Seven centuries of reconstructed Brahmaputra River discharge demonstrate underestimated high discharge and flood hazard frequency

The Brahmaputra River, a major tributary of the Ganga-Brahmaputra-Meghna river system, plays a crucial role in the lives and livelihoods of over 60 million people in Bangladesh, Northeast India, Bhutan, and Tibet. Its high discharge rates, driven by substantial annual precipitation and snowmelt from the Himalayas and Tibetan Plateau, provide essential benefits such as irrigation, fish stocks, and sediment deposition. However, the Brahmaputra is also a source of devastating floods, particularly during the monsoon season (July-September). These floods have severe consequences, resulting in fatalities, displacement, and food insecurity. While climate change projections suggest an intensified monsoon and increased flood risk due to factors like rising temperatures and glacial melt, the short observational record limits our understanding of the natural climate variability in the basin and makes it difficult to accurately assess the extent of future risk. This paper addresses this gap by developing a long-term tree-ring reconstruction of Brahmaputra discharge to provide a historical perspective on flood hazard in the region and to understand the natural range of variability compared to current and future projections.

Previous studies examining flood hazards in the Brahmaputra basin have been hampered by the relatively short and incomplete instrumental discharge records, typically spanning only a few decades. This limited data makes it challenging to assess the magnitude of future changes in the context of natural variability, especially at decadal and centennial timescales. Tree-ring reconstructions of hydroclimate offer a valuable tool to extend these short records and provide insights into past droughts and floods, allowing for a better understanding of the natural range of variability and the severity of recent climate extremes. Several studies have successfully used tree-ring data to reconstruct hydroclimate variables in other regions, offering a precedent for this approach in the Brahmaputra basin.

To reconstruct monsoon season (July-August-September, JAS) Brahmaputra River discharge, the researchers employed a Bayesian Regression model using 28 annually dated tree-ring series from locations proximate to the Brahmaputra watershed. These tree-ring series were selected based on their significant correlation (p < 0.1) with mean JAS streamflow at the Bahadurabad gauging station in Bangladesh. The model incorporates a nested approach, sequentially dropping shorter tree-ring series to develop multiple reconstruction ‘nests’, ultimately creating a 696-year reconstruction (1309–2004 CE). Model fidelity was assessed using metrics such as the calibration period coefficient of multiple determination (CRSQ), the validation period square of the Pearson correlation (VRSQ), validation period reduction of error (VRE), and validation period coefficient of efficiency (VCE). The historical flood events before 1956 were collated from existing literature. Superposed Epoch Analysis (SEA) was used to test the association between high discharge and documented flood years. Projections of future discharge were derived from an ensemble of 20 CMIP5 climate models, simulating historical (1850-2005 CE) and RCP8.5 scenarios (2006-2099 CE). Recurrence intervals of high discharge (above the 2007 flood level) were calculated using bootstrapping to compare the frequency of such events across the instrumental record, the reconstruction, and the CMIP5 projections.

The study's key findings include: 1. **Strong correlation between seasonal and sub-seasonal discharge:** A strong positive relationship was found between mean JAS discharge and the maximum 10-day mean discharge, validating the use of JAS discharge as a proxy for flood hazard. 2. **Instrumental period unusually dry:** The early instrumental period (1956-1986) was exceptionally dry compared to the seven-century reconstruction, ranking in the 13th percentile. 3. **Underestimation of flood hazard frequency:** The instrumental record significantly underestimates the recurrence frequency of high discharge events associated with flood hazard (by 24-38% compared to the reconstruction). 4. **Reconstruction reveals past variability:** The reconstruction reveals extended periods of above-normal discharge in the past, with no modern analogues, highlighting the limitations of relying solely on recent observations. 5. **High discharge during historical flood years:** While the reconstruction underestimates the instrumental record's highest discharge (1998), the high discharge in 1998 was still unusually high in the long-term context. 6. **Climate models project increased discharge:** CMIP5 models under the RCP8.5 scenario project a substantial increase in Brahmaputra River discharge by the end of the 21st century. 7. **Climate teleconnections weak:** No significant link between Brahmaputra discharge and ocean SST indices (ENSO, IOD) was found.

The findings demonstrate a substantial underestimation of flood hazard frequency in the Brahmaputra basin when relying on the relatively short instrumental record. The unusually dry conditions of the early instrumental period and the multi-decadal wet periods revealed by the reconstruction highlight the importance of considering long-term natural variability when assessing flood risks. The projected increases in discharge from CMIP5 models, coupled with the wetter conditions shown in the reconstruction, suggest that the future frequency of high discharge events and associated flood hazards could be far greater than currently estimated. This highlights the vulnerability of the region and the need for more robust risk assessments and mitigation strategies.

This study's long-term reconstruction of Brahmaputra River discharge provides crucial context for understanding past and future flood risks. The significant underestimation of high discharge and flood hazard frequency based on recent observations underlines the necessity of incorporating long-term hydroclimatic variability into flood risk assessments. Future research should aim to improve reconstruction skill through incorporation of additional proxies, enhanced sampling, and integration of other flood-related data sets to better capture the complexity of flood events. Continued monitoring, improved real-time discharge data across all basin states and enhancement of adaptive capacities are crucial to reduce the overall flood risk in the region.

The study's limitation is the focus on mean JAS discharge as a proxy for flood hazard. While high discharge is strongly associated with flood events, other factors (rainfall intensity, soil moisture, land use) also contribute to flooding. Furthermore, the reconstruction underestimates the extreme discharge events observed in the instrumental period. The analysis considers the likelihood of high discharge but does not fully address societal exposure and vulnerability.