Flash floods rank among the most catastrophic natural disasters worldwide, inflicting severe socio-economic, environmental, and human impacts. Consequently, accurately identifying areas at potential risk is of paramount importance. This study investigates the efficacy of Deep 1D-Convolutional Neural Networks (Deep 1D-CNN) in spatially predicting flash floods, with a specific focus on the frequent tropical cyclone-induced flash floods in Thanh Hoa province, North Central Vietnam. The Deep 1D-CNN was structured with four convolutional layers, two pooling layers, one flattened layer, and two fully connected layers, employing the ADAM algorithm for optimization and Mean Squared Error (MSE) for loss calculation. A geodatabase containing 2540 flash flood locations and 12 influencing factors was compiled using multi-source geospatial data. The database was used to train and check the model. The results indicate that the Deep 1D-CNN model achieved high predictive accuracy (90.2%), along with a Kappa value of 0.804 and an AUC (Area Under the Curve) of 0.969, surpassing the benchmark models such as SVM (Support Vector Machine) and LR (Logistic Regression). The study concludes that the Deep 1D-CNN model is a highly effective tool for modeling flash floods.

The main purpose of this paper is to assess the land use and land cover (LULC) changes for thirty years, from 1990–2020, in the Dong Thap Muoi, a flooded land area of the Mekong River Delta of Vietnam using Google Earth Engine and random forest algorithm. The specific purposes are: (1) determine the main LULC classes and (2) compute and analyze the magnitude and rate of changes for these LULC classes. For the above purposes, 128 Landsat images, topographic maps, land use status maps, cadastral maps, and ancillary data were collected and utilized to derive the LULC maps using the random forest classification algorithm. The overall accuracy of the LULC maps for 1990, 2000, 2010, and 2020 are 88.9, 83.5, 87.1, and 85.6%, respectively. The result showed that the unused land was dominant in 1990 with 28.9 % of the total area, but it was primarily converted to the paddy, a new dominant LULC class in 2020 (45.1%). The forest was reduced significantly from 14.4% in 1990 to only 5.5% of the total area in 2020. Whereas at the same time, the built-up increased from 0.3% to 6.2% of the total area. This research may help the authorities design exploitation policies for the Dong Thap Muoi’s socio-economic development and develop a new, stable, and sustainable ecosystem, promoting the advantages of the region, early forming a diversified agricultural structure.

Flash floods continue to emerge as a serious and growing natural hazard for many communities worldwide, especially in areas affected by tropical storms. These floods damage critical infrastructure and severely strain economic resources, underscoring the urgent need for advanced flood prediction tools. This study presents an innovative integrated machine learning approach, BCMO-RF, which merges Balancing Composite Motion Optimization (BCMO) with Random Forest (RF) to map flash flood susceptibility. In the BCMO-RF approach, the RF algorithm is applied to develop the flash flood model, while BCMO is used to explore and optimize the model's parameters. The study concentrates on areas in Thanh Hoa Province, Vietnam, frequently impacted by flash floods. Accordingly, various geospatial data sources were utilized to compile a geodatabase comprising 2,540 flash flood locations and 12 influencing factors. The geodatabase served as the basis for training and validating the BCMO-RF model. Results show that the BCMO-RF model attained high prediction accuracy (93.7%), achieving a Kappa coefficient of 0.874 and an AUC score of 0.988, outperforming the Deep Learning model benchmark. The study finds that the BCMO-RF model is reliable for accurately mapping areas susceptible to flash floods.