Flooding is a commonly occurring natural hazard on a global scale, resulting in significant negative impacts on human populations, infrastructure, ecological systems, and financial systems (Meshram et al., 2020). Floods have a terrible effect on the lives of millions of people every year and are generally considered natural disaster that poses the most significant risk to human life worldwide (Anusha & Bharathi, 2020). The Floods Directive of the United Nations defines a flood as "the temporary inundation of land that is typically not covered by water" (Floods Directive, 2011). As per the United Nations (UN), flooding has affected over 40% of the population by weather-related calamities (Amitrano et al., 2018). It is estimated that over 12,000 hydrological catastrophes have been registered worldwide between 1970 and 2015. These events resulted in the loss of over 3,530,000 human lives, impacting more than 6.7 billion individuals, and incurred damages exceeding US$ 2,600 billion (Qiu et al., 2021). Location, flood readiness, flood mitigation strategies, and response plans are the essential factors that impact the severity of flood disasters (Joyce et al., 2009). The Indian Himalayan region has experienced exceptional flooding recently, including the Ganga River floods in 2010 (Bhatt & Rao, 2016) and the Jhelum floods in 2014 (Bhatt & Rao, 2017). Floods will have long-lasting effects on a place if prolonged heavy rainfall occurs or if massive amounts of water flow downstream (Rahman and Thakur, 2018). The escalating frequency and severity of flooding worldwide have been related to climate change and the corresponding increase in sea levels. As a result, it is recognized that the risk of flooding will not diminish; instead, floods are expected to become more frequent and intense, threatening across the global regions (Tavares et al., 2019). Flood risk is a widespread issue with varying degrees of impact on all nations. On a worldwide scale, flooding is a widespread natural hazard that negatively impacts infrastructure, human populations, ecological systems, and financial systems (Meshram et al., 2020). Devastating floods, which historically occurred once per hundred years, now hit emerging nations hard every year. The main causes include insufficient housing and infrastructure, subpar warning systems, a lack of funding, and a lack of readiness (Borah et al., 2018).
Effective response during flood occurrences depends on monitoring and charting the amount of flooding (Shen et al., 2019). A poor method of monitoring floods is to map them based on river water levels. According to Uddin et al. (2019), the hydrological model approach to flood mapping was favoured above the satellite-based technique. The provision of flood-related data, encompassing the spatial and temporal coordinates of inundated regions, flood magnitude, susceptibility, hazard-prone localities, and secure areas, is of the utmost significance in disaster response activities (Vanama et al., 2020). Flood mapping was instrumental in planning, prevention, and mitigation exercises (Tripathi et al., 2020). Real-time mapping of floods can help disaster management decision-makers to enhance mitigation and response strategies.
In contrast, long-term monitoring and periodic mapping can aid in developing suitable policies and readiness (Mudi et al., 2022). Satellite-based flood assessment is a critical input during, before, and after the flood occurrence for determining the extent and severity of the flood (Mishra et al., 2016). Multiple earth observation satellites provide much remote sensing data, enabling real-time water inundation mapping and monitoring(Parida et al., 2022). For precise flood mapping and monitoring, numerous researchers have explored the potential of diverse remote sensing datasets and approaches (Gao et al., 2018). In recent years, scholars have emphasized the ability of Earth observation (EO) datasets obtained from remote sensing to generate flood maps for real-time monitoring of floods (Schumann, 2018). The enhanced spatial and temporal coverage provided by the latest generation of Earth Observation (EO) satellites has significantly improved the capacity for flood mapping and monitoring (R. Kumar, 2019). On the other hand, due to the high data rate of these satellites, it is necessary to have high-performance computing facilities. These facilities are required to fulfil several prerequisites, including adequate storage space, dependable internet access, sufficient system hardware and software, and efficient algorithms for flood mapping (Bucur et al., 2018).
Flood monitoring and mapping in near real-time are now frequently done using Earth observation (EO) databases (Brivio et al., 2002). Active and passive sensors, which operate within the electromagnetic spectrum's microwave, infrared, visible, and thermal portions, provide crucial data about flooding regions (Sanyal & Lu, 2004). Clouds and hazy weather, typical during floods, make it hard for optical observations to be accurate (Heimhuber et al., 2017). Microwave data, specifically SAR data, are commonly employed for mapping flood areas in addition to optical data (Domeneghetti et al., 2019). In contrast to optical sensors, which rely on solar electromagnetic energy. Synthetic Aperture Radar (SAR) uses its energy to transmit a signal and receives backscatter of surface objects (Ajmar et al., 2017). SAR system operates at extended wavelengths and can penetrate through atmospheric obstructions such as clouds, rain showers, and fog, thereby enabling the monitoring of flood events (Halder & Bandyopadhyay, 2022). In addition, inundated regions are frequently calm, resulting in a smooth water surface that returns less signal to satellites. As a result, the flooded parts appear darker on the radar image compared to other land areas (Amarnath & Rajah, 2016). These qualities of SAR contribute significantly to mapping the flood extent and precise measurements of rivers, lakes, and wetlands (Ohki et al. 2016). During instances of flooding, the availability of SAR images is unfortunately reduced due to the revisit period of the SAR sensor (Bovenga et al., 2018). Typically, SAR images with VV and VH polarizations are used to monitor floodwater, and their applicability for mapping flood inundation was investigated (Agnihotri et al., 2019). Currently, operational satellites, namely RADARSAT-2, TerraSAR, ALOS-2, and Sentinel-1, provide SAR data with global uniformity (Benzougagh et al., 2022). Sentinel-1, TerraSAR, RADARSAT-2, and ALOS are SAR satellites that operate in different frequency bands. Sentinel-1 and RADARSAT-2 operate within the C band, ALOS-2 utilizes the L band frequency range, and TerraSAR operates in the X band. On the other hand, ALOS-2, RADARSAT-2, and TerraSAR operate commercially, necessitating an expensive and time-consuming procurement process. As a result, Sentinel-1 SAR images are favoured in mapping floods because they are readily accessible shortly after data collection, are freely accessible via the web, and retain relatively frequent observations (Uddin et al., 2021).
Instead of obtaining and processing the Sentinel-1 data on a local device, cloud-based technologies like GEE allow users to access and manipulate massive amounts of Sentinel-1 data in real-time (Gorelick et al., 2017). Google's computing infrastructure processes data in parallel, making the processing much more efficient and giving end users many new options. In recent years, the GEE cloud platform has become increasingly popular for remote sensing applications, especially flood mapping (L. Kumar & Mutanga, 2018). Time-series earth observation data from the GEE catalogue has been used in various studies for diverse practical purposes, such as worldwide forest change, urban expansion mapping, global surface water exploration, and global forest observation (Pekel et al. 2016). The GEE and Sentinel datasets were used in relatively recent work by Uddin et al. (2019) to analyze the floods in Bangladesh in 2017 by creating the LULC map needed for flood area validation. Other studies demonstrate that the GEE cloud platform can effectively manage vast multitemporal earth observation data (Hong et al., 2015). Vanama et al., 2020 effectively mapped flood area zones during the Kerala floods in August 2018 in the GEE platform using Otsu's thresholding technique by applying it to Sentinel-1 data to obtain high precision. The findings of these investigations suggest that the GEE platform offers several benefits, such as access to multitemporal Earth Observation datasets, a parallel processing framework, and efficient management of vast amounts of data. The evidence suggests that the fast flood mapping system, which operates on cloud-based technology, should possess specific characteristics. These include being easily accessible, freely accessible, displaying high levels of efficiency and accuracy, and being user-friendly for decision-makers and stakeholders.
The GEE cloud platform and Sentinel-1 SAR data were used in this work to estimate the flood mapping inundation of the Baitarani River Basin. The significant floods that occurred in the basin between 2018 and 2022 were evaluated in this research. The primary objectives of this research are (i) LULC mapping of the Baitarani River Basin using Sentinel 2 and different Machine Learning techniques and (ii) To analyze various polarization combinations of VV and VH and to find the optimal polarization that differentiated water and non-water bodies, (iii) Creation of flood inundation maps using optimum threshold value and polarization, (iv) Validation of flood inundation maps using NDWI maps obtained from Sentinel-2 imagery. Section 2 of this research article discusses the study region and data collection. Section 3 encompasses a detailed discussion of the methodology adopted in this study. Section 4 presents the Results and discussions part of this study. The summary of conclusions is presented in section 5.