1 Introduction

River catchment systems are complex owing to multiple influencing factors (e.g., increasing global population, land use change, climate, and other anthropogenic disturbances) and the alteration of hydrological processes is a major challenge in many basins across the world (Gyamfi et al. 2016). Hydrological processes are often affected by many complex factors (e.g., land use, soil properties, topography, and climate). Therefore, accurate prediction of the hydrological processes is essential to ameliorate the understanding of the impact of natural and anthropogenic disturbances on hydrological regimes. The hydrological models are efficient tools which assist to forecast water resource changes and to create water management strategies for sustainable development of human society (Viviroli et al. 2009). The functions of river catchment are significantly influenced by the climate change and overexploitation of water resources. The impact of climate change and adaptation measures is perceived as a major contemporary global concern (IPCC 2014). Global rise in temperature and unpredictability in precipitation both spatially and temporally, as well as changes in certainty of this variance is more likely to occur in coming century (Kharin et al. 2013). Change in climate may lead to change in hydrological processes in the river. Hydrological cycle processes are influenced by climatic changes in multiple ways, which have the characteristics of complex, non-linear and dynamic (Liu et al. 2014). The hydrological processes are very sensitive to climate and its variability. In particular, increase in temperature can accelerate evapotranspiration process leading to lower water availability, whereas increase in precipitation without prominent changes in evapotranspiration process may lead to increase water availability. Increase in precipitation in upstream regions would cause higher discharge, but precipitation in downstream regions may even decrease. The most pronounced changes in hydrological processes in the recent decades are observed across all parts of India. Furthermore, the potential impact of climate change is evident in the Himalayan region, where runoff is dominant because of steep hill slope (Immerzeel et al. 2012). In view of complexity of the climate variability, climate change related impact assessments studies are typically enforced in continental and regional extents. However, these are not the scales at which most hydrological analyses are implemented. For instance, a large-scale analysis ignores detailed hydrological processes and as a consequence, the localized impacts are not efficiently represented at such scale (Varis et al. 2004). The hydrological responses in response to climate, therefore, can be studied at small spatial scales. There are many hydrodynamics, such as, the variable infiltration capacity (VIC), Hydrologiska Byråns Vattenbalansavdelning (HBV) model, Water – Global Analysis and Prognosis (WaterGAP), Lund-Potsdam-Jena managed Land (LPJmL) model, and the soil and water assessment tool (SWAT), HEC-RAS, and MIKE FLOOD are available, which could effectively be utilized in climate change studies. Complicated models provide detailed hydrologic processes, but the number of inputs and parameters tend to result in uncertainty (Her and Chaubey 2015). Notably, the complexity of model softened cannot be used in data-scarce environment as it requires larger number of hydrological input parameters. Studies have evaluated GCM climate models and reported that a small rise in global temperature can have significant impacts on river discharge (Olli et al. 2004). However, climate model related uncertainty is larger in monsoon dominated basins like the Ganges in India which obscured the river discharge sensitivity (Hattermann et al. 2018). This is mostly owing to the uncertainty in projected precipitation trends over Ganges river basin. Most of the studies have discussed the uncertainty in projected river discharge and water availability under different climate scenarios (Giuntoli et al. 2015; Buda et al. 2017; Eisner et al. 2017; Vetter et al. 2017) and findings suggested that GCMs contribute the largest uncertainty. Besides GCM uncertainty, hydrological models related uncertainty is higher in regions where evapotranspiration have a substantial influence on the hydrological regime, for instance, during the dry season in the Ganges basin (Pechlivanidis et al. 2017; Hattermann et al. 2018). Dankers et al. (2014) analyzed nine global hydrological models and suggested that hydrological model related uncertainty can be predominated over GCM related uncertainty especially in regions where snow melt and soil freezing processes are prevailed.

Hydrological models make the hydrological systems simpler. They forecast hydrological changes and hydrological phenomena to compute the hydrological effects and formulate various policies on water resource management (Mirza 2003). By analyzing the mutual relationship between climate parameters and surface parameters, a mathematical relationship is established in order to represent various hydrological phenomena (Gosain et al. 2009). Different hydrological models, such as, VIC, MIKE-SHE and SWAT have been developed in the field of hydrological modelling to obtain precise assessments on various hydrological phenomena (Devi et al. 2015). SWAT has been widely used among the hydrologists in order to analyze the key hydrological processes and the long-term impacts of climate change on various hydrological parameters (Bouraoui et al. 2005). This model is extensively used to predict runoff and soil loss, water quality modelling, assessing effect of land use and climate change (Gassman et al. 2007). The SWAT was used to quantify the uncertainty of stream flow and sediment in the Daning and Jingchuan River Basin in China (Shen et al. 2012; Zhao et al. 2018). The SWAT model has been extensively used in various river basins (e.g., Tapi Basin in northern part of the Deccan Plateau, Tungabhadra catchment in the Western Ghats, Karnataka, Tons river basin in central India) of India for hydrological modelling and water resource management (Singh et al. 2013; Shivhare et al. 2014; Sisay et al. 2017; Kumar et al. 2017).

The Manipur River basin is the largest river basin in Manipur valley and plays a vital role in socio-economic condition of this region. In last few decades, there has been a significant change in climatic condition due to deforestation, rapid urbanization and various other natural and anthropogenic factors. Manipur River basin is frequently affected by floods, landslides and other natural hazards in the recent years. Future climate change will bring new challenges to water resources management and sustainable utilization of resources. There are only a few known studies about the hydrological regime of Manipur River basin and none of these studies have addressed climate change impact assessment on hydrological process of the basin. River discharge contributing to Loktak Lake catchment was simulated using Hybrid SHE-SWAT model (Eliza et al. 2018) and emphasized the use of support vector machine (SVM) and SHE-SWAT model in estimating the river discharge in unguaged river basin. Impact of prescribed global warming on runoff from the headwater catchments and their implications for the water level regime of Loktak Lake was modelled using MIKE-SHE (Singh et al. 2010). This study suggested occurrence of predominantly higher water levels within Loktak Lake in the future, especially during the monsoon period. Higher lake water levels are, however, likely to lead to further habitat degradation within the internationally important Loktak Lake ecosystem as well as inundation of rural communities. This study is the first attempt to quantify the combined effects of future projected LULC change dynamics and future projected meteorological parameters on hydrological regime of Manipur River basin. A hydrological model like SWAT will help us to quantify the challenges because of future climate changes in the future. The objectives of this study are: (1) understanding the interrelationships between various complex hydrological factors in the river basin, (2) to investigate the uncertainty associated with the future climate prediction, and (3) to assess implication of impact of climate change on hydrological regime.

2 Study area

Manipur River basin is the largest river basin in the Manipur valley nestled in the Indo-Myanmar Himalayan ranges. Manipur river basin is located between 24° to 25°25′ North (latitude) and 93°36′ to 94°27′ East (longitude) (figure 1). Hydrologically, this river basin has been divided in nine sub-basins: Heirok, Nambul, Khuga, Iril, Imphal, Kongba, Thoubal, Western, and Sekmai. Each sub-basin comprises of hilly terrain along with plain regions, varying altitude and topographical slope. One of the important sub-catchments of the basin is Loktak Lake, a unique ecosystem in the state of Manipur (Ramsar Bureau 2016; LDA 2003). Distinctive feature of this lake is the presence of herbaceous floating biomass locally known as Phumdis (LDA 2003). The lake covers 287 km2 (LDA 2003) and has total catchment area of 5020 km2.

Figure 1
figure 1

Manipur River basin with river network, meteorological station and outlet point at Ithai.

Full size image

Rock types are largely Tertiary and Cretaceous sedimentary formations, mainly igneous intrusions and minor metamorphic (NBSS and LUP 2001). Soils found in the catchment area are mostly silty and clayey by texture (NBSS and LUP 2001). Thickness of the soil is more in the hilly terrain region under thick vegetation cover, but less elsewhere as it is subjected to erosion. Basin is characterized by tropical to semi-tropical type of climate. Basin receives an average rainfall of 1,350 mm. The annual temperature ranges between 12° and 31°C and relative humidity ranges from 51 to 81% (Directorate of Environment (GoM) 2013).

3 Materials and methods

3.1 Data acquisition and analysis

SWAT model requires digital elevation model (DEM), soil data, LULC data and meteorological data as data input to set up the model. Observed discharge data was used to perform model calibration and validation. LULC data was obtained by performing supervised classification on Landsat 8 OLI Level1 image. Data obtained from different sources have been enlisted in table 1. Model was set up using the above data sets for the period between 1999 and 2017. There is severe data scarcity in Manipur River basin, due to which there is unavailability of direct discharge data. Water level data was used to derive the discharge by using stage-discharge curve (figure 2). The monthly water level data measured from 2008 to 2017 at the outlet point (Ithai) was collected from the NHPC Power Station Loktak Project. In order to calibrate the model, discharge data was used.

Table 1 Input data and source.
Full size table
Figure 2
figure 2

(a) Stage-discharge curve for wet season and (b) stage-discharge curve for dry season.

Full size image

3.2 SWAT model

Soil and water assessment tool is a physical-based model with spatial explicit parameterization (Neitsch et al. 2011). Development of the SWAT model was started in 1990s (Neitsch et al. 2002). In SWAT basin is divided into sub-basin and sub-basin is further divided into hydrological response unit (HRUs) based on soil type and LULC. Hydrologic processes in SWAT are based on water balance equation (Arnold and Fohrer 2005).

$$ SW_{t} = SW_{0} + \sum\limits_{i = 0}^{n} {(R_{\text{day}} {-}Q_{\text{surf}} {-}E_{a} {-}W_{\text{seep}} - Q_{gw} )}, $$
(1)

where SWt is the final soil water content after t days (mm); SW0 is the initial soil water content on day i (mm); Rday is the precipitation amount on day i (mm); Qsurf is the surface runoff amount on day i (mm); Ea is the amount of evapotranspiration on day i (mm); Wseep is the amount of percolation and flow entering the vadose zone of the soil profile on day i (mm); Qgw is the amount of the return flow on day i (mm); and t is the time (days).

Manning’s equation is used for the computation of watershed concentration time (Neitsch et al. 2011) and for computing the PET Penman–Monteith method is used (Baymani and Han 2013). The detailed depictions of SWAT are given in manual (Neitsch et al. 2011).

3.3 Evaluation of performance of SWAT model

To quantify the reliability of model, it is necessary to evaluate its output results. There are several model performance statistical indices in order to evaluate model performance. Such outputs are acceptable whose model performance statistical indices fall within the permissible range (Moriasi et al. 2007). Statistical methods used in this study to evaluate the model performance are co-efficient of determinacy (R2), Nash–Sutcliffe efficiency (NSE) and Kling-Gupta efficiency (KGE). Details of these methods are available in Gupta et al. (2009), Nash and Sutcliffe (1970) and Abbaspour (2015).

3.4 General circulation model and RCP scenarios

The Intergovernmental Panel on Climate Change (IPCC) has defined a series of representative concentration pathways (RCPs) scenarios for future projection, based on the coupled model intercomparison project (CMIP5) (van Vuuren et al. 2011). In order to quantify the uncertainty, four GCM models HadCM3, ECHAM-MPI5, GFDL CM3 and MIROC ESM with two RCP scenarios (RCP 2.6 and RCP 8.5) were analyzed. In this study, CMIP5 datasets of four GCM was downloaded from portal, for two different time periods 2046–2064 and 2081–2099. By performing uncertainty analysis on different general circulation model (GCM), two best GCM models, i.e., ECHAM and HadCM3 for Manipur River basin were selected based on vintage and validity. Precipitation, minimum temperature and maximum temperature for the future time period under two different scenarios were downscaled using statistical downscaling method.

Future hydrological processes in the study catchment were simulated using the calibrated SWAT model, using inputs from the GCMs for 2050s and 2090s decades and the projected LULC. We used the output of the SWAT simulations forced by the observed data as a baseline condition for the future simulation forced by the GCM dataset.

4 Results

4.1 Land use and land cover change analysis (2017–2030)

The main objective of this study is to assess the combined impact of change in LULC dynamics and projected meteorological condition on hydrological regime of Manipur River basin. The future projected LULC dataset developed by Anand and Oinam (2020) (figure 3) was used as input data in this study.

Figure 3
figure 3

Land use land cover map for the year 2017 and 2030. Source: Anand and Oinam (2020).

Full size image

Future projected LULC map for 2030 was developed using land change modeler (LCM) in TerrSet from the past LULC map (figure 3). LULC for different time period was analyzed and change detection was performed. Various driving factors, environmental policies and government norms were taken into consideration for predicting LULC for the year 2030. Percentage change in different LULC classes is tabulated in table 2. It can be observed that area under forests is getting decreased by 1.46%. Decrease in the area under forest is mainly concentrated around the foothills mainly because of conversion of area under forest to agriculture (Jhum). Major changes can be observed in built-up area which is increasing drastically by 10.66%. The detailed analysis of future projected can be found from literature (Anand and Oinam 2020).

Table 2 LULC change assessment between the year 2017 and 2030 (Source: Anand and Oinam (2020)).
Full size table

4.2 Model calibration and validation

Sensitivity, calibration and validation are required prior to using the outputs for future analysis. Sensitivity analysis was carried out on different sets of parameters. The SWAT model was calibrated between the time period 2010 and 2014 and validated between the time period 2015 and 2017 at the outlet point of Manipur River basin. In order to develop appropriate water, soil and LULC condition, a warm-up period of 2 years (2008–2009) is provided for calibration (table 3) shows the calibrated parameters with their ranges. Calibration and validation result of discharge of Manipur River watershed showed good performance of model (figure 4) and the R2 was 0.78 and 0.75 during the calibration and validation period, respectively (table 4). Results of different water balance components such as discharge, evapotranspiration, water yield, etc., obtained from calibrated model between the time period 1999 and 2017 was used as a baseline reference data in order to compare with the future simulated results of different GCM models for 2050s and 2090s decade.

Table 3 Calibrated parameter values with their ranges.
Full size table
Figure 4
figure 4

Calibration and validation result of SWAT model of Manipur River basin at Ithai (2008–2017).

Full size image
Table 4 Statistical measures of SWAT model parameters.
Full size table

4.3 Projected precipitation

Downscaled data at 10 km2 spatial resolution was used for the analysis of future temperature and precipitation. Four GCM models with two RCP scenarios for the time period 2046–2064 and 2081–2099 were then compared with baseline scenario (1999–2017).

In all the RCPs scenarios, there is increase in precipitation on temporal scale, i.e., precipitation for 2090s decade is more as compared to 2050s (figure 5). Projected increase in precipitation will lead to rise in discharge in all the scenarios. Projected increase in precipitation is most likely to be higher during monsoon season (June to September) and post-monsoon season (October to November) as compared to that of dry (December to February) and pre-monsoon season (March to May). In our projection, all the scenarios show the increase in precipitation in monsoon and post-monsoon season. This follows the similar trends observed by India Meteorological Department (IMD) in past 60 years in Manipur, India (Rathore et al. 2013). Maximum increase in precipitation of 58.83% was observed under RCP 8.5 during 2090s decade, whereas the minimum increase in precipitation of 43.8% was observed under RCP 2.6 during 2050s decade.

Figure 5
figure 5

Monthly seasonal variation and uncertainty in precipitation under RCP 2.6 and RCP 8.5 for the time period 2046–2064 and 2081–2099 across four GCMs and historical scenario (1999–2017).

Full size image

Higher amount of rainfall and uncertainty could aggravate disasters like landslides in the hilly region and floods in the low lying regions of Manipur River basin. Changes in the frequencies of higher rainfall may lead to soil erosion and land degradation which may reduce the area under agriculture and forest cover.

4.4 Projected temperature

Temperature is also one of the critical parameters in climate science. Unlike precipitation, the amount of uncertainties involved in temperature is less. Temperature is most likely to increase progressively with time for RCP 2.6 and RCP 8.5 in all the GCM model (figure 6). The maximum rise in temperature is for the time period 2081–2099 under the RCP 8.5 scenario, whereas the minimum increase in temperature is for 2046–2064 time period under RCP 2.6 scenario. There is no indication of reduction in annual mean temperature. This follows the trend of global projections for the northern hemisphere, this will lead to increase in potential evapotranspiration and actual evapotranspiration (Immerzeel et al. 2012). Projected temperature and uncertainty in four GCM models has been shown in figure 6. The mean temperature is likely to be increased by 0.42°C to 3.7°C in the coming decade. Variation in climate parameters under RCP 2.6 and RCP 8.5 has been tabulated in tables 5 and 6, respectively.

Figure 6
figure 6

Monthly seasonal variation and uncertainty in mean temperature under RCP 2.6 and RCP 8.5 for the time period 2046–2064 and 2081–2099 across four GCMs and historical scenario (1999–2017).

Full size image
Table 5 Variation in climate parameters under RCP 2.6.
Full size table
Table 6 Variation in climate parameters under RCP 8.5.
Full size table

4.5 Impact on hydrological parameters

4.5.1 Impact on discharge

Different water balance components and change in LULC may affect discharge in the river. There is increase in discharge at the outlet point under all RCPs scenario. Increase of 57.79 m3/s was observed for the highest scenario RCP 8.5 in the time period 2081–2099, whereas increase of 9.94 m3/s was observed for the lowest scenario RCP 2.6 in the time period 2046–2064. In monsoon, post-monsoon and dry season discharge is likely to get increased (figure 7). Whereas in pre-monsoon season, it shows a decreasing trend as compared to the baseline scenario because of shift in monsoon season towards the later half of the year. Increase in the area under built-up and water bodies is also one of the reasons in the increase of discharge in Manipur River basin.

Figure 7
figure 7

Seasonal variation in runoff under RCP 2.6 and RCP 8.5.

Full size image

4.5.2 Impact on water balance components

Different parameters of water balance contribute to overall hydrological cycle of the watershed. Water balance refers to net amount of water contributed by sub-basins and HRUs to the stream flow. Water balance components are also influenced by LULC of the area under the basin. Increase in the area under built-up and water bodies leads to increase in surface runoff. In this research, impacts of climate change on various water balance parameters namely, runoff, precipitation, evapotranspiration and water yield (WY) have been analyzed in figure 8. Future predicted water balance parameters for 2050s and 2090s decades under the RCP 2.6 and RCP 8.5 were compared with calibrated SWAT model output (1999–2017) as baseline scenario or reference data, due to the unavailability of observed data of various water balance parameters. Using the calibrated SWAT model output as a baseline scenario is justifiable due to the lack of observed data set for different water balance components. Water balance components for RCP 2.6 and RCP 8.5 have been shown in figure 8.

Figure 8
figure 8

Water balance component under RCP 2.6 and RCP 8.5 for ECHAM and HadCM3 and historical scenario (1999–2017).

Full size image

4.5.3 Impact on evapotranspiration

Evapotranspiration is likely to increase due to rise in temperature both spatially and temporally. There is a maximum increase of 54.5% under RCP 8.5 for 2090s decade, whereas for 2050s decade evapotranspiration is likely to get increase by 47.2% under RCP 2.6. Evapotranspiration is more pronounced in the northeastern side of the basin in higher elevation as compared to lower elevation. Rate of evapotranspiration appears to be rising progressively in the future time period with progressive increase in temperature in the basin. Combining the effect of rise in temperature with precipitation will have significant effect on the evapotranspiration (Stonefelt et al. 2000). Spatial variation of evapotranspiration in the basin has been represented in figure 9. Area under the forest cover in higher elevation is more susceptible to evapotranspiration in the basin.

Figure 9
figure 9

Spatial variation of evapotranspiration in Manipur River basin.

Full size image

4.5.4 Impact on water yield

Water yield includes the direct surface runoff, groundwater flow, lateral flow, transmission losses and pond abstraction. Increase in precipitation will lead to increase in water yield. In comparison to precipitation and temperature, other parameters like, wind speed, solar radiation and relative humidity have less significant effect on water yield. Water yield is projected to be increased minimum by 63.05% in 2050s decade and 89.8% under RCP 8.5 in the latter half of this century. Increase in the water yield is more in the lower reaches of the basin near to Loktak Lake as compared to the upper reaches of the basin. Increase in water yield in the lower elevation which lead to inundation of low lying area near to the lake. This indicates that area around Loktak Lake is more susceptible to change in climate than the other regions in the basin. Spatial distribution of water yield in the basin is represented in figure 10. Variation in various water balance components, such as, runoff, evapotranspiration and water yield under RCP 2.6 and RCP 8.5 for different time periods is tabulated in tables 7 and 8.

Figure 10
figure 10

Spatial variation of water yield in Manipur River basin.

Full size image
Table 7 Variation in hydrological parameters under RCP 2.6.
Full size table
Table 8 Variation in hydrological parameters under RCP 8.5.
Full size table

5 Discussions and conclusion

The effect of climate change on water resources is a serious concern in the northeastern part of India in the lesser Himalayan Region. Rising temperature and increased precipitation have serious impacts on water balance components on both spatial and temporal scale. Hilly areas of Manipur are responsible for more precipitation. Major part of the Manipur river basin lies in the Manipur valley. Because of this when the moisture laden winds enter the valley, since the valley is land locked geographically from all four sides by the hills, most of the wind carrying moisture gets trapped in between the mountain ranges resulting in more precipitation in the basin. Because of the increase in precipitation, simulated discharge by the SWAT model for both ECHAM and HadCM3 under RCP 2.6 and RCP 8.5 showed increase in discharge compared to base line scenario. The increase in discharge can have both positive and negative influence in the future period for water resource management and planning. The basin is currently being used to produce hydroelectric power, through the hydropower plant with an installed capacity of 105 MW. Increase in discharge will better its efficiency of production of hydroelectric power during the post-monsoon season where both precipitation and discharge are more likely to increase. Although increase in discharge during monsoon season does not affect the hydropower generation efficiency, it can have adverse effect in the areas around Loktak Lake which is vulnerable to impacts of climate change. Increase in discharge can lead to increase in the water level of the lake beyond its capacity; this may lead to flooding in the low lying areas around the lake (Singh et al. 2010; Anand et al. 2019).

There does not seem to be a problem of water availability in the Manipur River basin in this century considering a projected increase in precipitation, discharge and water yield. As the northern hilly regions of the basin under the forest cover is highly susceptible to climate change, heavy precipitation in the region increases the risk of landslides in the future. The key findings of this study indicate that the annual temperature is predicted to increase by 0.68°–2.84°C, whereas annual average precipitation would increase by 43.78–58.31% by the end of 21st century. Precipitation is expected to increase more during monsoon and post-monsoon season as compared to pre-monsoon and dry season. Change in meteorological condition and LULC will lead to increase in runoff, evapotranspiration and water yield by 57.79 m3/s, 54.59% and 89.82%, respectively, under RCP 8.5. Whereas under RCP 2.6 runoff, evapotranspiration and water yield is likely to get increased by 27.46 m3/s, 53.49% and 73.63%, respectively, by 2090’s. Stagnation of water in the low lying wetland areas near Loktak lake further increases the risk of flooding in the future. Our research findings could contribute to an effective management of water resource in Manipur River basin considering the effect of climate change in the basin. Benefits can be harnessed from climate change in the sector of water demand and supply, and in generation of hydroelectric power, but it is hard to ignore the adverse effects of climate change. Since, Manipur River basin lies in extreme data scarce region, due to the lack of observed discharge data for longer time period, 3 yrs stage-discharge data was used to develop the rating curve, whereas 7 yrs stage data was used for calibration and 3 yrs data was used for validation of hydrological model. This is one of the limitations of this study as the basin is ungauged and has extreme data scarcity. This study takes into consideration the four GCM models with first level of LULC classification for climate change impact assessment. Future studies can be done by incorporating Regional Climate Models (RCMs) and second level of LULC classification with high resolution dataset in order to quantify the impact of future climate change.