The southern plain of Nepal recognized as the ‘granary of Nepal’, confronts recurrent monsoon-induced flooding, posing a substantial threat to its pivotal role as a major agricultural contributor to the national economy. As an analysis, this study employs advanced satellite imagery to delineate historical floods in nine flood-prone transboundary basins and compares the rainfall-induced model-based inundation in the West Rapti Basin (WRB) to validate the result. The extent of flooding was mapped between 2015 and 2022 using Sentinel-1 Synthetic Aperture Radar data processed on Google Earth Engine. Hydrodynamic modelling centred on the WRB, incorporated daily measured precipitation data with varying return periods over a 10 m resolution digital elevation model generated through an in situ survey. The model was calibrated for the August 2017 flood event with Nash–Sutcliffe efficiency greater than 70% and validation reasonably with satellite-derived flood maps with Cohen's Kappa value of 0.58 and an overall accuracy metric of 0.84. This synergic approach integrates climatology, remote sensing data, and hydraulics to monitor transboundary river floods in Nepal where precise hydro-meteorological data are limited, thus, offering continuous all-weather monitoring.

  • Satellite imagery (processed on Google Earth Engine) was used to study inundation in nine flood-prone transboundary rivers of Nepal.

  • Hydraulic modelling in the West Rapti River (gauged basin) validates the accuracy of flood prediction tools against satellite-maps.

  • Climatology, remote sensing, and hydraulics-mixed multidisciplinary approach validates the application of remote sensing for flood inundation studies, especially in data-scarce regions.

Floods, recognized as the most common and devastating natural disasters globally, inflict substantial economic, agricultural, and human losses (Cred 2018). These catastrophic events pose a significant threat to infrastructure, agriculture, and natural habitats, impacting the lives of those affected. Timely and accurate mapping of floods is thereby imperative for emergency responders and decision-makers who rely on this information to allocate resources and evacuate affected areas. In the southern Terai region and valleys of Nepal, characterized by diverse topography, floods and associated inundation are recurrent and widespread (Smith et al. 2017; Thapa et al. 2020). This region, encompassing the entire 1,700 km Nepal–India border, experiences severe flooding annually during the monsoon season. Furthermore, the construction of infrastructures such as high embankment roads (with very few cross-water drainages) and barrages (immediately downstream of transboundary) to hold the monsoon floods on the other side of the border has resulted in a backwater effect on Nepal's border causing severe inundation resulting in substantial damage to agriculture and infrastructure (Sharma et al. 2019; Gupta et al. 2021). The impact on agriculture, particularly crops such as paddy and maize, leads to massive losses in agricultural land and property (Gautam et al. 2023). In the context of Nepal, the issue of flooding is noticeable in low-relief mountains and the Mahabharat range amplifying towards Siwaliks where a severe flash flood is prominent during monsoon season. Increased river flows in the lower Siwaliks during these storms create massive flash floods in the southern plains resulting in excessive sedimentation and inundation towards transboundary regions based on observations from the past (Kafle 2020). Despite the attention given to nine key rivers in the region, flood, and associated issues persist due to insufficient evidence and challenges in management, which is a reason why flood modelling and mapping is essential for understanding and responding to these disasters (Merz et al. 2007; Mudashiru et al. 2021). This complication in gathering evidence escalates due to scarce data-measuring facilities on the southern transboundary where there is a serious concern of cross-border intervention highlighting the application of remote sensing in these regions.

Hydraulic modelling, with adequate resources and validation, has become a trending practice in flood inundation mapping and impact assessment (Aryal et al. 2020; Devkota et al. 2020; Thapa et al. 2020). However, for large rivers lacking adequate observational records, satellite-based products such as Sentinel 1 offer valuable information for delineating flood extent maps and assessing impacts (Psomiadis 2016; Sharma et al. 2019b) due to their ability to capture images through cloud cover and even at night, proving indispensable in all-weather, day-night monitoring (Maggioni & Massari 2018; Shen et al. 2019). The high spatial resolution of Sentinel 1 enables the detection of changes caused by floods, including the expansion of water bodies and the erosion of riverbanks. Beyond mapping, Sentinel 1 also monitors flood progress over time, aiding emergency responders in planning and anticipating impacts on communities and infrastructure (Manjusree et al. 2012). Hydro-meteorological characteristics, encompassing elements such as precipitation, evaporation, humidity, wind speed, and temperature, play a crucial role in flood dynamics due to their influence on the extent and severity of flood inundation (Estrany et al. 2020). Therefore, understanding and monitoring these characteristics are vital for predicting and assessing flood risks, issuing early warnings, and implementing effective mitigation measures. Traditional flood extent modelling involves numerical models based on the conservation of mass and momentum principles, which provide accurate simulations; however, significant computational resources and expertise are required to achieve relevant accuracy. The reduction in computational extensiveness commonly achieved through simplification of complex equations further adds up uncertainty in the model.

The robustness in the application of advanced machine learning (ML) models such as long-short-term-memory, multilayer perceptron-firefly algorithm, support vector machine, relevant vector machine, adaptive neuro-fuzzy inference system, and extreme ML among others, has been much more pronounced for flood modelling and forecasting (Araghinejad et al. 2011; Chen et al. 2015; Atiquzzaman & Kandasamy 2016; Yaseen et al. 2017, 2019). However, the reliability of these models extensively depends upon the quantity of available training datasets, which adds to uncertainty for poorly gauged regions such as Nepal. Hence, the precision and reliability of flood models that can simulate small inflow–outflow datasets should also be assessed for accurately predicting and validating flood extents, particularly considering the observed overestimation in flooding extent using hydraulic modelling approaches (Giustarini et al. 2012). Refinements are required in hydraulic modelling approaches to ensure their applicability and accuracy across diverse transboundary river basins (Hong Quang et al. 2019). Similarly, flood dynamics vary across different transboundary river basins in the region, and adjustments are necessary for hydraulic modelling tools to capture these variations (Hostache et al. 2009). Therefore, the question of how to improve flood impact assessments on the data-scarce transboundary region for better evaluation of the vulnerable agricultural areas under various flooding scenarios, and the ways of enhancing the overall assessments and mitigation has become crucial.

To address these questions, the precise and reliable flood modelling tools infused with satellite-derived mapping products can aid in the context of transboundary river basins (Ali et al. 2021) in the southern plain of Nepal, which are facing a challenge of physical data scarcity to perform numerical modelling. For that, remote sensing technologies, including radar data, have emerged as valuable tools in flood mapping due to their ability to offer high-resolution, timely, and cost-effective data (Yan et al. 2015). Besides that, radar data, penetrating clouds, and fog have proven particularly useful in regions such as Bangladesh and Nepal, where floods coincide with the monsoon period, obstructing optical image acquisition (Borah et al. 2018; Sharma et al. 2019b). Although not needed due to available cloud-free images in our case, the uncertainties within the available synthetic aperture radar (SAR) products due to cloud noise, and scattering delays must be addressed and filtered upon assessment to eliminate ambiguous results.

This research aims to address existing gaps in flood modelling tools and enhance their accuracy for the southern plain of Nepal's transboundary river basins by focusing on improved flood mapping techniques, integrating advanced satellite technologies such as Sentinel 1, and considering the hydro-meteorological characteristics influencing flood dynamics. Furthermore, the study focuses on the West Rapti River basin, leaving room for potential variations in flood dynamics across the other eight transboundary basins. Addressing this gap involves refining and validating the hydraulic modelling tools to ensure their applicability across diverse river basins, providing a more comprehensive understanding of flood behaviour in the region. Similarly, depth quantification is attainable using hydraulic models if quality hydro-meteorological input data are available for validating the model results; however, the gap still exists for quantifying depth from remote sensing applications. Thus, a coupled approach is required to justify the application of remote sensing products for flood extent mapping in transboundary regions with serious concern of data scarcity, which has been done in this study. This study, therefore, seeks to get future contributions for quantifying depth, duration, and flood monitoring through the use of high-resolution satellite products, which will be revolutionary for effective emergency response planning.

Study area

The focus of this study encompasses nine river systems lying in the southern part with drainage areas ranging from 120 to 6,380 km2 inside Nepal (Figure 1). Most of the medium river systems originate from the Siwalik region, however, large river systems originate from the mountain regions. Three basins are gauged (i.e., having observed discharge data) while the remaining are ungauged. Key features of the selected nine river systems are provided in Table 1 within those river systems (Mahakali; West Rapti; Banganga; Lal Bakaiya; Bagmati; Lakhandehi; Ratu; Khado; and Kamala).
Table 1

Key features/characteristics of the nine selected river systems in Nepal

SNRiver nameLat (°N)Lon (°E)ProvinceOutlet districtGauged/ungaugedOriginates fromDrainage area (km2)District drained(Rural) Municipality
LeftRight
Mahakali 28.83 80.11 Sudurpaschim Kanchanpur Gauged High mountains 15,587 (in Nepal: 5,087) Darchula, Baitadi, Dadeldhura, Kanchanpur Several Several 
West Rapti 28.17 81.38 Lumbini Banke Gauged Middle mountains 6,380 Pyuthan, Rolpa, Arghakhachi, Dang, Banke Rapti Sonari Duduwa and Nepalgunj 
Banganga 27.47 82.96 Lumbini Kapilvastu Ungauged Siwaliks 994 Arghakanchi, Kapilvastu Maharajgunj Maharajgunj 
Lal Bakaiya 26.76 85.26 Bagmati and Madhesh Rautahat Ungauged Siwaliks 985 Makwanpur, Bara, Rautahat Rajdevi and Yemunamai Gaur, and Rajpur 
Bagmati 26.75 85.34 Bagmati and Madhesh Rautahat Gauged Hill 3,231 Kathmandu, Lalitpur, Bhaktapur, Makwanpur, Kavrepalanchowk, Sindhuli, Rautahat, Sarlahi Balara, and Ram Nagar Rajdevi, and Durga Bhagwati 
Lakhandehi 26.82 85.51 Madhesh Sarlahi Ungauged Terai 194 Sarlahi Malangawa Kaudena 
Ratu 26.63 85.78 Bagmati and Madhesh Mahottari Ungauged Siwaliks 758 Mahottari, Dhanusha Jaleshwor Jaleshwor 
Khado 26.46 86.77 Madhesh Saptari Ungauged Siwaliks 120 Saptari Tilthi Koiladi Chhinnamasta 
Kamala 26.61 86.15 Bagmati, Madhesh, and Province 1 Siraha Ungauged Hill 2,074 Sindhuli, Udayapur, Dhanusha, Siraha Siraha, and Kalyanpur Janaknandani, and Kamala 
SNRiver nameLat (°N)Lon (°E)ProvinceOutlet districtGauged/ungaugedOriginates fromDrainage area (km2)District drained(Rural) Municipality
LeftRight
Mahakali 28.83 80.11 Sudurpaschim Kanchanpur Gauged High mountains 15,587 (in Nepal: 5,087) Darchula, Baitadi, Dadeldhura, Kanchanpur Several Several 
West Rapti 28.17 81.38 Lumbini Banke Gauged Middle mountains 6,380 Pyuthan, Rolpa, Arghakhachi, Dang, Banke Rapti Sonari Duduwa and Nepalgunj 
Banganga 27.47 82.96 Lumbini Kapilvastu Ungauged Siwaliks 994 Arghakanchi, Kapilvastu Maharajgunj Maharajgunj 
Lal Bakaiya 26.76 85.26 Bagmati and Madhesh Rautahat Ungauged Siwaliks 985 Makwanpur, Bara, Rautahat Rajdevi and Yemunamai Gaur, and Rajpur 
Bagmati 26.75 85.34 Bagmati and Madhesh Rautahat Gauged Hill 3,231 Kathmandu, Lalitpur, Bhaktapur, Makwanpur, Kavrepalanchowk, Sindhuli, Rautahat, Sarlahi Balara, and Ram Nagar Rajdevi, and Durga Bhagwati 
Lakhandehi 26.82 85.51 Madhesh Sarlahi Ungauged Terai 194 Sarlahi Malangawa Kaudena 
Ratu 26.63 85.78 Bagmati and Madhesh Mahottari Ungauged Siwaliks 758 Mahottari, Dhanusha Jaleshwor Jaleshwor 
Khado 26.46 86.77 Madhesh Saptari Ungauged Siwaliks 120 Saptari Tilthi Koiladi Chhinnamasta 
Kamala 26.61 86.15 Bagmati, Madhesh, and Province 1 Siraha Ungauged Hill 2,074 Sindhuli, Udayapur, Dhanusha, Siraha Siraha, and Kalyanpur Janaknandani, and Kamala 
Figure 1

Location of the nine selected river systems in Nepal and topographical variation: (a) Mahakali, (b) West Rapti, (c) Banganga, (d) Lal Bakaiya, (e) Bagmati, (f) Lakhandehi, (g) Ratu, (h) Khado, and (i) Kamala.

Figure 1

Location of the nine selected river systems in Nepal and topographical variation: (a) Mahakali, (b) West Rapti, (c) Banganga, (d) Lal Bakaiya, (e) Bagmati, (f) Lakhandehi, (g) Ratu, (h) Khado, and (i) Kamala.

Close modal

Data and sources

In this study, available hydro-meteorological data were retrieved from the Department of Hydrology and Meteorology (DHM), Nepal. For satellite-based flood mapping and studying the inundation pattern of rivers, radar images from Sentinel-1 and optical imageries from Landsat were accessed from the Google Earth Engine (GEE) platform. Similarly, for detailed hydraulic modelling of the floodplain, required data such as the digital elevation model (DEM) from AW3D30 (downloaded from the JAXA portal), land use/cover data maintained by the International Centre for Integrated Mountain Development (ICIMOD), and soil data from the soil and terrain database (SOTER) were also utilized. Table 2 shows the overall database and its potential sources.

Table 2

Sources and scales of the database used in this study

DataSourcesResolutionLengthProcessing Tool
Radar image Sentinel − 1 10 m 2015–Present GEE, python 
Optical image Landsat 5, 7, 8 30 m 1987–Present GEE, python 
DEM ALOS World 3D 30 m (AW3D30) 30 m  Quantum Geographic Information System (QGIS), HEC-RAS 
Soil Soil and terrain database (SOTER) 1:1,000,000  QGIS, HEC-RAS 
Land use/cover ICIMOD 30 m 2000–2020 QGIS, HEC-RAS 
Daily discharge, precipitation DHM Daily 1980–2017 MS-Excel, R, Climpact2 
DataSourcesResolutionLengthProcessing Tool
Radar image Sentinel − 1 10 m 2015–Present GEE, python 
Optical image Landsat 5, 7, 8 30 m 1987–Present GEE, python 
DEM ALOS World 3D 30 m (AW3D30) 30 m  Quantum Geographic Information System (QGIS), HEC-RAS 
Soil Soil and terrain database (SOTER) 1:1,000,000  QGIS, HEC-RAS 
Land use/cover ICIMOD 30 m 2000–2020 QGIS, HEC-RAS 
Daily discharge, precipitation DHM Daily 1980–2017 MS-Excel, R, Climpact2 

To understand and characterize extreme rainfall in the basins, daily precipitation data acquired from the DHM were used after quality assessment. The screened data were then used to analyse the flood frequency and compute return periods for daily maximum rainfall. Furthermore, a climate extreme analysis was conducted to understand the trends in various extreme rainfall indices. For delineating flood extents, the Sentinel-1 data from 2015 to 2022 for each month of the monsoon season (June–September) were pre-processed and extracted using an adaptive histogram-thresholding algorithm (Prakash et al. 2022). Next, the West Rapti River was chosen as the study area for detailed hydraulic modelling due to the sufficient availability of hydro-meteorological data, and cloud-free satellite images during the flooding period. Besides that, a cross-section survey was conducted near the border region and a 10 m resolution DEM was developed from the survey points. The generated DEM was merged with AW3D 30 m resolution DEM for the whole basin and a 2-D rain-on-grid model was set up for the basin in Hydrologic Engineering Center-River Analysis System (HEC-RAS). The model was simulated and calibrated for the 2017 August flooding event and validated with the satellite-derived flood map. The validated model was then used to simulate flooding extent and hazards for various return period scenarios.

Climatic extreme analysis

Observed daily precipitation (P) data were collected from the DHM in and around the vicinity of the nine watersheds. Data quality was assessed based on the percentage of missing values and exploratory analysis plots, such as hyetographs and average annual as well as monthly values. The time series data were reviewed for discrepancies using visual plots and statistical analysis, and the stations with good-quality data were retained for further investigation after extensive exploratory analysis. Based on the quality of the data, a suitable data length was chosen for filling and subsequent analysis. Thereafter, missing precipitation values were filled using the normal ratio method (Armanuos et al. 2020).

Rainfall frequency analysis

Rainfall frequency analysis helps in understanding the frequency and magnitude of extreme rainfall events, which is crucial for the identification of the proper timing of flood extent. One of the commonly used probability distributions in rainfall frequency analysis is the Gumbel distribution. The Gumbel distribution, also known as the Type I extreme value distribution, is often applied to model the maximum or minimum values of a random variable, such as rainfall, in a given period. The Gumbel distribution is characterized by two parameters: location (μ) and scale (β). The location parameter represents the position of the distribution's mode or peak, while the scale parameter determines the spread or variability of the distribution. These parameters can be estimated using various statistical methods, such as the method of moments or maximum likelihood estimation. The Gumbel distribution was fitted for an annual time series of maximum 24-h peak rainfall for each station used in the analysis. The fitted distribution was then used to estimate the corresponding return periods for observed peak rainfall magnitude.

Characterizing climatic extremes

A total of 27 indices are considered to be core indices as per Zhang (2020). They are based on daily temperature values or daily precipitation amounts. Sixteen of them are based on temperature and 11 are based on precipitation. These indices can be calculated with the help of the Climpact2 tool. The indices based on precipitation have been listed in Table 3. Only precipitation indices are used in this study.

Table 3

ETCCDI precipitation indices (source: Zhang & Yang (2004))

IndicesDescription
Rx1day Monthly maximum 1-day precipitation 
Rx5day Monthly maximum consecutive 5-day precipitation 
R20 mm Annual count of days when PRCP ≥ 20 mm 
CDD Maximum length of dry spell, the maximum number of consecutive dry days with RR < 1 mm 
CWD Maximum length of the wet spell, the maximum number of consecutive wet days with RR ≥ 1 mm 
R95pTOT Annual total PRCP when RR > 95p 
R99pTOT Annual total PRCP when RR > 99p 
PRCPTOT Annual total precipitation on wet days 
IndicesDescription
Rx1day Monthly maximum 1-day precipitation 
Rx5day Monthly maximum consecutive 5-day precipitation 
R20 mm Annual count of days when PRCP ≥ 20 mm 
CDD Maximum length of dry spell, the maximum number of consecutive dry days with RR < 1 mm 
CWD Maximum length of the wet spell, the maximum number of consecutive wet days with RR ≥ 1 mm 
R95pTOT Annual total PRCP when RR > 95p 
R99pTOT Annual total PRCP when RR > 99p 
PRCPTOT Annual total precipitation on wet days 

Note: PRCP, Precipitation; RR, daily precipitation; 95p = 95th percentile; 99p = 99th percentile.

Flood extent mapping based on satellite data

Radar images from the Sentinel-1 mission, available on the GEE platform, were used for mapping flood/inundation extents. GEE is a cloud-based platform for remote sensing and geospatial analysis. It is designed to allow users to access, process, and analyse large volumes of geospatial data quickly and efficiently (Pham-Duc et al. 2023; Prodromou et al. 2023). The Sentinel-1 image collection in GEE contains Level-1 ground range detected scenes that have been processed to provide backscatter coefficient values in decibels (dB). The backscatter coefficient represents the radar cross-section of the target, or how much the terrain reflects the incident microwave radiation towards or away from the SAR sensor. This value is important because it depends on the physical characteristics of the terrain, such as its geometry and electromagnetic properties. To derive the backscatter coefficient in each pixel, Earth Engine uses a series of pre-processing steps, including applying an orbit file, removing noise and invalid data, applying radiometric calibration values, and performing terrain correction.

The methodological framework for extracting flooding extent from Sentinel-1 imagery is shown in Figure 2. The entire image collection available in the GEE platform is first filtered to include only images of 4 months (June–September) from the monsoon season. Each image in this filtered collection undergoes a pre-processing that includes border noise removal, slope correction, and speckle filtering (Singh et al. 2022). The Sentinel-1 images available before April 2017 have very low backscatter values at the borders, causing them to be identified as water during the later stage of the process. Hence, these border noises were removed by thresholding very small values of backscatter (−35 dB). The shadow effects due to terrain also cause the image to have low backscatter values causing it to be seen as water. This effect was reduced by incorporating slope information from MERIT DEM (Yang et al. 2022). The height above nearest drainage information was used to remove all the water pixels that were misidentified at locations with steep slopes (Shilengwe et al. 2023). Finally, the pre-processing stage includes the removal of speckles through a ‘Refined-Lee’ speckle filtering algorithm (Ferraggine & Villar 2020).
Figure 2

Methodology of flood mapping from Sentinel-1 imagery adopted in this study.

Figure 2

Methodology of flood mapping from Sentinel-1 imagery adopted in this study.

Close modal

The pre-processed images are then input to a surface water extraction algorithm. The algorithm we used is called edge-otsu. It is a histogram-based adaptive thresholding algorithm. A common approach to extracting water from Sentinel-1 imagery is thresholding with a value below which all pixels are water and all others are land (Markert et al. 2020). The threshold value is often determined using the distribution of the pixels. However, the histogram must be bimodal to adequately represent equal areas of land and water. To achieve this, the implemented algorithm first detects the water edges using a canny edge detector. The edges are then buffered equally on both sides to include equal areas of land and water ensuring bimodality in the obtained histogram. The Otsu thresholding is performed for pixels sampled from the buffered region.

The surface water extracted in this way also includes permanent water bodies, including lakes or large irrigation canals. It is crucial to classify permanent and actual flooding in the generated maps. To do so, the same methodology was applied to the dry period of the year. A median composite of the images from all the remaining months of the year was used as a reference image. The reference image underwent the same procedure of water extraction, which was classified as permanent water. Finally, a difference in surface water image and permanent water produced the final flooding map. These flood maps were then aggregated at a monthly scale so that they would span the whole basin and statistics such as area extent could be computed. The final products are monthly spatial maps and time series of flooding extent from 2015 to 2022.

Flood inundation model-based mapping

For this study, a two-dimensional rain-on-grid model was developed in the HEC-RAS, which enables the incorporation of a hydrologic model within the hydraulic modelling framework (Costabile et al. 2020). The new HEC-RAS version six features the inclusion of spatially varying precipitation using various interpolation methods. Daily precipitation data from 16 ground-based meteorological stations within and around the basin were acquired and used as meteorological input to the HEC-RAS model. For terrain, a DEM AW3D with a spatial resolution of 30 m was used in the model. However, in the study extent near the border, a detailed cross-section survey was carried out. The acquired survey points were then converted to DEM using interpolation techniques in ArcGIS software. The resulting DEM had a spatial resolution of 10 m and was merged with the 30 m AW3D DEM creating a seamless DEM for the whole study basin. Model development in HEC-RAS was initiated by setting up an external boundary condition (outflow) near the Indo-Nepal border. A precipitation inflow (at meteorological gauge locations) for the entire 2D flow area and a normal depth for the outflow boundary condition were established. Precipitation inflow was provided as per the meteorological gauge coordinates. The land use and infiltration layers were then created; these were the key calibration parameters for matching the flow characteristics. A variable Manning's ‘n’ value, percentage imperviousness, and infiltration parameters were used based on different land cover classes. An initial value of Manning's n was first adopted using the parameter range recommended by HEC-RAS software. The model was then calibrated and validated to mimic the observed discharge and rating curves at Bagasoti and Jalkundi gauging stations for the flood events of 1999, 2014, and 2017 AD. This was attained based on simulation for different land cover types through a rigorous trial-and-error method.

For the event-based simulation, we selected the floods of August 2017 since they had the largest flooding extent based on Sentinel-1 flood mapping and historical climate analysis. The HEC-RAS model was run from 1 to 30 August 2017. The mapping, hydrograph, and detailed output intervals were taken as 1 day whereas the computation interval was set as 10 min with variation as per courant condition.

The maximum flood inundation boundary was then extracted from the HEC-RAS software and compared with the Sentinel-1-based flood map. The comparison was done using the simple but effective Kappa coefficient and overall accuracy metrics (Vieira et al. 2010). The Kappa coefficient (Equation (1)) also known as Cohen's Kappa (Cohen 1960) is a statistical measure used to assess the level of agreement between two raters or observers beyond what would be expected by chance and is given by:
(1)
where represents the observed agreement between the raters, which is the proportion of agreement observed beyond chance and represents the expected agreement due to chance, which is calculated by assuming that the raters are randomly assigning categories. The Kappa coefficient ranges from −1 to 1. A value of 1 indicates perfect agreement, 0 indicates agreement equivalent to chance, and −1 indicates total disagreement. Interpretation of the Kappa coefficient can vary depending on the field of study. In general, a value above 0.75 is considered excellent agreement, values between 0.4 and 0.75 indicate fair to good agreement, and values below 0.4 indicate poor agreement (Sun 2011). The overall accuracy metric is also a common evaluation measure used to assess the performance of the classification model. It is computed based on the proportion of true positive and true negative in a confusion matrix. It represents the proportion of correctly classified instances out of the total number of instances in the dataset. The significance of this study lies in the comparison of satellite products with results from the hydraulic model and implementing the open-source satellite imagery for ungauged transboundary rivers. Numerous evaluation metrics for data classification can be found upon review. However, taking into consideration the strength and limitation of the Kappa and overall accuracy metrics (Foody 2002; Sun 2011; Hossin & Sulaiman 2015; Congalton & Green 2019), and the expertise available to perform this analysis, we adopted those as performance indices for evaluation of image comparison. Kappa metrics have been particularly applied for image comparison showing promising results in multiple fields ranging from health informatics to remote sensing (though the use of these statistics in remote sensing has been countered) (Pontius & Millones 2011; Foody 2020). However, today, multiple sophisticated tools have evolved, making use of MLs and artificial intelligence (AI) capabilities for image comparison with better performance compared to the Kappa coefficient.

There are also practices to present results in the form of scatter plots, and Taylor and Violin plots. However, they were not suitable for this study due to the binary classification (i.e., flooded and non-flooded areas) of data in our case.

Climatic extremes

In the exploratory analysis, we considered a total of 36 meteorological stations distributed across the West Rapti Basin (WRB) (11 stations), Mahakali (3 stations), Lakhandehi (5 stations), Bagmati (4 stations), Bakaiya (3 stations), Ratu (3 stations), Khado (1 station), Banganga (1 station), and Kamala (5 stations) watersheds. The selection of stations prioritized those with minimal missing data, ensuring spatial distribution throughout the respective basins. Our frequency analysis focused on the annual time series of 24-h peak rainfall magnitude for these selected stations, encompassing the period from 2015 to 2020. The study revealed two significant flood events in most basins in August 2017 and July 2019. Notably, eight stations exhibited the highest return period during 12–14 August 2017, while nine stations recorded the maximum rainfall around 13 July 2019. Figure 3 presents a spatial map illustrating rainfall extreme indices across the nine study watersheds. The examination of rainfall extremes indicated an increasing trend in most stations. For detailed insights into the extreme analysis results for each watershed, refer to the sub-headings corresponding to the respective basins.
Figure 3

Trends in extreme rainfall indices for selected nine watersheds.

Figure 3

Trends in extreme rainfall indices for selected nine watersheds.

Close modal

The climatic indices for the Mahakali watershed, specifically RX1day, RX5day, R95pTOT, and R99pTOT, display an increasing trend at both stations, with station 105 showing a significant increase across all four heavy precipitation indices. Cumulative Dry Days (CDD) and Cumulative Wet Days (CWD) exhibit mixed trends at both stations, with CWD significantly increasing in station 105 and decreasing in station 214 (Figure 3). R20 mm shows a varied trend. Total annual wet day precipitation (PRCPTOT) demonstrates a significant increasing trend for station 105 and a decreasing trend for station 214.

In the WRB, heavy precipitation indices increase in stations below 1,000 m but decline at higher elevations. CDD generally increases, while CWD varies across regions. R20 mm decreases in a majority of stations, indicating fewer high precipitation days. PRCPTOT increases in stations below 1,000 m but varies at higher elevations. For the Banganga basin, climatic indices representing heavy precipitation exhibit a predominantly non-significant decreasing trend in the representative station. CDD increases, indicating more dry days, while CWD decreases, suggesting fewer wet days. R20 mm significantly decreases, indicating a decline in high precipitation over time. PRCPTOT also decreases, indicating a reduction in total annual precipitation. In the Lalbakaiya basin, extreme indices such as CDD and CWD increase, while all other extreme indices show a decreasing trend, suggesting decreasing rainfall and increasing dry days. Slightly increasing trends in consecutive wet days imply an increase in days with rainfall greater or equal to 1 mm. In the Bagmati basin, consecutive dry days decrease, but maximum one-day precipitation increases in most parts except the middle portion. Indices signifying heavy precipitation show non-significant increasing trends in stations below 150 m, while CDD increases in all stations. R20 mm decreases in most stations, indicating a reduction in high precipitation days. PRCPTOT shows mixed trends. Lakhandehi basin experiences non-significant increasing trends in heavy precipitation indices for stations below 150 m and increasing CDD in all stations. CWD exhibits mixed trends, and R20 mm decreases in most stations, signifying fewer high precipitation days. Ratu River data reveal an increasing trend in maximum 1-day and maximum consecutive 5-day rainfall, while the number of days with rainfall ≥10, 20, and 25 mm decreases. Dry days increase, and wet days decrease. Annual and seasonal rainfall decreases, except in the pre-monsoon season. The Kamala basin sees decreasing trends in extreme indices at higher elevations and an increasing trend at lower elevations. CDD increases throughout the basin, while R10, R20, and R25 mm decrease, indicating more dry days and fewer days with higher rainfall. Khado River data shows non-significant decreasing trends in heavy precipitation indices, increasing CDD, decreasing CWD, and a non-significant decrease in R20 mm and PRCPTOT, suggesting decreasing high precipitation over time.

Flood extent mapping based on satellite data

The comprehensive analysis of flood extents across multiple river basins in the southern plains of Nepal provides valuable insights into the spatio-temporal patterns and variability of flooding events. The examination of nine river systems highlights the distinct characteristics and challenges associated with each basin.

In the Mahakali Basin, the monthly flood extents during the monsoon season (June, July, August, and September) were determined, averaging 62.5, 38.3, 35.6, and 33.3 sq.km, respectively. The largest flood extent was recorded in June 2022, covering an area of 126.85 sq.km, followed by the June 2015 flood with an area of 99.18 sq.km. An analysis of the 2015 June flooding event indicated a return period of 3.12 years. Moving to the West Rapti River basin, the average flooded area was highest in August (18.1 sq.km), while June, July, and September recorded areas of 9.52, 16.5, and 17.7 sq.km, respectively. The maximum flood extent occurred in August 2017, covering 44.22 sq.km, with an equivalent return period of 40.51 years (Figures 4 and 5). The Banganga River basin experienced extensive flooding in June and July, with average flooded areas of 17.8 and 19.3 km2, respectively. The maximum flooding, observed in June 2019 (41.782 km2), was equivalent to a 1-year return period. In the Lalbakaiya River basin, where dam-induced water levels contribute to widespread floods, the highest average monthly flooding area occurred in July (96.5 km2), followed by August (40 km2), June (17 km2), and September (9.77 km2). The largest flood event in July 2019 covered 188.08 km2, with a return period of 12.82 years. The Bagmati River basin experienced its largest flood in July, covering an area of 118 km2, with average flooding areas during June, August, and September at 18.7, 57, and 18 km2, respectively. The July 2019 flood, with an area of 260 km2, had a return period of 185.40 years. The Lakhandehi River basin recorded its highest average monthly flood extent in July (28.7 km2), with flooding areas of 4.23, 15.2, and 4.31 km2 in June, August, and September, respectively. The maximum flooding occurred in July 2019 (61.65 km2), equivalent to a 185.40-year return period. In the Ratu River basin, the largest flooding typically occurred in July (27.4 km2), while June, August, and September recorded areas of 6.18, 4.62, and 4.05 km2, respectively. The maximum recorded flood extent was in July 2020 (48.80 km2), equivalent to a 1.43-year return period. For the Kamala River basin, the most flooded month was July, averaging an area of 31.6 km2, followed by June (9.75 km2), August (17.0 km2), and September (7.39 km2). The extensive inundation in July 2020 covered 61.15 km2, with a return period of 54.44 years. Finally, the Khando River basin experienced its largest flooding in July, with an average area of 7.20 km2, while June, August, and September recorded areas of 0.87, 4.19, and 1.15 km2, respectively. The largest recorded flood in July 2020 covered approximately 12.5 km2, with return periods of 1.77, 4.39, and 12.68 years for events in July 2020, July 2019, and August 2017, respectively. Additional details and spatial maps are available in the Supplementary Materials.
Figure 4

Spatio-temporal map of monthly flood extents for the Rapti River basin from 2015 to 2018.

Figure 4

Spatio-temporal map of monthly flood extents for the Rapti River basin from 2015 to 2018.

Close modal
Figure 5

Flooding near the border region on 13 August 2017 was caused by heavy rainfall with a return period of 40 years.

Figure 5

Flooding near the border region on 13 August 2017 was caused by heavy rainfall with a return period of 40 years.

Close modal

Model-based flood inundation mapping

The survey data obtained from point measurements were utilized to construct a DEM with a resolution of 10 m, which was subsequently merged with the 30 m resolution DEM from AW3D for the West Rapti River at the identical transboundary flood map location generated through satellite data. DEM generation from survey data and the integration with AW3D were executed using terrain modification tools within HEC-RAS. The model was then deployed to simulate the August 2017 flood event. Fine-tuning of key parameters, such as Manning's ‘n’ and imperviousness percentage, was conducted through rigorous trial-and-error to closely align the inundation extent with the flood map derived from Sentinel-1 satellite data. The conclusive calibrated parameters for various land cover classes are detailed in Table 4. A comparative analysis for binary classification (flooded vs. non-flooded areas) was carried out between the maximum inundation map produced by HEC-RAS and the Sentinel-1 flood map. The assessment revealed a Kappa coefficient of 0.58 (moderate agreement), and an overall accuracy metric of 84.2% (very good performance), which depicts the proportion of true positives and true negatives in a confusion matrix. The hydraulic model was calibrated for discharge and rating curves at the gauging station (Bagasoti and Jalkundi) with an Nash–Sutcliffe efficiency (NSE) value greater than 70%. This shows the HEC-RAS's ability to emulate the observed flood characteristics successfully. However, the estimation of flood depth from satellite-based products still remains a challenge in the scientific community. Figure 6 visually depicts the maximum inundation extent generated by the calibrated HEC-RAS model in comparison to the satellite-based flood map. The model demonstrated a successful prediction of flood inundation at the downstream junction. However, challenges emerged due to the suboptimal resolution of the DEM in the upstream section and interpolation errors associated with the DEM generated from survey points, resulting in an overestimation of flood area within the model. Addressing these intricacies is essential for enhancing the accuracy of flood models in predicting inundation extents, particularly in upstream regions. For better applicability of the satellite products, problems such as border noise, speckle effect, and cloud interference should be dealt with to minimize the associated uncertainties. This further strengthens its applicability for policy-level implications.
Table 4

Calibrated values for Manning's n, per cent impervious, and infiltration parameters

Manning's n
Landcover classRangeCalibrationPer cent imperviousMaximum deficit (mm)Initial deficit (mm)Percolation rate (mm/h)
NoData n/a 0.001 100 
Grassland 0.02–0.05 0.05 25 10 1.25 
Water 0.025–0.05 0.05 95 
Snow n/a 0.035 98 
Forest 0.08–0.20 0.13 10 25 15 4.25 
RiverBed 0.07–0.15 0.12 85 15 10 1.2 
Built 0.12–0.20 0.12 85 10 0.2 
Cropland 0.02–0.05 0.05 15 0.65 
BareRock 0.023–0.03 0.03 60 10 0.45 
Manning's n
Landcover classRangeCalibrationPer cent imperviousMaximum deficit (mm)Initial deficit (mm)Percolation rate (mm/h)
NoData n/a 0.001 100 
Grassland 0.02–0.05 0.05 25 10 1.25 
Water 0.025–0.05 0.05 95 
Snow n/a 0.035 98 
Forest 0.08–0.20 0.13 10 25 15 4.25 
RiverBed 0.07–0.15 0.12 85 15 10 1.2 
Built 0.12–0.20 0.12 85 10 0.2 
Cropland 0.02–0.05 0.05 15 0.65 
BareRock 0.023–0.03 0.03 60 10 0.45 
Figure 6

Validation of satellite-based and model-derived flood inundation maps for the 2017 August flood event. (a) Flood extent from sentinel-1, (b) flood extent from HEC-RAS, and (c) agreement between satellite and model.

Figure 6

Validation of satellite-based and model-derived flood inundation maps for the 2017 August flood event. (a) Flood extent from sentinel-1, (b) flood extent from HEC-RAS, and (c) agreement between satellite and model.

Close modal

In the Mahakali basin, there are notable variations in observed monthly flood extents during the monsoon season, emphasizing the dynamic nature of flooding events. The peak flood extent recorded in June 2022 highlights the need for up-to-date monitoring and predictive models. The return period analysis of the 2015 June flood further emphasizes the importance of understanding historical flood events. In the context of the Banganga River basin, despite limited rainfall station coverage, substantial flooding occurs during the monsoon months. The June 2019 flood event, with its short return period, underscores the importance of considering local factors and topographical nuances in flood risk assessments. The reliance on satellite-based mapping, as demonstrated in this study, proves crucial in regions with sparse ground-level monitoring. The Lalbakaiya River basin displays unique flooding patterns due to fluctuating water levels caused by small dams, particularly pronounced due to the limited excess waterway passage across the transboundary line following the construction of embanked roads. Similar effects are observed in a major river at a transboundary location during flood events. The largest flood in July 2019, coupled with its lengthy return period, necessitates a nuanced approach to floodplain management, considering both climatic and anthropogenic factors. Similarly, in the Bagmati River basin, the observed patterns of flooding highlight the importance of detailed spatial analysis. The July 2019 flood, characterized by its extensive coverage and exceptionally long return period, emphasizes the need for enhanced floodplain modelling and preparedness measures. The Lakhandehi River basin, in the July 2019 flood event, with its substantial return period, underscores the complexity of flood dynamics in this region. The Ratu Watershed, characterized as a rainfed seasonal river, presents unique challenges in flood prediction and management. The observed decrease in rainfall events with higher return periods implies a shifting precipitation pattern, influencing flood dynamics. Spatial maps depicting monthly and yearly flooding extents provide a comprehensive view of flood events, aiding in the identification of vulnerable areas. In the Kamala River basin, the detailed analysis of flood extents and return periods underscores the importance of considering both temporal and spatial dimensions in flood risk assessments. The July 2020 flood, with its substantial return period, indicates the need for adaptive strategies. Finally, the Khado River basin, characterized by its non-perennial nature, exhibits varying flood extents during the monsoon season. The observed flooding events in July 2020, July 2019, and August 2017, each with distinct return periods, highlight the need for basin-specific flood management strategies.

Uncertainties in climate extremes and flood inundation mapping

The rainfall extreme analysis in nine watersheds across Nepal reveals varying trends and patterns. While some regions experienced increased rainfall intensity and frequency, others witnessed decreases. These shifts may have significant implications for water resources management, flood risk assessment, and agricultural practices in these areas. The majority of basins (except Mahakali) showed an increase in CDD, which validates our findings from Subba et al. (2019). On the contrary, there was an asynchronous variation in CWD across studied regions, limiting our ability to recommend effective mitigation strategies. Specifically, the WRB experiences an increase in heavy precipitation indices at stations below 1,000 m but a decline at higher elevations. R20 mm decreases in a majority of stations, indicating fewer high precipitation days. PRCPTOT increases in stations below 1,000 m but varies at higher elevations. Thus, the climate extreme analysis in WRB, characterized by its diverse topography, presents unique challenges in flood management. However, the limited number of meteorologic ground stations available in some of the study basins limits the overall understanding of climatic extremes from this study. Furthermore, the uncertainties add up from neglecting the shift of the East Asia Summer/Winter Monsoon cycle to minimize research complexity. These are complicated yet indispensable for the analysis of climate extremes.

The inundation study across nine river basins in Nepal's southern plains revealed distinct flood patterns where flooding typically peaked during monsoon season, the majority in the month of July with some exceptions. The largest recorded flood extent varied considerably, ranging from 12.5 km2 in the Khando basin to 260 km2 in the Bagmati basin. Return periods, indicating flood severity, also differed significantly across basins as per flood events. The highest flood extent observed in WRB on August 2017, with a considerable return period, emphasizes the severity of extreme flooding events. While we show that the flood extent map derived from radar shows high accuracy as compared to HEC-RAS hydraulic modelling, both of these approaches come with their own limitations and uncertainties. Employing the radar-based flood mapping algorithm in densely forested areas might give erroneous results as the C-band cannot penetrate the canopy structure (Chapman et al. 2015). There are also uncertainties pertaining to the subsequent processes of radar-based flood mapping like in choice of filtering methods and initial threshold determination. Similarly, flood inundation mapping using hydraulic models also possesses uncertainty at various stages, including uncertainty in forcing, model structure, and calibration approaches. The poor gauging facilities near transboundary regions, have added uncertainty in validating the results. The DEM used in this study was interpolated using point-based measurement, which could also affect the modelling results. So future research endeavours should prioritize the enhancement of DEMs, possibly through higher-resolution surveys, to improve the accuracy of hydraulic models. Flooding in Rapti Barrage, India, located immediately downstream of the study boundary can be referred to from the social media video, extracted from YouTube (Bazar 2017), which shows huge flooding in August 2017. However, estimation and validation of flood depth from this particular media source still remains a challenge from both modelling and the satellite-based product's perspective and is considered a limitation of this study.

Creating synergies in under-observed regions

The integration of advanced satellite technologies, such as Sentinel-1, coupled with detailed hydro-meteorological data-driven models, provides a holistic understanding of flood dynamics in transboundary river basins. However, addressing the challenges related to DEM resolution and satellite data interpolation is crucial for minimizing inaccuracies, particularly in upstream areas. Thus, recognizing the limitations and uncertainties in both hydraulic models and satellite-based products, the findings from this study suggest the application of satellite products in cases of fast response during the flooding period especially for ungauged transboundary river systems. The study provides a foundation for broader implications across transboundary river basins, aiding policymakers and emergency responders in developing robust flood risk management strategies.

In regions characterized by data scarcity and with challenges in capturing real-time hydro-meteorological as well as physical terrain data across transboundary areas, the satellite-generated real-time flood map offers more accurate information on river floodplains. Besides that, the computational intricacy of a data-driven model seems to be a reason why researchers are more inclined towards remote sensing products (Ali et al. 2021; Lamada et al. 2023; Shilengwe et al. 2023). For under-observed regions, transferring the hydrological parameters from nearby gauged regions based on the catchment's hydrologic similarity index might help estimate the flood extent to compare satellite-derived flood extent (Burn & Boorman 1993; Oudin et al. 2008; Narbondo et al. 2020; Karki et al. 2023). Also, integrating ML algorithms for precise flood extent predictions, particularly in challenging terrains, holds promise. In conclusion, this research significantly advances our understanding of flood dynamics in the southern plain of Nepal, with an emphasis on the West Rapti River basin. The integration of satellite technologies, hydro-meteorological data, and hydraulic modelling has laid a foundation for improved flood risk management, providing valuable insights on the use of satellite-based products for future studies and interventions, particularly in poorly gauged regions.

The southern plain of Nepal is affected by widespread flooding during the monsoon season of each year. Regarded as the ‘granary of Nepal’, this region, however, is a very important part of the country with substantial contribution to the nation's economy through large agricultural production. It is thus important to study the flooding behaviour and assess its impacts in these regions. This study therefore leverages high-resolution and open-source satellite imagery to delineate past floods in nine flood-prone transboundary basins: Mahakali, West Rapti, Banganga, Lal Bakaiya, Bagmati, Lakhandehi, Ratu, Kamala, and Khando. The study involves assessing rainfall characteristics, creating real-time flood inundation maps from satellite data, performing intricate hydraulic modelling in one of the nine basins using cross-sectional survey data and scrutinizing and validating the model and satellite-based outputs at a transboundary location.

The rainfall frequency analysis identified two major flooding events in most of the basins: during August 2017 and July 2019. The extreme analysis also showed increasing trends of rainfall extremes in most of the basins. The Sentinel-1-derived flood mapping was consistent with the rainfall frequency analysis and showed large flood extents during the two widespread events. Two key outputs of the satellite flood mapping were the spatio-temporal flood inundation map for each month of monsoon season from 2015 to 2022 as well as the monthly time series of the inundation area. Among the analysed nine transboundary basins, the WRB was selected due to clear image availability, larger cross sections of the channel (easily captured through remote sensing products), available input data for hydraulic modelling (rainfall, discharge), and relevant works of literature for justification of the provided inputs. The event-based flood modelling in the WRB using HEC-RAS 2D could identify the flooding in the downstream zone near the transboundary. The extent of monthly maximum flood inundation resulted in an overall accuracy of 84.3% and a Kappa metric of 0.58. While the hydraulic model seemed to accurately pinpoint downstream flooding, it tended to slightly overestimate the overall extent compared to SAR-derived flood maps. This shows the limitation of remote sensing products in capturing peak flood events (maximum inundation duration and extent), and other associated characteristics such as flood depth.

The use of SAR products in transboundary flood studies has proven helpful, but only after a thorough assessment of errors and noise to minimize the uncertainty, where the data constraints still prevail. So, this study brings a multidisciplinary approach to monitoring flooding in the transboundary rivers of Nepal by combining multiple facets from climatology and hydraulics to river morphology. The derived flood inundation maps can be used to prioritize highly affected and frequent flooding zones. These maps can also help in identifying proper sites for river training as well as flood protection measures. In the absence of accurate and reliable data on hydro-meteorological variables, satellite remote sensing-based products, especially SAR contain all-time all-weather data, which is highly useful for monitoring flooding. Notwithstanding, the quantification of depth and duration of the flooding is still a limitation of this study and is earmarked for future research. Let us hope for the satellite products to evolve towards the quantification of these primary and intricate challenges in the near future.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.

Ali
S.
,
Cheema
M. J. M.
,
Waqas
M. M.
,
Waseem
M.
,
Leta
M. K.
,
Qamar
M. U.
,
Awan
U. K.
,
Bilal
M.
&
Rahman
M. H. U.
(
2021
)
Flood mitigation in the transboundary Chenab river basin: A basin-wise approach from flood forecasting to management
,
Remote Sensing
,
13
(
19
).
doi:10.3390/rs13193916
.
Araghinejad
S.
,
Azmi
M.
&
Kholghi
M.
(
2011
)
Application of artificial neural network ensembles in probabilistic hydrological forecasting
,
Journal of Hydrology
,
407
(
1–4
),
94
104
.
doi:10.1016/J.JHYDROL.2011.07.011
.
Armanuos
A. M.
,
Al-Ansari
N.
&
Yaseen
Z. M.
(
2020
)
Cross assessment of twenty-one different methods for missing precipitation data estimation
,
Atmosphere
,
11
(
4
).
doi:10.3390/ATMOS11040389
.
Aryal
D.
,
Wang
L.
,
Adhikari
T. R.
,
Zhou
J.
,
Li
X.
,
Shrestha
M.
,
Wang
Y.
&
Chen
D.
(
2020
)
A model-based flood hazard mapping on the southern slope of Himalaya
,
Water
,
12
(
2
),
540
.
Atiquzzaman
M.
&
Kandasamy
J.
(
2016
)
Prediction of hydrological time-series using extreme learning machine
,
Journal of Hydroinformatics
,
18
(
2
),
345
353
.
doi:10.2166/HYDRO.2015.020
.
Bazar
J.
(
2017
)
Extent of flooding at Rapti Barrage (India) [Broadcast]. Available at: https://www.youtube.com/watch?v=Avf7pGiNjQI.
Borah
S. B.
,
Sivasankar
T.
,
Ramya
M. N. S.
&
Raju
P. L. N.
(
2018
)
Flood inundation mapping and monitoring in Kaziranga National Park, Assam using Sentinel-1 SAR data
,
Environmental Monitoring and Assessment
,
190
(
9
).
doi:10.1007/S10661-018-6893-Y
.
Burn
D. H.
&
Boorman
D. B.
(
1993
)
Estimation of hydrological parameters at ungauged catchments
,
Journal of Hydrology
,
143
(
3–4
),
429
454
.
doi:10.1016/0022-1694(93)90203-L
.
Chapman
B.
,
McDonald
K.
,
Shimada
M.
,
Rosenqvist
A.
,
Schroeder
R.
&
Hess
L.
(
2015
)
Mapping regional inundation with spaceborne L-Band SAR
,
Remote Sensing
,
7
(
5
),
5440
5470
.
Chen
X. Y.
,
Chau
K. W.
&
Busari
A. O.
(
2015
)
A comparative study of population-based optimization algorithms for downstream river flow forecasting by a hybrid neural network model
,
Engineering Applications of Artificial Intelligence
,
46
,
258
268
.
doi:10.1016/J.ENGAPPAI.2015.09.010
.
Cohen
J.
(
1960
)
A coefficient of agreement for nominal scales
,
Educational and Psychological Measurement
,
20
(
1
),
37
46
.
Congalton
R. G.
&
Green
K.
(
2019
)
Assessing the Accuracy of Remotely Sensed Data : Principles and Practices
, 3rd edn.
Assessing the Accuracy of Remotely Sensed Data
.
doi:10.1201/9780429052729
.
Costabile
P.
,
Costanzo
C.
,
Ferraro
D.
,
Macchione
F.
&
Petaccia
G.
(
2020
)
Performances of the new HEC-RAS version 5 for 2-D hydrodynamic-based rainfall-runoff simulations at basin scale: Comparison with a state-of-the art model
,
Water
,
12
(
9
).
doi:10.3390/w12092326
.
Cred
U.
(
2018
)
Economic Losses, Poverty & Disasters 1998–2017
, vol.
33
.
Brussels
:
Université Catholique de Louvain (UCL)
.
Devkota
R.
,
Bhattarai
U.
,
Devkota
L.
&
Maraseni
T. N.
(
2020
)
Assessing the past and adapting to future floods: A hydro-social analysis
,
Climatic Change
,
163
(
2
),
1065
1082
.
Estrany
J.
,
Ruiz-Pérez
M.
,
Mutzner
R.
,
Fortesa
J.
,
Nácher-Rodríguez
B.
,
Tomàs-Burguera
M.
,
García-Comendador
J.
,
Peña
X.
,
Calvo-Cases
A.
&
Vallés-Morán
F. J.
(
2020
)
Hydrogeomorphological analysis and modelling for a comprehensive understanding of flash-flood damage processes: The 9 October 2018 event in northeastern Mallorca
,
Natural Hazards and Earth System Sciences
,
20
(
8
),
2195
2220
.
Ferraggine
V. E.
&
Villar
S. A
. (
2020
) ‘
A robust version of the Lee filter for speckle reduction and contrast enhancement applied to side scan sonar images’
,
2020 IEEE Congreso Bienal de Argentina, ARGENCON 2020–2020 IEEE Biennial Congress of Argentina, ARGENCON 2020
.
doi:10.1109/ARGENCON49523.2020.9505346
.
Foody
G. M.
(
2002
)
Status of land cover classification accuracy assessment
,
Remote Sensing of Environment
,
80
(
1
),
185
201
.
doi:10.1016/S0034-4257(01)00295-4
.
Gautam
D.
,
Adhikari
R.
,
Gautam
S.
,
Pandey
V. P.
,
Thapa
B. R.
,
Lamichhane
S.
,
Talchabhadel
R.
,
Thapa
S.
,
Niraula
S.
,
Aryal
K. R.
&
Lamsal
P.
(
2023
)
Unzipping flood vulnerability and functionality loss: Tale of struggle for existence of riparian buildings
,
Natural Hazards
,
119
(
2
),
989
1009
.
Giustarini
L.
,
Matgen
P.
,
Hostache
R.
&
Dostert
J.
(
2012
)
From SAR-based flood mapping to water level data assimilation into hydraulic models
.
Remote Sensing for Agriculture, Ecosystems, and Hydrology XIV
,
8531
,
186
197
.
Gupta
N.
,
Dahal
S.
,
Kumar
A.
,
Kumar
C.
,
Kumar
M.
,
Maharjan
A.
,
Mishra
D.
,
Mohanty
A.
,
Navaraj
A.
,
Pandey
S.
,
Prakash
A.
,
Prasad
E.
,
Shrestha
K.
,
Shrestha
M. S.
,
Subedi
R.
,
Subedi
T.
,
Tiwary
R.
,
Tuladhar
R.
&
Unni
A.
(
2021
)
Rich water, poor people: Potential for transboundary flood management between Nepal and India
,
Current Research in Environmental Sustainability
,
3
,
100031
.
doi:10.1016/J.CRSUST.2021.100031
.
Hong Quang
N.
,
Tuan
V. A.
,
Thi Thu Hang
L.
,
Manh Hung
N.
,
Thi The
D.
,
Thi Dieu
D.
,
Duc Anh
N.
&
Hackney
C. R.
(
2019
)
Hydrological/hydraulic modeling-based thresholding of multi-SAR remote sensing data for flood monitoring in regions of the Vietnamese Lower Mekong River Basin
,
Water
,
12
(
1
),
71
.
Hossin
M.
&
Sulaiman
M. N.
(
2015
)
A review on evaluation metrics for data classification evaluations
,
International Journal of Data Mining & Knowledge Management Process
,
5
(
2
),
01
11
.
doi:10.5121/IJDKP.2015.5201
.
Hostache
R.
,
Matgen
P.
,
Schumann
G.
,
Puech
C.
,
Hoffmann
L.
&
Pfister
L.
(
2009
)
Water level estimation and reduction of hydraulic model calibration uncertainties using satellite SAR images of floods
,
IEEE Transactions on Geoscience and Remote Sensing
,
47
(
2
),
431
441
.
Kafle
M. R.
(
2020
)
Comparative study of flood impacts and damages by major trans-boundary rivers in Nepal using MCDM
,
Journal of Water Resources and Pollution Studies
,
5
,
7
17
.
Karki
N.
,
Shakya
N. M.
,
Pandey
V. P.
,
Devkota
L. P.
,
Pradhan
A. M. S.
&
Lamichhane
S.
(
2023
)
Comparative performance of regionalization methods for model parameterization in ungauged Himalayan watersheds
,
Journal of Hydrology: Regional Studies
,
47
,
101359
.
Lamada
M.
,
Arfan
A.
&
Afdal Haris
N.
(
2023
)
‘Flood modelling using integration of multi-data analysis and HEC-RAS model in Mata Allo River, Sulawesi’, International Conference on Sciences, Mathematics, and Education (ICoSMEd), 400. doi:10.1051/e3sconf/202340001004
.
Maggioni
V.
&
Massari
C.
(
2018
)
On the performance of satellite precipitation products in riverine flood modeling: A review
,
Journal of Hydrology
,
558
,
214
224
.
Manjusree
P.
,
Prasanna Kumar
L.
,
Bhatt
C. M.
,
Rao
G. S.
&
Bhanumurthy
V.
(
2012
)
Optimization of threshold ranges for rapid flood inundation mapping by evaluating backscatter profiles of high incidence angle SAR images
,
International Journal of Disaster Risk Science
,
3
,
113
122
.
Markert
K. N.
,
Markert
A. M.
,
Mayer
T.
,
Nauman
C.
,
Haag
A.
,
Poortinga
A.
,
Bhandari
B.
,
Thwal
N. S.
,
Kunlamai
T.
,
Chishtie
F.
,
Kwant
M.
,
Phongsapan
K.
,
Clinton
N.
,
Towashiraporn
P.
&
Saah
D.
(
2020
)
Comparing Sentinel-1 surface water mapping algorithms and radiometric terrain correction processing in Southeast Asia utilizing Google Earth Engine
,
Remote Sensing
,
12
(
15
).
doi:10.3390/RS12152469
.
Merz
B.
,
Thieken
A. H.
&
Gocht
M
. (
2007
)
Flood Risk Mapping at The Local Scale: Concepts and Challenges
. In
M. J. F. and H. J. W. Begum Selina & Stive (Ed.)
,
Flood Risk Management in Europe: Innovation in Policy and Practice
, (pp.
231
251
).
Springer Netherlands. https://doi.org/10.1007/978-1-4020-4200-3_13
Mudashiru
R. B.
,
Sabtu
N.
,
Abustan
I.
&
Balogun
W.
(
2021
)
Flood hazard mapping methods: A review
,
Journal of Hydrology
,
603
,
126846
.
Narbondo
S.
,
Gorgoglione
A.
,
Crisci
M.
&
Chreties
C.
(
2020
)
Enhancing physical similarity approach to predict runoff in ungauged watersheds in sub-tropical regions
,
Water
,
12
(
2
),
528
.
Oudin
L.
,
Andréassian
V.
,
Perrin
C.
,
Michel
C.
&
Le Moine
N.
(
2008
)
Spatial proximity, physical similarity, regression and ungaged catchments: A comparison of regionalization approaches based on 913 French catchments
,
Water Resources Research
,
44
(
3
).
doi:10.1029/2007WR006240
.
Pham-Duc
B.
,
Nguyen
H.
,
Phan
H.
&
Tran-Anh
Q.
(
2023
)
Trends and applications of Google Earth Engine in remote sensing and earth science research: A bibliometric analysis using Scopus database
,
Earth Science Informatics
,
16
(
3
),
2355
2371
.
doi:10.1007/S12145-023-01035-2/FIGURES/8
.
Pontius
R. G.
&
Millones
M.
(
2011
)
Death to Kappa: Birth of quantity disagreement and allocation disagreement for accuracy assessment
,
International Journal of Remote Sensing
,
32
(
15
),
4407
4429
.
doi:10.1080/01431161.2011.552923
.
Prakash
A. J.
,
Kumar
S.
,
Behera
M. D.
,
Das
P.
,
Kumar
A.
&
Srivastava
P. K.
(
2022
)
Impact of extreme weather events on cropland inundation over Indian subcontinent
,
Environmental Monitoring and Assessment
,
195
(
1
),
50
.
doi:10.1007/s10661-022-10553-3
.
Prodromou
M.
,
Theocharidis
C.
,
Fotiou
K.
,
Argyriou
A.
,
Polydorou
T.
,
Hadjimitsis
D.
&
Tzouvaras
M.
(
2023
) ‘
Fusion of Sentinel-1 and Sentinel-2 satellite imagery to rapidly detect landslides through Google Earth Engine
’,
EGU23
.
doi:10.5194/EGUSPHERE-EGU23-12618
.
Psomiadis
E.
(
2016
) ‘
Flash flood area mapping utilising SENTINEL-1 radar data
’,
Earth Resources and Environmental Remote Sensing/GIS Applications VII
, vol.
10005
, pp.
382
392
.
Sharma
T. P. P.
,
Zhang
J.
,
Koju
U. A.
,
Zhang
S.
,
Bai
Y.
&
Suwal
M. K.
(
2019
)
Review
of flood disaster studies in Nepal: A remote sensing perspective
,
International Journal of Disaster Risk Reduction
,
34
,
18
27
.
Shen
X.
,
Wang
D.
,
Mao
K.
,
Anagnostou
E.
&
Hong
Y.
(
2019
)
Inundation extent mapping by synthetic aperture radar: A review
,
Remote Sensing
,
11
(
7
),
879
.
Shilengwe
C.
,
Nyimbili
P. H.
,
Msendo
R.
,
Banda
F.
,
Mukupa
W.
&
Erden
T.
(
2023
)
Synthetic aperture radar and optical sensor techniques using Google Earth Engine for flood monitoring and damage assessment – A case study of Mumbwa District, Zambia
,
Zambia ICT Journal
,
7
(
1
),
7
15
.
doi:10.33260/ZICTJOURNAL.V7I1.122
.
Singh
H.
,
Kochhar
A.
,
Litoria
P. K.
&
Pateriya
B.
(
2022
)
Analysis of Long-Term Rainfall Trends Over Punjab State Derived from CHIRPS Data in the Google Earth Engine Platform
, pp.
481
492
.
doi:10.1007/978-981-16-5747-4_41
.
Smith
P. J.
,
Brown
S.
&
Dugar
S.
(
2017
)
Community-based early warning systems for flood risk mitigation in Nepal
,
Natural Hazards and Earth System Sciences
,
17
(
3
),
423
437
.
Subba
S.
,
Ma
Y.
&
Ma
W.
(
2019
)
Spatial and temporal analysis of precipitation extremities of Eastern Nepal in the last two decades (1997–2016)
,
Journal of Geophysical Research: Atmospheres
,
124
(
14
).
doi:10.1029/2019JD030639
.
Sun
S.
(
2011
)
Meta-analysis of Cohen's Kappa
,
Health Services and Outcomes Research Methodology
,
11
(
3–4
),
145
163
.
doi:10.1007/S10742-011-0077-3/METRICS
.
Thapa
S.
,
Shrestha
A.
,
Lamichhane
S.
,
Adhikari
R.
&
Gautam
D.
(
2020
)
Catchment-scale flood hazard mapping and flood vulnerability analysis of residential buildings: The case of Khando River in eastern Nepal
,
Journal of Hydrology: Regional Studies
,
30
,
100704
.
Vieira
S. M.
,
Kaymak
U.
&
Sousa
J. M. C.
(
2010
) ‘
Cohen's Kappa coefficient as a performance measure for feature selection
’,
International Conference on Fuzzy Systems
, pp.
1
8
.
doi:10.1109/FUZZY.2010.5584447
.
Yan
K.
,
Di Baldassarre
G.
,
Solomatine
D. P.
&
Schumann
G. J.-P.
(
2015
)
A review of low-cost space-borne data for flood modelling: Topography, flood extent and water level
,
Hydrological Processes
,
29
(
15
),
3368
3387
.
Yang
X.
,
Zuo
X.
,
Xie
W.
,
Li
Y.
,
Guo
S.
&
Zhang
H.
(
2022
)
A correction method of NDVI topographic shadow effect for rugged terrain
,
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
,
15
,
8456
8472
.
doi:10.1109/JSTARS.2022.3193419
.
Yaseen
Z. M.
,
Ebtehaj
I.
,
Bonakdari
H.
,
Deo
R. C.
,
Danandeh Mehr
A.
,
Mohtar
W. H. M. W.
,
Diop
L.
,
El-shafie
A.
&
Singh
V. P.
(
2017
)
Novel approach for streamflow forecasting using a hybrid ANFIS-FFA model
,
Journal of Hydrology
,
554
,
263
276
.
doi:10.1016/J.JHYDROL.2017.09.007
.
Yaseen
Z. M.
,
Sulaiman
S. O.
,
Deo
R. C.
&
Chau
K. W.
(
2019
)
An enhanced extreme learning machine model for river flow forecasting: State-of-the-art, practical applications in water resource engineering area and future research direction
,
Journal of Hydrology
,
569
,
387
408
.
doi:10.1016/J.JHYDROL.2018.11.069
.
Zhang
X
. (
2020
)
ETCCDI climate change indices. Available online at: https://etccdi.pacificclimate.org/
.
Zhang
X.
&
Yang
F.
(
2004
)
ETCCDI/CRD climate change indices. Available online at: http://etccdi.pacificclimate.org/
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).

Supplementary data