Quantifying channel planform and physical habitat dynamics on a large braided river using satellite data - The Brahmaputra, India

Abstract

This paper presents the results of a trialling of the use of Indian Remote Sensing (IRS) satellite imagery for mapping the geomorphology and physical habitat of the Brahmaputra River in Assam, India. The study was undertaken on a 110 km reach as the river emerges out of the Himalayas; a reach with a complex braid pattern and high levels of channel instability. Image analysis on four sets of IRS data encompassing a period of 13 years (1990-2002) was undertaken to detect basic changes in the extent and pattern of low flow channels, vegetated islands, exposed sand bars and floodplain vegetation. Simple unsupervised classification techniques were applied to the images. The water surface of the low flow braided channel network, however was mapped more accurately using the GROW facility within ENVI 4.1; a technique based upon mapping areas within a given number of standard deviations in terms of pixel values within a training area. Subsequent to this unsupervised classification of the water surface area allowed classification of the water in to three water types based on depth and separation of isolated pools and backwaters from the channels conveying flow. Overall classification accuracies of 82.5% were achieved in relation to mapping physical habitat; varying between 86 and 95% for the four different dates. Analysis of the nature of channel planform and habitat change showed that there has been a 3.7% increase in the active channel area consisting of exposed sediment and water within the river corridor. This substantiated a general awareness of an increase in the width and reduction in depth of the Brahmaputra River over the last few decades. The analysis also depicts the occurrence of a major avulsion and the highly dynamic nature of the braided channel network and sand bars.Overall the study demonstrated that simple classification methods when applied to satellite data, capturing imagery in the visible and near infra-red, can be used to measure important changes in the geomorphology and physical habitats that make up the fluvial system of the Brahmaputra River. The approach may also be more widely applicable to large river systems comprising a mosaic of water, exposed sediment and spectrally different vegetation communities. Copyright ¸ 2008 John Wiley & Sons, Ltd.