Integrated hydrological modelling of two contrasting watersheds with a terminal reservoir in the Upper Tapi River basin, India Open Access

Water is vital for life but is limited and unevenly distributed in space and time. Understanding the movement of water and its interactions are challenging tasks in various phases of hydrological processes. Nearly one-third of the world's population may face water shortage by the year 2050 due to qualitative and quantitative water reductions caused by rapid industrialisation, urbanisation, population increase, desertification, global warming, etc. (Costa et al. 2003; McColl & Aggett 2007). Climate change–induced extreme weather events (floods/droughts) and seasonality of monsoon rainfall, especially in developing nations like India, inflict substantial socioeconomic disruptions and question the temporal availability of water (Monirul & Mirza 2003). To deal with these multifaceted water resources management issues, the dams and reservoirs evolved with human civilisation, embarking more adaptability on the one hand; on the other, they interfered with the natural hydrological processes (Zhang et al. 2012). This necessitates undertaking an integrated hydrological modelling (IHM) approach, including hydrological processes of watersheds, reservoirs, and water exchange more holistically and comprehensively to better understand water resource systems for their sustainable and efficient water usage.

The computer-based numerical model invariably represents complex physical hydrologic processes and their interactions which is used to analyse water movement, distribution, and behaviour in the watersheds (Kim & Parajuli 2012). For instance, Hydrological Simulation Program-Fortran (HSPF; Donigian et al. 1995), MIKE-Système Hydrologic Europeèn (MIKE-SHE; DHI 2017), variable infiltration capacity (VIC; Hamman et al. 2018) model, soil and water assessment tool (SWAT; Neitsch et al. 2011), etc. are the models used by hydrologists for a variety of applications. The SWAT is a widely used physically based, computationally efficient, continuous hydrological model to mimic watershed processes (Neitsch et al. 2011). The calibration of the hydrological model is one of the most crucial processes to ensure an accurate representation of actual watershed processes (Sharma et al. 2022). Researchers have emphasised multisite and multivariable calibration approaches owing to their numerous hydrological improvements over single-site calibration (Daggupati et al. 2015; Shah et al. 2021). Though the SWAT is extensively explored for simulating streamflow, sediment, and pollutant loading for different hydrological conditions across the globe (Thampi et al. 2010; Piniewski et al. 2017; Setyorini et al. 2017; Abbaspour et al. 2018; Munoth & Goyal 2019, 2020a; Wang & Cao 2021; Rostami et al. 2022), the literature lacks a comprehensive application of the SWAT for IHM of watershed hydrology and reservoir processes using multisite and multivariable calibration approaches.

The Upper Tapi River basin (UTRB) in India is a physio-climatically heterogeneous region due to its unique geographical setting and relief (Sharma et al. 2019a). In IHM, the model complexities increase when the watersheds encompass large-scale physioclimatic heterogeneity (Birkel & Barahona 2019). In recent decades, the Hatnur reservoir, a terminal reservoir of the UTRB, and the upstream watersheds have observed a shift towards increasing water stress conditions (Sharma et al. 2019b). The head watershed, i.e., Burhanpur, having high runoff coefficients, generates very high runoff, making the UTRB susceptible to flooding (Sharma et al. 2019a) and prone to excessive erosion (Resmi et al. 2020; Ramani et al. 2021). Notably, the 2006 flood in Surat city in the Lower Tapi basin was responsible for the death of 150 people and losses of 210 billion Rupees (Jibhakate et al. 2023b) due to heavy runoff contributions from the UTRB. Gehlot et al. (2023) are also concerned about the transition of the UTRB towards dry-warm climatology in the near future (2021–2050), which would escalate the present water stress conditions in the region. Chandra et al. (2014, 2016) and Munoth & Goyal (2020b, 2022) studied the hydrologic response of the headwater regions without considering the hydrologic responses of its largest tributary and terminal reservoir.

Therefore, by considering the limitations of previous studies, as stated above, and the importance of the regional hydrologic assessments, the present study adds another dimension to the existing knowledge on the hydrologic behaviour of two contrasting watersheds and their terminal reservoir using an IHM approach at a daily timescale. The primary objective of the present study is to develop an integrated framework of the SWAT model and calibrate the same to capture the complex and diverse characteristics of the UTRB, Burhanpur and Purna watersheds, with their terminal reservoir, i.e., Hatnur reservoir. The study also emphasises the simulation of the reservoir in the IHM framework and spatial variability of water balance components for harnessing the water resources potential of the study region. Such analyses would help in effectively planning and managing the water resources for meeting the irrigation and domestic demands in the command area of the reservoir in the Middle Tapi River basin (MTRB) and the prevention of floods in the Lower Tapi River basin (LTRB).

The Hatnur reservoir is a terminal reservoir of the UTRB, which was constructed in 1982 with a gross and live storage capacity of 388.0 and 255.0 million cubic meters (MCM), respectively. The reservoir was commissioned in 1983 to serve the purpose of irrigation and water supply in the basin. The gross command area and the culturable command area of the Hatnur reservoir are 59,150 and 47,350 ha, respectively. The full reservoir level (FRL) at RL 214.0 m corresponds to the gross storage capacity of 388.0 MCM. The new operation strategy of the Hatnur reservoir advises no water impoundment until August 21, with a minimum reservoir level of RL 209.5 m and a corresponding storage capacity of 180.0 MCM. The reservoir is usually filled up to the FRL by October 15.

The climate data were procured from the India Meteorological Department (IMD), Pune. The station-based rainfall datasets have been used for the model simulation after infilling the missing rainfall data using the inverse distance weighting method (Gehlot et al. 2021). The gridded temperature data of IMD (Srivastava et al. 2009), available at 1.0° × 1.0°, have been used after bilinear interpolation to 0.5° × 0.5° (Jibhakate et al. 2023a). The streamflow data at Burhanpur and Yerli gauging stations were procured from the CWC, Surat Division. The inflow/outflow to/from the Hatnur reservoir was obtained from the Tapi Irrigation Development Corporation (TIDC), Jalgaon. These were used for the calibration and validation of the developed hydrologic model.

The Upper Tapi River flows up to the Hatnur reservoir, and the Purna River merges into it around 8 km upstream of the reservoir. The Hatnur reservoir was simulated in an integrated manner by extending the SWAT model of the Burhanpur watershed to the reservoir by defining a terminal reservoir and an inlet point for adding discharge of the Purna watershed at the confluence of the Upper Tapi River and the Purna River. The SWAT model for the Purna watershed was developed separately to account for large-scale topographical, morphological, and hydroclimatic heterogeneity in the UTRB.

Since the Hatnur reservoir is a controlled reservoir, the target release method was used to simulate the reservoir outflows. The reservoir parameters were obtained from the field agency, i.e., TIDC, Jalgaon. The parameters defined for the reservoir module in the SWAT are shown in Table 2. The monthly target storage for the flood period (July–October) was provided based on the existing rule curve of the reservoir. In contrast, the target storage of the FRL was used for the remaining months (non-flood period) to maximise the reservoir storage. The discharges from the reservoir for municipal and agricultural needs are abstracted from the storage for consumptive use.

The Sequential Uncertainty Fitting algorithm (SUFI-2; Abbaspour 2015) in SWAT-CUP (Calibration and Uncertainty Program) calibrates the SWAT model. A multimeric approach is used to calibrate and validate the developed model for Burhanpur using the observed streamflow at the Burhanpur gauging site, inflow/outflow to/from the Hatnur reservoir. At the same time, the model for the Purna watershed is calibrated and validated using the measured streamflow at the Yerli gauging site. The warm-up, calibration, and validation periods are 1994–1997, 1998–2007, and 2008–2013, respectively. To better represent the watershed hydrology of the heterogeneous catchments of the UTRB, the parameters related to the basin, main and tributary channels, soil, HRUs, groundwater, and watershed management were identified, and a subset of these parameters is used to calibrate the developed models for two watersheds.

The areal rainfall duration curve (Figure 4(b)) signifies that the Burhanpur watershed is more susceptible to high-magnitude rainfall extremes and more rainy days (when daily rainfall equals or exceeds 2.5 mm) than the Purna watershed. The average annual rainfall in the Burhanpur and Purna watersheds is 994.5 and 817.1 mm, respectively. The seasonal distribution of annual rainfall in the former (latter) watershed during monsoon (JJAS), post-monsoon (ON), winter (DJF), and pre-monsoon (MAM) is 92.7% (87.5%), 5.0% (8.7%), 1.4% (2.2%), and 1.0% (1.7%), respectively. The narrow valley and the elevated Betul Plateau of the Burhanpur watershed led to a strong orographic effect, resulting in higher rainfall during the southwest monsoon. Interestingly, the Purna watershed has a higher post-monsoon contribution than the Burhanpur watershed. The higher elevation of Satpura hills can be a plausible reason for preventing the northeast monsoon in the Burhanpur watershed.

The varying rainfall characteristics of these watersheds lead to substantially different streamflow characteristics (Figure 4(c)). The Burhanpur watershed has dominant forest cover, while the Purna watershed has agricultural land. Both catchments have nearly three-quarters of the soil cover of clayey texture. The steep topography of the Burhanpur watershed leads to large streamflow peaks compared to the flatter Purna watershed. The Upper Tapi River flows nearly 85% of the time of the year, showing the significant contribution of the baseflow. The persistence of the forest land cover helps detain the runoff, leading to percolation and, thus, prolonging the baseflow assistance in the Burhanpur watershed. On the other hand, the alluvium plains of the Purna watershed provide good potential for groundwater reserve in shallow aquifers. The large number of detention structures in the Purna watersheds is also responsible for the availability of the flows in the Purna River (Sharma et al. 2019a). The storage capacity of the downstream Hatnur reservoir has reduced significantly due to the reservoir sedimentation with prime contribution from the Burhanpur watershed due to high erosion-prone areas in the watershed (Resmi et al. 2020; Ramani et al. 2021). The reservoir greatly influences the low flows in the river, which are evident from the flow duration curves of inflow and outflow from the Hatnur reservoir, as shown in Figure 4(d). Due to the limited storage capacity of the terminal reservoir, i.e., 388.0 MCM, the reservoir does not affect the high flows and passes on to the downstream MTRB without much interference. Sharma et al. (2021) showed moderate alterations in the natural regime of the Tapi River due to the construction of the Hatnur dam. Because of physiographic and hydroclimatic heterogeneity and the influence of human artefacts on the hydrological regime, IHM and investigations are required to understand the watershed dynamics of the UTRB.

The Purna watershed has more human interventions in the form of detention structures. In contrast, the Burhanpur watershed resembles a virgin catchment. This enables more parameterisation efforts in calibrating the hydrologic model for the former watershed vis-à-vis the latter. The hydrologic modelling of the headwater catchment of the UTRB, i.e., Burhanpur watershed, was performed by integrating the Hatnur reservoir. Additional parameters related to the reservoir are considered during the calibration process to account for the integrated multisite and multivariable calibration processes. The multisite and multivariable calibration approach is adopted, considering its numerous advantages over single-site calibration. It considers the impacts of spatial heterogeneity in the watershed characteristics during semi-automated calibration and parametric sensitivity assessments of the hydrological models.

The parameters in Table 3 are related to the fundamental processes of water circulation in the hydrological cycle, namely, groundwater (.gw), land management (.mgt), soil (.sol), main channel (.rte), hydrological response units (.hru), tributary channels (.sub), and basin parameters (.bsn). Sensitive parameters, at a 5% significance level, comprise the groundwater return flow/base flow, land management, soil, primary and tributary channel, basin, and HRU processes (Table 3). However, the top five sensitive parameters for the Purna watershed are the threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw), groundwater ‘revap’ coefficient (GW_REVAP.gw), effective hydraulic conductivity in the main channel in mm/hr (CH_K2.rte), SCS runoff curve number (CN2.mgt), and deep aquifer percolation fraction (RCHRG_DP.gw). It can be observed that the parameters from the baseflow and main channel dominate among the family of sensitive parameters. On the other hand, hydrologically sensitive parameters for the Burhanpur watershed (Table 3) are RCHRG_DP.gw, GW_DEALY.gw, CN2.mgt, SLSOIL.hru, and ALPHA_BF.gw. The groundwater and soil slope parameters were sensitive because a large portion of this sub-basin is covered with forest area and has clay loamy and clay soils. Also, the upper part of the catchment has a fractured rock area, which is favourable to return flow. The clay particles can attract and retain water molecules due to the polar nature of water molecules. These types of soils have larger amounts of water for plant uptake. Therefore, there may be a chance that the deep-rooted plants can uptake water from the shallow aquifer itself. The CN2 is a function of soil permeability, slope, land use, and antecedent moisture conditions, and its values show the runoff characteristics of the basin.

It is evident from the performance indicators (Table 4) and the flow hydrographs (Figure 5) that the developed hydrological model is reliable for predicting inflow into the terminal reservoir of the UTRB, i.e., the Hatnur reservoir. The outflow simulated using the target storage approach in the reservoir module of the SWAT matched well with observed releases from the reservoir. Also, the uncertainty in the model simulated outflows shows that the model encapsulated nearly 70% (80%) values of the observations during calibration (validation) at a relative thickness of 0.07 (0.09). Since the reservoir module of the SWAT simulated the watershed processes on a daily timescale, the provision for target storage in the reservoir module is only at the end of the month. It could be one of the significant factors influencing the simulation of the outflow from the Hatnur reservoir. In addition, the model underestimates the streamflow in the Burhanpur watershed, particularly the peak flows. Since the SWAT simulates the watershed processes at the HRU scale, the meteorological variables are considered uniformly at the sub-watershed scale, restricting the ability of the model to account for the local scale weather patterns. It can also be due to the non-availability of sufficient rain gauge stations to capture the orographic effect in the Burhanpur watershed.

The Burhanpur watershed has steep topography and rocky strata with prominent geological units like the Deccan trap. In contrast, the Purna watershed has a flat topography, with alluvium, favouring agricultural practices in the watershed (Sharma et al. 2019a). The relief features of these catchments enable quick (delayed) disposal of the runoff generated in response to the rainfall event, leading to a steep (flat) recession of the flow hydrographs in the Burhanpur (Purna) watershed, leading to high (low) runoff coefficient. The higher CN of the Burhanpur watershed represents the high runoff potential of the catchment with relatively lower evapotranspiration. The vice versa holds equally good for the Purna watershed. Therefore, the calibrated parameters of the two watersheds of the UTRB represent their physiographic heterogeneity. The hydrological sensitivity of these parameters can be attributed to their (Burhanpur vis-à-vis Purna watersheds) subhumid vis-à-vis semi-arid climatology, narrow-steep-rocky valley vis-à-vis wide-plain-alluvium surface, and forest-dominated land cover vis-à-vis agricultural land use, respectively. For instance, the ALPHA_BF, representing the baseflow parameter, is high for the Burhanpur watershed. At the same time, GW_DELAY is lower. On the contrary, the values of ALPHA_BF are lower, and GW_DELAY is higher in the Purna watershed. The deep percolation represented by RCHRG_DP is increased in the Burhanpur watershed compared to the Purna watershed due to forest cover in the former watershed. These contrasting parameters agree with the contrasting physiography of these two catchments.

The surface runoff and water yield in the Purna (Burhanpur) watershed is 12.8% (17.7%) and 16.0% (37.3%) of the annual rainfall. It shows that the Burhanpur watershed has a significant contribution from the baseflow, which was evidenced through the flow duration curve (Figure 4(c)). The surface runoff and water yield in the sub-watersheds are directly proportional to the areal rainfall (Figure 6). The Pearson's correlation coefficient between areal rainfall and surface runoff (water yield) is 0.8 (0.9). The annual average streamflow potential of the Burhanpur watershed (expressed as m³/km²/mm of rainfall) is 568.7, 4.2 times higher than the streamflow potential of the Purna watershed. The mean annual water yield depth and its CV in the Purna (Burhanpur) watershed are 127.7 mm (357.2 mm) and 36.4% (57.6%), respectively. The high CV value in the Burhanpur watershed directly influences its diverse topography from the origin of the river to the terminal reservoir. The watershed in the Satpura ranges' foothills has more rainfall than on the foothills of the Gwaligarh hills. It is due to their presence on the windward side where moisture-laden southwest monsoon trade winds strike off.

The outcomes of the present study would have direct relevance to field applications. The IHM of the upstream watershed and terminal Hatnur reservoir would help predict daily inflows (outflows) into (from) the Hatnur reservoir for the forecasted rainfall. The anticipated inflows into the reservoir would be very useful for supplying the water into the canal for fulfilling the domestic and irrigation requirements of the command area in the MTRB. Also, the releases from the Hatnur reservoir would be useful for predicting inflows into the downstream Ukai reservoir. Knowing the inflows into the Ukai reservoir would help the dam authority to have planned releases to prevent excessive flooding in the downstream regions of the LTRB, especially Surat City. It is worth noting that Surat city is one of the fastest-growing cities in the world, and the 2006 flood alone led to the loss of 210 billion Rupees. The developed integrated model also provides spatial and temporal distributions of rainfall based on existing watershed characteristics. This would help plan watershed management practices and their impacts on the water balance components. Such information would alleviate the water stress conditions in the region and attain sustainable development goals. Thus, the present study brings invaluable insight into the current understanding of the hydrological processes with contrasting watershed characteristics.

The hydrological processes in the UTRB, having two physio-climatically heterogeneous watersheds, i.e., Purna and Burhanpur, and their terminal Hatnur reservoir, have been modelled in an integrated manner using the SWAT to predict the daily inflow (outflow) into (from) the Hatnur reservoir. The hydrologic modelling of the headwater catchment of the UTRB, i.e., the Burhanpur watershed, was performed by integrating the Hatnur reservoir and outflows from the Purna watershed. Extensive multisite and multivariable (streamflow, inflow/outflow) calibration (1998–2007) and validation (2008–2013) approaches have been employed to investigate the hydrological influence of model parameters on the watershed processes, including their sensitivity and uncertainty analyses. The key conclusions drawn from the present study are as follows:

• The areal rainfall over the Burhanpur and Purna watersheds deviates nominally, except for the extremes; however, their flow duration curves varied significantly in high-, medium-, and low-flow regimes. The sizable change in the model output (i.e., streamflow) for nearly similar inputs (i.e., rainfall and other meteorological variables) highlighted the substantial influence of the contrasting nature of these two watersheds in terms of topography, geology, soil, and LULC.

• The parametric sensitivity assessment for the contrasting watersheds reflected the physiographical heterogeneity of the two watersheds. It is also revealed that hydrological processes in the Burhanpur and Purna watersheds are primarily affected by RCHRG_DP.gw (deep aquifer percolation fraction), GW_DEALY.gw (groundwater delay), CN2.mgt (curve number), SLSOIL.hru (slope length for lateral subsurface flow), and ALPHA_BF.gw (baseflow alpha factor); and GWQMN.gw (the threshold depth of water in the shallow aquifer required for return flow to occur), GW_REVAP.gw (groundwater ‘revap’ coefficient), CH_K2.rte (effective hydraulic conductivity in the main channel), CN2.mgt, and RCHRG_DP.gw, respectively.

• The hydrological sensitivity of these parameters can be attributed to their (Burhanpur vis-à-vis Purna watersheds) subhumid vis-à-vis semi-arid climatology, narrow-steep-rocky valley vis-à-vis wide-plain-alluvium surface, and forest-dominated land cover vis-à-vis agricultural land use, respectively.

• The graphical and quantitative evaluation of the integrated SWAT model has shown satisfactory simulation capability (NSE and R² ≥ 0.71) under the multisite and multivariable calibration and validation framework for the prediction of daily inflow into and outflow from the Hatnur reservoir, including the hydrological response of Burhanpur and Purna watersheds.

• The Burhanpur watershed showed a large CV, compared to the Purna watershed, in the spatial distribution of water balance components over sub-watersheds due to its diverse topography from the origin of the river to the terminal reservoir.

Nonetheless, integrating sediment yield, reservoir sedimentation modelling, and details of minor hydraulic and detention structures with the existing model would make the simulation of hydrological processes more realistic. Including more rain gauges would alleviate the modelling of better-watershed dynamics of the Burhanpur watershed. These limitations can be considered as a future direction for the present study.

The authors acknowledge the financial support received from the Indian National Committee on Climate Change (INCCC) sponsored research project ‘Impact of Climate Change on Water Resources of Tapi Basin’ under the Department of Water Resources, River Development & Ganga Rejuvenation, Ministry of Jal Shakti, Government of India. The authors also appreciate the infrastructural support provided by the Centre of Excellence (CoE) on ‘Water Resources and Flood Management’, TEQIP-II, Ministry of Education, Government of India. The authors are also thankful to India Meteorological Department (IMD), Pune; Central Water Commission (CWC), Tapi Division-Surat, National Remote Sensing Centre (NRSC), Hyderabad; National Bureau of Soil Survey and Land Use Planning (NBSS&LUP), Nagpur; Tapi Irrigation Development Corporation (TIDC), Jalgaon; and USGS EarthExplorer portal for providing the required data for the present study. Acknowledgements are due to the esteemed editor and reviewers for providing valuable and thoughtful suggestions in improving the quality of this manuscript.

P.M. and L.K.G.: Conceptualization, methodology, model development and data analysis, writing-original draft; P.L.P., S.K., P.V.T., and R.G.: Conceptualization, data curation, software, supervision, writing-review and editing.

This work was supported by the Indian National Committee on Climate Change (INCCC), Department of Water Resources, River Development & Ganga Rejuvenation, Ministry of Jal Shakti, Government of India vide their letter no. 16/22/2016-R&D/3059-3076 dated November 7, 2016.

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

The authors declare there is no conflict.