Potential Impacts of Tsangpo Lake-Burst Megafloods and Their Preservation in the Bengal Basin and Delta System

Abstract:

Large, glacially-dammed lakes formed via the impoundment of the Tsangpo River in Tibet led to lake-burst floods during the late Pleistocene and at least two intervals in the early and late Holocene. We present the first critical examination of the potential effects that the Holocene lake drainages had on the downstream Bengal delta and their preservation in the geologic record. Based on stratigraphic evidence from cores drilled across the delta, digital elevation models, seismic data, and hydraulic flow calculations, we propose that lake-burst floods could be responsible for (a) triggering short-lived avulsion events of the Brahmaputra River into the Sylhet basin, (b) the formation of two apparent overflow channels on the Madhupur Terrace, and (c) the deposition of a large, mass transport deposit in the submarine Swatch of No Ground canyon system. Comparing the early and late Holocene events, we expect the distribution of the floodwaters and their deposits in the two intervals to differ sharply owing to major differences in flood volume and the paleotopography of the delta. Despite much higher discharge, the early Holocene floods were largely accommodated within the vast lowstand valley of the Brahmaputra, with some spillover into the Sylhet basin. In contrast, the late Holocene floods likely spread over a larger area due to the relatively even, low-gradient topography. Offshore, a 40 m thick chaotic, semi-transparent seismic facies observed in the canyon corresponds temporally with the early Holocene floods and is tentatively interpreted as a subaqueous debris flow generated by the flood pulse directed to the canyon via the lowstand river valley. We examine the theoretical preservation potential of rare flood events in light of signal shredding mechanisms to help explain why evidence of the larger, early Holocene floods is preserved whereas we have found no clear signal from the younger floods.