1 Introduction

Most of the North–West Himalayan (NWH) river basins are susceptible to numerous catastrophic events like earthquakes, floods, drought, landslides and cloudbursts on account of its distinctive geo-climatic and socio-economic conditions (Bhatt et al. 2010; Bisht et al. 2018; Patro et al. 2009). Almost 12% of the Indian land area suffers from the floods (UNDP 2017). The Indian Himalayan region ecosystem is intrinsically elusive and is characterized by diversified mountain ecosystems, which are tectonically unstable and vulnerable to natural hazards (Arrowsmith and Inbakaran 2002; Buckley et al. 2000; Fort et al. 2010; Geneletti and Dawa 2009; Prasad et al. 2016; Ruiz-Villanueva et al. 2017). A flash-floods is usually caused by extreme rainfall in a small epoch over a relatively smaller area resulting in river stage overtopping embankments, landslides and mudflows and can further cause bridges collapse, damages to buildings, psychological harm to people (Hapuarachchi et al. 2011). It mainly occurs in streams and minor catchments of about a 100 sq. km (Kelsh 2001). These catchments respond quickly to intense rainfall rates due to the presence of steep slopes, saturated soils and soil crusting or human-made structures-induced impermeable surfaces (Hapuarachchi et al. 2011).

Flash-floods mapping is an essential tool for flood preparedness, which can be used to monitor local planning and reduce the vulnerability of susceptible exposures such as urban communities, infrastructure and the agricultural area (Kazakis et al. 2015a, b; Seejataet al. 2018; Sumiet al. 2013). Traditional approach which combines hydrologic–hydraulic modelling and the GIS have been widely employed for flood hazard evaluation and mapping. In order to explain the flood wave movement and to simulate accurate flood hazard in the floodplains, the usage of 1-dimensional (1D), 2-dimensional (2D) and 1D–2D coupled flood modelling approaches offers a detailed explanation of the hydraulic actions of river flow dynamics (Afshari et al. 2018; Dhote et al. 2018, 2019; Rahman and Ali 2016; Thakur et al. 2020). In contrast to data extensive and high computation time modelling approach, RS and GIS techniques have been extensively explored to assess the flash-floods behaviour with high reliability even in a remote location with harsh climate and larger flood-affected areas (Azmeri et al. 2016; Bisht et al. 2018; Elkhrachy 2015; Henry et al. 2006; Hunter et al. 2007; Tanguy et al. 2017). Remote sensing technology is beneficial in terms of spatial, spectral, radiometric and temporal data availability over traditional techniques. With the introduction of new, improved sensors, platforms and implementation techniques, one can better assess the flood variability and extent (Thakur et al. 2016). So far, various studies have been done to map the flash-floods in different countries, such as in China (Liang et al. 2011), Egypt (Mashaly and Ghoneim 2018; Taha et al. 2017), Greece (Kazakis et al. 2015a), India (Panda 2014), Kingdom of Saudi Arabia (Elkhrachy 2015), Pakistan (Memon et al. 2015), Thailand (Seejata et al. 2018) and in various European countries (Gaume et al. 2009). Multiple remote sensing derived indexes such as NDVI (Tucker 1979), modified normalized differential vegetation index (mNDVI) (Jurgens 1997) were used in the past to determine the harms caused to the vegetation and forest area. The historical studies have shown that unwise human interaction with the geo-environment of the area is the main reasons for the damages in terms of life and property (Sah and Mazari 2007).

The morphometric factors of the basin area create the basis for most of the flash-floods-prone area mapping (Gabr and El Bastawesy 2015). For a successful identification of a flash-flood within a catchment, all the hydrologic and hydraulic processes must be identified. The static physical properties such as land use land cover (LULC), soil types, time of travel, slope, roughness coefficient and drainage density together with time-varying properties such as soil moisture could indicate the flash-floods potential of heavy rainfall. These properties influence the catchment response to flash-floods. CN grid which is a function of hydrologic soil group (HSG), LULC and hydrologic condition, represents the human interventions along with the predominant soil moisture of the study region. Thus, this layer could directly indicate the flash-floods possibility of an area. The terrain surface indicator, i.e. digital elevation model (DEM) is one of the most crucial parameters for flash-floods mapping as it allows to extract slope and drainage information of the region. The scale and precision of these topographic layers directly influence the accuracy of flash-floods mapping analysis. (Elkhrachy 2015; Seejata et al. 2018).

The factors triggering flash-floods are needed to be linked with a multi-criteria decision-making approach in a GIS environment. One of the techniques to estimate the relative importance of the factors is AHP (Saaty 1980) that uses a multi-level hierarchical structure of criteria, sub-criteria, objectives and alternatives to solve the complex decision problems and has been used in many applications. Triantaphyllou and Mann (1995) exhibited the importance of AHP in many engineering applications. Aneesh and Deka (2015) used AHP technique to estimate the groundwater potential recharge zones of Bengaluru Urban district. Elkhrachy (2015) determined relative impact weight of factors causing flash-floods using AHP to prepare a flash-floods map of Najran city, Saudi Arabia. Kazakis et al. (2015a) developed a method named FIGUSED and gave relative importance of each parameter using AHP to assess flood hazard areas on a regional scale. Oikonomidis et al. (2015) incorporated AHP to map the potential groundwater zones of Thessaly, Greece. Seejata et al. (2018) assessed the flood hazard areas in Sukhothai province, Thailand using AHP model.

The disaster of flash-floods occurs almost every year in Himachal Pradesh (HP), India, which are accelerated by seasonal hydro-meteorological conditions of the state (Bisht et al. 2018; Fort et al. 2010; Prasad et al. 2016; Ruiz-Villanueva et al. 2017). Extensive surface runoff of shorter duration on small spatial scales causes flash-floods, which combined with deforestation, population growth, land use degradation and economic development act as an accelerating agent in the vulnerability of hazards. Recent flash-flood events recorded in the state of Himachal Pradesh are shown in Table 1.

Table 1 Flash-flood occurrence in Himachal Pradesh from 1990 to 2018
Full size table

Previously, intensive work has been done on the flood hazard mapping on a basin-scale considering rainfall intensity, slope, elevation, river density, soil and land use (Elkhrachy 2015; Kazakis et al. 2015b). However, present work aims to map the flash-floods-prone river reaches in Beas river basin, HP by incorporating Soil Conservation Service Curve Number (SCS CN) as a most critical parameter to cause flash-flooding. In this paper, various layers that encourage flood hazard were incorporated into a GIS environment using multi-criteria decision-making technique to generate flood hazard index. Later, flood hazard index map was integrated with the geomorphology map and river buffer to identify the flash-floods-prone river reaches. The generated flash-floods map was validated using historical reported events, field observations and remote sensing data. The generated flash-floods map can be a basis for effective and efficient floods monitoring, management and mitigation strategies in the Beas river basin.

2 Study area and data

The study area of present research work is Beas river basin, HP, India, with an outlet at Thalout and having a catchment area of about 4964 km2 as shown in Fig. 1.

Fig. 1
figure 1

Overview of Beas river basin

Full size image

It is one of the major tributaries of the Indus River. It is characterized by a narrow North–South valley formed by Beas River and its tributary valleys formed by Sainj, Tirthan and Parvati rivers. The Eastern and northern tributaries of the Beas are perennial and snow-fed, while Southern is seasonal. The sources of surface water in this region are Bhrigu Lake (4280 m) and Beas Kund (3680 m) and one hot spring at Vashist (2000 m) and one at Manikaran (1680 m) in Parvati tributary. The area experiences varied climatic behaviour and low average monthly temperatures due to its elevation differences (938–6570 m). The maximum and minimum temperature during summer is about 25 °C and 10 °C, respectively. During winter, the average maximum temperature reaches around 15 °C, and minimum falls around 0.5 °C (Jain et al. 2015). The area receives an average annual rainfall of about 1085 mm and its flow is maximum during monsoon months. The risk of floods, landslides and mudflows cause severe damage and loss of life in a localized area.

The various dataset (remote sensing, meteorological and physiographical) utilized in the study were collected from several agencies and geo-portals. The details are given in Table 2. Radiometric terrain-corrected (RTC) DEM of ALOS PALSAR having vertical linear error at 90th percentile (LE90) as16 m and horizontal circular error at 90th percentile (CE90) as 20 m was used to extract the morphological factors of the study domain. To process all the procured data and to prepare required thematic layers, ArcMap 10.2, Google Earth 7.3, Google Earth Engine and HEC-Geo HMS (USACE 2016) were used.

Table 2 Various dataset used in the study
Full size table

3 Methodology

The workflow of the present work is shown in Fig. 2. Methodology part is divided into 3 sections. At first, different parameters were selected based on case studies with similar characteristics (Elkhrachy 2015; Fekadu 2018; Kazakis et al. 2015a; Oikonomidis et al. 2015; Seejata et al. 2018). Flood hazard index (FHI) map was generated by implementing AHP on these parameters in a GIS environment. Then, flash-floods-prone river reaches were identified by integrating FHI with a geomorphological map and river buffer and lastly produced results were validated using ground data, historical reports, literature and remote sensing-based index.

Fig. 2
figure 2

Flowchart adopted to generated flash-floods map (inputs, outputs and process); Wi = the weight of each factor

Full size image

3.1 Flood hazard index generation

A flood hazard index is a statistical tool that is used to identify the flood probability, frequency, magnitude, location and areas affected causing a risk to life, health and properties (Federal Emergency Management Agency 2016). Numerous studies have utilized Flood Hazard Index-based approach to prepare a flash-flood map (Azmeri et al. 2016; Dash and Sar 2020; El-Magd et al. 2010; Kazakis et al. 2015a; Papaioannou et al. 2015; Rawat et al. 2012; Seejata et al. 2018).

The 6 factors that have a significant influence in mapping flood hazard index were identified. The selected factors are rainfall intensity (R), curve number GRID (CN), time of travel (T), surface slope (S), Manning’s coefficient (M) and drainage density (D). The AHP model was implemented to estimate the normalized weight (W) of each parameter (Sect. 3.2).

Later, each parameter was classified to five hazard levels defined by its rating score ranging between 2 (for minimum influence) and 10 (for maximum influence). Finally, the flood hazard index was calculated using Eq. 1.

$${\text{FHI}} = \mathop \sum \limits_{i = 1}^{n} Wi \times p = R \times W1 + CN \times W2 + T \times W3 + S \times W4 + M \times W5 + D \times W6,$$
(1)

where n = number of parameters, Wi = weight of each parameter, p = Parameter used (in terms of rating score).

3.2 Analytical hierarchy process (AHP)

The AHP provides a suitable technique for solving complex multi-criteria decision problems. It is a mean of breaking the problem into a hierarchy of sub-problems which can more simply be understood and subjectively assessed (Saaty 1980).

In this study, to assess the potential of each factor related to flood susceptibility, the scale value ranging from 1 to 9 was assigned in the decision matrix. This 9-point scale examines the off-diagonal relationship among the parameters taken. The value 1 corresponds to ‘Equal important’, 3 denotes ‘Moderately important’, 5 denotes ‘Important’, 7 denotes ‘Very important’ and 9 denotes ‘Absolutely important’. The pair-wise comparison of each factor was carried out to produce normalized weight (W). The pair-wise comparison using a 6 × 6 matrix is shown in Tables 3 and 4.

Table 3 Pair-wise comparison of parameters
Full size table
Table 4 Estimation of normalized weights (W)
Full size table

The rainfall intensity and curve number were the factors having the highest normalized weight (34.52), least weight was observed for drainage density (2.8).

3.2.1 Consistency check

The consistency of the created decision matrix (Sect. 3.2) was evaluated using the following index:

$${\text{CR}} = \frac{{{\text{CI}}}}{{{\text{RI}}}},$$
(2)

where CR: consistency ratio; CI: consistency index; RI: random index.

The acceptable CR must be < 0.1. The values of RI are tabulated in Table 5 (Alonso et al. 2006; Lane and Verdini 1989). The RI value depends upon the number of factors (n) used in the AHP. Thus, for 6 factors, RI comes out to be 1.25.

Table 5 Random index (RI) for calculating consistency ratio (CR)
Full size table

CI is calculated using Eq. 3.

$${\text{CI}} = \frac{{\lambda_{{{\text{max}}}} - n}}{n - 1},$$
(3)

where λmax: maximum eigenvalue of the comparison matrix, n: the number of factors.

After analysing Table 3, the value of λmax comes out to be 6.107 and CI is computed to be 0.0214. Eventually, using Eq. 2, the calculated Consistency ratio is 0.017, which is lower than the threshold 0.1; this approves that the weights are consistent.

After assessing the consistency of decision matrix, all parameters were reclassified into 5 hazard levels viz. very high, high, moderate, low and very low (more details in Sect. 4). The rating score was given for each risk level (Table 6) ranging between 2 and 10.

Table 6 Classification of the parameters and their rating score
Full size table

3.3 Flash-floods-prone river reaches mapping and validation

The FHI map obtained by implementing the AHP model was masked for identification of affected river reaches. The masking was carried out by taking a buffer along the rivers of different stream order (higher the order greater the buffer width) within the floodplain boundary of geomorphology map. The masked-out layers were then validated using reported locations of flash-floods obtained from published reports, news agencies.

An attempt was also made to validate the flash-floods map by assessing the spectral signature change post-historical flash-flood events using NDVI at identified flash-floods-prone river reaches. The NDVI time series for a period of 1990–2018 was retrieved using Google Earth Engine. To further validate the result, a detailed field visit was carried out of the study region during October 2018, soon after the flash-flood disaster.

4 Results and discussion

4.1 Factors influencing flash-floods

4.1.1 Rainfall intensity

Rainfall intensity is one of the essential factors in determining the flood hazard. The higher rate of (spatial and temporal) rainfall typically increases the risk of flash-floods susceptibility. Rainfall intensity was represented using the modified Fournier index (MFI) (Luis and Gonza 2010).

The modified Fournier index (MFI) (Eq. 4) can be expressed as the product of Precipitation Concentration Index (PCI) (Eq. 5) (Oliver Oliver, 1980) and Total Annual Precipitation (Pt), i.e. \({\text{MFI}} = {\text{ PCI }} \times P_{{\text{t}}}\) (Apaydin and Erpul 2006).

$${\text{MFI}} = \mathop \sum \limits_{i = 1}^{12} \frac{{p_{i}^{2} }}{{P_{{\text{t}}} }}$$
(4)
$${\text{PCI}} = \frac{{\mathop \sum \nolimits_{i = 1}^{12} p_{i}^{2} }}{{\left( {\mathop \sum \nolimits_{i = 1}^{12} p_{i} } \right)^{2} }},$$
(5)

where \({p}_{i}\) = monthly rainfall; Pt = total annual precipitation.

The monthly rainfall data of 06 in situ ground stations (Banjar, Bhuntar, Janeheli, Larji, Manali and Sainj) and 12 gridded stations (1990–2013) was analysed to estimate PCI and later MFI (Fig. 3).

Fig. 3
figure 3

MFI of all the 18 stations, i.e. 06 in situ rain gauges and 12 IMD rain grid points (Sti = Station number ranging from 0 to 11)

Full size image

The spatial distribution of PCI and MFI was computed by applying Inverse Distance Weighting (Liu 2012; Lu and Wong 2008; Mei et al. 2017) interpolation on the respective values of the stations. The PCI and MFI maps of the study region are shown in Fig. 4a and b, respectively. The average MFI ranges from 135 to 280 (Table 6), with the higher values located in the north-western part of the study area.

Fig. 4
figure 4

a Precipitation concentration index (PCI) map; b modified Fournier index (MFI) map; c CN grid map; d time of travel map; e slope map; f manning’s coefficient map; g drainage density map; h geomorphology map

Full size image

4.1.2 CN grid

To calculate curve number (CN) grid as another most essential factor, land use land cover map, soil map, soil–vegetation–land (SVL) complex, CN look-up table and fill DEM were used as an input in HEC-GeoHMS extension of ArcGIS. The SVL complex was produced from the vector overlay of LULC and HSG maps through a spatial Union.

CN Grid accounts for the type of vegetation, built-up areas, soil type and texture, which determine the water holding and infiltration capacity of the area and consequently affect the flood susceptibility. As a general rule, areas having high CN values are more likely to be flooded. It is important to note that CN is not defined for perennial ice/snow areas; it indicated runoff potential from rainfall when soil is not frozen. Thus we have restricted the flood hazard analysis at high elevation zones of the Beas river basin characterized by permanent snow cover area. CN values range from 69 to 100 (Table 6), with higher values in the Eastern and Northern parts of the study area. CN map of the study region is shown in Fig. 4c.

4.1.3 Time of travel

For the calculation of the time of travel, HEC-GeoHMS extension of ArcGIS is utilized for obtaining river and basin characteristics like River length, slope, longest flow-path and basin centroid (USDA 1986). This method assumes the flow of water through a watershed in the form of open channel flow, sheet flow and shallow concentrated flow, and their combination. The Log Pearson type-III distribution method was adopted to estimate the maximum rainfall of 2-year return period, which is used as an input parameter for the generation of TR55 time of travel. The time of travel in this work is estimated at sub-watershed scale with stream's junctions at outflow points. The flow accumulation grid and stream length map are used as input GIS layers. The spatial distribution of maximum rainfall of 6 ground stations and 12 gridded stations from 1990 to 2013, obtained from Log Pearson’s method were subjected to zonal distribution for estimation of sub-basin-wise rainfall and final time of travel.

Flash-floods is characterized by rapid flow; thus, the time of travel indirectly represents the flow velocity and topography of the area. Hence, a lesser time of travel is assigned the higher rating. The time of travel ranges from 0 to 5.48 h (Table 6). Figure 4d shows the time of the travel map of the study region.

4.1.4 Slope

The slope of an area impacts the amount of infiltration as well as surface runoff. Flash-floods occur where the slope gets steeper, so, it is considered as the fourth of the critical parameters. It was observed that the slope in percent rise ranges from 0 to 550 (Table 6). The slope map of the study region is shown in Fig. 4e.

4.1.5 Manning’s coefficient

Manning’s coefficient, an empirical value, represents the surface roughness characteristics of the area and is dependent on land use. Less surface roughness indicates faster flow velocity and less infiltration, and it is considered as the fifth of the crucial parameters. The surface roughness ranges from 0.001 to 0.15 (Table 6). Manning’s coefficient map of the study region is shown in Fig. 4f. As it is clear from the figure that the forest and vegetation are given the highest roughness value, and minimum roughness is assigned to rocky or snow/glaciated areas.

4.1.6 Drainage density

Drainage density, morphological factor, is the ratio of the length of all channels within the basin and area of the basin, and higher density indicates higher flow accumulation. It is considered as the 6th of the essential parameters, and it ranges from 0 to 26.1 (Table 6). The drainage density map of the study region is shown in Fig. 4g.

4.2 Flood hazard index using a multi-criteria technique

The flood hazard index map (Fig. 5) was governed by rainfall intensity and CN grid equally due to high weights assigned during the AHP procedure. The spatial pattern of the computed flood hazard raster was categorized into five levels of hazard using Natural Break method (Dash and Sar 2020), namely very low, low, moderate, high and very high. The result shows that very high and high zones of flood area cover around 79 and 436 sq. km, respectively, i.e. 1.6 and 8.84%, respectively, of the total area.

Fig. 5
figure 5

Flood hazard index map

Full size image

4.3 Identification of flash-floods affected river reaches

As the areas located near to the main channels are more prone to flash-flooding, the resultant flood hazard map was masked according to geomorphological map and stream buffer.

Geomorphology map (Fig. 4h) gives a hint of evolution and origin of topographic as well as subsurface features such as structural origins, denudation origin, fluvial and glacial origin, and water bodies. As the first step of masking, fluvial origin features were used as a masking layer of flash-floods-affected stretches categorization.

Later, stream order-wise buffer radius (Table 7 and Fig. 6) was defined to implement the second and final step of masking to identify flash-floods-prone river reaches.

Table 7 Stream order-wise buffer radius
Full size table
Fig. 6
figure 6

Stream order-wise buffer radius

Full size image

The prepared flash-floods map after application of masking is shown in Fig. 7.

Fig. 7
figure 7

Flash-floods-prone river reaches in Beas river basin

Full size image

The stream order-wise flash-floods zones and their area according to hazard level are shown in Table 8. It is observed that very high and high flash-floods are mostly experienced by stream order 2, followed by stream order 1, 3 and 4, respectively.

Table 8 Stream order-wise flash-flood zones analysis
Full size table

The analysis shows that almost 3% of the total area lies in the very high flash-floods zone. The river stretches namely Bahang to Manali (Beas), Kullu to Bhuntar (Beas) and Manikaran–Kheer-Ganga (Parvati; refer Figs. 7, 8 and 9 to locate the locations) have been categorized into very high and high flash-floods zones. As there is a need of the optimal design of structural measures for flood control from the local water managers, this case study can be helpful to prioritize the flash-floods-prone river reaches.

Fig. 8
figure 8

Photographs from the field survey conducted in October 2018 along a stretch from Bahang to Bhuntar. a floodplain at Dobhi bridge; b flood mark (4.5 m) at Manali bus stand c flood mark (4 m) at Manali city

Full size image
Fig. 9
figure 9

Locations of the river stretches for NDVI estimation

Full size image

4.4 Validation

The Beas River experienced a flash-flood event on 22 September 2018 along the stretch from Kullu to Bhuntar. The detailed post-flash-flood survey (Fig. 8) was conducted in October 2018 to validate the generated flash-floods map. The flood levels were estimated from in situ flood marks, and relevant data gathered such as field photographs, geographic coordinates of the damaged area and severity of the event information from local citizens.

The historical reports for the period 1990–2018 obtained from the various research papers, news agency and published reports of Himachal Pradesh State Disaster Management Authority were combined to validate the produced flash-floods map (Table 1).

According to Chandel et al. (2014), flood scenario got worsened during 1992–1996, the Beas river experienced a massive downpour in Kullu, Mandi and Kangra districts. According to the Tribune (newspaper, 19th September 1995), the river stretches between Bhuntar and Manali changed its course and caused massive destruction in Kullu valley on both banks of the river. The extreme flash-floods events were reported in 2000, 2001, 2005 and 2007; vast stretches of cultivable Land and crops were washed away. During 2011–2018, the Beas river stretch between Manali and Kullu was the most affected by floods. The loss to public and private property was enormous. All the locations mentioned above (refer Figs. 7, 8 and 9 to locate the locations) fall well within the mapped flash-floods river reaches, thus imply the validation of computed results.

The NDVI trend from 1990 to 2018 was computed using Google Earth Engine (Gorelick et al. 2017) for the identified flash-floods-prone rive reaches (Table 9 and Fig. 9). Landsat satellite images of cloud cover less than 10% were used for NDVI estimation.

Table 9 River reaches identified for NDVI time series analysis
Full size table

From the NDVI time series analysis (Fig. 10) for river stretches such as Solang to Bahang, Manali to Haripur, Dobhi to Biasar, Kullu to Bhuntar and Sainj to Thallot, decreasing trend of NDVI was observed, depicting the region has suffered a vegetation loss due to floods. Also, along with the Kullu to Bhuntar stretch, the NDVI shows a sudden drop in the year 2005, the year which experienced one of the significant flash-flood events.

Fig. 10
figure 10

NDVI trend: a Solang–Bahang stretch; b Manali–Haripur stretch; c Dobhi–Biasar stretch; d Kullu–Bhuntar stretch indicating decreasing trend

Full size image

The present study confirms that AHP could be used in the GIS environment as an efficient method to determine and monitor flash-floods hazards zones. This methodological approach was inspired by various previous work (El-Magd et al. 2010; Gabr and Bastawesy 2015; Kim and Kim 2014; Dash and Sar 2020; Mera et al. 2015; Seejata et al. 2018) and it is clear that the flash-floods hazard is correlated with the combined intervention of many different parameters. However, results can be improved by using high-resolution spatio-temporal images (Rafieeinasab et al. 2015) and hydraulic/hydrologic modelling simulations for efficient flash-floods management and monitoring (Talisay et al. 2019).

5 Conclusion

The primary purpose of the study is to identify flash-floods affected river stretches in the Beas river basin, Himachal Pradesh, India, using a multi-criteria analysis approach specifically AHP model, which facilitates the multi-source data combinations. The adopted methodology spatially analyses the 6 physical parameters, namely Rainfall intensity, CN grid, time of travel, slope, Manning’s coefficient and drainage density. After the application of the AHP model, the higher weights were assigned to rainfall intensity and CN grid while lower to drainage density. The raster calculation in GIS in the environment using assigned weights and hazard level-wise rating score and post masking results to visualize flash-floods-prone river reaches.

The present case study in Beas river basin has revealed the flash-floods-prone areas. The results depicted that the river reaches in the north and east of the Beas basin are susceptible to flash-floods which are mainly governed by heavy rainfall intensity and high runoff characteristics. The river stretches namely Bahang–Manali, Kullu–Bhuntar and Manikaran–Kheer-Ganga have been categorized into very high and high flash-floods hazard zones. Decreasing trend of NDVI was observed for river reaches falling within the very high and high zones indicating the vegetation loss post successive flash-floods events.

The very high and high flash-floods hazard is most experienced by river order 2, then followed by river order 1, 3 and 4, respectively. It infers that tributaries are more prone to floods and might contribute to flash-flood events significantly in this area.

The reliability of the application of this methodology is further confirmed by the reported locations of flash-floods, field survey conducted in October 2018 and remote sensing data. The catastrophic products of the present study can be used to priorities the selection of flash-floods-prone river reaches to construct structural measures of flood protection. This knowledge of the region with high and very high flash-floods hazard zones would allow the government water managers to save public lives and properties. The resulting map can, therefore, provide decision-makers with technical regulatory measures and recommendations for future anticipatory intervention, better planning on land use and control of flash-floods risk.