Geo-spatial analyses of meandering rivers, assessing past and future impacts on bank landforms and LULC changes

Meandering rivers have a winding, curved path, unlike straight or braided rivers. These rivers form as the watercourse continuously erodes outer banks and deposits sediment on inner banks, creating a series of bends or meanders. This meandering process is a key aspect of river morphology, which studies the shapes and structures of river channels. The constant reshaping of meandering rivers exemplifies the dynamic interactions between water flow, sediment transport, and the surrounding landscape in fluvial systems. River morphology changes on a spatial and temporal scale as a result of a river's frequent meandering and varied courses in the alluvial stretch. The term ‘river morphology’ is used to describe both the cross-sectional shape of a river channel and the process by which a river's section changes as a result of sedimentation and erosion from the banks and beds. The alluvial river basin often undergoes bank erosion (Ercan & Younis, 2009). The floodplain is one of the most easily accessible land resources with benefits for human culture, but due to significant land-use changes and riverbank erosion, floodplains are deteriorating globally (Hazarika et al., 2015; Debnath et al., 2017). The two main elements that could influence the hydrological process of catchments are changes in land use and climate (Shahid et al., 2018), where land use land cover (LULC) analysis is the common terminology to represent the Earth's surface cover due to natural and man-made activities (Anandkumar et al., 2019). Over the last few decades, various researchers have investigated the reasons and processes behind the changing nature of morphological characters. The LULC on the bank adjacent to the urban body is changed by the channel erosion–accretion process, putting the inhabitants at risk (Thakur et al., 2012; Hasanuzzaman et al., 2022). Regional soil erosion rates and sediment transport to China's Lushi River Basin were greatly affected by minor land-use changes (Wang et al., 2012). The lives of people living in floodplains were directly affected by the river dynamics in the upper Brahmaputra plains, which in turn affected land utilization (Hazarika et al., 2015). The effects of deforestation include flooding and river meandering (Barasa & Pereram, 2018; Adnan et al., 2019). The stream erosion–accretion changes the lower urban limit of the valley region next to the bank, putting residents at risk (Thakur et al., 2012). Urbanization was therefore one of the main contributing factors related to changes in the morphology of rivers, as evidenced by literature. However, many studies have employed remote sensing and GIS approaches for some important rivers in order to detect spatiotemporal variations in river erosion (Kotoky et al., 2005; Archana et al., 2012; Sinha & Ghosh, 2012) and their effect on river morphology. River morphological changes can be reliably and efficiently mapped and investigated using an RS and GIS platform with field verification (SAC & Brahmaputra Board, 1996; Bhakal et al., 2005; Kotoky et al., 2005; Gogoi & Goswami, 2014; Langat et al., 2019; Verma et al., 2021). Over two million people were displaced from their homes in Bangladesh between 1970 and 2000 due to bank erosion along the Padma and Jamuna Rivers (Islam & Rashid, 2011). There has been a significant morphological shift in the Ganga River near Allahabad, India (Pati et al., 2008). Per the National Disaster Management Authority of India (NDMA, 2014), Bihar and Assam are two of the states in India that have experienced substantial flooding and erosion. A study conducted in 1996 by the Space Application Centre (SAC), Ahmedabad, India, and the Brahmaputra Board, India, aimed to define and demarcate the island portions that had altered along the bank line as a result of the river's dynamic behaviour, as well as to assess the effects of river erosion on Majuli Island.

However, the complexity of channel migration prevents accurate prediction of changes in a particular meander (Hooke, 2007). Thus, the United States Geological Survey (USGS)'s digital shoreline analysis system (DSAS) is a highly acceptable method for measuring and predicting right and left bank line positions (Ashraf & Shakir, 2018; Jana, 2021). Generally, the DSAS model is applied to model seashores. The DSAS derivation of the past rate of change trends as an indicator of future trends has been used to anticipate patterns of shoreline behaviour, presuming consistency in the physical, natural, or anthropogenic factors driving the historical change observed at the location (Oyedotun, 2014). The purpose of this study is to use DSAS to quantify the rate of changes in river migration that happened along the meandering river bank and their effects on river morphology.

Conversely, the effects of LULC transitions and predicting river morphology, particularly for meandering rivers, are still largely unexplored in the past literature. LULC fluctuations through time and space can be caused by either direct or indirect causes. The relationship between LULC changes (Ahmed, 2012; Maviza & Ahmed, 2020), riverbank erosion–accretion, and LULC change as a result of channel morphoposition changes has been established using modern geospatial tools and techniques (Debnath et al., 2017; Jana, 2021), and their future prediction has been established using satellite images (Mansour et al., 2019; Nurwanda & Honjo, 2019). Even so, much study has gone into finding the most accurate machine learning classifier for future LULC mapping (Camargo et al., 2019; Jamali, 2019). Models of LULC include GEOMOD, CA–Markov, Markov chain, and others. The CA–Markov model correctly predicts and identifies LULC changes, according to Xiao et al. (2012) and Du et al. (2020).

Conversely, Barak River channel migration and changes in land usage and land cover are frequently noted (Nath & Ghosh, 2022c). The discharge and sediment carrying capacity and morphological vulnerability of the Barak River have recently increased (Annayat & Sil, 2020a; Nath & Ghosh, 2022a). In addition, significant bank shifting in the Barak River was investigated (Das, 2012; Nath & Ghosh, 2022a, b, c), with the study indicating that it will have serious effects on the economy and people's standard of living soon (Annayat & Sil, 2020b). Increased urbanization in recent decades has also been a source of concern for the valley (Nath & Ghosh, 2022b). According to the previous literature (Debnath et al., 2017; Jana, 2021; Nath & Ghosh, 2022c), land use, land cover, and river morphology are closely linked.

River morphology is also significantly affected by frequent flooding. There have been five floods exceeding the danger levels in Assam's Barak River between 1976 and 1986. The basin has seen numerous floods since 2016, especially in Silchar City. The 2004 and 2012 floods caused widespread damage and economic hardships, highlighting the vulnerability of flood management infrastructure. A fundamental geomorphological phenomenon known as ‘oxbow formation’ is linked to the recurrent flooding that causes a river's path to shift dynamically, especially in meandering rivers such as the Barak. The river's innate propensity to erode its banks and deposit sediments causes the formation of oxbow lakes and causes major morphological changes over time. Numerous oxbow formations, including Satkarakandi, Sunarigram, Baskandi, Chanpur, Sibnarayanpur, Puranabazar, Fulbari, Rupairbali, Dungripar, Salchapra, Salghat, Chiribazar, and Ratanpur, were noted along the Barak flood plain (Gupta & Devi, 2014). One of the most important factors in the formation of oxbow lakes is soil erosion, which frequently results in high morphological changes (Ijafiya & Yonnana, 2018). Between 1918 and 2003, nine significant oxbows were detected (Laskar & Phukon, 2012). The most recent observation of an oxbow took place near Sonaimukh between 1990 and 2000. As a result, following that phase of oxbow development, a high susceptibility was noted downstream. In order to limit the impact of future oxbow development and to prevent river movement, standard bank protection techniques are employed. All through the Barak River region, the government has implemented a number of protection measures, such as riprap, porcupine, and Geo-tube, to curb the activities associated with the shifting behaviour. It was observed that in those protected but susceptible areas, river migration occurred considerably less rapidly. But according to Choudhury et al., (2014), just 57% of the 3.50 lakh hectares of flood-prone land in the valley have access to any kind of flood damage protection. The valley includes 26 major sluice gates and 738 km of embankments along the main river and its tributaries to mitigate the effects of flooding (Choudhury et al., 2014). However, because the majority of these embankments are older than they were intended to be, major failures frequently occur during monsoon seasons, causing substantial flood damage. Many parts of the floodplain are at a lower height than the tops of the nearby riverbanks, which were traditionally produced by local silt.

River morphology is directly influenced by river migration activity. Long-term bank line migration analysis over the seashore model was identified by the DSAS model. However, it is significantly interesting for rivers, particularly with regard to the issue of meandering rivers. This kind of extremely meandering river requires a long-term river migration prediction, for which morphological assessment of the past and future can be used to identify the areas. Moreover, although the literature suggests that there are many links between land use, land cover, and river shape, LULC studies are crucial for managing water and land resources, especially in regions such as Barak that have a lot of meandering and are susceptible to flooding. Even though the Barak River channel has obviously changed, no study has been done to identify the cause and nature of these changes or to forecast long-term changes in the area's river morphology and land usage.

While each river section's shifting rate was used to evolve the rate of shifting using the Linear Regression Rate (LRR) and End Point Rate (EPR) models, the Markov Chain was utilized to investigate and anticipate changes in LULC across time and forecast future changes. The LRR and EPR models were utilized to forecast the future location of the riverbank line by considering the historical rate of river movement. Therefore, CA–Markov and DSAS models were used to forecast land-use changes and their consequences on river morphology have been investigated in the present research. Given the importance of the Barak River and its catchment region, the present study is necessary to understand better its dynamics and the true meandering characteristics of the meandering river and provide a new concept for futuristic modelling for meandering river management.

The USGS contributed to the multispectral remote sensing and Landsat data used in this study. Landsat images were taken every 10 years from 1990 to 2020 (Table 1).

For a better understanding of the vulnerability of the study area, four sections were considered during the study from Section 1 in the upstream portion to Section 4 in the downstream portion of the study area. For the river, morphological analysis of layer stacking was done for Landsat Images to digitize the river path over time in ArcGIS. The shifting of river bank lines along the river was traced using the DSAS tool in ArcMap. In the DSAS model, the shifting rate of each section of the river was used to evolve the rate of shifting using the LRR and EPR models. The future position of the riverbank line was predicted by considering the past shifting rate of the river using the LRR and EPR models. Validation of the forecasting model was done by considering the error between the observed position and the predicted position of the river.

The present study employed the ArcGIS extension tool DSAS to estimate the rate of bank line erosion and accretion using reference retrieved baselines and auto-generated transects. For the DSAS output, two additional models were used: EPR for current bank line erosion–accretion and LRR for future bank line estimates.

The DSAS model was utilized for calculating future riverbank erosion–accretion, shifting, and bank line location. However, before making future predictions, the model must be tested using current data (Mukhopadhyay et al., 2012; Jana, 2021). The least-squares method (fitting a regression line) was used to evaluate the outcome of this attempt to predict channel shifting and bank line position (Thieler et al., 2009). To estimate riverbank shifting rates, the LRR and EPR models were used. In the case of the LRR model, baselines are defined for the temporal spans of bank shifting as (1990–2000), (1990–2010), and (1990–2020), respectively, using the channel midpoint position of the previous year within each span.

Based on the transition between 1990 and 2010, the model predicted the river position for the year 2020. The model performance has been checked using performance indices such as MAPE and t-test performance. The model was further used for long-term prediction (2033 and 2043).

A 3-km flood plain buffer zone from both banks was taken for the analysis shown in Figure 1(b). A supervised land-use map was generated over the study area for calculating changes in transformation over the flood plain. A land-use map was generated using support vector machine (SVM) supervised classification. After that, the prediction process was followed.

Changes in the LULC classes in the catchment were predicted for short-term (2022) and long-term (2033, 2043) scales based on the transitions that had occurred in the basin during 2005–2017. The LCM generates the future temporal–spatial allocation of LULC classes in the catchment area in three basic stages. In the first stage, the historical multitemporal LULC maps (t₁ and t₂) of the study area were used to analyse the previous LULC change pattern. In this stage, the major transition of each LULC class to other classes is considered. Then, the LULC transition potentials are modelled from the analysed land-use/cover change pattern.

A is the Markov transition matrix use as LULC transition between two periods (i, j). Each line factor was required to be 0–1, and each rate is a non-negative quantity in the transition matrix. The relative frequency of transitions recognized throughout the entire period is the estimate of the Markov chain (2005–2017), and the outcome of the estimation can be used for prediction (Mondal et al., 2016). Three future years were predicted in this study, 2022, 2033, and 2043.

The LULC transition potential indicates the chance of a certain LULC class switching to another LULC class. The vulnerable portion identified by the land-change modeller (LCM) was further analysed regarding river morphological changes. By including specific static and dynamic driving variables, such as slope, DEM, distance from the road, and LULC disturbance map, the LCM was created. The transition potential was modelled by incorporating certain static and dynamic driver variables such as slope, aspect, and distance from urban areas. The significance of the variable potential test was done by Cramer V and p-test. The LULC model calibrates data from 2005 to 2017 to anticipate future land-use changes, and it validates with LULC observed by supervised classification. Three future years were predicted in this study, 2022, 2033, and 2043.

A future morphological assessment based on predicted river position and LULC cover was analysed simultaneously, which will aid in an early warning for land and river management in the study region.

The study area was divided into four sections (Figure 1). In those sections, various transects (depending on topography) were created to assess riverbank migration/erosion–accretion rate. To understand river migration over the study area, a location-wise analysis was performed.

The location-wise river migration calculation for Section 1 is summarized in Table 2.

The geological investigation of the study area indicated that Section 2 is one of the vulnerable sections within the Barak floodplain. The location-wise river migration calculation for Section 2 is summarized in Table 3.

Only 22.77 m of average shifting and 42.54 m maximum shifting were observed for Arunachal. But a rise in vulnerability for both the directions of Jhatingamukh has had an impact during the last decade. An average BCI of 69.20 m and a maximum BCI of 196.52 m were found for this location. The same value for NBM and BCI within the upper, middle, and lower sections (transit numbers 23–34, 44–85, 113–151, 171–235) of the location has been found. EPR, NBM, and BCI indicate recent vulnerability for those portions within the location.

Section 4, from Ghagrapar to Kalinagar, was the most unstable in the study area, according to the migratory studies. The study area's most unstable locations were Salchapra Pt3, Sripur Pt1, Polarpar, and Kalinagar. The outcome of migration analysis over Section 4 is provided in Table 5.

The most important road connecting Silchar NH 37 is situated very near the location (Pachgram). Due to high morphological changes till 2010, especially between 2000 and 2010, some vulnerable location was protected by the protection structure. The protection structure performs effectively towards limiting the morphological vulnerability. The vulnerability of the left side significantly reduced during the last decade. NBM and BCI indicate the same value for some upper and lower portions (transit numbers 1–6, 47–66, 70–71) within the location. Therefore, those locations were recently affected due to the shifting of the river.

The overall error and sectional MAPE show good agreement for all sections according to the model's validation. The p-performances across all locations were accepted (P < T_c (5%)). After successfully verifying the output of the prediction model, predictions for the following two decades have now been produced (2033, 2043).

The output of the prediction model for the sections is shown in Figure 7.

For Section 2, the upcoming shifting for Govindpur will be 89.7 m for up to the next decade and the average shifting will increase up to 195.85 m up to 2043. However, the most significant shifting will appear for Sonabarighat Pt 2. The land will shift 197.98 m by 2043 and 90.58 m over the next decade (2033). Kasipur Grant will be the most stable and Govindpur, Sonabarighat Pt2, and Sonabarighat Pt3 remain vulnerable in the upcoming decades (Figure 8(b)).

For Section 3, the future prediction of river position is shown in Figure 7, and the outcome of the prediction analysis is shown in Figure 8(c). Per future predictions for Section 3, the maximum upstream of the section will remain stable for the next two decades. Arunachal will be the most stable location within the section, with only 12.92 m of shifting predicted to affect the location by 2033 and 22.17 m on average shifting by 2043. Migration will be more affected towards the downstream portion of the section. The maximum migration over the section was found at Ranghat. The average shifting for the next decade will be 56.31 m over the location and it could increase by up to 96.8 m on average by 2043. The overall migration over the location will mostly be stable upstream, while downstream sections especially in Jhatingamukh, Ranaghat, and Algapur will be affected by the presence of some river migration activities.

The predicted river position for Section 4 is shown in Figure 7 and the outcome is notified in Figure 8(d). The most severely affected section for the river migration prediction model within the study area has been found to be Section 4. Per the future prediction of river position, it indicates Sripur 1 will significantly be affected up to 2043. The average rate of migration will be 289.43 over the next decade and 575.01 m up to 2043. Comparing other locations of the study area, the rate of vulnerability for Section 4 will be significantly high for the upcoming decade. Especially for Sripur 1, Polarpar and Kalinagar will be significantly affected by the position changes of the Barak River over the locations.

For LULC forecasting analysis, three land-use maps were generated for the years 2005, 2017, and 2022 using SVM techniques.

Urbanization has influenced the land significantly. More than 66% more settlements were observed within the investigation period. Agriculture land was raised from 90.18 to 136.34 km². More than 51% of agricultural land growth was observed during the investigation. The area covered under forest land has changed from 151.16 to 144.21 km². Almost 4.6% of land suffers due to deforestation during the period. A significant decremented trend for bare land areas was observed. More than 6% of water land has changed to other classes within the study period.

Simultaneously, a land-use map was prepared for the year 2022. The outcome of the land-use analysis for 2022 is presented in Table 8. The change between 2017 and 2022 has also been investigated. The land-use map used for the study is shown in Figure 9.

Between 2017 and 2022, similarly to earlier LULC calculations (2005–2017), the rate of growth of the urbanized area is very significant. Almost 21.28% more land was urbanized during this period. Along with settlement increment, the growth of agricultural land also increased during this phase of the study. More than 6% increment was observed for agricultural land. Deforestation has affected almost 5.57% of the land. The high rate of urbanization is a major concern for the study area.

In this study, the accuracy of LULC maps with randomly selected points was assessed using the kappa coefficient technique. The points were selected to reflect each of the LULC classes throughout the study area.

For evaluation, the overall accuracy, kappa statistics, producers, and user accuracy were used. For the years 2022, 2017, and 2005, the overall accuracy of classified images show 90.6%, 89.33%, 86.66% and the kappa statistics 0.88, 0.86, and 0.83, respectively. The accuracy matrix is presented in Table 9.

The influencing driver factors are based on spatial analysis and can be added to the model as static or dynamic components. Slope, DEM, distance from the river, distance from the road, and evidence likelihood map variables were included in the model based on Cramer's V. Cramer's V value was used to identify the significance of a particular variable over the changes. Table 10 shows Cramer's V and p-value for each explanatory variable.

The slope, DEM, and evidence likelihood maps had a V value of 0.008, 0.3, and 0.55, respectively. The distance from the river and the urban area had a V value of 0.19 and 0.17, respectively. The effect of slope on LULC change in the study area is not critical, as evidenced by low Cramer's V values for variables such as slopes. To simulate the transitions in this study, all driver variables were utilized.

As can be noted, the degree of freedom (df) = 4, significance level (α), and critical value of chi-square (χ²) (χdf, α) from table = 9.48 (for significance level 5%). As a result, Table 5 presents the χ² test values, which reflect the CA–Markov model's validation and acceptance for LULC map prediction. The actual and predicted LULC map was visualized. The χ² test was validated with the χ² tabulated value for checking the significance level. The predicting model performance is very good as the χ² value suggests.

Water bodies, forestland, and bare land will gradually decrease, while agricultural land and settlement area will gradually increase. Settlement increment was one of the most dominating changes over the land-use changes analyses indicated. In comparison with 2022, a land-use area of more than 31% of the settlement will affect the land by 2043. The rise of agricultural activities was also significant over the past. More than 20% of the land will convert to agriculture class within the next two decades. The loss of forest land will be constant over the next two decades as well. Deforestation will occur over almost 7.5% of the land by 2043. The bare land class will also follow a similar trend. Almost 26% more bare land will change into other classes up to 2043. A constant minor reduction of water bodies will also affect the land over the next two decades.

In geological cause investigation, it has shown that the growth of urbanization along with agricultural practices and deforestation is critical from the perspective of river stability. The growth rate of those vulnerable land-use classes will significantly affect the land per the prediction model outcome indicated. Therefore, it will be a serious concern for the perspective of land and river stability. It will cause more morphological changes in the Barak River.

This study has demonstrated the applicability and capabilities of Earth observatory technology by providing a detailed evaluation of temporal and spatial changes in river channel dynamics, as well as LULC changes. The Barak River's bank lines have continuously changed positions, and its LULC pattern has greatly changed, according to the multitemporal data analysis. This study investigated the dynamic changes in river bank line positions in sensitive locations, as well as the hazardous state of nearby settlements and infrastructures as a result of excessive bank erosion. There were four sections to the study area. The river migration study shows that the river migrates slowly between Sections 1 and 3, with the effect of the LULC influencing parameter being relatively low. These two sections were more stable due to less urbanization and agricultural practices. Sections 4 and 2 were recognized as the most vulnerable sections. For Section 2, urban growth has affected the riverbanks, leading to increased agricultural activity. One of the key factors for a high vulnerability rate in Section 4 is a lack of forestation and increased urbanization. Conversely, the DSAS and CA–Markov-based automated methodology is used in this study as an alternate method that successfully and reliably assesses and predicts geomorphic processes (river migration and LULC patterns) at an acceptable spatiotemporal scale. Actual bank line positions (2020) versus expected bank line (2020) and actual LULC (2022) versus predicted LULC (2022) demonstrate the level of accuracy. This study was validated using RMSE, Student's t-test (riverbank migration), χ² test, and kappa coefficient (LULC Maps). LULC forecasting for the study area reveals rapid urbanization up to 2043. Settlement increment was one of the most dominating changes over the land-use changes analysis indicates. In comparison with 2022, a land-use area of more than 31% of the settlement will affect the land up to 2043. The rise of agricultural activities is also predicted to be significant compared to the past. More than 20% of the land will be converted to agriculture class within the next two decades. The loss of forest land will be constant over the next two decades as well. Deforestation will appear in almost 7.5% of the land within 2043. The bare land class will also follow a similar trend. Almost 26% more bare land will change into other classes by 2043. A constant minor reduction of water areas will also affect the land over the next two decades. Localized bank protection cannot provide a long-term solution to the problem of executing and maintaining protection activities as erosion becomes more complex. With the right conservation measures and land-use management, this sort of highly migratory river might be protected. As a result, this study will assist in assessing the long-term influence of LULC on river migration and the morphology of meandering rivers. It explains the importance of developing and implementing sustainable watershed development policies to maintain the urbanization effect on the vulnerable catchment area. Proper urban planning is crucial for a city's future expansion in addition to limiting the future impact of river morphology in order to safeguard highly vulnerable areas from urban growth and river vulnerability, as predicted by this study. The results of this study will assist the communities that manage rivers and land in determining the factors that contribute to morphological vulnerability and in developing strategies for preserving the right conditions that will prevent changes in the morphology in the future.

The overall efficacy of these measures is often hampered by issues such as inadequate funding, poor maintenance, insufficient community involvement, and lack of enforcement. Therefore, it is essential to establish an increasing amount of protection measures, particularly in the regions highlighted by this research, in order to reduce the chance of future high vulnerability.

Moreover, incorporating geospatial data and the analytical findings from the study into river management practices offers a powerful toolset for anticipating, mitigating, and managing the impacts of river migration. These tools enable a proactive approach, reducing the vulnerability of both communities and ecosystems to the dynamic changes in river systems such as the Barak River. The study's use of geospatial data can significantly improve river migration management, including predictive modelling, early warning systems, informed land-use planning, targeted mitigation strategies, monitoring, and increased community engagement.

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

All authors contributed to the study conception and design. Material preparation, data collection, and analyses were performed by A.N. and S.G. The first draft of the manuscript was written by A.N. and the final manuscript was checked by S.G. All authors read and approved the final paper.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.