Spatial modeling of river bank shifting and associated LULC changes of the Kaljani River in Himalayan foothills

Channel dynamics is an inherent characteristic of river in the floodplain region. It has some significant impacts on the ecosystem and human life. GIS-based, DSAS and CA-Markov models are efficient techniques to measure historical and predictive changes following channel shifting and LULC change. In this study, forty-eight years (1972–2020) of earth observatory data were taken to demarcate and detect the channel bank position and LULC change along the Kaljani River, located at the eastern Himalayan foothill. During 1998–2008, a very high rate of erosion has been taken place on both the bankline, which are about −4.48 m/y (left bank) and −3.48 m/y (right bank), respectively. The overall result of the predicted bankline represents that the bulky expansion will occur along the left bank, and sediment accretion will take place at the right bank. Among the three zones, both banks of zone ‘A’ (lower part of the river) are worst affected in the past and present and will follow the same trends in future. The LULC change of all six classes from 1972 to 1998 is very high compared with the changes between 1998 and 2020. Moreover, long profile, hypsometric curve value, and the Soil Conservation Service Curve Number (SCS-CN) value have a significant help in understanding and identifying consequences reasons. The accuracy level is validated by the actual bankline positions (2020) with predicted bankline (2020) and actual LULC (2020) to predicted LULC (2020) empirically with RMSE and statistical test. The accuracy level of this study is conducted with the Kappa statistics for LULC map of 2020, and the result is 87.57%, and bankline shifting RMSE varies from 0.007 to 0.176. Therefore, the prediction output serves as the spatial guidelines for monitoring future trends of channel shift and land use planning management.


Introduction
Systematic assessment on river bank erosion-accretion and its position are very complex and challenging tasks in the field of geomorphology (Lawler 1993). The morphological position, changes of the river channel, especially its lateral migration, erosion-accretion, and bank instability, are the most important research themes in geomorphology and engineering platforms nowadays (Ren et al. 2018;Langat et al. 2019;Suhaimi et al. 2018;Langat et al. 2019). The effect of extreme bank erosion contributes almost 90 per cent of sediment loads in tropical rivers (Steel and Milliken 2013). In the foothill's region, a common characteristic of the alluvial river basin area is bank erosion and its associated problems (Ercan and Younis 2009). Similarly, river banks raise significant economic and environmental issues such as loss of infrastructure and agricultural land due to rampant bank erosion (Bhunia et al. 2016; Ashraf and Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/ s00477-021-02147-1. Shakir 2018), which creates significant threat to local people. The vast dimension of floodplain areas is one of the most attractive land resources for human society but worldwide these plains are facing land degradation by riverbank erosion and huge land use alteration (Hazarika et al. 2015;Bhunia et al. 2016;Debnath et al. 2017). The channel erosion-accretion process changes LULC on the bank adjacent village area that has caused hazardous exposure to the local people. (Thakur et al. 2012;Hasanuzzaman et al. 2021). At present, earth observatory tools like remote sensing (RS) and geographic information systems (GIS) have played an important role in various fields of geography (Wang and Mei 2016;Wang and Xu 2018). But, in the morphological field, the use of various GIS tools is limited, especially for the study of riverbank shifting and future prediction. Therefore, some models are applied to predict of bankline position like higher-order polynomial, exponential model, cyclic series models (Li et al. 2001). The DSAS model is most valuable and straight forward (Li et al. 2001). This present study tries to relate both channel morpho-position changes (Debnath et al. 2017;Lawler 1993;Jana 2019) and LULC changes (Ahmed 2012;Mondal et al. 2016;Maviza and Ahmed 2020) with their future prediction (Miall et al. 2018;Mansour et al. 2019;Nurwanda and Honjo 2019). The hydro-geo-morphic data on the multi-temporal scale (long, intermediate, and short-term) use modern geospatial tools and techniques. Therefore, the digital shoreline analysis system (DSAS) is a highly acceptable method developed by the United States Geological Survey (USGS). This model can accurately measure the rate and predication of different river bank line positions (Right and left bank separately) (Thieler et al. 2009;Kankara et al. 2015;Bhunia et al. 2016;Ashraf and Shakir, 2018;Jana, 2019). Generally, the DSAS model has been used in the context of the sea shoreline. However, in river study, the right and left bank can be separately mapped with a higher degree of accuracy. The CA-Markov model can predict LULC changes, and it can also delineate the LULC change pattern (Xiao et al. 2012;Hamad et al. 2018;Mansour et al. 2019;Nath et al. 2020;Wang et al. 2020;Du et al. 2020;Hasanuzzaman et al. 2021). Thus, this research is significant and rare in its content. This work has been developed to determine the historical bank line and calculate erosion-accretion and LULC change using the field verification graphical information and mathematical methods. The effort to measure erosion-accretion and LULC change of Kaljani River adjacent village area is absent, making this study as exceptional. According to the annual flood report, Kaljani River area is geomorphologically active for its erosionaccretion nature (West Bengal annual flood report 2013; 2014; 2016). Therefore, this work attempts to measure the erosion-accretion and LULC change over four decades and predicts future trends for the year of 2025, 2035, and 2045. Thus, channel and LULC change management are necessary for the sustainable use of land resources, protection of people's property and lives (Thakur et al. 2012;Debnath et al. 2017).

Study area
The Kaljani River is a tributary of Torsha River, originated in Bhutan at the foothills of the Himalayas, and it flows from north to south via Bhutan and India; confluences with Torsha River that again joins in the mighty Brahmaputra River and eventually merges with Padma River to reach the Bay of Bengal. The Kaljani River flowed the undulating Bhutan Himalaya terrain with the alluvial fans. The Terai plain thus covers both the 'Bhabar' and continuous plains of 'Terai' downhill. The major tributaries of the Kaljani River are Dima, Nonai, etc. The study area is located on the seismo-tectonic unstable foothill terrain of Bhutan Himalaya, which is produced an array of magnificent landscapes, both physical & cultural involving multiple cycles of fluvial erosions (Goswami et al. 2012).
The Kaljani River is situated in the Eastern and North-Eastern parts of Alipurduar and Cooch Behar district in West Bengal. The region extends from 26°43′08'' N to 26°1 6′30'' N latitude and 89°25′17'' E to 89°34′56'' E longitude. The length of the river within the study area is 80.5 km. The authors have identified 45 Mouzas (smallest administrative unit for revenue collection) along the Kaljani River buffer zone (adjacent village areas). These adjacent village areas have a maximum stretch of 5.88 km and a minimum of 1.64 km from the channel centerline. For the simplification of the study, the adjoining village area is divided into three zones, namely (A) Deocharai to Ambari stretch, (B) Ambari to Dakshin Paitkapara stretch, and (C) Dakshin Paitkapara to Gabaur Bachhra forest stretch spacing around 26.8 km (Fig. 1).

Database preparation
In this study, Landsat MSS (Multispectral Scanner), TM (Thematic Mapper), ETM?(Enhanced Thematic Mapper Plus), and OLI (Operational Land Imager) datasets have been collected for 1972, 1987, 1998, 2008, and 2020 and used to demarcate the channel bank lines and LULC change detection (Table 1). To maintain the data quality, all images have been co-registered using the first-order polynomial model with the accuracy of root mean square error (RMSE) of less than 0.5 pixels with a minimum number (here it is 5) of ground control points (GCPs). Based on the axial length of river stretch and meander nature, the entire selected course of Kaljani River has been divided into three distinct zones (A, B, and C) with an extension of 26.8 km length (Fig. 1). Afterwards, around 378 transects (on both banks) are generated in each of these zones to estimate the riverbank shifting/ erosion-accretion rate. Copernicus satellite imagery has also been used for measuring river bank embankment length and verification  of primary data. The work has been carried out as per the following methodology (Fig. 2).

Bank lines extraction
This process has been adopted to assess earlier bankline position with the help of selected satellite images. Authors have used the normalized difference water index (NDWI) after McFeeters (1996); Haque et al. (2020) and modified normalized difference water index (MNDWI) (Xu 2006) for bank line extraction based on Eqs. 1 and 2, which employed green and NIR bands for segregation of land from water, where pixels for water features are assigned as '1' and for land as '0' to achieve a binary image.
To estimate the MNDWI, the MIR band of Landsat 7 and SWIR band of Landsat 5 and 8 along the green band are used. The technique for calculating the MNDWI was given by Xu (2006) as: The NDWI and MNDWI final images are used for digitizing the bankline position (right and left banklines separately) by ArcGIS software (de Bethune et al. 1998;Jana, 2019).

Estimation of erosion-deposition rate and its prediction
In the present work, the Digital Shoreline Analysis System (DSAS) extension tool of ArcGIS is used to assess the banklines' erosion-accretion rate. Subsequently, its prediction is also estimated by using the reference extracted baselines and auto-generated transects. For the DSAS based statistical output, two further models have been employed, the End Point Rate (EPR) model for computing present erosion-accretion of the bank lines or shifting rate and the Linear Regression (LRR) model to estimate the shifting of future banklines. The rate of change in the position of banklines are frequently applied to summarize the historical bankline shifting and their future prediction. The model is based on the assumption that the observed periodical rate of change of bankline position is the best estimate for prediction of the future bankline (Fenster et al. 1993) and no prior knowledge regarding the flow discharge or sediment transport is required because the cumulative effect of all underlined processes is assumed to be captured in the position history (Li et al. 2001).
In End Point Rate (EPR) model, based on the availability of data, the studied time period is divided into four temporal datasets, i.e., 1972-1987, 1987-1998, 1998-2008, and 2008-2020 (Fig. 3). The superimposed technique has been portrayed for each dataset to demarcate bankline positions and achieve a final line of overlapping visualization. This line is found out as a superimpose line.
Afterward, a buffer of 100 m distance from the superimpose line is used to drawn towards the right for the right bank and left for the left bank to demarcate baselines. Therefore, transects have been placed at a 50 m gap on the baseline. These transects are created at the acute angle to the baseline up to 3.5 km distance away from both banks. These transects are auto-generated with±0.5 m uncertainties depending on the orientation of the baselines. Moreover, around 1135 transects are placed along the baseline with 50 m spacing to cover the entire selected tract of the Kaljani River (about 80.5 km) (Fig. 3).

EPR ¼
Distance of bankline movement Time between earlier and recent images ð4Þ In EPR model, previous and recent data of two banklines are needed for this calculation and do not require any prior knowledge regarding hydraulic interference or sediment transport. Moreover, the model uses data of two years at a single time. For example, the model calculates the  1972and 1987. After that, 1987-1998then 1998and finally 2008 to calculate the riverbank erosion-accretion rate, which depict the shifting trend over periods. EPR is applied to calculate the rate of bank line migration and understand the erosion-accretion nature Jana 2019). Therefore, we have used the 'Y' for the earlier (Y eb ) and the recent (Y rb ) bank line. In this attempt, it is used as 'Y' to denote the projected bank line position. which is estimated by following Eq. 5 where, X is the time interval (X eb À X rb ) between earlier bankline (X eb ) and recent bankline (X rb ), a EPR is model intercept, b EPR denotes the rate of riverbank shifting (slope or regression coefficient).
On the other hand, EPR intercept is calculated by Eq. 6.
The rate of bankline migration for a given set of transects, the b EPR is calculated by the Eq. 7 3.3.2 LRR model for predicting the bankline erosionaccretion/shifting rate LRR model uses statistics of model generated baseline, which is demarcated by the temporal period of bankline migration. It has shown the bank position of the subsequent year of selected timespan. Therefore, the channel side position of the data set 2020 is considered as a common baseline to all sets. The result of this attempt has been scrutinized by the least-square method (fitting a regression line) to predict the channel shifting and bank line position (Thieler et al. 2009). For this, a regression line is placed to all linear series, points along a particular user transect. Afterwards, the river bankline migration rate is estimated by fitting the least-square regression lines. This process has been used to all banklines for a particular transect. Therefore, this method is used for predicting the position in short-term (2025), intermediate-term (2035) and long-term (2045) basis with a period of 5, 15 and 25 years, respectively. Moreover, the position of the bank line of 2020 is predicted for accuracy assessment. Then the value of EPR is used to predict the future riverbank positions (Y pb ). This is because the predicted riverbank position (X pb ) can extend beyond the recent riverbank (either left or right). Hence, Eq. 5 is modified and formulated through LRR by the following Eq. 8.

LULC change analysis
In this study, five Landsat images are used, specifically 1972 (MSS), 1987(TM), 1998(TM), 2008(ETM?) and 2020 (OLI), for estimation of LULC imprints of the study area. For analysis of LULC change, existing bank landscape of the study area have been classified into six classes such as (i) waterbodies, (ii) dense forest, (iii) open forest, (iv) agricultural land, (v) built-up area, and (vi) fallow land. The maximum likelihood algorithm, a very useful and common supervised classification method (Debnath et al. 2017;Wang et al. 2020;Du et al. 2020), is used for this classification using ArcGIS environment. Numerous LULC change detection techniques are successfully used for monitoring land-use temporal variation (Kaufmann and Seto 2001;Maviza and Ahmed 2020). This method has developed an array of "from-to" matrix-like, pixel conversion matrix, area conversion matrix, and percentages conversion matrix on a pixel-by-pixel basis.

Prediction of LULC change using CA-Markov model
The stochastic based CA Markov model, a popular model for LULC changes prediction, has been employed by the 'Terrset' software package. CA filter along with Markov chain strategy developed CA model. The CA model can be stated as follows: where, S is the set of limited and discrete cellular states, N is the Cellular field, t and t ? 1 indicate the different times, and ʄ is the transformation rule of cellular states in local space. Markov model is depicted of LULC change predication of following mathematically conditional probability formula.
where, S(t) is the system status at the time of t, S(t?1) is the system status at the time of t?1; P ij is the transition probability matrix in a state which is calculated as follows: where, P denote the transition probability; P ij stands for the probability of converting from current state i to another state j in next time. P n is the state of probability of any time.
The transition matrix needs each line factor 0 to 1, and each rate is a non-negative quantity. The estimate of the Markov chain is the relative frequency of transitions recognized the whole period and result of the estimation can be used for prediction (Mondal et al. 2016). This study predicts four future years, such as 2020, 2025, 2035, and 2045. The transition probability matrix has been calculated from 1972 to 2008 for the prediction of LULC of 2020. Also, the transition probability matrix has also been calculated of 1972-2020 to predict LULC of 2025, 2035, and 2045.

Model validation methods
DSAS model has been used for estimating the future riverbank erosion-accretion, shifting and future bankline position. But before the future prediction, the model has to be validated with the current circumstances Jana 2019). Therefore, the LRR method predicts future bank line position based on EPR (slope), interval, and intercept value.
Moreover, the LRR method has been used for the 2020 future bankline prediction. This 2020 predicted bankline is verified with the actual bank line 2020, which was digitized from the 2020 satellite image. Also, the 2020 satellite image was verified with 100 GPS location points (100 transects).The positional error in the above mentioned model-based outputs is estimated by RMSE. It is carried out using Eq. 12.
where, X mb and Y mb are model estimated bankline, and X ab and Y ab are the actual bankline in X (time) and Y (position) coordinates the sample points. In this study, the kappa index (Congralton 1991; Keshtkar et al. 2017) and chi-square (χ 2 ) test (Nath et al. 2020) statistics have been developed for the accuracy assessment of LULC images. The kappa coefficient has been employed for the LULC image classification accuracy level tested with the actual points from field verification and high-resolution Copernicus satellite data. The kappa coefficient has calculated using the following formula (Congalton and Green 2009); where, i is the class number, n is the total number of classified pixels that are being compared to actual data, n ii is the number of pixels belonging to the actual data class i, that were classified with a class i, C i is the total number of classified pixels belonging to class i and G i is the total number of actual data pixels belonging to class i. For the chi-square (χ 2 ) test, the actual land use of 2020 was compared with the predicted 2020 land use based on the CA-Markov model (Nath et al. 2020). We assume that the area statistics of predicted and actual image are the same.

DSAS based riverbank migration/erosionaccretion rate
The entire course of Kaljani River has been segmented into three distinct zones (A, B, and C) at an interval of 26.8 km ( Fig. 1) based on the axial length of river stretch and meandering nature. Afterwards, around 378 transects (on both banks) have been generated in zone A, B, and C to estimate of riverbank migration/ erosion-accretion rate.
The riverbank shifting trend is estimated (Figs. 4 and 5) by considering 48 years of data . The result depicts that in 1972-1987 the average rate of bank line shifting in zone A is-4.61 m/y (erosion) at the left bank, and the right bank is 4.22 m/y (accretion) respectively. In zone B, the rate of average bank line shifting is 0.57 m /y at the left bank and 5.88 m/y at the right bank. The rate of average bank line shifting in zone C is 2.59 m/y and 2.0 m/ y for the left and right banks, respectively. In this period, the overall average shifting of the left and right bank is 2.54 m/y and 2.00 m/y, respectively and the positive shifting (accretion) are also observed on both the banks in each zone, namely A, B and C. The accretion rate and high no of transect of accretion indicate that the channel is narrowing, caused by sediment accretion in both the bank lines. Therefore, the dynamicity of the river is very high, especially in zone A compared to other zones. During 1987-1998, the average channel migration rate in zone A is recorded as-4.21 m/y (erosion) in the left bank and 13.22 m/y (accretion) in the right bank. In this zone, no of transects of accretion is 205 (left bank) and 190 (right bank), which also indicates the process of channel narrowing due to high sediment accretion on right banks. In zone B, the rate of average shifting is −0.16 m/y (erosion) for the left bank and 1.03 m/y (accretion) for the right bank. The channel migrates towards the left bank due to more erosion on the left side (207 transects out of 378 transects) and accretion on the right side (171 transects out of 378 transects). Zone C is resulted in an average channel migration rate of 1.13 m/y for the left bank and −0.10 m/y for the right bank, respectively. In this zone, river moves the right side due to more erosion on the right bank (219 transects out of 378 transects). In this time frame, the average channel migration rate is 1.13 m/y (left bank) and −0.10 m/y (right bank), respectively. In this time, the right bank is experienced extensive erosion and excessive sedimentation (accretion) along the left bank.
During 1998-2008, the average riverbank migration rate in zone A is 8.23 m/y (very high accretion) on the left and −1.50 m/y (erosion) on the right bank. It indicates the channel shift towards the right (erosion), and the left bank observes a huge accretion condition. Zone B comprises significant accretion in-4.47 m/y on the left bank and 11.44 m/y on the right bank (Fig. 4). Therefore, the river channel is narrowing by inward sedimentation. In addition, zone C passes through a relatively higher erosion rate at Fig. 4 DSAS model derived riverbank migration rate (accretion*erosion) during the periods of (a) 1972-1987, (b) 1987-1998, (c) 1998-2008, and (d) 2008-2020 at three selected zones (A, B, and C) −4.48 m/y (left bank) and −3.84 m/y (right bank), respectively. The correspondences of high erosion at both banks indicate the channel widening. In this period, an overall negative (erosion) trend on both bank lines have been observed, which are about −4.48 m/y (left bank) and −3.48 m/y (right bank), respectively (Fig. 4). This result indicates the river course widening triggered by persistent erosion on both banks.
In the time frame 2008-2020, zone A records a higher erosion rate on both banks (−0.23 m/y at left bank and −3.79 at the right bank). Out of a total of 378 transects, the 209 transects in the left bank and 186 in the right bank are erosional transects in this zone. This result uncovers channel widening through the river channel shifting. In zone B, high accretion (2.54 m/y) is recorded in the left bank and high erosion (−2.83 m/y) in the right bank. This zone also conforms to the shifting of the channel towards the right bank line in response to the higher rate of erosion on the right bank and a relatively meagre rate of accretion on the left bankline. Zone C shows the leftward shifting of the river course with an average riverbank shifting rate of −0.44 m/y at the left bank and 2.12 m/y at the right bank. The nature of river shifting indicates the leftward shifting of the channel. In general, during the period 2008-2020, the overall migration rate is −0.44 m/y at the left bank and 2.44 at the right bank, indicating the river channel's leftward shifting (Fig. 5). As a result, the river channel migrates towards the left bank with large extents of sedimentation at the right bank.

Temporal variation of LULC
The temporal variation of LULC categories for 1972, 1987, 1998, 2008, and 2020 is depicted in Fig. 6. Table 2 illustrates the LULC change of the Kaljani River adjacent village area. There has been a significant increasing trend for agricultural land, open forest, and built-up area, whereas a decreasing trend is observed for water bodies, dense forest, and fallow lands. The change of all six classes from 1972 to 1998 is recorded very high when compared with the change between 1998 and 2020. A gradual increase rate from 1987-1998, 1998-2008, and 2008-2020 in open forest, agricultural and built-up areas have been observed, while water bodies and dense forest have decreased rapidly (Table 2). Throughout the given investigation period, the area matrixes of the different LULC types reveal that the dense forest and water bodies have been converted into other land classes, especially water bodies converted to agricultural land, built-up area, and fallow land gradually (Table S1).

Model based prediction of bankline shifting
The result depicts in the Short-term prediction (from 2020 to 2025) that the average rate of bankline shifting in zone A will be −2.61 m/y at left and the right bank will be 4.37 m/ y, respectively. In zone B, the average bankline shifting rates will be −1.07 m/y (left bank) and 4.69 m/y (right bank). The rates of average bank line shifting in zone C will be −2.09 m/y and 3.04 m/y for the left and right banks, respectively. In this period, the overall average shifting of the left and right bank will be −1.90 m/y and 4.00 m/y, respectively. The negative shifting (erosion) will be observed on the left bank in A, B, and C zones. The predicted bank line in 2020 is validated with the help of empirical data from field survey (2020) and image-based (2020) assessment (Fig. 7). Therefore, the ability of the model in evaluation and validation is stretched with field observation.
The result of the medium-range prediction (from 2020 to 2035) of the average channel migration rate in zone A will be −3.15 m/y for the left bank and 2.78 m/y for the right bank. In zone B, the rate of average shifting will be 0.22 m/ y for the left bank and 4.53 m/y for the right bank. Zone C resulted in an average channel migration rate of −0.18 m/y and 0.55 m/y for the left and right banks, respectively. In this time frame, the overall average channel migration rate will be −0.96 m/y (left bank) and 2.56 m/y (right bank), respectively. This result is equiponderant to the intermediate rates of bankline shifting through erosion-accretion between the predicted bankline of 2025 and 2035 (Fig. 7). During 2020-2045 (Fig. 7), the average riverbank migration rate in zone A will be −5.36 m/y (very high erosion) on the left and 1.47 m/y (accretion) on the right bank. It indicates the channel shift towards the left bank (erosion), and the right bank observes accretion condition. Zone B comprises significant erosion in −0.55 m/y on the left bank and 3.51 m/y on the right bank (Fig. 4). In addition, Zone C resulted in an average channel migration rate will be 0.26 m/y and −0.39 m/y for the left and right banks, respectively. In this period, the overall negative (erosion) trend on the left bankline will be about −1.77 m/y and 1.48 m/y at the left and right, respectively. It is also recorded that the absolute bankline migration in the Kaljani River is significant and as high as for the intermediate predicted period from 2035 to 2045. Therefore, to a large extent erosion, and accretion overreach between the actual and predicted bankline positions in 2020 and 2045. In this b Fig. 5 Distribution of DSAS model derived riverbank erosion and deposition rate along transects during the different study periods, (a) 1972-1987, (c) 1987-1998, (e) 1998-2008, (g) 2008-2020, (i) 2020-2025, (k) 2020-2035and (m) 2020-2045 at the left bank and (b) 1972-1987, (d) 1987-1998, (f) 1998-2008, (h) 2008-20,202, (j) 2020-2025, (l) 2020-2035, and (n) 2020-2045 at the right bank period, the left bank will be experiencing extensive erosion, and the right bank will experience excessive sedimentation (accretion). However, such kind of bankline migration may become exceptional and extensive in the future due to the varied nature of different drivers of bankline erosion and accretion process. The overall result of the predicted bankline represents that bulky expansion will occur along the left bank, and sediment accretion will take place at the right bank. Based on the overlaid analysis of bank shifting and mouza distribution, the result suggests that Uttar Paitkapara, Jaigir Chilakhana, Chhatoa, Kaljani, Bhelapeta, Dakshin Latabari, Nimtijhora Tea Garden, Kholta, Chalnipak, Bhelakopa Dwitia khanda, Ambari, Dakshin Paitkapara, Chalnipak, and Amlaguri Dwitia khanda mouzas are vulnerable to erosion, which is captured in the actual rate of change relevant to predict the future riverbank position. Also, the accretion process is active in Kholta, Ambari, Dakshin Paitkapara, Chengpara,  Bhuchungmari, Jaigir Chilakhana, Uttar Paitkapara, Chapatali, AiraniChitalia, Chhatoa, Paschim Salbari, Pukuritola, Dhopguri, BholarDabri (CT), Amlaguri, and GabaurBachhra mouza. The Ambari and Jaigir Chilakhana mouzas are the most dynamic mouzas for both erosion and accretion effects.

Predicted LULC 2025, 2035, and 2045
The transition probability matrix is generated using LULC maps of 2008 and 2020. The CA-Markov method used these results for LULC prediction maps (2020, 2025, 2035, and 2045) (Fig. 6). This work has revealed that if the present spatio-temporal LULC change trends continue, only 12.91 per cent of the total dense forest will be remain by 2045.

Model validation and evaluation
RMSE and t-test are adopted for the model validation in estimated bank lines (left and right), which is accurate between actual and predicted bank lines (Table 4). The positional errors at each transect point are placed by error  (Table 4) with an overall mean error of 0.05. The t-test results reveal that the model has a good prediction capacity (p\0.05). So, this result is accurately matched with the predicted bank line position corresponding with the actual bank line position (Table 4). The Kappa statistics result for the years 2020 is 87.57%. This level of accuracy indicates that the classification and prediction LULC maps are acceptable. Therefore, Table 5 depicts the chi-square (χ 2 ) test result values that indicate the validation and acceptance of the CA-Markov model for the LULC maps prediction.

Discussion
The Kaljani River has continuously changed the LULC of its buffer area through the erosion-accretion process over time. The riverbank erosion and accretion are strongly related to river migration and channel expansion. This study shows that from 1972 to 1998, large-scale bank erosion-accretion and channel migration took place. In this river course, both banks of zone 'A' are affected and, areas located in Jaigir chilakhana, Chilakhana, Bhelakopa dwitia khanda, Amlaguri, Bhelapeta, Panisala mouzas have experienced a large amount of erosion-accretion and channel migration. The result of erosion and accretion represents that channel widening takes place in this zone through continuous and overturning erosion. After 1998,  the nature of bank erosion and accretion has changed. Also, the LULC result has depicted high erosion from 1972 to 1998, and after 1998 the change has gradually decreased. On the other hand, zone 'B' and 'C' are comparatively unaffected areas where the LULC has the least altered. In this analysis, dense forest and water bodies are reduced over time, which has caused a change in channel behaviour (Fig. 8). The built-up area and agricultural land have been eroded and converted into accretion land, which has been used for both built-up and agricultural purposes. But some areas remain as fallow category due to the presence of a high percentage of sand deposition that cannot be used for other purposes. Therefore, agricultural land and open forest areas have erosion on one bank and accretion at the opposite side. Owing to riverbank erosion-accretion, this type of LULC change is very significant for socio-economic aspects. As a result, local people have to migrate to another place to change their livelihood patterns. Later, with the modification of the socio-cultural environment and adjustment of LULC change, a new symbiotic ecosystem was developed in the study area.
In this investigation, predicted bankline and LULC results reveal that dense forest, settlement, agricultural land, and open forest land will face a significant threat due to the future of bankline positions. At present time, the channel migration rate and LULC change rate cannot be accurately predicted because of changing tectonic activity, population growth, and climate change. But it is included that riverbank migration and land utilization ideas will improve the river adjacent mouza area. Thus, it is helped for restoration and management and uplift the socio-economic condition of the riparian peoples in the future. Therefore, the output of the predictions in this work could serve as spatial guidelines for monitoring future trends of channel migration and LULC dynamics and address threats and deterioration of river adjacent village areas. Moreover, the prediction of the river bankline position and LULC pattern is meaningfully possible by applying the CA-Markov and DSAS models.

Reasons for channel shifting and LULC change
We have observed that both natural and anthropological dimensions play an important role in chenging the river shifting behaviour. The channel migration and erosiondeposition of the Kaljani River may be the results of various reasons, i.e., massive floods, flow discharge, nature of bedload, bar formation, change of thalweg between the banks, plate tectonics movement, human intervention, etc.
In the Kaljani river, the gradual formation of mid-channel bars changes the river flow pattern and influences both banks' erosion-deposition and channel migration rate (Maiti 2016). The fluctuations of discharge water in different seasons, specifically in the rainy season are the origin of the massive and frequent floods every year or every two years. These processes can change bank stability and channel morphology (Vanacker et al. 2005;Wohl 2006;Struck et al. 2015;Qazi and Rai 2018). The acquired results have uncovered that, under the extreme rainfall and frequent floods and the geomorphic response to change the rates of erosion-deposition. The average annual rainfall of the study region is 3444.05 mm during the period 1989(District Disaster Management Report 2009. We have studied 20-year rainfall data, which presents extreme rainfall events that lead to high magnitude floods (Fig. 9). Channel migration and erosion-depositions were recorded very high in those years. The floods can trigger sudden changes in the rivers channel. The Kaljani River has been facing floods almost every year or every second year (Starkel et al. 2008 gal, 2007, and annual flood report, Govt. of West Bengal, 2010, 2014, 2016. The magnitudes of the floods are gradually decreasing over time, but some floods are very massive. In 1993, extreme floods caused the highest discharge ever for the Kaljani River, which was recorded at 140,234.53 cusecs (3971 cumecs), (WAPCOS 2003). Thus, it can be said that flood events are one of the most vital factors for change of channel migration and erosion-deposition rates. The fluctuations of the sedimentation rate of the different seasons are the very important factors contributing to morphological changes in the Kaljani River migration and erosion deposition (Mandal et al. 2017). During the monsoon, excessive yearly sedimentation has uplifted the river's channel bed and resistance to the free flow of water (Dey and Mandal 2019). This problem makes the system highly hazardous during the rainy season. This time, high river discharge with massive velocity increases the amount of river water and creates huge pressure on the Kaljani Riverbank. (Bjorklund 2015). The Kaljani River is bearing the imprint of active tectonics of the region as they lie in the zone of the Himalayan Frontal Fault, the most active thrust belt of the Himalayas (Das 2004;Goswami et al. 2012). Tectonic changes are depicted by the responses made in the adjoining morphology of this river behaviour. In our analysis, the total erosion is greater than the deposition, indicating this river still erodes. The hypsometric curve represents the different stages of the evolution of erosional landform (Strahler 1952). The basin age can be analyzed through the hypsometric value. The value of hypsometric integral close to 0 is highly eroded, and 1 is quietly eroded regions (Schumm, 1956;Strahler 1952). We have calculated the hypsometric curve integral. The value (0.13) indicates the Kaljani river basin is now at the old stage. The basin relief and size variations are represented in the long profiles of the river. The distances and elevations were divided by maximum basin relief and the total stream length, respectively, to define the long profile (Lee and Tsai 2010). Thus, the presence of breaks in the river long profile depicts that a strong structural effect is somehow present in the river course. The long profile represents that the maximum portion of the river basin is included in plain land (Fig. 10). The hypsometric curve value and the long profile properties provide adequate evidence of this river's low energy, but the study results differ. This may be due to the energy boost in the river during extreme rainfall and floods. River discharge data can be played a significant role in the concluded the reason for channel migration. But the main limitation of this work, discharge data of this river is not available from any governmental or private authorities. So, we have calculated the SCS-CN. SCS-CN method is determined based on soil group, hydrologic conditions, vegetation types, and agricultural treatment (USDA 1972   the important cause of the Kaljani Riverbank erosion (Fig. 9). In this investigation, it is observed that many mouzas with immense population pressure are vulnerable, where most of the people are engaged in agricultural activities. Therefore, many embankment installations along the Kaljani River for flood control can be modified by the erosionaccretion and adopted LULC patterns. Embankments along the river affect river channel morphology and its flow dynamics. As we know that the man-made embankment increases the stress on the riverbed, so it becomes more vulnerable to erosion than protecting it (Yao et al. 2011). Thus, it is an important factor for the morphologic changes of river floodplains. In our study area, the most active banks have been stabilized, especially along the right side.
The field survey and Copernicus satellite image show the existence of man-made embankment in 27 places at the right bank and 13 places at the left bank, along with 20 bridges (Fig. 11). In the right bank, the erosion rate has been decreased because the continuity of the embankment is extended here, but the left bank experienced more erosion due to shorter length of bank embankments. The riverbank erosion-accretion, channel migration, and LULC pattern change observation are important variables for planners, environmentalists, policymakers for understanding and formulating the needed and appropriate channel design schemes of vulnerable areas of Kaljani River.

Conclusion
This study has demonstrated the application and capability of earth observatory technology that has generated a detailed evaluation of temporal and spatial changes in river channel dynamics and adjustment of LULC of the Kaljani River. The multi-temporal data analysis reveals that the Kaljani River has continuously changed the positions of its bank lines and modified its LULC pattern significantly. In this present river course, the three zones' both banks of zone 'A' (lower part of the river) are severely affected (Jaigir chilakhana, Chilakhana, Bhelakopad witia khanda, Amlaguri, Bhelapeta, Panisala mouzas). This zone has experienced a large amount of erosion-accretion and its impact on LULC has been spotted during the study period. Therefore, the overall result of the predicted bankline has depicted that Ambari and Jaigir Chilakhana mouza are the most dynamic mouzas for both erosion and accretion effects. This study assesses the dynamic change of river bankline positions in vulnerable areas and the endangered condition of the nearby settlements and infrastructures due to high bank erosion. Moreover, the long profile, hypsometric curve value, and the Soil Conservation Service Curve Number (SCS-CN) value have been a significant help in understanding and identifying of consequence reasons. Therefore, in the present research, the DSAS and CA-Markov based automated approach is employed as an alternative way that successfully and accurately measures and has predicted the geomorphic processes (erosion-accretion and LULC patterns) at an appropriate spatio-temporal scale. The level of accuracy is validated by the actual bankline positions (2020) with predicted bankline (2020) and actual LULC (2020) to predicted LULC (2020) empirically. Also, RMSE, and Students t-test (for riverbank migration) and Chi-Square Test, kappa coefficient (for LULC Maps) are adopted to validate of this work. The accuracy level of this study, the Kappa statistics for the LULC (2020) map is 87.57per cent and bankline shifting RMSE varies from 0.007 to 0.176 m. Moreover, it will benefit for engineers, planners and administrators to take the required management or river restoration plans.