Climate change impact on streamflow and suspended sediment load in the flood-prone river basin

Climate change is a major impediment to achieving sustainable development goals, particularly in nations like India with limited ability to adapt and moderate expertise in managing natural resources. In its sixth Assessment Report (AR6), the Intergovernmental Panel on Climate Change (IPCC) stated that greenhouse gas emissions from human activity have contributed to an average global temperature increase of 1.1 °C or more since 1850–1900. The first two decades of the 21st century (2001–2020) saw an increase in global surface temperature of 0.99 [0.84–1.10] °C, and the last decade (2011–2020) saw an increase of 1.09 [0.95–1.20] °C since 1850–1900. The IPCC also concluded that over the course of the next 20 years, this increase will likely exceed 1.5 °C (IPCC 2021). Climate change is predicted to exacerbate the water cycle, causing changes in streamflow and watershed erosion because of more uneven occurrences of rainfall over time and geography (Nearing et al. 2004). Further study on the changes in surface runoff and soil erosion against the backdrop of climate shift is essential to determine the relationship between changing climate and alteration in the streamflow and suspended sediment load (SSL) at a regional scale.

The output of the Global Climate Models (GCMs) based on the Coupled Model Intercomparison Project Phase 5 (CMIP5) has been widely employed to anticipate future climate change and to assess how climate change may affect basin response (Shrestha et al. 2016; Jaiswal et al. 2017; Bisht et al. 2020; Ma et al. 2021). However, most CMIP5 GCMs, as well as their ensemble averages, have failed to describe the patterns in Indian summer monsoon rainfall, indicating poor reliability of rainfall estimates (Saha et al. 2014; Jain et al. 2019). The new climate scenarios based on the combinations of shared socioeconomic pathways (SSPs) and representative concentration pathways (RCPs) have been incorporated into the CMIP6 as part of revisions to the framework for climate modelling research (Van Vuuren et al. 2014; Riahi et al. 2017). A more improved global carbon cycle, aerosol effects, atmospheric chemistry, terrestrial and marine biogeochemistry, and other processes are included in CMIP6 model, which further enhances each model's physical process and simulation capability (Eyring et al. 2016). Also, the performance of the CMIP6 models for the Indian summer monsoon was reported to have improved, with a notable decrease in bias (Gusain et al. 2020). There are a few studies that used CMIP6 model projections to investigate climate change impact on the hydro-sedimentology of flood-prone Mahanadi River Basin (MRB) (Pandey et al. 2022). Despite improvements in CMIP6, there is still a substantial level of error in climate change scenarios from various GCMs because of inter-model differences, scenario uncertainty, and internal climate variability. Therefore, multi-model ensembles have been used to capture the wide range of uncertainty leading to increased confidence in the output when examining catchment responses to climate change (Jain et al. 2019; Ma et al. 2021).

The MRB in India is anticipated to be the most adversely affected Indian River basin by the intensity of floods brought on by changing climate due to its proximity to the Bay of Bengal (Pandey et al. 2022). The Hirakud dam and the Mundali weir close to the delta's outlet separate the MRB into three zones: the upper, middle, and delta regions. The upper zone serves as the catchment area of Hirakud dam. The middle Mahanadi Basin is located between the Hirakud dam and the Delta region. Over the previous 10 years, there have been five big floods in the basin (Jena et al. 2014). According to the study of historical flood records from 1969 to 2011, streamflow from the middle reaches caused 69% of big floods, the Hirakud dam and middle zone produced 23%, and Hirakud discharges alone caused 8% (Parhi et al. 2012). Given the possible negative implications of climate change, a growing body of research has concentrated on how it affects streamflow regimes (including floods) in the MRB (Asokan et al. 2016; Jaiswal et al. 2017; Bisht et al. 2020; Khatun et al. 2022; Pandey et al. 2022). Nevertheless, much research has not been done on how climate change may affect the amount of sediment produced in the MRB (Dunn et al. 2018). Using the global hydrogeomorphic model – WBMsed, Dunn et al. (2018) projected the twenty-first century river sediment export to the three deltas in the natural period (without any human interference) from the recent past. Twelve different environmental change scenarios were considered, each of which included a single period for the construction of future reservoirs and combinations of four climates and three socioeconomic paths. The anticipated sediment delivery for the Mahanadi delta increased from 23 to 40 Mt/year (+77%) between the pristine and recent periods, and then it declined from 7 to 25 Mt/year in the last decade of the twenty-first century. Although the WBMsed provides reasonable results globally, it may not always consider local inputs and processes since it is a global model with inputs of very low precision (Dunn et al. 2018). Therefore, given the scarcity of climate change assessment investigation and the significance of the catchment, there is a need to evaluate suspended sediment dynamics against the backdrop of climate change at a local scale using catchment level hydro-sedimentological models. Recent years have seen the development of numerous such models, such as the AGNPS (Agricultural Non-Point Source model), SWAT (Soil and Water Assessment Tool), and HSPF (Hydrologic Simulation Program – Fortran) (Young et al. 1989; Bicknell et al. 1996; Arnold et al. 1998). However, the SWAT hydrological program is one of the most tried-and-true, catchment scales, hydrological response unit based, semi-physical, and continuous hydrological programs created to comprehend the implication of climate shift and anthropogenic practices on the watershed process (Tan et al. 2020).

Given the foregoing explanation, the present research was carried out to examine the effects of potential climate change under the CMIP6 model projections on the streamflow and SSL from the Kantamal catchment situated within the middle zone of the MRB. This was accomplished by using the SWAT, a widely used hydrological model, in conjunction with climate projections produced by five bias-corrected GCMs for SSP245 and SSP585 climate scenarios for the near future (NF, 2020–2046) and mid future (MF, 2047–2073). The following specific goals were set: (1) To study how the catchment's climate is impacted by climate change, particularly in relation to changes in rainfall and diurnal temperature range (DTR). (2) To investigate how climate change affects ET, streamflow, and SSL on monthly, seasonal, and annual timeframes. Findings of the study may be helpful in enhancing the understanding of the effects of a changing climate on streamflow and SSL in the flood-prone river basin.

The SWAT model was set up for the area by using Digital Elevation Model (DEM), land use–land cover (LULC), soil, and weather information. The Shuttle Radar Topography Mission website (http://earthexplorer.usgs.gov) provided the DEM (90 m spatial resolution). The Food and Agriculture Organization (FAO) soil database was used to get soil information at a geographical scale of 1 km by 1 km. LULC of 2005 at a 100 m by 100 m resolution were retrieved from the Distributed Active Archive Center at Oak Ridge National Laboratory (ORNL DAAC) (Das et al. 2018). The India Meteorological Department (IMD), Pune, provided the gridded rainfall (0.25° × 0.25°) and temperature (1° × 1°) data used in this investigation (Pai et al. 2014). Bilinear interpolation is used to re-grid the 1° × 1° temperature observation to 0.25° × 0.25° to maintain the constant spatial resolution. The daily discharge and sediment data for the period 2000 to 2014, measured at the Kantamal gauging station, were collected from Central Water Commission (CWC), Bhubaneswar.

Monthly discharge and SSL were used to calibrate the model for the years 2003 to 2011 and to validate it for the years 2012 to 2014. Sensitivity analysis, calibration, and validation were carried out using the SWAT-CUP program. Even though SWAT includes more than 200 parameters, we chose only 22 based on sensitivity analysis result for the SWAT model calibration (Table 2). The 15 most sensitive parameters, in decreasing order, are CH_N2, PRF_BSN, SPCON, CH_K2, ALPHA_BNK, SOL_AWC, SPEXP, ESCO, GW_REVAP, EPCO, CN2, CH_K1, ADJ_PKR, GW_DELAY, and SOL_K. Eight out of 15 sensitive parameters are related to channel processes which implied that the overland flow process is not the dominant driver of streamflow and erosion in this catchment. Fitted parameters are optimized parameters which represent watershed characteristics of MRB. All the fitted parameters are within their physically meaningful range which increases the confidence in the calibrated models. Groundwater delay time (GW_DELAY) for the catchment is defined as the average travel time required for percolated water from the bottom of the soil profile to join the groundwater table through the vadose zone. Its calibrated value is 102.75 days which implied that the water table is at a shallow depth. Groundwater re-evaporation coefficient (GW_REVAP) refers to the rate of movement of water from an unconfined aquifer to the root zone of the soil profile. Its optimum value was found to be 0.1482 which shows that the groundwater re-evaporation is a significant process in the study area and the saturated zone is not very far below the land surface. Baseflow days are the number of days for baseflow recession to decline through one log cycle. The baseflow recession constant (V_ALPHA_BF) is inversely related to baseflow days and is a direct index of groundwater flow response to changes in recharge. Its optimum value is 0.3345 which indicates slow response. Deep aquifer percolation fraction (V_RCHRG_DP) is fitted to 0.0578 which implies minimal loss of water from catchment to deep aquifer recharge. Curve number (R_CN2) is optimized to 0.127 which results in final CN > 82 for every HRU. This represents a high surface runoff condition present in the study area. Optimized soil parameters also refer to moderately high runoff potential as 97.72% of the study area lies in hydrologic soil group of C. Manning ‘n’ for main channels (V_CH_N2) is 0.12 which resembles the roughness of natural stream with heavy timber or brushes. The effective hydraulic conductivity of the main channel (V_CH_K2) is 34.975 mm/h, which indicates that the bed and bank material of stream is sand and gravel mixture with low slit-clay content and it supports a moderately high loss rate. The baseflow alpha factor for bank storage (V_ALPHA_BNK) is 0.4659, indicating the moderate recession of bank storage. Bank storage contributes flow to the main channel or reach within the sub-basin. Effective hydraulic conductivity of the tributary channel (V_CH_K1) controls transmission loss from surface runoff as it flows to the main channel in the sub-basin. Its optimized value is 91.6038 mm/h which indicates a high loss rate. The channel erodibility factor (V_CH_COV1) is conceptually similar to the soil erodibility factor in the USLE equation. It is a function of bed/bank material and its optimized value is 0.322 which indicates that the channel provides a moderate high resistance to erosion. Channel cover factor (V_CH_COV2) is defined as the ratio of degradation from a channel with a specified vegetative cover to the corresponding degradation from a channel with no vegetative cover. The vegetation affects degradation by reducing the streamflow velocity and consequently its erosive power near the bed/bank. Its optimised value is 0.484 which indicates that moderate protection is offered by vegetative cover in the stream. Soil evaporation compensation factor (V_ESCO) is calibrated at 0.6215 which indicates a moderate contribution of lower layer soil to meet soil evaporative demand. Plant uptake compensation factor (EPCO) is calibrated at 0.6155 which again shows the moderate contribution of lower layer soil to meet plant water uptake demand. Peak flow adjustment factor in sub-basin for sediment routing (V_ADJ_PKR), peak flow adjustment factor in the main channel for sediment routing (V_PRF_BSN), linear sediment re-entrainment factors for sediment routing (V_SPCON), and exponential sediment re-entrainment factors for sediment routing (V_SPEXP) are calibrated at 1.1479, 0.5044, 0.0072, and 1.3196, respectively, which shows the moderate degradation of the river reach. These parameters are related to sediment routing by the simplified Bagnold equation. The lowest operationally demanding of the SWAT-CUP optimization software is the SUFI-2. It is based on the inverse modelling approach for simultaneous optimization and uncertainty analysis (Wu & Chen 2015) and was used for calibration and validation in this work. The sequential (SQN) calibration method has been employed in most research to simulate SSL and streamflow. However, the Sequential-Simultaneous (SQN_SML) method was tested as superior to the conventional SQN calibration method in terms of improved predictive skills and reduced parameter equifinality (Ranjan & Mishra 2022). Hence, the SQN_SML calibration approach was used in this study.

Climate change is most visible in rainfall and surface air temperature changes. As a result, the change in annual basin averaged climatic information for two future periods; near future (NF, 2020–2046), and mid future (MF, 2047–2073); with respect to the historical period (1985–2014) was analysed and depicted in Table 4. Expected rainfall was compared to the base period rainfall and the results show that three out of five models projected wetter conditions in the NF for both SSP245 and SSP585 scenarios. Also, all five GCMs, except MPI-ESM1-2-HR, predicted a wetter future in the mid future for both scenarios. This finding agreed with other studies which show the lesser uncertainty in the direction of projected rainfall in MF than NF for MRB (Jin et al. 2018; Bisht et al. 2020). For NF and MF, respectively, the percentage change in rainfall under SSP245 (SSP585) ranges from −7.75 to 11.08% (−12.97 to 13.73%) and −0.64 to 27.40% (2.8 to 23.49%). The result, using ensemble means of all five models, showed an increased NF rainfall of 2.86% (2.19%) and increased MF rainfall of 13.35% (16.98%) under SSP245 (SSP585) scenarios. Thus, MRB may suffer a wetter climate in both NF and MF provided future climate is unfolded following either of the SSPs. The DTR, defined as the difference between T_max and T_min, showed a declining trend, irrespective of climatic scenarios and analysis period. The drop in DTR has been identified as an important indicator of how the climate is changing around the world (Braganza et al. 2004). The change in DTR varied from −0.44 to −0.01 °C (−0.47 to −0.15 °C) for NF and from −0.53 to −0.12 °C (−0.91 to −0.4 °C) for MF under SSP245 (SSP585). An increase in projected minimum temperatures was more than the increase in projected maximum temperatures, resulting in a decline in the DTR in all future scenarios. This finding is in line with the previous studies which indicated that climate change was closely associated with a greater increase in the minimum temperature than the maximum temperature, and thus a widespread decline in DTR was observed in almost all parts of the globe during the past decades (Sun et al. 2021; Liu et al. 2022). Also, DTR in the MRB was reported to be declining by Bisht et al. (2020).

Climate change impact analysis of the annual hydro-sedimentological fluxes in the Kantamal catchment under the SSP245 and SSP585 scenarios with five GCMs revealed mixed signals for both streamflow and SSL; however, the directions of change for streamflow and SSL were the same (Table 5). Also, change in rainfall is accompanied by proportional change in streamflow both in terms of magnitude and direction.

NorESM2-MM projected the highest increase in streamflow (69.72%) and sediment load (106.44%) in MF under the SSP245 scenario as a result of the highest projected rainfall (Table 4). For NF and MF, respectively, the percentage change in streamflow under SSP245 (SSP585) ranges from −13.70 to 26.28% (−22.63 to 35.18%) and −5.47 to 69.72% (5.09 to 54.44%). Similarly, the percentage change in SSL under SSP245 (SSP585) ranges from −19.31 to 27.17% (−26.92 to 56.16%) and −17.19 to 106.44% (5.85 to 65.59%) for NF and MF, respectively. Uncertainty in the rainfall projection (range of projected rainfall) is found to be lower than that of streamflow and uncertainty in projected SSL is greater than that of streamflow under both SSPs and timeframes. Propagation of uncertainty down the cascading modelling chain may be attributed to the enlargement of error in SSL. Under SSP245 (SSP585), the mean ensemble result showed an increase in streamflow by 6.38% (6.06%) in the NF and an increase by 29.78% (36.90%) for the MF, as well as an increase in SSL by 4.43% (7.89%) for the NF and an increase by 37% (46.43%) for the MF. Our results corroborated with Pandey et al. (2022) which reported much higher streamflow under CMIP6 climate projection in the MRB that may result in frequent future floods.

Given that the MRB is among the most climate-vulnerable regions, it is crucial to comprehend how climate change may affect the MRB's hydro-sedimentological processes. Hence, this study analysed the response of streamflow and SSL to changing climate by using the sequential-simultaneous method followed by the calibrated SWAT model. Bias-corrected historical and future climatic data from five GCMs under SSP245 and SSP585 pathways were utilized in the SWAT model for the baseline (1985–2014), near future (NF, 2020–2046), and mid future (MF, 2047–2071) periods. Mean ensemble GCMs scenarios utilised research findings showing that MRB will experience a warmer and wetter climate in future with less uncertainty in MF than in NF. Consequently, streamflow and SSL are also projected to increase with a higher anomaly in the MF than in the NF under both SSPs. Uncertainty in projected SSL (46.48%–123.63%) is found to be greater than the error in streamflow (39.98–75.19%) which is greater than the uncertainty in projected rainfall (18.83–28.04%). Under all seasons except the summer, streamflow and SSL demonstrated an upward trend. This suggests that the flood problem in the Kantamal catchment may worsen in the future, given the increase in streamflow under most scenarios which has also been reported by Pandey et al. (2022). Furthermore, the increasing trend of SSL may reduce the storage capacity of the river body, resulting in frequent flooding of the Kantamal catchment and increasing the likelihood of delta stability against the sea level rise. A more pronounced increase of rainfall in the post-monsoon in comparison to the monsoon season may aggravate the flooding situation in the downstream deltaic region as the catchment is already saturated with monsoonal rainfall, so opportunities for abstraction loss are much less in the post-monsoon season. This may cause an increased frequency of flooding in the post-monsoon season in the near future as compared to monsoonal floods. Historically, only 6 of 41 monsoonal flood events have occurred in the first half of October from 2005 to 2013 (Ghosh et al. 2019). Precipitation and ET, respectively, were found to be the dominant drivers of streamflow generation in post-monsoon and winter seasons under the changing climate. The major takeaways of this study are as follows:

• (1)

  Although the study area does not have any major reservoirs, release from the Hirakud reservoir which lies upstream of the study area may worsen the flooding situation in the already vulnerable delta region. Therefore, this study provides important input to reservoir operation planning of Hirakud dam under the potential future climate change condition. Monthly and seasonal projections of evapotranspiration, streamflow, and SSL under most recent CMIP6-SSP scenarios may be helpful in adapting the rule curve of the reservoir in view of the changing climate. As the highest increase in rainfall, streamflow, and SSL are expected in the post-monsoon season, instead of the monsoon season, the rule curve may be tweaked to provide additional flood control reservoir storage for mitigating the flood hazard.

• (2)

  Seasonality of rainfall and ET are projected to be more amplified, i.e. wet season may get wetter while dry season may get drier. This may lead to flood-drought alteration events in the MRB, especially in the near future. As the study area is predominantly agricultural (55% of basin area) with the major crop rice grown throughout the year, this study is helpful in adapted irrigation/agricultural planning in view of the projected flood-drought situation.

• (3)

  The projected change in SSL reveals the severity of the erosion issues in the catchment. Waterbodies may become more silted as a result of these changes. Geomorphological changes in rivers and floodplains may also have an adverse effect on riparian ecosystems and communities. Intricate sediment management techniques such as best management practices may become more necessary as a result of the possible increase in erosion, which could potentially impede agricultural expansion. These anticipated scenarios offer substantial guidance for natural resource planning and management, enhancing the environmental, social, and economic well-being of the catchment.

Concerning streamflow and SSL in the flood-prone river basin, this work offers a valuable assessment of the consequences of a changing climate under the latest CMIP6 model projections. By advancing knowledge of adaptability in the context of a changing climate, the findings may also help put into practice and plan streamflow and sediment management measures. Additionally, the study is limited by the uncertainties associated with the emission scenario, GCMs, bias-correction technique, and hydrological modelling. Uncertainty due to the choice of hydrological model may be mitigated through the multi-model approach (Georgakakos et al. 2004). This also facilitates the identification of a better model for large-scale catchment sediment simulation. The copula-based zero inflated rainfall bias-correction method may be employed to mitigate the error due to the choice of bias correction as this technique is found to be better in preserving both the mean and extreme rainfall characteristics for a geography characterized by tropical monsoon-type rainfall pattern with a large proportion of daily rainfall time-series data of zero values (Maity et al. 2019; Suman & Maity 2021). Therefore, one of the critical goals of our future study should be to investigate uncertainty and improve confidence in the accuracy of the climate change impact outcome of flood-prone basin. The effect of future LULC change is not considered in our study. Several previous studies have demonstrated that climate change has greater impacts on catchment behaviour than land use change in the large river basin (Wagner et al. 2013; Ma et al. 2021; Sreedevi et al. 2022). However, changes in land use do affect the hydro-sedimentological process of the basin. Therefore, it is essential to gain an understanding of independent effects of LULC and climate change on hydro-sedimentological fluxes as well as their combined effects to lessen soil erosion during monsoon season and maintain water availability during the rest of the year.

The authors would like to acknowledge the Central Water Commission (CWC), Bhubaneswar and India Meteorological Department (IMD) for providing the necessary data to carry out this research. The data used in this study can be accessed by these agencies after fulfilling their data-sharing policy. The climate projection data have been collected from https://zenodo.org/record/3987736#.Y34U0nZBxPY. The fellowship to the first author provided by the Ministry of Human Resources Department (MHRD), Government of India is highly acknowledged.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.