Modelling flood susceptibility based on deep learning coupling with ensemble learning models

Mapping flood susceptibility with PROMETHEE multi-criteria analysis method

On a global scale, flooding is the most devastating natural hazard with an increasingly negative impact on humans. It is necessary to accurately detect flood-prone areas. This research introduces and evaluates the Preference Ranking Organization METHod for Enrichment Evaluation (PROMETHEE) integrated with GIS in the field of flood susceptibility in comparison with two conventional multi-criteria decision analysis (MCDA) methods: analytical hierarchy process (AHP) and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). The Spercheios river basin in Greece, which is a highly susceptible area, was selected as a case study. The application of these approaches and the completion of the study requires the creation of a geospatial database consisting of eight flood conditioning factors (elevation, slope, NDVI, TWI, geology, LULC, distance to river network, rainfall) and a flood inventory of flood (564 sites) and non-flood locations for validation. The weighting of the factors is based on the AHP method. The output values were imported into GIS and interpolated to map the flood susceptibility zones. The models were evaluated by area under the curve (AUC) and the statistical metrics of accuracy, root mean squared error (RMSE), and frequency ratio (FR). The PROMETHEE model is proven to be the most efficient with AUC = 97.21%. Statistical metrics confirm the superiority of PROMETHEE with 87.54% accuracy and 0.12 RMSE. The output maps revealed that the regions most prone to flooding are arable land in lowland areas with low gradients and quaternary formations. Very high susceptible zone covers approximately 15.00-19.50% of the total area and have the greatest FR values. The susceptibility maps need to be considered in the preparation of a flood risk management plan and utilized as a tool to mitigate the adverse impacts of floods.

Flood susceptibility assessment of the Agartala Urban Watershed, India, using Machine Learning Algorithm

Frequent floods are a severe threat to the well-being of people the world over. This is particularly severe in developing countries like India where tropical monsoon climate prevails. Recently, flood hazard susceptibility mapping has become a popular tool to mitigate the effects of this threat. Therefore, the present study utilized four distinctive Machine Learning algorithms i.e., K-Nearest Neighbor, Decision Tree, Naive Bayes, and Random Forest to estimate flood susceptibility zones in the Agartala Urban Watershed of Tripura, India. The latter experiences debilitating floods during the monsoon season. A multicollinearity test was conducted to examine the collinearity of the chosen flood conditioning factors, and it was seen that none of the factors were compromised by multicollinearity. Results showed that around three-fourths of the AUW area was classified as moderate to very high flood-prone zones, while over 20 percent was between low and very low flood-prone zones. The models applied performed well with ROC-AUC scores greater than 70 percent and MAE, MSE, and RMSE scores less than 30 percent. DT and RF algorithms were suggested for places with similar physical characteristics based on their outstanding performance and the training datasets. The study provides valuable insights to policymakers, administrative authorities, and local stakeholders to cope with floods and enhance flood prevention measures as a climate change adaptation strategy in the AUW.

A Unified Efficient Deep Learning Architecture for Rapid Safety Objects Classification Using Normalized Quantization-Aware Learning

The efficient recognition and classification of personal protective equipment are essential for ensuring the safety of personnel in complex industrial settings. Using the existing methods, manually performing macro-level classification and identification of personnel in intricate spheres is tedious, time-consuming, and inefficient. The availability of several artificial intelligence models in recent times presents a new paradigm shift in object classification and tracking in complex settings. In this study, several compact and efficient deep learning model architectures are explored, and a new efficient model is constructed by fusing the learning capabilities of the individual, efficient models for better object feature learning and optimal inferencing. The proposed model ensures rapid identification of personnel in complex working environments for appropriate safety measures. The new model construct follows the contributory learning theory whereby each fussed model brings its learned features that are then combined to obtain a more accurate and rapid model using normalized quantization-aware learning. The major contribution of the work is the introduction of a normalized quantization-aware learning strategy to fuse the features learned by each of the contributing models. During the investigation, a separable convolutional driven model was constructed as a base model, and then the various efficient architectures were combined for the rapid identification and classification of the various hardhat classes used in complex industrial settings. A remarkable rapid classification and accuracy were recorded with the new resultant model.