Changes in flow and sediment load of poorly gauged Brahmaputra river basin under an extreme climate scenario

Bangladesh, the lower riparian country of Ganges–Brahmaputra–Meghna (GBM) river basins, is highly vulnerable to flood, storm surge, cyclone, erosion and other natural disasters. Bangladesh contributes only a 7% area of 1.7 million km² of GBM basins. Although Bangladesh encompasses a very small part of that large area, the amount of discharge and sediment flow from the upstream countries passed through Bangladesh towards the Bay of Bengal are relatively very high. The GBM river system is the third-largest freshwater outlet to the world's oceans (Chowdhury & Ward 2004). The Brahmaputra River Basin (BRB) is the second-largest basin of the GBM basins and carries the highest discharge and sediment load. The mean annual discharge of the Brahmaputra river (also known as Jamuna river in Bangladesh) is about 20,000 m³/s (Immerzeel 2008). Massive sediment loads yield from the Brahmaputra river every year which varies from 590 Mt/year (FAP 1996) to 792 Mt/year (Islam et al. 1999) as estimated in several previous studies.

In the past, several studies have developed different modeling systems for assessing the water resources of the GBM Basins (Nishat & Rahman 2009; Akbor et al. 2014). Ghosh & Dutta (2011) used a single projected climate scenario and found that the projected climate change can increase the peak flow of the Brahmaputra by about 28%. In a study by Gain et al. (2011), a daily discharge time series has been constructed by using a discharge-weighted ensemble based on the inputs from the 12 General Circulation Models (GCMs). Then, these data have been used to a global hydrological model and the results show that climate change is likely to improve dry season conditions in the Lower BRB. Ghosh & Dutta (2012) analyzed the impact of climate change on the Tezpur, Guwahati and Dhubri stations of the BRB using a macro-scaled distributed hydrological model. This study revealed the increase in peak flow and flow duration during the pre-monsoon (March, April, May) and monsoon (June, July, August, September) seasons for future climate change. Although changes in time of peak during pre-monsoon might increase, the chance of time of peak during monsoon might be the same. Paul (2014) used Soil & Water Assessment Tool (SWAT) to set up a semi-distributed hydrological model of the BRB for assessing the future flow of BRB using Quantifying Uncertainty in Model Predictions (QUMP) ensembles experiments of Providing REgional Climates for Impacts Studies (PRECIS) for the early century (2011–2040), mid-century (2041–2070) and end century (2071–2099). In this study, the change in monthly flow for August has increased to 4, 5 and 8% for the years of 2020s, 2050s and 2080s, respectively. Darby et al. (2015) found an increase of 25–30% in sediment load from the BRB considering variants of the SRES A1B emissions scenario. Alam (2015) analyzed the impact of climate change on streamflow of BRB using a physically-based semi-distributed hydrological model, SWAT. Another study by Alam et al. (2016) investigated the sensitivity of the discharge of the Brahmaputra river to climate change and found that changes in discharge are almost linearly related to changes in precipitation and temperature. An increase in mean annual streamflow in the 2020s, 2050s and 2080s for BRB has been found in this study. Considering land use and climate change scenarios, Pervez & Henebry (2015) assessed changes in the future freshwater availability in the BRB. Mohammed et al. (2017) considered bias-corrected data from an ensemble of 11 climate projections with the Representative Concentration Pathway (RCP) 8.5 from the Coordinated Regional Climate Downscaling Experiment-South Asia (CORDEX-SA) database. They found an increase and a decrease in mean monthly discharges during pre-monsoon months and post-monsoon months, respectively, for the BRB. Results show that floods are likely to become more frequent in the future, and their magnitude will also become more severe. The average timing of floods is projected to shift earlier compared to the present hydrological regime.

To make the best use of water resources of the BRB, engineers, as well as policymakers, need to know the future changes in both sediment load and flow. However, little is known regarding the impact of climate change on the sediment regime of the BRB. Sediments contain essential nutrients and material for ecosystems and agricultural lands and it is vital from environmental, economic and social perspectives (Apitz 2012). The quantity and quality of sediments are affected due to natural variability in hydrological conditions, as well as changes in land use, water use and climate (Chalov et al. 2015). Changes in sediment loads in the future considering alteration of climatic variables need to be assessed to control and manage sediment flows in large catchments properly, especially where people's livelihood depends on the river systems and their natural processes. Therefore, it is obvious to understand how a change in the global climate could affect regional water supplies and the amount of sediment load. A one-dimensional model Hydrologic Engineering Center-River Analysis System (HEC-RAS) has been used by Fischer et al. (2017) considering both present and future climatic conditions for estimating annual sediment load of the BRB. At Bahadurabad Transit (SW 46.9 L, the downstream gauging station of BRB in Bangladesh), the annual sediment load was found to vary from 260 to 720 Mt/year which indicates considerable uncertainty (Fischer et al. 2017). On the other hand, Rahman et al. (2018) analyzed Bangladesh Water Development Board (BWDB) (period: 1982–2000) and Flood Action Plan 24 (period: 1993–1995) data, plotted together with secondary data and found a downward trend of 6 Mt/year in the sediment flux for the BRB. However, none of the existing studies has used the new generation of high-resolution downscaled projections of the Coupled Model Intercomparison Project Phase 5 (CMIP5) models by the CORDEX-SA to assess their performance in simulating sediment load for the BRB.

This study focused not only on changes in discharges from the BRB in the 2020s, 2050s and 2080s but also on the sediment yield from this basin in the future considering climate change. The RCP 8.5 climate change scenario, which represents the highest CO₂ concentrations in the atmosphere and the highest rise in global temperatures out of the four concentration scenarios defined within the Fifth Assessment Report (AR5) of the IPCC (IPCC 2013), has been considered for impact analysis on the flows and sediment load of the BRB. The key objectives of this present study are to (1) calibrate and validate the HEC-HMS model for the BRB using available observed data of the baseline period, (2) use the calibrated models to simulate future river discharges and sediment load for the respective climate projections, (3) analyze the simulated future discharges to estimate the changes in monthly, seasonal and annual flow and (4) analyze the simulated future sediment loads to estimate the changes in monthly, seasonal and annual sediment loads.

The BRB covers a catchment area of about 543,400 km² (Joint Rivers Commission Bangladesh-JRCB 2019) of which 50% lies in China, 36% in India, 7% in Bangladesh and 7% in Bhutan. Figure 1 shows the watersheds and major river networks of the BRB. The BRB is categorized into three different physiographic zones: Tibetan Plateau, Himalayan Belt and the floodplain (Immerzeel 2008). The major river of this catchment, the Brahmaputra–Jamuna, is originated in the Himalayas and passes through China, India and entering Bangladesh through the northern boundary after traveling for about 3,000 km. The average bed slope of this river is 7.5 cm/km (CEGIS 2010). The climate of the basin is monsoon driven with a distinct wet season from June to September, which accounts for 60–70% of the annual rainfall. In 1998, the highest recorded peak flow at Bahadurabad Transit station (SW 46.9 L) of the Brahmaputra River was 102,534 m³/s (Mirza et al. 2003). The mean annual sediment load was about 555 Mt/year which was measured during 1966–1969 by the BWDB. The total suspended load was about 610 Mt/year according to Coleman (1969). Most of the sediment is composed of silt and transport as a suspended load while around 15–25% are of sand which transports as a bed load. The size of bed material (d₅₀) at Bahadurabad Transit (SW 46.9 L) has an average value of 0.20 mm (Sarker 2009).

Hydrological models like SWAT, HEC-HMS, Variable Infiltration Capacity (VIC), Hydrologiska Byråns Vattenbalansavdelning (HBV), MIKE-Système Hydrologique Européen (MIKE-SHE) are widely used for catchment routing and watershed analysis. Among them, SWAT, HEC-HMS and MIKE-SHE models have surface erosion computing components that measure erosion due to rainfall or surface runoff. In these models, the Universal Soil Loss Equation (USLE) or Modified Universal Soil Loss Equation (MUSLE) is applied to compute surface erosion. Both SWAT and HEC-HMS have a sediment routing component which represents sediment erosion and deposition processes within rivers. In this study, HEC-HMS has been used for simulating continuous rainfall–runoff processes and sediment loads of the BRB.

Geospatial data, such as elevation, land cover and soil types, are collected from different global geospatial repositories. To delineate the catchment area and river network for HEC-HMS modeling, Digital Elevation Model (DEM) with 90 m resolution from Shuttle Radar Topography Mission (SRTM) is used as shown in Figure 2. Land cover map of 500 m resolution has been collected from the United States Geological Survey (USGS). Land cover maps have been developed from the analysis of Moderate Resolution Imaging Spectroradiometer (MODIS) images for the period 2001–2010. The digital soil map with a scale of 1:5,000,000 of the year 2007 has been collected from the Food and Agricultural Organization.

The HEC-HMS model requires daily values of precipitation, average temperature, solar radiation, relative humidity and wind speed. Meteorological data over the BRB have been collected from Prediction of Worldwide Energy Resource (POWER) of the National Aeronautics and Space Administration (NASA) (http://power.larc.nasa.gov) for the climate normal period 1983–2010. Bahadurabad Transit (SW 46.9 L) has been used as the outlet of BRB modeling which has also been considered as the calibration, as well as validation point. Discharge and sediment load of this station have been collected from BWDB (BWDB) for the period 1983–2010. Future precipitation and temperature data of RCP8.5 scenarios have been collected from the CORDEX-SA domain database for the period 1971–2099. The Institute of Water and Flood Management (IWFM) of Bangladesh University of Engineering and Technology (BUET) has provided these RCP8.5 scenarios data.

The watershed and stream network of BRB has been delineated using Arc Hydro tools (a version that works with Arc GIS 10.1). The output files from terrain pre-processing have been used to create input files for the HEC-HMS models using the HEC-Geospatial Hydrologic Modeling Extension (HEC-GeoHMS). Thiessen polygon method is used to achieve accurate estimation of the spatial distribution of rainfall which gives the average precipitation over an area. Penman–Monteith (Zotarelli et al. 2010) method has been selected to calculate evapotranspiration. As Soil Moisture Accounting (SMA) can simulate runoff volume, both wet and dry weather conditions, it has been used for continuous runoff simulation. Considering the characteristics of BRB as a data-poor basin, Snyder's Unit Hydrograph has been selected for separating direct runoff. The linear reservoir model which is consistent with the SMA algorithm has been used to simulate total base flow for the watershed. For capturing the sediment load in a reach, it is necessary to represent the river network like natural channels. The Muskingum–Cunge model has been used for routing the channel flow.

The widely used MUSLE has been selected to simulate erosion from surface runoff. The sediment transport capacity of the flow can be calculated from the flow parameters and sediment properties. If the sediment transport capacity at a section is greater than the available sediment in streamflow, the erosion from stream bed will occur. If the sediment transport capacity is less than available sediment in streamflow, the deposition will occur to the reach bed. Among the available methods for estimating transport potential function (Scharffenberg 2015) for a reach in HEC-HMS, Engelund Hansen method (Engelund & Hansen 1967) has been used for all streams. This equation has been found useful to estimate sediment transport with considerable accuracy for Brahmaputra/Jamuna river by FAP 24 (1996). The volume ratio method has been applied to link the transport of sediment to the transport of flow in the reach. The estimated parameters from previous literature and data analysis used in the BRB model are presented in Table 1. HEC-HMS model has been simulated at a daily time step for all purposes, i.e., for calibration, validation and simulation of future discharges and sediment load.

Calibration and validation of the HEC-HMS model have been conducted by estimating errors through different indices. The Nash Sutcliffe Efficiency (NSE) determines the relative magnitude of the residual variance compared to the measured data variance (Nash & Sutcliffe 1970). NSE values vary between 1 to −∞ where a value of 1 is the optimum value. Percent bias (PBIAS) measures the average tendency of the simulated data to be larger or smaller than their observed counterparts. The optimal value of PBIAS is 0.0 (varies between −∞ to ∞), with low-magnitude values indicating accurate model simulation. Positive values indicate model underestimation bias, and negative values indicate model overestimation bias (Gupta et al. 1999). RMSE is one of the commonly used error-index statistics (Chu & Shirmohammadi 2004). RSR is calculated as the ratio of the RMSE and the standard deviation of measured data. RSR values varied between 0 and ∞, where 0 is the optimum value. The general reported ranges for NSE, PBIAS and RSR for calibration and validation processes are stated by Moriasi et al. (2007) and it has shown in Table 2.

Initially, the model has been simulated with the base data collected and estimated from the literature review (Coleman 1969, Goswami 1985, Jung et al. 2010, Mahanta et al. 2014, Rawls et al. 1982) and GIS analysis, respectively, and Table 1 represents the base data. After the first experiment, the NSE has been found as 0.16 which is not within the satisfactory range. Later, a total of nine parameters of the SMA algorithm have been considered in the manual calibration of the model and optimize the parameters. In the calibration process, the selected parameters and variables of the model are adjusted to make the model output match observations. Sensitivity analysis of the initial run of the model is performed for selecting the parameter which should be optimized first in manual calibration. In the sensitivity analysis, out of the nine SMA parameters, one parameter at a time is varied and analyzed from −40 to 40% with increments of 10%, keeping all other parameters constant. The output runoff volume values have been analyzed to determine variation for the initial estimates of the parameters. A greater percentage change in the simulated volumes represents a greater sensitivity of the selected parameter. In the case of the sediment model, most of the parameters have been collected and estimated from secondary literature (Coleman 1969, Scharffenberg 2015) and GIS analysis, respectively. Manning's n, the property of river is one of the major calibration factors for both discharge and sediment load simulation and it has been calibrated during the surface runoff calibration.

An 11-member ensemble from three Regional Climate Models (RCMs) forced by eight GCMs has been selected from the CORDEX-SA domain database. The selection of the ensemble members is based on the availability in the database at the time of the study. All the projections have been made by forcing an RCM using a GCM with the RCP 8.5 climate scenario (van Vuuren et al. 2011). This scenario represents no change in the current trend of greenhouse gas emissions. So far, the actual trend in emission is also found to follow this pathway (Friedlingstein et al. 2014; Piontek et al. 2014). Changes in annual average precipitation for the periods of 2010–2039, 2040–2069 and 2070–2099 from the climatic base period (1980–2009) for all three RCMs forced by eight GCMs have been analyzed and presented in Table 3. Specific models are selected for impact analysis on sediment load considering driest, moderate and wettest conditions in the future. The RCM SMHI-RCA4 forced by GCM IPSL-CM5A-MR has shown that the percentage increase in precipitation will be 11.8, 15.4 and 33.5% for the 2020s, 2050s and 2080s, respectively. These values are the highest among all the models. Hence, this model has been considered to produce the wettest condition scenarios in which the highest precipitation will occur in the future. The RCM REMO2009 forced by the GCM MPI-M-MPI-ESM-LR_MPI-CSC model has shown that the percentage decrease in precipitation will be 0.7, 0.9 and 8.9% for the 2020s, 2050s and 2080s, respectively, and is regarded as the driest model. The RCM SMHI-RCA4 forced by GCM MPI-M-MPI-ESM-LR has shown moderate condition scenarios of precipitation in the future.

In a sensitivity analysis of SMA parameters, the percentage changes in simulated volume have been plotted against the percentage variation of each SMA parameter and it is shown in Figure 3. In this study, soil percolation has been found as the most sensitive parameter for simulated streamflow during the calibration period. On the other hand, Ground Water 2 (GW2) storage has been found as the least sensitive parameter. Each parameter has been ranked according to their sensitivity for the change in simulated volume, as shown in Table 4. Therefore, among the SMA parameters, the value of soil percolation has been calibrated first and Ground Water 2 (GW2) storage has been adjusted last during the calibration process through manual calibration.

The model has been simulated for a 30-year baseline period from 1980 to 2009 to assess changes in flow and sediment load in the future. Therefore, data up until 2010 have been used for model calibration and validation. The surface runoff model has been calibrated at Bahadurabad Transit (SW 46.9 L) for the daily runoff of 1983 to 1996. In calibration, the NSE value has been found as 0.65. PBIAS and RMSE observations standard deviation ratio (RSR) has been found as −20.92 and 0.59, respectively. The surface runoff model has been validated for the 1997–2010 period. During validation, NSE, PBIAS and RSR values are found as 0.54, −23.40 and 0.68, respectively. The coefficient of correlation (R²) of calibration and validation period is found as 0.76 and 0.72, respectively. These statistics have been calculated by using simulated and observed runoff values on a daily basis and the values of these statistics demonstrate that the developed hydrological model using the HEC-HMS lies in the satisfactory region in both calibration and validation stages (Figure 4) based on historical measured data for BRB. The ranges of these statistical parameters can be found in Moriasi et al. (2007), which establishes the basis for developing a sediment model to generate a continuous sediment discharge at Bahadurabad Transit (SW 46.9 L).

The simulated sediment load for corresponding dates of observation at Bahadurabad Transit (SW 46.9 L) has been plotted in Figure 5 and coefficient of correlation (R²) is found as 0.6 which provides a satisfactory level for the surface erosion and sediment routing model. The simulated mean annual sediment load at Bahadurabad Transit (SW 46.9 L) from the calibrated and validated HEC-HMS model is about 370.2 Mt/year. This is very close to the estimated sediment load, i.e., 368 Mt/year, using 1D HEC-RAS at Bahadurabad Transit (SW 46.9 L) of the Brahmaputra river by Fischer et al. (2017). More specifically, for the period 1993–1996, the simulated sediment load from the model is about 379.1 Mt/year, which is close to the sediment load, i.e., 402 Mt/year, estimated by Delft Hydraulics (FAP 24 1996). Some of the major Dams (Zangmu DAM, Dagu DAM, Jiexu DAM and so on) in Brahmaputra river have been constructed after 2010 (Source: https://en.wikipedia.org/wiki/Zangmu_Dam) and this model could not incorporate the Dam operation due to lack of operation data, therefore, HEC-HMS model for BRB has been calibrated and validated for the period 1983–2010.

The monthly mean flow in the baseline period and future change in flow from baseline at Bahadurabad Transit (SW 46.9 L) for the 2020s, 2050s and 2080s are shown in Figure 6. In all three projections, the highest discharge will occur in July ranging from 52,000 to 96,000 m³/s. In the 2020s, 2050s and 2080s, the changes in the percentage of flow are highest during March, March and April and the values are 350, 385 and 350% (Supplementary Material, Table 2, Table 4 and Table 6), respectively, however, the contribution of flow in these months is less compared to the months in the monsoon season.

Changes in flow are analyzed for pre-monsoon (March–May, MAM), monsoon (June–September, JJAS), post-monsoon (October–November, ON) and dry (December–February, DJF) seasons. The volume of seasonal flow and change in the seasonal flow at Bahadurabad transit for the 2020s, 2050s and 2080s projections from baseline period are shown in Figure 7. The quartiles of the pre-monsoon (MAM) discharges will be increasing from the 2020s towards the 2050s and 2080s. The seasonal flow will be increased to 94, 150 and 245% for the 2020s, 2050s and 2080s, respectively. The inter-quartile range of the monsoon (JJAS) discharges has been found to be increasing from the 2020s towards the 2050s and 2080s. Changes in seasonal flow for the 2020s, 2050s and 2080s from baseline will be 14, 35 and 57%, respectively. The highest changes in flow will be observed in the pre-monsoon seasons in the future, however, the contribution in flow during pre-monsoon season is less compared with the monsoon season.

On the other hand, for post-monsoon (ON) and dry (DJF) seasons, the change in the inter-quartiles of discharges will be decreasing from the 2020s towards the 2050s and 2080s. For post-monsoon, change in seasonal flow for the 2020s, 2050s and 2080s from baseline are 11, 37 and 40%, respectively, and for the dry season, these values are 23, 71 and 75%, respectively.

The differences in mean annual streamflow simulated by the ensemble members forced by GCMs and RCMs to the climate normal period are shown in box plots of Figure 8. The percentage changes in mean annual flow from the baseline period for the 2020s, 2050s and 2080s are 18, 27 and 61%, respectively. A gradual increase in mean annual flow has been found from the 2020s to the 2080s. The inter-quartile ranges between the 75th and 25th percentile of the box plot for the 2020s to 2080s indicate the uncertainty associated with the projected streamflow of BRB increases as we go into the distant future.

The amount of monthly sediment load and future change of sediment load from baseline at Bahadurabad Transit (SW 46.9 L) for the 2020s, 2050s and 2080s projections are shown in Figure 9. In the 2020s, the highest sediment load (Mt/month) values will occur in August which will be ranging from 18 to 147 Mt/month. However, the highest sediment load will occur in July in both the 2050s and 2080s. The amount of the highest sediment load will be varying from 60 to 182 Mt/month for the 2050s and from 55 to 223 Mt/month for the 2080s, respectively. The difference of maximum and minimum range of sediment load (Mt/month) is highest for March for the 2020s and 2050s while in May for the 2080s. For the 2020s, 2050s and 2050s, the values of the highest changes in the percentage of sediment load are 710, 610 and 1,200%, respectively. Though high-level uncertainty has been found for these months, the amount of sediment load is very minor in these months. Therefore, it has a very low impact on total sediment load considering the other months.

Change in seasonal sediment load from baseline at Bahadurabad transit (SW46.9 L) of BRB has been assessed. The volume of seasonal sediment load for the 2020s, 2050s and 2080s are shown in Figure 10. Changes in the pre-monsoon (MAM) and monsoon (JJAS) sediment loads are found to be increasing from the 2020s towards the 2050s and 2080s projections. Seasonal sediment load increases of the 2020s, 2050s and 2080s are found as 182, 441 and 1,170%, respectively, during the pre-monsoon season. On the other hand, changes in seasonal sediment load during the monsoon period for the 2020s, 2050s and 2080s are found as 40, 74 and 111%, respectively. In the post-monsoon (ON) and dry (DJF) season, the change in the quartiles' seasonal sediment load is found to be decreasing from the 2020s towards the 2050s and 2080s. Sediment load will be changed for the 2020s, 2050s and 2080s are 61, 83 and 92% during post-monsoon and these values are 36, 98 and 54% for the dry season, respectively.

Differences in mean annual sediment load are described in the box plots as shown in Figure 11. The percentage changes in mean annual sediment load for the 2020s, 2050s and 2080s from baseline period are 34, 67 and 115%, respectively. A gradual increase in mean annual flow and sediment load has been found from the 2020s to the 2080s. The inter-quartile ranges (between 25th and 75th percentile) of the box plot indicate the uncertainty of the projected changes and it is increased with time.

In this study, a hydrological model which includes both surface runoff and sediment routing models has been developed for the BRB using an open source-based model, HEC-HMS. The coefficient of correlation (R²) values are 0.76 for calibration and 0.72 for the validation period, respectively, and have shown a satisfactory correlation between simulated and observed discharge at Bahadurabad Transit (SW 46.9 L). NSE, PBIAS and RSR values are 0.65, −20.92, 0.59 and 0.54, −23.40, 0.68, respectively, for calibration and validation of the hydrological model which is within the acceptable range. Surface erosion and sediment model has been calibrated from 1983 to 1996 and validated from 1997 to 2010. It has been found that the coefficient of correlation (R²) is found as 0.6 which lies in a satisfactory level and the mean annual sediment load at Bahadurabad transit for BRB was about 370.2 Mt/year for 1983–2010. More specifically, from 1993 to 1996, the mean annual sediment load was 379.1 Mt/year, which is close to the sediment load, i.e., 402 Mt/year, estimated by Delft Hydraulics (FAP 24 1996). Using the calibrated and validated model, both streamflow and sediment load are generated at the Bahadurabad Transit (SW 46.9 L) for the base period (1980–2009) and three future periods (viz., the 2020s (2010–2039), 2050s (2040–2069) and 2080s (2070–2099)).

Percentage changes in mean annual flow and sediment load at Bahadurabad transit (SW 46.9 L) for the 2020s, 2050s and 2080s from baseline period are found as 18, 27, 61% and are 34, 67, 115%, respectively. The inter-quartile ranges between the 75th and 25th percentile of the box plot from the 2020s to 2080s indicated uncertainty associated with the projected flow and sediment load will be increased with time in the future. Uncertainty in the change in seasonal discharge and seasonal sediment load in pre-monsoon will be higher than the monsoon in the future; however, the contribution in flow and sediment load during the pre-monsoon season is less compared with that of the monsoon season. The highest sediment load (Mt/month) values will occur in August, ranging from 18 to 147 Mt/month for the 2020s. On the other hand, the highest sediment load (Mt/month) values will occur in July for the 2050s and 2080s and the values are varying from 60 to 182 Mt/month and from 55 to 223 Mt/month, respectively.

The engineers, planners and policymakers can get an idea about the plausible changes in flow and sediment load of the Brahmaputra river considering high emission scenarios. Results of this study would help to plan and design climate-resilient infrastructures, such as bridges, regulators, embankments, etc. Many sediment management-related studies, such as the planning and design of capital dredging for navigational channels, bank protection works and analysis of the stability of river islands (char), would be benefited from the findings of this study. Improvement of calibration of surface erosion and sediment model can be done by regularly measured sediment load data at different parts of the catchment. A recent land-use map derived from high-resolution satellite images could improve the accuracy of the model.

The climate projection data have been collected from the High-End cLimatic Impacts and eXtremes (HELIX) project which received funding from the European Union Seventh Framework Programme FP7/2007–2013 under the grant agreement no. 603864 (HELIX: High-End cLimate Impacts and eXtremes; http://www.helixclimate.eu).

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