The Dai Loc shear zone in central Vietnam contains granulite-facies rocks and is a key area for studying the Early Paleozoic metamorphic evolution of the Indochina Block. An integrated study of in-situ geochronology, trace element geochemistry, and microtextural analysis was conducted to decipher the metamorphic evolution of this high-grade unit. Monazites from the two granulite samples display three distinct chemical domains, whose trace element compositions closely correlate with garnet growth and breakdown. Yttrium- and heavy rare-earth element (HREE)-rich monazite core domains are interpreted to have formed with limited garnet growth, recording a discrete growth episode during prograde metamorphism at ~435 Ma. Y- and HREE-poor domains are linked to significant garnet growth during peak conditions at ~420 Ma. The elevated Y+HREE concentrations in the outermost rim domains indicate their formation during garnet breakdown and likely date the retrograde metamorphism to ~390 Ma. These U–Pb monazite ages align well with the U-Pb zircon ages from granulites and syn-metamorphic granitoids in the study area, reinforcing the inferred metamorphic timeline. The results of this petrochronological study highlight the importance of integrating petrology with trace element data from major and accessory phases to link geochronological data to metamorphic P–T paths.

This study evaluates the performance of multispectral optical sensors onboard PlanetScope (PS) and Sentinel-2 satellites in mapping burned areas resulting from a small forest fire that occurred on 21 March, 2025, in Nghiem Mountain, northern Vietnam. Cloud-free pre- and post-fire imagery acquired on the same dates (January, 17 and 12 May, 2025) were used to compute the differenced Normalized Difference Vegetation Index (dNDVI) using Red and Near-Infrared surface reflectance. A threshold value (T = 0.10), selected after analyzing the dNDVI histograms, was applied to classify burned (dNDVI > T) and unburned regions (dNDVI ≤ T). Results showed a strong spatial correlation between dNDVI maps derived from both satellites (R = 0.97), although Sentinel-2 tends to yield slightly higher dNDVI values than PS satellites. The burned area estimated from PS was 20.225 ha, while Sentinel‑2 produced a similar estimate of 20.622 ha, a difference of less than 2% and in close agreement with the official damage assessment report (~20 ha). Most discrepancies occurred along fire boundaries, where mixed pixels and spectral heterogeneity are expected. Our results demonstrate the effectiveness of Sentinel-2 and PS satellite imagery for mapping burned areas from small-scale fires, which is essential for forest management. Despite several limitations, including dependence on clear-sky conditionss and the lack of a ground-based validation dataset, the proposed approach provides a timely and cost-effective solution for wildfire mapping at small scales, particularly important in remote regions.

Methods for enhancing and estimating parameters of potential fields from gravimetric and magnetometric surveys typically utilize the vertical derivative (VDR) of the potential field. These derivatives amplify the high-frequency content of the field, which can be caused by shallow bodies or survey noise. Regularization methods generate approximations of derivatives that must reconcile two objectives: reducing the effect of high-frequency amplification and providing an accurate approximation. Achieving this balance is crucial for methods that require vertical derivatives of successive order, such as Taylor-series implementations of downward continuation and the enhanced horizontal derivative (EHD) filter. This paper evaluates the performance of several vertical-derivative methods for downward continuation and EHD filters using both noise-free and noisy synthetic data. In addition, gravity data over the SW Sub-basin are considered, and the findings are compared with seismic data. Our results show that the β-VDR method provides more accurate and stable derivatives under noisy conditions.

The Permian-Triassic igneous rocks of the Truong Son Fold Belt, on the northeastern margin of the Indochina block, formed during Paleotethyan subduction and the subsequent collision with the South China block. Systematic variations in zircon ages, geochemistry, and isotopic compositions are observed from northern Laos toward the Song Ma Suture zone, the collisional boundary with the South China block. In the Xiang Khuang-Muang Khoun (MK) area, 170–200 km from the suture, magmatic rocks (271–253 Ma) include gabbros and I-type granitoids with relatively higher εNd(t) values (-1.5 to -9) and lower 87Sr/86Sri (0.704–0.717). Toward the suture, in the Nam Phao-Kim Cuong area, granitoids dated at 260–251 Ma are predominantly S-type, highly peraluminous granites with intermediate isotopic εNdi values (-7.4 to -9) and 87Sr/86Sri values (0.7115–0.7285). In the Sam Neua (SN) area, closest to the suture, granitoids dated at 251–244 Ma are primarily I-type and minorly S-type, with highly enriched isotopic compositions (εNdi of -8.4 to -14; 87Sr/86Sri between 0.708 and 0.775). The trace-element chemistry of all granitoids indicates volcanic-arc affinities, though signatures also suggest intraplate and post-collisional influences. The association of granitoids with coeval gabbro-diorites in the MK area suggests binary mixing between mantle magmas and Mesoproterozoic crust-derived melts. In contrast, felsic magmas in the SN area likely reflect melting of diverse crustal components, including Paleoproterozoic continental crust of South China affinity. The suture zone-ward younging of magmatism is consistent with slab rollback during the final stages of continental collision.

Over the past few decades, urban expansion has accelerated worldwide. This process can increase future flood risks due to local changes in hydrological conditions and the increased exposure and vulnerability of communities in flood-prone areas. Therefore, assessing the impact of urban expansion on flood susceptibility is an important task that can support local authorities in urban planning and in mitigating flood impacts. The objective of this study was to assess the impact of urban expansion on flood susceptibility in Hanoi using machine learning models: Deep Neural Networks (DNN), Adaptive Boosting (ADB), Extreme Gradient Boosting (XGB), and Random Forest (RF). A total of 1058 flood points and 14 conditioning factors corresponding to 2014 and 2024 were used as input to the models. Statistical indices, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Area Under the Curve (AUC), and Coefficient of Determination (R2) were used to evaluate the performance of the proposed model. The results showed that the DNN model achieved the highest performance in assessing the impact of urban expansion on flood susceptibility (AUC=0.92), followed by XGB (0.91), ADB (0.86), and RF (0.82). During 2014–2024, urban expansion combined with the impacts of climate change has significantly increased the areas susceptible to flooding. In Hanoi, areas in the "high" and "very high" flood-susceptibility categories have been expanding continuously, accounting for about 25% of the total study area. In contrast, the "medium" group has a slight decreasing trend, while the "low" and "very low" areas have narrowed. This shows that urban expansion is increasing the area prone to flooding. The results of this study provide a solid scientific basis, supporting planners and policymakers in identifying limitations in current flood risk adaptation measures and in developing more appropriate spatial and temporal strategies to minimize flood impacts.

Northern Vietnam’s karst landscapes offer an untapped potential for paleo-monsoon research, complementing the extensive speleothem records of Southwest China. Here, we surveyed 51 caves across seven provinces, targeting those with humidity exceeding 95% from 2017 to 2024, and collected 127 broken stalagmites to evaluate their stable-isotope potential (δ18O, δ13C) as paleoclimate proxies. Focusing on caves in Hoa Binh and nearby karst-rich regions, we applied Hendy tests to 56 subsamples of eight layers on four stalagmites, NS3, HS3, HS7, and HS16, to assess the isotopic equilibrium conditions. The deposition intervals of the four stalagmites, determined using U/Th dating, range from 36 to 60 thousand years ago (ka). Small 1-sigma variations of ±0.04−0.20‰ in coeval δ18O values across all eight layers suggest deposition under near oxygen isotope equilibrium. Combined with fast growth rates exceeding 0.09 mm/year, this evidence suggests high-resolution potential for paleohydroclimate reconstruction using stalagmite δ18O data. However, one-sigma variations of ±0.04−0.71‰ of coeval δ13C data reflect relatively large carbon isotopic fractionation during the degassing process. It suggests that stalagmite δ13C records from these caves should be carefully evaluated before use in paleoclimate reconstructions.

In geotechnical engineering, an accurate estimation of maximum dry density (MDD) is essential to ensure the stability of geotechnical structures such as roads, embankments, and foundations. While traditional laboratory methods, such as the Proctor compaction test, are reliable, they are often labor-intensive and time-consuming. Therefore, the main aim of this study is to develop efficient data-driven models, including XGBoost (XGB), CatBoost (CAB), and Bagging (BAG), for rapid and reliable estimation of MDD using easily measurable soil properties. A dataset of 214 soil samples collected from the Van Don-Mong Cai expressway construction project (Vietnam) comprising eight key input variables was used: gravel content, coarse and fine sand contents, silt and clay content, optimum moisture content, liquid limit, plastic limit, and plasticity index. Model performance was evaluated using R², RMSE, MAE, and a Taylor diagram. Results indicate that the Grid Search-optimized BAG model achieved the best performance, with R² values of 0.94 and 0.81 for the training and test datasets, respectively, and the lowest RMSE and MAE. Optimized CAB showed comparable performance, while XGB exhibited relatively lower generalization capability. Optimized CAB yielded similar results, whereas optimized XGB performed worse. The significance of this study lies in demonstrating that ensemble learning models, particularly Bagging, can provide accurate, physically interpretable predictions of MDD, thereby reducing reliance on extensive laboratory testing. The novelty of this work lies in the systematic comparison of optimized ensemble models using a real construction dataset, combined with interpretability analysis via partial dependence plots consistent with established soil mechanics principles. These findings highlight the potential of optimized machine learning models as practical tools for modern geotechnical engineering applications.

Bathymetry mapping plays a critical role in coastal zone management, marine conservation, and navigation safety. With the increasing availability of high-resolution satellite imagery, such as PlanetScope (3−5 m), remote sensing-based bathymetry retrieval offers a cost-effective and scalable alternative to traditional in-situ surveys. This study explores the capability of PlanetScope imagery to retrieve a wide range of bathymetry (-0.5 − ~ -40 m) in the southern area of the Nha Trang Marine Protected Area (MPA), Vietnam - an ecologically significant and dynamic coastal region. We conduct a comprehensive comparison between traditional approaches, including the Stumpf ratio model and Multiple Linear Regression (MLR), and a suite of advanced machine learning (ML) algorithms, including Random Forest (RF), Support Vector Machine (SVM), Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting (XGB), CatBoost (CB), and Gradient Boosting (GB). Among these, RF achieved the highest performance with an R2 of 0.85, RMSE of 2.66 m, and MAE of 1.85 m, significantly outperforming the Stumpf model (R2 = 0.29) and MLR (R2 = 0.57). This study represents one of the most extensive model comparisons to date for satellite-derived bathymetry using PlanetScope data, offering a benchmark for future applications in tropical coastal environments. Results underscore the potential of machine learning to advance spatially detailed and accurate bathymetric mapping from space.

In this research, two newly hybrid machine learning (ML) models, including Decorate Ensemble-based Partial Decision Trees (D-PART) and Bagging Ensemble-based Partial Decision Trees (B-PART), were applied to generate an accurate landslide susceptibility map for the Muong Te area, Lai Chau Province, Vietnam. The performance of the novel models was compared with two single benchmark models, namely Partial Decision Trees (PART) and Logistic Regression (LR), using the popular area under the Receiver Operating Characteristic (ROC) curve (AUC) metric. To construct the training and validation datasets, a spatial database was developed comprising ten landslide conditioning factors associated with the area's topographic, geological, structural, and hydrological characteristics, along with 248 documented historical and recent landslide occurrences. The OneR technique was applied to prioritize the most influential factors and to improve the model's performance. The evaluation results demonstrate that D-PART yielded the strongest predictive performance, with an AUC of 0.801, followed by B-PART (0.795), PART (0.758), and Logistic Regression (0.736). Thus, the novel hybrid model D-PART is a promising technique for constructing a reliable landslide susceptibility map, which can be used for effective planning and management of landslides in landslide-prone areas.

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.