Satellites Reveal a River on the Move: Three Decades of Erosion and Accretion Along the Jamuna

Along the banks of the Jamuna River in Bangladesh, the ground itself is in constant motion. A new study published in Discover Geoscience has quantified, with unprecedented detail, just how restless one of South Asia’s most dynamic braided rivers really is. Drawing on thirty years of Landsat satellite imagery from 1995 to 2024, combined with bathymetric cross-sections and interviews with riverside residents, researchers from the University of Dhaka found that roughly 55,436 hectares of land were eroded along the Jamuna’s banks near Islampur and Dewanganj in Jamalpur district, while about 53,496 hectares of new land accreted elsewhere along the channel. The near-perfect balance between loss and gain is not a sign of stability. Instead, it reveals a river that continuously tears down one bank while building another, migrating steadily eastward at an average rate of 60 to 100 meters per year and reshaping the lives of everyone in its path.

The Jamuna, a major distributary of the Brahmaputra, is one of the most morphologically dynamic braided rivers in the world. Its waters carry an enormous sediment load delivered from the Himalayas, and its mean annual discharge of approximately 20,541 cubic meters per second, recorded by the Bangladesh Water Development Board, swings dramatically between monsoon floods and dry-season lows. This combination of high sediment supply and seasonal flow variability drives rapid channel adjustments: sandbars form and vanish, the thalweg, the deepest line of the riverbed, wanders across the channel, and entire banklines retreat or advance within a single year. For communities in Jamalpur district, this variability is a chronic socio-economic hazard, with erosion hotspots repeatedly consuming farmland, homes, and rural infrastructure.

To capture these changes systematically, the research team assembled a methodological framework that blends satellite remote sensing, morphological measurements, and field validation. Multi-temporal imagery from Landsat 5, covering 1994 to 2012, and Landsat 8, covering 2013 to 2024, was processed through Google Earth Engine, a cloud-based platform that makes decadal-scale analysis of large regions computationally feasible. The team calculated the Normalized Difference Water Index, or NDWI, for each image, using the ratio of the green and near-infrared bands to classify pixels as water or non-water and thereby extract the river’s bankline. Although NDWI is a widely used and practical approach for large-scale river analysis, the researchers acknowledged its limitations, including mixed pixels and shallow-water effects, and selected thresholds through careful visual comparison with the original imagery.

Once banklines were extracted, the team imported them into ArcGIS using a projected coordinate system and computed annual areas of erosion, accretion, and unchanged land by comparing overlapping regions between consecutive years. The year-by-year results show a river in near-constant flux. Between 1995 and 2005, the channel was highly unstable, with severe erosion episodes in 1995, 1998, and 2000 accompanied by substantial accretion zones, indicating simultaneous bank retreat and land formation elsewhere. From 2006 onward, stable bank segments increased considerably, but erosion surged again between 2014 and 2021, a renewal the authors attribute to a combination of natural hydrological variability and human pressures such as unregulated sand extraction and floodplain deforestation. Across the full study period, mean annual erosion was 1,911.61 hectares, while mean annual accretion reached 1,844.71 hectares.

What sets this study apart from most satellite-based erosion assessments is its second dimension: the vertical. Planform imagery alone cannot reveal whether a riverbed is silting up or scouring down, yet that vertical balance determines navigable depth. To capture it, the researchers analyzed bathymetric cross-sections from four permanent monitoring stations operated by the Bangladesh Water Development Board, designated RMJ12.1, RMJ13, RMJ13.1, and RMJ14, all located within identified erosion and accretion hotspots. Bed elevations measured at 20-meter horizontal intervals were integrated using the trapezoidal rule, a numerical method that estimates the area beneath irregular curves, allowing the team to compute cross-sectional area and average depth for each survey and track how they changed over time.

The cross-sectional profiles revealed striking spatial and temporal variability in riverbed elevation. At station RMJ12.1, pronounced deepening concentrated in the mid-channel zone, with the deepest scour recorded in 2005 followed by partial recovery in later years. RMJ13 displayed the largest vertical oscillations, with localized deep scours whose positions shifted over time, a signature of frequent thalweg migration. RMJ13.1 showed alternating zones of bed lowering and sediment building, while RMJ14 exhibited significant vertical irregularity with periodic deep scouring interspersed with stable intervals. Linear regression of mean bed elevation across all four stations indicated a consistent downward trend, pointing to progressive channel degradation over the long term, punctuated by short-term disturbances such as localized scouring events.

Satellite and bathymetric findings were then tested against the ground. The researchers conducted key informant interviews with more than fifty residents, boat operators, and local officials, and carried out systematic transect walks along erosion-prone stretches, photographing near-vertical scarps, exposed soil layers, bank collapse, and failed protective works. These observations corroborated the satellite-derived maps of erosion-dominated zones. They also exposed a paradox that matters enormously for the region’s economy: while the banks keep eroding, recent years have seen a relative dominance of accretion, with sediment filling the channel, raising the bed, and shrinking navigable depth, especially during the dry season when discharge is lowest.

The navigational consequences are severe. According to seventy-five percent of community members interviewed, shallow river conditions force cargo vessels to lighten their loads or take longer alternative routes, driving transport costs up by an estimated 40 to 60 percent during the dry season. Journey times have lengthened by 60 to 70 percent in many cases as boats slow to avoid grounding or detour through deeper channels. Dredging demands have grown correspondingly, adding further expense. For a region that depends on the Jamuna to move bulk goods and agricultural produce, the creeping sedimentation represents a slow-motion economic squeeze that compounds the more visible drama of bank collapse and displacement.

The study’s findings align with earlier research on the Jamuna, which documented net accretion of roughly 125 square kilometers along the reach between Bahadurabad and Sarisabari over four decades, and erosion of about 832 square kilometers along the eastern bank, confirming the river’s persistent eastward migration. The authors argue that their integrated framework, combining multi-temporal satellite imagery, cross-sectional morphological data, and community-based validation, offers a robust template that could be extended to other sediment-rich braided rivers in Bangladesh, including the Teesta, Dharla, and Lower Padma. They also note the limitations of their approach: medium-resolution Landsat imagery introduces positional uncertainty in bankline delineation, gaps in bathymetric records make it difficult to tie specific changes to individual floods, and the small number of monitoring stations may not capture the full spatial heterogeneity of the channel.

Looking forward, the researchers propose augmenting the framework with high-resolution data from Sentinel-2 or drone-based surveys, hydrodynamic and sediment transport modeling, and machine learning approaches capable of predicting future erosion-accretion patterns and flagging vulnerable reaches before they fail. In a delta nation where tens of millions of people live on land built and rebuilt by rivers, the ability to anticipate where the water will go next is not an academic luxury. This study demonstrates that three decades of freely available satellite data, married to careful fieldwork, can turn one of the world’s most restless rivers from an unpredictable threat into a system that can be monitored, understood, and planned for, one bankline at a time.

Subject of Research: Spatiotemporal riverbank erosion, accretion, and navigability change along the braided Jamuna River in Bangladesh

Article Title: Assessment of spatiotemporal erosion and accretion patterns in the Jamuna riverbanks using satellite and morphological data

Article References: Haque, M. Z., Niha, F. T., Rahman, M. M., & Zaman, M. A. (2026). Assessment of spatiotemporal erosion and accretion patterns in the Jamuna riverbanks using satellite and morphological data. Discover Geoscience, 4(1), Article 385. https://doi.org/10.1007/s44288-026-00760-7

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00760-7

Keywords: Jamuna River, riverbank erosion, accretion, braided river, remote sensing, Landsat, Google Earth Engine, NDWI, bathymetry, navigability, Bangladesh, geomorphology