Abstract

Climate change impact on flow regimes in the Gomti River basin, India was studied using the Soil and Water Assessment Tool (SWAT) driven by climate change scenarios generated from multiple general circulation model (GCM) projections. The SWAT-CUP (SWAT-Calibration and Uncertainty Programs) was used for calibration and validation of SWAT using multi-site data. Climate change scenarios were generated from multiple GCM projections using the hybrid-delta ensemble method. Calibration of SWAT using the nine most sensitive parameters showed that the model performed reasonably well with P-factor >0.7 and R-factor <1.0. The annual rainfall is projected to increase by 3.4–4.5, 4.7–10.0, and 5.0–18.0% during the 2020s, 2050s, and 2080s respectively under different Representative Concentration Pathways (RCPs). There is a decrease in rainfall during the winter season. The annual streamflow is projected to increase by 1–9, 1–22, and 2–38% during the 2020s, 2050s, and 2080s, respectively. However, winter and summer streamflow is projected to decrease. Magnitude and frequency of high flows is also projected to increase in the range of 3.6–27.3 and 12–87%, respectively under different RCPs. The results of this study will be helpful in developing suitable water management adaptation plans for the study basin.

INTRODUCTION

Global climate change caused by increasing concentrations of greenhouse gases as well as anthropogenic activities is likely to cause intensification of the global hydrological cycle (Arnell 1999). This may result in changes in water resource availability in almost all regions of the world (Intergovernmental Panel on Climate Change (IPCC) 2013). Climate change impacts on water resources have been attributed mainly to associated changes in the dominant climate variables of precipitation and temperature (Hansen et al. 2006; Islam et al. 2012a). For India, a warming of 1.5–4.3 °C for the 2080s (2071–2100), as compared to the baseline period of 1961–1990, has been projected under the different Representative Concentration Pathways (RCPs) (Chaturvedi et al. 2012). Similarly, precipitation is projected to increase over almost all regions of India except for a few regions in short-term projections (2030s). All-India annual precipitation is projected to increase by 6–14% under different RCPs during the 2080s compared with the 1961–1990 baseline. Such a change in temperature and rainfall would affect water availability for different sectors, particularly agriculture, with serious implications for livelihood security. Therefore, quantitative assessment of climate change impact on the spatial and temporal variability of water resources availability in different river basins is vital for understanding potential water resources problems and preparing basin specific adaptation plans.

Since hydrological conditions vary from region to region, the impact of climate change on regional water resources availability will vary from one river basin to another. Thus, for preparing local adaptation plans it is of utmost importance to understand the hydrological responses of a river basin (Bisht et al. 2018) in the backdrop of climate change. The Gomti River, an alluvial river of the Indo-Gangetic plains, is an important tributary of the River Ganges. The discharge of the Gomti River is mainly due to monsoon rainfall and is characterized by sluggish flow throughout the year except during the monsoon (rainy) season (Dutta et al. 2011). Based on the trend analysis of historical data (1982–2012), Abeysingha et al. (2015b) reported a significant increasing trend in annual streamflow at an upstream gauging station, and a gradually decreasing trend in annual and seasonal streamflow from the midstream to the downstream of the river. Thus, temporal and spatial variations in rainfall under the changing climate scenarios may further alter the flow condition and water resources availability in the Gomti River. Dutta et al. (2015) pointed out the need for a restoration plan for the Gomti River basin to improve poor water quality and poor water flow in the river.

Impacts of climate change on hydrology and water resources have been widely studied, mainly using water balance models coupled with general circulation model (GCM) projections. The Soil and Water Assessment Tool (SWAT) is one of the most widely used models for simulating basin hydrology and assessing the effect of land use and climate change at basin scale (Arnold et al. 1998). The SWAT model has been applied by several researchers to investigate the climate change impacts on hydrology and water resources availability in Indian River basins using different GCM projected climate change scenarios (e.g. Gosain et al. 2011; Mishra & Lilhare 2016; Abeysingha et al. 2017). Abeysingha et al. (2015a) calibrated and validated SWAT model for the Gomti River basin using the historical rainfall and temperature data for the period 1985–2010. The model was calibrated manually by matching the observed streamflow data for four gauging stations (Neemsar, Sultanpur, Jaunpur and Maighat), and calibration was performed sequentially from the uppermost sub-basin (Neemsar) to the downstream sub-basin (Maighat). The calibrated model was applied for climate change impacts assessment on streamflow, and wheat and rice production (Abeysingha et al. 2016, 2017) using the MIROC3.2 GCM climate change projection for A2 (high), A1B (medium) and B1 (low) emission scenarios.

Manual calibration of hydrological models is more subjective and time consuming and its success highly depends on the experience of the modeler and their knowledge of the study basin, detailed understanding of the model and its assumptions (Kannan et al. 2008). Auto-calibration greatly reduces the uncertainty compared to the manual calibration in SWAT modeling (Van Liew et al. 2005). Further, calibration of hydrological models using multi-site data from spatially distributed gauging stations has been found to improve the flow simulation by capturing the spatially distributed change (spatial heterogeneity and discontinuities) in the watershed (Niraula et al. 2015). Therefore, in the present study, we used the SWAT Calibration and Uncertainty Programs (SWAT-CUP) for calibration and validation of the SWAT model. Moreover, calibration and validation were carried out using the multi-site (i.e. four gauging stations) streamflow data and all the four gauging stations data were used simultaneously during the auto-calibration process.

Most of the studies on climate change impact on hydrology and water resources in the Indian River basins are based on selected GCM/RCM projections (e.g. Gosain et al. 2011; Raje et al. 2014; Abeysingha et al. 2017). To reduce the uncertainty associated with individual GCM projections, climate change projections from multiple GCMs and/or ensemble of multiple GCMs and emission scenarios is generally preferred (Christensen & Lettenmaier 2007; Islam et al. 2012b; Vandana et al. 2018). Further, climate change scenarios are continuously evolving over time. The IPCC Fifth Assessment Report (AR5) is based on RCPs, which are based on greenhouse gas concentration trajectories resulting from projections of radiative forcing. The Coupled Model Inter-comparison Project phase 5 (CMIP5) models include advances in parameterization of physical processes, representation of new physical processes, and increases in model resolution (IPCC 2013). Sonali et al. (2017) reported enhancement in skills of CMIP5 models as compared to CMIP3 models in simulating the seasonal cycle of temperatures. Keeping in view the above facts, an attempt has been made to study the climate change impact on streamflow over the Gomti River basin using the SWAT model driven by climate change scenarios generated from the IPCC Fifth Assessment Report's RCP based climate change projections. As the climate change is likely to impact extreme events, the effect of climate change on high (flood) and low flows has also been studied. Simulations were made for the projected climate change scenarios for four different RCPs and three different future periods of the 2020s (2010–2039), 2050s (2040–2069) and 2080s (2070–2099). The four RCPs, namely, RCP8.5, RCP6.0, RCP4.5, and RCP2.6 indicate radiative forcing of 8.5, 6.0, 4.5 and 2.6 W/m2, respectively, by the year 2100 (Van Vuuren et al. 2011). Simulation runs were also made for the baseline scenarios of no changes in rainfall and temperature, and the changes in the future periods were computed against the baseline scenarios.

METHODS

Study area

The Gomti River basin is a sub-basin of the Ganga River basin and is located in north India (Figure 1). The basin is located between latitudes of 25°23′13″ to 28°46′59″ N and longitudes of 79°57′ 34″ to 83°11′13″ E, and has a catchment area of 30,437 km2 (Dutta et al. 2011). The topography of the catchment is undulating with high terrain at the upstream end of the basin, and the elevation ranges from 58 to 238 m above mean sea level (Figure 2). The climate of the basin ranges from semi-arid to sub-humid tropical climate. The basin receives an annual rainfall varying from 850 to 1,100 mm with about 75% of the total annual rainfall occurring between June and September due to the south-west monsoon (Abeysingha et al. 2015b). The flow in the Gomti River is mainly influenced by the intensity and duration of the monsoon rainfall. Thus, any change in magnitude and intensity of rainfall under the projected climate change scenarios will have a reflective effect on streamflow hydrograph, frequency and magnitude of flood events, and water resources availability.

Figure 1

The Gomti River basin with sub-basin boundaries and its gauging stations, weather stations, IMD grid points.

Figure 1

The Gomti River basin with sub-basin boundaries and its gauging stations, weather stations, IMD grid points.

Figure 2

DEM, land cover and soil map of the Gomti River basin, India.

Figure 2

DEM, land cover and soil map of the Gomti River basin, India.

Data

Daily rainfall and air temperature data of 12 districts, covering the entire Gomti River basin, for the period 1982–2010 were obtained from the National Initiative on Climate Resilient Agriculture (NICRA) project website (available at www.nicra-icar.in/nicrarevised/ on a request basis only). Long-term (1972–2011) daily weather data (rainfall, temperature, relative humidity, wind speed etc.) for Lucknow, Sultanpur and Jaunpur stations were also obtained from the India Meteorological Department (IMD). These data were used as weather statistics for the weather generator (WGEN) model of SWAT. As there were some missing data during different periods for different locations in the above datasets, we used high resolution 0.25 × 0.25° IMD gridded rainfall (Pai et al. 2014) and 1.0 × 1.0° resolution (Srivastava et al. 2009) temperature data for continuous simulation of hydrological model for climate change impact assessment. The high resolution 0.25 × 0.25° gridded rainfall data were developed using the well-tested inverse distance weighted interpolation (IDW) scheme proposed by Shepard (1968), and quality controlled rainfall data from more than 6995 rain gauge stations spread over India (Pai et al. 2014). The high resolution gridded dataset captured the climatological and variability features of rainfall very well and also effectively captured the orographic rainfall in the Western Ghats and in Northeastern India (Pai et al. 2014). Similarly, high resolution 1.0 × 1.0° temperature data were developed using the temperature data of 395 quality controlled stations and a modified version of the Shepard's angular distance weighting algorithm (Srivastava et al. 2009). The gridded IMD dataset has been extensively used in climate related research and applications, such as hydrological modeling for impact assessment (Mishra & Lilhare 2016; Vandana et al. 2018), monsoon variability and trend assessment (Goswami et al. 2006; Bisht et al. 2018) and temperature change analysis (Sonali et al. 2017). As we used Coupled Model Inter-comparison Project phase 5 (CMIP5) bias corrected and spatially disaggregated (BCSD) monthly projections at 0.5 × 0.5° resolution, rainfall data for the study basin corresponding to the CMIP5 BCSD grid points were extracted from 0.25 × 0.25° resolution IMD rainfall dataset. The IMD gridded temperature data available at 1 × 1° resolution were regridded at 0.5 × 0.5° resolutions to match with the CMIP5 BCSD grid points. Streamflow data of four spatially distributed gauging stations, namely, Neemsar, Sultanpur, Jaunpur and Maighat, were collected from the Central Water Commission (CWC), Lucknow and Varanasi, India. As 10 days average streamflow data were available, monthly streamflow data (computed from 10 days average streamflow data) were used for calibration and validation of SWAT.

The Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) of 90 × 90 m resolution (Jarvis et al. 2008) was used for delineation of the basin, sub-basins and stream networks. For land use and land cover (LULC) map of the study basin, we used 56 × 56 m resolution International Water Management Institute (IWMI) LULC map derived from the satellite remote sensing (mainly from AVHRR 10-km (band 1, 2, 4), and GTOPO30 1-km satellite sensor) data (Thenkabail et al. 2009). The soil map of the Ganga river basin at 78 × 78 m spatial resolution, developed by the National Bureau of Soil Survey and Land Use Planning (NBSS&LUP), Nagpur, India was obtained from the www.gisserver.civil.iitd.ac.in/grbmp/ website (accessed 2 May 2015). The data on cropping pattern, irrigation and other management inputs on planting, harvesting were obtained from the available literature pertaining to the area (Hobbs et al. 1992; Gangwar & Singh 2011) and field survey of selected area of the basin.

In this study, the IPCC Fifth Assessment Report's climate change projections based on new emission scenarios called RCPs were used. BCSD monthly projections at 0.5 × 0.5° resolutions, from the World Climate Research Program's (WRCP's) Coupled Model Inter-comparison Project phase 5 (CMIP5) multi-model dataset for the period 1950–2099, were obtained from ftp://gdo-dcp.ucllnl.org/pub/dcp/archive/cmip5/global_mon. As more than 20 modeling groups participated in CMIP5, climate projections were available from a number of GCMs and the number of runs (realizations) per model also varied. Further, climate projections were available for different simulation periods (near term (up to 2035) and long term (up to 2100)) and variables (precipitation, Tmax, Tmin, Tav, etc.). We selected models/projections with long-term simulations (up to 2100) having precipitation, Tmax, Tmin data. Thus, we selected 51, 61, 34, and 64 projections (runs) for RCP2.6, RCP4.5, RCP6.0, and RCP8.5, respectively, for generating climate change scenarios (Table 1).

Table 1

GCM projections used for generating climate change scenarios

Modeling center or group, countryModel nameNumber of runsa for RCP
2.64.56.08.5
Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology (CSIRO-BOM), Australia ACCESS1,0   
Beijing Climate Center, China Meteorological Administration (BCC), China BCC-CSM1.1 
BCC- CSM1.1(M)   
College of Global Change and Earth System Science, Beijing Normal University (GCESS), China BNU-ESM  
Canadian Centre for Climate Modelling and Analysis (CCCma), Canada CANESM2  
National Center for Atmospheric Research, USA CCSM4 
Community Earth System Model contributors, USA CESM1-BGC   
CESM1-CAM5 
Centro Euro-Mediterraneo per I Cambiamenti Climatici (CMCC), Italy CMCC-CM   
Centre National de Recherches Météorologiques/Centre Européen de Recherche et Formation Avancée en Calcul Scientifique (CNRM-CERFACS), France CNRM-CM5   
Commonwealth Scientific and Industrial Research Organization and Queensland Climate Change Centre of Excellence (CSIRO-QCCCE), Australia CSIRO-MK3.6.0 10 10 10 10 
EC-EARTH consortium EC-EARTH  
LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences and CESS, Tsinghua University (LASG-CESS), China FGOALS-G2  
The First Institute of Oceanography, SOA, China FIO-ESM 
NOAA Geophysical Fluid Dynamics Laboratory (NOAA GFDL), USA GFDL-CM3 
GFDL-ESM2G 
GFDL-ESM2M 
NASA Goddard Institute for Space Studies (NASA GISS), USA GISS-E2-H-CC    
GISS-E2-R 
GISS-E2-R-CC    
National Institute of Meteorological Research/Korea Meteorological Administration (NIMR/KMA), South Korea HADGEM2-AO 
Met Office Hadley Centre (additional HadGEM2-ES realizations contributed by Instituto Nacional de Pesquisas Espaciais) (MOHC/ INPE), UK HADGEM2-ES 
Institute for Numerical Mathematics (INM), Russia INMCM4   
Institut Pierre-Simon Laplace (IPSL), France IPSL-CM5A-LR 
IPSL-CM5A-MR 
IPSL-CM5B-LR   
Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The University of Tokyo), and National Institute for Environmental Studies (MIROC), Japan MIROC-ESM 
MIROC5 
MIROC-ESM-CHEM 
Max-Planck-Institut für Meteorologie (Max Planck Institute for Meteorology) (MPI-M), Germany MPI-ESM-LR   
MPI-ESM-MR   
Meteorological Research Institute (MRI), Japan MRI-CGCM3   
Norwegian Climate Centre (NCC), Norway NORESM1-M 
Total models  24 30 17 31 
Total runs  51 61 34 64 
Modeling center or group, countryModel nameNumber of runsa for RCP
2.64.56.08.5
Commonwealth Scientific and Industrial Research Organization and Bureau of Meteorology (CSIRO-BOM), Australia ACCESS1,0   
Beijing Climate Center, China Meteorological Administration (BCC), China BCC-CSM1.1 
BCC- CSM1.1(M)   
College of Global Change and Earth System Science, Beijing Normal University (GCESS), China BNU-ESM  
Canadian Centre for Climate Modelling and Analysis (CCCma), Canada CANESM2  
National Center for Atmospheric Research, USA CCSM4 
Community Earth System Model contributors, USA CESM1-BGC   
CESM1-CAM5 
Centro Euro-Mediterraneo per I Cambiamenti Climatici (CMCC), Italy CMCC-CM   
Centre National de Recherches Météorologiques/Centre Européen de Recherche et Formation Avancée en Calcul Scientifique (CNRM-CERFACS), France CNRM-CM5   
Commonwealth Scientific and Industrial Research Organization and Queensland Climate Change Centre of Excellence (CSIRO-QCCCE), Australia CSIRO-MK3.6.0 10 10 10 10 
EC-EARTH consortium EC-EARTH  
LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences and CESS, Tsinghua University (LASG-CESS), China FGOALS-G2  
The First Institute of Oceanography, SOA, China FIO-ESM 
NOAA Geophysical Fluid Dynamics Laboratory (NOAA GFDL), USA GFDL-CM3 
GFDL-ESM2G 
GFDL-ESM2M 
NASA Goddard Institute for Space Studies (NASA GISS), USA GISS-E2-H-CC    
GISS-E2-R 
GISS-E2-R-CC    
National Institute of Meteorological Research/Korea Meteorological Administration (NIMR/KMA), South Korea HADGEM2-AO 
Met Office Hadley Centre (additional HadGEM2-ES realizations contributed by Instituto Nacional de Pesquisas Espaciais) (MOHC/ INPE), UK HADGEM2-ES 
Institute for Numerical Mathematics (INM), Russia INMCM4   
Institut Pierre-Simon Laplace (IPSL), France IPSL-CM5A-LR 
IPSL-CM5A-MR 
IPSL-CM5B-LR   
Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The University of Tokyo), and National Institute for Environmental Studies (MIROC), Japan MIROC-ESM 
MIROC5 
MIROC-ESM-CHEM 
Max-Planck-Institut für Meteorologie (Max Planck Institute for Meteorology) (MPI-M), Germany MPI-ESM-LR   
MPI-ESM-MR   
Meteorological Research Institute (MRI), Japan MRI-CGCM3   
Norwegian Climate Centre (NCC), Norway NORESM1-M 
Total models  24 30 17 31 
Total runs  51 61 34 64 

aIn CMIP5, ensemble members are identified by rNiMpL, whereas r, i, and p are realization, initialization, and perturbation physics, respectively. Run reflects N from a given CMIP5 projections rNiMpL identifier.

Soil and Water Assessment Tool

SWAT is a physically based, semi-distributed watershed model developed to predict the impact of land management practices and climatic change on water, sediment and agricultural chemical yields over long periods of time (Arnold et al. 1998; Neitsch et al. 2011). This model has been widely used for long-term simulations in agricultural watersheds. The major model components include hydrology, erosion/sedimentation, crop growth, nutrient, pesticide, agricultural management, channel and pond/reservoir routing, and weather generation (Arnold et al. 1998). The hydrologic components in SWAT include surface runoff, infiltration, evapotranspiration, lateral flows, tile drainage, percolation/deep seepage, shallow aquifer contribution to streamflow for a nearby stream (base flow), and recharge by seepage from surface water bodies (Neitsch et al. 2011). SWAT also has the automatic calibration and uncertainty analysis capability via the SWAT-Calibration and Uncertainty Programs (SWAT-CUP).

In order to simulate hydrologic processes in SWAT, the watershed is divided into sub-watersheds, which are further segregated into unique hydrological response units (HRUs). These HRUs represent the unique combination of land use, soil type, and slope to describe the basin's physical heterogeneity. The model calculations are performed on an HRU basis, and flow and water quality variables are routed from HRUs to sub-watersheds and subsequently to the watershed outlet (Neitsch et al. 2011). The SWAT model simulates the land portion of the hydrologic cycle based on a water mass balance. Soil water balance in each HRU is represented as (Arnold et al. 1998; Neitsch et al. 2011): 
formula
(1)
where SWt = final soil water content (mm water); SW0 = initial soil water content in day i (mm water); t is the time (days); Rday = amount of precipitation in day i (mm water); Qsurf = amount of surface runoff in day i (mm water); Ea = amount of evapotranspiration in day i (mm water); Wseep = amount of water entering the vadose zone from the soil profile in day i (mm water); and Qgw = amount of return flow in day i (mm water).

In SWAT, surface runoff can be estimated using either the USDA Soil Conservation Service curve number (SCS-CN) method or the Green–Ampt infiltration method. As water management affects the hydrologic water balance, SWAT includes several water management practices such as irrigation, tile drainage, impounded/depression area, loading from nonpoint sources, etc. Irrigation water can be applied either manually or using auto-application option. Auto-irrigation triggers irrigation events according to a water stress threshold. The detailed description of the SWAT model along with equations used for computation of different components of the basin hydrology is described in the theoretical documentation of SWAT (Neitsch et al. 2011). In the present study we used the latest version of SWAT (SWAT-2012.10_1.14) within the ArcGIS (ver.10.1) interface.

SWAT model setup

The Gomti River basin was divided into 21 sub-basins during the SWAT watershed delineation and sub-basin discretization process. It was further divided into 296 hydrological response units during HRU definition and analysis process. In this study, the USDA Soil Conservation Service curve number procedure and variable storage coefficient method was selected for estimation of runoff and channel routing, respectively. The Penman–Monteith method was selected for estimation of evapotranspiration. SWAT uses the Ritchie (1972) methodology for estimating actual evapotranspiration.

For SWAT model setup, we used soil properties from the NBSS&LUP soil map (Figure 2). Soils of the basin were predominantly alluvial, deep soil. Agriculture is the predominant land use in the basin (Thenkabail et al. 2009). Singh et al. (2011) reported rice-wheat followed by sugarcane, rice-pulses, and sugarcane-wheat as the major cropping pattern in the state of Uttar Pradesh in terms of area. Considering the above cropping pattern data, crops were assigned to different IWMI land use categories (Figure 2). Thus, in the model setup, land use category R-08 was assigned to irrigated rice (kharif) and wheat (rabi) cropping system. The R-02 category was assigned to rice (kharif) and pulses (rabi) and R-03 category was assigned to sugarcane (annual) crop. Land use types R-08, R-02, and R-03 occupied 59.58, 32.45, and 1.38% area of the basin, respectively. The potholes are recommended for paddy field simulation in the theoretical documentation for SWAT (Neitsch et al. 2011). Hence, HRUs with paddy fields were considered as potholes, and pothole parameters were incorporated for paddy cultivated areas. Considering the farm operation in the area, impound operation was given before planting, and release operation was given 7 days before harvesting of paddy for rice growing HRUs. The maximum volume of water stored in the pothole was set to 60 mm and the fraction of area that drains to the pothole was set as 0.8.

The water source for the simulation of the rice, pulses, and sugarcane was assigned as the canal water in HRUs where Sharda Sahayak canal system is located. SWAT recognized this source as an outside unlimited source. For the other areas, where the canal is not located, the source of irrigation for rice was considered to be a shallow aquifer located in the same sub-basin. The only sources of wheat irrigation are shallow aquifers, assumed to be located in the wheat growing sub-basins. The HRUs under sugarcane and pulse (lentil) crops were irrigated similar to that of the area under wheat. The auto-irrigation option was selected to irrigate the crops, and plant water demand was set as a water-stress identifier. The automatic fertilization option was selected for fertilizing the crops.

Calibration and validation of SWAT model

The SWAT model was parameterized using the observed streamflow data of four spatially distributed gauging stations. The area contributing to the Neemsar, Sultanpur, Jaunpur and Maighat gauging stations are referred to as Neemsar, Sultanpur, Jaunpur and Maighat sub-basins, respectively. The Maighat gauging station is located at the most downstream end of the basin, and thus represents the flow from the entire Gomti basin. The model was calibrated and validated by comparing the observed and simulated monthly streamflow for the period 1990–2000 and 2001–2005, respectively. A warm-up/spin-up period from 1979 to 1989 was used to minimize the effect of the user's estimate of initial values and to bring the hydrological processes in the basin at an equilibrium condition.

Calibration, validation, and sensitivity analysis were performed using the SWAT-CUP (version 5.1.6) (Abbaspour 2014), which utilizes the Sequential Uncertainty Fitting (SUFI-2) algorithm. The SUFI-2 algorithm is a widely used tool for combined calibration and uncertainty analysis of the SWAT model. It accounts for uncertainties due to uncertainty in driving variables (e.g. rainfall), conceptual model, parameters, and measured data (Abbaspour 2014). The algorithm aims to minimize the width of the uncertainty bound and encloses as many observations as possible. In SUFI-2, the simulation uncertainties are quantified by P-factor and R-factor. The P-factor is defined as the percentage of the observed data bracketed by the 95% prediction uncertainty (95PPU), and is calculated at the 2.5 and 97.5% levels of the cumulative distribution of an output variable obtained through Latin hypercube sampling (Abbaspour 2014). The R-factor is the ratio of the average thickness of the 95PPU band and the standard deviation of the measured data (Abbaspour 2014). Theoretically, the value of the P-factor ranges between 0 and 100%, and a value of 100% indicates exact correspondence of simulation and measured data. The R-factor values range between 0 and infinity, with zero indicating exact correspondence between the simulated and measured data. These two indices indicate strength of calibration considering the parameter uncertainty.

Moriasi et al. (2007) suggested application of a combination of graphical techniques and dimensionless and error index statistics for model evaluation. Hence, in this study the Nash–Sutcliffe efficiency (NSE), coefficient of determination (R2), percent bias (PBIAS), and RSR (RMSE-observations standard deviation ratio) statistics were used for assessing model performance during calibration and validation of SWAT. Different equations used for computing NSE, R2, PBIAS and RSR are described below.

Nash–Sutcliffe efficiency (NSE)

NSE indicates how well the plot of observed versus simulated data fits the 1:1 line, and is computed as (Moriasi et al. 2007): 
formula
(2)
where Qo = observed flow, Qs = simulated flow, and n = total number of observation.

The value of NSE ranges between –∞ and 1, and 1 is the optimal value.

Coefficient of determination (R2)

R2 represents the trend similarity between the observed and the simulated data, and is computed using the following equation: 
formula
(3)

The R2 value ranges from 0 to 1, with higher R2 values indicating better model performance.

Percent bias (PBIAS)

PBIAS measures the average tendency of the simulated data to be larger or smaller than the observed data, and is computed using the following equation (Moriasi et al. 2007): 
formula
(4)

The lower the value of PBIAS, the better is the model simulation. The optimal value of PBIAS is 0.0.

RMSE-observations standard deviation ratio (RSR)

RSR is computed as a ratio of root mean squared error and standard deviation of the observed data, using the following equation (Moriasi et al. 2007): 
formula
(5)

Lower RSR indicates lower RMSE, hence, better model simulation performance. The optimal value of RSR is 0.

Climate change scenario generation

The delta change method is the most commonly used method for generating future climate scenarios. This method does not consider variability or change in time series behavior in the future. The hybrid-delta method, on the other hand, considers inter-annual variability for each month (Hamlet et al. 2010; Islam et al. 2012b; Tohver et al. 2014). This method applies a different scaling factor to each month of the historic time series based on where it falls in the probability distribution of monthly values (Dickerson-Lange & Mitchell 2014; Vandana et al. 2018). In this method, BCSD monthly GCM data (historical as well as future) were first disaggregated into individual calendar months. The cumulative distribution functions (CDFs) were then developed for each month for historical and future time periods (2020s, 2050s, and 2080s). For creating an ensemble of multiple GCMs/runs, data from multiple GCMs/runs were used for developing historical and future CDFs. Similarly, the CDFs for the observed time series data (1976–2005) were also developed, and the non-exceedance probability for each of the observed data were computed. Thus, for n number of year, n values of non-exceedance probability for each month were obtained. Instead of bias correcting GCM data to match observation, Quantile mapping (Wood et al. 2002) was applied here to re-map the observations onto the bias-corrected GCM data (historical and future CDF) for each month to obtain the historic and future GCM projected data (rainfall and temperature) corresponding to the non-exceedance probability of observed data. The difference between the future and historical temperature values was then computed to obtain the change factor. Thus, for each month n different change factors for n number of years were computed. This process is repeated for all the twelve months and for all the grid points. A step-by-step procedure for generating climate change scenarios using the hybrid-delta method is described in Tohver et al. (2014). In this study, the hybrid-delta method was used for generation of climate change scenarios from multiple GCM projections for four different RCPs.

RESULTS AND DISCUSSION

SWAT calibration and validation

We used observed streamflow data of four spatially distributed gauging stations for calibration and validation of the SWAT model. Sensitivity analysis was first performed using the SWAT-CUP to identify the most sensitive model parameters. Based on the literature survey and land use pattern of the basin, a total of 17 model parameters were considered for the sensitivity analysis (Table 2). Using the t-stat and p-value, sensitive parameters were identified (Table 2) for further calibration and validation of the model. The t-stat provides a measure of sensitivity (larger absolute values indicate more sensitivity) and the p-value determines the significance of sensitivity (smaller value suggest a higher level of significance). A comparison of sensitivity analysis results of this study with that of other river basins of India (Singh et al. 2013; Uniyal et al. 2015) showed that ALPHA_BF, CH_N2, CN2, ESCO, GWQMIN, SOL_AWC, SOL_K are the common sensitive parameters in all the three river basins (Tungabhadra, Baitarani and Gomti), though their sensitivities (sensitivity rank) varied from one river basin to another. For example, CN2 was found to be the most sensitive parameter for Tungabhadra (Singh et al. 2013), but for Baitarani (Uniyal et al. 2015) and Gomti River (present study) basins it was found to be less sensitive with ranks of 6 and 9, respectively. The curve number is a function of antecedent moisture condition. The low sensitivity of CN2 could be attributed to updating the CN2 value for each day of simulation based on available water content in the soil profile. Thus, water balance components will not be greatly affected due to a change in the initial CN2 value (Kannan et al. 2007; Shen et al. 2012). Based on the results of sensitivity analysis nine most sensitive parameters (Table 2), including CN2, were considered for calibration and validation of SWAT.

Table 2

Sensitivity analysis results with ranking of SWAT parameters for the Gomti River basin

ParameterDescriptiont-statap-valuebSensitivity rank
EPCO Plant uptake compensation factor 35.96 0.00 
ALPHA_BF Baseflow alpha factor (1/days) −6.81 0.00 
SURLAG Surface runoff lag coefficient −2.59 0.01 
CH_N2 Manning's n value for the main channel 2.30 0.02 
SOL_AWC (1,2) Available water capacity of the first two soil layers (mm/mm) 2.28 0.02 
GWQMN Threshold depth of water in the shallow aquifer for return flow to occur (mm H2O) 2.07 0.04 
SOL_K(1,2) Saturated hydraulic conductivity of first two soil layers (mm/h) 1.56 0.12 
ESCO Soil evaporation compensation factor 1.52 0.13 
CN2 Initial SCS runoff curve number for moisture condition II −1.33 0.14 
CH_K2 Channel effective hydraulic conductivity (mm/h) 0.83 0.41 10 
RCHRG_DP Deep aquifer percolation fraction 0.81 0.42 11 
CANMX Maximum canopy storage (mm) −0.55 0.58 12 
POT_VOLX Maximum volume of water stored in the pothole (mm) −0.50 0.61 13 
OV_N Manning's ‘n’ value for overland flow 0.43 0.67 14 
SOL_BD(1,2) Moist bulk density (Mg/m30.19 0.85 15 
POT_FR Fraction of HRU area that drains into the pothole 0.15 0.88 16 
GW_DELAY Groundwater delay (day) −0.06 0.95 17 
ParameterDescriptiont-statap-valuebSensitivity rank
EPCO Plant uptake compensation factor 35.96 0.00 
ALPHA_BF Baseflow alpha factor (1/days) −6.81 0.00 
SURLAG Surface runoff lag coefficient −2.59 0.01 
CH_N2 Manning's n value for the main channel 2.30 0.02 
SOL_AWC (1,2) Available water capacity of the first two soil layers (mm/mm) 2.28 0.02 
GWQMN Threshold depth of water in the shallow aquifer for return flow to occur (mm H2O) 2.07 0.04 
SOL_K(1,2) Saturated hydraulic conductivity of first two soil layers (mm/h) 1.56 0.12 
ESCO Soil evaporation compensation factor 1.52 0.13 
CN2 Initial SCS runoff curve number for moisture condition II −1.33 0.14 
CH_K2 Channel effective hydraulic conductivity (mm/h) 0.83 0.41 10 
RCHRG_DP Deep aquifer percolation fraction 0.81 0.42 11 
CANMX Maximum canopy storage (mm) −0.55 0.58 12 
POT_VOLX Maximum volume of water stored in the pothole (mm) −0.50 0.61 13 
OV_N Manning's ‘n’ value for overland flow 0.43 0.67 14 
SOL_BD(1,2) Moist bulk density (Mg/m30.19 0.85 15 
POT_FR Fraction of HRU area that drains into the pothole 0.15 0.88 16 
GW_DELAY Groundwater delay (day) −0.06 0.95 17 

at-stat indicates parameter sensitivity (larger t-value indicates more sensitive parameter).

bp-value indicates the significance of the t-value (smaller p-value indicates less chance of a parameter being accidentally assigned as sensitive).

Multi-site calibration and validation of the SWAT model resulted in ‘P-factor’ in the range of 0.67–0.87 and 0.66–0.74 during calibration and validation periods, respectively (Table 3). Thus, the percentage of observed streamflow data being bracketed by 95PPU varied in the range of 67–87 and 66–74% during calibration and validation periods, respectively, depending upon the sub-basin. For the Gomti River basin (represented by Maighat gauging station), the percentage of observed streamflow data being bracketed by 95PPU is 87 and 74% during calibration and validation periods, respectively. The ‘R-factor’ varied from 0.83 to 1.06 and 0.72 to 0.95 during calibration and validation periods, respectively. The R-factor for the Gomti River basin was 0.84 and 0.95 during calibration and validation periods, respectively. For discharge simulation, Abbaspour (2014) recommended the ‘P-factor’ value of >0.7 and the R-factor value of around 1.0 depending on the situation. Thus, the uncertainty associated with this multi-site calibration and validation of the SWAT model is within an acceptable range.

Table 3

SWAT model calibration and validation statistics

Sub-basinsCalibration
Validation
P-factorR-factorR2NSEPBIASRSRP-factorR-factorR2NSEPBIASRSR
Neemsar 0.67 1.06 0.73 0.61 −19.1 0.62 0.67 0.72 0.75 0.72 −10.0 0.53 
Sultanpur 0.72 0.83 0.68 0.55 27.8 0.68 0.70 0.82 0.63 0.60 10.6 0.63 
Jaunpur 0.76 0.83 0.68 0.61 25.2 0.63 0.66 0.77 0.57 0.53 17.6 0.69 
Maighat 0.87 0.84 0.73 0.68 20.1 0.56 0.74 0.95 0.49 0.35 10.5 0.78 
Sub-basinsCalibration
Validation
P-factorR-factorR2NSEPBIASRSRP-factorR-factorR2NSEPBIASRSR
Neemsar 0.67 1.06 0.73 0.61 −19.1 0.62 0.67 0.72 0.75 0.72 −10.0 0.53 
Sultanpur 0.72 0.83 0.68 0.55 27.8 0.68 0.70 0.82 0.63 0.60 10.6 0.63 
Jaunpur 0.76 0.83 0.68 0.61 25.2 0.63 0.66 0.77 0.57 0.53 17.6 0.69 
Maighat 0.87 0.84 0.73 0.68 20.1 0.56 0.74 0.95 0.49 0.35 10.5 0.78 

The NSE and R2 values varied in the range of 0.55–0.68 and 0.68–0.73, respectively, during the calibration period (Table 3). During the validation period, the NSE and R2 values ranged from 0.35–0.72 and 0.49–0.75, respectively. The PBIAS for different sub-basins varied from −19.1 to 27.8 during the calibration period and it varied from −10.0 to 17.6 during the validation period. The RSR varied from 0.56 to 0.68 and 0.53 to 0.78 during the calibration and validation periods, respectively. As per the model performance rating suggested by Moriasi et al. (2007), model simulation performance can be judged as ‘satisfactory’ if NSE >0.50 and RSR ≤0.70, PBIAS <± 25. Based on these criteria, the model performance can be rated as ‘satisfactory’ during calibration periods for all the sub-basins. Though PBIAS >25 (unsatisfactory) for Sultanpur sub-basin, based on NSE and RSR values model performance it can be rated as ‘satisfactory’. During the validation period, the model performed satisfactorily for all the sub-basins except Maighat sub-basin with NSE <0.5 and RSR ≥0.7. Parajuli (2010) categorized model performance based on R2 for monthly streamflow as excellent (≥0.90), very good (0.75–0.89), good (0.50–0.74), fair (0.25–0.49), poor (0–0.24), and unsatisfactory (<0). According to these criteria, model performance can be rated as ‘good’ during the calibration period. During the validation period, model performance varied from ‘very good’ to ‘fair’. Simulation results depicted as hydrographs showed a similar trend between the observed and simulated hydrographs of mean monthly streamflow, but could not capture some of the peak flows (Figure 3). In general, there is good agreement between observed and simulated streamflow values during both calibration and validation periods.

Figure 3

SWAT model calibration and validation hydrograph of different sub-basins of the Gomti River basin, India.

Figure 3

SWAT model calibration and validation hydrograph of different sub-basins of the Gomti River basin, India.

For Maighat sub-basin, the NSE value during the validation period is not as good as the NSE during the calibration period, though 74% of observed streamflow data are being bracketed by 95PPU. Comparatively, poor performance of the model during the validation period may be due to the fact that SWAT could not simulate some of the peak discharges and also the low flows (Figure 4) reasonably well. As differences between the observed and predicted values are calculated as squared values, NSE leads to an overestimation of the model performance during peak flows and an underestimation during low flow conditions. Poor performance of the model could also be attributed to quality of observed streamflow data (Mishra & Lilhare 2016). As observed streamflow data for river basins with trans-boundary water sharing issues are restricted to public sharing, 10 days average streamflow data were available. We used monthly streamflow data (computed from 10 days average streamflow) for model calibration and validation, and the model could not capture some of the peak flows and low flows. Though the NSE value at the outlet of the Gomti basin was low, the NSE values for the upstream sub-basins were greater than 0.50. For the upper-most sub-basins NSE is 0.72. Natural streamflow in most of the river basins in India is affected due to the presence of different water management structures (dams, barrages and reservoirs), and the Gomti River is no exception. Mishra & Lilhare (2016) selected gauging stations located at the upstream region of the basin that are least affected by the presence of dams and reservoirs. They also reported lower NSE using SWAT for Cauvery (0.38 for calibration and 0.32 for validation), Krishna (0.34 for calibration) and Sabarmati (0.34 for calibration) river basin while studying hydrologic sensitivity of Indian sub-continental river basins to climate change. Similarly, Mango et al. (2011) reported an NSE value of <0.5 for both calibration (0.43) and validation (0.23) periods, and applied the calibrated SWAT model for assessing response of the Mara River basin to future land use and climate change scenarios. Niraula et al. (2015) reported that model calibration does not significantly affect the relative change in streamflow due to climate change, and it may not be necessary to spend time and money in model calibration if the objective is to analyze the relative change in streamflow due to climate change. As the aim of this study is to provide an assessment of climate change impact on flow regime with respect to baseline climate forcing, we have presented relative changes in annual and seasonal streamflow, high and low flow, etc. Therefore, the calibrated model can be applied for assessing relative changes in water balance components based on long-term simulations.

Figure 4

Scatter plot of observed and simulated monthly streamflow at Miaghat gauging station.

Figure 4

Scatter plot of observed and simulated monthly streamflow at Miaghat gauging station.

CLIMATE CHANGE SCENARIOS

Projected changes in temperature over the basin

As rising temperature due to global warming is likely to impact evapotranspiration demand and water resources availability in the basin, projected changes in maximum temperature (Tmax) and minimum temperature (Tmin) in the study basin were analyzed. As shown in Figure 5, there is an increase in maximum as well as minimum temperature in the basin during future periods, but the increase in Tmin is greater as compared to the increase in Tmax. The increase in the Tmax at different locations of the basin varied in the range of 0.8–1.0, 1.4–2.5, and 1.5–4.3 °C during the 2020s, 2050s and 2080s, whereas the increase in the Tmin varied in the range of 1.0–1.2, 1.6–3.4 and 1.6–4.8 °C during the 2020s, 2050s and 2080s, respectively. In general, Neemsar sub-basin (upstream) experienced greater warming than the other sub-basins as well as the entire Gomti basin.

Figure 5

Projected changes in annual maximum (Tmax) and minimum temperature (Tmin) during future periods at different sub-basins: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Figure 5

Projected changes in annual maximum (Tmax) and minimum temperature (Tmin) during future periods at different sub-basins: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Projected changes in rainfall over the basin

As shown in Figure 6, there is an increase in annual rainfall over the basin during all the three future periods under all the four climate change scenarios. The annual rainfall in different sub-basin is projected to increase by 3.4–4.5, 4.7–10.0, and 5.0–18.0% during the 2020s, 2050s, and 2080s, respectively. Though there is an increase in annual rainfall during all the three future periods, the seasonal analysis showed a decrease in rainfall during winter months (January–February) in most of the scenarios during the future periods (Figure 7). This decrease in winter season rainfall over the basin varied in the range of 1.9–4.6 and 0.4–8.8% during the 2050s and 2080s, respectively. During the 2020s, the basin experienced a decrease (2.3–3.5%) in winter rainfall under RCP2.6 and RCP4.5 scenarios, and an increase (0.4–2.9%) under the remaining two RCPs. The increase in rainfall is mostly during monsoon (JJAS) and post-monsoon (OND) seasons. However, the post-monsoon season experienced a greater increase than that of the monsoon season. During the monsoon season, the increase in rainfall varied from 2.7–4.6, 4.5–10.0, and 4.8–17.9% during the 2020s, 2050s, and 2080s, respectively under different RCPs. The increase in rainfall during post-monsoon season varied from 8.4–13.2, 14.5–21.0, and 15.6–42.1% during the 2020s, 2050s, and 2080s, respectively under different RCPs. Though there is an increase in annual rainfall, monsoon as well as post-monsoon rainfall, temporal and spatial variation in rainfall would affect seasonal water resource availability, crop water demand in the different sub-basins, and the rainfed winter crops in the basin. Increases in temperature, as large as 4.8 °C during the 2080s, may further aggravate the water scarcity situation, particularly during summer months.

Figure 6

Projected changes in annual rainfall during future periods at different sub-basins: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Figure 6

Projected changes in annual rainfall during future periods at different sub-basins: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Figure 7

Projected changes in seasonal rainfall over the basin under different climate change scenarios. JJAS: Jun to Sep; OND: Oct to Dec; JF: Jan to Feb; MAM: Mar to May.

Figure 7

Projected changes in seasonal rainfall over the basin under different climate change scenarios. JJAS: Jun to Sep; OND: Oct to Dec; JF: Jan to Feb; MAM: Mar to May.

Changes in streamflow

Simulation results showed an increase in annual streamflow in the basin at all four sub-basins with a maximum increase at Neemsar followed by Sultanpur, Jaunpur and Maighat (Figure 8). Changes in annual streamflow under different RCPs varied in the range of 1–9, 1–22 and 2–38% during the 2020s, 2050s, and 2080s, respectively. During the 2020s and 2050s, RCP6.0-based climate change projections resulted in a minimum change in annual streamflow. During the 2080s, the minimum and maximum increase in annual streamflow resulted from RCP2.6 and RCP8.5 based climate change scenarios, respectively. The increase in annual streamflow is mainly due to the increase in rainfall in the basin. Simulation results indicated comparatively higher changes at the upper most sub-basin (i.e. Neemsar sub-basin), and this may be due to the steeper terrain of the Neemsar sub-basin.

Figure 8

Changes in annual streamflow in different sub-basins under different climate change scenarios: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Figure 8

Changes in annual streamflow in different sub-basins under different climate change scenarios: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Seasonal analysis showed an increase in streamflow in the monsoon (June–September) as well as in the post-monsoon (October–December) seasons, and a decrease in streamflow in the winter (January–February) and summer (March–May) seasons in the Gomti River basin (Figure 9). A similar trend was found in all the sub-basins except Neemsar sub-basin (Table 4). At Neemsar sub-basin, which is located at the upstream end of the Gomti River basin, there is an increase in streamflow in all four seasons. Steeper terrain (121–255 m) and a comparatively smaller sub-basin with lower time of concentration may be attributed to transformation of more rainfall into runoff, and hence a greater increase in streamflow at Neemsar sub-basin. The contribution of base flow from shallow aquifer, due to seepage and irrigation return flow from the Sharda canal network, may also be attributed to an increase in streamflow at Neemsar sub-basin. An increase in streamflow during the monsoon season is likely to increase the risk of flooding in low-lying areas. A decrease in streamflow during winter months may result in over-extraction of groundwater for irrigating Rabi (winter) crops. Further, a decrease in streamflow during the winter season may also increase the effect of pollution in the already polluted Gomti river due to lower dilution effects. Increased pollution may also be harmful to the riverine ecosystem.

Table 4

Changes in seasonal streamflow in different sub-basins of the Gomti River basin under different climate change scenarios

RCPsNeemsar
Sultanpur
Jaunpur
2020s2050s2080s2020s2050s2080s2020s2050s2080s
Monsoon season (JJAS) 
 RCP2.6 9.3 13.3 10.4 4.0 4.6 2.7 3.5 4.0 2.3 
 RCP4.5 7.7 20.2 25.5 2.7 7.8 10.8 2.3 8.9 10.2 
 RCP6.0 6.1 10.4 28.4 1.5 2.2 12.1 1.2 1.8 11.4 
 RCP8.5 8.0 24.0 41.2 2.4 8.5 17.6 2.1 7.8 16.4 
Post-monsoon (OND) 
 RCP2.6 9.5 14.7 11.6 5.8 8.5 6.0 5.5 8.2 5.7 
 RCP4.5 9.1 20.5 23.3 5.2 11.1 13.4 5.0 10.1 13.1 
 RCP6.0 8.1 9.6 25.5 4.3 4.9 14.1 4.1 4.6 13.8 
 RCP8.5 8.4 23.6 40.2 4.3 12.1 21.5 4.0 11.7 21.0 
Winter (JF) 
 RCP2.6 2.8 4.5 3.1 −1.8 −4.6 −5.9 −2.1 −4.5 −5.9 
 RCP4.5 2.4 7.9 10.7 −2.8 −5.3 −5.3 −3.0 −5.6 −4.9 
 RCP6.0 2.2 1.9 11.0 −3.0 −8.2 −6.9 −3.1 −8.0 −6.4 
 RCP8.5 2.0 8.7 15.0 −3.5 −7.9 −9.4 −3.5 −7.3 −8.5 
Summer (MAM) 
 RCP2.6 7.8 10.4 7.1 −4.1 −6.5 −8.2 −4.7 −7.5 −9.2 
 RCP4.5 7.4 13.8 14.8 −4.7 −7.1 −7.8 −5.4 −9.1 −9.0 
 RCP6.0 5.8 8.0 15.8 −5.2 −9.9 −8.3 −5.9 −10.9 −9.8 
 RCP8.5 7.0 13.9 19.5 −5.7 −9.4 −9.8 −6.3 −10.8 −11.8 
RCPsNeemsar
Sultanpur
Jaunpur
2020s2050s2080s2020s2050s2080s2020s2050s2080s
Monsoon season (JJAS) 
 RCP2.6 9.3 13.3 10.4 4.0 4.6 2.7 3.5 4.0 2.3 
 RCP4.5 7.7 20.2 25.5 2.7 7.8 10.8 2.3 8.9 10.2 
 RCP6.0 6.1 10.4 28.4 1.5 2.2 12.1 1.2 1.8 11.4 
 RCP8.5 8.0 24.0 41.2 2.4 8.5 17.6 2.1 7.8 16.4 
Post-monsoon (OND) 
 RCP2.6 9.5 14.7 11.6 5.8 8.5 6.0 5.5 8.2 5.7 
 RCP4.5 9.1 20.5 23.3 5.2 11.1 13.4 5.0 10.1 13.1 
 RCP6.0 8.1 9.6 25.5 4.3 4.9 14.1 4.1 4.6 13.8 
 RCP8.5 8.4 23.6 40.2 4.3 12.1 21.5 4.0 11.7 21.0 
Winter (JF) 
 RCP2.6 2.8 4.5 3.1 −1.8 −4.6 −5.9 −2.1 −4.5 −5.9 
 RCP4.5 2.4 7.9 10.7 −2.8 −5.3 −5.3 −3.0 −5.6 −4.9 
 RCP6.0 2.2 1.9 11.0 −3.0 −8.2 −6.9 −3.1 −8.0 −6.4 
 RCP8.5 2.0 8.7 15.0 −3.5 −7.9 −9.4 −3.5 −7.3 −8.5 
Summer (MAM) 
 RCP2.6 7.8 10.4 7.1 −4.1 −6.5 −8.2 −4.7 −7.5 −9.2 
 RCP4.5 7.4 13.8 14.8 −4.7 −7.1 −7.8 −5.4 −9.1 −9.0 
 RCP6.0 5.8 8.0 15.8 −5.2 −9.9 −8.3 −5.9 −10.9 −9.8 
 RCP8.5 7.0 13.9 19.5 −5.7 −9.4 −9.8 −6.3 −10.8 −11.8 

JJAS: Jun–Sep; OND: Oct–Dec; JF: Jan and Feb; MAM: Mar–May.

Figure 9

Changes in seasonal streamflow in the Gomti River basin under different climate change scenarios: JJAS: Jun to Sep; OND: Oct to Dec; JF: Jan-Feb; MAM: Mar to May.

Figure 9

Changes in seasonal streamflow in the Gomti River basin under different climate change scenarios: JJAS: Jun to Sep; OND: Oct to Dec; JF: Jan-Feb; MAM: Mar to May.

Impacts of climate change on surface water resources

In this study, we considered surface water resources (SWR) as the water flow contribution from sub-basins to generate streamflow. The water yield (WYLD) in the SWAT model is defined as the summation of the surface water flow (Qsurf), lateral flow contribution to streamflow (Qlat), and the water that returns to the stream from the shallow aquifer, also known as groundwater contribution (Qgw), minus the total loss of water from the tributary channels (Arnold et al. 2012). The loss through tributary channel is transmission loss through the bed which finally reaches the shallow aquifer as recharge. Since there were no considerable surface water reservoirs in the basin, we considered WYLD in each sub-basin as SWRs.

Similar to annual streamflow, there is an increase in annual water yield in the basin and the increase in annual water yield varied from 0.9 to 8.7, 0.8 to 22.2, and 1.8 to 38.8% during the 2020s, 2050s, and 2080s, respectively (Figure 10). These are very much in agreement with the increase in annual streamflow. This increase in annual water yield as well as streamflow is mainly due to the increase in annual rainfall in the basin. The spatial variation in SWRs as well as streamflow in the basin suggests the need for development of location specific adaptation strategies, particularly for crop planning.

Figure 10

Change in annual water yield in different sub-basins under different climate change scenarios: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Figure 10

Change in annual water yield in different sub-basins under different climate change scenarios: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Flow-duration curve

A flow-duration curve (FDC) describes the relationship between the frequency and magnitude of streamflow for a particular river basin (Vogel & Fennessey 1994), and depicts the complete range of river discharges from high flow (flood events) to low flows. In order to assess climate change impact on high (Q5) and low (Q95) flows, flow-duration curves were developed and Q5 and Q95 values were estimated to evaluate changes in high and low flow characteristics, respectively. For evaluating changes in the frequency of high and low flows, thresholds for high (Q5) and low flows (Q95) were computed from the baseline simulation. The numbers of daily streamflow events above (Q5) or below (Q95) the baseline values were then calculated to find high and low flow frequencies (Gellens & Roulin 1998).

Simulation results showed an increase in high flows (Q5) in the basin during all the three future periods under all the four RCP based climate change scenarios (Figure 11). The increase in high flows, as compared to baseline, varied in the ranges of 6.8–44.7, 2.9–25.9, 3.2–25.9, and 3.6–27.3%, at Neemsar, Sultanpur, Jaunpur and Maighat sub-basins, respectively. This increase in high flow magnitude is the maximum during the 2080s under RCP8.5 scenarios. Further, the increase in high flow is greater at the upstream sub-basin (Neemsar) as compared to the entire Gomti River basin. Based on the threshold baseline Q5 values of 115.2, 485.7, 598.9, and 921.1 m3/s at Neemsar, Sultanpur, Jaunpur, and Maighat sub-basins, respectively, the number of days exceeding these threshold values are computed as 493 days (spread over 30 years) for all the four sub-basins. Using this baseline value, changes in number of days exceeding this threshold value were found to vary in the range of 11–78, 10–97, 14–93, and 12–87% at Neemsar, Sultanpur, Jaunpur, and Maighat sub-basins, respectively. This increase in the number of high (flood) flow days in the basin under future climate change scenarios indicates a need for suitable adaptation measures, both engineering and agronomic measures, to avoid losses and damage to life as well as agricultural crops.

Figure 11

Changes in high (Q5) and low (Q95) flows and its frequencies in different sub-basins under different climate change scenarios: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Figure 11

Changes in high (Q5) and low (Q95) flows and its frequencies in different sub-basins under different climate change scenarios: (a) Neemsar, (b) Sultanpur, (c) Jaunpur, and (d) Maighat.

Changes in low flow within the basin, computed using Q95 as an index, remained within ±5% at all the sub-basins (Figure 11), except at Neemsar sub-basin. Neemsar sub-basin is likely to experience an increase (0.6–32.3%) in low flows during all the three future periods under all the four climate change scenarios. Changes in low flows varied in the range of −0.9 to 2.2, −1.2 to 1.3, and −3.2 to 1.8% at Sultanpur, Jaunpur, and Maighat sub-basins, respectively. In general, there is a decrease in low flow under RCP6.0 and RCP8.5 based climate change scenarios. There is a decrease in low flow during the 2050s as well as the 2080s over the Gomti basin. During the 2020s, there is an increase in low flow in the Gomti basin but this increase in low flow is less than 3%. Similar to high flows, the number of days with less than baseline Q95 was estimated to find the changes in low flow days. Changes in low flow days varied in the range of −12 to 4, −7 to 6, and −7 to 14% at Sultanpur, Jaunpur, and Maighat sub-basins, respectively. As there is an increase in low flow at Neemsar sub-basin, there is a decrease (1.2–25%) in number of low flow days in this sub-basin. Though there is little or no change in low flow (Q95) values, further analysis using Q90 as low flow index indicated a decrease in low flow (Q90) values at Maighat (4.3–8.6%) and Jaunpur (0.04–4.7%) and a corresponding increase in low flow days at Maighat (8–17%) and Jaunpur (0–11%) sub-basin. At Sultanpur there is a decrease (0.3–3.4%) in Q90 values in most scenarios, with the exception of RCP4.5 scenarios during the 2050s and 2080s and RCP8.5 scenarios during 2080s. However, at the Neemsar sub-basin, there is an increase in Q90 values and corresponding decrease in low flow days. This increase in low flow at the Neemsar sub-basin may be attributed to the contribution of base flow from the shallow aquifer zone of the alluvial plane. This result also corresponds with seasonal streamflow change at Neemsar, which is increasing in all the four seasons. These results indicate that changes in low flow in the Gomti River basin may not be of immediate concern. However, due to over-extraction of groundwater to meet the demands of increasing population and intensive agriculture, there is a reduction in base-flows and several tributaries of the Gomti river basin are becoming drier during the non-monsoon months (Dutta et al. 2015). Hence, while preparing a long-term adaptation plan, due attention is needed to maintain environmental flows and ecosystem services.

The results presented in this study indicate plausible changes in the streamflow in the Gomti River basin under the RCPs-based climate change projections. The RCPs provide a range of emission and concentrations that lead to radiative forcing levels of 2.6, 4.5, 6.0 and 8.5 W/m2 by the end of the century. The Gomti River is one of the important tributaries of the Ganga River system. The primary source of water in the Gomti River basin is precipitation during the monsoon season. As the basin is covered through a canal network, seepage and irrigation return flow contribute to groundwater recharge in the canal command areas. However, the basin is experiencing water stress conditions due to increasing human population along with increased agricultural activities. Our analysis showed an increase in both maximum and minimum temperatures in the basin. This increase in temperature will increase the evapotranspiration and subsequently increase the irrigation water demand. Rainfall during winter months is also projected to decrease. Thus, rising temperature and decreasing rainfall during the winter season will aggravate the water scarcity problem, affecting the Rabi (winter) crops in the area. The Ganga River basin is more sensitive to changes in precipitation than that of temperature, and increased precipitation along with warming will largely lead to an increase in surface runoff in all the sub-basins in Ganga River basin (Mishra & Lilhare 2016). Our analysis also showed an increase in rainfall and associated increase streamflow, especially during the monsoon and post-monsoon season. The upper and middle segment of the Gomti River is severely affected by flooding during the monsoon season due to drainage congestion. The increase in streamflow due to climate change would further increase the magnitude and frequency of flood, and subsequent damage to crops, livestock and other properties. As the basin faces the dual problem of excess water during the monsoon season and water scarcity during the non-monsoon period, water harvesting and storing excess water in ponds and reservoirs during the monsoon will mitigate the water scarcity problem during the non-monsoon season, reduce the flood risk during the monsoon season, and improve groundwater conditions (Sikka et al. 2018).

SWAT is a well-established model for long-term simulations in watersheds dominated by agricultural land uses. The model has also been applied to watersheds containing paddy fields. The SCS-CN method does not consider surface storage (ponded water), which is an important component to model runoff in the paddy field. Thus, the estimated runoff using the SCS-CN method may not be realistic as in the paddy field runoff is produced as a result of the overflow process. The impounded fields for rice production are hydrologically similar to the potholes. The pothole module allows water ponding and also has impound/release operation, irrigation operation and cropping operation, all of which are readily applicable to paddy field management. However, in the pothole module, the pothole is assumed to be cone shaped and thus surface area varies with the change in volume of impounded water. This can lead to underestimation of evapotranspiration as the paddy field has an almost constant surface area. In the pothole module seepage is calculated only when the soil water content of the profile is below the field capacity, but in the paddy field seepage occurs continuously when there is ponded water. Thus, this approach can lead to underestimation of the seepage volume. Though both the SCS-CN and pothole modules have limitations to completely represent the characteristics of paddy fields and address the complicated water management practice in the paddy field, many researchers have applied SWAT to watersheds containing paddy fields and most of their studies resulted in high NSE values at the watershed scale (Kang et al. 2006; Sakaguchi et al. 2014; Jeong et al. 2016).

Hydrological simulation studies are subject to uncertainty due to model structure and model parameterization. In this study, multi-site data from spatially distributed gauging stations were used for better model calibration and validation by capturing the spatially distributed changes in the basin. To reduce the uncertainties associated with the model parameter and structure, sensitivity analysis and uncertainty analysis were performed using the SUFI-2 algorithm of SWAT-CUP. However, in this study month streamflow data were used to calibrate and validate the SWAT hydrological model. With the availability of daily streamflow data, the model performance in simulating the basin hydrology could further be improved. The study assumes that the calibrated hydrologic model will perform well under the future climate change scenarios too. In this study, hydrological modeling was carried out assuming static land use and land cover (LULC) conditions during future periods. The RCPs cover a very wide range of land-use scenario projections (Van Vuuren et al. 2011). Several studies have shown that the effect of climate change was more significant in controlling the hydrological processes than that of the LULC change (Ki et al. 2013; Chawla & Mujumdar 2015). Chawla & Mujumdar (2015) reported that in the upstream reaches of the Ganga basin in India, referred to as the upper Ganga basin (UGB), climate change contributes more (>90%) to the simulated streamflow. However, changes in LULC may aggravate the problems of increased seasonal variability in streamflow caused by climate change (Ki et al. 2013). Hence, future research work should also consider the assessment of changes in streamflow caused by the separate and combined impacts of future climate and LULC changes in the basin for sustainable water resource planning and management.

SUMMARY AND CONCLUSIONS

Intensification of the global hydrological cycle due to global climate change is likely to change water resources availability in most regions of the world. This study investigates the climate change impact on flow regimes in the Gomti River basin, India. A simulation study was carried out using the Soil and Water Assessment Tool (SWAT) driven by RCP-based ensemble climate change scenarios generated from multiple GCM projections. Calibration, validation, and sensitivity analysis were performed using the Sequential Uncertainty Fitting (SUFI-2) algorithm of the SWAT Calibration and Uncertainty Programs (SWAT-CUP).

The SWAT model calibration and uncertainty analysis showed P-factor values of 0.87 and 0.74 and R-factor values of 0.84 and 0.95 during calibration and validation periods, respectively, which are well within the acceptable limit (P-factor >0.7 and R-factor <1.0). The Gomti River basin is likely to experience warmer climate with the maximum temperature (Tmax) and the minimum temperature (Tmin) increasing in the range of 1.5–4.3 and 1.6–4.8 °C during the 2080s, respectively, under different RCPs. The annual rainfall is projected to increase by 3.4–4.5, 4.7–10.0, and 5.0–18.0% during the 2020s, 2050s, and 2080s, respectively. However, there is a decrease in rainfall during the winter season in most of the scenarios, and this decrease in winter season rainfall over the basin varied in the range of 1.9–4.6 and 0.4–8.8% during the 2050s and 2080s, respectively. Simulation results projected an increase in annual streamflow in the range of 1–9, 1–22 and 2–38% during the 2020s, 2050s, and 2080s, respectively. Similar to annual streamflow, monsoon (June–September) as well as post-monsoon (October–December) season streamflow is likely to increase over the basin, but winter (January and February) and summer (March–May) season streamflow is projected to decrease under different RCPs. Though there is a decrease in streamflow during winter and summer seasons in the basin, there is an increase in winter and summer season streamflow in the uppermost sub-basin (i.e. Neemsar sub-basin). The magnitude and frequency of high (flood) flows in the Gomti River is likely to increase in the range of 3.6–27.3 and 12–87%, respectively, under future climate change scenarios. Changes in low flow (Q95) varied within ±5% over the basin. Simulation results indicated spatial and temporal variation in streamflow over the basin. These results provide valuable information regarding changes in the magnitude of streamflow under future climatic scenarios, and could be used for preparing location-specific adaptation measures, agricultural crop planning, and developing suitable water management strategies.

ACKNOWLEDGEMENTS

The authors would like to thank the Central Water Commission, Government of India for providing streamflow data, NICRA, CRIDA and IMD for providing meteorological data, and to the Department of Agriculture, Uttar Pradesh for providing crop production data. The authors are grateful to the anonymous reviewers for their constructive comments which helped to greatly improve the quality of the article.

REFERENCES

REFERENCES
Abbaspour
K. C.
2014
SWAT-CUP 2012: SWAT Calibration and Uncertainty Analysis Programs – A User Manual
.
Swiss Federal Institute of Aquatic Science and Technology
,
Eawag
.
Abeysingha
N. S.
Singh
M.
Sehgal
V. K.
Khanna
M.
Pathak
H.
2015a
Analysis of trends in streamflow and its linkages with rainfall and anthropogenic factors in Gomti River basin of North India
.
Theor. Appl. Climatol.
123
(
3–4
),
785
799
.
Abeysingha
N. S.
Singh
M.
Sehgal
V. K.
Khanna
M.
Pathak
H.
Jayakody
P.
Srinivasan
R.
2015b
Assessment of water yield and evapotranspiration over 1985 to 2010 in the Gomti River basin in India using the SWAT model
.
Curr. Sci.
108
(
12
),
2202
2212
.
Abeysingha
N. S.
Singh
M.
Islam
A.
Khanna
M.
Sehgal
V. K.
Pathak
H.
2017
Impacts of climate change on stream flow in the Gomti River Basin of India
.
J. Agric. Eng. ISAE
54
(
4
),
49
63
.
Arnell
N. W.
1999
Climate change and global water resources
.
Glob. Environ. Chang.
9
,
31
49
.
Arnold
J. G.
Srinivasan
R.
Muttiah
S.
Williams
J. R.
1998
Large-area hydrologic modeling and assessment: Part I. Model development
.
J. Am. Water Resour. Assoc.
34
(
1
),
73
89
.
Arnold
J. G.
Kiniry
J. R.
Srinivasan
R.
Williams
J. R.
Haney
E. B.
Neitsch
S. L.
2012
Soil and Water Assessment Tool: Input/Output Documentation
,
Version 2012, TR-439
.
Texas Water Resources Institute
,
Temple, TX
,
USA
.
Bisht
D. S.
Chatterjee
C.
Raghuwanshi
N. S.
Sridhar
V.
2018
Spatio-temporal trends of rainfall across Indian river basins
.
Theor. Appl. Climatol.
132
(
1–2
),
419
436
.
Chaturvedi
R. K.
Joshi
J.
Jayaraman
M.
Bala
G.
Ravindranath
N. H.
2012
Multi-model climate change projections for India under representative concentration pathways
.
Curr. Sci.
103
,
791
802
.
Chawla
I.
Mujumdar
P.
2015
Isolating the impacts of land use and climate change on streamflow
.
Hydrol. Earth Syst. Sci.
19
,
3633
3651
.
Dutta
V. D.
Srivastava
R. K.
Yunus
M.
Ahmed
S.
Pathak
V. V.
Rai
L.
Prasad
N.
2011
Restoration plan of Gomti River with designated best use classification of surface water quality based on river expedition, monitoring and quality assessment
.
Earth Sci. India
4
(
III
),
80
104
.
Dutta
V.
Kumar
R.
Sharma
U.
2015
Assessment of human-induced impacts on hydrological regime of Gomti river basin, India
.
Manage. Environ. Q. Int. J.
26
(
5
),
631
649
.
Gangwar
B.
Singh
A. K.
2011
Efficient Alternate Cropping System
.
Project Directorate for Farming Systems Research, Indian Council of Agricultural Research
.
Modipuram, Meerut-250 110 (UP),
India
, pp.
300
339
.
Gosain
A. K.
Rao
S.
Arora
A.
2011
Climate change impact assessment of water resources of India
.
Curr. Sci.
101
(
3
),
356
371
.
Goswami
B. N.
Venugopal
V.
Sengupta
D.
Madhusoodanan
M. S.
Xavier
P. K.
2006
Increasing trend of extreme rain events over India in a warming environment
.
Science
314
,
1442
1445
.
Hamlet
A. F.
Salathe
E. P.
Carrasco
P.
2010
Statistical Downscaling Techniques for Global Climate Model Simulations of Temperature and Precipitation with Application to Water Resources Planning Studies
. .
Hansen
J.
Sato
M.
Ruedy
R.
Lo
K.
Lea
D.
Medina-Elizade
M.
2006
Global temperature change–European Environment Agency (EEA)
.
Proc. Natl Acad. Sci. USA
103
,
14288
14293
.
Hobbs
P. R.
Hettel
G. P.
Singh
R. K.
Singh
R. P.
Harrington
L. W.
Singh
V. P.
Pillai
K. G.
1992
Rice-Wheat Cropping Systems in Faizabad District Pradesh, India: Exploratory Surveys of Farmers’ Practices and Problems and Needs for of Uttar Further Research
.
CIMMYT
,
Mexico, DF
.
IPCC
2013
Climate Change 2013: The Physical Science Basis
. In:
Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(
Stocker
T. F.
Qin
D.
Plattner
G.-K.
Tignor
M.
Allen
S. K.
Boschung
J.
Nauels
A.
Xia
Y.
Bex
V.
Midgley
P. M.
, eds).
Cambridge University Press
,
Cambridge, NY
,
USA
,
1535
pp.
Islam
A.
Sikka
A. K.
Saha
B.
Singh
A.
2012a
Streamflow response to climate change in the Brahmani River Basin, India
.
Water Resour. Manage.
26
,
1409
1424
.
Jarvis
A.
Reuter
H. I.
Nelson
A.
Guevara
E.
2008
Hollow Filled SRTM for the Globe
,
version 4, available from CGIAR-CSI SRTM 90 m database. Available from
:
http://srtm.csi.cgiar.org (accessed 20 June 20103
).
Kang
M. S.
Park
S. W.
Lee
J. J.
Yoo
K. H.
2006
Applying SWAT for TMDL programs to a small watershed containing rice paddy fields
.
Agric. Water Manage.
79
(
1
),
72
92
.
Moriasi
D. N.
Arnold
J. G.
Van Liew
M. W.
Bingner
R. L.
Harmel
R. D.
Veith
T. L.
2007
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations
.
Trans. ASABE
50
(
3
),
885
900
.
Neitsch
S. L.
Arnold
J. G.
Kiniry
J. R.
Williams
J. R.
2011
Soil and Water Assessment Tool Theoretical Documentation Version 2009
.
Grassland, Soil and Water Research Laboratory, Agricultural Research Service and Blackland Research Center, Texas Agricultural Experiment Station
,
College Station, Texas
.
Pai
D. S.
Sridhar
L.
Rajeevan
M.
Sreejith
O. P.
Satbhai
N. S.
Mukhopadhyay
B.
2014
Development of a new high spatial resolution (0.25° × 0.25°) long period (1901–2010) daily gridded rainfall data set over India and its comparison with existing data sets over the region
.
Mausam
65
,
1
18
.
Shepard
D.
1968
A two-dimensional interpolation function for irregularly spaced data
. In:
Proc. 1968 ACM Nat. Conf. ACM
,
New York, USA
, pp.
517
524
.
Sikka
A. K.
Islam
A.
Rao
K. V.
2018
Climate-smart land and water management for sustainable agriculture
.
Irrig. Drain.
67
,
72
81
.
Singh
N. J.
Kudrat
M.
Jain
K.
Pandey
K.
2011
Cropping pattern of Uttar Pradesh using IRS-P6 (AWiFS) data
.
Int. J. Remote Sens.
32
(
16
),
4511
4526
.
Singh
V.
Bankar
N. K.
Salunkhe
S. S.
Bera
A. K.
Sharma
J. R.
2013
Hydrological stream flow modelling on Tungabhadra catchment: parameterization and uncertainty analysis using SWAT CUP
.
Curr. Sci.
104
(
9
),
1187
1199
.
Srivastava
A. K.
Rajeevan
M.
Kshirsagar
S. R.
2009
Development of a high resolution daily gridded temperature data set (1969–2005) for the Indian region
.
Atmos. Sci. Lett.
10
,
249
254
.
Thenkabail
P. S.
Biradar
C. M.
Turral
H.
Noojipady
P.
Li
Y. J.
Vithanage
J.
Dheeravath
V.
Velpuri
M.
Schull
M.
Cai
X. L.
Dutta
R.
2009
An Irrigated Area Map of the World Derived from Remote Sensing
.
Research Report 105
.
International Water Management Institute
,
Colombo
,
Sri Lanka
,
65
p. .
Van Liew
M. W.
Arnold
J. G.
Bosch
D. D.
2005
Problems and potential of auto calibrating a hydrologic model
.
Trans. ASABE
48
(
3
),
1025
1040
.
Van Vuuren
D. P.
Edmonds
J.
Kainuma
M.
Riahi
K.
Thomson
A.
Hibbard
K.
Hurtt
G. C.
Kram
T.
Krey
V.
Lamarque
J. F.
Masui
T.
Meinshausen
M.
Nakicenovic
N.
Smith
S. J.
Rose
S. K.
2011
The representative concentration pathways: an overview
.
Clim. Change
109
,
5
31
.
Vandana
K.
Islam
A.
Parth Sarthi
P.
Sikka
A. K.
Kapil
H.
2018
Assessment of potential impact of climate change on streamflow: a case study of the Brahmani River Basin, India
.
J. Water Clim. Change
(
In press
).
doi:10.2166/wcc.2018.129
.
Vogel
R. M.
Fennessey
N. M.
1994
Flow-duration curves. I: New interpretation and confidence intervals
.
J. Water Resour. Plann. Manage.
120
(
4
),
485
504
.
Wood
A. W.
Maurer
E. P.
Kumar
A.
Lettenmaier
D. P.
2002
Long range experimental hydrologic forecasting for the eastern U.S
.
J. Geophys. Res.
107
(
D20
),
4429
.