Evaluation of the Water Budget Components of the Brahmaputra River Basin Using Satellite Data

Abstract

The water budget can be described as the volume of water that enters a land area, remains stored within it, and eventually exits the land system during a specific time interval. The water budget of a river basin can be represented by equating key components of the hydrological cycle, which include precipitation, actual evapotranspiration (ET), runoff (Q), and changes in terrestrial water storage. The current research is centred on the assessment of the water budget elements within the Brahmaputra river basin by utilizing satellite-derived data. The motivation for this PhD research is grounded in the complex and transboundary nature of the Brahmaputra River basin, which extends through several countries. A key challenge is the scarcity of hydrometeorological data within the basin, making it difficult to conduct comprehensive hydrological studies. To address this data deficiency, the study turns to space-borne data, as it can offer a more complete and cohesive view of the basin's water budget. The satellite precipitation data were evaluated against updated Brahmaputra River basin gauge data. We also assessed different precipitation data to determine the risk of hydrometeorological variables using dependence measures. We further assessed precipitation data for reconstructing significant water budget variables and innovative trend analysis (ITA) of those variables. The study Five daily satellite precipitation products were evaluated against an updated Asian Precipitation Highly-Resolved Observational Data Integration Towards Evaluation of the Extreme Events version 2 (APHRODITE v2) using categorical and continuous metrics. Global precipitation measurement (GPM) resulted in the program known as the Integrated Multi-satellitE Retrievals for GPM (IMERG) was found to be the best-performing product daily, considering the spatial and temporal mean for the whole time series. The Climate Prediction Center (CPC) Morphing technique (CMORPH) was found to be the best- performing product considering the evaluation of metrics on a seasonal basis. The soil moisture to rain (SM2RAIN) of the European Space Agency (ESA) climate changeinitiative (CCI) precipitation product was found to be the least-performing product on all counts. Given precipitation quantity, the conditional bivariate copula concept predicted evapotranspiration, the Gravity Recovery and Climate Experiment terrestrial water storage change (GRACE TWSC), and river discharge. The optimal copula is Frank for all three precipitation-TWSC pairs, the European Centre for Medium-Range Weather and Forecasting (ECMWF) reanalysis ET (ERA5-ET) and ERA5-ET, and Clayton for the remaining pairs. Pearson's linear and Spearman's rank correlations for all the pairs of variables are significant for observed and simulated values. The non-exceedance probability of all the dependent variables (lower percentile) decreases with increased precipitation. However, the exceedance probability of the same variables (upper percentile) increases gradually with increased precipitation. The water budget equation of a large basin based on the conservation of mass was used to reconstruct TWSC, ET, and runoff for the Brahmaputra basin. The reconstructed water budget variables are further assessed using a correlation coefficient to know the linear strength. Also, error metrics like absolute mean error and bias were used to determine how far we can see the variation, such as reconstruction against a gauge or quasi-gauge data. The ERA5-derived TWSCs and Qs tend to provide the highest linear strength expressed in the correlation coefficient on a monthly and seasonal basis. To a greater extent, the Tropical Rainfall Measuring Mission (TRMM), IMERG and the Climate Hazards group Infrared Precipitation with Stations (CHIRPS) also depict a closer correlation coefficient to that of ERA5-derived TWSCs and Qs. The linear strength of derived ETs shows that the inherent uncertainties in the water budget variables did not reconstruct ETs well. On a monthly basis, the TRMM-based TWSC reconstruction was the most optimal, and IMERG-driven ET and runoff were the most optimal. SM2RAIN-driven TWSC, ET, and Q were reported to be the least optimal. For most of the seasons, it was either TRMM or IMERG with the least error. However, the error in terms of the percentage of gauge precipitation for winter and post-monsoon seasons is staggeringly high. Even on a seasonal basis, SM2RAIN was iithe least performing. Overall, TRMM, IMERG, and CHIRPS show much lesser uncertainties than other precipitation datasets, as evidenced by the raincloud plots. The recently developed ITA and innovative polygon trend analysis (IPTA) were used to determine the trend of individual months, including the sub-trends based on different clusters (low, medium, and high) for complete time series and transition of trend between months, respectively. In addition, the traditional Mann-Kendall (MK) test was also conducted to compare the findings of trends. The monthly precipitation of seven precipitation data, four evapotranspiration data, river basin discharge, and GRACE TWSC were used in the study. The present findings are consistent, as reported in several studies on ITA and a few on sub-trends. What was commonly observed in all the water budget variables is the higher percentage of months detecting either increasing or decreasing significant trends using ITA compared to the classical MK test, which in most cases could not detect any significant trend. The sub-trends provided us with the trends in each of the three clusters. Only APHRODITE, TRMM, and IMERG showed more than 2.5 mm/month decreasing trend in the high category. Numerically, ETs showed insignificant trend variation in all the clusters. Discharge of the basin shows a high decreasing trend in the high cluster (339.01 m3/s) and a decreasing trend in the low cluster by a rate of 176.79 m3/s. Similarly, GRACE TWSC shows a decreasing trend of 7.75 mm/month in the high cluster and an 8.69 mm/month decreasing trend in the low cluster.