A hydrological modelling-based approach for vulnerable area identification under changing climate scenarios

The hydrologic behaviour of the Brahmani River basin (BRB) (39,633.90 km), India was assessed for the base period (1970–1999) and future climate scenarios (2050) using the Soil and Water Assessment Tool (SWAT). Monthly streamflow data of 2000–2009 and 2010–2012 was used for calibration and validation, respectively, and performed satisfactorily with Nash-Sutcliffe Efficiency (ENS) of 0.52–0.55. The projected future climatic outcomes of the HadGEM2-ES model indicated that minimum temperature, maximum temperature, and precipitation may increase by 1.11–3.72 C, 0.27–2.89 C, and 16–263 mm, respectively, by 2050. The mean annual streamflow over the basin may increase by 20.86, 11.29, 4.45, and 37.94% under representative concentration pathway (RCP) 2.6, 4.5, 6.0, and 8.5, respectively, whereas the sediment yield is likely to increase by 23.34, 10.53, 2.45, and 27.62% under RCP 2.6, 4.5, 6.0, and 8.5, respectively, signifying RCP 8.5 to be the most adverse scenario for the BRB. Moreover, a tenfold increase in environmental flow (defined as Q90) by the mid-century period is expected under the RCP 8.5 scenario. The vulnerable area assessment revealed that the increase in moderate and high erosionprone regions will be more prevalent in the mid-century. The methodology developed herein could be successfully implemented for identification and prioritization of critical zones in worldwide river basins. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wcc.2020.202 ://iwaponline.com/jwcc/article-pdf/12/2/433/866131/jwc0120433.pdf Sonam S. Dash School of Water Resources, IIT Kharagpur, WB, 721302 Kharagpur, India Dipaka R. Sena Uday Mandal (corresponding author) Gopal Kumar Prasant K. Mishra Hydrology and Engineering Division, ICAR-Indian Institute of Soil and Water Conservation-Dehradun, Dehradun, Uttarakhand, 248195, India E-mail: uget.uday.mandal@gmail.com Anil Kumar G. B. Pant University of Agriculture & Technology, Pantnagar, 263145, Uttarakhand, India Monika Rawat Amity University, Noida, 201301, India


INTRODUCTION
Water is becoming one of the scarcest natural resources in densely populated countries such as India despite the fact that they are endowed with many seasonal and perennial rivers often causing flood havoc. This warrants close attention of policy makers and implementing agencies responsible for the implementation of sustainable water conservation and management practices. For efficient management of water resources in various sectors such as domestic, irrigation, and industrial sectors, and to improve flood control, drainage, and water quality, an accurate forecast of streamflow is essential. The need for management activities becomes intense in case of severe alteration in the climate and land use pattern of a region. A little variability in the rainfall pattern may cause a substantial impact on agriculture and allied sectors by alteration of the supply-demand scenario in agriculture with changes in streamflow and sediment yield. In the current state-of-the-art, although there are many hydrological modelling-based climate studies in the literature (Hassan ; Padhiary et al. ), no such study is capable of fully mapping the variability in precipitation and temperature of the BRB under changing climate scenarios. Therefore, widespread knowledge about the hydrological fluxes with acceptable certainty can be implemented for possible future impact assessment of BRB. In this study, the output of the HadGEM2-ES climate model is adapted for flow simulation, which is an implementation of CMIP5 centennial simulations.
The BRB is a traditional hub of agricultural activities, but it is converting into an industrial hub as a consequence of rapid urbanization, resulting in significant increase in demand for water. Historically, without prior potential hydrological analysis, the major portion of the available inflow was allocated to less important sectors (Guntner et al. ). This leads to a significant reduction in the flow volume at the outlet and intensified water demand at the downstream of the reservoir. Thus, it is necessary to identify the vulnerable areas from the perspective of water demand and severe hydrological alterations. Attempts have been made to quantify the flood hazard (Dash et al. ) and potential soil erosion-prone regions (Chen et al. ) individually. The vulnerable areas are a combination of flood and erosion-prone regions because of the complex degree of association between them, which was not addressed in previous studies.
In light of the above research gaps, this study has been carried out with the following specific objectives: (i) to estimate the streamflow and sediment yield fluxes over the BRB under four future climate scenarios; and (ii) to identify the vulnerable sub-basins of the BRB by analysing the streamflow and sediment yield under both present and future climate scenarios. The outcomes of this study will be of great interest to policy makers in selecting the locations for implementing best management practices (BMPs), in order to develop a sustainable operational plan for management of water resources in future climate change scenarios. This paper is organized as follows: In the next two sections information is presented about the study area and the methodology followed in this study is described. The important outcomes of this research are then presented (Results).
The paper concludes with important discussions pertaining to the results obtained.

STUDY AREA
The Brahmani River basin is encompassed between longi- where SW t is the final soil water content at period t (mm); SW 0 is the initial soil water content (mm); t is the time (no. of days); R daily is the amount of precipitation on i th day (mm); Q sur is the amount of surface runoff on i th day (mm); E a is the amount of evapotranspiration on i th day (mm); W seep is the amount of water as recharge to groundwater from the soil profile on i th day (mm); Q gw is the amount of return flow on i th day (mm).
In SWAT, various hydrological processes are character- Though the SWAT model can accommodate the soil information of ten layers as input, due to data unavailability, information of only two layers was used in this study. The land use data prepared by National Remote Sensing Centre (NRSC) at a spatial scale of 1:250,000 was utilized as input for HRU delineation. Observed weather data, precipitation (mm), minimum and maximum temperature ( C), wind speed (m/s), solar radiation (MJ/m 2 ), and relative Individual sub-basin outlets and main catchment outlet were defined and 126 sub-basins were delineated. Land use and soil layers were reclassified as per the user-defined classes. Slope was obtained from the input DEM layer.
Five slope classes were defined for the reclassification purpose. The HRU thresholds were defined as 10% for land use, 15% for soil, and 10% for slope classes to limit the number of HRUs and improve computational efficiency of the model. After overlaying these three layers with the predefined unique thresholds, 2,286 HRUs were generated.
Six weather parameters, namely, precipitation, minimum and maximum temperature, solar radiation, wind speed, and relative humidity were provided as input on a daily basis. The data for the period 1980-2013 obtained from eight meteorological stations (i.e. Altuma, Gomlai, Panposh, Indupur, Jenapur, Talcher, Tilga, and Jarikela) were used as the inputs.
The model simulation was carried out for a period of 34 years, starting from 1 January 1980 to 31 December 2013.
Three years of warm-up period was included for improving the simulation performance. To account for the heterogeneous distribution of rainfall over the study area, a skewed normal distribution pattern of rainfall was preferred.
The parameterization approach was followed to account for spatial heterogeneity conceptually. Some model parameters, which could not be obtained directly from the present data- analysis. Pictorial representation of the adopted methodology for SWAT model simulation is shown in Figure 3.

Model calibration and validation
Instead of manual calibration, which is more time consuming and often fails to identify the inter-parameter sensitivity, an automatic calibration approach of SWAT-CUP tool (http://swat.tamu.edu/software/swat-cup/) was used. The Sequential Uncertainty Fitting (SUFI-2) algorithm was adopted for the calibration. An initial two years (1998)(1999) was used as warm-up period and a period of ten   (Table 1), to classify the erosion-prone regions.

Goodness of fit assessment
The calibrated parameters were updated in the raw model respectively.
Moreover, co-linearity between simulated and observed values was interpreted using the R 2 , and it ranges from À1 to 1.
where n is total number of observed data, O i and P i are observed and simulated data at time i, O 0 and P 0 are the mean of observed and simulated data. The nearer the value of E NS and R 2 to 1, the better the model tends to perform in modelling and capturing the dynamics. E NS generally lies between À∞ and 1; more positive values

Sensitivity analysis
Model  Table 2. Sensitivity analysis operation was performed for both the calibration and warm-up periods. Four iterations were performed in the SWAT-CUP. Out of 22 parameters, ten were found most sensitive during the calibration processes and were ranked according to the objective function value; that is, P-value and absolute t-stat between observed and simulated streamflow values. Streamflow was found to be impinged by both surface water and groundwater parameters of the study basin (Table 2) indicating diverse hydrological variability in the study area.
The most important baseflow calibration parameter is baseflow alpha factor (ALPHA_BF), which explains the contribution of groundwater flow to variation in the recharge.
ALPHA_BF also bears a direct relation with another groundwater calibration parameter, groundwater recession constant. A higher value of these two parameters indicates a quick response to groundwater recharge in the basin.
Due to lack of knowledge regarding the basin hydrology, the complete calibration ranges of ALPHA_BF (i.e. 0-1) was considered in this study. The Manning's n-value for main channel (CH_N2) was found to be the third most sen-

Uncertainty analysis
The magnitude of P-factor and R-factor were found to be different for different gauging locations (Table 3). During the calibration process, the value of P-factor and R-factor at two different gauging stations was slightly higher than 0.5 which indicates the close agreement between observed and simulated values. The 95% prediction uncertainty  were obtained by the regional approach of river flow prediction at the gauging stations. Therefore, they were not affected by any water storage structures or reservoirs located in the upstream. The unavailability of consistent data of reservoir inflow-outflow at a daily timescale for simulation of runoff was a constraint and main reason for uncertainty.

Model calibration and validation
The initial simulation showed an E NS value of 0.12 and R 2 of

Sub-basin-wise variation in streamflow and sediment yield
To conserve soil and water by implementing best management practices (BMPs), it is essential to identify and prioritize the critical sub-basins. The impact of climate change on hydrology was quantified by comparing the calibrated SWAT model outputs of base periods  and mid-century period, 2050. The preliminary assumption of this study was that land use, soil type, and agricultural practices remained unchanged in the study area for the respectively for each RCP (Figures 7 and 8). As per the value of proposed vulnerability using an equal weight approach for runoff and sediment yield, sub-basins were categorized into five vulnerability classes such as slight, low, moderate, high, and extreme. The areas under five vulnerability classes were calculated sub-basin-wise for both present and future climate change scenarios and are presented in Table 5. During the base period, the area under extreme and high vulnerability classes were 3.5% and no substantial change is found under projected climate change scenarios.

Identification of vulnerable area
The highest increase in area is indicated under the moder- Approximately 30% of basin area comes under the moderate to extreme vulnerability class and needs priority attention.
Only 5.5% of the area is categorized under severe and high erosion-prone zone (Figure 9). Spatial heterogeneity regarding projected climate change and hydrological alterations in BRB is also corroborated from the predicted reduction in the vulnerability of some patches against the overall increased vulnerability.
In the projected climate condition, around 55% of the total area is under the moderate to extremely vulnerable class which is almost double that of the base scenario.
Though no considerable change is observed in case of severe erosion-prone areas, an increase of around 12% is observed in the case of high erosion-prone areas. The percentage of slight erosion-prone areas reduced from an initial value of 24.5%-15.7%.      From the water availability aspect over BRB, an improved situation is expected in the mid-century. However, the real problem arises when erosion-prone areas come into the picture. Though no appreciable change in the area susceptible to severe erosion is predicted, there is a likelihood of an increase in moderate and high erosion-prone areas. If not attended properly, the increased erosion potential may negate the positive effect of increased water availability for crop production. Due to the adverse soil erosion potential in the mid-21st century, a contradictory scenario would be created when the benefit of a large amount of available water for irrigation would not be actualized to a satisfactory extent due to the reduced availability of fertile land.
As described earlier, the BRB is going to be an industrial hub in the near future, causing a significant reduction of the

CONCLUSION
The hydrology of BRB was quantified for the baseline period  as well as for future climate (mid-century, 2050).
The SWAT model could be successfully used for simulation of BRB hydrology. Integration with the SWAT-CUP tool and SUFI-2 algorithm for calibration and validation was found to be very effective. A good agreement between observed and simulated streamflow during calibration and validation proves the applicability of the SWAT model at river basin-scale under a limited data availability scenario.
Bias-corrected outputs of GCM (HadGEM2-ES) were used to assess the impact of changes on runoff and sediment yield. The variability of climatic factors is found to be at a maximum during the RCP 8.5 scenario for the mid-century.
Overall, runoff and soil loss are predicted to increase under future climate change conditions. However, a decrease in runoff and sediment loss is indicated from some patches in upper catchments of the BRB. Increase in runoff and sediment yield indicated from mid-and lower catchments may be attributed to increased rainfall magnitude, altered rainfall pattern (Figure 6), interaction with topography, and land use under future climate. An approximate ten-fold increase in environmental flow (Q 90 ) is expected under projected climate scenarios. The increased flow can be managed to fulfil the ever-increasing demand of various sectors and can be of great support for ongoing urbanization. The area under moderate and high vulnerability classes is likely to increase from 25 to 37% and 2 to 14%, respectively, by the mid-century. A substantial area under the low and slightly vulnerable classes is likely to be converted to the moderate and highly vulnerable classes under the changed climate. The area in need of soil and water conservation treatment is likely to double (55% of total area) under future climate change context. The critical sub-basins identified in this study can be taken up as a priority for introducing best management practices (BMPs) to sustain agricultural productivity and judicious water resource management. Conclusively, the threat to hydroelectric plants for power production is imminent due to climate change scenarios across all the RCPs in BRB.
This study provides a generalized framework for the identification of the vulnerable areas in a watershed/river basin in the context of climate change. The outcomes of this research would act as a guiding tool for policy makers to identify the locations where implementation of suitable best management practices (BMPs) may alleviate future adverse conditions to a greater extent. In future studies, more complex frameworks can be integrated with the existing approach to identify the vulnerable regions in both climate change and land use/land cover change scenarios.