2.1 Study site and its geomorphological setting

River Mujnai (62.45 km) is a tributary of river Jaldhaka which originates from Hantapara uplands (Southern part of Bhutan hill) near Madarihat in the district of Alipurduar (West Bengal) (Figure 1). It is flowing on the vast tract of Alipurduar and Coochbehar districts of West Bengal. The entire unit of Mujnai basin lies on Quaternary (Holocene) alluvial sedimentary foothill zone of Himalaya. This region of North Bengal is recognized as Dooars (Doors of Bhutan) (Rudra, 2018). When this sub-Himalayan river crosses the Himalayan Frontal Thrust (HFT); it debouches on the Terai piedmont zone. Sudden drop of channel gradient creates large-scale sediment accumulation on the channel bed and forms multiple types of dynamic channel bars (Ghosh & Bera, 2022). Rivers like Rehti, Titi, Pagli are being drained to the Mujnai in Alipurduar district of west Bengal. The average annual rainfall of the Alipurduar and Coochbehar district is 3160 mm and 3049 mm respectively (IMD). Foothill zone is also tectonically very active and dynamic (Ghosh & Bera, 2022).

This floodplain is characterized by more or less flat surface and it is dissected by numerous tributaries and distributaries while formation of spill channel is due to drainage congestion (Ghosh & Bera, 2022). This foothill zone makes interlacing drainage system and channel avulsion is occurred due to different fluvio-hydrological factors (Ghosh et al., 2022a). For the current research, the study reach has been divided into three segments (based on channel geometry and geomorphic uniformity) such as Segment–I, Segment–II, and Segment–III. Random sampling (N-26) techniques along both bank sides (Left and right) of river Mujnai have been chosen to improve the accuracy of the result. Most of the previous studies had been considered minimum 20 random samples to assess the river bank vulnerability study and sampling number depends on nature of the river bank and severity of bank erosion.

2.2 Design of new Bank Erosion Susceptibility Index (BESI) model

BESI is the modified form Bank Erosion Hazard Index (BEHI) method. The original BEHI method was propounded by Rosgen (2001). Rosgen's BEHI method is completely based on primary data. Rosgen (2001) considered five principal parameters to identify the stream bank vulnerability zones. These parameters are bank height/bank full height, root depth/bank height, root density, surface protection (including vegetation) and bank angle. Here, some modification should be required for the better performance of the model in addition with some important parameters (erosivity factors) which are responsible for river bank erosion. Similarly, modification is necessary to identify the river bank erosion potential hotspots with improvement of accuracy level of the bank erosion susceptibility model. Previously, we selected these above-mentioned sample sites and finally prepared a bank erosion vulnerability index map. But after monsoon, we visited these sites for practical verification and we got reverse results, that is, moderate bank erosion susceptibility sites were severely affected due to large-scale bank failure and somewhere high bank erosion susceptibility sites (according to Rosgen) were not affected or collapsed. We thoroughly investigated the reasons behind the inverse results and all sample sites have been thoroughly examined. After seasonal investigation, three important factors such as bank material or bank stratification, subsurface hydrological condition and riparian vegetation have been considered for BESI mapping. Here, bank material or bank stratification means composition of river bank. It is composed of sand, silt, clay, boulder, gravel, pebble and cobble, and so forth with various sedimentary structures (cross-bedding, trough and lamination). Sand-bearing bank layer is more prone to erosion compared with clay due to low cohesiveness or stickiness of sand whereas gravel, pebble and cobble have comparatively moderate to high resistant power. Subsurface hydrological condition is classified into five important groups such as dry, damp, wet, dripping and flowing. Dripping and flowing subsurface hydrological conditions accelerate bank erosion rate due to continuous infiltration and percolation processes through the porous medium. Riparian vegetation plays significant role to protect river bank erosion and failure. Deep-rooted trees and few grass species can tightly hold the bank materials. Hence, therefore, BEHI method has been slightly modified. Weightage value of different sub-categories (bank material, subsurface hydrological condition and riparian vegetation) has been properly assigned as per their relative significance. Finally, new map was prepared using Bank erosion Susceptibility Index (BESI) method and it shows high accuracy and precision level for spatiotemporal landuse planning.

In case of BESI method, authors have added three more important susceptibility factors such as bank material, subsurface hydrological condition and riparian vegetation for better performance of the model. Here, bank height (BH) is the vertical depth from top of the bank to the channel bed while the bank full height (BFH) is generally measured by visual estimation of inundation of the river bank during over bank flow or flood. Root depth (RD) means the length of plant roots into the soil while root density (RD₁) denotes the percentage of root in a given unit area of the bank surface. Similarly, surface protection (SP) signifies various natural (including natural vegetation) and anthropogenic impoundment on the bank whereas bank angle (BA) is the angle of inclination of bank with respect to surface. It is measured by Clinometer. Bank material (BM) is the composition of river bank (consolidated, semi-consolidated, non-cohesive materials and hard rock) which is shaped through the geological evolution over time. Sub-surface hydrological condition (SSHC) of river bank means the various hydrological characteristics like flowing, dripping, dry, damp and wet while riparian vegetation (RV) is the existing plants and grasses over the river or stream bank. The ratio of bank height and bank full height (BH/BFH) and ratio between root depth and bank height (RD/BH) have been considered for this new model like BESI. Data regarding river bank susceptibility were obtained through field observation and 26 random samples have been collected along the river from both sides uniformly (Supplementary File S1). Table 1 indicated the individual assigned BESI scores and total weightage value. The total susceptibility values have been categorized into six classes such as very low (8.0–15.20), low (16.0–31.12), moderate (32.0–47.20), high (48.0–63.20), very high (64.0–72.0), and extreme (73.0–80.0).

2.3 Sedimentary bank facies (SBF) analysis

River bank materials, their composition, sedimentological structures and fluvial hydraulics play significant role to erode river bank particularly noncohesive bank (Bera et al., 2019; Lawler et al., 1999). So, it can be studied for the susceptibility and magnitude of river bank erosion through spatiotemporal scale. Here, sedimentary bank facies study has been done in the post-monsoon (December) season of 2021. Three specific bank facies sites have been chosen due to presence of distinct layers sedimentary structures and Site–I is located at Jaldhaka–Mujnai confluence point of Mukuldanga (Coochbehar District). Site–II is located at Pashim Falakata (near Mujnai Railway Bridge) while Site–III is located at Harinathpur Village of Falakata on the left bank of the Mujnai river.

The bank facies study has been conducted in two phases (i) during field and (ii) post field. During the field phase, the facies trench (the prominent layers of lithological units and structures of both the river bank of Mujnai) was identified and measured systematically. Soil samples have also been collected from different prominent layers of bank facies sites to know about the nature of erodibility of the bank. Soil samples have been tested in the laboratory of National Bureau of Soil Science and Land use Planning (NBSSLUP). During post field phase, diagrams of litho facies or river bank facies (Composition, structure and bioturbation) have been prepared on LogPlot version 7.4 software platforms.

2.4 Temporal bank line extraction

In this present study, Topographical sheet of United States Army Map (1955) has been used. Here, firstly the map has been geo-referenced through geo-coding system and projected on GIS environment. MSS (1980), TM (1990), TM (2000), ETM+ (2010), and OLI (2021) multispectral satellite images have been used for delineation of different temporal bank lines of the river. Images have been collected from the USGS earth Explorer (https://earthexplorer.usgs.gov/) site. All the satellite data (Table 2) layers have been stacked and geo-rectified in the Erdas Imagine 2014 platform and projected to the UTM (Universal Tranverse Marcator) projection system on WGS-1984 datum using Arc GIS 10.3.1.

2.4.1 NDWI and MNDWI

For the bank line extraction of river Mujnai, satellite images have been processed through different band combinations. Water line extraction has been done through spectral indices like Normalized Difference Water Index (NDWI) and Modified Normalized Difference Water Index (MNDWI) and the extracted raster data has been converted into vector data.

MNDWI formula was formerly propounded by Xu (2006). Later on, many researchers have worldwide applied and also slightly modified the formula for better result (Gazi et al., 2020; Jiang et al., 2014).

Basically, NDWI is a method to delineate the water body from the different bands of satellite imageries and it differentiates the land border from water bodies through the reflectance and absorption of spectral signatures. It is following as

()

()

where, ρ denotes the reflectance value from the band of Landsat images. However, in case of Landsat TM/ETM+ data, the band combination is Blue (Band-1), Green (Band-2), Red (Band-3) whereas NIR (near Infra Red) is Band-4.SWIR1 and SWIR 2 are bands 5 and 7, respectively.

2.5 Measurement of lateral migration of bank lines, width, channel length, and sinuosity index

Lateral migration or oscillation of river bank is a fluvio-hydrological process. Demarcation of bank lines, lateral migration of channel and changing channel pattern have been computed from different data (topographical maps and satellite images) applying geospatial techniques. Remote sensing and GIS techniques are very helpful for the study of fluvio-hydrological condition of river (Gogoi & Goswami, 2013; Thakur et al., 2011). The process of manual digitization has been performed on the GIS platform for all the consecutive years (Rhoads & Kenworthy, 1995). The entire studied stretch of river Mujnai has been divided into three segments (Figure 2) where 16 cross-sections (CS) have been drawn for the measurement of lateral shifting of bank line in different years. The technique of bank line extraction has been adopted from the given formula (Giardino & Lee, 2011).

()

Where, Dm represents distance of channel migration in m and T1 and T2 denote the years of created polygon for river.

First, layers of multiple polygons of river Mujnai have been created to measure the channel width while polygons of changes of river width have been calculated distinctly for each cross-section of three segments through the measuring tool in Arc 10.3.1 GIS software.

Channel length and channel sinuosity refer the geometric attribute of a river. Channel length within its actual geometry has been measured through different datasets. Sinuosity index is the ratio between actual channel length and valley length of river and it was measured following Schumm's equation (Schumm, 1963).

()

The lateral migration or movement of bank line of River Mujnai has been measured orthogonally with proper geometry. Quantification of bank line shifting has been assessed at 16 cross sections of three segments of river Mujnai. Total erosion and accretion areas have been calculated through geometry calculator in Arc GIS platform to comprehend the nature of erosion and accretion (Baki & Gan, 2012).

3.1 Result of BESI model

26 sampling sites have been selected for the running of BESI model. Bank erosion potentiality zonation map at village level has been prepared (Figure 3) based on calculated BESI value of each site (Table 3). Total 17 villages of Alipuduar district have been assessed through the collected samples. The results denoted that out of total studied villages, 28.74% areas fall under the High, 25.55% areas come under Moderate and 22.90% and 8.03% come under very High and Extreme erosion prone zones as per BESI computation. Paschim Deogaoan, Uttar Deoagon, and Dalimpur villages are highly vulnerable to bank erosion. Falakata and Barodoba villages are comparatively less vulnerable rather Dhulagaon, Hedayetnagar and Purba Jhar Beltali falls under very high vulnerable zone.

Out of 26 sample sites, three sites are located at Paschim and Uttar Deogaon on the right bank of river Mujnai. Here, bank height is more than 2.2 m and bank angle is very high (more than 80°) due to periodic toe erosion of the bank. Consequently, deep rooted grass, shrubs like vegetation and bank protection (like bamboo fence, wooden log. etc.) have been noticed at the Site no. 4, 5, and 6 (left bank). River bank materials are mainly dominated with wet sand and silt most of the sites of river Mujnai.

3.2 Analysis of sedimentary bank facies (SBF)

Spatial variability of the river bank erosion depends on geological setting of the study area. Mujnai river basin is the subbasin of river Jaldhaka. This Himalayan foreland basin is basically composed of Quaternary alluvium deposits and it is tectonically very active and fast changing landscape. When Himalayan Rivers cross the Himalayan Frontal Thrust (HFT), they debouch on the piedmont zone as well as quaternary alluvial flood plain. Different sized sediments with gravel, pebble and cobbles are being deposited on vast floodplain due to sudden drop of channel gradient. Altitude of this plain varies from 50 to 75 m. Here, the SBF site–1 (right bank of river Mujnai), the L1 (0.65 m) is dominated by sand and silt with presence of mica (muscovite) while sub-surface hydrological condition shows semi wet as well as damp (Figure 4a). Similarly, the second layer (L2–1.05 m) of this facies is dominated by clay and silt whereas this layer is very much compacted in nature and it is connected with flowing water. During bankfull discharge, the upper layer is very susceptible to erosion particularly during monsoonal months. The upper layer is partially bioturbated by micro soil animals. This animal activities help to erode particularly during flood.

Around 2.55 m height of left bank of river Mujnai (Site 2) has been exposed through continuous bank failure. It is situated near Paschim Falakata village of Alipurduar district. We have observed three distinct layers at the facies site (Figure 4b). Hence, the upper most layer (L1) is contained of sand and silt and it is indistinctly laminated. This layer is dominated by plant roots and grasses. This layer is (1.1 m) characterized with wet subsurface hydrological condition. Similarly, 0.45 m thick layer 2 (L2) is dominated by silt and clay whereas, the thin mud bands and signatures of oxidation have been found within this horizon and subsurface hydrology shows wet condition. Accordingly, the lower layer or L3 (1.0 m thickness) is also composed of semi consolidated sand and silt (more than 85%) and small proportion of clay (15%) (Table 4). Muscovite and biotite minerals have also found in the lower layer which is connected with post monsoonal water level of river. Erosion capacity of this layer is very high due to lack of compactness or semi-consolidated nature of these materials.

Five prominent horizons have been identified at SBF Site 3 (Figure 4c). Total height of the facies was approximately 2.14 m, where top layer or L1 (0.45 m) is basically composed of sand and silt along with sparse vegetation (Partheniam hysterophorous) and grasses. It is bioturbated along with development of minor cracks. Sub-surface hydrological imprint shows damp condition. Surface crown of cracks has also been developed far away from the bank. L2 (0.26 m) is characterized by dry laminated semi compacted sands. Similarly, L3 (0.23 m) is contained by silt and clay while this layer is also slightly oxidized and it is absolutely bioturbated with damp hydrological condition. L4 (0.16 m) is composed of loose sands which signify indistinct lamination along with dry and damp hydrological conditions. Consequently, the lowermost layer or L5 (1.04 m) is also contained by clay (37%) with silt (42%) and sand (21%) (Figure 5) which shows indistinct lamination. This layer is also characterized by wet hydrological condition due to directly connected with flowing water.

3.3 Channel length and Sinuosity Index (SI)

Channel length of three segments of river Mujnai has been assessed through geospatial technique. The average length of Segment–I is 9.81 km whereas the total length of segment–II and III has been measured around 21.20 km and 13.79 km respectively. After 1990, the channel length of Segment–I (7.20 km) was decreased as a result the SI (1.15) of river channel Mujnai was also decreased. Similarly, after 2010, channel length of segment–II was 21.87 km and SI was 1.72, while Segment–II is being reduced its length after 2010 whereas, the SI was 2.00. It has shown in Table 5 that the channel sinuosity has been increased in 2021 in three segments of river Mujnai. The channel length and sinuosity was 9.71 km and 1.77 in Segment–I, whereas, 19.44 km was the channel length of Segment–II with SI of 2.14 while Segment–III has been measured its flowing path as 14.25 km with SI of 2.59. Channel length is directly proportional with SI.

3.4 Changes of river width

Width of Mujnai River has been significantly changed over the study period. Over the 66 years (1955–2021) of time span, the average width of the channel was 90.06 m whereas the average width of Segment I, II and III was 90.74, 90.61, and 90.13 m, respectively. The maximum width (avg: 129.69 m) of Mujnai was measured in the year 1980. Figure 6 shows that the width of the river has been gradually decreased from the year 1980 to 1990 and further it was increased after 1990. After 2000, river Mujnai was flowing within the narrow channel path while further it has been increased in the year 2021(average width 92.99 m). In the year 1980, CS-4 section showed the maximum width which was 170.63 m (within Segment–I) whereas, in Segment–II, and III, maximum width showed across the section CS-6 and CS-13, respectively. Maximum width in Segment–I, II, and III was noticed across the section CS-4 (99.58 m), CS-11 (97.50 m), and CS-15 (110.30 m) in the year 1990. The maximum width was measured in the section CS-4 (103.72 m), CS-10 (110.63 m), and CS-12 (96.52 m) in the year 2000 in three segments (I, II, and III) of river Mujnai. Although in the year 2010, the average channel width of Mujnai (78.38 m) was decreased but the maximum width was noticed across the section CS-4 (107.26), CS-11 (104.61 m), and CS-15 (102.47 m) in Segment I, II, and III. However, in the recent year (2021), river Mujnai is flowing within comparatively wide channel path rather than previous decades. So, maximum channel width of three segments has been noticed across the sections CS-4 (107.26 m), CS-10 (113.20 m), and CS-15 (102.47 m) accordingly (Online Supporting Information File S2).

3.5 Accretion - erosion dynamics and lateral migration of bank line

River Mujnai oscillates its flow path and shifted bank lines over the entire study period. Figure 7 showing cross section wise lateral movement of channel in three segments (I, II, & III). The maximum length of shifting (320.48 m in Segment–I) has been noticed in section CS-1(right bank) in between the years 1955 and 1980. Right and left bank of Mujnai were shifted across the section CS–3 during 1980 and 1990 while, in between 1990 and 2000, it has shown that maximum lateral movement across section CS–5 in left bank (239.43 m) and minimum movement (64.93 m) in the right bank (CS–2) were calculated. Highest average shifting rate (40.30 m/y) noticed during the period 1980 to 1990 in the left bank of river Mujnai (Segment–I). Accordingly, in between 2000 and 2010, the left and right bank were avulsed slightly and the average rate of channel movement was 3.90 m/y (left bank) and 6.94 m/y (right bank). Hence, in between the years 2010 and 2021, the channel shifted rapidly rather than previous year which was measured as 10.59 m/y for the left and 8.22 m/y for the right. However, in segment–II, both the right and left bank were migrated quickly (avg. migration rate was 14.30 m/y and 17.02 m/y) during the period 1955 to 1980. Figure 7 showed that during 1980–1990, the highest oscillation of right bank (746.49 m) and left bank (634.62) have been recorded across the section CS-11 while the maximum shifting of left and right bank (during 1990–2000) has been measured across the section CS-10. Consequently, the minimum rate of shifting (10.20 m) across section CS-11 in the left bank was noticed. Furthermore, in between 2000 and 2010, the left bank of the channel was migrated with the rate of 12.21 m/y but the right bank was moved comparatively slower rate (5.40 m/y). The bank line of river Mujnai was shifted speedily during the period of 2010–2021 whereas the average shifting rate of right and left bank was 2.27 m/y and 28.92 m/y, respectively (Online Supporting Information File S3).

On the other hand, both the bank line gradually and concurrently shifted within the segment–I during the period 1955–1980. Thus, the significant shifting of right bank was noticed across section CS-13 (159.82 m) and the shifting distance of left bank was 146.37 m (CS-13). Similarly, the maximum rate of migration in the left bank (131.91 m/y) and right bank (117.45 m/y) has been measured in the segment–III during 1980–1990. Maximum shifting of left bank (39.20 m) and right bank (68.39 m) across section CS-12 has been recorded for the one decade (2000–2010) but section CS-13 showed the maximum left and right bank movement during 2010–2021. However, in this period, the rate of migration of left and right bank of segment–III was 41.13 m (Cs-12) and 47.89 m (Cs-13).