AN INTEGRATED FRAMEWORK OF RECONSTRUCTING FLUVIAL ARCHITECTURE FOR HYDRAULIC MODELLING: APPLICATION OVER A DATA-SCARCE HIMALAYAN RIVER IN INDIA

Abstract

The demand for sand is rising steadily on a global scale. One of the main sources of aggregates required to meet the demands of the quickly expanding construction sector is sand mining. Despite being effective flood control methods, mining for sand and gravel is known to have a detrimental effect on the riverine ecology. Given the state of commercial aggregate mining today, one of the numerous challenges that will surface in the near future is the safety of hydraulic structures in excavated riverbeds. Numerous researchers have emphasized the paucity of comprehensive studies about commercial in-stream mining and its effects on the river systems' hydro morphology. hydrodynamic modeling methods typically rely on detailed bathymetric data of river channels, often represented by a series of cross-section profiles distributed along the river segment of interest. However, such detailed data is available only for a limited number of rivers. Consequently, one common challenge faced in many studies is how to approximate channel geometry when detailed data is lacking. This research attempts to develop a framework for the reconstruction of fluvial architecture i.e. cross sections shapes and dimensions, slope and roughness, essential for the hydraulic modelling of the sand mining affected river reach. This study utilized bed level data over 30 years of a cross section to generate a trendline which was further utilized to estimate bed level at different cross sections along the river during different years (1992-2022). The elevation of critical points along the cross section are estimated utilizing the slope of the river at corresponding points according to the surveyed cross sections data of 2022. And then joining these points with parabolic curves provided us with the cross sections along the river reach for various years during 30 years (1992-2022). The 1D hydraulic model was set up in MIKE+ with reconstructed cross sections along the river reach for years 1992, 1997, 2002, 2007, 2012, 2017 and 2022. The model was calibrated and validated with the June-July and August-September discharge data respectively from ERA5 with Manning’s roughness coefficient, n as a model calibration parameter. Manning’s n as 0.033 provided best performance indices with RMSE as 0.1333 m3/s, R2 as 0.9875, NSE as 0.9891, IOA as 0.9971. The coefficient of determination (R2), Nash-Sutcliffe efficiency (E) and index of agreement (IOA) were found to be 0.9876, 0.9892 and 0.9971 respectively for the validation of the model. The simulated hydrographs for different years model set up at the chainage of 4500 m along the river and with 100 years return period flood hydrograph as input at the upstream boundary have come out to be almost identical. Water depth for the years 1992, 1997, 2002, 2007, 2012, 2017 and 2022 at 4500 m chainage are coming out to be close to 4 m except for 2022 i.e. 4.1 m, 4.2 m, 4.4 m, 4.3 m, 4.3 m, 4.6 m and 5.8 m, which is increasing with time.