Abstract
Hydrological models play a key role in simulating and assessing climate and land use/cover (LULC) change impacts on hydrology in a watershed. In this study, the impact of climate and LULC change was investigated using the Soil and Water Assessment Tool (SWAT) model. The simulated and observed streamflow showed a good agreement. Both Nash–Sutcliffe Efficiency (NSE) and coefficient of determination (R2) were found to be greater than 0.7 during the calibration (1985–2002) and validation (2003–2012) period. The water balance components were simulated with inputs from downscaled Global Climate Models (GCMs) data (i.e., future scenario (2030–2100) relative to a baseline period (1974–2004)) under RCP4.5 and RCP8.5, and hypothetical generated LULC change scenarios. All GCMs projected an increase in temperature over the Kabul River Basin (KRB), whereas there was a lack of agreement on projected precipitation among GCMs under both emission and future scenarios. Water yield (WYLD) and evapotranspiration (ET) were projected to decrease in the 21st century. Average annual WYLD was projected to increase under the agriculture-dominant scenario, whereas it decreased under forest and grassland-dominant scenarios. These results are valuable for relevant agencies and stakeholders to adopt measures to counter the negative impacts of climate and LULC change on water resources.
HIGHLIGHTS
The projected temperature consistently increases in the 21st century.
Most of the GCMs projected a decrease in annual and winter precipitation in the KRB.
The simulated water balance components show a higher impact by climate than LULC change scenarios.
The water yield was increased under the agriculture-dominant scenario, whereas it decreased under forest and grassland-dominant scenarios.
Graphical Abstract
INTRODUCTION
Climate change is considered a significant variable that influences changing streamflow worldwide (Somorowska 2017; Yeh 2020). Climate change is a global concern because of its impact on the reduction in terrestrial water storage particularly in the Global South which led to droughts and desertification (Pokhrel et al. 2021). Climate change will have a major impact on seasonal flow regimes (i.e., by more than 10%) on 90% of global land area by the 2050s compared to alteration in flows by dams and water withdrawals (Doll & Zhang 2010). Climate change causes the redistribution of water resources, which may lead to increases in the frequency and intensity of drought and flood disasters (Li et al. 2013; Farooq et al. 2022). Significant attention needs to be paid to the impact of climate change on river flows from rivers fed by ice or snow and glaciers where these act as the main source of flow in arid and semi-arid regions (Nepal et al. 2013). The upstream reserves of snow and ice in the Himalayas and Hindu Kush are more likely to be affected by climate change, jeopardizing food security and availability of water in the Kabul River (Immerzeel et al. 2010; Lutz et al. 2016).
Hydrological models are representative of the simplification of real world processes (Schurz et al. 2019). The lack of field measurements is a major challenge in the model's construction and calibration. The up-to-date data on land and water resources in the Kabul River Basin (KRB) are limited or of poor quality (Mack et al. 2013). In arid and semi-arid regions, readily available remote sensing data are a viable option for rainfall–runoff modelling, as well as to choose the best model to quantify the unique hydrology of the region. It is also essential to choose a model which represents environmental processes appropriately and is capable of simulating a possible impact of future changes on hydrological processes (Chiew & Vaze 2015). A modeler can use his experience, knowledge and assessment of the study area to adjust several model parameters during the model-building process to improve simulations. Profound knowledge of the relevant hydrological processes is vital to draw informed conclusions. A well-known issue of equifinality is common in hydrological models. Furthermore, simulation of climate and land use/cover (LULC) changes are very uncertain (Chen et al. 2019). However, a commonly adopted approach of scenario-based modelling is carried out to meet this challenge in the KRB.
The SWAT model is the most widely used model in environmental studies because it is open access, computationally efficient and user friendly which strongly promotes its applications in a wide range of hydrologic and/or environmental problems. It is used to simulate water quantity, water quality and climate change on annual, monthly, daily and sub-hourly levels over long periods at a basin scale (Sanchez-Gomez et al. 2022). It can be used to simulate hydrological processes under the influence of future changes (climate and LULC changes) on different variables of the water cycle (Wagner et al. 2017; Schurz et al. 2019). Over 30 years of work has resulted in the development of supporting software, interfaces, plug-ins and several other tools (e.g., SWAT + , SWAT-CUP, VIZSWAT, MWSWAT, SWATplusR, R SWAT and SWAT viewer). However, a wide range of data input, parameters and their spatial variability is a major constraint to use in situ observations to run the SWAT model (Moriasi et al. 2015). The SWAT model is weak to simulate ground water systems in geologically heterogeneous basins (Nguyen & Dietrich 2018). Manual calibration and sensitivity analysis is time-intensive when applying the model to complex catchments with numerous hydrological response units (HRUs).
Global climate models (GCMs) are the primary tools that provide information about climate on global, hemispheric and continental scales (Trzaska & Schnarr 2014). Input data from GCMs are mostly used to assess the impact of climate change on the hydrologic cycle. Previous works on climate change impact on hydrology in the Himalayan region demonstrate a significant variability in the projected outcomes (Tahir et al. 2011; Lutz et al. 2014; Ndhlovu & Woyessa 2021). In the KRB, projected climate change results showed increases (+2.4 to 3.3%) and decreases (−5 to 8%) of streamflow at different gauging stations under RCP4.5 and RCP8.5 in the near future (Akhtar et al. 2021). The projected climate showed an increase in annual temperature and precipitation in most parts of the Nile Basin (Daniel Mengistu et al. 2021). These changes affected the hydrological regime in the basin by increasing potential evapotranspiration up to +27%, and surface runoff (+14%) by the end of the 21st century. However, water yield and base flow are estimated to decrease in contrast to increase in surface runoff. Similarly, climatic changes are likely to have implications in the KRB, where more than 80% of the flow is dependent upon snow/glacier melt during late spring and summer (Lutz et al. 2016; Akhtar et al. 2021).
LULC changes significantly impact the hydrological components such as evapotranspiration, groundwater recharge, soil infiltration and surface runoff generation (Glavan & Pintar 2012; Ozturk et al. 2013). A number of studies have been conducted to evaluate the impact of LULC change on streamflow across the globe (Zare et al. 2017; Shirmohammadi et al. 2020; Khoi et al. 2021; Ni et al. 2021). In order to understand the LULC change impact on hydrology, most researchers adopt a scenario-based approach (Wu et al. 2015; Aduah et al. 2017; Haghighi et al. 2020), creating scenarios of land cover conversions (Niu & Sivakumar 2014). However, the results of LULC change impact on watershed hydrology remained contradictory. For example, Malmer et al. (2010) argued that tree planting in the tropics for the purpose of carbon sequestration often competes with scarce water resources. Forests act as ‘pumps’ by increasing rates of evapotranspiration and ‘sponges’ by enhancing infiltration and soil moisture retention capacities (Peña-Arancibia et al. 2019). Deforestation results in reduction of leaf area, changes in albedo and reduction in rooting depth which all contribute to reduced evapotranspiration which ultimately affect streamflow (Levy et al. 2018; Sokolova et al. 2019; Posada-Marín & Salazar 2022). Mao & Cherkauer (2009) simulated vegetation-dependent changes in response to hydrology and noted that deforestation driven by urban expansion and large-scale conversion of land to agriculture resulted in an increase in total runoff. Welde & Gebremariam (2017) simulated the hydrologic response to LULC change by using the SWAT model in the Tekeze Dam watershed in northern Ethiopia by creating three LULC change scenarios, and observed that increasing bare land and conversion to agriculture resulted in increased streamflow and sediment yield (Guzha et al. 2018).
Rapid urbanization, population growth and economic development increases the demand for food, energy and water resources across the globe (Aghsaei et al. 2020). This has evidently led to considerable LULC changes especially in developing countries. The KRB is a main supplier of freshwater to its inhabitants where impact of climate and LULC change on hydrology is of great importance for sustainable water supply to domestic, agriculture and industrial sectors. More than 44% of the population of Afghanistan is directly or indirectly employed in the agriculture sector (ILOSTAT 2021). It is critical to investigate the effects of climate and LULC change scenarios on water balance components in the KRB where scientific research related to impact assessment on water resources is lacking. In previous studies, the climate change impact on hydrology in the KRB was investigated with climate models’ data to run hydrological models. However, no studies are available (to the best of my knowledge) to examine the impact of both climate and hypothetical LULC change scenarios on water balance components in the study basin. This study evaluates the potential impact of climate and LULC changes on hydrology to differentiate the major driver of change at play in the KRB. In this study, a novel framework was proposed to quantify LULC change impact on water balance components by generating hypothetical LULC change scenarios. This study also investigates the potential future changes in temperature and precipitation and their impact on streamflow. The simulation of projected water balance components under climate and LULC change scenarios is all important for better water resource planning and management particularly for data scarce environments where the major contribution to flow comes from the snow and ice melt.
Study area
The location map of the study area. The soil classes are Lithosols (3503, 3512, 3525, 3712, 3731), Cambisols (3667, 3668, 3673), Xerosols (3870, 3871) and Glacier (6998).
The location map of the study area. The soil classes are Lithosols (3503, 3512, 3525, 3712, 3731), Cambisols (3667, 3668, 3673), Xerosols (3870, 3871) and Glacier (6998).
MATERIALS AND METHODS
Methodological framework for impact assessment of climate and LULC change scenarios on water balance components in the KRB using the SWAT model.
Methodological framework for impact assessment of climate and LULC change scenarios on water balance components in the KRB using the SWAT model.
Climate change impact on water balance components
Input data from five GCMs (GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A, MIROC and NoerESM1-M) along with two emission scenarios (RCP4.5 and RCP8.5) for future projections (2030–2100) relative to the historical period (1980–2004) were selected to run the SWAT model in the KRB (Table 1). Medium (RCP4.5) and high (RCP8.5) emission scenarios which are more realistic and worst case scenario, respectively, were selected for this study. GCMs often misrepresent extremes in climatic variables such as temperature and precipitation. Due to the typically coarse spatial resolution of GCMs (approximately 150–300 km), they cannot account for the fine-scale heterogeneity driven by landscape features (e.g., mountains, water bodies) or components of climate systems, such as convective clouds and coastal breezes, that have resolutions finer than 100–500 km. To compute local details from global models, empirical or statistical downscaling is needed. To do this, it is assumed that the local climate is specified by interactions between large-scale atmospheric characteristics and local topography (mountains, landscape properties, etc.), which can be modeled through downscaling processes to derive fine-scale climate information. Essentially, the downscaling process adds information to coarse GCM outputs to better understand local climatic conditions at a finer resolution. Climate information is required for future planning typically for regions that are smaller than approximately 100 km grid cell. Therefore, downscaling and upscaling techniques have been developed to obtain the relevant information on regional-scale climate change impact studies (Trzaska & Schnarr 2014).
The CMIP5 models used in this study
Model . | Resolution . | Institution ID . | Key references . |
---|---|---|---|
GFDL-ESM2M | 2.5 × 2L24(M45) | GFDL | Delworth et al. (2006) |
IPSL-CM5A-LR | 3.75 × 1.89L39 | MOHC | Hourdin et al. (2013) |
HadGEM2-ES | 1.875 × 1.24L38(N96) | IPSL | Bellouin et al. (2011) |
MIROC5 | 1.40625 × 1.40625L40(T85) | MIROC | Watanabe et al. (2011) |
NorESM1-M | 2.5 × 1.9L26(F19) | NCC | Bentsen et al. (2013) |
Model . | Resolution . | Institution ID . | Key references . |
---|---|---|---|
GFDL-ESM2M | 2.5 × 2L24(M45) | GFDL | Delworth et al. (2006) |
IPSL-CM5A-LR | 3.75 × 1.89L39 | MOHC | Hourdin et al. (2013) |
HadGEM2-ES | 1.875 × 1.24L38(N96) | IPSL | Bellouin et al. (2011) |
MIROC5 | 1.40625 × 1.40625L40(T85) | MIROC | Watanabe et al. (2011) |
NorESM1-M | 2.5 × 1.9L26(F19) | NCC | Bentsen et al. (2013) |
A large number of GCMs are available nowadays in the CMIP5 archive but identifying a set of representative climate models for a particular region is still a cumbersome task. The selection of climate models for climate change impact assessment studies of a particular region is not a straightforward process (Lutz et al. 2016). In many studies, only a single GCM was used without a reasonable justification and often based on the country of research. For example, the HADCM3, CGCM3.1 and CSIRO GCM are commonly used by UK, Canadian researchers and Australia, respectively (Holman et al. 2009; Austin et al. 2010). However, a better approach is to employ more than one GCM for a particular area to provide a mean of estimates and simulate all aspects of precipitation dynamics satisfactorily (Lutz et al. 2016).
The raw climate outputs from GCMs are not recommended due to their limited spatial resolution, incomplete knowledge of climate system processes and simplified physics and thermodynamic processes (Mehrotra & Sharma 2019). It has been found that a hydrological model run with raw data can produce unrealistic results (Hansen et al. 2001; Harding et al. 2014). Biases or errors in raw climate model outputs can also be due to historical observations (Ramirez-villegas et al. 2013). Hence, it is crucial to bias correct GCM simulations in order to produce future climate projections. Therefore, a key requirement for developing a statistical downscaling model is a reliable daily observation of climate variables at a high spatial resolution (Dahri et al. 2018; Scott et al. 2019). The lack of observed data in the region means that, in this case, reanalysis data or model reconstructed products are needed to explore the spatial and temporal characteristics of hydro-meteorological processes (Table 2).
Global climate gridded and reanalysis datasets
Dataset . | Variables . | Start date . | End date . |
---|---|---|---|
NCEP/NCAR Reanalysis | Atmosphere, Air Temperature, Geopotential Height | Jan-1948 | Sep-2019 |
GPCC | Precipitation | 1901 | 2009 |
NCEP CFSR | Temperature, Precipitation | Jan-1979 | Aug-2019 |
ERA-Interim | Temperature, Precipitation | Jan-1979 | Jun-2019 |
ERA-5 | Temperature, Precipitation | Jan-1979 | Dec-2018 |
JRA-55 | Temperature, Precipitation | Jan-1958 | Sep-2019 |
CERA-20C | Temperature, Precipitation | Jan-2001 | Dec-2010 |
TRMM | Precipitation | 1998 | 2010 |
CRU_TS 4.01 | 2 m temperature, Precipitation | Jan-2001 | Dec-2016 |
GPCP V2.3 | Precipitation | Jan-1979 | Near present |
CMAP (STD) | Precipitation | Jan-1880 | Near present |
NOAA Global Anomalies | 2 m temperature | Jan-1880 | Near present |
APHRODITE | Precipitation | Jan-1951 | Near present |
Dataset . | Variables . | Start date . | End date . |
---|---|---|---|
NCEP/NCAR Reanalysis | Atmosphere, Air Temperature, Geopotential Height | Jan-1948 | Sep-2019 |
GPCC | Precipitation | 1901 | 2009 |
NCEP CFSR | Temperature, Precipitation | Jan-1979 | Aug-2019 |
ERA-Interim | Temperature, Precipitation | Jan-1979 | Jun-2019 |
ERA-5 | Temperature, Precipitation | Jan-1979 | Dec-2018 |
JRA-55 | Temperature, Precipitation | Jan-1958 | Sep-2019 |
CERA-20C | Temperature, Precipitation | Jan-2001 | Dec-2010 |
TRMM | Precipitation | 1998 | 2010 |
CRU_TS 4.01 | 2 m temperature, Precipitation | Jan-2001 | Dec-2016 |
GPCP V2.3 | Precipitation | Jan-1979 | Near present |
CMAP (STD) | Precipitation | Jan-1880 | Near present |
NOAA Global Anomalies | 2 m temperature | Jan-1880 | Near present |
APHRODITE | Precipitation | Jan-1951 | Near present |
The data from Climate Forecast System Reanalysis (CFSR) were used to identify biases in the historic GCMs data using climate change toolkit (Vaghefi et al. 2017). CFSR data were chosen for downscaling and bias correction of GCMs due to their high resolution (∼38 km2) which meets the study requirements. Furthermore, the CFSR data of minimum temperature, maximum temperature and precipitation over the KRB were used for constructing the downscaling models to ensure that bias corrected temperature and precipitation will not project unrealistic results. The future data were bias corrected against CFSR data during 1980–2013.
LULC change impact on water balance components
The hypothetical LULC change scenarios were generated to investigate LULC change impact on water balance components such as evapotranspiration (ET), soil water storage (SW) and water yield (WYLD). Hypothetical LULC change scenarios were generated by converting semi-natural ecosystems to agriculture ecosystems, forest cover, grassland ecosystems and urban expanded areas to assess their potential impact on the hydrology of the KRB.
To quantify the impact of LULC change, a calibrated and validated SWAT model was forced with observed meteorological data during 1990–2013 using hypothetical LULC change scenarios to simulate water balance components. The simulated water balance components with baseline scenario were compared with four hypothetical LULC change scenarios to quantify the impact on the hydrological regime of the KRB.
An LULC map of the KRB was derived from the global map of 300 m spatial resolution for the year 2012, downloaded from the European Space Agency (ESA) and formatted by SWAT as row crops (AGRR), generic crops (AGRL), grasslands/herbaceous (RANGE), range shrub land (RANGB), deciduous forest (FRSTD), evergreen forest (FRSTE), mixed forest (FRST), bare rock (SWRN), urban medium density (URML) and water including snow and ice (WATR) as shown in Figure 4. Similar classes in the baseline LULC map were merged to generate the baseline scenario of six major classes of agriculture, grassland, forest, bare, urban and water (Table 3).
LULC change in the KRB from original scenario to four hypothetical scenarios by increasing 30% agriculture land, 30% forest land, 34% grassland and 5% urban area
LULC change Scenarios . | Baseline Scenario . | Scenario-1 (Agriculture) . | Scenario-2 (Forest) . | Scenario-3 (Grassland) . | Scenario-4 (Urban) . |
---|---|---|---|---|---|
Area | % change | % change | % change | % change | % change |
Agriculture (AGRR) | 25 | 55 | 25 | 5 | 25 |
Forest (FRST) | 8.2 | 8.2 | 38 | 0 | 8.2 |
Grassland (GRASS) | 58 | 28 | 28 | 86 | 58 |
Urban (URBN) | 1 | 1 | 1 | 1 | 6 |
Barren land (BARN) | 5 | 5 | 5 | 5 | 0 |
Water (WATR) | 2.4 | 2.4 | 2.4 | 2.4 | 2.4 |
LULC change Scenarios . | Baseline Scenario . | Scenario-1 (Agriculture) . | Scenario-2 (Forest) . | Scenario-3 (Grassland) . | Scenario-4 (Urban) . |
---|---|---|---|---|---|
Area | % change | % change | % change | % change | % change |
Agriculture (AGRR) | 25 | 55 | 25 | 5 | 25 |
Forest (FRST) | 8.2 | 8.2 | 38 | 0 | 8.2 |
Grassland (GRASS) | 58 | 28 | 28 | 86 | 58 |
Urban (URBN) | 1 | 1 | 1 | 1 | 6 |
Barren land (BARN) | 5 | 5 | 5 | 5 | 0 |
Water (WATR) | 2.4 | 2.4 | 2.4 | 2.4 | 2.4 |
Maps of hypothetically generated LULC change scenarios over the KRB. Agriculture, forest, grassland, and urban are the dominant LULC classes in (a) scenario-1, (b) scenario-2, (c) scenario-3 and (d) scenario-4, respectively.
Maps of hypothetically generated LULC change scenarios over the KRB. Agriculture, forest, grassland, and urban are the dominant LULC classes in (a) scenario-1, (b) scenario-2, (c) scenario-3 and (d) scenario-4, respectively.
It is important to recognize that not all barren land would be suitable for conversion to urban and that there would need to be considerable effort and investment to convert 30% grassland to forest. However, afforestation schemes across the globe are ambitious (Spanning 2018) and urban expansion of 5% not unreasonable. Therefore, although significant assumptions have been made, and the scenarios are only hypothetical, they are still of value given current political ambitions.
SWAT modelling
A common approach followed in most climate and LULC change impact studies is to obtain data from GCMs and LULC maps, and then run a hydrological model to simulate streamflow for projected scenarios. The main goal of this study was to utilize a calibrated SWAT model to estimate future water availability by simulation of the water balance components under climate and LULC change scenarios over the KRB.
Model inputs
Major classes of model inputs such as HRU definition, climate data, calibration and validation data and climate projection data are shown in Table 4. SWAT utilizes gridded datasets to properly define HRUs. The watershed and river network is delineated using digital elevation model (DEM) data from Shuttle Radar Topographic Mission (SRTM). The land use data from Globcover and soil data from FAO are used to generate LULC classes and soil types (Figure 1). The soil types are classified as loam (3503–3525, 3673, 3712) clay-loam (3667–3668, 3871), silt-loam (3870) and glacier soil (6998).
The description, source and scale/resolution of data used in SWAT
HRU definition data . | Data source . | Scale/Resolution . |
---|---|---|
Digital Elevation Model | Shuttle Radar Topographic Mission (SRTM) | Grid cell 90 × 90 m |
Land use | Glob Land Cover | 300 × 300 m |
Soil | FAO-UNESCO Global Soil Map | Scale: 1:5,000,000 |
Climate data Precipitation, max./min. Temperature relative humidity, solar radiation. wind speed | Climate Forecast System Reanalysis (CFSR) | Grid cell 38 km |
Calibration and Validation Data | ||
Discharge | Water and Power Development Authority (WAPDA) | |
Climate Projection Data | Coupled Model Inter-comparison Project (CMIP5) | Grid cell 38 km |
HRU definition data . | Data source . | Scale/Resolution . |
---|---|---|
Digital Elevation Model | Shuttle Radar Topographic Mission (SRTM) | Grid cell 90 × 90 m |
Land use | Glob Land Cover | 300 × 300 m |
Soil | FAO-UNESCO Global Soil Map | Scale: 1:5,000,000 |
Climate data Precipitation, max./min. Temperature relative humidity, solar radiation. wind speed | Climate Forecast System Reanalysis (CFSR) | Grid cell 38 km |
Calibration and Validation Data | ||
Discharge | Water and Power Development Authority (WAPDA) | |
Climate Projection Data | Coupled Model Inter-comparison Project (CMIP5) | Grid cell 38 km |
The climate data (e.g., temperature, precipitation, relative humidity, solar radiation and wind speed) were downloaded from Global Weather Data for SWAT (https://globalweather.tamu.edu/) in a SWAT file format. The monthly discharge data for model calibration and validation were obtained from Water and Power Development Authority (WAPDA) at the Nowshera outlet of the KRB located in Pakistan. The climate projection data were downloaded from water weather energy ecosystem (https://www.2w2e.com/).
Model setup
Description of the SWAT model input commands and time period of discharge data
Description . | KRB . |
---|---|
Minimum sub-basin threshold area | 10,000 ha |
HRU threshold (slope/land use/soil) | 15/12/12 |
Number of sub-basins | 11 |
Slope classes | 0–20/20–25/ > 25% |
Number of HRUs | 204 |
Number of elevation bands | 10 |
Warm-up period | 1979–1984 (6 years) |
Calibration period | 1985–2002 (18 years) |
Validation period | 2003–2012 (10 years) |
Description . | KRB . |
---|---|
Minimum sub-basin threshold area | 10,000 ha |
HRU threshold (slope/land use/soil) | 15/12/12 |
Number of sub-basins | 11 |
Slope classes | 0–20/20–25/ > 25% |
Number of HRUs | 204 |
Number of elevation bands | 10 |
Warm-up period | 1979–1984 (6 years) |
Calibration period | 1985–2002 (18 years) |
Validation period | 2003–2012 (10 years) |
SWAT users usually used SWAT-CUP (SWAT Calibration and Uncertainty Program) (Abbaspour et al. 2007) for automatic calibration due to ease and efficiency (Arnold et al. 2012). The set of all model parameters for calibration of runoff in SWAT-CUP produced unrealistic values especially for snow parameters (e.g., SMFMN and SMFMX) as well as precipitation and temperature lapse rates (e.g., PLAPS and TLAPS). These erratic values in parameters could be due to fitting the model with flow only or calibrating too many parameters at once. According to Abbaspour et al. (2017), snow parameters and lapse rates should be fitted separately before other flow parameters. In this study, lapse rates were fitted first before snow parameters (SFTMP, SMTMP, SMFMN, SMFMX and TIMP) and at the end all the remaining parameters were fitted to calibrate the SWAT model.
Sensitivity analysis, calibration and validation
After fitting PLAPS, TLAPS and snow parameters separately, the calibration procedure for the remaining parameters was conducted in the SWAT-CUP (Table 6). A sensitivity analysis was performed in SWAT-CUP using Sequential Uncertainty Fitting Algorithm-Version2 (SUFI-2) algorithm which is based on Latin hypercube sampling (McKay et al. 1979). A global sensitivity analysis was conducted to identify the impact of each parameter on objective function. The results from the sensitivity analysis are tabulated in Table 7. The parameter sensitivity in the SUFI-2 algorithm is measured by p-value and t-stat. The larger value of t-stat represents higher sensitivity of parameters, while the significance of parameter sensitivity increases with p-value closer to zero. The most sensitive parameters selected from the group of parameters were CN2, GW_DELAY, REVAPMN, ALPHA_BF, GW_REVAP, ALPHA_BNK and ESCO.
Fitted values of model parameters in the KRB
Parameters . | Fitted value . | Min value . | Max value . |
---|---|---|---|
V__TLAPS | −8.80 | −10.00 | −2.00 |
V__PLAPS | 87.80 | 0.00 | 100.00 |
V__SFTMP.bsn | −0.14 | −5.00 | 5.00 |
V__SMTMP.bsn | −1.10 | −5.00 | 5.00 |
V__SMFMX.bsn | 3.85 | 0.00 | 5.00 |
V__SMFMN.bsn | 1.61 | 0.00 | 5.00 |
V__TIMP.bsn | 0.15 | 0.00 | 1.00 |
R__CN2.mgt | 0.01 | 0.00 | 0.20 |
V__GW_DELAY.gw | 35.71 | 0 | 450 |
V__REVAPMN.gw | 207.29 | 0.00 | 500 |
V__ALPHA_BF.gw | 0.27 | 0.00 | 1 |
V__GW_REVAP.gw | 0.06 | 0.02 | 0.2 |
V__ALPHA_BNK.rte | 0.07 | 0 | 1 |
V__ESCO.hru | 0.87 | 0.00 | 1 |
Parameters . | Fitted value . | Min value . | Max value . |
---|---|---|---|
V__TLAPS | −8.80 | −10.00 | −2.00 |
V__PLAPS | 87.80 | 0.00 | 100.00 |
V__SFTMP.bsn | −0.14 | −5.00 | 5.00 |
V__SMTMP.bsn | −1.10 | −5.00 | 5.00 |
V__SMFMX.bsn | 3.85 | 0.00 | 5.00 |
V__SMFMN.bsn | 1.61 | 0.00 | 5.00 |
V__TIMP.bsn | 0.15 | 0.00 | 1.00 |
R__CN2.mgt | 0.01 | 0.00 | 0.20 |
V__GW_DELAY.gw | 35.71 | 0 | 450 |
V__REVAPMN.gw | 207.29 | 0.00 | 500 |
V__ALPHA_BF.gw | 0.27 | 0.00 | 1 |
V__GW_REVAP.gw | 0.06 | 0.02 | 0.2 |
V__ALPHA_BNK.rte | 0.07 | 0 | 1 |
V__ESCO.hru | 0.87 | 0.00 | 1 |
Global parameter sensitivity analysis conducted in SWAT-CUP
Parameter name . | Description . | t-stat . | p-Value . |
---|---|---|---|
R__CN2.mgt | SCS runoff curve number | −11.029 | 0.000 |
V__GW_DELAY.gw | Groundwater delay | 4.785 | 0.000 |
V__REVAPMN.gw | Threshold water depth in the shallow aquifer for ‘revap’ to occur | 3.802 | 0.000 |
V__ALPHA_BF.gw | Base flow alpha factor | 2.478 | 0.014 |
V__GW_REVAP.gw | Groundwater ‘revap’ coefficient | −2.385 | 0.018 |
V__ALPHA_BNK.rte | Baseflow alpha factor for bank storage (days) | 2.347 | 0.020 |
V__ESCO.hru | Soil evaporation compensation factor | −1.909 | 0.057 |
R__SOL_BD(..).sol | Soil moist bulk density | −1.291 | 0.198 |
V__EPCO.bsn | Plant evaporation compensation factor | −1.240 | 0.215 |
V__CH_K2.rte | Effective hydraulic conductivity in main channel alluvium | −1.115 | 0.266 |
V__GWQMN.gw | Threshold water depth in the shallow aquifer for return flow to occur | −1.055 | 0.292 |
V__BIOMIX.mgt | Biological mixing efficiency | 0.930 | 0.353 |
R__SOL_AWC(..).sol | Available water capacity of the soil layer | −0.925 | 0.356 |
R__CANMX.hru | Maximum canopy storage | 0.442 | 0.658 |
V__SURLAG.bsn | Surface runoff lag coefficient | 0.371 | 0.711 |
V__CH_N2.rte | Manning's ‘n’ value for the main channel | 0.253 | 0.800 |
R__SOL_K(..).sol | Saturated hydraulic conductivity | −0.017 | 0.987 |
Parameter name . | Description . | t-stat . | p-Value . |
---|---|---|---|
R__CN2.mgt | SCS runoff curve number | −11.029 | 0.000 |
V__GW_DELAY.gw | Groundwater delay | 4.785 | 0.000 |
V__REVAPMN.gw | Threshold water depth in the shallow aquifer for ‘revap’ to occur | 3.802 | 0.000 |
V__ALPHA_BF.gw | Base flow alpha factor | 2.478 | 0.014 |
V__GW_REVAP.gw | Groundwater ‘revap’ coefficient | −2.385 | 0.018 |
V__ALPHA_BNK.rte | Baseflow alpha factor for bank storage (days) | 2.347 | 0.020 |
V__ESCO.hru | Soil evaporation compensation factor | −1.909 | 0.057 |
R__SOL_BD(..).sol | Soil moist bulk density | −1.291 | 0.198 |
V__EPCO.bsn | Plant evaporation compensation factor | −1.240 | 0.215 |
V__CH_K2.rte | Effective hydraulic conductivity in main channel alluvium | −1.115 | 0.266 |
V__GWQMN.gw | Threshold water depth in the shallow aquifer for return flow to occur | −1.055 | 0.292 |
V__BIOMIX.mgt | Biological mixing efficiency | 0.930 | 0.353 |
R__SOL_AWC(..).sol | Available water capacity of the soil layer | −0.925 | 0.356 |
R__CANMX.hru | Maximum canopy storage | 0.442 | 0.658 |
V__SURLAG.bsn | Surface runoff lag coefficient | 0.371 | 0.711 |
V__CH_N2.rte | Manning's ‘n’ value for the main channel | 0.253 | 0.800 |
R__SOL_K(..).sol | Saturated hydraulic conductivity | −0.017 | 0.987 |
Most sensitive parameters based on p-values (<0.05) are represented by bold.
The selected most sensitive model parameters were utilized for automatic calibration of streamflow with SUFI-2 algorithm in SWAT-CUP. Each model parameter was assigned either relative (r) or replace (v) change method with upper and lower limits. The term ‘r’ is the relative change in parameter with a given range while the term ‘v’ is the replacement value of the model parameter with assigned value. The model parameters fitted values obtained in two iterations of 1,000 runs each with Nash–Sutcliffe Efficiency (NSE) value set as the objective function. The model parameters range/value obtained after successful calibration are provided in Table 6.
The water balance components were simulated based on HRUs then added to each sub-watershed and routed to the outlet of the watershed. In the third step, the model was run based on prepared weather data files and HRU information defined in the previous steps. The model was run initially for a ‘warm-up period’ of 6 years. The streamflow was simulated first for the calibration period (1985–2002) in the KRB, before calibration and validation using the SWAT-CUP SUFI2 algorithm.
Performance evaluation
The SWAT model was evaluated by using three performance evaluation parameters available in the SUFI-2 algorithm: NSE, the coefficient of determination (R2), and Percent Bias (PBIAS) (Krause & Boyle 2005) as shown in Table 8. NSE is one of the most commonly used criteria for calibration of hydrological models (Nash & Sutcliffe 1970) and describes the goodness-of-fit of the observed and simulated plots. Its value ranges from 0 to 1, with values >0.5 considered acceptable and 1 indicating a perfect match between the observed and simulated data. The values of NSE are ranked from very good (0.75 < NSE ≤ 1.00) good (0.65 < NSE ≤ 0.75), satisfactory (0.50 < NSE ≤ 0.65) to unsatisfactory (NSE ≤ 0.50).
Model performance evaluation values of observed and simulated discharge at the KRB outlet used for model calibration and validation
Outlet . | Calibration (1985–2002) . | Validation (2003–2012) . | ||||
---|---|---|---|---|---|---|
Nowshera . | R2 . | NSE . | PBIAS (%) . | R2 . | NSE . | PBIAS (%) . |
0.83 | 0.79 | 15.9 | 0.79 | 0.71 | 7.6 |
Outlet . | Calibration (1985–2002) . | Validation (2003–2012) . | ||||
---|---|---|---|---|---|---|
Nowshera . | R2 . | NSE . | PBIAS (%) . | R2 . | NSE . | PBIAS (%) . |
0.83 | 0.79 | 15.9 | 0.79 | 0.71 | 7.6 |
The coefficient of variation (R2) values range from 0 to 1, with values higher than 0.5 considered acceptable. The performance of the hydrological model based on R2 ranges from very good (0.80 < R2 ≤ 1.00), good (0.70 < R2 ≤ 0.80), to satisfactory (0.60 < R2 ≤ 0.70) and unsatisfactory (R2 < 0.60) The PBIAS is an error index which calculates the average tendency of the simulated data to be either larger or smaller than their measured counterparts, with an optimal value of zero (Gupta et al. 1999). The performance of the hydrological model based on PBIAS values is considered as very good (PBIAS < ±10), good (±10 ≤ PBIAS < ±15), satisfactory (±15 ≤ PBIAS < ±25) and unsatisfactory (PBIAS ≥ ±25).
RESULTS
The SWAT model was run by input data from five downscaled and bias corrected GCMs and hypothetically generated LULC change scenarios to investigate the climate and LULC change impact on water balance components spatially averaged over the KRB.
Projected changes in annual and seasonal precipitation
Average annual and seasonal changes in Tmax: (a) RCP 4.5 and (b) RCP 8.5; in Tmin: (c) RCP 4.5 and (d) RCP 8.5; and in precipitation (e) RCP4.5 and (f) RCP8.5 with reference to the baseline period (1974–2004) averaged over two time slice of 70 years (2030–2099) for the 21st century, respectively.
Average annual and seasonal changes in Tmax: (a) RCP 4.5 and (b) RCP 8.5; in Tmin: (c) RCP 4.5 and (d) RCP 8.5; and in precipitation (e) RCP4.5 and (f) RCP8.5 with reference to the baseline period (1974–2004) averaged over two time slice of 70 years (2030–2099) for the 21st century, respectively.
Annual projected precipitation over the KRB was widespread among GCMs from +7% increase by HadGEM2-ES to −24% decrease by IPSL-CM5-LR with respect to baseline under RCP8.5 in the 21st century (Figure 8). Similarly, projected precipitation varied from +5% increase by HadGEM2-ES to −19% decrease by IPSL-CM5-LR under RCP4.5. Only one GCM (i.e., HadGEM2-ES) projected an increase in annual precipitation while all remaining GCMs projected decrease in annual precipitation over the KRB. The change in annual projected precipitation was less than <± 10% by all GCMs except HadGEM2-ES > -20% under both emission scenarios.
The seasonal change in precipitation was much higher than annual over the KRB. For instance, the highest variation in projected precipitation was found during summer (+55 to −31% under RCP4.5 and +66 to −31% under RCP8.5 by HadGEM2-ES and IPSL-CM5-LR, respectively) with respect to baseline over the KRB. Three GCMs (i.e., HadGEM2-ES, MIROC and NoerESM1-M) projected an increase in precipitation, whereas two GCMs (i.e., GFDL-ESM2M and IPSL-CM5-LR) projected a decrease in precipitation during summer. Similarly, three GCMs (HadGEM2-ES, IPSL-CM5-LR and NoerESM1-M) projected decrease in precipitation, whereas two GCMs (GFDL-ESM2M and MIROC) projected increase in precipitation during autumn. All GCMs projected a decrease in precipitation except HadGEM2-ES and GFDL-ESM2M which projected an increase in winter and spring precipitation, respectively, under RCP4.5.
Projected changes in annual and seasonal temperature
In the case of temperature projections, higher agreement was observed among GCMs under both emission scenarios in the 21st century (Figure 8). All GCMs projected an increase in Tmax and Tmin with respect to baseline. The highest increase in annual Tmax was projected by IPSL-CM5-LR (3.5–5.7 °C under RCP4.5 and RCP8.5, respectively), while Tmin by MIROC (3.3–5 °C under RCP4.5 and RCP8.5, respectively). Similarly the lowest increase in annual Tmax and Tmin was projected by GFDL-ESM2M from 2 to 1.96 °C and 3.7 to 3.3 °C under RCP4.5 and RCP8.5, respectively.
The seasonal projected increase in Tmax and Tmin showed higher disparity among GCMs than annual temperature projections (Figure 8). For instance, GFDL-ESM2M projected the lowest increase in Tmax (1.7 and 3.5 °C) while MIROC projected the highest rise in Tmax (4.1 and 6.3 °C) under RCP4.5 and RCP8.5 during spring, respectively. Similarly, GFDL-ESM2M projected the lowest increase in Tmin (1.7–3.1 °C), whereas MIROC projected the highest increase in Tmin (3.6 and 5.3 °C) under RCP4.5 and RCP8.5 in spring, respectively.
Climate change impact on hydrological components
Projected change in water yield (WYLD), evapotranspiration (ET) and soil moisture (SW) under RCP4.5 (a, c and e) and RCP8.5 (b, d and f), respectively, over the KRB.
Projected change in water yield (WYLD), evapotranspiration (ET) and soil moisture (SW) under RCP4.5 (a, c and e) and RCP8.5 (b, d and f), respectively, over the KRB.
All GCMs projected a decrease in ET in the 21st century. The projected decrease in ET by various GCMs varied from −19 to −34% and −13 to −27.5% under RCP4.5 and RCP8.5, respectively. In contrast to annual projected ET, a higher disparity was observed among various GCMs in seasonal projected ET. For instance, all GCMs projected increase in ET from +12 to +27.8% under RCP4.5 in spring, and a decrease in ET during winter and autumn under RCP4.5 and RCP8.5. Furthermore, GCMs projected ET in summer varied from +15.8% (HadGEM2-ES) increase to −28% decrease (IPSL-CM5-LR) under RCP4.5 and +17.7% increase (HadGEM2-ES) to −39% decrease (IPSL-CM5-LR) under RCP8.5.
All GCMs projected an increase in annual soil moisture under both emission scenarios in the 21st century. However, the annual projected soil moisture varied greatly among GCMs (+10% by IPSL-CM5-LR to +68% increase by HadGEM2-ES and +1% increase by IPSL-CM5-LR to +69% increase by HadGEM2-ES under RCP4.5 and RCP8.5, respectively). Similarly, all GCMs projected an increase in soil moisture during winter, spring and autumn but a decrease during summer except HadGEM2-ES.
LULC change impact on water balance components
Average annual change in (a) water yield (WYLD), (b) evapotranspiration (ET), and (c) soil water storage (SW) under hypothetical LULC change scenarios from the baseline scenario in the KRB during 1990–2013.
Average annual change in (a) water yield (WYLD), (b) evapotranspiration (ET), and (c) soil water storage (SW) under hypothetical LULC change scenarios from the baseline scenario in the KRB during 1990–2013.
In scenario-2, where forest cover was a dominant hypothetical land cover (38%), WYLD varied greatly from −6.6% decrease in 2001 to −23% decrease in 2011 from the baseline scenario. In scenario-2, ET and SW were decreased during the simulated period (1990–2013).
In scenario-3 (e.g., grassland as dominant hypothetical land cover), WYLD and ET were projected to decrease while SW was projected to increase continuously during all years (1990–2013). The average annual WYLD was projected to increase under agriculture as dominant hypothetical land cover, while WYLD was projected to decrease under forest and grassland as dominant hypothetical LULC change scenarios.
Average monthly change in (a) water yield (WYLD), (b) evapotranspiration (ET), and (c) soil water storage (SW) under hypothetical LULC change scenarios from the baseline scenario in the KRB during 1990–2013.
Average monthly change in (a) water yield (WYLD), (b) evapotranspiration (ET), and (c) soil water storage (SW) under hypothetical LULC change scenarios from the baseline scenario in the KRB during 1990–2013.
A higher increase in WYLD was projected during summer months with a decrease in WYLD during winter months mostly in agriculture and grass dominated land use (i.e., scenario-1 and scenario-3). However, future forest dominated hypothetical land cover (i.e., scenario-2) would result in a decrease in WYLD during all months especially higher decrease during spring and autumn. Therefore, the impact of future LULC change scenarios on the intensity of seasonal water balance components is more conspicuous than projected changes in average annual water balance components. The conversion of barren land to urban (e.g., 5% increase in urban population) would have a negligible impact on future WYLD from the KRB.
DISCUSSION
For the sustainable use of water resources, it is critical to predict and simulate climate and LULC change impact on future hydrology and quantify water balance components at a watershed scale. In this study, water balance components were simulated using the SWAT model to better understand the water cycle in KRB along with studying the impact of climate and LULC changes on the hydrological cycle.
Climate change projections and impact on hydrology
Precipitation projections
In previous studies, precipitation over the KRB was projected to increase in the 21st century (Khan et al. 2015; Hayat et al. 2019). It was observed that spread of precipitation projected by each GCM over the KRB was largely due to the complex climate and topography of the region (Turner & Annamalai 2012; Lutz et al. 2016). Although subject to large uncertainty, discrepancies in inter-model projections were maximum for precipitation compared to temperature. For instance, HadGEM2-ES projected an increase in average annual precipitation (+6 to +7%), while IPSL-CM5-LR projected a decrease in average annual precipitation (−19 to −24%) (Figure 8). In previous studies, the annual projected precipitation ranged from +5 to +25% over the Western Himalayas (Akhtar et al. 2009; Immerzeel et al. 2010; Forsythe et al. 2014; Lutz et al. 2016). However, some studies projected annual precipitation ranges from +40% increase to −20% decrease in the Western Himalayas and KRB (Chaturvedi et al. 2014; Lutz et al. 2014). In this study, outputs from the four GCMs projected a decrease in precipitation in average annual precipitation (−4 to −24%), which is much different than precipitation projections of the above studies. Similarly, Akhtar et al. (2021) projected a decrease in mean annual precipitation at four rain gauge stations (ranges from −2.1 to −5.0%), and an increase at Nawabad station (6.2%) in the KRB. The projected decrease in annual precipitation was even larger for individual GCMs (i.e., +5 to −24%) in the KRB. Hassanyar & Tsutsumi (2017) projected a slight decrease (−2.2%), while Haider (2018) reported a decrease up to −50% in average annual precipitation in the KRB towards the end of the 21st century. So, these discrepancies in precipitation projections as evident from this study and in literature lead to higher uncertainty in streamflow simulations for the future in the KRB.
The spread of seasonal precipitation projections in the complex topography of the Western Himalayas were very large and difficult to simulate (Turner & Annamalai 2012). Summer monsoon precipitation showed a remarkable spread over the KRB (Figure 8). For instance, the summer monsoon precipitation was projected from +66% increase to −30% decrease. The intensity in the monsoon precipitation was projected to increase at the end of the century over the KRB which was similar to projected precipitation in previous studies (Palazzi et al. 2013; Babur et al. 2016; Lutz et al. 2016; Azmat et al. 2018; Bokhari et al. 2018; Sidiqi et al. 2018; Jury et al. 2020). Sperber & Annamalai (2014) explained that no single model can satisfactorily represent the monsoon regime. In this study, large uncertainties in precipitation projections and trends were more or less similar to what was found in the analysis by Palazzi et al. (2015) from 32 CMIP5 GCMs over the HinduKush Karakoram Himalaya region.
In this study, GCMs show a satisfactory performance in projection of winter precipitation regime (i.e., +25 to −27%) over the KRB. The winter precipitation was projected to decrease in the 21st century which is consistent with earlier studies in this region (Rajbhandari et al. 2014; Babur et al. 2016; Bokhari et al. 2018; Sidiqi et al. 2018). In the KRB, seasonal pattern of precipitation projections showed a higher increase in winter (+25 to 27%) than summer (+6 to +8%) precipitation, but a reduction in spring precipitation (Akhtar et al. 2021). Furthermore, part of the precipitation during winter shows a shift to summer or monsoon rains in the eastern part of the KRB (Akhtar et al. 2021). This shift in winter precipitation needs an urgent response from policymakers and stakeholders to anticipate the water scarcity for winter crops by constructing water storage, dams and reservoirs. However, current research shows that projected precipitation in the KRB is highly complex and uncertain; there were significant conflicts among various GCMs and RCPs scenarios in the annual and seasonal precipitation projections. The results of precipitation projected by various GCMs indicate large discrepancies in the KRB. Therefore, future predictions of water availability in this region are highly uncertain in the long run.
Temperature projections and impact on hydrology
In this study, the projected average increase in annual temperature ranges from 3.0 to 5.2 °C at the end of the century, while the global average annual warming was projected from 1.8 to 4.4 °C under RCPs (RCP4.5 and RCP8.5) by the end of the century (Knutti & Sedláček 2013). The increase in mean annual temperature (4.8 °C) was projected at the end of the century in the KRB (Sidiqi & Shrestha 2021). The projected warming in the Upper Indus Basin (UIB), which includes the KRB, is higher than other parts of the world which ranges from 2.1 to 8 °C by the end of the century (Forsythe et al. 2014; Rajbhandari et al. 2014; Ali et al. 2015; Lutz et al. 2016; Azmat et al. 2018; Jury et al. 2020).
A rise in temperature was projected in the KRB during winter and spring season (Figure 8) which is consistent with earlier studies (Azmat et al. 2018; Hasson et al. 2019). In this study, the highest increase in temperature was projected during winter by the end of the century which is consistent with the results in other studies (Sidiqi & Shrestha 2021). This rise in temperature during the winter and spring season could cause seasonal solid precipitation to decrease in future. Snowmelt was projected to decrease slightly, with a more liquid precipitation during winter under future warming in the UIB (Lutz et al. 2014). According to Bokhari et al. (2018), annual and seasonal warming in the UIB may negatively affect snow accumulation during the winter and has the potential to accelerate glacier melting during summer. This rise in temperature will impede accumulation of snow in the KRB and resulting decrease in the contribution of snow melt to WYLD.
In this study, mean annual WYLD was projected to decrease by all GCMs under both emission scenarios in the KRB. Similarly, Akhtar et al. (2021) projected a decrease in mean annual streamflow in the KRB. This decrease in mean annual streamflow may be attributed to higher warming by the end of the century. The rising temperature will have multiple implications for snow melting and timing of discharge in the KRB. The rise in temperature is alarming for snow accumulation and snow melting in this region. Winter warming may reduce solid precipitation and snow accumulation in the basin, which will ultimately decrease spring flow which is mostly dependent on snow/ice melt in the KRB. Similarly, rising temperature during spring may trigger early melting of snow which will impact on the timing of peak flows from the KRB. Most of the GCMs projected increase in precipitation during summer, while increasing temperature may lead to more liquid than solid precipitation during winter.
LULC change impact on hydrology
Alden Hibbert formulated three hypotheses on the relation between forest cover and water yield: ‘(1) Reduction of forest cover increases water yield. (2) Establishment of forest cover on sparsely vegetated land decreases water yield. (3) Response to treatment is highly variable, and, for the most part, unpredictable’ (Hibbert 1967, p. 535). However, results from various studies varied widely from 50% decrease to 200% increase in water yield (Adams et al. 2012). Goeking & Tarboton (2020) observed that forest disturbance may result in increase, decrease or no change in streamflow. Several studies observed a decrease in WYLD and enhanced evapotranspiration under afforestation scenario (Bi et al. 2009; Kumar et al. 2021). Similarly, an increase in 30% area of forest land resulted in a decrease in monthly WYLD (i.e., −24.2% in October) in the KRB (Figure 10). This decrease in WYLD was much higher during spring and autumn seasons. However, simulation of water balance components by using agriculture as dominant LULC would increase WYLD in the KRB, and afforestation could impact on reduction of water availability from the KRB. The expansion of agriculture land over vegetation area results in an increase in surface flow following rainfall events. However, forests generate less surface runoff to satisfy soil moisture deficit in forests lands after rainfall events (Markham & Anderson 2021). This decrease in WYLD could be attributed to an increase in infiltration. Numerous studies reported a decrease in WYLD and increase in ET under afforestation at watershed scale (Suarez et al. 2014; Mwangi et al. 2016; Guzha 2018). These results suggest that KRB could face a decrease in WYLD in future under the afforestation scenario. Similarly, the KRB would be more exposed to water stress due to enhanced evapotranspiration and water resource managers should pay attention to sustainable water resource management. On the other hand WYLD was projected to increase under the dominant agriculture scenario. This increase in WYLD was also observed in the Tekeze dam watershed in northern Ethiopia where an increase in bare land and agriculture land resulted in an increase in annual and seasonal stream flows (Welde & Gebremariam 2017).
In previous studies, it is well documented that deforestation in a watershed leads to net increase in WYLD from the basin (Arancibia 2013; Welde & Gebremariam 2017; Garcia et al. 2018; Guzha et al. 2018). The areas occupied by fast-growing forest plantations are mainly related to the decrease in water availability to downstream users (Dijk & Keenan 2007; Garcia et al. 2018). These results are consistent with decrease in WYLD during all months in the forest dominated scenario (i.e., scenario-2) from the KRB (Figure 11). The LULC change impact was higher on monthly than annual streamflow (Khoi et al. 2021). Similarly, the results from the KRB also show a higher impact of LULC change on monthly than annual simulated water balance components (Figures 10 and 11).
CONCLUSIONS
This study was conducted to investigate the impact of climate and LULC change scenarios on water balance components in the KRB. All GCMs projected an increase in temperature, while agreement on precipitation projection was lacking among GCMs for the 21st century under RCP4.5 and RCP8.5. Most of the GCMs projected a higher warming during winter and spring seasons in the KRB. The spread of average annual precipitation projections among GCMs is very large, and it is even larger for seasonal precipitation. Winter precipitation projections are more satisfactorily projected than summer precipitation. The higher variations in the simulated water balance components in the KRB are a result of greater variability in the projected climate variables and their responses. For instance, KRB show large uncertainties in the simulation of water balance components which translated from the higher variations in the projected precipitation. The future climate change will result in a decrease in WYLD, with higher flows during winter and earlier spring snowmelt. This reduction in water availability is projected during the critical irrigation period (i.e., summer season). The situation will deteriorate in the future due to decrease in projected WYLD, which could grow worse for water supply to irrigation and power generation in the 21st century. Therefore, water management and adaptation policies would be necessary for sustainable use of water resources in the KRB.
Future hypothetical LULC change scenarios under dominant agriculture land would result in an increase in average annual WYLD (+9%), while afforestation would result a decrease in average annual WYLD (−15%) in the KRB. Under grassland-dominant LULC change scenarios, WYLD was projected to increase during summer (e.g., +33% in June) and to decrease during winter (e.g., −24% in December and February). The impact of climate and LULC change scenarios on water balance components is much higher in terms of seasonal changes than average annual changes. It can be concluded that future LULC change under extreme scenarios will have a less significant impact on water balance components in the KRB compared to water balance simulation under projected climate change. Therefore, climate change will be the dominant driver in hydrological processes in the KRB.
The predicted changes in water balance components from the KRB would have implications for hydroelectricity, industry, downstream irrigation and crop patterns, as well as society in general. Therefore, a better understanding of the climate and LULC change impact on water resources is critical for decision-making in water-dependent sectors. In this regard, scenario-based predictions can be utilized to bridge the knowledge gap for near- and longer-term changes in the hydrology of the KRB. Accurate prediction is still difficult due to the complex topography, climate system and poor capture of this complexity in GCMs over this region. Furthermore, many activities such as construction sites, reservoirs, dams and water abstraction schemes are not accounted for in model building. Despite this, policymakers and stakeholders benefit from the projected potential upcoming changes in streamflow for better water resource management in the KRB. These predicted changes to patterns of water balance components will help in planning to mitigate and improve management of water resources for sustainable water supply under climate and LULC change scenarios to address the current water crisis in the region.
It is suggested that there should be further investigations into the use of the SWAT model to simulate LULC change impact on hydrology at fine spatial scales. This would aid in understanding the impact of detailed LULC categories on streamflow at watershed scales. The calibrated and validated SWAT model can be used to support adaptive water management and to assess and predict other watershed components, such as the impact of climate and LULC on water quality and sediment yield. The performance of the SWAT model can be enhanced by increasing calibration points in the KRB.
DATA AVAILABILITY STATEMENT
Data cannot be made publicly available; readers should contact the corresponding author for details.
CONFLICT OF INTEREST
The authors declare there is no conflict.