STUDY ON THE RIVER BANK EROSION FOR THE ALLUVIAL PLAINS OF THE BARAK RIVER, NORTHEAST INDIA
Loading...
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Department of Civil, NITS
Abstract
Riverbank erosion is a significant and distinct problem in the floodplains of alluvial rivers
in North East India. The Barak River has experienced an alarming increase in bank erosion
rate over the last few decades, resulting in embankment breaches and biodiversity loss, but
there is a dearth of field studies to evaluate riverbank fluvial erosion. This study aims to
assess the river bank erosion and Vulnerability Assessment of the Barak River in India using
an in-situ submerged Jet Erosion Test (JET). The riverbank erosion was estimated for a span
of the riverbank on one side of the stream using the excess shear stress equation and
impinging jet theory. A total of Forty in-situ submerged JET tests were performed at ten
locations to determine the soil erodibility parameters (i.e., critical shear stress and erodibility
coefficient) of the composite riverbanks along the river. The upper reach of the Barak River
up to Badarpur Ghat in Assam was considered for the study area. The results show that the
spatial variation of the erodibility parameters in the river banks varies significantly, with the
magnitude of critical shear stress 𝜏𝑐 and erodibility coefficient 𝑘𝑑 of the composite river
bank soils ranging from 0.1 to 55 Pa and 0.1 to 10 cm3/N-s, respectively. The analysis of
erosion-vulnerable sites based on the JET test showed that the soil could be classified as very
erodible to moderately erodible, which is highly vulnerable to erosion. Annual bank erosion
was computed using measured erodibility parameters and stage record data, which shows
the bank erosion of the Barak River will occur in many places, particularly at the critical
area. The measured bank erosion was compared with the observed satellite imagery map for
the 2000-2020. The results of the in-situ jet tests using the excess shear stress approach are
more reliable in predicting erosion rates than the empirical method and are also helpful in
identifying river bank locations that are more vulnerable to erosion.
Estimating the magnitude of bank erosion is essential for planning any bank
protection work. However, conducting comprehensive in-situ surveys to measure bank
erosion rates for a large area is time-consuming, this study integrates machine learning (ML)
techniques to predict spatial riverbank erosion across the entire study area. In-situ JET
measurements were used to determine erosion rates at specific sites, and key influencing
factors such as developed shear stress, bankfull depth, bankfull days, reach slope, discharge,
vegetation cover, soil texture, bank material size, bank slope, and curvature were identified.
The dataset was randomly split for training and testing, and four ML algorithms random
forest (RF), decision trees (DT), multiple linear regression (MLR), and gradient boosting
regressor (GBR) were employed to establish the correlation between bank erosion and the
influencing factors, and the predictive performance of each ML model was analyzed using
statistical indices such as R-squared, NSE, RMSE, and NRMSE. Among the four ML models
tested, the Random Forest (RF) model demonstrated better performance (R-squared: 0.96,
RMSE: 2.34, NRMSE: 0.14, NSE: 0.96) and was chosen as the optimal methodology to
calculate the spatial distribution of river bank erosion throughout the study area. The
developed shear stress and bankfull depth are the most influential variables affecting river
bank erosion. Moreover, discharge, percentage of clay, mean bank material size (D50), reach
slope, and bank angle are other essential variables as well. Findings also show that the areas
with high-developed shear stress and steep slopes have relatively high bank erosion rates.
The optimal (RF) model was applied to estimate the riverbank erosion rate over the entire
study area, and the resultant values were then spatially mapped in GIS to visualize annual
bank erosion distribution. The developed bank erosion map was verified through additional
submerged JET data
This study demonstrates that integrating the submerged JET test and ML techniques
has proven highly effective for the spatial modeling of riverbank erosion, and the proposed
methodology can be implemented elsewhere. However, a primary challenge in
implementing this methodology is identifying and selecting the ideal input variables and
obtaining sufficient observations to adequately explain bank erosion variability that may not
be readily available in all locations. Another advantage lies in the ability to integrate ML
inputs into GIS to map the spatial estimated riverbank erosion rates and serve as a tool for
zoning bank erosion hazards. Therefore, the methodology adopted in this present study can
be used in similar environments to map bank erosion and prioritize bank stabilization
treatments or other preventive measures. By combining field-based measurements with
advanced ML techniques, this research contributes to a deeper understanding of riverbank
erosion dynamics and provides a robust framework for future erosion assessment and
management strategies.