3 Materials and methods
This study uses Landsat imageries (MSS, TM, ETM+, and OLI) of the last four decades (1976-2018), mostly at decadal intervals. The image selection strategy is based mainly on the availability of cloud-free data in the region. Post-monsoon satellite imageries are downloaded (from the USGS website) to avoid cloud-content data. Several Landsat images are required to cover the entire Brahmaputra River, which is sometimes non-feasible due to lack of cloud-free data for a year under evaluation. To overcome such situations, the images of contiguous months of successive years are considered having similar hydrologic conditions (lean phase period). The downloaded imageries are stacked in ERDAS Imagine and finally mosaiced to get a composite image of the entire Brahmaputra River. The composite image is further utilized to digitize the bankline of the Brahmaputra for the study periods.
The ambiguous nature of true bankline identification of the Brahmaputra River is well documented (Hassan et al., 1997; Sarker et al., 2014). Due to complex anabranching, true bankline consideration is always a challenge, where Brahmaputra is concerned. In this study, active anabranches that rejoin the Brahmaputra are considered part of the river, though flanking channels are mostly avoided. Crevasse splay and overbank sediment spills are very common in the Brahmaputra. Such geomorphic traces fade out during the driest periods and are relatively easy to distinguish. Therefore, the satellite images of successive months of the relatively dry period are used to differentiate between such geomorphic impressions and true bankline. To reduce the abrupt over or underestimation of bankline dynamics, the delineated banklines of successive study year images are standardized with each other. Further, the banklines of the Brahmaputra utilized for the study periods are digitized (Fig. 2). To document the dynamics of the Brahmaputra River, the entire digitized banklines in Bangladesh and India is divided into equidistant segments. For this purpose, the downstream point where Brahmaputra was confluenced with Ganga (Padma) in 1976 is taken as reference. The river is divided into ten equal reaches, demarcated as R1, R2… and R10, from reference point to the upstream point of the Brahmaputra, which lies near Laikaghat, and was the area of the concourse of the Dihang, Dibang, and Lohit dating back to 1976 (Fig. 1). Subsequently, the reach-wise comparison is performed to evaluate the land erosion and accretion.
The digitized banklines for study periods are also used to measure the width of the Brahmaputra River. For this purpose, lateral lines at an interval of ~1 km are drawn for the entire stretch of the river. While drawing the cross-section lines, the orientation of the Brahmaputra River is kept in mind. The cross-sections for river width measurement are drawn perpendicular to the direction of Brahmaputra channel.
This work evaluates the major eroding sections of Brahmaputra. Such active sections are identified based on the severity of erosion along the river. The natures of the erosion, i.e., progressive or random, are interpreted from multi-temporal images. Planform characteristics of Brahmaputra are also interpreted from satellite imageries. It includes identifying the dominant channels, braided and nodal belts, and their nature. The Second order channels, which represent the dominant channels, of Brahmaputra are interpreted based on Bristow’s channel hierarchy system (Bristow, 1987; Bristow and Best, 1993). Bristow (1987) described the morphology of Brahmaputra as consisting of first, second, and third order channels. The first order channel comprises the whole river, which may include several second order channels. The second order channel has low stage third order channels within them. This morphological description is used by several researchers working in the Brahmaputra (Thorne et al., 1993; Sarker et al., 2014).
Multiple approaches, including satellite image-based interpretations, fieldwork, and meta-analysis, are performed to evaluate the major drivers of erosion along the active sections of Brahmaputra River. It has inferences from- (a) previous independent research work literature, (b) major seismic activities, (c) floodplain lowlands, and (d) trajectories and sustenance of second order channels near banklines. Shillong and Assam earthquakes of 1897 and 1950 have largely affected the recent channel evolution of Brahmaputra (Oldham, 1899; Goswami, 1985). The possible role of those well-known earthquakes in Brahmaputra River dynamics is discussed. Corona photographs of 1961 are utilized to interpret the effect of the 1950 earthquake along the upper Brahmaputra reaches. It includes identification of sediment lobates ( associated with the earthquake )along the eastern end of Brahmaputra plains. Elevation profiles are drawn using SRTM DEM to identify the lowlands along the floodplains of Brahmaputra. Flow course and the sustenance of the second order channels are identified in ArcGIS to assess their role in erosional activities.