Purpose. Mining-induced land subsidence is a significant concern in areas with extensive underground mining activities. Therefore, the prediction of land subsidence is crucial for effective land management and infrastructure planning. This research applies an artificial neural network (ANN) to predict land subsidence over the Mong Duong underground coal mine in Quang Ninh, Vietnam Methods. In the ANN model proposed in this research, four features are used as the model inputs to predict land subsi-dence, i.e., model outputs. These features include the positions of ground points in the direction of the trough main cross-section, the distance from the chamber (goaf) center to the ground monitoring points, the accumulated exploitation volume of extraction space, and the measured/recorded time. The entire dataset of 12 measured epochs, covering 22 months with a 2-month repetition time period, is divided into the training set for the first 9 measured epochs and the test set for the last 3 measured epochs. k-fold cross-validation is first applied to the training set to determine the best model hyperparameters, which are then adopted to predict land subsidence in the test set. Findings. The best model hyperparameters are found to be 5 hidden layers, 64 hidden nodes and 240 iterated epochs. Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) of the predicted land subsidence depend on the time separated between the last measured epoch and the predicted epoch. Within 2 months from the last measurements, RMSE and MAE are at 22 and 13 mm for Epoch 10, which increase to 31 and 20 mm for Epoch 11 (4 months from the last measurement) and 37 and 24 mm for Epoch 12 (6 months from the last measurement). Originality. A new ANN model with associated “optimal” hyperparameters to predict underground mining-induced land subsidence is proposed in this research. Practical implications. The ANN model proposed in this research is a good and convenient tool for estimating mining-induced land subsidence, which can be applied to underground mines in Quang Ninh province, Vietnam.

Fluvial floods endure as one of the most catastrophic weather-induced disasters worldwide, leading to numerous fatalities each year and significantly impacting socio-economic development and the environment. Hence, the research and development of new methods and algorithms focused on improving fluvial flood prediction and devising robust flood management strategies are essential. This study explores and assesses the potential application of 1D-Convolution Neural Networks (1D-CNN) for spatial prediction of fluvial flood in the Quang Nam province, a high-frequency tropical cyclone area in central Vietnam. To this end, a geospatial database with 4156 fluvial flood locations and 12 flood indicators was considered. The ADAM algorithm and the MSE loss function were used to train the 1D-CNN model, whereas popular performance metrics, such as Accuracy (Acc), Kappa, and AUC, were used to measure the performance. The results indicated remarkable performance by the 1D-CNN model, achieving high prediction accuracy with metrics such as Acc = 90.7%, Kappa = 0.814, and AUC = 0.963. Notably, the proposed 1D-CNN model outperformed benchmark models, including DeepNN, SVM, and LR. This achievement underscores the promise and innovation brought by 1D-CNN in the realm of susceptibility mapping for fluvial floods.

The Mekong Delta and Ho Chi Minh City in Vietnam are recognized as areas significantly impacted by land subsidence. This phenomenon has led to notable consequences, including increased vulnerability to issues such as saline intrusion and tidal flooding. GNSS-CORS technology, known for its capability to provide continuous time-series data, plays a crucial role in accurately monitoring changes in the land surface. Despite the existence of traditional algorithms for analyzing continuous measurement data collected through GNSS-CORS technology, their effectiveness is constrained by challenges in handling diverse input data and limitations in forecasting future displacements. Consequently, there is a growing trend towards the adoption of artificial intelligence techniques, particularly artificial neural networks (ANN), for predicting Up component in GNSS time-serries daily solution. This study leverages data from the CTHO GNSS CORS station located in the Mekong Delta to evaluate proposed models. An innovative hybrid approach, which integrates the Moving Average Filter (MAF) and Multilayer Perceptron Neural Network (MLPNN), is introduced to enhance the accuracy of forecasting. Performance evaluation metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE) are utilized to assess the effectiveness of the models. Results demonstrate the superior performance of the MLPNN model, achieving high prediction accuracy with metrics including MAE = 0.001, MSE = 0.000, and RMSE = 0.002. This research underscores the robustness of the proposed model in forecasting GNSS time-serries daily solution, highlighting its potential for practical applications in geodetic research.

Purpose. The research aims to address the challenges posed by the integration of diverse methods, focusing on data collection techniques and level of detail (LoD) considerations, which facilitates the creation of detailed 3D models. The CityGML standard is employed for its ability to represent complex urban features, adapted here for mining environments. Methodology. Combining Unmanned Aerial Vehicle (UAV) and Terrestrial Laser Scanning (TLS) technologies to collect data for open-pit and underground coal mines. These data are processed to generate point clouds, which are then used to create 3D models of mining structures using Sketchup and REVIT. Finally, these models are converted into the CityGML standard using FME SAFE software. Findings. Through the use of Unmanned Aerial Vehicle and Terrestrial Laser Scanning technologies, precise point cloud data for open-pit and underground structures are acquired. CityGML serves as a suitable framework for digital mine representation, offering standardized data organization and exchange. The proposed methodology optimizes data collection and processing procedures, ensuring accuracy and efficiency in model creation. Notably, the study introduces a nuanced approach to LoD selection, considering the complexity and specific requirements of different mining structures. Originality. The article innovatively combines UAV and TLS technologies with the CityGML standard to create comprehensive 3D GIS models for coal mines operating with both open-pit and underground methods, addressing the unique challenges of modeling diverse mining structures and terrain features. Practical value. The practical value of the article lies in its provision of a systematic approach using UAV and TLS technologies, coupled with the CityGML standard, to create accurate 3D GIS models for coal mines employing both open-pit and underground methods. This methodology enhances mine management efficiency, resource estimation accuracy, and safety assessment capabilities.