­A developed approach to detect flooded areas (case study: Firozkoh county in Tehran province)

doi:10.21203/rs.3.rs-4823672/v1

Download PDF

Research Article

A developed approach to detect flooded areas (case study: Firozkoh county in Tehran province)

https://doi.org/10.21203/rs.3.rs-4823672/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Floods have severe consequences on infrastructure, agriculture, human lives, and the economy, making them one of the most destructive water-related disasters. To effectively manage this disaster, monitoring and analysis are crucial. In this research, a developed approach was presented to identify flooded areas, with a focus on Firozkoh county in Tehran province, Iran. Specifically, a flood event that occurred on July 28th, 2022, was investigated.

To detect flooded areas, Sentinel1 images before and after the flood were collected and preprocessed. These areas appear in dark tones in radar images. However, it is important to note that water bodies such as rivers, lakes, and reservoirs exhibit similar characteristics, and shadows can be mistakenly classified as flooded areas. Therefore, the proposed approach in this study deal with these challenges. At the end, a total of 1369.29 hectares was identified as flooded areas in the case study. To validate the accuracy of the results, TWI and SPI indexes maps were computed and overlaid. This comparison confirmed an accuracy rate of 67%. Additionally, the tweets posted during the flood were examined. Most of them had a hashtag or comment about Mozdaran, a village in Firozkoh county and it is approved by the heat map of detected flooded areas. Overall, this study provides valuable insights into the detection and analysis of flooded areas, offering potential strategies for managing such disasters effectively.

Disaster Management

Floods

GIS

Remote Sensing

Sentinel

Water-related disasters, such as storms, floods, and droughts, account for 90% of natural disasters, with flooding alone making up 54% of these incidents (Mehmood et al. 2021). Floods are considered one of the most devastating, frequent, and costly natural disasters worldwide (Kiran et al. 2019; Sun et al. 2023; Teng et al. 2017; Tarpanelli et al. 2022).

Floods can be triggered by heavy rainfall, rapid snowmelt, dam failures, and inadequate water management (Singh and Pandey, 2021). The frequency of flood events has been on the rise in recent years due to climate change (Tavus et al. 2020). Additionally, factors such as population growth, urban development, and changes in rainfall patterns have amplified the impacts of flooding (Moharrami et al. 2021). Insufficient infrastructure, rapid urbanization, and inadequate consideration of environmental and weather conditions further exacerbate flood risks, posing adverse effects on social, economic, and environmental structures, as well as endangering human lives. They affect numerous individuals and result in significant economic losses (Tavus et al. 2020; Sun et al. 2023; Moharrami et al. 2021; Tarpanelli et al. 2022; Vavassori et al. 2022). Moreover, they disrupt transportation and communication systems, such as transportation lines and communication networks (Azizian and Brocca, 2020; Pourghasemi et al. 2020; Hou et al. 2021), and cause destruction to agricultural lands (Solaimani et al. 2020; Entezari et al. 2020).

Since the start of 2020, floods in South Asia alone have impacted over 17.5 million people, resulting in 1000 fatalities and billions of dollars in economic losses (Mehmood et al. 2021). In Europe, from 1980 to 2021, there have been 614 flood events, leading to over 4000 deaths, affecting 12 million individuals, and causing damages worth 141 billion dollars (Tarpanelli et al. 2022). Furthermore, according to the Global Assessment Report, floods affected 76 million people between 1997 and 2017 (Singh and Pandey, 2021).

Hence, flood monitoring and management play a crucial role in mitigating the impact of floods (Nghia et al. 2022). To effectively manage the risks associated with floods, it is essential to utilize flood warning systems, analyze past flood events, assess the spatial and temporal patterns of floods, estimate the intensity and damage caused by flooding, and determine the extent of destruction (Moharrami et al. 2021; Pourghasemi et al. 2020; Vanama et al. 2020). Accurately estimating the area affected by flooding is vital for evaluating the extent of damage, determining compensation, and making informed decisions for planning in flood-prone areas (Singh and Pandey, 2021).

Additionally, creating flood maps for modeling, risk assessment, and hazard analysis is necessary (Tavus et al. 2020). By implementing appropriate flood measures and crisis management strategies, the impact of flood damage can be minimized (Kiran et al. 2019; Vavassori et al. 2022). Furthermore, identifying areas at risk of flooding enables efficient allocation of resources and investments, such as infrastructure upgrades, and facilitates informed land use planning in the long run (Mehmood et al. 2021; Ghosh et al. 2022). The primary objective of this study is to estimate the inundation area resulting from the flood that occurred on the 28th July 2022, in Firuzkoh County. The research focuses on developing an advanced approach for this purpose.

The literature review encompasses studies conducted between 2017 and 2023, specifically related to estimating and calculating flood inundation areas. The methodology section outlines the proposed research method, and the implementation and results are subsequently discussed.

Previous studies have employed various methods, such as experimental and hydrodynamic models, to assess and estimate the extent of flood inundation (Teng et al. 2017). In a different approach, Azizian and Brocca (2020) utilized a topographic map with a scale of 1:1000 and an interpolation method to generate a Digital Elevation Model (DEM) and channel geometry. This allowed them to simulate floods and create an inundation map.

In order to prepare a flood sensitivity map, Entezari et al. (2020) classified flood data into two groups: training and validation. They considered 11 factors that influence flood occurrences, including geology, land use, distance from rivers, drainage density, slope, aspect, land curvature, topographic wetness index (TWI), altitude classes, and average rainfall. Using the frequency ratio model, weight of evidence, and flood point density in each layer, they calculated the weight of each layer. Finally, by employing the overlapping method, they obtained the final flood sensitivity map for Kermanshah province in Iran. This map was divided into five categories, ranging from areas with very low to very high sensitivity.

River gauge data and simulation models can be utilized to predict flooded areas, although they may not pinpoint the exact locations of flood areas. Conversely, conducting land surveys of flooded areas is a daunting task, especially for vast regions. Satellite flood maps, on the other hand, serve as a fundamental tool for determining the spatial distribution and extent of floods. Consequently, recent focus has shifted towards methods that rely on remote sensing, optical, and radar images to gather valuable information for creating flood inundation maps.

There are two main types of satellite observations used for flood monitoring: optical and radar images. Optical sensors like MODIS, AVHRR, Landsat, Spot, Iconos, GeoEye, Sentinel 2, and WorldView offer medium to high-resolution images for monitoring flooding events both before and after a crisis, aiding in the extraction of flood maps.

Mehmood et al. (2021) employed Landsat 5, 7, and 8 to map flood areas, evaluating the results by creating an uncertainty matrix for nine flood events worldwide, with an overall accuracy ranging between 74 and 89 percent. Landsat 8, in particular, plays a crucial role in preparing flood inundation maps, with its archive accessible to the public since 2008.

Goyal et al. (2023) conducted an analysis of 64 sites in Ramsar China to study their inundation patterns. They utilized Landsat images from 1991 to 2020 to obtain a time series of data. To achieve this, they utilized three images from Landsat 5, 7, and 8. Landsat 5 and 7 consisted of 4 visible and near-infrared bands, two shortwave bands, and one thermal band. On the other hand, Landsat 8 includes 11 bands.

Supervised classification algorithms, such as NDWI², are often used to extract water from optical images (Vavassori et al., 2022). This algorithm is based on the principle that water absorbs reflections in near-infrared and beyond wavelengths (Mehmood et al., 2021). However, according to Tarpanelli et al. (2022), although NDWI is a well-known water index that utilizes green and NIR bands and is effective in most cases, it tends to overestimate water classes in residential areas (Tavus et al., 2020).

To address this limitation, a more sensitive index called MNDWI³ was introduced in 2006 to detect water complications. MNDWI is computed based on the green and SWIR bands and is used to highlight water bodies. It also eliminates features that are commonly associated with water bodies in other water mapping indicators (Mehmood et al., 2021; Tavus et al., 2020).

NDWI and MNDWI are widely used band ratios for generating surface water maps. These indicators rely on the green and near-infrared spectral values of the electromagnetic spectrum. Both MNDWI and NDWI range between -1 and +1, with values above zero indicating water (Sivanpillai et al., 2021).

After a flood event, the values of NDWI and MNDWI in the image are higher compared to before the flood. By applying a threshold value, it is possible to separate pixels corresponding to flooded areas from non-flooded areas (Sivanpillai et al., 2021).

In some studies, the NDVI⁴ index is used to estimate the extent of vegetation damage. For instance, Shrestha et al. (2013) examined the impact of floods in 2006, 2008, and 2011 on agricultural products in the Missouri and Iowa regions using MODIS satellite images.

Solaimani et al. (2020) introduced a method to identify the extent of flood damage in Golestan province in April 2018. They utilized the NDVI index derived from Sentinel2 data during two time frames: early March 2017 and late April 2018. Sentinel2 offers high-resolution optical images globally, encompassing 13 spectral bands including visible, near-infrared, and short-wave infrared, with spatial resolutions of 10, 20, and 60 meters over a swath width of 290 km. Due to its limitations, Sentinel 2 can only capture floods in daylight and under favorable weather conditions, as sunlight cannot penetrate clouds in the visible spectrum (Sivanpillai et al. 2021; Tarpanelli et al. 2022; Vavassori et al. 2022). Consequently, the chances of detecting the maximum flood area are diminished when cloud cover interferes with satellite revisit times (Tarpanelli et al. 2022).

Active radar sensors have the capability to penetrate clouds, dust, and rain commonly found in flood scenarios, enabling observations both day and night. Hence, they are the preferred choice for generating flood inundation maps (Singh and Pandey, 2021; Vanama et al. 2020; Kiran et al. 2019; Solaimani et al. 2020; Singha et al. 2020; Nghia et al. 2022; Tavus et al. 2020; Ghosh et al. 2022; Teng et al. 2017).

Singh and Pandey (2021) employed Sentinel1 data to delineate flood-prone areas in Punjab before and after the August 2019 event. They applied filters to the Sentinel1 dataset to extract SAR⁵data in IW, VV, and VH polarizations in descending orbit modes, focusing on dates surrounding the flood event. Pre-flood analysis utilized SAR images from 13th March 2019 to 13th June 2019, while post-flood analysis utilized images from 21st August 2019 to 31st August 2019. The flood-affected areas were delineated by importing shape files into the GEE⁶ platform, resulting in mapping an area of 205.2 square kilometers as flood zones.

Kiran et al. (2019) utilized Sentinel1 data of level 1 GRD⁷ type and VV polar mode. The processing steps involved the utilization of the satellite orbit file, removal of thermal noise, calibration, salt and pepper filter, ground correction, and linear to decibel conversion.

Some studies have employed a combination of optical and radar data to monitor floods and generate maps of flooded areas (Tavus et al. 2020; Ghosh et al. 2022; Tarpanelli et al. 2022). Tavus et al. (2020) employed a combined approach using Sentinel 2 optical data and Sentinel 1 SAR data to create a flood map for the flood event that occurred on 8th 2018, in Erdoğan Province, Turkey. Tarpanelli et al. (2022) examined the effectiveness of using Sentinel1 and 2 for flood detection in Europe. They analyzed river discharge data in 2000 regions of Europe over a ten-year period. The findings revealed that, on average, 58% of flood events are detectable with Sentinel1. However, due to cloud cover, only 28% can be observed with Sentinel 2.

Since the introduction of the GEE platform in 2010, a web-based platform designed to address big data challenges and enhance satellite image processing for large-scale applications, many flood monitoring researchers have gravitated towards its usage. GEE is a cloud computing platform developed by Google to tackle big data analysis, enabling the processing of remote sensing data over extensive areas and long-term environmental monitoring (Ghosh et al. 2022). This advanced cloud-based platform provides users with access to a vast amount of data and facilitates their processing (Vanama et al. 2020; Nghia et al. 2022; Ghosh et al. 2022; Wu et al. 2019). The GEE data catalog encompasses diverse data from Landsat, Sentinel, Modis, and NAIP (Wu et al. 2019).

Vanama et al. (2020) introduced the GEE4FLOOD framework to generate flood maps on the GEE cloud platform. GEE4FLOOD efficiently manages extensive data by utilizing multi-time SAR images in GEE along with an automated algorithm for flood map creation.

Moharrami et al. (2021) monitored the flood event in March 2019 in Aqqala, a city in northern Iran, by employing Sentinel1 images. GRD level 1 products were acquired, and Otsu's threshold algorithm was applied to distinguish flood areas from other land covers in the region. A total of 8 scenes were chosen for flood monitoring, including one pre-flood, 6 during the flood, and one post-flood. The SRTM⁸ height digital model was utilized to adjust the image height. The flood map was overlaid on DEM and slope maps for analyzing the flooded areas' distribution.

Singha et al. (2020) examined the temporal and spatial patterns of floods in Bangladesh from 2014 to 2018 using SAR images on the GEE platform. Wu et al. (2019) developed an automated approach for flood detection through the integration of lidar data and multi-temporal aerial images on Google Earth Engine. The aerial images were classified using a machine learning algorithm to identify inundated areas. Following the extraction of water clusters from unsupervised classification, misclassified pixels, such as tree shadows, buildings, and other topographic features, were eliminated.

Pourghasemi et al. (2020) utilized machine learning techniques, including boosted regression tree (BRT) and generalized linear model, to assess flood risk and susceptibility in specific areas using Sentinel3 images on Google Earth Engine. The study focused on four prominent districts in Fars province, revealing high to very high flood risk levels for critical infrastructures like hospitals, pharmacies, fire stations, ATMs, gas stations, and mosques in Shiraz. The research methodology involved four key steps: 1) gathering flood data from Sentinel 3 images on the GEE platform, 2) identifying and evaluating influential factors such as elevation, slope, surface curvature, Topographic Wetness Index (TWI), proximity to rivers and roads, drainage density, lithology, precipitation, land cover, and soil characteristics, 3) generating a flood susceptibility map validated with performance indicators, and 4) assessing flood risk on key natural units and vital infrastructures.

DeVries et al. (2020) introduced an algorithm that leverages all accessible Sentinel1 images along with historical Landsat images and supplementary data sources within Google Earth Engine to generate flood inundation maps. Singh and Pandey (2021) conducted an analysis of Sentinel 1 SAR data using Google Earth Engine to produce a flood inundation map.

Scheip and Wegmann (2021) utilized an open-source application called Hazmapper in Google Earth Engine to create a risk map, enabling users to extract and analyze maps based on Sentinel or Landsat data. The NDVI index in HazMapper was employed to detect areas where vegetation was damaged after a natural disaster.

Nghia et al. (2022) established a logical model using GEE with Sentinel1 data and observed data from Tan Chu hydrological stations to estimate floods in the downstream Mekong River Basin.

Ghosh et al. (2022) conducted a web-based analysis to exhibit the spatial analytical capabilities of GEE in flood-affected regions, while also examining socio-demographic impacts. The results were validated using Sentinel2 data.

In recent times, there has been a growing trend in utilizing volunteered and public data for flood calculations. Vavassori et al. (2022) employed optical multispectral images and VGI⁹ to develop a flood inundation map using a semi-automated approach. To overcome the limitations of social media images, they utilized QField to gather the necessary metadata for image classification. QField is an open-source and user-friendly application that seamlessly integrates with QGIS, making it convenient for collecting spatial data. In cases where posts did not have a specific location, the analysis focused on the content of the post itself, including comments, hashtags, likes, and shares. Google Street View was employed to retrieve the location information. The NDWI index was then calculated to differentiate between water features, vegetation, and urban areas.

The research methodology presented in this study consists of four main components: data collection, preprocessing, processing, and validation (Fig. 1). To begin with, the sentinel1 data from two different dates, one before and one after the flood event, are gathered. Moving on to the next phase, the collected data undergoes pre-processing steps. Firstly, both datasets are clipped based on the bounding box of the case study. Subsequently, radiometric calibration is performed, followed by speckle filtering. The purpose of this filtering process is to eliminate salt and pepper noise, which can degrade image quality and hinder interpretation.

Afterwards, it is crucial to establish a threshold to differentiate between water and non-water areas. This threshold is determined based on the histogram, where pixels below the threshold are classified as water, while pixels above the threshold are classified as non-water. Flooded areas appear dark in SAR images due to the minimal return signal they produce. This characteristic allows them to be distinguished from other land covers such as vegetation, agriculture, barren land, or built-up areas, enabling the mapping of flood-affected regions (Moharrami et al., 2021; Singha et al., 2020; Kiran et al., 2019).

In radar images, the water surface, which reflects the radar, exhibits low backscatter values. This results in a dark tone in the images. However, in windy conditions or when vegetation is present, the water surface becomes rough, leading to increased re-diffusion and a decrease in contrast between flooded and non-flooded areas. Consequently, identifying flooded areas becomes more challenging (Tarpanelli et al., 2022).

The occurrence of surface roughness reversal happens when there is a certain level of roughness on the water's surface. These areas may appear dark, although not as dark as the surface of soft water. The rougher the surface, the more signal it reflects, resulting in a brighter appearance (Singh and Pandey, 2021). The threshold value for this phenomenon can be determined by analyzing the overall histogram (Moharrami et al., 2021; Singha et al., 2020; Kiran et al., 2019).

Afterward, terrain correction needs to be implemented to eliminate radiometric changes associated with the topography. This step involves introducing various parameters, such as the Digital Elevation Model (DEM). In this study, a one-second SRTM DEM and nearest neighbor interpolation were utilized. During the processing phase, two layers representing the detected flooded areas before and after the flood are subtracted to eliminate incorrectly classified pixels caused by shadows or other issues. Additionally, certain pixels that belong to water bodies like rivers and lakes need to be excluded. To accomplish this, a landuse layer is required. Water bodies are extracted from the landuse layer and subtracted from the remaining pixels.

To validate the obtained results, two methods were employed. Firstly, TWI and SPI indexes were computed and overlaid for further investigation. Additionally, posted tweets were taken into consideration. Figure 1 illustrates all the steps that were carried out in this study.

Firozkoh County in Tehran province was considered as the case study. This county is located in the northeast of Tehran Province (Fig. 2).

On 28th July 2022, a flood occurred in this county and caused a lot of damage and heavy human and financial losses. After this flood, many buildings were destroyed and thousands of tons of mud entered the path and the road (https://www.isna.ir/news/1401050705435).

To investigate the flood, two sentinel1 images were downloaded from the Copernicus hub. One of them is related to 29th July 2022 which means 1 day after the flood event. Another image is related to 17th July 2022 about 10 days before the flood. Then pre-processing steps were done using SNAP¹⁰ software. The above scenes were clipped by the bounding box of Firozkoh’s county and the radiometric calibration operatar was performed on both of them (Figs. 3a and 3b).

In the next step, a 7 x 7 Lee filter was applied to remove salt and pepper noise (Figs. 4a and 4b).

In the next level, geometric correction was done using the Range Doppler Terrain Correction operator, and SRTM 1Sec HGT elevation digital model was considered for elevation correction, and the bilinear interpolation method was used for DEM resampling and nearest neighbor method for image resampling.

Then binarization of the image was done to separate water from non-water by choosing a threshold limit. For this purpose, the histogram was analyzed and the threshold limit of 2 was selected for the image on 17th July (Fig. 9). But since the flood water is muddy and turbid, the dark effects of water cannot be seen. Therefore, the threshold value of 2.5 was considered for the image on 29th July (Fig. 5a and 5b).

One issue that should be considered is that the above-detected areas contain water bodies too. Therefore, the area of water bodies should be removed. For this purpose, 10-meter land cover data of Dynamic World was downloaded from the Google Earth Engine platform (Fig. 6) and then, water bodies pixels were extracted (Fig. 7).

The extracted water bodies were subtracted from detected flooded areas (Figs. 8a and 8b).

In the next level, the difference between Figs. 8a and 8b was computed to remove the areas that were wrongly considered as flooded ones (Fig. 9).

According to Fig. 9, 1369.29 hectares were detected as flooded areas.

To validate the detected flooded areas, flood risk zoning was done by considering two indexes; topographic wetness index (TWI) and Stream Power Index (SPI). To prepare the TWI map, the Fill operator is first used to correct the DEM of the area. Figure 10 shows the filled DEM of Firozkoh county.

Then it is necessary to calculate the flow direction layer (Fig. 11). In the next step, a flow accumulation layer was prepared (Fig. 12). In the next step, the slope layer is created in degree units, converted to radians and then the tangent of the radian slope map was prepared (Fig. 13).

After that, the modified accumulation flow map is created by adding 1 to the existing one (Fig. 14). Then the TWI index is obtained with Eq. 1:

Ln(modified_flow_accumulation/Tan(SlopeRadian) Eq. 1.

Next, the SPI layer was prepared following Eq. 2.

Ln(modified_flow_accumulation *Tan(SlopeRadian)) Eq. 2

TWI and SPI layers are shown in Figs. 15 and 16 respectively.

Based on previous studies, places that have TWI above 12 and SPI above 4.1 are at the risk of flood. So, two boolean raster layers were created (Figs. 17a and 17b) and were overlaid (Fig. 18).

In the next step, Figs. 17a and 17b were overlaid with the help of the SUM operator (Fig. 18).

Figure 9 and Fig. 18 were compared (Fig. 19) and it was observed that most of the detected flooded areas (red pixels in Fig. 9) are near to green lines.

To quantify the results, a 1 KM buffer was created around green lines and it was investigated the percentage of detected flooded areas that are located in the buffer zones (Fig. 20).

This investigation showed that 914.42 hectares which means 67% are located in the buffer zones. Also, for further investigation, we checked the tweets that were posted at that point of time, and a scraper tool was used to extract metadata of each tweet.

53 tweets were collected. One of them was related to Afghanistan and so it was deleted. Among of 52 remaining tweets, 31 of them were posted on the 29th and 30th of July 2022. Moreover, their hashtags and their contents were examined. Most of them were pointed to Mozdaran (the name of a village in Firozkoh’s county( and it corresponds with the heatmap of detected flooded areas (Fig. 21).

Floods are one of the disasters that lead to many negative effects on the economy, people’s lives, and the environment. Therefore, it should be monitored and analyzed. Different methods are used to estimate the extent of floods and their damages, one of the novel methods is to utilize satellite images, especially the radar ones. The advantage of radar is that its signals can penetrate clouds and dust.

In this research, the Sentinel1 data was used to investigate the extent of a flood that occurred in Firozkoh county of Tehran province on 28th July 2022. Flooded areas appear in dark tones in radar images, so they can be distinguished from other features. However, water bodies such as rivers, lakes, and water reservoirs appear in dark tones too and they cause overestimation. Therefore, to remove these water bodies, they were extracted from the land use layer and they were subtracted from detected flooded areas. Moreover, to remove areas that are wrongly classified as flooded areas, one image from about two weeks before the flood event was downloaded and the flooded area was extracted from it. Then the difference between detected flooded areas of these two dates was computed. The result of this investigation showed that 1369.29 hectares were influenced by this flood event.

To validate the results, two methods were applied. In the first method, SPI and TWI indexes were computed and places that are at risk of flood were identified and compared with the detected flooded areas. This comparison showed 67% agreement.

In the second method, posted tweets about this flood event were investigated and most of them have hashtags about Mozdaran that corresponds with the heat map of the flood event.

No funding was received for this research.

Author Contribution

Giti KhoshAmooz wrote the main manuscript text, collected data and did analysis.

Azizian, A., & Brocca, L. (2020). Determining the best remotely sensed DEM for flood inundation mapping in data sparse regions. International Journal of Remote Sensing, 41(5), 1884-1906.
DeVries, B., Huang, C., Armston, J., Huang, W., Jones, J. W., & Lang, M. W. (2020). Rapid and robust monitoring of flood events using Sentinel-1 and Landsat data on the Google Earth Engine. Remote Sensing of Environment, 240, 111664.
Entezari M, Jalilian T, Darvishi Khatooni J. Classification map of the sensitivity of flooding using the method of assessment frequency and weight of evidence in the Kermanshah Province. Journal of Spatial Analysis Environmental Hazards 2020; 6 (4) :143-162 URL: http://jsaeh.khu.ac.ir/article-1-2707-en.html.
Ghosh, S., Kumar, D., & Kumari, R. (2022). Cloud-based large-scale data retrieval, mapping, and analysis for land monitoring applications with google earth engine (GEE). Environmental Challenges, 9, 100605.
Goyal, M. K., Rakkasagi, S., Shaga, S., Zhang, T. C., Surampalli, R. Y., & Dubey, S. (2023). Spatiotemporal-based automated inundation mapping of Ramsar wetlands using Google Earth Engine. Scientific Reports, 13(1), 17324.
Hou, J., Zhou, N., Chen, G., Huang, M., & Bai, G. (2021). Rapid forecasting of urban flood inundation using multiple machine learning models. Natural Hazards, 108(2), 2335-2356.
Kiran, K. S., Manjusree, P., & Viswanadham, M. (2019). Sentinel-1 SAR data preparation for extraction of flood footprints-A case study. Disaster Advances, 12(12), 10-20.
Mehmood, H., Conway, C., & Perera, D. (2021). Mapping of flood areas using landsat with google earth engine cloud platform. Atmosphere, 12(7), 866.
Moharrami, M., Javanbakht, M., & Attarchi, S. (2021). Automatic flood detection using sentinel-1 images on the google earth engine. Environmental monitoring and assessment, 193, 1-17.
Nghia, B. P. Q., Pal, I., Chollacoop, N., & Mukhopadhyay, A. (2022). Applying Google earth engine for flood mapping and monitoring in the downstream provinces of Mekong River. Progress in Disaster Science, 14, 100235.
Pourghasemi, H. R., Amiri, M., Edalat, M., Ahrari, A. H., Panahi, M., Sadhasivam, N., & Lee, S. (2020). Assessment of urban infrastructures exposed to flood using susceptibility map and Google Earth Engine. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 1923-1937.
Scheip, C. M., & Wegmann, K. W. (2021). HazMapper: a global open-source natural hazard mapping application in Google Earth Engine. Natural Hazards and Earth System Sciences, 21(5), 1495-1511.
Shrestha, R., Di, L., Yu, G., Shao, Y., Kang, L., & Zhang, B. (2013, August). Detection of flood and its impact on crops using NDVI-Corn case. In 2013 Second international conference on agro-geoinformatics (Agro-geoinformatics)(pp. 200-204). IEEE.
Singha, M., Dong, J., Sarmah, S., You, N., Zhou, Y., Zhang, G., ... & Xiao, X. (2020). Identifying floods and flood-affected paddy rice fields in Bangladesh based on Sentinel-1 imagery and Google Earth Engine. ISPRS Journal of Photogrammetry and Remote Sensing, 166, 278-293.
Singh, G., & Pandey, A. (2021). Mapping Punjab flood using multi-temporal open-access synthetic aperture radar data in Google earth engine. Hydrological Extremes: River Hydraulics and Irrigation Water Management, 75-85.
Sivanpillai, R., Jacobs, K. M., Mattilio, C. M., & Piskorski, E. V. (2021). Rapid flood inundation mapping by differencing water indices from pre-and post-flood Landsat images. Frontiers of Earth Science, 15, 1-11.
Solaimani, K., Sharifipour, M., & Abdoli Boozhani, S. (2020). Flood Damage Detection Algorithm Using Sentinel-2 Images (Case Study: Golestan Flood of March 2019). Iranian journal of Ecohydrology, 7(2), 303-312. doi: 10.22059/walk.2020.292005.1233.
Sun, A. Y., Li, Z., Lee, W., Huang, Q., Scanlon, B. R., & Dawson, C. (2023). Rapid flood inundation forecast using Fourier neural operator. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 3733-3739).
Tarpanelli, A., Mondini, A. C., & Camici, S. (2022). Effectiveness of Sentinel-1 and Sentinel-2 for flood detection assessment in Europe. Natural Hazards and Earth System Sciences, 22(8), 2473-2489.
Tavus, B., Kocaman, S., Nefeslioglu, H. A., & Gokceoglu, C. (2020). A fusion approach for flood mapping using Sentinel-1 and Sentinel-2 datasets. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 43, 641-648.
Teng, J., Jakeman, A. J., Vaze, J., Croke, B. F., Dutta, D., & Kim, S. J. E. M. (2017). Flood inundation modelling: A review of methods, recent advances and uncertainty analysis. Environmental modelling & software, 90, 201-216.
Vanama, V. S. K., Mandal, D., & Rao, Y. S. (2020). GEE4FLOOD: rapid mapping of flood areas using temporal Sentinel-1 SAR images with Google Earth Engine cloud platform. Journal of Applied Remote Sensing, 14(3), 034505-034505.
Vavassori, A., Carrion, D., Zaragozi, B., & Migliaccio, F. (2022). VGI and Satellite Imagery Integration for Crisis Mapping of Flood Events. ISPRS International Journal of Geo-Information, 11(12), 611.
Wu, Q., Lane, C. R., Li, X., Zhao, K., Zhou, Y., Clinton, N., ... & Lang, M. W. (2019). Integrating LiDAR data and multi-temporal aerial imagery to map wetland inundation dynamics using Google Earth Engine. Remote sensing of environment, 228, 1-13.

Normalized Difference Water Index
Modified Normalized Difference Water Index
Normalize Difference Vegetation Index
Synthetic Aperture Radar
Google Earth Engine
Ground Range Detected
Shuttle Radar Topography Mission
Volunteered Geographic Information
Sentinel Application Platform

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

A developed approach to detect flooded areas (case study: Firozkoh county in Tehran province)

Status:

Version 1

Abstract

Figures

1. Introduction

2. Literature Review and Background

3. Methodology

4. Implementation

5. Discussion and Results

6. Conclusion

Declarations

Author Contribution

References

Footnotes

Additional Declarations

Status:

Version 1

­A developed approach to detect flooded areas (case study: Firozkoh county in Tehran province)

Status:

Version 1

Abstract

Figures

1. Introduction

2. Literature Review and Background

3. Methodology

4. Implementation

5. Discussion and Results

6. Conclusion

Declarations

Author Contribution

References

Footnotes

Additional Declarations

Status:

Version 1

A developed approach to detect flooded areas (case study: Firozkoh county in Tehran province)