Climate change impact assessment on hydrological fluxes based on ensemble GCM outputs: a case study in eastern Indian River Basin

The present study assessed the impact of climate change in the Anandapur catchment of Baitarani River basin, India, using the Soil and Water Assessment Tool (SWAT) hydrological model. The future climatic alterations under two Representative Concentration Pathways (RCPs), i.e. 4.5 and 8.5 scenarios, are quantified by an ensemble of two different CMIP5 models, i.e. CNRM-CM5.0, GFDL-CM3.0. The outcomes of this study reveal that the future rainfall and temperature may experience an increasing trend with gradual shifting of monsoon from mid-June to mid-May. The average annual streamflow experienced the highest increase during the period 2071–2095, whereas the highest average annual evapotranspiration (ET) is observed for the period 2046–2070 under both the RCPs and resulting in comparatively slower groundwater recharge (GWR) over the basin. In order to implement suitable adaptation strategies for a possible flood scenario on the concerned study basin, three critical sub-basins, namely, sub-basin 1, 4, and 5, were identified. Furthermore, the altered streamflow and ET dynamics may result in a significant shifting in the conventional agricultural practice in the coming future time scales. Conclusively, the outcomes of this study have potential implications for policy makers in formulating the policies related to sustainable water resources management in future scenarios. doi: 10.2166/wcc.2019.080 om http://iwaponline.com/jwcc/article-pdf/11/4/1676/830179/jwc0111676.pdf er 2021 Jagadish Padhiary (corresponding author) Kanhu Charan Patra Department of Civil Engineering, National Institute of Technology, Rourkela, Odisha, India E-mail: jagadishpadhiary@gmail.com Sonam Sandeep Dash School of Water Resources, Indian Institute of Technology Kharagpur, West Bengal, India A. Uday Kumar Civil Engineering Department, National Institute of Technology, Warangal, 506004, India


INTRODUCTION
Water is the most important natural resource for human beings due to its wide application in the field of domestic use, irrigation water supply and hydroelectric power generation (Chan ). Global climate change has exerted a considerable impact on the hydrological cycle that has subsequently affected the available water resources (Everett et al. ). The variation in precipitation patterns, floods, droughts, and evapotranspiration rate over different regions is caused by climate change (Frederick & (Sankarasubramanian & Vogel ). Hence, a basin-level hydrologic analysis is very important to evaluate the sensitivity of a basin to climate change scenarios and to develop better water management systems and climate adaption strategies.
In the current state of the art, General Circulation Models (GCMs) have proven to be the most reliable tools to predict future water resources changes in a hydrological basin (Xu ). Recently, the future climate change has been projected by using different GCMs (Tebaldi & Knutti ). The multi-model mean gives a better simulation of climate variables than a single model projection (Gleckler et  Based on the identified research gaps, the specific objectives of this study are: (1) to set up the SWAT model for the study catchment during the base period within the acceptable uncertainty limits; (2) to evaluate the future climate change on different hydrological fluxes in spatial and temporal scale by using downscaled meteorological data from an ensemble of two GCMs. The multi-model ensemble based approach will be helpful in formulating a more realistic future climate scenario for the concerned study area. The sub-basin scale water resources assessment will be helpful in formulating sustainable management policies under changing climate and land use scenarios. The integrated hydrological model based futuristic water balance assessment will formulate a flexible modelling framework and can be extended to many worldwide river catchments. Thus the outcomes of this study will be beneficial for policymakers to accurately estimate the available water resources in future climate scenarios and, subsequently, the supply-demand mechanism in a water distribution system can be managed more effectively.

METHODOLOGY Study area and data
The Anandapur catchment of Baitarani River Basin, which is situated in between 85 0 0 0″ to 86 30 0 0″ E longitude and 21 0 0 0″ to 22 30 0 0″N latitude is selected as the study area and is situated in the Keonjhar district of Odisha state, India ( Figure 1). The catchment area is 8,645 km 2 with topographic elevation ranging from 32 to 1,181 m above mean sea level (MSL). The basin experiences an undulated topography with average slope varying between 0 and 2%.
Average rainfall in the basin is 1,628 mm with sub-humid tropical climate predominating over the complete basin.
where SW t is the final soil water content (mm) at the end of day i, SW 0 is the initial soil water content at the beginning of day i (mm), R day is the amount of precipitation on day i (mm), Q surf is the amount of surface runoff on day i (mm), E a is the amount of evapotranspiration on day i (mm), W seep is the amount of water entering the vadose zone from the soil profile on day i (mm), Q gw is the amount of return flow on day i (mm), and t is the time of day.

Model set-up, calibration, and validation for streamflow
The DEM, land use, soil data and weather data (rainfall and temperature) are used as input to SWAT for hydrological modelling in the Anandapur catchment. The catchment is delineated into sub-basins using DEM data. Further, the sub-basins are divided into Hydrologic Response Units In this study the Nash-Sutcliffe efficiency (NSE) (Equation (2)), coefficient of determination (R 2 ) (Equation (3)) and percent bias (PBIAS) (Equation (4)) statistical indicators are used in the model performance evaluation. Performance of the model is good when the PBIAS is within ±15%, NSE is above 0.75 (Moriasi et al. ) and R 2 is close to one: where O i is the ith observed data, S i is the ith predicted value, P i is the ith predicted data, O is the mean of measured data, P is the mean of model estimated data, and N is the total number of simulation periods.

Sensitivity analysis
Sensitivity analysis is one of the pre-processing steps that helps to understand the change in model outputs concerning  Parameter uncertainty was expressed in terms of a 95% prediction uncertainty (95PPU) band. The lower limit of 95PPU is 2.5% and the upper limit is 97.5%. Initially, parameter uncertainty remains large, but after each iteration parameter uncertainty decreases. In the case of a more sensitive parameter, there is a large uncertainty reduction in comparison to that for the less sensitive parameter. The 95PPU is quantified by the P-factor (0-1) and the R-factor (0-∞). The P-factor is the percentage measured data bracketed by the band of 95PPU, whereas the R-factor is the ratio of the average width of the band to the standard deviation of the corresponding measured variable (Abbaspour et al. ). When the P-factor is 1 and R-factor is 0, the simulated value perfectly matches the observed value (Abbaspour ).  Relative Error (MARE) as given in Equations (5) and (6):

Multi
where P hist is the historical period precipitation, GCM hist is historical GCM simulated precipitation, W i is initial weight, n is number of GCM models and i is the model index.
Further, by multiplying the computed weight with the future precipitation CDF, the weighed mean CDF is calculated. Then the subsequent MARE and weight are computed until the final weight for the different GCM remains the same as that of the previous iteration.

Bias correction and downscaling
Before entering the meteorological inputs to the hydrological model, the bias correction of GCM output is essential due to the associated systematic and random model errors Furthermore, the scatter plot shown in Figure 2(b) indicates a greater degree of correlation among the GCM and downscaled precipitation data for the period 2071-2095: where P obs and P wet are observed and weighted precipitation, F wet is the CDF of P wet and F À1 obs is the inverse CDF corresponding to P obs .
The Statistical Downscaling Model (SDSM) based hybrid approach that forces synoptic-scale climate variables into station scale variables using a suitable statistical relationship is adopted in this study for downscaling the GCM outputs. The station-wise predictor variables were chosen from the partial correlation coefficient value carried out at 5% level of significance. The downscaling process was further enhanced by using the Climate Forecast System Reanalysis (CFSR) data products as the predictor variable. The commonly identified predictor variables for multiple stations include precipitation, relative humidity, temperature at 700 hPa pressure level and atmospheric pressure.

Historical climate and land use and land cover (LULC) trend analysis
The Mann-Kendall test, a popular non-parametric trend analysis approach, has been used in this study to detect the trend in the historical time series of the above three variables. The outcomes of this analysis reveal that no significant trend is present for either temperature or precipitation time series and is evidenced by a Sen's slope value of 9.41 mm/y and 0.021 C, respectively. To confirm the role of weather variables on controlling the groundwater recharge flux, the trend analysis approach was further extended for the groundwater recharge (GWR) and, certainly, the results also indicate absence of any significant trend. In general, it can be inferred from this analysis that the historical period was quite normal from the context of both the meteorological and hydrological variability and was supported further  Table 2. From Table 2 it can be surmised that there is no significant change in land use over the past three decades. Therefore, it can be assumed that the future water balance components are mainly affected by climate change and accordingly the following analyses were carried out.

Impact of climate change on precipitation and temperature
The variation in all the previously mentioned meteorological variables sought to affect the future water resources of Anandapur catchment (AC) both quantitatively and qualitatively. However, due to availability of information of only precipitation and temperature for the future projections, this study is confined to evaluate the role of only these two variables in assessing the future water resources. Thus the variability of precipitation and temperature for the AC was analysed for the RCP 4.5 and RCP 8.5 scenarios in three future time slices of 2021-2045, 2046-2070 and 2071-2095 (Figures 5 and 6). It can be envisaged from      Table 3. It can be surmised that the rate of increase in the average minimum temperature will be higher than that of average maximum temperature in the future climate change scenario. The highest increase in maximum temperature is found to be 1.2 and 1.6 C for    Here, the ET is estimated using the Penman-Monteith method that depends on various climatic variables including temperature, wind speed, solar radiation, relative humidity, and ground heat flux and air density. Hence, though the increase in precipitation is less in the period 2046-2070 than the period 2021-2045, the ET is maximum in the period 2046-2070 due to increase in temperature. On the contrary, the increase in ET is less at the end of the period due to a lower increase in rainfall than the other two periods.
The spatial distribution of streamflow is shown in  seems to alter the present practice of rice transplantation during the end of July to the middle of August in the study basin to prevent the acute water shortage during the cropping period. It can be inferred from  include only three sub-basins (sub-basins 3, 4, and 7), considering their overall magnitude among all the future time scales, and special emphasis must be given to these three sub-basins while formulating management policies.

CONCLUSIONS
The impact of climate change on water resources in the Anandapur catchment has been analysed using the future climatic dataset under different scenarios. It has been noticed that the spatial and temporal variation of water balance components are affected due to extreme climatic alterations. Both the streamflow and ET are found to be increasing, whereas the GWR is decreasing in future time periods. The streamflow is attributed mainly due to precipitation and the ET is influenced by both precipitation and temperature. The slower GWR is attributed to more ET losses and increased urbanization in the future time periods, thereby leading to substantial depletion of the water table.
Overall, the decreasing trend of water availability in the catchment is likely to intensify further with fewer runoff losses in the form of streamflow. On the other hand, the crop is expected to suffer from moisture stress owing to frequent alternate dry spells and unexpected inundation due to frequent flash floods, due to intensive rainfall across the basin during monsoon season. Frequent irrigation is required during the non-monsoon period because of increased ET in this period. Therefore, for achieving sustainable crop production in order to meet the food grain demand of a growing population, the following suggestions may be incorporated by policymakers: • Climate-resilient cropping pattern is to be adapted to utilize the monsoon rain effectively.
• More storage structures across the basin are required to be built to restrict the streamflow and harvest the excess water for irrigation during severe dry spells.
• Heavy duty crops grown at present across the basin may be substituted by light duty ones or low duration varieties to mitigate the higher irrigation demand of the crops accentuated under the increasing trend of evapotranspiration.
• Emphasis on more afforestation, and soil and water conservation measures would help in reducing the runoff losses across the basin and thereby increase the in-situ soil moisture storage and groundwater recharge.