVN 43:2025/BTNMT), several samples exceeded the U.S. EPA (1997) Threshold Effects Level (TEL), particularly for As, Pb, Hg, Cu, and Ni. The geo-accumulation index identified Pb as the most enriched element, followed by As and Zn. Despite localized exceedances, the overall ecological risk associated with heavy metals was classified as low. In contrast, microplastic contamination posed a more prominent ecological threat. MP concentrations ranged from 3,600 to 9,000 items/kg (mean: 5,908±1,790) in Ba Lat and from 1,900 to 4,800 items/kg (mean: 3,858.3±832.8) in Bach Dang, surpassing levels reported in previous regional studies. The dominant particle types were small-sized fibers (< 2 mm) and fragments (0.05–0.2 mm²), which are likely to have greater bioavailability and ecological impact. A Potential Ecological Risk Index (PERI), incorporating MP abundance, polymer types, and hazard scores, indicated high-to-dangerous risk levels in Bach Dang and medium-to-high risk levels in Ba Lat. These findings highlight the urgent need for effective mitigation strategies, including improved plastic waste management and routine MP monitoring, particularly in ecologically sensitive areas such as aquaculture zones, coastal habitats, and salt production sites.

The Xa Loi gold deposit, located within the Truong Son Fold Belt of central Vietnam, provides a critical case study for understanding magmatic-hydrothermal gold systems in Southeast Asia. This study integrates sulfur (δ³⁴S), oxygen-hydrogen (δ¹⁸O–δD) isotopes, microthermometry, and stepwise-crushing gas-inclusion analyses to constrain the origin, evolution, and physicochemical conditions of the ore-forming fluid. Pyrite exhibits a narrow δ³⁴S range from –0.49‰ to +2.90‰, indicating a homogeneous sulfur source derived from magmatic degassing under moderately oxidizing conditions. Quartz δ¹⁸O and fluid δD values plot entirely within global magmatic fields, demonstrating that the aqueous component of the hydrothermal system was sourced from a deep, evolved felsic-intermediate magma with no meteoric contribution. Fluid-inclusion gases define a H₂O–CO₂–N₂–CH₄ system characterized by CO₂/CH₄ > 1 and elevated N₂/Ar ratios (≈400–1100), with all data plotting within the Magmatic-Evolved Magmatic fields of the CO₂/CH₄–N₂/Ar diagram. These signatures indicate a CO₂-rich, moderately oxidized magmatic fluid that underwent limited interaction with N-bearing crustal rocks during ascent. The progressive transition from SFI to Mixing and PFI inclusion populations records increasing oxidation consistent with magmatic degassing. Collectively, the S–O–H isotope systematics and gas compositions reveal that the Xa Loi gold deposit was formed by a magmatic-dominated hybrid fluid, in which sulfur, water, CO₂, and the conditions required for Au transport were supplied directly by an evolved magma and only slightly modified by crustal interaction. This genetic model highlights the role of deep magmatic processes in the formation of gold deposits along the Truong Son Fold Belt.

Land-use and land-cover (LULC) change, as well as landscape fragmentation, significantly impact biodiversity conservation in biosphere reserves. This paper aims to analyze LULC change intensity and landscape fragmentation, and to test their relationship within the Western Nghe An Biosphere Reserve (WNABR) from 1990 to 2020. LULC data from 1990, 2000, 2010, and 2020, covering 13 LULC types, were analyzed for LULC change, LULC change intensity at the interval and categorical levels, and landscape fragmentation using landscape metrics at class and landscape levels. Pearson's correlation coefficient was then employed to examine the relationship between land change intensity and landscape fragmentation. The results indicate that from 1990 to 2020, high-developed areas, low-developed areas, and bamboo areas consistently increased, while plantation land, scrub/shrub, grassland/herbaceous vegetation, other croplands, and rice paddies declined. Evergreen broadleaf forest areas remained stable and dominant. Land change intensity was above average from 1990 to 2000, then gradually decreased from 2000 to 2020. Evergreen broadleaf forests, open water, other croplands, and rice paddies exhibited lower land change intensity than other types. There is a general trend of decreasing fragmentation within the WNABR. Evergreen broadleaf forest and plantation land are composed of large, less fragmented patches, while the remaining LULC types are typically fragmented into small, scattered patches. LULC change intensity and landscape fragmentation increased from the core zone to the buffer and transition zones. There is a significant positive linear relationship between LULC change intensity and the number of patches (NP).

Mixed Rossby-gravity (MRG) waves are key equatorial disturbances that modulate convection and rainfall across the Maritime Continent, yet their representation in regional models remains underexplored. Accurate simulation of MRG waves induced vertical structure is critical for improving forecasts of tropical weather variability and associated hydrometeorological hazards. This study evaluates the Weather Research and Forecasting (WRF) model's ability to simulate vertical atmospheric profiles during an MRG wave event on the southern coast of West Java from 26 to 29 September 2022, using radiosonde observations at 06, 12, and 18 local time (LT). Eleven model configurations were evaluated, differing in domain schemes (two vs. three nesting steps), vertical resolution (33, 45, 60, 80, and 100 levels), and input data sources (Real-Time Global Forecast System [GFS] vs. Final Operational Global Analysis [FNL]). All model configurations used in the simulation have a fixed physics scheme parameterization. The simulations were compared with radiosonde observations and evaluated statistically using the correlation coefficient (R) and Normalized Mean Absolute Error (NMAE). The analysis demonstrates that the WRF model effectively captures MRG wave dynamics by simulating key atmospheric variables, including pressure (P), temperature (T), relative humidity (RH), zonal (U) and meridional (V) wind anomalies in strong agreement with observations. P is well represented, exhibiting the highest R (0.81), whereas RH is the lowest (0.21), likely reflecting the model's inability to capture fine‑scale observed moisture variations. Configurations that utilized a two-step nesting domain and the FNL input demonstrated the best performance, achieving higher R values and lower NMAE. Input data had a notable impact on model performance: the FNL analysis improved R by ~36% and reduced NMAE by ~12% compared to GFS, likely due to FNL's assimilation of observational data, which reduces uncertainty. Moreover, a domain scheme with a smaller outer domain and fewer nesting steps also improved R by ~36% and reduced NMAE by ~12%, suggesting that simpler domain configurations help limit error propagation. Additionally, increasing the vertical resolution from 33 to 100 levels enhanced the simulation of MRG wave structures, improving R by ~45% and reducing NMAE by ~22%. These findings enhance the understanding of MRG wave dynamics and offer valuable insights for improving regional weather forecasting.