Hydrological modelling of largest braided river of India using MIKE Hydro River software with rainfall runoff, hydrodynamic and snowmelt modules

Rainfall-Runoff (RR) estimation plays an extremely vital role in water resource planning, flood forecasting, pollution control, inter-basin water movement, decision-making, and policy formation, among other things. Precipitation distribution, evaporation, transpiration, abstraction, watershed topography, and soil types are all implicit and explicit components that influence the RR process in modelling. The runoff discharge and flow rate at a river location vary dramatically over the time, depending on seasonal rainfall, watershed characteristics, and a variety of other factors. Several models have been established to simulate hydrological processes such as RR, and they are classified as physical, conceptual, and black box models. Surface water modelling, such as runoff, is one of the most often used hydrology studies for estimating peak river flow or the hydrograph generated by actual or hypothetical rainfall (Sitterson et al. 2017). MIKE Hydro River NAM model is a deterministic, lumped conceptual RR model that is a series of connected mathematical statements that describes the characteristics of the land area of the hydrological cycle in a simplified quantitative form (Agrawal & Deshmukh 2016). Lumped conceptual models need a large quantity of calibration data as well as user experience. Physical distribution-based models are no longer appropriate since they require a huge quantity of data on the topography, soil, vegetation, and geological properties of the watershed. The accuracy of empirical black box models is similarly influenced by the quality of observed data, and they are helpful operational tools when there are lacking of meteorological data (Bojkow 2001).

The rational technique (Mc Pherson 1969), the soil conservation service-curve number method (Maidment 1993), and the Green Ampt Method are three well-known RR models (Green & Ampt 1911). The MIKE Hydro River NAM (Nedbør Afstrømnings-Model), MIKE SHE (Systeme Hydrologique Europeen), and WatBal (An Integrated Water Balance Model for Climate Impact Assessment of River Basin Runoff) models were verified by water resource decision makers in three catchments of Zimbabwe (Refsgaard & Knudsen 1996), where at least 1 year's data were available for calibration. The Danish Hydraulic Institute (DHI) developed the MIKE Hydro River NAM hydrological model in 1972. It is an integrated and conceptual model of RR that can simulate surface, subsurface, and base flow (DHI 2017). Shamsudin & Hashim (2002) used the NAM model to estimate runoff rates in the Liang River in Malaysia's northern region, and they found that the predicted values by the NAM mode were in line with historical data. The performance of Artificial Neural Network (ANN) and MIKE Hydro River NAM models was compared by Lipiwattanakarn et al. (2004). They initiate that the ANN model was better at simulating the discharge peak, but the MIKE Hydro River NAM model was better at simulating the base flow or discharge. RR modelling is highly complicated because of its non-linear and multi-dimensional character (Lipiwattanakarn et al. 2004). Liu et al. (2007) proposed a unique sensitivity analysis approach for the MIKE Hydro River NAM RR model, which highlighted the sensitivity analysis challenge in a multi-objective context. Model calibration is essential as the parameters of such models cannot be obtained directly from quantifiable watershed characteristics. Doulgeris et al. (2008) employed MIKE Hydro River NAM to simulate RR processes in the Strymonas river and Lake Kerkini for water resources management purposes. Makungo et al. (2010) generated runoff hydrographs for the un-gauged Nzhelele river using the MIKE Hydro River NAM model and the Australian Water Balance Model (AWBM). Water resources planning and management, as well as the operation of water resources systems, can benefit from the simulated runoff hydrographs. Ferdous (2010) has developed the hydrological model system for the lower Rideau River sub-watershed, Ontario, Canada using MIKE Hydro River NAM hydrological model. Doulgeris et al. (2012) used the MIKE Hydro River NAM model to investigate the RR relationship in the Strymonas river watershed. Rahman et al. (2012) have developed a flood forecasting system for the large Jamuneswari river basin, in Bangladesh using the MIKE Hydro River NAM model. The MIKE Hydro River NAM hydrological model comprises nine parameters and simulates the RR process in the catchment. Some of the parameters can therefore be estimated from physical catchment data, but the final parameter estimation must be performed by calibration using concurrent input and output time series (DHI 2017).

The MIKE Hydro River NAM hydrological model including the snowmelt module was used in this study to improve the calibration of NAM parameters. The calibration and validation procedures of the model were carried out to provide a satisfactory estimation. The specific objectives of this study are stated below.

• To develop an integrated hydrological and hydrodynamic modelling system covering the tributaries and the mainstream Brahmaputra and its flood plain.

• To develop a hydrological model (RR model) for generating runoffs from all the catchments of the Brahmaputra River basin.

• To develop a hydrodynamic model for routing the flow through the main Brahmaputra River, to compute flows, water levels and flood maps.

• To calibrate and validate a hydrological model of the Brahmaputra River basin.

The Brahmaputra River originates in the Chemayungdung mountain ranges near the Mansarovar lake in the Mount Kailash range at an elevation of 5,150 m. It flows across southern Tibet for nearly 2,900 km, passing through the Himalayas, down through the Assam plain, and eventually emptying into the Bay of Bengal. In India, it flows for 916 km. The Brahmaputra basin covers a total area of 5,80,000 km² and is shared by four countries of Tibet, Bhutan, India, and Bangladesh. Tibet owns 50.5% of the entire basin area, India 33.6%, Bangladesh 8.1%, and Bhutan 7.8%. In India, the Brahmaputra basin accounts for roughly 5.9% of the country's total land area. As it travels through one of the world's deepest gorges in the Himalayas and enters the Assam plain, the river drops dramatically in elevation, depositing massive volumes of sediment downstream (Pareta 2021a). The river runs in extensively braided channels (several channels that split off and rejoin) between the Assam and Bangladesh plains, separated by small islands (Pareta 2021b). Since the previous century, the Brahmaputra's (riverbed) size has been steadily increasing. According to reports from the WRD Assam, the river Brahmaputra was spread over 3,998 km² in 1920, 6,003 km² in 2008, and as per Landsat-9 OLI-2 satellite imageries analysis, it has increased to 6,613 km² in 2022.

The Brahmaputra River is characterized by a huge and fluctuating flow due to its reliance on both the monsoon and snow-glacial melt. The monthly average flow rate was fluctuating from 3,244 m³/s in March to 44,752 m³/s in July. The average annual flow rate is 19,824 m³/s, which is the world's fourth highest after Amazon (South America), Congo (Africa), and Orinoco (South America). The highest daily discharge in the Brahmaputra River was 72,726 m³/s in August 1962 at Pandu, while the lowest daily discharge was 1,757 m³/s in February 1968 at the same location (Mahanta et al. 2014). The Central Govt. and Assam Govt. carried out the hydrological observations (HO) in the sub-basin of the Brahmaputra River. In this basin, the Central Water Commission (CWC) manages 108 HO sites, while the Assam Govt and the Brahmaputra Board keep gauge data at 80 locations, gauge-discharge data at 15 locations, and gauge, discharge, and sediment data at 25 locations. CWC has operated 27 flood forecasting stations in the Brahmaputra basin.

The data have been collected from secondary sources for the entire Brahmaputra basin to fulfil the objectives of the study. The collected data were analysed using appropriate analytical procedures. It includes, but is not limited to, rainfall and evaporation from fixed stations, precipitation and snow cover from satellite images, rainfall forecasts from numerical simulations, river cross-sections, river embankments and available high-resolution Digital Elevation Model (DEM), and satellite images. The various critical hydrological data such as rainfall, discharge, water level (WL) and additional relevant information were collected for the study from different sources. The meteorological data such as minimum, maximum, and mean dew point temperature, mean wind speed and hours of bright sunshine were also collected. The collected meteorological data was used as the input for ET₀ calculator (Raes et al. 2009), a software developed by the land and water division of the Food and Agriculture Organization of the United Nations (FAO 1998), to calculate the potential evapotranspiration (ET₀).

Shuttle Radar Topography Mission (SRTM) DEM data with 30 m spatial resolution have been downloaded from https://earthexplorer.usgs.gov/ of the year 2014 for the entire Brahmaputra basin covering an area of 503.85 thousand km². This dataset has been used for the delineation of basin boundary, sub-basin (sub-catchment) boundary, and detailed drainage network. Apart from SRTM DEM data, ALOS PALSAR – Radiometric Terrain Correction (RTC) DEM data with 12.5 m spatial resolution has been downloaded from https://asf.alaska.edu/data-sets/sar-data-sets/alos-palsar/ of the year 2014 for only lower Brahmaputra valley for verification of cross-section data of topography, while bathymetry data have been verified with available bathymetry data and literatures, i.e. Pareta (2021c).

Survey of India (SoI) toposheets at 1:50,000 scale have been downloaded from SoI website at https://onlinemaps.surveyofindia.gov.in/. A total of 171 SoI toposheets have been downloaded, which covered the lower Brahmaputra valley in Assam. Future, these toposheets have been geo-processed and have been used for verification of the basin boundary, sub-basin (sub-catchment) boundary, and drainage network of the Brahmaputra River.

Landsat-5 Thematic Mapper (TM), Landsat-7 Enhanced Thematic Mapper Plus (ETM+), and Landsat-8 Operational Land Imager (OLI) satellite imageries with 30 m spatial resolution, including Thermal Infrared Sensor (TIRS) band (spatial resolution of L-5, L-7 is 90 m, and L-8 is 60 m) have been downloaded from 2001 to 2021 at https://earthexplorer.usgs.gov/ for snow cover and temperature mapping of the entire Brahmaputra basin. These snow cover area and temperature data have been verified from available literatures, i.e. Barman & Bhattacharjya (2015).

For this study, several satellite remote sensing-based precipitation data have been reviewed. A list of available satellite-based precipitation data is shown in Table 1.

The Tropical Rainfall Measuring Mission (TRMM), and The Global Forecast System (GFS) rainfall data with 27.75 km (0.25°) spatial resolution have been used. TRMM is a joint mission between NASA and the Japan Aerospace Exploration Agency (JAXA) to study rainfall for weather and climate research. TRMM satellite was launched in late November 1997 with a design life of just 3 years, however, produced over 17 years of valuable scientific data. There are several gaps in TRMM data for the years 1998 and 1999. It provides seamless data from 2000 to 2014. The GFS is a global numerical weather prediction system containing a global computer model and variational analysis run by the US National Weather Service (NWS). The mathematical model is run four times a day and produces forecasts for up to 16 days in advance, but with decreased spatial resolution after 10 days. TRMM data have been downloaded from 2000 to 2014, and GFS data have been downloaded from 2015 to 2021.

Observed water levels from 2015 to 2017 have been collected from Water Resource Department, Assam for 23 stations. These data were published in the Daily Flood Bulletin of Assam. Afterthought from 2018 to 2021, the observed WL has been collected from CWC at https://ffs.tamcnhp.com/.

The available cross-section data have been obtained from the Water Resource Department, Assam for 65 transects. Future, these cross-section data have been updated with available bathometry data.

For calibration and verification of the model, the global flood monitoring system (GFMS) discharge data with 13.87 km (0.125°) spatial resolution has been used. The GFMS data has been downloaded from 2017 to 2021 (5 years) at http://flood.umd.edu/. River stage (height) is easily measured and widely collected. River discharge is more difficult to measure, and therefore has limited availability. A ‘rating-curve’ relationship is typically developed using stage and discharge measurements at the same location. Discharge data at Dibrugarh, Neamatighat, Bhomoraguri, Guwahati, Goalpara, and Dhubri locations have been assessed using stage-discharge curves correlating GFMS discharge and observed water levels. Rating-curve-based discharge estimates improve gauge-to-gauge correlation approaches, and remove unphysical functional relations between the upstream-downstream gauge due to local channel geomorphology.

In MIKE Hydro River, several modules were available such as NAM (Nedbør-Afstromings Model), UHM (Urban Hydrological Model), Time-Area and Kinematic Wave, but for this study, NAM RR module including snow melt has been used.

The NAM hydrological model is part of the MIKE Hydro River systems’ RR module, which simulates RR processes at the catchment scale. The NAM model is a deterministic and lumped conceptual model with relatively low input data requirements. The NAM model is a continuous moisture accounting hydrological model that can predict long-term water balance and runoff as well as predict short-term inflows. The input data required for a NAM model are meteorological data, catchment parameters definition, initial conditions definition, and discharge/WL data for model calibration and validation. Precipitation, potential evapotranspiration, temperature, and radiation time series are the fundamental meteorological data needs in case snow accumulation and melt are to be modelled. On this basis, the model generates a time series of catchment runoff as well as data on other aspects of the hydrological cycle's land phase, such as soil moisture content and groundwater (GW) recharge.

Precipitation has the most significant impact on the generated flows in a hydrological model (Pareta 2021d). The RR base model was established based on TRMM data, which is later calibrated for GFS data. For calibration, it was decided to use those rainfall stations for the NAM modelling that had discharge data available from 2017 to 2021. In that way, six stations on the Brahmaputra were found suitable for the calibration of the NAM model. In most cases, the time period of availability of rainfall and discharge did not match. In such cases, runoff is based on TRMM rainfall data only. The NAM model has been calibrated with three hourly time series GFS data for the years 2017–2021. Automatic procedures have been developed to download quantitative precipitation forecast (QPF) rainfall data, convert it into MIKE Hydro River time-series format (*.dfs0 file format) and used it in the NAM model simulation.

Snowmelt runoff is an important component of the hydrological cycle and has a significant effect on the water balance. The energy balance technique and the degree-day method are the two most prominent methodologies for modelling snowmelt. Energy fluxes within the snowpack are simulated using the energy balance approach. This approach requires a lot of data and cannot be utilized if you do not have enough. On the other hand, the degree-day technique is based on a basic temperature index approach that assumes temperature is a primary driving component in snowmelt processes. While the index approach has some limitations, it is widely employed due to its simplicity and ease to use.

Accounting for the energy balance at the ground surface leads to the computation of snowmelt using the energy balance approach. It considers all stored, outgoing, and entering energies before calculating the net incoming energy. If the net incoming energy is positive, the system will gain heat, which will eventually cause the snowpack to melt. The amount of energy contributed to the system determines the rate and quantity of snowmelt.

When the air temperature is above a predetermined temperature level (base temperature, T_b) defined by the user, the precipitation falls as rain in the degree-day-based snow module. When the air temperature falls below the base temperature, the precipitation is assumed to accumulate as snow. When the air temperature climbs above the base temperature, the accumulated snow begins to melt and melt water is generated. This melt water is maintained as liquid water in the snow storage until it surpasses the snow storage's water retention capacity.

The runoff simulation can be enhanced by separating the catchment into smaller zones and computing the snow storage for each zone in places where temperature, precipitation, and snow cover vary greatly within a single catchment. The melt water in each altitude zone is calculated using the degree-day approach by the altitude-distributed snow model, which is included in the RR-NAM model. Because hydro-meteorological data in mountain basins is scarce, the module also offers tools for distributing meteorological data according to altitude. The degree-day approach is particularly useful in open and sparsely wooded regions. For hilly and thickly wooded locations, energy-balancing strategies are preferable (DHI 2017). However, a well-parameterized degree-day technique, like the one in MIKE Hydro River, can typically be calibrated for all climatic situations.

The delineation of the basin, sub-catchments, and identification of the river network all require a DEM with sufficient spatial resolution and coverage of the whole research region. The accuracy of the DEM is crucial because inaccuracies that could be generated by the terrain model itself require careful consideration (Pareta & Pareta 2011, 2012). The author has analysed the four global topographic datasets such as SRTM DEM, ASTER DEM, Cartosat DEM, and JAXA DSM with 30 m spatial resolution, and found that SRTM, DEM data with 30 m spatial resolution is the most accurate for characterization of the basin, and sub-catchments. Therefore, that data has been used for extraction of micro-watershed, basin, sub-catchments area and study area boundary. Hydrology tool of Spatial Analyst Tool and/or ArcHydro Tool in ESRI ArcGIS-10x software (Scopel 2014) has been used for this analysis. The author has identified 61 most important tributaries of the Brahmaputra River in Assam, those having a catchment area of more than 1,000 km², hence Brahmaputra basin has been divided into 61 sub-catchments for hydrological modelling.

Snow often accumulates at higher elevations (2,500 m amsl), when the temperature is less than the base temperature (normally between +4 °C and −4 °C). To regulate the temperature in each altitude zone, the dry and wet lapse rate approach is utilized. In this study, a 30 m SRTM DEM was used to derive elevation zones. The GIS tool is used for the automatic extraction of the area in different elevation zones for each of the snow-fed sub-catchments. Using this tool, a summary table containing area-altitude distribution is generated, which is further used to generate a hypsometric curve (Pareta & Pareta 2011). The hypsometric curve is used to estimate zonal mean elevation as an input to the RR-NAM snowmelt hydrological model.

Hydrologists usually want to know how much water is stored as snow in a basin. The distribution of snow, its state, and the presence of liquid water in it will all be of interest to hydrologists. In general, all these snow indicators are difficult to quantify and vary from place to place, especially in mountainous areas (Najafzadeh 2004; Najafzadeh et al. 2022). Sensor technology has advanced to the point that equipment that can withstand extreme weather and hard winter conditions is now available. Even though practical implementations of such systems exist, manual snow surveys are still required owing to station calibration and updates. Remote sensing and geographic information systems (GIS) are two emerging technologies that are rapidly being employed in snow hydrology. It can be used to analyse observed snow and meteorological data, execute necessary up and/or down scaling operations, prepare input variables for hydrological models, determine the parameters required by these models, and show the model runs’ outcomes. Snow data such as snow-covered areas and snow water equivalent to snowmelt runoff forecast may be obtained in real-time via remote sensing, which is a crucial component for a dynamic occurrence like snow (Pareta 2021e).

Satellite remote sensing is becoming a more common alternative in hydrological applications due to lower prices and larger coverage regions (Samantha 2004). The spatial and temporal resolutions, the number of spectral bands, the basin area, and the climatological circumstances are the main parameters for choosing the best sensor in satellite platform selections. Satellites with high repetition rates are crucial in operational snow cover monitoring because it has provided a cloud-free view. Landsat and NOAA satellite imageries are more accessible in this regard than aerial photographs. Even though due to the spatial resolution of NOAA satellites, it can be employed in small-scale area assessments, there are numerous successful applications (Pareta 2021f). Landsat-8/9 OLI can be used in small catchment areas. Both NOAA and Landsat, on the other hand, operate in the visible and near-infrared regions of the electromagnetic spectrum. Landsat-8 OLI and Landsat-9 OLI-2 satellite imageries were used to map the snow cover in the Brahmaputra basin (outlet at the Assam–Bangladesh border).

The Normalized Difference Snow Index (NDSI) is used to distinguish snow from other land-cover types (Kelly et al. 2003; Rees 2006; Armstrong 2010). NDSI method is generally used for snow cover mapping using satellite remote sensing data (Hall et al. 1995, 2002; Kulkarni et al. 2002, 2006; Negi & Thakur 2008). In the visible (green) and shortwave infrared (SWIR) regions, NDSI employs the high and low reflectance of snow, respectively. Furthermore, cloud reflectance stays high in the SWIR region, allowing NDSI to distinguish between snow and clouds. The NDSI scale ranges from −1 to +1 and is defined as: NDSI = (Green − SWIR)/(Green + SWIR). Where, Green and SWIR are the reflectances of the green and SWIR bands, respectively. Snow has a high reflectance in band-3 (0.525–0.600 μm, visible green) and a low reflectance in band-6 (1.560–1.660 μm, shortwave near-infrared) of the Landsat-8/9 OLI instrument. If the NDSI score is more than 0.4, snow is typically presumed to be present (Dozier 1984, 1989; Hall et al. 1995). The snow-covered area (SCA) calculation can be described as: NDSI ≥ 0.4. Although the threshold for the SCA is usually set to 0.4, this figure is not applicable to all land-cover types. Recent research suggests that the ideal threshold value for different land-cover types fluctuates periodically (Rees & Collins 2006), particularly in forested regions.

Sub-catchment-wise SCA for the year 2021 is given in Table 4.

The hydrodynamic river model takes the RR from the NAM and carries out a continuous routing of the flow and flood waves through the main rivers and reservoirs of the basin. The model outputs are discharge and water levels. For application to short-term flood forecasting, data assimilation is also incorporated at all the real-time discharge and WL stations.

The stream network of the Brahmaputra basin was developed from SRTM DEM data by using ArcGIS toolbox (Roux et al. 2013). These networks have been updated and verified with the help of SoI toposheets, and the best available Google Earth satellite imageries. The stream network of the Brahmaputra basin has been classified according to the Strahler classification (Strahler 1957) to find out the mainstream in each sub-catchment. Therefore, the author has selected 24 most important branches, those having available hydrological data such as discharge data, WL data, etc. for hydrological modelling.

The collected 65 cross-section data from Water Resource Department, Assam has been verified with ALOS PALSAR DEM data (topography), and available bathymetry data and literatures. As Brahmaputra River is a very dynamic river in nature, and it has changed its bed level in every monsoon, and every flood wave. So, it is required to update the cross-section data with the most recent available datasets. Consequently, the author has obtained available recent bathymetry surveys (2018) at three locations, i.e. Dibrugarh, Kaziranga and Palasbari-Gumi, which have been used to update the cross-section data. The essential topographical information necessary for hydraulic modelling is river cross-sections, which include flood plain levels. As a result, for all rivers in the model domain, an updated river and floodplain cross-section is required from time to time.

A river link is a stretch of river, a section of a branch, or a point where the catchment's runoff is dispersed to the river. As a result, how the flow from the RR simulation is employed in the hydrodynamic simulation is controlled by the river link's spatial extent. The runoff will be linked to the hydrodynamic model as a lateral inflow in a broad catchment, either at a sequence of single locations or as a dispersed source (DHI 2017). A river link is defined with the two following groups of parameters.

River Link: For the hydrodynamic simulation, the parameters below regulate where the runoff is applied to the river branches.

• • Branch name: The branch where the runoff is applied is selected from this list.

• • Link type: Three-link types can be selected.

• • o Point: the supplied runoff is added to the branch as a point source. A single chainage is used to define this source.

  • o Distributed: a reach along the branch receives the appropriate runoff. The reach's placement is determined by an upstream and downstream chainage.

  • o Entire branch: because the specified discharge is dispersed throughout the whole branch, no chainages are required.

• • Upstream chainage: this chainage describes the position of the point source where the runoff is applied for a point link type. This chainage describes the upstream extent of the reach where the flow is employed for a Distributed link type.

• • Downstream chainage: this chainage determines the downstream extent of the reach where the runoff is applied for a Distributed link type.

Catchment Definition: The runoff value, which is considered part of the connection, is controlled by the settings below.

• Catchment name: from the list, the catchment for which the runoff is calculated is chosen.

• Linked area type: this parameter determines whether the river link portion of the catchment is expressed as an absolute area or as a percentage of the total area.

• Linked area: this option specifies the catchment region that is considered for the river connection. The overall catchment's runoff multiplied by a ratio of the stated area to the entire catchment's area will result in the runoff for this river connection.

• Linked fraction: this parameter specifies the percentage of the catchment that is considered for the river connection. The entire catchment runoff multiplied by the proportion will be the runoff for this river connection. As a result, a fraction of one means that the whole catchment's flow is consumed.

The NAM RR model's catchment runoffs are employed as upstream bounds and intermediate inflows. The RR model is automatically coupled to the hydrodynamic model.

In river modelling applications, boundary conditions are required to solve the commonly used flow and transport equations. All ‘open-end’ sites in the river network must have boundary conditions at the very least. This is where there are no upstream or downstream related river branches on a river branch. At all open borders, an HD model requires inflow or WL conditions. In this model, total of 25 standard boundaries have been steeped as open boundaries. Out of which 24 are defined as type ‘discharge’ that are getting runoff from connected upstream catchments. The downstream boundary is defined as type ‘Q/h relation’.

The NAM model is constructed with nine parameters, representing the surface-zone, root zone, snowmelt data, ground water data, initial conditions, and irrigated area. In this study, the snowmelt component has been incorporated. The snowmelt module uses temperature time series as an input, usually mean daily temperature time series is used. The NAM module has two sub-modules (a) simple snowmelt module which has a constant single value for C_snow parameter and (b) extended snowmelt module, in which C_snow may be defined for each month. In this study, the latter module is implemented. The description of NAM model parameters, and snowmelt parameters are presented in Table 6.

Starting water volumes in the surface and root zone storages, as well as initial values of overland flow, interflow, and baseflow, are required for the NAM model. The starting value of the snow storage should also be given if the snow module is used. When the simulation time is smaller than a year, the parameters established in the model are significantly dependent on the beginning conditions. If an RR simulation begins at the end of a dry season (e.g. the end of May for much of India), it is frequently sufficient to put all beginning variables around zero. In the case of a snowmelt model, however, the initial storage of snow at the end of the monsoon or the beginning of the winter season can be taken to be zero (e.g. end of September in India). However, it should be remembered that the soil will be completely wet at this point. By recording the proper moisture contents of the root zone and baseflow at the same point in the year at the time the current simulation begins, improved estimates of the beginning conditions may be acquired from prior simulations done for a longer data record (several years). The brief description of NAM model parameters used here are described below.

• (i)

  Maximum water content in surface storage (U_max): It displays the cumulative total water content of vegetation-covered interception storage, surface depression storage, and storage in the top few centimetres of soil. The typical range is 10–20 mm.

• (ii)

  Maximum water content in root zone storage (L_max): It shows the maximum amount of soil moisture that is accessible for plant transpiration in the root zone. Typical ranges range from 50 to 300 mm.

• (iii)

  Overland flow runoff coefficient (CQOF): It controls how much extra rainfall will go overland and how much will infiltrate the ground. The values range from 0.0 to 1.0.

• (iv)

  Time constant for routing interflow (CKIF): It determines the interflow quantity, which falls off with increasing time constants. Values between 500 and 1,000 h are typical.

• (v)

  Time constants for routing overland flow (CK1, CK2): They determine how hydrograph peaks should be shaped. Two linear reservoirs (serially connected) with various time constants, given in [hours], are used for the routing. Small time constants are used to replicate high, abrupt peaks, while large values of these parameters are used to represent low, later-occurring peaks. Values between 3 and 48 h are typical.

• (vi)

  Root zone threshold value for overland flow (TOF): It establishes the proportional value of the root zone moisture content (L/L_max) at which overland flow is produced. At the start of a wet season, where an increase in the parameter value may postpone the initiation of runoff as overland flow, TOF has the greatest impact. The permissible threshold value ranges from 0 to 70% of L_max, with a maximum of 0.99.

• (vii)

  Root zone threshold value for interflow (TIF): It establishes the proportional value of the root zone moisture content (L/L_max) over which interflow is produced.

• (viii)

  Root zone threshold value for GW recharge (TG): It establishes the proportion of the root zone's moisture content (L/L_max) over which GW recharge is produced. Less recharge to the GW store is the main effect of increased TG. Threshold values range from 0 to 70% of L_max, with a maximum value of 0.99 permitted.

• (ix)

  Time constant for routing baseflow (CKBF): In dry seasons, it can be calculated from the hydrograph recession. Rarely, the measured recession eventually takes on a slower recession structure. A second GW reservoir might be added to approximate this; some enlarged components are discussed below.

• • Ratio of GW-area to catchment area (Carea): The ratio of the topographic surface water catchment area to the GW catchment area is specified under the General tab and is described by this parameter. A portion of the entering water may drain to another catchment due to local geological conditions. A Carea of less than one describes this water loss. The normal value is 1.0.

• • Specific yield of GW reservoir (Sy): Except in rare circumstances where the GW level is utilized for NAM calibration, this parameter should be left at its default setting. This may be necessary, for instance, in riparian areas where GW drainage has a significant impact on the seasonal variance of river levels nearby. Specific yield Sy and GW outflow level GWLBFO data, which can change over time, are needed in order to simulate GW level variation. The value of Sy varies by soil type and is frequently determined by hydro-geological information, such as test pumping. Typically, clay values range from 0.01 to 0.10 and sand values from 0.10 to 0.30.

• • Maximum GW-depth causing baseflow (GWLBF0): It shows the measurement in metres between the typical catchment surface level and the river's lowest WL. Except in rare circumstances where the GW level is used for NAM calibration, this parameter should be left at its default setting.

The snow module simulates the accumulation and melting of snow in a NAM catchment. There are two degree-day methods that can be used: a straightforward lumped computation or a more sophisticated distributed method. The constant degree-day coefficient and the base temperature (snow/rain) are the only two global factors used by the basic degree-day technique. With different snowmelt parameters, temperature, and precipitation input for each elevation zone, the user can select a number of elevation zones within a catchment using the distributed approach. The snowmelt parameters are:

Constant degree-day coefficient (C_snow): The melting rate of the snow storage is determined by the temperature difference above the base temperature divided by the degree-day coefficient. Snowfall rates typically range between 2 and 4 mm per day/C.

Base temperature (snow/rain) (T_o): Only when the temperature is below the base temperature is the precipitation kept in the snow storage, whereas in higher temperature conditions, it is bypassed and stored on the surface (U). Typically, the base temperature is 0°C or very close to it.

Since the parameters of the MIKE Hydro River NAM could not be directly determined from quantifiable sub-catchment variables, model calibration is required. The process of calibration involves changing model parameters to minimize the discrepancy between the observed and simulated streamflow. Model calibration for the MIKE Hydro River NAM model is done manually.

In manual calibration, a visual assessment is made by contrasting the measured and the expected discharge, leading to a trial-and-error parameter modification. The model was run in auto-calibration mode while the default model settings were maintained (Amir et al. 2013). The calibration was repeated with extraordinarily little variations after multiple manual calibrations. The calibration was examined for the coefficient of determination (R²), water balance error (%WBL) values, and visually assessed for the level of agreement between simulated and observed runoff to ensure the accuracy of the results. Model validation entails assessing the calibrated model's performance. When a model's accuracy and predictive power throughout the verification phase are shown to be within acceptable bounds, it is considered to have been validated (Refsgaard & Knudsen 1996). Utilizing the updated set of observed data and the parameters that were calibrated in the preceding stage, verification is conducted.

The model's calibration focused on the routing of river inputs. It is a two-step process to calibrate the river and flood plain resistance:

• Obtain the closest match between observed and simulated discharges; and

• Using the above-found discharge boundaries, achieve the greatest feasible match between observed and simulated water levels.

Adjusting NAM parameters to produce the greatest feasible match between observed and simulated discharge calibrates the RR model. The hydrodynamic model is calibrated by changing slope and roughness along and across the river in order to get the greatest feasible match between observed and simulated water levels using the discharge boundaries described above.

In addition to the timing and form of flood hydrographs, the calibration will focus on medium to high-flow situations with regard to peak levels. The calibration did not prioritize the prediction of low flows since this needs thorough information on the regulation occurring along the river during the dry season, which was not accessible. Both graphical and numerical measurements were used to evaluate the calibrated model in general. The calibration plots are included in the graphical evaluation, and the overall WBL and the overall form of the hydrograph based on the coefficient of determination R² are numerical measurements. R² = 1 and WBL = 0% indicate a perfect match.

The comparison reveals a good agreement between the simulated and real data as well as between the timing, rate, and volume of the hydrograph's shape. Hafezparast et al. (2013) used the MIKE Hydro River NAM model to predict streamflow in the Sarsoo river basin and reported similar results. They discovered that the observed and simulated flow levels agreed very well. Amir et al. (2013) used the MIKE 11-NAM model in research in Fitzroy Basin, Australia, and discovered that there is good hydrograph agreement between observed and simulated discharge, demonstrating the model's capacity to replicate streamflow in the basin. Odiyo et al. (2012) used the MIKE 11 NAM model to simulate the streamflow in the Latonyanda River Quaternary catchment (LRQ). They discovered that, with the exception of a few low flows and under-predicted peak occurrences, the observed and simulated streamflow for the LRQ watershed corresponded well. As shown in Figure 9, the coefficient of determination (R²) and %WBL for the model calibration were obtained at 0.864 and <6%, respectively, indicating that the model calibrated well and could replicate the catchment's runoff. It was found that the averages of the simulated and observed basin discharge were in good agreement.

The validity of the model had been examined using the time series from 2000 to 2021. The collection of model parameters obtained after calibration was used to mimic the runoff during validation. For the model validation, the observed coefficient of determination (R²) and %WBL were 0.63 and 6.3%, respectively. The model was able to calibrate runoff outside of the calibration periods, as evidenced by the high degree of agreement between the averages of the simulated and observed basin runoff volumes for the validation period. RR model (NAM) calibrated parameters are available with the author and can be obtained by e-mail.

The availability of historic discharge data for snow-fed catchments is given in Table 7.

Table 7

Discharge data available for calibration for snow-fed catchments

View Large

In Table 8, T0 is mostly N/A, 0 or/and −1, because when the temperature is below the base temperature, precipitation is stored in the snow, but when the temperature is greater, it bypasses the snow storage and is stored on the surface (U). The basal temperature is often zero degrees Celsius or very near to it. R² and WBL for snow-fed catchments are given in Table 9.

The %WBL and coefficient of determination (R²) for the model calibration were obtained at 0.643–0.864 and <6%, respectively, showing that the model calibrated successfully and could mimic the catchment's runoff. It was discovered that there was good agreement between the averages of the simulated and observed basin discharge.

Plotting EI and R² against the appropriate model parameters allowed for an analysis of the output data. The most significant and sensitive model parameters were discovered to be CQOF, CK1,2 and L_max. The remaining parameters, however, were discovered to be non-sensitive. Because of the high disparities in EI and R² magnitude they cause, the significant parameters are CQOF, CK1,2, and L_max. The findings were backed up by a number of writers, including Yadav et al. (2019), who used MIKE-NAM 11 to estimate RR in the Sher River basin and found that the model's CQOF and CK1,2 parameters were the most sensitive. According to Wakigar (2017), the MIKE11-NAM model's U_max,L_max, CQOF, and CK1,2 parameters were the most capable of forecasting streamflow in the Upper Guder Catchment. In this work, sensitivity analysis significantly contributed to the increase in calibration by identifying the parameters that needed to be prioritized.

Basin and sub-catchment scales modelling of hydrological processes are important for different purposes including water resources planning, development, and management. This study focusses on the hydrological modelling of the largest braided river (Brahmaputra River) in India using MIKE Hydro River software. The QPF by GFS has been incorporated into this study. SRTM DEM data; ALOS PALSAR – RTC DEM data; SoI toposheets at 1:50,000 scale; Landsat-5 TM, Landsat-7 Enhanced Thematic Mapper Plus (ETM+), and Landsat-8 OLI satellite imageries with TIRS band; TRMM data; QPF by GFS; Observed WL from WRD, Assam and CWC; and GFMS discharge data have been extensively used.

Basin, sub-catchment, and drainage network has been extracted from SRTM DEM data, and identified 61 most important tributaries of the Brahmaputra River in Assam, hence Brahmaputra basin has been divided into 61 sub-catchments for hydrological modelling. The Tuting-Siang sub-catchment has been analysed with an area-elevation map, area-altitude distribution zones and percentage of area in each elevation zone (hypsometric curve). Snow cover mapping of the Brahmaputra basin has been done by using Landsat satellite imageries. The total SCA in 2021 is 151,905.37 km². To improve lead times at Dibrugarh, a hydrodynamic river model with Tibet's catchment has been setup up to upstream. There are 24 branches altogether in this model. The model considers each important tributary that affects the Brahmaputra outflow.

In this work, the capability of the MIKE Hydro River NAM model to simulate streamflow in the Brahmaputra basin was assessed and has been setup for 61 sub-catchment and calibrated against the measured discharge at five stations. Overall, during both the calibration and validation periods, the model successfully simulated daily streamflow during the monsoon period. Results showed that the model's EI, R², and %WBL during the calibration period (2017–2019) were 0.871, 0.893, and 0.786%, respectively, whereas they were 0.533 and 0.143% during the validation period (2020–2021). An excellent match between the simulated and observed data, as well as a solid overall agreement in the shape of the hydrograph with respect to timing, rate, and volume, demonstrated the model's competence. It should be emphasized that the model's performance might be enhanced yet more by adding more stations that can accurately record the spatial distribution of rainfall in the Brahmaputra basin. The MIKE Hydro River NAM model is a valuable tool to simulate streamflow, especially in locations with limited data availability, due to its simplicity and minimal data needs. The model's best (perfect) performance is indicated by an EI of 1. During the performance analysis, the author measured an average EI of 0.85. The model is effective enough to replicate the catchment's runoff based on the results of both evaluation criteria

The RR-NAM model including the snowmelt component has been setup for 29 sub-catchments and calibrated against the measured discharge at 11 stations. During the calibration procedure, the optimum values of the nine RR-NAM parameters were obtained by considering a good agreement with observed runoff in terms of timing, volume, and shape of the hydrograph. The performance of the model has been evaluated by comparing simulated runoff with measured discharge using the coefficient of determination (R²). It has been observed that the value of R² is ranging between 0.643 and 0.864. This can be deemed acceptable for the purposes of this study. In addition to R², the overall WBL has also been checked, and it has also been observed to be less than 6% error.

The TRMM/GFS satellite-based rainfall data has been used, without any correction, as input to build the hydrological model. However, model performance can be improved if rain-gauge data is available for bias correction of TRMM/GFS precipitation data. Due to the lack of local rainfall data for bias correction of satellite-based rainfall data, It may be possible for inaccurate discharge rating curves at river gauging stations. The model performance can also be improved if observed discharge would be available for the missing sub-catchments. If extended records of discharge from the existing stations can be retrieved this will enhance the possibility of improving the model calibration. Although PET does not play a significant role in the current study, the model performance can also be improved if daily PET data is incorporated. Despite the uncertainties that are unavoidable in hydrological modelling it is judged that the calibrated RR-NAM model can be used for the Flood Forecasting and Early Warning System design and water resources management and planning of the Brahmaputra basin.

The author is grateful to Managing Director, DHI (India) Water & Environment Pvt Ltd, New Delhi India for providing the necessary facilities to carry out this work.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The author declares there is no conflict.