Minimisation of Overestimation of River Flows in 1D- Hydrodynamic Modeling

One-dimensional hydrodynamic models overestimate river cross-section derived from freely available SRTM DEMs. The present study aims to minimize the overestimation of river flow. DEM-extracted cross-sections obtained from 30 m and 90 m resolutions show higher elevation values than the actual river cross sections of Krishna and Bhima rivers, India. To minimize the overestimation of the river flow, DEM-extracted cross-sections are modified using known cross-section of the river. The corrections for cross sections extracted from DEM, are obtained by subtracting the DEM-derived cross-sections from a known cross-section of the river. Monsoons flows that occurred in years 2006 and 2009 in Krishna and Bhimariver have been used for modeling. The MIKE HYDRO River model performance with modified DEM-extracted cross-sections of river improves as the correlation coefficient, root mean square error, index of agreement, Nash Sutcliffe efficiency & Percentage deviation in peak (%) values are improved.


INTRODUCTION
Among the natural hazards (cyclones, floods, landslides, drought, volcanic eruptions, and earthquakes) occurring in developing countries, flood is the most prominent for study as it is the most frequent disastrous event. Many a time, it causes a heavy toll on human lives, loss to economy, geographical and geological damage to the locality. India is one of the worst flood-affected countries and the Himalayan Rivers have a large contribution in causing floods in different states. The main causes of floods in India are due to overtopping of river banks because of sedimentation and low carrying capacity, heavy precipitations, rise in water level due to cyclones and high tides, poor drainage, and management practices. To minimize the devastation caused by floods, information of the disastrous event for improving decision making, planning design, and construction activities in the flood-prone areas plays a vital role. Such information can be inferred from the mathematical modeling which simulates the spatial and temporal characteristics of the flood events.
One-dimensional hydrodynamic models like HEC-RAS, ISIS 1D, MIKE 11, MIKE HYDRO River, SOBEK, TUFLOW 1D, are the simplest ones and require less computation time and effort but does not give good results with complex rivers. These models simulate the channel and floodplain as a series of lateral-sections perpendicular to the flow direction and solve either the complete or some approximation of the one-dimensional shallow water equations (SWEs) (Bates & Roo (2000)). 2D models solve the two-dimensional SWEs using the method of characteristics, finite element, or finite difference method. Models such as DIVAST, HEC-RAS 2D, ISIS 2D, JFLOW, LISFLOOD-FP, MIKE 21, RMA2, SOBEK 2D, TELEMAC 2D, TUFLOW 2D require an accurate representation of bathymetry, higher computational cost, and time (Horritt & Bates (2002); Mignot et al. (2006)) but with the advancement of remote sensing technology, these models are gaining popularity (Bates (2004)). 3D models such as DELFT 3D, MIKE 3, TELEMAC 3D, TUFLOW 3D are complex and have high data requirements than the simpler models giving realistic results ( Hunter et al. (2007)). Also, there are 1D/2D coupled models used in literature like MIKE FLOOD, MIKE URBAN, SOBEK Suite, TUFLOW Classic used by researchers (Fan et al. (2017); Kadam & Sen (2012); Lin et al. (2006);  for flood inundation modeling. One of the advantages of hydrodynamic models is that they have a direct linkage to hydrology and can account for hydraulic features and structures.
There is difficulty in hydrodynamic modeling in developing countries because of the lack of data availability. The problem is encountered because of the non-availability of the highresolution digital elevation models (DEMs), cross-section data, and a suitable set of data for calibration and validation. 1D models can simulate accurately because of the accuracy of describing the hydraulic behavior of streams and rivers (Chen & Liu, 2017). Also, if high-resolution data is available 1D models are able to perform very well in larger stream domains (Bates (2004)). So the need has arisen for using high-resolution DEMs because low-resolution DEMs have the limitation of dropping out important features that affects fluid dynamics (Cook & Merwade (2009) ;Marks & Bates (2000); Omer et al. (2003); Werner (2001)).
As the hydrodynamic modeling results (e.g. Lane, 2006, Horritt et. al. 2006) depend on the spatial resolution of DEM, one-dimensional hydrodynamic models overestimate simulated river level due to inherent overestimation of elevation of riverbed cross-section derived from freely available SRTM DEMs. Thus, the present study tries to minimise this overestimation of river flow using MIKE HYDRO river. To perform 1D hydrodynamic modeling, on the available data set (pertaining to the Krishna and Bhima River) MIKE HYDRO River model was chosen because of less complexity of the river system, and less computational time. In 2009 heavy rain caused a flash flood in the north Karnataka and Rayalaseema region of Andhra Pradesh that affected nearly 2 million people and claimed 210 lives (Padmanabhan, 2009

Study area
In Peninsular India, Krishna Basin is the second-largest eastward draining interstate river basin that lies between the longitudes 73̊ 21'00''E and 81̊ 09'00''E and the latitudes 13̊

Data used
The data used in this study for hydrodynamic modeling in MIKE Hydro (River) are DEMs, For the open end boundaries in the model, at the upstream boundaries a discharge file was supplied, and a constant water level was specified as the d/s boundary condition. The discharge and water level data collected from CWC were represented as observed data in this study and the data predicted by the model were referred to as simulated data.
The SRTM (Shuttle Radar Topographic Mission) DEM of the study area of 30 m and 90 m resolution were downloaded from www.csi.cgiar.org and https://earthexplorer.usgs.gov (Reston, 1993) These downloaded files were made mosaic to get the complete DEM of the study area. The downloaded DEM needed to be preprocessed first to fill the voids and then used for tracing of rivers and extraction of cross-sections for the study.

Extraction of cross-sections from DEM
River cross-sections are of prime importance for the hydrodynamic modeling of a river.
Cross-sections of the Krishna and Bhima Rivers were extracted from the DEMs. These DEMs were projected at the topographic map at projection system GCS-WGS-1984 and all the digital elevation data were preprocessed before extracting cross-section. Rivers were added in the model using the add trace option available for digitizing the rivers keeping the topographic map of the study area as the base map. Cross-sections were then generated at  (1) and (2) using an implicit, finite-difference scheme (Abbott & Ionescu, 1967): (1) ii) Root mean square error (RMSE): Where O i is the observed water level at the ith hour , P i the predicted or simulated water level, n is total number of observations; O avg is average of observed water level; is the predicted peak and is the observed peak of water level.   (Pan et al., 2013) which indicates that the model does not perform well with the DEM extracted cross-sections. The overpridiction of flows are due to extracted cross-sections from DEMs.


This method is computationally efficient as compared to the method adopted by