Flood Prediction and Forecasting Using Anns and Fuzzy Logic Model in Sylhet, Bangladesh

Flood Prediction and Forecasting Using Anns and Fuzzy Logic Model in Sylhet, Bangladesh | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Flood Prediction and Forecasting Using Anns and Fuzzy Logic Model in Sylhet, Bangladesh Mayaz Uddin Gazi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6882638/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Floods are a significant threat to the environment, human life, and infrastructure in Sylhet, Bangladesh. The region's unique geography and climate make it prone to frequent and severe floods, which can cause significant economic and social losses. To address this issue, this study proposes an intelligent flood management system that combines the power of artificial neural networks (ANNs) for rainfall prediction and fuzzy logic for flood forecasting. The system is designed to provide accurate and real-time predictions of rainfall and flood events, enabling effective decision-making and mitigation strategies. The ANNs are trained using historical rainfall data to predict future rainfall patterns, whereas the fuzzy logic component uses the predicted rainfall data to forecast flood events. The system was tested on real-world data from Sylhet and demonstrated high accuracy in predicting rainfall and flood events. The ANN model was developed via a feed-forward backprop network with three input variables and one output variable (rainfall). A TRAINLM training function, LEARNGDM adaptation learning function, and MSE performance function were used. The ANN architecture consisted of two hidden layers with eight neurons each, with LOGSIG and PURELIN transfer functions for the first and second layers, respectively. The fuzzy logic component employs a Mamdani-type fuzzy inference system (FIS) with twelve rules, using rainfall and river level as inputs and flooding as the output. Triangular (TRIMF) and trapezoidal (TRAPMF) membership functions were utilized. The results of the ANN model revealed a mean square error (MSE) with a suitable curve and a correlation coefficient (R) over 0.9, indicating a strong correlation between the predicted and actual values. Additionally, we obtain an outstanding mean absolute error (MAE) value. The hybrid approach combining ANNs and fuzzy logic demonstrated high accuracy in flood forecasting, outperforming traditional methods. The proposed system can provide early warning of flood events, enabling timely mitigation measures and reducing the impact on communities. Civil Engineering Flood forecasting Artificial neural networks Feed-forward backdrop Fuzzy logic Hybrid approach Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 1. Introduction As a riverine country with complex weather, floods are common during the monsoon season in this country. From 1972–2017, Bangladesh faced 86 flood events [ 1 ]. Excessive rainfall in the monsoon season is one of the most prominent causes of frequent flood occurrence [ 2 ]. As Sylhet is situated at the banks of the Surma and Kushiyara rivers, floods have become a frequent natural disaster for Sylhet in recent years. In 2022, Sylhet faced the worst flood conditions in 122 years, which affected more than 7.2 million people and caused an overall economic loss of 722.24 dollars in different sectors [ 3 ]. Therefore, for flood forecasting and forecasting, it is necessary to improve resource allocation and response coordination and develop flood management techniques. Machine learning (ML) and deep learning with fuzzy logic offer the ability to provide instant and up-to-date information. ML models have the ability to offer real-time flood predictions and multistep-ahead forecasts, aspects that are deemed essential for providing timely warnings and effective disaster management [ 4 ]. As machine learning has shown outstanding ability to process data, detect patterns and predict events, many researchers are using machine learning models to predict weather conditions, weather anomalies and natural disasters. Past studies have shown the multidimensional application of artificial neural networks (ANNs) in precipitation prediction and promising success in accurately predicting rainfall in advance [ 5 – 6 ]. Fuzzy logic systems combined with neural network models are being used by researchers for effectively forecasting flood events [ 7 – 8 ]. 1.1 Artificial Neural Networks (ANNs) A different approach to flow forecasting has been created in recent years, relying on the use of artificial neural networks (ANNs) [ 9 ]. The ANN is an ML algorithm that is created to simulate the functions of the human brain and its capacity to acquire new skills [ 10 ]. ANNs stand out from other types of computer intelligence by not being rule-based, such as expert systems. Instead, ANNs are designed to learn and generalize the connections between inputs and outputs. The concept of ANNs was initially influenced by the functioning of the human brain. Recent advancements in ANN technology have transformed it into a practical mathematical tool that shares certain resemblances with the human brain. ANNs maintain two key brain-like traits: the capacity to learn and the ability to generalize from incomplete data [ 11 ]. The architectures of ANNs, such as the backpropagation neural network (BPN), multilayer perceptron (MLP), and feedforward neural network (FFNN), have been noted for their effectiveness in predicting chaotic phenomena and demonstrating considerable efficiency in forecasting rainfall [ 12 ]. Another study demonstrated that artificial neural networks (ANNs) are efficient in short-term rainfall prediction when equipped with optimal spatial and temporal factors. The reduction in lag time and the utilization of suitable spatial variables contribute to improvements in the precision of short-term forecasts [ 13 ]. 1.2 Adaptive neuro-fuzzy inference system (ANFIS) The root of the fuzzy logic approach dates back to 1965, when Lotfi Zadeh presented the fuzzy set hypothesis and its applications. Since that point, the fuzzy logic concept has been used in a wide range of applications in different spaces, such as estimation, prediction, forecasting, control, inexact thinking, design acknowledgment, restorative computing, mechanical technology, optimization, and mechanical design [ 14 – 15 ]. AI-based models, such as the ANFIS model developed via a hybrid training algorithm, have potential for predicting floods and providing useful techniques for flood control departments [ 16 ]. Hence, the utilization of ANFIS is highly appropriate for current flood forecasting studies. 1.3 Hybrid Model Benefits The integration of hybrid models, particularly the combination of artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFIS), represents a significant advance in flood forecasting methods. ANNs, machine learning algorithms inspired by the functionality of human brain-inspired methods, are known for their ability to learn and generalize from incomplete datasets, making them well suited for modeling and predicting complex, chaotic phenomena such as precipitation patterns [ 17 ]. The hybrid model leverages the flexibility of ANFIS and the learning capacity of ANNs to provide a robust predictive tool for flood forecasting, making it ideal for the precise needs of flood control departments [ 18 ]. Together, these hybrid models represent a convergence of computational intelligence techniques that improve predictive capabilities in flood forecasting, supporting proactive flooding. 2. Methodology Figure 1 illustrates the method employed to complete the entirety of the research via artificial neural networks. To select the ideal combination for the ANN model, we first gathered the historical rainfall data and preprocessed it. 2.1 Data collection A comprehensive dataset comprising 32 years of historical rainfall data, ranging from 1990–2022, is utilized for this study [ 19 ]. This dataset was collected from the Power Data Access Viewer. The POWER meteorological data are predictions or observations given by NASA's GMAO MERRA-2 assimilation model [ 20 ]. This dataset encompasses a total of 12,054 data points, each representing daily records. The dataset includes three key meteorological variables, namely, temperature, humidity, and historical rainfall data, which are selected as input features for the prediction model. The target variable is the daily precipitation amount, measured in millimeters. To develop and evaluate our prediction model, the dataset was split into three subsets: 70% of the data (8,438 records) were designated for training, 15% (1,808 records) for validation, and the remaining 15% (1,808 records) for testing. This division ensures that the model can be trained, validated, and tested effectively, facilitating robust performance evaluation. 2.2 Artificial neural networks In this study, an artificial neural network (ANN) is meticulously designed as a feed-forward backpropagation network. This choice is driven by the model’s capacity to efficiently learn and map complex nonlinear relationships from input data to outputs, making it particularly well suited for the variability inherent in meteorological phenomena. The ANN model is built with three carefully chosen input neurons, each representing a key predictive feature: temperature, humidity, and historical rainfall data. These features were selected on the basis of their strong correlation with rainfall patterns, as identified through extensive data analysis. The incorporation of these features enables the network to capture a comprehensive snapshot of the atmospheric conditions that influence rainfall. Figure 2 illustrates the network architecture, which includes two hidden layers with eight neurons. The rationale for this configuration is rooted in the need to balance model complexity with computational efficiency and generalizability. By employing two hidden layers, the network can progressively abstract and capture higher-order features from the input data, while the eight neurons in each layer ensure adequate capacity for learning without overfitting. This architectural design ensures the model’s robustness and predictive accuracy, providing a reliable tool for forecasting rainfall under varying climatic conditions. In the first hidden layer, we employed the log-sigmoid (LOGSIG) transfer function, which maps input values into a range between 0 and 1, enabling the modeling of nonlinear relationships between the input features. The second hidden layer uses the pure linear (PURELIN) transfer function, maintaining the linear characteristics of the data as they progress through the network. The output layer consists of a single neuron that generates the predicted precipitation value. As shown in Fig. 2 , the network was trained via the Levenberg‒Marquardt (TRAINLM) algorithm, which is well regarded for its efficiency and quick convergence. During the training process, the weights and biases were adjusted via the gradient descent with momentum (LEARNGDM) adaptation learning function. To evaluate the model's performance during training, we used the mean squared error (MSE) function, which measures the average squared difference between the predicted and actual precipitation values. Throughout the training phase, 15% of the dataset was used for validation to monitor and refine the model, helping to prevent overfitting. The model's generalization ability was subsequently tested on the remaining 15% of the data, ensuring its reliability in predicting precipitation from unseen data. 2.3 Adaptive neuro-fuzzy inference system (ANFIS) For the flood forecasting component of this study, we integrated a Mamdani-type fuzzy inference system (FIS). This system is adept at managing uncertainties and incorporating expert knowledge into the forecasting process. As illustrated in Fig. 3 , the FIS uses two primary inputs: the predicted rainfall from the ANN model and the observed river level, both of which are critical indicators of flood potential. The fuzzy sets for these inputs were defined via Triangular (TRIMF) and Trapezoidal (TRAPMF) membership functions. Figures 4 & 5 visualize the application of the triangular membership functions to represent typical and moderate input ranges, whereas Fig. 6 portrays the application of the trapezoidal functions for extended ranges, capturing greater variability in the data. These functions provide a robust mechanism to handle the imprecision inherent in meteorological and hydrological data. A total set of twelve fuzzy rules was formulated that map the input variables—predicted rainfall and river level—to the output variable, which is the flood risk level. The output of the FIS is a degree of certainty regarding the likelihood of flooding, which is used to assess flood risk levels. 3. Results and Discussion A number of metrics, such as the correlation coefficient (R) and mean square error (MSE), were used to assess the performance of the ANN model. On the basis of a comparison of the actual and predicted rainfall data for Sylhet, Bangladesh, these metrics were computed. The performance assessment shows promising results with respect to the accuracy of the ANN model in predicting rainfall data, which was later used in the fuzzy logic model to forecast floods of different severity levels. 3.1 Performance of the artificial neural network (ANN) model Figures 7 , 8 , 9 , and 10 present the correlation coefficient values for the datasets utilized in the model, including the training, test, and validation phases. Specifically, the correlation coefficients are 0.93 for the training dataset, 0.91 for the test dataset, and 0.94 for the validation dataset. These values reflect a robust positive correlation between the predicted and actual values throughout all stages of the model's evaluation. The correlation coefficient (r) measures the strength and direction of the linear relationship between two variables, with values ranging from − 1 to 1. A coefficient close to 1 indicates a strong positive linear relationship, where the predicted values closely align with the actual values. In this study, the high correlation coefficients observed across the datasets—0.93, 0.91, and 0.94—demonstrate a high degree of concordance between the model's predictions and the actual observed data. The strong positive correlation across these datasets signifies that the artificial neural network (ANN) model effectively captures the underlying patterns and relationships within the data. The consistency of these high correlation values across different datasets (training, test, and validation) underscores the model's reliability and generalizability to new, unseen data. As shown in Fig. 11, the ANN model produced a satisfactory MSE curve, which shows that the model's predictions and the actual rainfall values were fairly close. The model appears to be highly accurate in capturing the patterns found in the rainfall data, as indicated by the low mean square error (MSE). 3.2 Performance of the ANFIS On the basis of inputs such as river level and the anticipated rainfall data from the ANN model, the fuzzy logic component was created to forecast flood events. Twelve fuzzy rules were used by the Mamdani-type fuzzy inference system (FIS) in this study to interpret the input data and produce an assessment of flood risk. Inputs: The system utilized two main inputs: predicted rainfall and river water levels. These inputs were translated into fuzzy sets via triangular (TRIMF) and trapezoidal (TRAPMF) membership functions. Output: The output was the flood risk level, which was categorized into various levels, such as no flood, moderate flood, minor flood and major flood. Figure 12 shows the correlations among the river level, rainfall level and flood level. The x-axis shows the river level, the y-axis shows rainfall, and the z-axis represents the flood level. The surface is divided into color-coded regions, and the surface height represents the flood level for a given river level and rainfall. The plot above shows how the river level combined with rainfall impacts the extent of flooding, whereby higher regions indicate increased flood levels. Figure 13 provides a comprehensive representation of flood conditions by correlating rainfall intensity with corresponding river levels. The figure categorizes precipitation into three different levels: low, moderate, and heavy. Each category is further analyzed in conjunction with river levels classified as normal, cautious, warning, and danger. The diagram provides a nuanced understanding of how these variables interact to produce specific flooding outcomes. In particular, the resulting flooding for various combinations of rainfall and river levels can be classified into the categories of no flooding, minor flooding, moderate flooding, and severe flooding. This visual representation highlights the dynamic relationship between meteorological and hydrological factors in determining flood severity. 4. Conclusion This study critically examines rainfall prediction through a hybrid model that integrates artificial neural networks (ANNs) and fuzzy logic to forecast flooding. We categorized flood risk into four levels—minor, moderate, major, and no flood—using rainfall and water level data. The innovative combination of ANNs and fuzzy logic yielded promising outcomes, establishing a robust framework for rapid and precise flood forecasting. The primary aim was to improve the accuracy and timeliness of flood predictions. While somewhat effective, traditional methods often lack the refined accuracy offered by modern hybrid techniques. Our results indicated a correlation in ANNs exceeding 0.93, reflecting a strong predictive capability and the ANN's adeptness at modeling complex nonlinear environmental relationships. Furthermore, the favorable mean squared error (MSE) curve highlights the model’s effectiveness in minimizing prediction errors. The incorporation of fuzzy logic enhanced the sophistication of our model. Fuzzy logic's capacity to manage imprecise and uncertain information has proven particularly beneficial for flood forecasting, where conditions may fluctuate rapidly. By establishing fuzzy sets and rules, we categorized potential flood events, enabling a more detailed and realistic evaluation of flood risk. The practical implications of this research are considerable. The ability to predict floods swiftly and accurately allows authorities and communities to take proactive measures to mitigate flood impacts, ultimately preserving lives and reducing property damage. The efficacy of the hybrid model suggests that it could serve as a viable alternative or complement to current flood forecasting systems, particularly in regions prone to severe flooding. Nonetheless, this study has its limitations. A significant gap identified is the absence of discharge data from Cherrapunji, an area known for heavy rainfall, which could enhance the model's predictive accuracy. References (PDF) Natural Disasters and Management Systems of Bangladesh from 1972 to 2017: Special Focus on Flood. Accessed: Jul. 01, 2024. [Online]. Available: https://www.researchgate.net/publication/330360643_Natural_Disasters_and_Management_Systems_of_Bangladesh_from_1972_to_2017_Special_Focus_on_Flood A-M D, Nishigaki M, Komatsu M (2003) Floods in Bangladesh. A Comparative Hydrological Investigation on Two Catastrophic Events Community-led Housing Recovery Assessment_Bangladesh Flood 2022, Accessed: Jul. 06, 2024. [Online]. Available: https://www.undp.org/sites/g/files/zskgke326/files/2022-12/Community-led%20Housing%20Recovery%20Assessment_Bangladesh%20Flood%202022.pdf Nevo S et al Flood forecasting with machine learning models in an operational framework, 10.5194/hess-2021-554 Lesnussa YA, Mustamu CG, Kondo Lembang F, Talakua MW, APPLICATION OF BACKPROPAGATION NEURAL NETWORKS IN PREDICTING RAINFALL DATA IN AMBON CITY (Aug. 2018) Int J Artif Intell Res 2(2). 10.29099/ijair.v2i2.59 Yadav P, Sagar A (Mar. 2019) Rainfall Prediction Using Artificial Neural Network (ANN) for tarai Region of Uttarakhand. Curr J Appl Sci Technol 1–7. 10.9734/cjast/2019/v33i530096 Mukerji A, Chatterjee C, Raghuwanshi NS, Using FF, Neuro-Fuzzy ANN, Models N-GA (Jun. 2009) J Hydrol Eng 14(6):647–652. 10.1061/(ASCE)HE.1943-5584.0000040 Janál P, Kozel T, FUZZY LOGIC BASED FLASH FLOOD FORECAST (2020), in Proceedings of the XXVII Conference of the Danubian Countries on Hydrological Forecasting and Hydrological Bases of Water Management , Nika-Tsentr, Oct. 10.15407/uhmi.conference.01.10 Rao AR, Govindaraju RS (2000) Artificial neural networks in hydrology. Kluwer Academic Elsafi SH (2014) Artificial neural networks (ANNs) for flood forecasting at Dongola Station in the River Nile, Sudan. Alexandria Eng J 53(3):655–662 Hewitson Bruce RG C. and, Crane (1994) Precipitation Controls in Southern Mexico. In: Hewitson Bruce RG C. and, Crane (eds) in Neural Nets: Applications in Geography. Springer Netherlands, Dordrecht, pp 121–143. doi: 10.1007/978-94-011-1122-5_7 . Lakshminarayana SV (2020) Rainfall Forecasting using Artificial Neural Networks (ANNs): A Comprehensive Literature Review, Indian Journal of Pure & Applied Biosciences , vol. 8, no. 4, pp. 589–599, Aug. 10.18782/2582-2845.8250 Luk KC, Ball JE, Sharma A (Jan. 2000) A study of optimal model lag and spatial inputs to artificial neural network for rainfall forecasting. J Hydrol (Amst) 227:1–4. 10.1016/S0022-1694(99)00165-1 Matreata M Overview of the artificial neural networks and fuzzy logic applications in operational hydrological forecasting systems Rath S, Nayak PC, Chatterjee C (2013) Hierarchical neurofuzzy model for real-time flood forecasting, International Journal of River Basin Management , vol. 11, no. 3, pp. 253–268, Sep. 10.1080/15715124.2013.798329 Tabbussum R, Dar AQ (May 2021) Performance evaluation of artificial intelligence paradigms—artificial neural networks, fuzzy logic, and adaptive neuro-fuzzy inference system for flood prediction. Environ Sci Pollut Res 28(20):25265–25282. 10.1007/s11356-021-12410-1 Bihl T, Young WA, Moyer A, Frimel S (Oct. 2022) Artificial Neural Networks and Data Science. Encyclopedia Data Sci Mach Learn 899–921. 10.4018/978-1-7998-9220-5.CH052 Joyice MH, Katta. Valli Sri Vidya, Lankalapalli. Vijaya Lakshmi, Murala. Juilath, Ketha. Prajwala, and, Vasarao PS (2024) Identifying Flood Prediction using Machine Learning Techniques, Int J Innov Sci Res Technol , pp. 144–147, Mar. 10.38124/IJISRT/IJISRT24MAR112 Weather Parameter Dataset Accessed: Aug. 12, 2024. [Online]. Available: https://www.kaggle.com/datasets/mdnurhossen/weather-report-in-bangladesh-1990-2022 POWER | DAV. Accessed Aug. 13, 2024. [Online]. Available: https://power.larc.nasa.gov/data-access-viewer/ Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6882638","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":470537478,"identity":"f3e2909d-b46b-4f73-a9c8-1417ce89ef35","order_by":0,"name":"Mayaz Uddin Gazi","email":"data:image/png;base64,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","orcid":"https://orcid.org/0009-0004-3221-3286","institution":"Sylhet Engineering College","correspondingAuthor":true,"prefix":"","firstName":"Mayaz","middleName":"Uddin","lastName":"Gazi","suffix":""}],"badges":[],"createdAt":"2025-06-12 18:19:09","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-6882638/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6882638/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":84521630,"identity":"543a8f3e-6b78-4b58-b041-89f46a309d23","added_by":"auto","created_at":"2025-06-13 03:48:47","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":82487,"visible":true,"origin":"","legend":"\u003cp\u003eANN diagram\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/8a432dc83bcf2d0d46d2a432.png"},{"id":84522352,"identity":"9698c975-1229-4992-8832-5d2aedf96779","added_by":"auto","created_at":"2025-06-13 04:04:47","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":140651,"visible":true,"origin":"","legend":"\u003cp\u003eANN training\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/4a5c80669fe0841c70484617.png"},{"id":84522239,"identity":"c57e1fa8-0dd2-4562-ad98-a0951efcd10a","added_by":"auto","created_at":"2025-06-13 03:56:47","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":117965,"visible":true,"origin":"","legend":"\u003cp\u003eRainfall Range\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/5fd381103312f57278ee50e2.png"},{"id":84522244,"identity":"d6e70c58-9376-4619-a86c-9c98340bca13","added_by":"auto","created_at":"2025-06-13 03:56:47","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":128498,"visible":true,"origin":"","legend":"\u003cp\u003eMembership Functions\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/983550482c35086d7c5eae23.png"},{"id":84523124,"identity":"9e4d593d-c3cb-4b60-8ded-68606648e788","added_by":"auto","created_at":"2025-06-13 04:12:47","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":135643,"visible":true,"origin":"","legend":"\u003cp\u003eFIS input: river level\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/d9b4cd4696ce7c0d4089b367.png"},{"id":84521635,"identity":"39aac1af-35fd-42ea-9591-31b85477423c","added_by":"auto","created_at":"2025-06-13 03:48:47","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":123770,"visible":true,"origin":"","legend":"\u003cp\u003eFIS output: Flood\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/874bf78091789816985cb2bb.png"},{"id":84521636,"identity":"2fd10ab0-9554-4ec8-bf4e-4367a7019439","added_by":"auto","created_at":"2025-06-13 03:48:47","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":27571,"visible":true,"origin":"","legend":"\u003cp\u003eTraining Result\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/d62fda62356248dc03e203ff.png"},{"id":84522354,"identity":"641ae253-b98d-475c-aebb-66f5b43ddb6a","added_by":"auto","created_at":"2025-06-13 04:04:47","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":26888,"visible":true,"origin":"","legend":"\u003cp\u003eTesting Result\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/28f9283702fb82ef5ef228a4.png"},{"id":84521642,"identity":"1d330f06-c0eb-477c-9227-34ce2552cfcf","added_by":"auto","created_at":"2025-06-13 03:48:47","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":27096,"visible":true,"origin":"","legend":"\u003cp\u003eValidation Result\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/e64ec43914f5a7692b51b0b2.png"},{"id":84521643,"identity":"ab1ce6bd-3dbf-472e-b61b-45e7028763d3","added_by":"auto","created_at":"2025-06-13 03:48:47","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":91720,"visible":true,"origin":"","legend":"\u003cp\u003eOverall Result\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/a2c3a4c083bdefe09d7614ae.png"},{"id":84521637,"identity":"b2de2421-0833-4c20-ac7f-c73e4eff4b64","added_by":"auto","created_at":"2025-06-13 03:48:47","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":39041,"visible":true,"origin":"","legend":"\u003cp\u003eMean square error curve (MSE)\u003c/p\u003e","description":"","filename":"11.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/eeeab27eae14cd9c4bdf56b2.png"},{"id":84521647,"identity":"1ff19137-5eae-403f-a306-f00c4066a78d","added_by":"auto","created_at":"2025-06-13 03:48:48","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":60358,"visible":true,"origin":"","legend":"\u003cp\u003e3D visualization of the fuzzy logic model\u003c/p\u003e","description":"","filename":"12.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/70aa7017d6180f87f132702c.png"},{"id":84522247,"identity":"87e755fb-a7da-4c83-ac0f-1d2c8459b66d","added_by":"auto","created_at":"2025-06-13 03:56:47","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":138712,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eFuzzy logic results based on inputs\u003c/em\u003e\u003c/p\u003e","description":"","filename":"13.png","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/b3d85132a374001b9ff59799.png"},{"id":84523324,"identity":"59a6694e-26e8-44ed-8195-dd38675996db","added_by":"auto","created_at":"2025-06-13 04:20:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1468021,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6882638/v1/4a935daa-c3f2-4c17-b856-f9e900b8275b.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eFlood Prediction and Forecasting Using Anns and Fuzzy Logic Model in Sylhet, Bangladesh\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eAs a riverine country with complex weather, floods are common during the monsoon season in this country. From 1972\u0026ndash;2017, Bangladesh faced 86 flood events [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Excessive rainfall in the monsoon season is one of the most prominent causes of frequent flood occurrence [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. As Sylhet is situated at the banks of the Surma and Kushiyara rivers, floods have become a frequent natural disaster for Sylhet in recent years. In 2022, Sylhet faced the worst flood conditions in 122 years, which affected more than 7.2\u0026nbsp;million people and caused an overall economic loss of 722.24 dollars in different sectors [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTherefore, for flood forecasting and forecasting, it is necessary to improve resource allocation and response coordination and develop flood management techniques. Machine learning (ML) and deep learning with fuzzy logic offer the ability to provide instant and up-to-date information. ML models have the ability to offer real-time flood predictions and multistep-ahead forecasts, aspects that are deemed essential for providing timely warnings and effective disaster management [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAs machine learning has shown outstanding ability to process data, detect patterns and predict events, many researchers are using machine learning models to predict weather conditions, weather anomalies and natural disasters. Past studies have shown the multidimensional application of artificial neural networks (ANNs) in precipitation prediction and promising success in accurately predicting rainfall in advance [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Fuzzy logic systems combined with neural network models are being used by researchers for effectively forecasting flood events [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Artificial Neural Networks (ANNs)\u003c/h2\u003e \u003cp\u003eA different approach to flow forecasting has been created in recent years, relying on the use of artificial neural networks (ANNs) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The ANN is an ML algorithm that is created to simulate the functions of the human brain and its capacity to acquire new skills [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. ANNs stand out from other types of computer intelligence by not being rule-based, such as expert systems. Instead, ANNs are designed to learn and generalize the connections between inputs and outputs. The concept of ANNs was initially influenced by the functioning of the human brain.\u003c/p\u003e \u003cp\u003eRecent advancements in ANN technology have transformed it into a practical mathematical tool that shares certain resemblances with the human brain. ANNs maintain two key brain-like traits: the capacity to learn and the ability to generalize from incomplete data [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The architectures of ANNs, such as the backpropagation neural network (BPN), multilayer perceptron (MLP), and feedforward neural network (FFNN), have been noted for their effectiveness in predicting chaotic phenomena and demonstrating considerable efficiency in forecasting rainfall [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAnother study demonstrated that artificial neural networks (ANNs) are efficient in short-term rainfall prediction when equipped with optimal spatial and temporal factors. The reduction in lag time and the utilization of suitable spatial variables contribute to improvements in the precision of short-term forecasts [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.2 Adaptive neuro-fuzzy inference system (ANFIS)\u003c/h2\u003e \u003cp\u003eThe root of the fuzzy logic approach dates back to 1965, when Lotfi Zadeh presented the fuzzy set hypothesis and its applications. Since that point, the fuzzy logic concept has been used in a wide range of applications in different spaces, such as estimation, prediction, forecasting, control, inexact thinking, design acknowledgment, restorative computing, mechanical technology, optimization, and mechanical design [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAI-based models, such as the ANFIS model developed via a hybrid training algorithm, have potential for predicting floods and providing useful techniques for flood control departments [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Hence, the utilization of ANFIS is highly appropriate for current flood forecasting studies.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e1.3 Hybrid Model Benefits\u003c/h2\u003e \u003cp\u003eThe integration of hybrid models, particularly the combination of artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFIS), represents a significant advance in flood forecasting methods. ANNs, machine learning algorithms inspired by the functionality of human brain-inspired methods, are known for their ability to learn and generalize from incomplete datasets, making them well suited for modeling and predicting complex, chaotic phenomena such as precipitation patterns [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The hybrid model leverages the flexibility of ANFIS and the learning capacity of ANNs to provide a robust predictive tool for flood forecasting, making it ideal for the precise needs of flood control departments [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Together, these hybrid models represent a convergence of computational intelligence techniques that improve predictive capabilities in flood forecasting, supporting proactive flooding.\u003c/p\u003e \u003c/div\u003e"},{"header":"2. Methodology","content":"\u003cp\u003eFigure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the method employed to complete the entirety of the research via artificial neural networks. To select the ideal combination for the ANN model, we first gathered the historical rainfall data and preprocessed it.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n\u003ch2\u003e2.1 Data collection\u003c/h2\u003e\n\u003cp\u003eA comprehensive dataset comprising 32 years of historical rainfall data, ranging from 1990\u0026ndash;2022, is utilized for this study [\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThis dataset was collected from the Power Data Access Viewer. The POWER meteorological data are predictions or observations given by NASA's GMAO MERRA-2 assimilation model [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e]. This dataset encompasses a total of 12,054 data points, each representing daily records. The dataset includes three key meteorological variables, namely, temperature, humidity, and historical rainfall data, which are selected as input features for the prediction model. The target variable is the daily precipitation amount, measured in millimeters.\u003c/p\u003e\n\u003cp\u003eTo develop and evaluate our prediction model, the dataset was split into three subsets: 70% of the data (8,438 records) were designated for training, 15% (1,808 records) for validation, and the remaining 15% (1,808 records) for testing. This division ensures that the model can be trained, validated, and tested effectively, facilitating robust performance evaluation.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n\u003ch2\u003e2.2 Artificial neural networks\u003c/h2\u003e\n\u003cp\u003eIn this study, an artificial neural network (ANN) is meticulously designed as a feed-forward backpropagation network. This choice is driven by the model\u0026rsquo;s capacity to efficiently learn and map complex nonlinear relationships from input data to outputs, making it particularly well suited for the variability inherent in meteorological phenomena.\u003c/p\u003e\n\u003cp\u003eThe ANN model is built with three carefully chosen input neurons, each representing a key predictive feature: temperature, humidity, and historical rainfall data. These features were selected on the basis of their strong correlation with rainfall patterns, as identified through extensive data analysis. The incorporation of these features enables the network to capture a comprehensive snapshot of the atmospheric conditions that influence rainfall.\u003c/p\u003e\n\u003cp\u003eFigure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the network architecture, which includes two hidden layers with eight neurons. The rationale for this configuration is rooted in the need to balance model complexity with computational efficiency and generalizability. By employing two hidden layers, the network can progressively abstract and capture higher-order features from the input data, while the eight neurons in each layer ensure adequate capacity for learning without overfitting. This architectural design ensures the model\u0026rsquo;s robustness and predictive accuracy, providing a reliable tool for forecasting rainfall under varying climatic conditions.\u003c/p\u003e\n\u003cp\u003eIn the first hidden layer, we employed the log-sigmoid (LOGSIG) transfer function, which maps input values into a range between 0 and 1, enabling the modeling of nonlinear relationships between the input features. The second hidden layer uses the pure linear (PURELIN) transfer function, maintaining the linear characteristics of the data as they progress through the network. The output layer consists of a single neuron that generates the predicted precipitation value.\u003c/p\u003e\n\u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e, the network was trained via the Levenberg‒Marquardt (TRAINLM) algorithm, which is well regarded for its efficiency and quick convergence. During the training process, the weights and biases were adjusted via the gradient descent with momentum (LEARNGDM) adaptation learning function. To evaluate the model's performance during training, we used the mean squared error (MSE) function, which measures the average squared difference between the predicted and actual precipitation values.\u003c/p\u003e\n\u003cp\u003eThroughout the training phase, 15% of the dataset was used for validation to monitor and refine the model, helping to prevent overfitting. The model's generalization ability was subsequently tested on the remaining 15% of the data, ensuring its reliability in predicting precipitation from unseen data.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n\u003ch2\u003e2.3 Adaptive neuro-fuzzy inference system (ANFIS)\u003c/h2\u003e\n\u003cp\u003eFor the flood forecasting component of this study, we integrated a Mamdani-type fuzzy inference system (FIS). This system is adept at managing uncertainties and incorporating expert knowledge into the forecasting process. As illustrated in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, the FIS uses two primary inputs: the predicted rainfall from the ANN model and the observed river level, both of which are critical indicators of flood potential.\u003c/p\u003e\n\u003cp\u003eThe fuzzy sets for these inputs were defined via Triangular (TRIMF) and Trapezoidal (TRAPMF) membership functions. Figures\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e \u0026amp; \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e visualize the application of the triangular membership functions to represent typical and moderate input ranges, whereas Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e portrays the application of the trapezoidal functions for extended ranges, capturing greater variability in the data.\u003c/p\u003e\n\u003cp\u003eThese functions provide a robust mechanism to handle the imprecision inherent in meteorological and hydrological data.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA total set of twelve fuzzy rules was formulated that map the input variables\u0026mdash;predicted rainfall and river level\u0026mdash;to the output variable, which is the flood risk level. The output of the FIS is a degree of certainty regarding the likelihood of flooding, which is used to assess flood risk levels.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Results and Discussion","content":"\u003cp\u003eA number of metrics, such as the correlation coefficient (R) and mean square error (MSE), were used to assess the performance of the ANN model. On the basis of a comparison of the actual and predicted rainfall data for Sylhet, Bangladesh, these metrics were computed. The performance assessment shows promising results with respect to the accuracy of the ANN model in predicting rainfall data, which was later used in the fuzzy logic model to forecast floods of different severity levels.\u003c/p\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n\u003ch2\u003e3.1 Performance of the artificial neural network (ANN) model\u003c/h2\u003e\n\u003cp\u003eFigures\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e, \u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e, \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e, and \u003cspan class=\"InternalRef\"\u003e10\u003c/span\u003e present the correlation coefficient values for the datasets utilized in the model, including the training, test, and validation phases. Specifically, the correlation coefficients are 0.93 for the training dataset, 0.91 for the test dataset, and 0.94 for the validation dataset. These values reflect a robust positive correlation between the predicted and actual values throughout all stages of the model's evaluation.\u003c/p\u003e\n\u003cp\u003eThe correlation coefficient (r) measures the strength and direction of the linear relationship between two variables, with values ranging from \u0026minus;\u0026thinsp;1 to 1. A coefficient close to 1 indicates a strong positive linear relationship, where the predicted values closely align with the actual values. In this study, the high correlation coefficients observed across the datasets\u0026mdash;0.93, 0.91, and 0.94\u0026mdash;demonstrate a high degree of concordance between the model's predictions and the actual observed data.\u003c/p\u003e\n\u003cp\u003eThe strong positive correlation across these datasets signifies that the artificial neural network (ANN) model effectively captures the underlying patterns and relationships within the data. The consistency of these high correlation values across different datasets (training, test, and validation) underscores the model's reliability and generalizability to new, unseen data.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs shown in Fig.\u0026nbsp;11, the ANN model produced a satisfactory MSE curve, which shows that the model's predictions and the actual rainfall values were fairly close. The model appears to be highly accurate in capturing the patterns found in the rainfall data, as indicated by the low mean square error (MSE).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n\u003ch2\u003e3.2 Performance of the ANFIS\u003c/h2\u003e\n\u003cp\u003eOn the basis of inputs such as river level and the anticipated rainfall data from the ANN model, the fuzzy logic component was created to forecast flood events. Twelve fuzzy rules were used by the Mamdani-type fuzzy inference system (FIS) in this study to interpret the input data and produce an assessment of flood risk.\u003c/p\u003e\n\u003cp\u003eInputs: The system utilized two main inputs: predicted rainfall and river water levels. These inputs were translated into fuzzy sets via triangular (TRIMF) and trapezoidal (TRAPMF) membership functions.\u003c/p\u003e\n\u003cp\u003eOutput: The output was the flood risk level, which was categorized into various levels, such as no flood, moderate flood, minor flood and major flood.\u003c/p\u003e\n\u003cp\u003eFigure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e12\u003c/span\u003e shows the correlations among the river level, rainfall level and flood level. The x-axis shows the river level, the y-axis shows rainfall, and the z-axis represents the flood level. The surface is divided into color-coded regions, and the surface height represents the flood level for a given river level and rainfall. The plot above shows how the river level combined with rainfall impacts the extent of flooding, whereby higher regions indicate increased flood levels.\u003c/p\u003e\n\u003cp\u003eFigure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e13\u003c/span\u003e provides a comprehensive representation of flood conditions by correlating rainfall intensity with corresponding river levels. The figure categorizes precipitation into three different levels: low, moderate, and heavy. Each category is further analyzed in conjunction with river levels classified as normal, cautious, warning, and danger. The diagram provides a nuanced understanding of how these variables interact to produce specific flooding outcomes. In particular, the resulting flooding for various combinations of rainfall and river levels can be classified into the categories of no flooding, minor flooding, moderate flooding, and severe flooding. This visual representation highlights the dynamic relationship between meteorological and hydrological factors in determining flood severity.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eThis study critically examines rainfall prediction through a hybrid model that integrates artificial neural networks (ANNs) and fuzzy logic to forecast flooding. We categorized flood risk into four levels\u0026mdash;minor, moderate, major, and no flood\u0026mdash;using rainfall and water level data. The innovative combination of ANNs and fuzzy logic yielded promising outcomes, establishing a robust framework for rapid and precise flood forecasting.\u003c/p\u003e \u003cp\u003eThe primary aim was to improve the accuracy and timeliness of flood predictions. While somewhat effective, traditional methods often lack the refined accuracy offered by modern hybrid techniques. Our results indicated a correlation in ANNs exceeding 0.93, reflecting a strong predictive capability and the ANN's adeptness at modeling complex nonlinear environmental relationships. Furthermore, the favorable mean squared error (MSE) curve highlights the model\u0026rsquo;s effectiveness in minimizing prediction errors.\u003c/p\u003e \u003cp\u003eThe incorporation of fuzzy logic enhanced the sophistication of our model. Fuzzy logic's capacity to manage imprecise and uncertain information has proven particularly beneficial for flood forecasting, where conditions may fluctuate rapidly. By establishing fuzzy sets and rules, we categorized potential flood events, enabling a more detailed and realistic evaluation of flood risk.\u003c/p\u003e \u003cp\u003eThe practical implications of this research are considerable. The ability to predict floods swiftly and accurately allows authorities and communities to take proactive measures to mitigate flood impacts, ultimately preserving lives and reducing property damage. The efficacy of the hybrid model suggests that it could serve as a viable alternative or complement to current flood forecasting systems, particularly in regions prone to severe flooding.\u003c/p\u003e \u003cp\u003eNonetheless, this study has its limitations. A significant gap identified is the absence of discharge data from Cherrapunji, an area known for heavy rainfall, which could enhance the model's predictive accuracy.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e(PDF) Natural Disasters and Management Systems of Bangladesh from 1972 to 2017: Special Focus on Flood. Accessed: Jul. 01, 2024. [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.researchgate.net/publication/330360643_Natural_Disasters_and_Management_Systems_of_Bangladesh_from_1972_to_2017_Special_Focus_on_Flood\u003c/span\u003e\u003cspan address=\"https://www.researchgate.net/publication/330360643_Natural_Disasters_and_Management_Systems_of_Bangladesh_from_1972_to_2017_Special_Focus_on_Flood\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eA-M D, Nishigaki M, Komatsu M (2003) Floods in Bangladesh. A Comparative Hydrological Investigation on Two Catastrophic Events\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCommunity-led Housing Recovery Assessment_Bangladesh Flood 2022, Accessed: Jul. 06, 2024. [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.undp.org/sites/g/files/zskgke326/files/2022-12/Community-led%20Housing%20Recovery%20Assessment_Bangladesh%20Flood%202022.pdf\u003c/span\u003e\u003cspan address=\"https://www.undp.org/sites/g/files/zskgke326/files/2022-12/Community-led%20Housing%20Recovery%20Assessment_Bangladesh%20Flood%202022.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNevo S et al Flood forecasting with machine learning models in an operational framework, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.5194/hess-2021-554\u003c/span\u003e\u003cspan address=\"10.5194/hess-2021-554\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLesnussa YA, Mustamu CG, Kondo Lembang F, Talakua MW, APPLICATION OF BACKPROPAGATION NEURAL NETWORKS IN PREDICTING RAINFALL DATA IN AMBON CITY (Aug. 2018) Int J Artif Intell Res 2(2). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.29099/ijair.v2i2.59\u003c/span\u003e\u003cspan address=\"10.29099/ijair.v2i2.59\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYadav P, Sagar A (Mar. 2019) Rainfall Prediction Using Artificial Neural Network (ANN) for tarai Region of Uttarakhand. Curr J Appl Sci Technol 1\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.9734/cjast/2019/v33i530096\u003c/span\u003e\u003cspan address=\"10.9734/cjast/2019/v33i530096\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMukerji A, Chatterjee C, Raghuwanshi NS, Using FF, Neuro-Fuzzy ANN, Models N-GA (Jun. 2009) J Hydrol Eng 14(6):647\u0026ndash;652. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1061/(ASCE)HE.1943-5584.0000040\u003c/span\u003e\u003cspan address=\"10.1061/(ASCE)HE.1943-5584.0000040\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJan\u0026aacute;l P, Kozel T, FUZZY LOGIC BASED FLASH FLOOD FORECAST (2020), in \u003cem\u003eProceedings of the XXVII Conference of the Danubian Countries on Hydrological Forecasting and Hydrological Bases of Water Management\u003c/em\u003e, Nika-Tsentr, Oct. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.15407/uhmi.conference.01.10\u003c/span\u003e\u003cspan address=\"10.15407/uhmi.conference.01.10\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRao AR, Govindaraju RS (2000) Artificial neural networks in hydrology. Kluwer Academic\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElsafi SH (2014) Artificial neural networks (ANNs) for flood forecasting at Dongola Station in the River Nile, Sudan. Alexandria Eng J 53(3):655\u0026ndash;662\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHewitson Bruce RG C. and, Crane (1994) Precipitation Controls in Southern Mexico. In: Hewitson Bruce RG C. and, Crane (eds) in Neural Nets: Applications in Geography. Springer Netherlands, Dordrecht, pp 121\u0026ndash;143. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/978-94-011-1122-5_7\u003c/span\u003e\u003cspan address=\"10.1007/978-94-011-1122-5_7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLakshminarayana SV (2020) Rainfall Forecasting using Artificial Neural Networks (ANNs): A Comprehensive Literature Review, \u003cem\u003eIndian Journal of Pure \u0026amp; Applied Biosciences\u003c/em\u003e, vol. 8, no. 4, pp. 589\u0026ndash;599, Aug. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.18782/2582-2845.8250\u003c/span\u003e\u003cspan address=\"10.18782/2582-2845.8250\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuk KC, Ball JE, Sharma A (Jan. 2000) A study of optimal model lag and spatial inputs to artificial neural network for rainfall forecasting. J Hydrol (Amst) 227:1\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0022-1694(99)00165-1\u003c/span\u003e\u003cspan address=\"10.1016/S0022-1694(99)00165-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatreata M Overview of the artificial neural networks and fuzzy logic applications in operational hydrological forecasting systems\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRath S, Nayak PC, Chatterjee C (2013) Hierarchical neurofuzzy model for real-time flood forecasting, \u003cem\u003eInternational Journal of River Basin Management\u003c/em\u003e, vol. 11, no. 3, pp. 253\u0026ndash;268, Sep. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/15715124.2013.798329\u003c/span\u003e\u003cspan address=\"10.1080/15715124.2013.798329\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTabbussum R, Dar AQ (May 2021) Performance evaluation of artificial intelligence paradigms\u0026mdash;artificial neural networks, fuzzy logic, and adaptive neuro-fuzzy inference system for flood prediction. Environ Sci Pollut Res 28(20):25265\u0026ndash;25282. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11356-021-12410-1\u003c/span\u003e\u003cspan address=\"10.1007/s11356-021-12410-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBihl T, Young WA, Moyer A, Frimel S (Oct. 2022) Artificial Neural Networks and Data Science. Encyclopedia Data Sci Mach Learn 899\u0026ndash;921. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4018/978-1-7998-9220-5.CH052\u003c/span\u003e\u003cspan address=\"10.4018/978-1-7998-9220-5.CH052\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJoyice MH, Katta. Valli Sri Vidya, Lankalapalli. Vijaya Lakshmi, Murala. Juilath, Ketha. Prajwala, and, Vasarao PS (2024) Identifying Flood Prediction using Machine Learning Techniques, \u003cem\u003eInt J Innov Sci Res Technol\u003c/em\u003e, pp. 144\u0026ndash;147, Mar. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.38124/IJISRT/IJISRT24MAR112\u003c/span\u003e\u003cspan address=\"10.38124/IJISRT/IJISRT24MAR112\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeather Parameter Dataset Accessed: Aug. 12, 2024. [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.kaggle.com/datasets/mdnurhossen/weather-report-in-bangladesh-1990-2022\u003c/span\u003e\u003cspan address=\"https://www.kaggle.com/datasets/mdnurhossen/weather-report-in-bangladesh-1990-2022\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePOWER | DAV. Accessed Aug. 13, 2024. [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://power.larc.nasa.gov/data-access-viewer/\u003c/span\u003e\u003cspan address=\"https://power.larc.nasa.gov/data-access-viewer/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"No funding","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Flood forecasting, Artificial neural networks, Feed-forward backdrop, Fuzzy logic, Hybrid approach","lastPublishedDoi":"10.21203/rs.3.rs-6882638/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6882638/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFloods are a significant threat to the environment, human life, and infrastructure in Sylhet, Bangladesh. The region's unique geography and climate make it prone to frequent and severe floods, which can cause significant economic and social losses. To address this issue, this study proposes an intelligent flood management system that combines the power of artificial neural networks (ANNs) for rainfall prediction and fuzzy logic for flood forecasting. The system is designed to provide accurate and real-time predictions of rainfall and flood events, enabling effective decision-making and mitigation strategies. The ANNs are trained using historical rainfall data to predict future rainfall patterns, whereas the fuzzy logic component uses the predicted rainfall data to forecast flood events. The system was tested on real-world data from Sylhet and demonstrated high accuracy in predicting rainfall and flood events. The ANN model was developed via a feed-forward backprop network with three input variables and one output variable (rainfall). A TRAINLM training function, LEARNGDM adaptation learning function, and MSE performance function were used. The ANN architecture consisted of two hidden layers with eight neurons each, with LOGSIG and PURELIN transfer functions for the first and second layers, respectively. The fuzzy logic component employs a Mamdani-type fuzzy inference system (FIS) with twelve rules, using rainfall and river level as inputs and flooding as the output. Triangular (TRIMF) and trapezoidal (TRAPMF) membership functions were utilized. The results of the ANN model revealed a mean square error (MSE) with a suitable curve and a correlation coefficient (R) over 0.9, indicating a strong correlation between the predicted and actual values. Additionally, we obtain an outstanding mean absolute error (MAE) value. The hybrid approach combining ANNs and fuzzy logic demonstrated high accuracy in flood forecasting, outperforming traditional methods. The proposed system can provide early warning of flood events, enabling timely mitigation measures and reducing the impact on communities.\u003c/p\u003e","manuscriptTitle":"Flood Prediction and Forecasting Using Anns and Fuzzy Logic Model in Sylhet, Bangladesh","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-13 03:48:43","doi":"10.21203/rs.3.rs-6882638/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a6f08f8a-cdbb-4aff-b3ae-35d6b53d909f","owner":[],"postedDate":"June 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":49982476,"name":"Civil Engineering"}],"tags":[],"updatedAt":"2025-06-13T03:48:43+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-13 03:48:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6882638","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6882638","identity":"rs-6882638","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}