Impact assessment of climate change on water resources of the Kokcha watershed: a sub-basin of the Amu Darya river basin in Afghanistan

Abbreviations

Definitions

GCM

General Circulation Model

IPCC

Intergovernmental Panel on Climate Change

LULC

land use/land cover

CC

climate change

LCC

land cover change

MEW

Ministry of Energy and Water

NAPA

National Adaptation Program of Action for climate change

UNEP

United Nations Environmental Program

USDA

United States Department of Agriculture

SW_t

final soil water content

SW₀

initial soil water content

JAM-STEC

Japan Agency for Marine-Earth Science and Technology

NIES

National Institute for Environmental Studies

SWAT-CUP

SWAT Calibration and Uncertainty Program

SUFI-2

Sequential Uncertainty Fitting version 2

STRM

Shuttle Radar Topography Mission

UN-FAO

United Nations Food and Agriculture Organization

DEM

Digital Elevation Model

TRMM

Tropical Rainfall Measurement Mission

NASA POWER

National Aeronautics and Space Administration Prediction of Worldwide Energy Resources

USGS

United States Geological Survey

Symbols

Definitions

t

time

Q_surf

runoff

e_a

actual vapor pressure

RCM

regional climate model

PET

potential evapotranspiration

E_a

evapotranspiration

P_i

the mean value of simulated values

HRUs

hydrological response units

R_day

precipitation

NSE

Nash–Sutcliff Efficiency

PBIAS

Percentage of Bias

R²

coefficient of determination

Q_gw

return flow

W_perc

percolation

G

soil heat flux density

γ

psychometric constant

Δ

slope of the saturation vapor pressure function

T_mean

daily mean temperature

u₂

mean wind speed at 2 m

mean value of the observed values

R_n

net radiation

In the 21st century, climate change (CC) is an important issue for scientists and water managers. The consequences of global warming are mostly water-related, besides heat wave occurrences, illnesses, and fatalities (Ekwueme & Agunwamba 2021). Nowadays, nearly 3.97 billion people were facing water shortage at least a month per year, and 0.5 billion people worldwide faced extreme water scarcity between 1996–2005 (Hussainzada & Lee 2021 cited Mekonnen & Hoekstra 2016). Food security is already affected by CC in arid and semi-arid regions (Jabal et al. 2022). Changes in the cryosphere can impact water resources, hydropower systems, irrigation, and urban areas due to variations in stream discharges, extreme floods, and landslides (Nivesh et al. 2022). CC impacts every region of the world differently, based on the geographic location and weather conditions of areas. Considering Asian countries, in locations of South Asia, there may be required dissimilar adaptation policies than western parts of this continent. Topography and elevation characteristics have their roles in CC, parts of the Himalayas mountain ranges and high altitudes are covered by permanent snow and glaciers that are extremely susceptible to temperature rise.

Southern regions of Asia which depend on monsoon rainfall (Aich et al. 2017) are a concern for CC specialists. In the Hindu Kush mountains area, the average temperature will change between 1–3.2 and 2–4.1 °C under the effects of the RCP4.5 scenario in the periods of 2006–2050 and 2006–2099, respectively. The temperature changes considering the RCP8.5 scenario will be 1.7–3.8 and 5.3–10.3 °C in the above-mentioned periods accordingly (Aich et al. 2017). Most rivers in the Himalayan region countries are originating from the snowmelt of higher mountains. The temperature rise in these areas can cause an increase in runoff in the basins which will be more effective than precipitation increases in the area. A study on the Lidder watershed considering different scenarios represented a 53% runoff increase in (April–September) considering a 2 °C degree temperature rise. However, with a 20% increase in precipitation, the runoff showed a 37% increase, and with a 2 °C temperature rise and 20% growth in snow-covered areas, the discharge represented a 67% increase (Kumar et al. 2022). The study shows the role of snowmelt in streamflow which is much more than precipitation, and for every degree of temperature rise, there will be a large amount of discharge in the rivers. However, the generated river flow from melting snow is not sustainable and the consequences are drought in the near and far future. In arid and semi-arid regions, drought is a serious problem that challenges developing countries and especially agrarian-based countries socially and economically (Nivesh et al. 2022). The changing climate can degrade the quality of river flow by both floods and reducing river discharge (Badrzadeh et al. 2022), which can cause health diseases, economic crises, and social conflicts. Therefore, the management of water resources considering CC is necessary.

There are many models and projections of climate data for the evaluation of CC; however, the General Circulation Model (GCM) is commonly utilized for the prediction of CC scenarios (IPCC 2007). Studies in different parts of the ground were conducted to simulate the streamflow under CC using the projected climate scenarios. Under the effects of CC in eastern Iran, a semi-arid region, the runoff was shown to have a decreasing trend in the mid to far future, and throughout the far future considering RCP4.5 and RCP8.5 scenarios (Rahimpour et al. 2021). In the Kaidu watershed which is located in a cold-arid region, in northwest China, temperature rise mainly represents the warming trend in the winter and spring by 2099. The runoff showed an increasing trend across the 21st century which is caused by precipitation increment in the spring and winter seasons (Sun et al. 2020). In addition to CC, land use/land cover (LULC) can influence the generation, quantity, and durability of streamflow in watersheds. Chen & Chang conducted a study to compare the impacts of CC and land cover change (LCC) on the Clackamas catchment in Oregon; the obtained result indicated the sensitivity of runoff in CC more than in LCC in this watershed. In the Clackamas watershed, the precipitation pattern varied from snowfall to rainfall during the winter. In addition, the peak flow timing was shifted 2–3 weeks earlier because of the warming condition in the catchment area (Chen & Chang 2021). In another research performed by Akhundzadah et al. (2020), in the Kunduz river basin, the average temperature rose while the mean precipitation and river discharge showed descending trends on the lower sides. However, the rising trend in river flow on the upper sides of the watershed represents the melting of snow/glaciers that existed on the catchment for a long time. The increased runoff due to snowmelt is not permanent and with the shrinkage of available glaciers in the high mountains, the river may face an extreme drought in the future. These studies manifest the effects of CC on water resources that are dissimilar in different areas.

Among developing countries, ‘Afghanistan is one of the most vulnerable countries to CC’ (Aich et al. 2017). The water resources in this country originate mostly from melting snow and glaciers which are extremely susceptible to temperature rise. The mean annual temperature in Afghanistan rose around 0.6 °C since 1960, while the mean precipitation decreased slightly (McSweeney et al. 2010). According to estimates made public by Afghanistan's Ministry of Energy and Water (MEW), from 1990–2015, the country lost 406.16 km² of its glaciated area in the higher mountains which counted as 13.8% of its total. These losses increased rapidly in recent years by 3.6% in 1990–2000, 4.7% in 2000–2010, and 6.25% in the years 2010–2015. The largest portion of river flow in this country arises from the Hindu Kush mountain ranges, an important tributary of the Amu Darya river basin (Barlow & Tippett 2008). Contributions of snowmelt in streamflow could play a key role in years with low precipitation. Changes in precipitation time and severity can cause flash floods which are so risky in the lowlands socially and economically. In the year 2022, there were 256 fatalities in addition to the demolition of thousands of households by flash floods that happened mostly in July and August in Afghanistan. It demonstrates a 75% increase when 147 persons lost their lives in the province of Nuristan in 2021, including 127 in a single incident (OCAH Afghanistan 2022). These disasters are symptoms of CC and can be intensified in the future if appropriate adaptation policies would not be considered. Also, flood early warning systems are required downstream. From machine learning and artificial intelligence methods, fuzzy-based methods are the most suitable for early warning and monitoring systems (Suwarno et al. 2021). The expected mean temperature rise in Afghanistan by 2070–2099 causes more rainfall than snowfall (Christensen et al. 2007). Based on the National Adaptation Program of Action for Climate Change (NAPA), the main climatic hazards such as drought, floods, warming, heat/cold waves, thunder, and monsoon wind are exacerbated by CC (UNEP 2009). Although it is projected to increase the amount of precipitation in the higher altitudes including the Hindu Kush mountains, the change in precipitation patterns can damage the regularity of flow in the rivers. The effects of CC on water resources vary across different regions (IPCC 2014). The impacts of global warming in every catchment should be investigated separately, and proper adaptation policies should be considered for watersheds and regions.

Watershed management can help to manage the issues of soil erosion, floods, CC, and water scarcity (Sarkar et al. 2022). The Kokcha watershed is an important tributary of the Amu Darya river basin which has a significant role in the water management of Afghanistan. There are many planned water infrastructures in the Kokcha watershed. So, it is significant to investigate the effects of CC on the water resources of this watershed. In recent years, some studies conducted about water management considering CC in different parts of Afghanistan. However, there is a very limited study to assess the effects of CC on the Kokcha watershed. Whereas, investigation of CC in this catchment is necessary due to being partly covered by snow and glaciers and having a crucial role in the country's economy. The main purpose of this study was to assess the impacts of CC on the Kokcha watershed.

Kokcha watershed is positioned in the northern regions of Afghanistan amid the 35.436°–36.463° Latitude and 69.481°–71.652° Longitude, it has a 20,139 km² area (Figure 1). The Kokcha river is one of the country's richest rivers, mostly fed by melting snow in the spring and summer months. Maximum and minimum temperatures in the Kokcha catchment vary from −33 °C in the midwinter to 24.5 °C in the summertime. The minimum and maximum flow in the Kokcha river occurs in the winter and summer, respectively. Throughout the year, mean monthly discharge is varying from a low of 62.5 m³/s in January to a maximum of 510 m³/s in July along history, the minimum, and maximum observed flow in this river was 42.5 m³/s in January 1973 and 748 m³/s in June 1978, respectively.

The upper sides of the study area include the Hindu Kush mountains are partly covered by permanent snow and glaciers. The temperature in Afghanistan is varying, and it is said to be the temperature rise over all of Afghanistan in the timespan of 1951–2010 was ‘1.8 °C, with the highest increase of 2.4 °C in the eastern regions and only 0.6 °C in the Hindu Kush’ mountains (Aich et al. 2017) where the Kokcha watershed is located.

Based on topography characteristics, the SWAT model divides the basin into multiple sub-basins and then into smaller cells called hydrological response units (HRUs), this division will happen according to LULC, soil maps, and slope which is homogenous in each HRU (Arnold et al. 2012). Required input data in the SWAT model are the digital elevation model (DEM), soil map, and LULC as spatial data, also weather data such as precipitation, maximum and minimum temperature, relative humidity, solar radiation, and wind speed in daily or semi-daily time step are used if available. In the absence of weather data, the SWAT model can generate the required data (Arnold et al. 2012).

In this study, all input data for the SWAT model were obtained from remote sensing sources, because of data deficiency in the area. The precipitation data were obtained from Tropical Rainfall Measurement Mission (TRMM) version 7 dataset, the maximum and minimum temperatures were acquired from the NASA POWER dataset. The reliability of these data was evaluated in the SWAT model by calibrating the simulated runoff to the observed runoff from the watershed outlet. To check the accuracy of the utilized data, model calibration and validation were carried out. The NASA POWER dataset was used to collect the remaining necessary information, which included wind speed, solar radiation, and relative humidity. DEM was acquired from the Shuttle Radar Topography Mission (STRM) of the United States Geological Survey (USGS) with a 1-arc second (∼30 m) resolution from www.earthexplorer.usgs.gov. The DEM is masked and projected in UTM_N42 projection using ArcGIS 10.5 to be ready for the SWAT model, the slope, and stream network data derived from DEM with standard flow accumulation. LULC data clipped from ESRI global LULC data set in the year 2020 that has a 10 m resolution (Figure 2a). The UN-FAO global soil map database version 3.6, which was finished in January 2003, was used to extract the soil map for the research area (Figure 2b). Observed river discharge provided by the MEW, Afghanistan.

Climate models are our main source of information regarding CC analysis, but there are systematic errors in the outputs of both the Global Climate Model (GCM) and Regional Climate Model (RCM) (IPCC 2014). The cause of these errors or biases is the limitation in the spatial resolution of data. Thus, the use of uncorrected data in the impact assessment of CC can often give unrealistic results. To overcome a large number of biases in climate models, many bias correction methods are available and for all methods, it should be remembered that the quality of observed data determines the quality of bias-corrected data (Copernicus). Bias correction is used to spatially downscale data to the point level (Maraun 2016). In this research, the precipitation and temperature were acquired from the GCM. The GCM is the model that is most frequently used to estimate various CC scenarios (Rahimpour et al. 2021 cited IPCC 2007, Afshar et al. 2017). Projected precipitation and maximum and minimum temperature data were acquired using the ‘MIROC5 model which has been cooperatively developed by the university of Tokyo, National Institute for Environmental Studies (NIES) and Japan Agency for Marine-Earth Science and Technology (JAM-STEC)’ (Rahimpour et al. 2021 cited Nozawa et al. 2007). Based on evaluation criteria in the SWAT calibration using the model efficiencies coefficients such as Nash–Sutcliffe Efficiency (NSE), percentage of bias (PBIAS), and coefficient of determination (R²), the MIROC5 model indicated the highest agreement with observational data (Rahimpour et al. 2021 cited Afshar et al., 2017). The precipitation and maximum and minimum temperature were downscaled and bias-corrected from the MIROC5 model of the GCM-CMIP5 project in the periods of 2050–2059 and 2090–2099 considering RCP4.5 and RCP8.5 scenarios. For converting climate data from GCMs grid points to the local point, the downscaling and bias correction process was performed using the CMhyd program. For this purpose, due to a lack of observed data in the study area, the precipitation and maximum and minimum temperature were used from TRMM and NASA POWER datasets, respectively.

Slope in the area is classified into five classes 0–15%, 15–30%, 30–50%, 50–70%, and 70–99% using multiple slope discretization. ‘To examine the differences in evapotranspiration and other hydrologic conditions for diverse land use, soil, and slope, SWAT sub-divide the watershed into regions with specific land use, soil maps, and slopes’ (Setegn et al. 2014). A minimum threshold area of 10, 5, and 5% for land use, soil maps, and slope classes, respectively, was determined for the number of HRUs in each sub-catchment. It means the area less than the defined percentage of the area is eliminated to minimize the number of HRUs while not overly compromising model accuracy (Glavan et al. 2013). Lang et al. (2017) conducted a study using ‘50 years of data from 90 meteorology stations in southwest China, to compare the five temperature-based methods of Hargreaves-Samani, Blaney-Criddle, Thornthwaite, Hamon’, and Linacre, and three radiation-based methods of Priestley-Taylor, Makkink, Abtew with Penman–Monteith at the yearly and seasonal scales. The result showed that Potential Evapotranspiration (PET) estimation methods' performance varies among regions, and radiation-based methods give better results than temperature-based methods. Among many methods for PET calculation, FAO Penman–Monteith is ‘usually recognized as an approved method for PET assessment’ (Lang et al. 2017). Penman–Monteith method was selected for the estimation of PET due to its better performance in the study area than the two other available methods of Priestley–Tylor and Hargreaves. Even though the Penman method has detailed requirements for meteorological data and is often constrained in many regions, the result from this method is evaluated better than other techniques. The Penman–Montieth method is calculated based on the following equation:

In the above equations; is the mean value of the observed values, is the ith observed value, is the ith simulated value, is the mean value of the simulated values and n is the total count of the sample pairs (Wang et al. 2018).

In the model calibration and validation using SWAT-CUP, parameter selection is the most time-consuming process to find appropriate parameters that fit the study area and give the best result. The modeler must find the best parameter value ranges for calibration to obtain the best outcome. The analyst will need some experience and hydrological expertise to determine the starting parameter ranges to be optimized. For considering the best parameters, in addition to the initial ranges, absolute parameter ranges should also be selected if necessary (Abbaspour et al. 2016). In this research, 13 sensitive parameters were chosen for model calibration and validation which are described in Table 1. Their sensitivity was done using global sensitivity analysis, and the resulting list of sensitive parameters was calibrated against the observed river discharge data from the Kokcha watershed's general outlet (Khwajaghar).

Sensitivity examination is the process of regulating the rate of variation in the model output according to changes in model input parameters (Abbaspour 2019). In this research, sensitivity analysis was carried out using global sensitivity analysis defined by P-value and t-stat. ‘The larger, in absolute value, the value of t-stat, and the smaller the p-value, the more sensitive the parameter’ (Abbaspour 2019).

In this research, the models were developed for 10 years to have a comparison to observed data, since the duration of the observed river discharge was available for 10 years (Figure 3). The model was calibrated and validated based on the observed streamflow in the periods of 2010–2014 and 2015–2019, respectively (Table 2). The model efficiency coefficients were obtained as R² = 0.77, NSE = 0.70 in the calibration, and R² = 0.82, NSE = 0.74 in validation. The calibration and validation were conducted to ensure the accuracy of obtained precipitation and temperature data from the TRMM and NASA POWER datasets accordingly. Based on the classification of the model efficiency coefficient (Moriasi et al. 2007), the result of calibration and validation was in a good class (Table 3).

The peak in river flow all around Afghanistan is happening in the spring season between April and May. The peak in Kokcha river flow under the impacts of CC shows an early shift in the spring throughout 2050–2059 and 2090–2099. A second flow peak will appear in July, which is caused by precipitation in this month. This change in runoff peak can have positive or negative effects on the catchment, which depends on the proper water management in the watershed. In addition to severe precipitation in July, the other reason for the second peak can be melting snow/glaciers since the area is partly covered by permanent snow and glaciers, which count as the main feeding source of the river during the summer season.

In this study and the previous studies that were conducted around the region (previously there is no similar study on the Kokcha watershed), the temperature change seems in line with the obtained results from past studies. Aich et al. (2017) investigated the temperature change in different regions of Afghanistan. The result of their study at the Hindu Kush mountains regions considering the RCP4.5 scenario indicated ranges of 1–3.2 and 2–4.1 °C in the periods of 2006–2050 and 2006–2099, respectively. Under the impacts of the RCP8.5 scenario, the temperature change will be 1.7–3.8 and 5.3–10.3 °C in the middle and end of the century, respectively. However, the acquired results from the Kokcha watershed are not out of the ranges of temperature change in the previous studies, but the regional variation of temperature is already known. Since the distribution of precipitation over the land is not regular, the rising temperature can cause intense drought due to increasing evapotranspiration in the area, although, the increased precipitation has been projected at higher altitudes. In the Kokcha watershed, the rising temperature in the winter months was unexpected. The temperature mostly increases in January and February when the precipitation would be at maximum. This increased precipitation can rise the runoff in the area while the water demand is at its minimum. The rising and decreasing tendency of temperature throughout the year is another issue that should be considered for water management. In the Kunduz watershed which is also a sub-basin of the Amu Darya in Afghanistan, a study by Akhundzadah et al. (2020) revealed the river discharge as heterogeneous for various parts of the watershed. In the Doab station which is located at the headwater of the watershed, the streamflow indicated an increase that will be mostly due to glacier melt in the area. In the lowland of Pul-i-Khumri station, the river discharge decreases as a result of precipitation decrease in the study area, even though, this is covered out by glacier melt on the upper sides.

In respect of temperature change, the study on the Kokcha watershed represented a similar result to the Kaido watershed in northwest China with a cold-arid region. However, in the Kaido watershed, precipitation indicated a rising trend in the winter and spring throughout the century, also the runoff increased as a result of precipitation increment. But, in the Kokcha basin, precipitation showed an extremely decreasing trend after January and February through the end of the spring season, also, the changes are different in the middle and at the end of the century. However, overall runoff shows a decreasing trend over the century. In the Kokcha watershed, the precipitation would not considerably change during 2090–2099, but the increasing evapotranspiration causes the decrease of overall river discharge in the area.

The validity of precipitation data has been evaluated using the SWAT model. The methodology limitation in this research can be the usage of remotely sensed precipitation data, which have been utilized as a reference for bias correction of future precipitation data. However, the observed data deficiency in Afghanistan is the main problem for researchers in this country. It is recommended to use observed data or remote sensing data with higher spatial resolution in future studies. Furthermore, the adaptation of water management to CC is strongly recommended in the Kokcha watershed.

CC is a serious issue for scientists and water managers in the 21st century. The consequences of global warming are mostly water-related which impacts every aspect of human life and the ecosystem. For overcoming this challenge, the impact assessment of CC on every watershed is necessary. This study is assigned to investigate the effects of CC on the water resources of the Kokcha watershed, mountainous and partly covered by snow and glaciers. The SWAT model performance was assessed between 2010 and 2019 with R² and NSE as model efficiency coefficients. The calibration and validation were performed based on the observed river discharge, and the coefficient values during the calibration were acquired as R² = 0.77 and NSE = 0.70, and in validation, R² = 0.82, and NSE = 0.74 which are both in a good class. The future climate data were obtained from GCM, the most commonly used source of future climate projection. The CMhyd program was used to extract rainfall and temperature values. Then the runoff simulation was done employing the SWAT model using the downscaled rainfall and temperature values of the validated data from the MIROC5 model. The climate data were utilized considering RCP4.5 and RCP8.5 scenarios in the 2050–2059 and 2090–2099 intervals.

Based on the RCP4.5 scenario, the average annual temperature rise is +2.47 and +2.85 °C in the middle and end of the century, respectively. However, the temperature shows an increasing trend mostly in the winter months than in summer. Precipitation shows an overall decreasing trend, with −30% in 2050–2059 and −17.17% at the end of the century. The precipitation with a flat peak starting from January and ending in May shortens its duration to a single peak in February. The second peak of precipitation takes shape in July, this change is slight during the middle of the century and will grow in the 2090–2099 period. The variations in temperature and precipitation impact all water balance elements and decreasing rainfall can challenge the water management system in the watershed. The snowfall in the winter months was the main source of streamflow that was feeding the Kokcha river throughout the year. Evaporation represents a rising trend throughout the year and decreases in July slightly. Runoff decreases throughout the century with −54.56% in 2050–2059 and −25.98% during 2090–2099 and an extreme drop in peak flow and shifting the peak time to earlier. Considering the RCP4.5 scenario, the combination of snow at the top of mountains will not be done and the river discharge during the winter months will rise, showing the change of snowfall to rainfall.

Considering the RCP8.5 scenario, the variation in water balance will become more severe with an excessive temperature rise of +3.38 °C in 2050–2059 and +5.51 °C at the end of the century. Although the precipitation change considering RCP8.5 is not much more (−9.28% in 2050–2059 and −4.52% in 2090–2099) in comparison to historical data, decreasing precipitation in the spring season can create a serious challenge in the area due to the maximum agricultural water demand in this season. Snowfall mostly changes to rainfall in the winter months, also average PET along the century will rise in the region. Less precipitation (−9.28 and −4.52%) change in the watershed caused −29.18 and −6.45% of runoff change in the 2050–2059 and 2090–2099 periods, respectively. Dropping the peak flow and shifting its time from May to April in the middle of the century, then the peak flow in the spring and winter seasons will decrease and a new peak flow takes shape in July that can make the regularity of the river more complicated in management.

This study represents the variations in water balance elements in the Kokcha catchment and the changes in runoff considering the RCP4.5 and RCP8.5 scenarios. Further investigations are recommended for the impact assessment of CC on available water infrastructures at the watershed. The result of this study can be used for the adaptation of water management to CC in the Kokcha watershed and similar regions.

S.A.A. assigned the topic, prepared the data, carried out the simulation, wrote the manuscript, and reviewed. S.Ö. supervised and reviewed. Both authors discussed the results.

We would like to thank Mr Khan Mohammad Takal for providing the hydrometeorological data of the study area.

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

The authors declare there is no conflict.