Machine Learning Approach to Predicting Floods in Bangladesh

Flood Prediction Using Machine Learning Models: Literature Review, 2018

Floods are among the most destructive natural disasters, which are highly complex to model. The research on the advancement of flood prediction models contributed to risk reduction, policy suggestion, minimization of the loss of human life, and reduction of the property damage associated with floods. To mimic the complex mathematical expressions of physical processes of floods, during the past two decades, machine learning (ML) methods contributed highly in the advancement of prediction systems providing better performance and cost-effective solutions. Due to the vast benefits and potential of ML, its popularity dramatically increased among hydrologists. Researchers through introducing novel ML methods and hybridizing of the existing ones aim at discovering more accurate and efficient prediction models. The main contribution of this paper is to demonstrate the state of the art of ML models in flood prediction and to give insight into the most suitable models. In this paper, the literature where ML models were benchmarked through a qualitative analysis of robustness, accuracy, effectiveness, and speed are particularly investigated to provide an extensive overview on the various ML algorithms used in the field. The performance comparison of ML models presents an in-depth understanding of the different techniques within the framework of a comprehensive evaluation and discussion. As a result, this paper introduces the most promising prediction methods for both long-term and short-term floods. Furthermore, the major trends in improving the quality of the flood prediction models are investigated. Among them, hybridization, data decomposition, algorithm ensemble, and model optimization are reported as the most effective strategies for the improvement of ML methods. This survey can be used as a guideline for hydrologists as well as climate scientists in choosing the proper ML method according to the prediction task. View Full-Text Keywords: flood prediction; flood forecasting; hydrologic model; rainfall–runoff, hybrid & ensemble machine learning; artificial neural network; support vector machine; natural hazards & disasters; adaptive neuro-fuzzy inference system (ANFIS); decision tree; survey; classification and regression trees (CART), data science; big data; artificial intelligence; soft computing; extreme event management; time series prediction

7th International Conference on Water and Flood Management-ICWFM 2019, 2019

VIII INTERNATIONAL ANNUAL CONFERENCE “INDUSTRIAL TECHNOLOGIES AND ENGINEERING” (ICITE 2021)

Floods in Malaysia are described as flash floods and monsoon floods by the Department of Irrigation and Drainage (DID). Forecasting is a specialized form of predictive analysis used to predict future trends or behavior from existing data. Weather forecasting predicts the weather in the future from current data circumstances. As flood forecasting models become more accurate, however, their capacity to accurately predict flooding decreased as the forecast continues. Therefore, to understand the different techniques used to forecast flood levels, a systematic review of the literature was conducted. This paper's main objective is to examine the most common variable used to forecast floods, utilizing the systematic review technique. From the main focus, we can identify the research questions, such as the most commonly used prediction method and its accuracy. In the end, two of the most common variables used for flood forecasting are rainfall and water level. This study's data can help others forecast floods using standard variables that yield the best accuracy possible.

2020

Flood is a highly chaotic, strongly emergent, and naturally occurring pattern. Flood prediction is important for building resilience to natural disasters by proactively managing and minimizing the impacts of natural hazards on our everyday lives. This project attempts to create a model that can accurately and efficiently map water level as a function of climate conditions to forecast flood levels at a given point in the future. Machine learning, as one of the Artificial Intelligence methods, contributes highly to the advancements of prediction systems to provide better performance and cost-effective solutions. Most of the current flood prediction models are primarily concerned with heavy rainfall and hurricanes. However, flooding in New Brunswick, commonly referred to as the spring flood, is primarily caused by rapid snowmelt, heavy rainfall and ice jams occurring during the spring (Figures 1 and 2). To create an accurate prediction, we implemented machine learning to discover a new...

2024

This study focused on forecasting flood susceptibility in regions within Ibadan Metropolis in Oyo state, South West Nigeria, using Logistic Regression and Random Forest Classifier models. Various features including drainage, slope, coordinates (X and Y), curvature, rainfall, flow accumulation, aspect, and topographic wetness index were utilized to predict flooding susceptibilities in this study. The Logistic Regression model had a training accuracy of 98.7%, and a test accuracy of 98.6%, while the Random Forest Classifier had a training accuracy of 100%, and a test accuracy of 100%. Both models gave good performances in the training and test sets, but the Random Forest Classifier outperformed the Logistic Regression model, hence it was chosen as the optimal model.

Water, 2018

Floods are among the most destructive natural disasters, which are highly complex to model. The research on the advancement of flood prediction models contributed to risk reduction, policy suggestion, minimization of the loss of human life, and reduction the property damage associated with floods. To mimic the complex mathematical expressions of physical processes of floods, during the past two decades, machine learning (ML) methods contributed highly in the advancement of prediction systems providing better performance and cost-effective solutions. Due to the vast benefits and potential of ML, its popularity dramatically increased among hydrologists. Researchers through introducing novel ML methods and hybridizing of the existing ones aim at discovering more accurate and efficient prediction models. The main contribution of this paper is to demonstrate the state of the art of ML models in flood prediction and to give insight into the most suitable models. In this paper, the literature where ML models were benchmarked through a qualitative analysis of robustness, accuracy, effectiveness, and speed are particularly investigated to provide an extensive overview on the various ML algorithms used in the field. The performance comparison of ML models presents an in-depth understanding of the different techniques within the framework of a comprehensive evaluation and discussion. As a result, this paper introduces the most promising prediction methods for both long-term and short-term floods. Furthermore, the major trends in improving the quality of the flood prediction models are investigated. Among them, hybridization, data decomposition, algorithm ensemble, and model optimization are reported as the most effective strategies for the improvement of ML methods. This survey can be used as a guideline for hydrologists as well as climate scientists in choosing the proper ML method according to the prediction task.

FLOOD HAZARD ESTIMATION AND EVALUATION IN LAGOS STATE USING MACHINE LEARNING TECHNIQUES, 2024

The term "flood" refers to the temporary overflow of water onto normally dry ground. There are few natural calamities as catastrophic as flooding. Excessive precipitation, snowfall, coastal storms, storm surges, and the failure of dams and other water management systems are all potential flooding causes. By definition, a flash flood occurs when a river overflows its natural levels, inundating the land around it for a brief period (Xie et al., 2020). As a result of humidity), topographical features (elevation, slope, land cover), and socioeconomic variables (population density, urbanization) is compiled. Feature selection and engineering techniques are applied to optimize model performance. Flooding is the most frequent and destructive natural catastrophe that may happen anywhere in the globe. The frequency and severity of flooding events have increased worldwide in recent years due to climate change and human activity. Flooding has caused widespread death and devastation of property, farms, and vegetation in several emerging Nigeria, including Lagos State, and has forced the relocation of many more. Flooding has been Lagos State’s most common natural disaster during the last decade. Modern machine learning methods have shown great promise for improving flood Estimation and Evaluation. The optimum machine learning algorithm for flood Estimation and Evaluation is debated. To reduce the harm caused by floods, finding better ways to anticipate their occurrence is crucial. This paper initially applied 4 machine learning algorithms (Support Vector Machine SVM, Classification and Regression Trees CART, K-Nearest Neighbors KNN, and 4. Generalized Linear Model Network GLMNET) on the default dataset. The results reveal fair accuracy (over 60%) and kappa values (< 0.4). The same ML algorithms were again applied to the transformed dataset using the Boxcox transformation technique; the accuracy and kappa values improved but not significantly. Finally, Models for predicting floods were implemented using 3 different ensemble algorithms: Bagged CART (Bootstrap Aggregating BAG), Random Forest (RF), and Stochastic Gradient Boosting ( Gradient Boosting Machine GBM). Compared to the other three models, the performance of RF (Area Under the Curve AUC = 0.93) and BAG (AUC = 0.92) indicated superior accuracy. Keywords: Artificial Neural Network, Flood Hazard Estimation, Gradient Boosting, Machine Learning, Random Forest, Support Vector Machine.

Contributing Paper to the 2019 UN Global Assessment on Disaster Risk Reduction, 2019

The objective of this paper is to illustrate how developing countries with limited datasets and capacity can utilize global hazard data to support risk-informed decision-making at the local level. Using urban hydrologic models for flood risk assessment requires the collection of intensive data for model calibration, and even after such an effort leads to considerable uncertainty for spatially specific risk assessment. The case study in this paper examines flooding events in Metro Manila, which routinely experiences the disasters. The authors explore whether relationships between flood occurrence in Manila and remotely-sensed environmental data, such as rainfall amounts and vegetation moisture, could be established using machine learning techniques such as visualization, decision tree, and logistic regression. The study demonstrated that a model that uses an appropriate rainfall index is better than one that uses only daily rainfall amounts or adds a vegetation index. The best predictive models are found to be: (i) one that uses rainfall type and rainfall amount; (ii) and one that integrates all the information, including rainfall amount, rainfall type and vegetation index. The case study demonstrates that globally available data used with machine learning techniques can be effective for local flood management.

Journal of Computer Science, 2021

Flood is one of the most destructive phenomena all over the world. Because the flooding uncertainties and the urgency to prepare for disaster management, three specific technique approaches are compared in this study to predict the flood occurrence based on historical rainfall data. The study involved the rainfall data in Kemaman, Terengganu between 2017 and 2018 extracted from the official portal of the state of Terengganu. The dataset covers daily rainfall reading between January to December of the particular year in millimeter (mm) per day along with flood risks occurrence. This prediction experiment will be conducted using three variations algorithms, which are Decision Tree, Naive Bayes and Support Vector Machine. The comparison using three different algorithms was used to define the best algorithms that work with historical rainfall datasets to predict flood in terms of accuracy, precision, recall and F1-score. In the future, the prediction results are hoped to alert government authorities to make an early strategy to handle flood problems in Malaysia by analyzing the rainfall pattern.

Floods are among the most destructive natural disasters, which are highly complex to model. The research on the advancement of flood prediction models contributed to risk reduction, policy suggestion, minimization of the loss of human life, and reduction the property damage associated with floods. To mimic the complex mathematical expressions of physical processes of floods, during the past two decades, machine learning (ML) methods contributed highly in the advancement of prediction systems providing better performance and cost-effective solutions. Due to the vast benefits and potential of ML, its popularity dramatically increased among hydrologists. Researchers through introducing novel ML methods and hybridizing of the existing ones aim at discovering more accurate and efficient prediction models. The main contribution of this paper is to demonstrate the state of the art of ML models in flood prediction and to give insight into the most suitable models. In this paper, the literature where ML models were benchmarked through a qualitative analysis of robustness, accuracy, effectiveness, and speed are particularly investigated to provide an extensive overview on the various ML algorithms used in the field. The performance comparison of ML models presents an in-depth understanding of the different techniques within the framework of a comprehensive evaluation and discussion. As a result, this paper introduces the most promising prediction methods for both long-term and short-term floods. Furthermore, the major trends in improving the quality of the flood prediction models are investigated. Among them, hybridization, data decomposition, algorithm ensemble, and model optimization are reported as the most effective strategies for the improvement of ML methods. This survey can be used as a guideline for hydrologists as well as climate scientists in choosing the proper ML method according to the prediction task conclusions.