Arab J Geosci DOI 10.1007/s12517-015-1985-2
ORIGINAL PAPER
Assessment of impact of climate change on water resources in a hilly river basin Dharmaveer Singh 1 & R. D. Gupta 2 & Sanjay K. Jain 3
Received: 19 January 2015 / Accepted: 1 June 2015 # Saudi Society for Geosciences 2015
Abstract The sensitivity of Sutlej river sub-basin (middle catchment) which is located in N-W Himalaya is investigated for its hydrologic response to potential changes in climate variability. The predictors of two global climate models (GCMs) that are found to perform well over Indian subcontinents are downscaled, and future time series of temperature (maximum and minimum) and precipitation is generated using statistical downscaling model (SDSM) under A1B, A2, and B2 emission scenarios. An overall increase in mean annual temperature and precipitation is predicted under both the models for future periods. The predicted increase in temperature is relatively higher for HadCM3 model compares to CGCM3 model whereas it is opposite for precipitation. The model also predicts considerable shift in monthly pattern of temperature and precipitation. Further, soil and water assessment tool (SWAT) is employed to appraise future changes in stream flow and water balance of the sub-basin under projected climate scenarios. The simulation results show that in future, increase in mean annual stream flow are likely to range from 1.3 to 7.8 % for CGCM3 model and 0.3 to 3.4 % for HadCM3 model, respectively. However, decrease in mean
* Dharmaveer Singh
[email protected] R. D. Gupta
[email protected] Sanjay K. Jain
[email protected] 1
GIS Cell, Motilal Nehru National Institute of Technology Allahabad, Allahabad 211004, India
2
Department of Civil Engineering, Motilal Nehru National Institute of Technology, Allahabad 211004, Uttar Pradesh, India
3
Water Resources Systems Division, National Institute of Hydrology, Roorkee 247667, India
monthly stream flow is predicted under scenarios of CGCM3 model (from January to May and from November to December) and HadCM3 model (from January to April and from October to December), respectively. Keywords Global climate model . Statistical downscaling model . Soil and water assessment tool . Stream flow
Introduction The alteration observed in hydrological cycle of the Earth is attributed to global climate change (Arnell 1999; Xu et al. 2011; Singh et al. 2014). Because of climate change, significant spatial variations in patterns (magnitude and intensity) of temperature as well as precipitation are observed over different regions of the world (IPCC 2007, 2013). Global warming has reduced annual snowpack accumulation, accelerated snowmelt processes, and increased water losses due to evapotranspiration (ET) which, in turn, impacts surface runoff, soil moisture, and recharge rates of the ground water and further renders local water availability and seasonal availability of water supply (Singh et al. 2007; Abu-Allaban et al. 2014). This may deepen the crisis in water sectors which are already in stress as substantial growths observed in population, agriculture, energy, and industrial sectors raised increased demand of water (Arnell et al. 2011; Koutroulis et al. 2013; Matonse et al. 2013). Mountains sometimes also known as Bwater towers^ of the world are major sources of river water. The major portion of their water is stored in the form of snow and glaciers. According to Viviroli et al. (2011), 23 % of mountain area worldwide are essential components of downstream water supply, while another 30 % have supportive role. Mountain snowpack and snowmelt is an important predictor of summer stream flow
Arab J Geosci
and constitutes the primary source of water for large population (Stewart 2009). The major river systems of the world which receive water from mountains are Indus, Ganga, Bramhputra, Irrawaddy, Mekong, Tarim, Changjiang, Yellow, Amazon, etc. About one sixth of the world’s population lives within the snowmelt dominated catchment with low reservoir storage (Barnett et al. 2005). According to the Fourth Assessment Report of Intergovernmental Panel on Climate Change (IPCC), the Himalayan river systems are more inclined to climate change (IPCC 2007). Large-scale retreat and shrinkage in the volume of the Himalayan glaciers are observed due to the enhanced rate of warming which, in turn, modify annual as well as seasonal pattern of flows of the major river systems, i.e., Indus, Ganga, and Brahmaputra (Ageta and Kadota 1992; Yamada et al. 1996). Decrease in mean annual flow of Sutlej river (tributary of Indus River) is observed over the period of 1970–2010 (Singh et al. 2014). The studies based on future scenarios expected that warming may lead to systematic change in the seasonality of snowmelt-dominated rivers (Bates et al. 2008). Shift in seasonality and decrease in the amount of glacial melt will cause a systematic reduction in water availability as well as a reduced buffering effect of glacier runoff during the dry season (Viviroli et al. 2011). This necessitates investigation of plausible impacts of climate change on the hydrology of a basin and its mitigation by adopting proper water resource management practices (Mantua et al. 2010). A number of studies relating the effect of future climate change on water resources have been undertaken in various parts of the globe. The study of Githui et al. (2009) conducted in Western Kenya using soil and water assessment tool (SWAT) predicts increase in stream flow due to projected rise in rainfall. In Latvia, assessment of future climate change (under A2 and B2 emission scenarios) over six river basins has been studied by Latkovska et al. (2012) using HBV model. The results derived from this study reveals decline in mean annual river flow in future (2071–2100) and also predict shift in seasonal pattern as maximum stream flow will occur in winter instead of spring. Another study, performed over Gallego river basin (Spain) by Majone et al. (2012) for A2 scenario using GEOTRANSF model shows that projected water availability for the Gallego is lower for the 2071–2100 period than for 1961–1990, with an increasing number of dry years. Significant changes (ranging from a 17 % decrease to 66 % increase under various emission scenarios) in mean annual flow of Nam Ou river (Laos) has been forecasted due to changes observed in projected temperature and precipitation (Shrestha et al. 2013). Similar kinds of studies have also been undertaken by Chattopadhyay and Jha (2014) for Haw river watershed in North Carolina (USA), Szépszó et al. (2014) for Rhine and Upper Danube rivers in Europe, Morán-Tejeda et al. (2014) for 27 mountain rivers in Spain, Tekle and Tadele (2014) for Bilate Watershed in Ethiopia, and
by Wu et al. (2014) for Heihe river basin in China, respectively. In the present study, sensitivity of Sutlej river basin (middle catchment) is investigated for its hydrologic response to potential changes in climate variability using SWAT model. The effort is build on an integrated approach of watershed modeling with various climate change scenarios. The station-based future time series of temperature and precipitation are generated from the output of global climate models (GCMs) using statistical downscaling model (SDSM) for three emission scenarios: A1B, A2, and B2. The remaining part of this paper is arranged as follows: study area is described and shown in the section BStudy area.^ The types of data and their sources are presented in the section BData sets.^ Methodology is elaborated in the section BMethodology.^ Results of this study are described and discussed in the section BResults and discussions.^ Finally, conclusion drawn from this study is presented in the section BConclusion.^
Study area Sutlej basin, a mountainous river basin is located in NorthWest Himalayan (NWH). The basin stretches from Mansarovar-Rakastal lakes (Western Tibet) in the northeast to Fazilka (Pakistan) in the west. The larger area of Sutlej basin falls under Indian territory while remaining in Tibet and Pakistan, respectively. The large disparities in the topographical relief have resulted in variety of climate causing different types of flora and fauna in the basin. Besides terrestrial flora and fauna, the Sutlej basin is very rich in aquatic life. The entire Sutlej river up to Govind Sagar reservoir and the tributaries of the river are home to 51 fish species belonging to 13 taxonomic families. The vegetation types in this region vary from tropical to alpine. The basin also plays a predominant role in the national energy supply (hydroelectricity) of India as it has potential of generating 10,268.5-MW hydroelectricity. However, only 3267.5 MW is being harnessed while the projects of 357 MW are under construction and 3944 MW is proposed to be installed on the various stages of the Sutlej river (SANDRP 2014). At the same time, it provides water for drinking and agriculture to the state of Himachal Pradesh. This study is performed over middle part of the basin extending from Rampur to Kasol. It has an area of 2457 km2 and located in the state of Himachal Pradesh between 31° 05′ 00″– 31° 39′ 26″ N to 76° 51′ 11″–77° 45′ 17″ E (Fig. 1). The study area presents an intricate mosaic of high mountain ranges, hills, and narrow deep valleys with altitude ranging from 500 m to more than 5000 m. In general, the region has been marked with undulating slopes ranging from 0° to 80° (Fig. 2). Comparatively, the plains lying along the course of Sutlej river and its tributaries have gentle slope (0°–10°). The remaining
Arab J Geosci
Fig. 1 Location map of study area
parts of the basin are characterized by moderate (>10° and <30°) to very steep (>30°) slopes. As one moves from east to west, decrease in relief as well as in slope is found. Climate of study area The climate of the study area varies between sub-tropical (below 2000 m) to temperate (above 2000 m) and characterized by distinct seasonality. There are four seasons: winter (December to March), pre-monsoon (April to June), monsoon (July to September), and post-monsoon (October to November). The major characteristics of prevailing climate in the study area are shown in Fig. 3. The long-term (1970–2005) mean annual temperature is recorded 21.2 °C. The mean monthly TMax varies between 18.6 to 35.7 °C and mean annual TMax is 28.3 °C. The mean monthly TMin ranges from 4.6 to 22.8 °C. The mean annual rainfall is 1033 mm, and around 55 % of total annual rainfall is received from southwest monsoon during monsoon season. The basin is highly sensitive to climate change, and this is confirmed by the studies of Jain et al. (2009) and Singh et al. (2015a). Besides, the variability observed in temperature and precipitation has altered the flow of Sutlej river and resulted to
the decrease in mean annual as well as summer stream flow, respectively (Bhutiyani et al. 2008; Singh et al. 2014). This could have serious implications on water resources. The decline in river flow may affect agriculture and electricity production and also intensify the problem related with drinking water. There is a growing concern how and to what extent future changes in regional temperature and precipitation will affect flow of Sutlej river. Therefore, the present study will provide useful insight to devise better strategy for the management of water resources in the sub-basin.
Data sets A wide range of data sets are used in this study. This is divided into five categories and presented under following subsections: Hydro-meteorological data The hydro-meteorological data used in the present study includes daily observed time series of maximum (TMax) and minimum (TMin) temperature, precipitation (PCP), relative
Arab J Geosci
Fig. 2 Major topographical characteristics within study area
humidity, wind speed, solar radiation, and stream flow (discharge). The temperature (TMax and TMin), precipitation, and discharge data are acquired for the period of 1970–2005 for three stations from Bhakara Beas Management Board
(BBMB), India (Fig. 1). However, relative humidity, wind speed, and solar radiation data on daily time step for the nearest grid of the BBMB’s station is obtained from the website of Global Weather Data for SWAT available at
Fig. 3 Climatic characteristics prevailed in study area (average of 1970–2005 data collected from three stations)
Arab J Geosci
http://globalweather.tamu.edu/. The geographical details of the stations considered for the study in Sutlej basin is provided in Table 1. Predictors: large-scale atmospheric circulation data The predictors which are large-scale atmospheric field data are grouped into two categories: observed predictors (NCEP/NCAR reanalysis gridded data sets) and modelled predictors (GCMs simulated gridded data). NCEP/NCAR reanalysis gridded data sets have been generated using a global data assimilation system based on historical data (Kalnay et al. 1996; Kistler et al. 2001). The NCEP/NCAR reanalysis data sets have a grid spacing of 1.9° latitude×1.9° longitude. The NCEP/NCAR reanalysis predictors have to be re-gridded to conform to the grid-spacing of GCMs. The re-gridded and standardized predictor variables for NCEP/NCAR and GCMs have been directly downloaded from the websites of Data Access Integration (DAI) (http://loki.qc.ec.gc.ca/DAI/ predictors-e.html) and Canadian Climate Impacts Scenarios (CCIS) (http://www.cics.uvic.ca/scenarios/index.cgi), respectively. The GCMs selected in this study are CGCM3 (3.75° latitude×3.75° longitude) and HadCM3 (2.5° latitude×3.75° longitude). CGCM3 is developed by Canadian Centre for Climate Modelling and Analysis, whereas HadCM3 by Hadley Centre for Climate Prediction and Research/Met Office, UK, respectively. The predictors were simulated under historical GHG and aerosol concentration experiment for twentieth century run (20C3M) as well as Special Report on Emission Scenarios (SRES) for future run. The future scenarios considered in this study are A1B and A2 for CGCM3 model and A2 and B2 for HadCM3 model, respectively. The predictor variables are available for period 1961–2100 for CGCM3 model, 1961– 2099 for HadCM3 model, and 1961–2001/2003 for NCEP/ NCAR, respectively. Elevation data Elevation data forms one of the most important sources of ancillary data that are used in hydrological studies. The most common form for the representation of elevation data in digital format is the grid type of raster form (Agarwal and Garg Table 1 Station
Kasol Sunni Rampur
2002). Further, digital elevation model (DEM) has been widely used for representing topographic characteristics of a terrain as it is easily accessible and simple in use. A DEM is a numerical representation of the spatial variation in the land surface elevation, which represents the land surface as a matrix of elevation values (Z) implicitly located by their geographic coordinates (X, Y). Any point in a DEM can be related to its neighboring cell if the data storage is regular (Garg 1991). In this study, DEM of 1″ generated from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is used. This has been downloaded from the website of ASTER Global Digital Elevation Model (GDEM) (http:// gdem.ersdac.jspacesystems.or.jp/). It provides tiles in GeoTIFF format with geographic lat/long coordinates and a 1″ (30 m) grid of elevation postings. It is referenced to the WGS84/EGM96 Geoid. The downloaded tiles of the DEMs are exported to IMG format under ERDAS Imagine 2010 platform. Further, these are mosaicked and re-projected on Projected Lambert Conformal Conical Projection and India_ 1975 datum.
Satellite imagery The classification of land use/land cover types is an integral part in any hydrological modeling. These influence surface as well as ground flows in the basin. The multi-spectral satellite imagery of 23.5-m spatial resolution of Indian Remote Sensing System (IRS)-1D is used in preparation of land use/land cover classes for the study area. The imagery was captured by Linear Imaging Self Scanning Sensor (LISS)-III loaded on IRS-1D satellite for 13 October 2008 (path 94 and row 48). Before analysis, the raw satellite imagery is registered using map (georeferenced Survey of India topographical map) to image registration process under the platform of ERDAS IMAGINE 2010. The root mean square (RMS) error is 0.432 of pixel size. The registered satellite imagery is classified using supervised classification technique. The maximum likelihood classifier is used under ERDAS IMAGINE 2010 environment. The classified image, i.e., the land use/land cover map is shown in Fig. 4. A total number of seven classes are classified, and these are given in Table 2. These are agricultural land (17.1 %), water (0.2 %), forest evergreen (27.9 %), forest deciduous
Geographic and climatic information of stations in Sutlej basin Latitude
31° 21′ 25″ 31° 14′ 15″ 31° 27′ 15″
Longitude
76° 52′ 42″ 77° 06′ 30″ 77° 38′ 40″
Elevation (m)
662 655 976
Average annual temperature (°C) TMax
TMin
28.55 29.02 27.17
16.79 12.27 13.66
Average annual PCP (cm)
Mean annual discharge (cumec)
Period
132.26 101.47 76.34
153150.20 143173.70 127275.29
1970–2005 1970–2005 1970–2005
Arab J Geosci Fig. 4 Land use/land cover map prepared from LISS-III imagery
(24.7 %), forest mixed (3.6 %), fallow and grasses (26.4 %), and snow (0.3 %). The land use/land cover classes are then reclassified according to SWAT data base.
Soil data The soil data and its attributes (soil texture, available water content, hydraulic conductivity, bulk density, and organic carbon content) on scale 1:500,000 are obtained from Food and Agricultural Organization (FAO) GeoNetwork portal (http:// www.fao.org/geonetwork/srv/en/main.home) for the study area. It is in vector format and converted into grid format (Fig. 5). The soil parameters are classified according to SWAT model geodatabase. The main soil categories which fall in Sutlej river sub-basin are dystric cambisols, dystric regosols, and eutric fluviosols. The characteristics of these soil types are shown along with land use/land cover types in Table 2. Table 2
Method for statistical downscaling SDSM is a combination of multiple regressions (MLR) and stochastic weather generator (SWG)-based downscaling methods (Wilby et al. 2002). SDSM has shown advantage over other statistical downscaling approaches such as weather generators and weather typing because of its better ability in describing inter-annual variability. It has widely been used throughout the world to downscale single-site scenarios of daily surface weather variables from predictors of GCMs for assessing hydrologic responses in climate change scenarios (Dibike and Coulibaly 2005; Gagnon et al. 2005; Aherne et al. 2008; Combalicer et al. 2010; Huang et al. 2011; Goyal et al. 2012). In SDSM, generation of station scale weather parameters is linearly conditioned by observed large-scale predictors of atmosphere (j=1, 2……n). The downscaled process is either unconditional or is conditional. The downscaling for the
Characteristics of land use/land cover and soil types
Land use/land cover types
Soil types
Methodology
Classes
Reclassified classes according to SWAT geodatabase
Approx. area (km2)
Area (%)
Agricultural land Water Forest evergreen Forest deciduous Forest mixed Fallow and grasses Snow Dystric cambisols Dystric regosols Eutric fluviosols
AGRL WATR FRSE FRSD FRST BSVG ICES Bd34 Rd30 Je75
421 5 685 605 89 644 8 1060 550 847
17.1 0.2 27.9 24.7 3.6 26.2 0.3 43.1 22.4 34.5
Arab J Geosci Fig. 5 Soil map of the study area
conditional process like daily PCP depends on an intermediate variable such as occurrence of wet day. The occurrence of wet day (Wi) on day i is linearly dependent on predictors Xij. n X
W i ¼ α0 þ
αjX ij
ð1Þ
j¼1
under the constraint 0≤Wi ≤1. The value of Wi varies according to prevailing large-scale weather conditions (represented by the predictor variables) between 0 and 1. The precipitation will occur if uniform random number r≤Wi. Wi is not a Boolean (0 or 1) number but is a continuous variable between 0 and 1 (Wilby and Dawson 2013). For example, on a day with high pressure, Wi might be equal to 0.2. Then, r is used to determine whether a rain day will actually occur depending upon whether r is less than or equal to 0.2. The amount of total PCP (Pi) downscaled on day i with return of wet day is shown in Eq. 2: P i ¼ β0 þ k
n X
β j X ij þ εi
ð2Þ
j¼1
where k is a transformation (fourth root, inverse normal, or logarithmic) which is applied as PCP data is skewed in nature. In case of unconditional processes like daily temperature (TMax and TMin), a direct linear relationship is established between the predict and Ui and selected NCEP/NCAR predictors Xij on individual sites such as the following: U i ¼ γ0 þ
n X
γ j X ij þ εi
ð3Þ
j¼1
where Ui is temperature on day i and Xij is selected NCEP/ NCAR predictors on day i. αj, βj, and γj are regression
coefficients estimated for each month using least-squares regression, and εi is model error. It is generated stochastically using a series of serially independent Gaussian numbers and is added to the deterministic components on daily basis. The major steps adopted for downscaling of TMax, TMin, and PCP involve (Singh et al. 2015b) (1) quality check, transformation, and screening of probable predictors; (2) calibration of monthly sub-model using station scale TMax, TMin, and PCP data and selected predictors of NCEP/NCAR; (3) generation of present and future time series for TMax, TMin, and PCP from the gridded data sets of NCEP/NCAR and GCMs (CGCM3 and HadCM3); and (4) statistical analysis of downscaled projected TMax, TMin, and PCP at each individual station.
Quality control check, transformation, and screening of probable predictors Station-based meteorological data may have errors in terms of missing records or outliers. Quality control check function is used to identify such errors prior to model calibration. The missing data may be replaced by a data identifier code, i.e., −999. In some cases, transformation of predictors or predictands may be of significant interests. SDSM provides facility to transform data before calibration using different types of transformations such as logarithm, power, inverse, lag, and binomial. After quality control check and transformation, screen variable operation is applied to select appropriate sets of observed predictors from the suite of NCEP/ NCAR reanalysis data sets based on scatter plots, correlation, and partial correlation statistics (Wilby and Dawson 2007).
Arab J Geosci
Calibration of monthly sub-model using station scale TMax, TMin, and PCP data and selected predictors of NCEP/NCAR SDSM is calibrated using observed station scale data (TMax, TMin, and PCP) and screened sets of observed predictors, i.e., NCEP/NCAR reanalysis data sets. SDSM offers three different types of sub-models: (1) monthly, (2) seasonal, and (3) annual for the downscaling of predictands (TMax, TMin, and PCP) from the large-scale predictors. The monthly submodel derives 12 different regression equations one for each month whereas seasonal sub-model generates 4 different regression equations one for each season. In case of annual submodel, a single regression equation is generated for all 12 months having same model parameters. The process involved in downscaling may be either unconditional (e.g., TMax, TMin) or conditional (e.g., PCP). There are two methods for optimizing SDSM: (1) dual simplex and (2) ordinary least squares. Both methods provide comparable results, but ordinary least squares is much faster and has been used in the present study. Further, monthly sub-model type is preferred because there are large monthly variations in TMax, TMin, and PCP at different stations within the study region. In monthly sub-model, identical sets of predictors and predictands generate different statistical values for each month. This may be attributed to the fact that different empirical relationships are constructed for different months of the year by this model. The values of E (%) are used to explain how and to what extent daily variations in predictands are determined by regional forcing. Generation of present and future time series for TMax, TMin, and PCP from the gridded data sets of NCEP/NCAR and GCMs After calibrating the model, Weather Generator function is applied to generate ensembles of synthetic daily time series of TMax, TMin, and PCP representing present climate from screened sets of NCEP/NCAR predictors. The synthetically generated daily time series of TMax, TMin, and PCP is compared (in terms of statistics) with observed records to know how close it is to the present climate. Finally, Scenario Generator function is used to simulate future time series of TMax, T Min , and PCP using output of GCMs (CGCM3 and HadCM3) on daily time-step under different emission scenarios. Statistical analysis of downscaled projected TMax, TMin, and PCP At the end, various statistical operations were performed on downscaled projected time series of TMax, TMin, and PCP in order to see changes observed in climate of the study area.
Method for hydrological modeling SWAT model developed by the US Department of Agriculture-Agricultural Research Services (USDA-ARS) is applied for hydrological modeling within the sub-basin. It is based on the works of J. G. Arnold, R. Srinivasan, and S. L. Neitsch (Arnold et al. 1998). SWAT is a continuous (daily or sub-daily), semi-distributed, process-based river basin model. It is designed to simulate surface flow, sub-surface flow, soil erosion, sediment deposition, nutrient fate, and movement in the basin using empirical and physically based equations (Bosch et al. 2004; Arnold and Fohrer 2005; Jayakrishnan et al. 2005; Krysanova and Arnold 2008). There are three important components: sub-basin, reservoir routing, and channel routing in a SWAT model. A complete description of the SWAT model can be found in Neitsch et al. (2005, 2011). Since the objective of the present research is to study hydrological aspects of the sub-basin, only the major aspects of hydrology, a sub-basin component in brief, is presented here. SWAT simulates the hydrology of the watershed in two phases. The land phase of the hydrologic cycle controls the amount of water, sediment, nutrient, and pesticide loadings to the main channel in each sub-basin. The water or routing phase of the hydrologic cycle controls the movement of water, sediment, nutrients, and pesticide loadings through the channel network of the watershed into the outlet (Shrestha et al. 2013). The equation representing water balance (all terms in mm H2O) for land phase of hydrological cycle is given as follows (Nossent 2012): Wt ¼ W0 þ
t X
ðPi−ET i−SQi−LQi−RQi−WUSi−WUSDi−DLiÞ
ð4Þ
i¼1
where Wt is the water content of land phase at time t, Wo is the initial soil water content, t is the time in days, Pi is precipitation on day i, ETi is the amount of evapotranspiration on day i, SQi is the amount of surface runoff on day i, LQi is the amount of lateral flow on day i, RQi is the amount of return flow on day i, WUSi is the amount of water pumped out of aquifer on day i, WUSDi is the amount of water pumped out of deep aquifer on day i, and DLi is the amount of water lost from the system through deep aquifer. SWAT offers three methods, namely, Penman-Monteith, Hargreaves, and Priestley-Taylor, for calculating amount of evapotranspiration in the basin. In comparison to other two methods, the Penman-Monteith method provides a better description of processes, and therefore is used in the present study. However, it requires large input data (Stehr et al. 2008). For estimating volume of surface runoff, either Soil Conservation Service Curve Number (SCS CN) method (USDA-SCS 1972) or Green and Ampt infiltration method is used in SWAT. In case of Ampt infiltration method, input data at a finer-than-daily time resolution are required, whereas
Arab J Geosci
SCS CN method can be applied using daily rainfall values (Johnson 1998). In this study, the SCS CN method is used due to availability of daily rainfall data. The SCS CN is a function of the soil permeability, land use, and the antecedent moisture condition. The CN varies nonlinearly from condition 1 (dry) at wilting point to condition 3 (wet) at field capacity and approaches 100 at saturation. The equation used for deriving surface runoff volume in SWAT using SCS curve number method is given as follows: SQt ¼
ðPt−IaÞ2 ðPt−Ia−S Þ
ð5Þ
where SQt is daily runoff, Pt is daily precipitation and S is retention parameter. Runoff will only occur when precipitation on day t is higher than initial abstraction Ia. These initial abstractions are often approximated as a fraction of the retention parameter, i.e., Ia =0.2*S. The retention parameter S depends on the soil type, land use, management, and slope but also changes over time due to variations in the soil water content. The retention parameter S is defined as
1000 S ¼ 25:4 −10 CN
ð6Þ
SWAT operates on three levels, viz., basin, sub-basin, and hydrological response unit (HRU). SWAT splits the basin into a number of sub-basins based on terrain, river channels and climatic conditions. Further, these sub-basins are divided into multiple HRUs based on spatial variability observed in slope, soil, and land use/land covers within the sub-basins (Cao et al. 2006). HRUs are the smallest homogeneous hydrological units within a sub-basin having spatial uniformity in soils, land use, topographic, and climatic data (Van Liew et al. 2005). It is the basic unit upon which SWAT simulate the water balance (Rostamian et al. 2008). The fluxes (e.g., water flow, sediment, or nutrient flows per unit area) initially computed at HRU are combined into the sub-basin output as a weighted sum as shown in Eq. 7 (Nossent 2012): OUT Sub ¼
n X OUT HRUi*area HRUi area sub i¼1
ð7Þ
where OUT_Sub is the total amount of measured flux (output) in the sub-basin, n is the number of HRUs in the sub-basin, OUT_HRUi is the output from ith HRU, area_HRUi is the surface area of ith HRU, area_sub is the area of sub-basin. The sub-basin outputs are then routed through the river reaches to the catchment outlet. In this study, Arc SWAT 9.0 version is integrated with Arc GIS 9.3 in order to simulate various hydrological components within the study area. The flow chart illustrating various steps followed in hydrological modeling using SWAT model is
given in Fig. 6. The major steps adopted in the present study for simulating various hydrological components are described under subsequent heads. INPUT to SWAT The inputs like DEM, land use/land cover map, soil map, meteorological data (maximum temperature, minimum temperature, precipitation, relative humidity, wind speed, and solar radiation), and stream flow data are required initially for model setup. Delineation of watershed A threshold of 200 ha is employed for delineating the watershed from ASTER DEM and generating drainage within the study area. The study area as a whole is divided into four subbasins and parameters such as slope gradient, slope length of terrain, channel slope, channel length, and width have been derived. The sub-basins delineated from ASTER DEM are shown in Fig. 7. Defining number of HRUs In this step, number of HRUs is generated by overlaying land use/land cover map and soil map according to defined slope classes. A total number of 80 HRUs classes have been generated based on thresholds of 5 % for land use/land cover, 20 % for soil, and 20 % for slope. Importing weather/ meteorological data SWAT requires meteorological data for defining climate within the basins. The data required include maximum temperature, minimum temperature, precipitation, relative humidity, wind speed, and solar radiation. In this study, meteorological data of three stations (Kasol, Sunni, and Rampur) of period 1970– 2005 (observed) and 2011–2040, 2041–2070, and 2071–2099 (projected) on daily time step is used for this purpose. Run developed SWAT model Finally, the developed SWAT model is run and its results of calibration and sensitivity analysis are checked. Once satisfactory results are obtained for calibration with identified sensitive parameters, the model is used to simulate the various hydrological components. Method for model evaluation and uncertainty analysis Five evaluation criteria are used to assess daily stream flow simulated by SWAT. These are coefficient of determination (R2), Nash-Sutcliffe coefficient of efficiency (NSE), percent
Arab J Geosci Fig. 6 Methodology adopted in hydrological modeling using SWAT
bias (PBIAS), sum of squares of the difference of the measured and simulated values after ranking (SSQR), and total mass balance controller (TMC). These are expressed as 12 n X ðOi −O ÞðPi −P Þ C B C B i¼1 C q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi R2 ¼ B q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Xn C B Xn 2 @ ðOi −O Þ Pi −P Þ2 A 0
n X
N SE ¼ 1−
ð8Þ
i¼1
i¼1
i¼1 n
X
ðOi −Pi Þ2 2 ðOi −O Þ
ð9Þ
i¼1
X n
PBIAS ¼
ðOi −Pi Þ
i¼1 n X
100
ð10Þ
ðOi Þ
i¼1
1X ðOi −Pi Þ2 n i¼1 2 n 3 X Pi 7 6 6 i¼1 7 6 T M C ¼ 100abs6 n −17 7 4X 5 Pi n
SSQR ¼
i¼1
ð11Þ
ð12Þ
where Pi are the predicted values, Oi are the observed values, n is the number of samples, Ō is the mean of the observed data, and Pi is the mean of predicted data, respectively. R2 explains correlation between the observed and simulated values and lies between 0 (poor) to 1 (best). Similar to R2, NSE is also used to represent predictive power of hydrological models. Its value lies between −∞ to 1 (Nash and Sutcliffe 1970). If the value of NSE is ≥0.75, the simulated result is considered to be good while for values of efficiency between 0.75 and 0.36, the simulation results are considered to be satisfactory and below 0.36, it is considered to be unsatisfactory (Motovilov et al. 1999). PBIAS computes average tendency of simulated flow, whether it is larger or smaller than the observed value. In optimal condition, it has value of 0 whereas a positive value indicates that the model underestimates the prediction and negative value shows overestimation of the prediction (Gupta et al. 1999). According to Van Liew et al. (2005), the values of PBIAS less than ±20 % is considered to be good while values between ±20 and ±40 % are considered to be satisfactory, and those greater than ±40 % are considered unsatisfactory. SSQR represents the fitting of the frequency distributions of the observed and simulated series whereas TMC estimates the deviation from the measured mass or volume (Van Griensven and Bauwens 2003). For uncertainty analysis in SWAT, the Sequential Uncertainty Fitting (SUFI-2) algorithm is applied where parameter uncertainty accounts for all sources of uncertainties such as
Arab J Geosci
divided by the standard deviation of the measured data) is the measure of uncertainty. The goodness of calibration and prediction uncertainty is judged on the basis of closeness of the p factor to 100 % (i.e., all observations bracketed by the prediction uncertainty) and the r factor to 1 (i.e., achievement of rather small uncertainty band) (Singh et al. 2013).
Results and discussions Projection of future time series of temperature (TMax and TMin) and precipitation
Fig. 7 Delineated sub-basins from ASTER DEM
uncertainty in driving variables (e.g., rainfall), conceptual model, parameters, and measured data. Uncertainty of input parameters is depicted as uniform distributions, while model output uncertainty is quantified by the 95 % prediction uncertainty (95PPU) calculated at 2.5 and 97.5 % levels of the cumulative distribution of output variables obtained through Latin hypercube sampling (Abbaspour et al. 2007). SUFI-2 starts by assuming a large parameter uncertainty so that the measured data initially fall within the 95 PPU, then it decrease this uncertainty in steps until two rules are satisfied: (1) the 95 PPU band brackets most of the observations, and (2) the average distance between the upper (at 97.5 % level) and the lower (at 2.5 % level) parts of the 95 PPU is small (Abbaspour et al. 2007; Shrestha et al. 2013). A detailed description of SUFI-2 optimizing algorithm is given in the study of Yang et al. (2008). The Bp factor^ (percentage of measured data bracketed by the 95 % prediction uncertainty (95 PPU)) along with Br factor^ (average thickness of the 95 PPU band Table 3
The developed SDSM model is applied to downscale and generate future scenarios of daily temperature (TMax and TMin) and precipitation from predictors of CGCM3 (SRES A2 and A1B) and HadCM3 (SRES A2 and B2) models for the period of 2011–2099. Further, the future period is grouped into 2020s (2011–2040), 2050s (2041–2070), and 2080s (2071–2099) for studying pattern of change in temperature and precipitation with reference to base line period (1971–2000). The baseline (1971–2000) corresponds to the observed data. The biases from daily time series of temperature and precipitation data are removed before the analysis (Singh et al. 2015a, b). For the analysis, daily values of temperature and precipitation are summed to obtain annual and monthly values at each station. Further, mean annual and monthly values of temperature and precipitation are attained from all the stations. Change in future annual and monthly temperature (TMax and TMin) The change in mean annual TMax and TMin in Sutlej river subbasin under scenarios A1B, A2 of CGCM3 model, and A2, B2 of HadCM3 model is given in Table 3. The rise in TMax and
Projected mean annual change in temperature (TMax and TMin) and precipitation under different emission scenarios in 2020s, 2050s, and 2080s
Model
Future time period
Scenario
Change in TMax (°C)
Change in TMin (°C)
Change in PCP (cm)
CGCM3
2020s
A1B A2 A1B A2 A1B A2 A2 B2 A2 B2 A2 B2
0.45 0.66 0.78 0.81 0.82 1.12 0.60 0.58 1.06 0.76 1.82 1.25
1.08 1.13 1.61 1.74 2.05 2.66 1.18 1.18 2.24 1.87 3.43 2.52
25.09 21.79 50.22 45.0 58.40 88.48 9.29 8.34 45.10 32.27 87.38 24.38
2050s 2080s HadCM3
2020s 2050s 2080s
Arab J Geosci
Fig. 8 Projected mean monthly change in TMax and TMin under A1B and A2 scenarios of CGCM3 model
TMin is predicted in all the future scenarios obtained from both the models. In case of CGCM3 model for TMax, increase is 0.45, 0.78, and 0.82 °C under A1B scenario and 0.66, 0.81, and 1.12 °C under A2 scenario for future periods of 2020s, 2050s, and 2080s, respectively. For TMin under scenarios A1B and A2, this is 1.08, 1.61, and 2.05 °C and 1.13, 1.74, and 2.66 °C in 2020s, 2050s, and 2080s, respectively. Similarly for HadCM3 model, increase in TMax under A2 scenario is 0.60, 1.06, and 1.82 °C and under B2 scenario is 0.58, 0.76, and 1.25 °C for all three future periods. For TMin under scenarios A2 and B2, it is 1.18, 2.24, and 3.43 °C and 1.18, 1.87, and 2.52 °C in 2020s, 2050s, and 2080s, respectively. The projected increment is high for A2 scenario because it has the highest concentration of carbon dioxide (CO2), i.e.,
850 ppm and it is followed by A1B (720 ppm) and B2 (450 ppm) scenarios, respectively. The change of pattern in mean monthly TMax and TMin under scenarios A1B and A2 of CGCM3 model with respect to base period for all three future periods is shown in Fig. 8. The significant increase in mean monthly TMax is predicted from January to May and from September to December under A1B and A2 scenarios. It is in the range of 0.02 to 1.63 °C, 0.68 to 1.69 °C, and 1.20 to 2.46 °C under A1B scenario and 0.05 to 2.10 °C, 0.44 to 2.26 °C, and 1.41 to 3 °C under A2 scenario in 2020s, 2050s, and 2080s, respectively. The highest increase in TMax (3 °C) is anticipated in month of March under A2 scenarios in 2080s. On the contrary, substantial decrease in TMax is predicted in months of June, July, and August under
Fig. 9 Projected mean monthly change in TMax and TMin under A2 and B2 scenarios of HadCM3 model
Arab J Geosci Fig. 10 Projected mean monthly change in precipitation under A1B and A2 scenarios of CGCM3 model
both the scenarios for future periods whereas highest decrease can be seen in the month of July. In case of TMin, increase is observed throughout year under scenarios A1B and A2 of CGCM3 model for all three future periods and it is more prominent in the month of October. The predicted increase in monthly TMin is in the range of 0.10 to 2.48 °C, 0.29 to 3.05 °C, and 0.49 to 4.38 °C under A1B scenario and 0.06 to 2.39 °C, 0.42 to 3.76 °C, and 0.75 to 5.10 °C under A2 scenario in 2020s, 2050s, and 2080s, respectively. Similarly, projected change in mean monthly TMax and T Min for future periods under A2 and B2 scenarios of HadCM3 model is shown in Fig. 9. The overall rise in mean monthly TMax is predicted from January to May and from September to December whereas decline in months of June, July, and August under scenarios A2 and B2 in 2020s, 2050s, and 2080s. The increase in TMax is in the range of 0.16 to 1.61 °C, 0.07 to 1.98 °C, and 0.70 to 1.79 °C under A2 scenario and 0.27 to 1.74 °C, 0.74 to 2.05 °C, and 0.76 to 2.62 °C under B2 scenario, respectively. The highest increase and decrease in TMax is expected in months of April and June, respectively. The increase observed in mean monthly TMin is in accordance with the results obtained from CGCM3 model. Fig. 11 Projected mean monthly change in precipitation under A2 and B2 scenarios of HadCM3 model
This is found in range of 0.35 to 2.58 °C, 1.10 to 3.45 °C, and 1.74 to 4.85 °C under A2 scenario and 0.51 to 2.34 °C, 0.96 to 3.30 °C, and 2.6 °C to 3.89 °C under B2 scenario, respectively. Change in future annual and monthly precipitation The analysis of downscaled precipitation projects rise in mean annual precipitation in Sutlej basin for the future periods (2020s, 2050s, and 2080s) under all scenarios of both the models (Table 3). This increase in precipitation may be attributed to an increase in the surface temperature which in turn may raise rate of evaporation leading to increased precipitation (Anandhi et al. 2008). The increase in mean annual precipitation (with respect to base period) is expected in range of 24.0 to 55.4 % under A1B scenario and 20.8 to 84.5 % under A2 scenario of CGCM3 model. For A1B and A2 scenarios, the maximum (55.4 and 84.5 %) and minimum (24.0 and 20.8 %) increase in mean annual precipitation is observed during 2020s and 2080s, respectively. Similarly for HadCM3 model, the increase is observed in range of 8.9 to 83.4 % and 8.0 to 30.8 % under A2 and B2 scenarios. Under both the
Arab J Geosci Table 4
Lists of parameters and their description used in SWAT model parameterization
Parameters
Description
Units
Parameters governing surface water response CN2 SCS runoff curve number ESCO Soil evaporation compensation factor EPCO Plant uptake compensation factor SOL_AWC Available soil water capacity SOL_K Saturated hydraulic conductivity SOL_Z Soil depth SOL_ALB Moist soil albedo Parameters governing subsurface water response GW_REVAP Groundwater Brevap^ coefficient REVAPMN Threshold depths of water in the shallow aquifer required for Brevap^ to occur GWQMN Threshold depths of water in the shallow aquifer required for return flow to occur GW_DELAY Groundwater delay ALPHA_BF Baseflow alpha factor or recession constant Parameters governing basin response SURLAG Surface runoff lag time CH_K2 Channel hydraulic conductivity CH_N2 Manning’s n value for the main channel Slope Slope
models, maximum increase in mean annual precipitation is reported for A2 scenario during 2020s, 2050s, and 2080s. Further, pattern of change in future mean monthly precipitation is shown in Fig. 10 for CGCM3 model under A1B and A2 scenarios. In this case, significant increase in precipitation with varying amount is projected in the months of April (1.2 to 2.1 cm under A1B and 0.6 to 2.7 cm under A2), June (1.7 to 13.2 cm under A1B and 1.3 to 23.3 cm under A2), July (7.3 to 14.7 cm under A1B and 5.1 to 18.1 cm under A2), August(7.4 to 14.1 cm under A1B and 7.1 to 17.2 cm under A2), September (4.9 to 11.0 cm under A1B and 4.3 to 18.4 cm under A2), Table 5 Lists of the ten most sensitive parameters used in calibration of SWAT model Parameters
Rank
Upper bound
Lower bound
Fitted values
CN2 ALPHA_BF ESCO REVAPMN SOL_AWC CH_K2 SOL_Z GW_REVAP CH_N2 SURLAG
1 2 3 4 5 6 7 8 9 10
0.04 0.90 0.95 500.30 0.15 86.20 0.14 0.30 0.16 12.09
−0.28 0.33 0.84 234.70 −0.34 1.45 −0.12 0.10 0.13 11.75
−0.28 0.76 0.95 451.69 0.09 3.40 0.07 0.14 0.14 12.07
None None None mm/mm mm/h mm Fraction None mm mm days days days mm/h None Degree
and October (0.7 to 2.4 cm under A1B and 2.0 to 2.6 cm under A2) during 2020s, 2050s, and 2080s, respectively. Besides, a slight increase in precipitation under A1B and A2 scenarios is also observed in months January and March during 2020s, 2050s, and 2080s followed by May and November during 2050s and 2080s. However, slight decrease (<1.0 cm) in precipitation is recorded during future periods in February under A1B and A2 scenarios and in December under A1B scenario, respectively. The results derived under A2 and B2 scenarios of HadCM3 model are shown in Fig. 11 and it shows large variability in pattern of monthly future precipitation. A small decrease (≤0.1 cm) in precipitation is observed during A2 and B2 scenarios in months of January, March, November, and December followed by considerable rise in July (4.3 to 14.2 cm under A2 and 4.5 to 12.9 cm under B2), August (3.7 to 8.2 cm under A2 and 3.7 to 8.2 cm under B2), and September (0.04 to 12.8 cm under A2 and 3.2 to 10.1 cm under B2) during 2020s, 2050s, and 2080s. The increase observed in projected precipitation during monsoon season (June, July, August, and September) under both the models could be the result of the projected intensification of the heat low over N-W India, the trough of low pressure over the Indo-Gangetic plains, and the land-ocean pressure gradient during the establishment phase of the monsoon (Kripalani et al. 2007). Besides, small increase in 2020s and 2050s is observed in precipitation followed by decrease in 2080s under A2 scenario in months of February
Arab J Geosci
Fig. 12 Correlation between daily observed and simulated stream flow measured at Kasol during calibration [11 (a)] and validation [11 (b)]
and October. However, under B2 scenario, increase in precipitation is observed in February during 2020s, 2050s, and 2080s followed by decrease in May and October, respectively. Sensitivity analysis Before calibration, a sensitivity analysis is performed for the control point, i.e., outlet of the sub-basin (Kasol in the present study) in order to find the lists of the most suitable model parameters. In SWAT-CUP, the two algorithms namely SUFI-2 and generalized likelihood uncertainty equation (GLUE) are used for the parameter optimization (Abbaspour et al. 2004, 2007). In the present study, SUFI-2 algorithm which provides comparable results with widely used autocalibration methods is applied for the sensitivity analysis, calibration, validation, and uncertainty analysis of SWAT models (Yang et al. 2008; Shrestha et al. 2013). Here, a total number of sixteen parameters which governs rainfall-runoff processes as shown in Table 4 are used for model parameterization. According to Eckhardt et al. (2005), the range of parameter values applied in the calibration process must be physically plausible so that the model can be Table 6 SWAT model performance for calibration (excluding warm-up period) and validation periods Period
R2
NSE
PBIAS
SSQR
TMC
Calibration (1975–2000) Validation (2001–2005)
0.96 0.93
0.94 0.90
10.17 % 19.60 %
6007.37 18,077.92
6.20 19.70
employed afterward for assessing the impact of change scenarios. The parameters are ranked according to their sensitivity determined by Latin Hypercube Sampling-One at A Time (LH-OAT) analysis method. Out of 16, the lists of ten most sensitive parameters which are used to calibrate the model along with their upper bound, lower bound, and fitted values are given in Table 5. For this study, SCS curve number (CN2) is found as the most sensitive parameter and it is followed by baseflow alpha factor (ALPHA_BF). Calibration and validation of SWAT model Calibration and validation of SWAT model are two critical issues in hydrological modeling. In calibration, the model parameters are adjusted so that simulated results have better agreement with observed/measured in the field. Split sample approach is adopted in model testing. The daily time series of 31-year data (1970–2000) including initial 5 years (1970– 1974) as Bwarm-up period^ is used for calibrating the model while the model is validated using independent data of periods 2001–2005. This warm-up period is used for the estimation of several parameters of the model, the initial values of which are not known (Bekiaris et al. 2005). The calibration is carried out both manually and automatically using SWAT-CUP software. A comparison of simulated daily stream flow with observed data is carried out at the outlet of the sub-basin (Kasol) for both calibration and validation period (Fig. 12). The results of calibration and validation are shown in Table 6. The simulated daily flow shows good agreement with the
Arab J Geosci Fig. 13 Uncertainty analysis results of stream flow for the calibration period
observed values for the calibration period with R2 =0.96, NS= 0.94, and PBIAS=10.17 %. For the validation period, the simulated and observed daily flows also revealed good agreement as indicated by the values of R2, NS, and PBIAS being 0.93, 0.90, and 19.60 %, respectively. The positive values of PBIAS obtained during calibration and validation periods indicate that the stream flow is under predicted by SWAT model. In addition, the model is not able to capture peak flows and mismatch in peak flows of simulated and observed flows are noticed. This may be attributed to precipitation data and also errors in the observed stream flow data, especially during high flows (Shrestha et al. 2013). The results of uncertainty analysis computed for daily stream flow during calibration period is shown in Fig. 13. The majority of the observed data is inside or very close to the predicted bands and this is also reflected by high values (70 %) of p factor. The value of r factor is also found reasonably good, and it is equal to 0.13. However, some events are outside the predicted bands and indicate under estimation of these events by the model. According to Abbaspour (2011) uncertainties in the model prediction may arise due uncertainties associated with model structure (conceptual model uncertainty), input data uncertainties and parameter uncertainties. In SUFI-2, these uncertainties sources are combined
Table 7 Mean annual water balance of the sub-basin for observed periods (1970–2000)
and total model uncertainty is evaluated by using two parameters (p factor and r factor) and represented by 95 PPUs. Estimation of water balance for observed periods (1970–2000) The mean annual water balance of the sub-basin simulated by SWAT for observed period (1970–2000) is given in Table 7. The study area as a whole receives 965.5 mm as precipitation out of which 572.5 mm is lost due to evapotranspiration (ET). Another, 188.42 and 54.16 mm of total precipitation is received in form of surface runoff and lateral flow, respectively. Some amount of precipitation falling on the ground is stored deep inside the ground through infiltration and percolation processes. The estimated mean annual water yield of the basin is 341.7 mm of the total precipitation falling within the study area. This is the amount of available water which can be utilized for drinking as well as irrigation purposes. However, large variations in mean monthly amounts of water yields are observed and this is shown in Table 8. The study area receives approximately 48 % of its total mean annual water yield during months of monsoon season (July, August, and September) where precipitation are maximum over the year. The highest mean monthly water yield is estimated for August
Hydrological components
Amount (mm)
Water balance ratio
Precipitation Snow fall Snow melt Surface runoff Lateral flow Groundwater (shallow aquifer) Revap Deep aquifer recharge Total aquifer recharge Total water yield Percolation out of soil Evapotranspiration (ET) Potential evapotranspiration (PET) Transmission Losses
965.50 0.06 0.06 188.42 54.16 100.47 22.06 30.63 153.14 341.70 151.66 572.50 2815.50 1.35
Stream flow/precipitation=0.36 Surface runoff/total flow=0.55 Base flow/total flow=0.45 Percolation/precipitation=0.16 Deep recharge/precipitation=0.03 ET/precipitation=0.59
Arab J Geosci Table 8
Mean monthly water balance of the sub-basin for observed periods (1970–2000)
Month
Rain (mm)
Snow fall (mm)
Surface flow (mm)
Lateral flow (mm)
Water yield (mm)
ET
PET
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
58.91 67.08 80.31 42.4 62.41 103.58 204.5 174.52 102.52 24.18 14.25 32.58
0.06 0 0 0 0 0 0 0 0 0 0 0
11.04 11.22 15.2 5.12 4.28 8.36 38.78 46.18 31.81 7.1 2.28 7.32
2.77 3.94 5.39 3.26 2.34 2.67 6.82 10.77 8.79 4.12 1.88 1.5
19.57 20.36 27.84 16.91 13.96 15.26 48.96 64.2 53.78 27.13 16.7 17.54
25.31 33.92 44.72 75.27 86.44 71.53 82.57 61.39 46.42 19.74 12.52 13.49
134.77 149.4 219.43 305.98 391.83 348.41 214.19 191.76 254.32 264.99 200.85 143.21
month (high amount of rainfall and low rate of evaporation) and the lowest for May month (high rate of evaporation and low amount of rainfall), respectively. Impact of climate change on water resources The calibrated SWAT model is applied further to simulate various hydrological components in order to study hydrological response of the sub-basin to potential changes in climate variability (temperature and precipitation). The future time series of temperature (TMax and TMin) and precipitation is generated from predictors of CGCM3 and HadCM3 models after downscaling (section BProjection of future time series of temperature (TMax and TMin) and precipitation^). Changes observed in stream flow and water balance components with respect to base period under different emission scenarios (A1B, A2, and B2) is discussed under sections BProjected change in stream flow^ and BProjected change in water balance for future periods (2020s, 2050s and 2080s),^ respectively. Projected change in stream flow Generally, persistent increase in mean annual flow of Sutlej river is predicted (measured at Kasol) by SWAT model
Table 9
for future periods under all emission scenarios of both the models except in 2020s where a slight decrease in mean annual stream flow is observed for B2 emission scenario of HadCM3 model (Table 9). Increase is higher under scenarios of CGCM3 model compare to HadCM3 model. The predicted increase in mean annual flow is 1.3, 3.8, and 5.8 % under A1B scenario and 1.3, 3.2, and 7.8 % under A2 scenario of CGCM3 model for future periods of 2020s, 2050s, and 2080s, respectively. For HadCM3 model, it is 0.3, 0.6, and 3.4 % under A2 scenario and −0.8, 1.6, and 0.9 % under B2 scenario. The maximum increase in mean annual flow is predicted during 2080s under A2 scenario of both the models. The increase observed in predicted stream flow is outcome of heavy precipitation predicted by SDSM model. The change in mean monthly pattern of stream flow with respect to base period under scenarios A1B and A2 of CGCM3 model for all three future periods is shown in Fig. 14. In general, decrease in mean monthly stream flow is predicted from January to May and from November to December. It is in the range of 1.90 to 10.70 °C, 0.40 to 12.02 %, and 0.13 to 12.53 % under A1B scenario and 0.51 to 10.78 %, 0.73 to 11.99 %, and 0.50 to 13.05 % under A2 scenario in 2020s, 2050s, and 2080s, respectively. The highest decrease in stream flow is anticipated in
Projected mean annual change in stream flow under different emission scenarios in 2020s, 2050s, and 2080s
CGCM3 model
HadCM3 model
Change in stream flow (%) under A1B scenario
Change in stream flow (%) under A2 scenario
Change in stream flow (%) under A2 scenario
Change in stream flow (%) under B2 scenario
2020s
2050s
2080s
2020s
2050s
2080s
2020s
2050s
2080s
2020s
2050s
2080s
1.3
3.8
5.8
1.3
3.2
7.8
0.3
0.6
3.4
−0.8
1.4
0.9
Arab J Geosci Fig. 14 Projected mean monthly change in stream flow under A1B and A2 scenarios of CGCM3 model
month of March (13.05 %) under A2 scenarios in 2080s. Opposite to this, overall increase in stream flow is predicted in months of June, July, August, and September, i.e., monsoon season under both the scenarios for future periods whereas highest increase can be seen in the month of July (5.84 to 14.83 % under A1B scenario and 5.64 to 20.05 % under A2 scenario). Similarly, projected change in mean monthly stream flow for future periods under A2 and B2 scenarios of HadCM3 model is shown in Fig. 15. The overall decrease in mean monthly stream flow is predicted from January to April and from October to December whereas increase in months of June, July, August, and September under scenarios A2 and B2 in 2020s, 2050s, and 2080s. The decrease in mean monthly stream flow is in range of 0.94 to 10.03 %, 1.99 to 11.91 %, and 2.16 to 13.47 % under A2 scenario and 1.37 to 10.99 %, 2.21 to 11.23 %, and 1.58 to 11.58 % under B2 scenario, respectively. However, increase in mean monthly stream flow is found in range of 3.15 to 4.29 %, 0.88 to 4.54 %, and 3.13 to 7.92 % under A2 scenario and 0.58 to 2.51 %, 0.99 to 6.70 %, and 2.63 to 6.27 % under B2 scenario, respectively.
Fig. 15 Projected mean monthly change in stream flow under A2 and B2 scenarios of HadCM3 model
Projected change in water balance for future periods (2020s, 2050s, and 2080s) The predicted mean annual water balance of study area for future periods under scenarios A1B and A2 of CGCM3 model is given in Table 10. It is observed that predicted rise in the range of 26.45 to 59.91 % under A1B scenario and 22.25 to 80.58 % under A2 scenario in precipitation have significant impact on various hydrological components within study area. The increase in surface runoff and total amount of water yield is predicted under both scenarios of CGCM3 model for future periods (excluding in 2020s under A2 scenario). The model based projected change in surface runoff for 2020s, 2050s, and 2080s is 24.80, 21.06, and 50.64 % under A1B scenario and −23.01, 13.09, and 83.68 % under A2 scenario whereas change in water yield is 25.3, 65.15, and 93 % and 19.24, 54.95, and 125.69 %, respectively. In addition, increase in ET (ranging from 27.19 to 52.92 % under A1B scenario and ranging from 34.28 to 67.70 % under A2 scenario) and Potential evapotranspiration (PET) is also predicted for future periods. However, decrease in volume of total aquifer
Arab J Geosci Table 10 Projected mean annual water balance of the sub-basin for future periods (2020s, 2050s, and 2080s) under A1B and A2 emission scenarios of CGCM3 model Hydrological components (all units in mm)
A1B scenario 2020s
A2 scenario 2050s
2080s
2020s
2050s
2080s
Precipitation
1220.9
1423.6
1544
1180.4
1375
1743.5
Snow fall Snow melt Surface runoff Lateral flow Groundwater (shallow aquifer) Revap Deep aquifer recharge Total aquifer recharge Total water yield Percolation out of soil Evapotranspiration (ET) Potential evapotranspiration (PET) Transmission losses
0 0 235.28 66.14 128.52 25.07 38.44 192.18 428.18 190.85 728.2 3001.4 1.77
0 0 228.11 295.55 43.45 5.68 2.59 51.71 564.34 48.97 852.3 2904.8 2.77
0 0 283.84 320.46 58.33 7.09 3.44 68.84 659.5 65.76 875.5 3093.3 3.12
0 0 145.07 240.7 23.51 22.06 1.43 28.58 407.45 26.68 768.8 3016.1 1.91
0 0 213.1 282 36.97 5.29 2.22 44.46 529.48 41.89 839.1 2908.8 1.91
0 0 346.09 359.58 69.16 8.29 4.08 81.51 771.19 77.89 960.1 3143.9 3.64
recharge is predicted under both the scenarios for future periods. More or less similar patterns are predicted for monthly water balance where the highest rise in surface runoff and water yield is observed during the monsoon seasons. Table 11 shows projected future status of different hydrological components of study area under A2 and B2 scenarios of HadCM3 model. The study area as a whole reveals rise in mean annual water yield under both scenarios of HadCM3 model for future periods except in 2020s where a slight decrease in mean annual water yield is predicted under B2
scenario. The projected change in water yield for 2020s, 2050s, and 2080s is 11.38, 14.65, and 37.47 % under A2 scenario and −4.82, 26.67, and 14.95 % under B2 scenario, respectively. Opposite to CGCM3 model, a decrease in surface runoff predicted under scenarios of HadCM3 model for the future periods. This may be linked with comparatively higher rise in projected mean annual temperature and lower rise in precipitation (Table 2) for HadCM3 model. In addition, similar patterns as observed in case of CGCM3 model are predicted for ET, PT and total aquifer recharge for future
Table 11 Projected mean annual water balance of the sub-basin for future periods (2020s, 2050s, and 2080s) under A2 and B2 emission scenarios of HadCM3 model Hydrological components (all units in mm)
Precipitation Snow fall Snow melt Surface runoff Lateral flow Groundwater (shallow aquifer) Revap Deep aquifer recharge Total aquifer recharge Total water yield Percolation out of soil Evapotranspiration (ET) Potential evapotranspiration (PET) Transmission losses
A2 scenario
B2 scenario
2020s
2050s
2080s
2020s
2050s
2080s
1071.2 0 0 150.54 212.05 19.58 2.73 1.17 23.47 380.61 21.91 688 3014.9 1.56
1089.4 0 0 160.08 216.14 17.21 2.7 1.05 20.94 391.77 19.29 695.2 2954.4 1.66
1144.4 0 0 226.25 220.4 24.88 3.64 1.5 30 469.76 28.25 669.8 3240.8 1.78
1000.3 0 0 111.27 201.5 13.87 2.3 0.85 17.01 325.2 15.57 673.4 3016.6 1.44
1166.6 0 0 177.29 234.7 22.75 3.39 1.38 27.5 432.86 25.65 729.5 2811 1.87
1082.8 0 0 156.74 218.01 19.32 3.14 1.18 23.63 392.38 21.95 687.4 3137.5 1.69
Arab J Geosci
periods; however, deviation in amount is noticed. The increase for ET likely ranges from 16.99 to 21.43 % under A2 scenario and 17.62 to 27.42 % under B2 scenario in future (Table 11).
Conclusion SWAT model is applied for studying hydrological behaviour of a hilly terrain basin (sub-basin of Sutlej basin) located in NW Himalaya for future periods. The model is calibrated with the ten most sensitive parameters (CN2, ALPHA_BF, ESCO, REVAPMN, SOL_AWC, CH_K2, SOL_Z, GW_REVAP, CH_N2, SURLAG) before simulating hydrology of the subbasin for the observed (1970–2000) as well as future periods. A comparison of simulated daily stream flow with observed data along with uncertainty analysis is carried out at outlet of the sub-basin (Kasol) for calibration period. The simulated daily flow show good agreement with the observed values for the calibration period with R2 = 0.96, NS = 0.94, and PBIAS=10.17 %. The results of uncertainty analysis indicate that majority of the observed data is inside or very close to the predicted bands, and this is also reflected by high values (70 %) of p factor. The model is also validated by using independent time series of observed stream flow for periods of five years (2001–2005). Again high correlation is observed between the simulated and observed daily as indicated by the values of (R2 =0.93), NS (0.90), and PBIAS (19.60 %). The results of calibration, validation, and uncertainty analysis imply that SWAT model can be applied to simulate future changes in stream flow and water balance of the basin due to eventual climate change. The future climate scenarios are generated after downscaling using SDSM model. These scenarios are generated for three different time periods 2020s (2011–2040), 2050s (2041–2070), and 2080s (2071–2099) by adopting multimodel (CGCM3 and HadCM3) and multi-scenarios (A1B, A2, and B2) approach. Increase in mean annual temperature and precipitation is predicted for the future periods under all scenarios of both the models. SWAT predicts increase with varying amount in mean annual stream flow (from a 1.3 to 7.8 % for CGCM3 model and 0.3 to 3.4 % for HadCM3 model), total water yield (from 25.3 to 125.69 % for CGCM3 model and 11.38 to 37.47 % for HadCM3 model) and ET (from 27.19 to 67.70 % for CGCM3 model to 16.99 to 27.42 % for HadCM3 model) under all scenarios of both the models for future periods but the same time decrease in amount of total aquifer recharge is predicted. However, decrease in mean monthly stream flow is predicted under scenarios of CGCM3 model (from January to May and from November to December) and HadCM3 model (from January to April and from October to December), respectively. It is observed that changes projected in different hydrological
components are highly dependent on the direction of the projected changes in precipitation. The results of this study may be helpful to development planners, decision makers, and other stakeholders when planning and implementing appropriate basin-wide water management strategies to adapt to climate change for Sutlej river basin. Acknowledgments Authors acknowledge the financial support in the form of fellowship provided by University Grant Commission (UGC), Government of India, to Mr. Dharmaveer Singh as Research Fellow for carrying out this research. Authors are also thankful to Bhakara Beas Management Board (BBMB), India, for providing the meteorological data used in the present work.
References Abbaspour KC (2011) SWAT-CUP4: SWAT calibration and uncertainty programs—a user manual. Swiss Federal Institute of Aquatic Science and Technology, Eawag Abbaspour KC, Johnson CA, Van Genuchten MT (2004) Estimating uncertain flow and transport parameters using a sequential uncertainty fitting procedure. Vadose Zone J 3:1340–1352 Abbaspour KC, Yang J, Maximov I et al (2007) Modelling hydrology and water quality in the pre-alpine/alpine Thur watershed using SWAT. J Hydrol 333:413–430 Abu-Allaban M, El-Naqa A, Jaber M, Hammouri N (2014) Water scarcity impact of climate change in semi-arid regions: a case study in Mujib basin, Jordan. Arab J Geosci. doi:10.1007/s12517-014-1266-5 Agarwal CS, Garg PK (2002) Textbook on remote sensing: in natural resources monitoring and management. AH Wheeler Publishing Co Ltd, pp 220 Ageta Y, Kadota T (1992) Predictions of changes of glacier mass balance in the Nepal Himalaya and Tibetan Plateau: a case study of air temperature increase for three glaciers. Ann Glaciol 16:89–94 Aherne J, Futter MN, Dillion PJ (2008) The impacts of future climate change and sulphur emission reductions on acidification recovery at Plastic Lake, Ontario. Hydrol Earth Syst Sci 12(2):383–392 Anandhi A, Shrinivas VV, Nanjundiah RS, Kumar DN (2008) Downscaling precipitation to river basin in India for IPCC SRES scenarios using support vector machine. Int J Climatol 28(3):401– 420. doi:10.1002/joc.1529 Arnell W (1999) The effect of climate change on hydrological regimes in Europe: A continental perspective. Glob Environ Chang 9:5–23 Arnell NW, van Vuuren DP, Isaac M (2011) The implications of climate policy for the impacts of climate change on global water resources. Glob Environ Chang 21:592–603 Arnold JG, Fohrer N (2005) SWAT2000: current capabilities and research opportunities in applied watershed modeling. Hydrol Process 19(3): 563–572. doi:10.1002/hyp.5611 Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area hydrologic modeling and assessment part I: Model development. J Am Water Resour Assoc 34(1):73–89. doi:10.1111/j.1752-1688. 1998.tb05961.x Barnett TP, Adam JC, Lettenmaier DP (2005) Potential impacts of a warming climate on water availability in snow-dominated regions. Nature 438:303–309 Bates BC, Kundzewicz ZW, Wu S, Palutikof JP (2008) Climate Change and Water. Technical Paper of the Intergovernmental Panel on Climate Change. http://ipcc.ch/pdf/technical-papers/climatechange-water-en.pdf. Accessed 30 Jul 2010
Arab J Geosci Bekiaris IG, Panagopoulos IN, Mimikou MA (2005) Application of the swat (soil and water assessment tool) model in the Ronnea catchment of Sweden. Global NEST 7(3):252–257 Bhutiyani MR, Kale VS, Pawar NJ (2008) Changing stream flow patterns in the rivers of northwestern Himalaya: Implications of global warming in the 20th century. Curr Sci 95(5):618–624 Bosch DD, Sheridan JM, Batten HL, Arnold JG (2004) Evaluation of the SWAT model on a coastal plain agricultural watershed. Trans ASAE 47(5):1493–1506. doi:10.13031/2013.17629 Cao W, Bowden WB, Davie T, Fenemor A (2006) Multi-variable and multi-site calibration and validation of SWAT in a large mountainous catchment with high spatial variability. Hydrol Process 20:1057– 1073 Chattopadhyay S, Jha MK (2014) Hydrological response due to projected climate variability in Haw River Watershed, North Carolina. Hydrol Sci J. doi:10.1080/02626667.2014.934823 Combalicer EA, Cruz RVO, Lee S, Im S (2010) Assessing climate change impacts on water balance in the Mount Makiling forest, Philippines. J Earth Syst Sci 119(3):265–283 Dibike YB, Coulibaly P (2005) Hydrologic impact of climate change in the Saguenay watershed: comparison of downscaling methods and hydrologic models. J Hydrol 307:145–163 Eckhardt K, Fohrer N, Frede HG (2005) Automatic model calibration. Hydrol Process 19:651–658 Gagnon S, Singh B, Roussselle J, Roy L (2005) An Application of the statistical downscaling model (SDSM) to simulate climatic data for stream flow modelling in Québec. Can Water Resour J 30(4):297– 314 Garg PK (1991) Development of a catchment scale erosion model for semiarid environment and its implementation through remote sensing. Ph.D. Thesis, University of Bristol, UK Githui F, Gitan W, Mutua F, Bauwens W (2009) Climate change impact on SWAT simulated stream flow in western Kenya. Int J Climatol 29(12):1823–1834. doi:10.1002/joc.1828 Goyal MK, Burn DH, Ojha CSP (2012) Statistical downscaling of temperatures under climate change scenarios for Thames river basin, Canada. Int J Glob Warm 4(1):13–30 Gupta HV, Sorooshian S, Yapo PO (1999) Status of automatic calibration for hydrologic models: comparison with multilevel expert calibration. J Hydrol Eng 4(2):135–143 Huang J, Zhang J, Zhang Z, Xu CY, Wang B, Yao J (2011) Estimation of future precipitation change in the Yangtze River basin by using statistical downscaling method. Stoch Env Res Risk A 25(6):781– 792 IPCC (2007) Summary for Policymakers. In: Parry ML, Canziani OF, Palutikof JP, van der Linden PJ and Hanson CE (eds) Climate change 2007: impacts, adaptation and vulnerability. Contribution of working Group II to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, UK, p 7–22 IPCC (2013) Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. In: Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds) Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, p 1535 Jain SK, Goswami A, Saraf AK (2009) Assessment of snowmelt runoff using remote sensing and effect of climate change on runoff. Water Resour Manag 24:1763–1777 Jayakrishnan R, Srinivasan R, Santhi C, Arnold JG (2005) Advances in the application of the SWAT model for water resources management. Hydrol Process 19(3):749–762. doi:10.1002/hyp.5624 Johnson RR (1998) An investigation of curve number applicability to watersheds in excess of 25000 hectares (250 km2). J Environ Hydrol. http://www.hydroweb.com/jeh/jeh1998/johnson.pdf. Accessed 10 Nov 2014
Kalnay E, Kanamitsu M, Kistler R et al (1996) The NCEP/NCAR 40year reanalysis project. Bull Am Meteorol Soc 77:437–471. doi:10. 1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2 Kistler R, Collins W, Saha S et al (2001) The NCEP–NCAR 50–year reanalysis: monthly means CD–ROM and documentation. Bull Am Meteorol Soc 82:247–267. doi:10.1175/1520-0477(2001) 082<0247:TNNYRM>2.3.CO;2 Koutroulis AG, Tsanis IK, Daliakopoulos IN, Jacob D (2013) Impact of climate change on water resources status: a case study for Crete Island, Greece. J Hydrol 479:146–158 Kripalani RH, Oh JH, Kulkarni A, Sabade SS, Chaudhari HS (2007) South Asian summer monsoon precipitation variability: coupled climate model simulations and projections under IPCC AR4. Theor Appl Climatol 90(3):133–159. doi:10.1007/s00704-006-0282-0 Krysanova V, Arnold JG (2008) Advances in ecohydrological modelling with SWAT—a review. Hydrol Sci J 53(5):939–947. doi:10.1623/ hysj.53.5.939 Latkovska I, Apsīte E, Elferts D, Kurpniece L (2012) Forecasted changes in the climate and the river runoff regime in Latvian river basins. Baltica 25(2):143–152 Majone B, Bovolo CI, Bellin A, Blenkinsop S, Fowler HJ (2012) Modeling the impacts of future climate change on water resources for the Ga´llego river basin (Spain). Water Resour Res. doi:10.1029/ 2011WR010985 Mantua NJ, Tohver I, Hamlet A (2010) Climate change impacts on stream flow extremes and summertime stream temperature and their possible consequences for freshwater salmon habitat in Washington State. Clim Chang. doi:10.1007/s10584-010-9845-2 Matonse AH, Pierson DC, Frei A, Zion MS, Anandhi A, Schneiderman E, Wright B (2013) Investigating the impact of climate change on New York City’s primary water supply. Clim Chang 116:437–456 Morán-Tejeda E, Lorenzo-Lacruz J, López-Moreno IJ, Rahman R, Beniston M (2014) Stream flow timing of Mountain Rivers in Spain: recent changes and future projections. J Hydrol 517:1114– 1127 Motovilov YG, Gottschalk L, Engeland K, Rodhe A (1999) Validation of distributed hydrological model against spatialobservations. Agric For Meteorol 98:257–277 Nash JE, Sutcliffe JV (1970) River flow forecasting through conceptual models, part I—a discussion of principles. J Hydrol 10:282–290 Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2005) Soil and Water Assessment Tool, Theoretical Documentation: Version 2005. Temple, TX.USDA Agricultural Research Service and Texas A&M Black land Research Centre. http://swat.tamu.edu/media/ 1292/swat2005theory.pdf. Accessed 15 Feb 2012. Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2011) Soil and Water Assessment Tool, Theoretical Documentation: Version 2009. Texas Water Resources Institute Technical Report No. 406. http://swat. tamu.edu/media/99192/swat2009-theory.pdf. Accessed 30 Sept 2012 Nossent J (2012) Sensitivity and uncertainty analysis in view of the parameter estimation of a SWAT model of the river Kleine Nete, Belgium. PhD thesis. http://www.vub.ac.be/infovoor/onderzoekers/ research/team_theses.php?team_code=HYDR. Accessed 14 Jul 2014 Rostamian R, Jaleh A, Afyuni M et al (2008) Application of a SWAT model for estimating runoff and sediment in two mountainous watersheds in central Iran. Hydrol Sci J 53(5):977–988 SANDRP (2014) Hydro-electric projects in Sutlej River Basin. http:// sandrp.in/basin_maps/Hydro_Electric_Projects_on_Sutlej_River_ in%20HP.pdf. Accessed 20 Oct 2014 Shrestha B, Babel MS, Maskey S, van Griensven A, Uhlenbrook S, Green A, Akkharath I (2013) Impact of climate change on sediment yield in the Mekong River basin: a case study of the Nam Ou basin, Lao PDR. Hydrol Earth Syst Sci 17:1–20. doi:10.5194/hess-17-12013
Arab J Geosci Singh AK, Yadava MG, Ramesh R (2007) High resolution monsoon records from land and the ocean: what have we learnt during the last decade? Jal Vigyan Sameeksha (Hydrol Rev) 22:177–190 Singh V, Bankar N, Salunkhe SS, Bera AK, Sharma JR (2013) Hydrological stream flow modelling on Tungabhadra catchment: parameterization and uncertainty analysis using SWAT CUP. Curr Sci 104(9):1187–1199 Singh D, Gupta RD, Jain SK (2014) Study of long-term trend in river discharge of Sutlej river N-W Himalayan region, India. Geogr Environ Sustain 3(7):87–96 Singh D, Gupta RD, Jain SK (2015a) Statistical analysis of long term spatial and temporal trends of temperature parameters over Sutlej river basin, India. J Earth Syst 124(1):17–35. doi:10.1007/s12040014-0530-0 Singh D, Jain SK, Gupta RD (2015b) Statistical Downscaling and Projection of Future Temperature and Precipitation Change in Middle Catchment of Sutlej River Basin, India. J Earth Syst (In Press) Stehr A, Debels P, Romero F, Alcayaga H (2008) Hydrological modelling with SWAT under conditions of limited data availability: evaluation of results from a Chilean case study. Hydrol Sci J 53(3):588–601. doi:10.1623/hysj.53.3.588 Stewart IT (2009) Changes in snowpack and snowmelt runoff for key mountain regions. Hydrol Proc 23:78–94 Szépszó G, Lingemann I, Klein B, Kovács M (2014) Impact of climate change on hydrological conditions of Rhine and Upper Danube rivers based on the results of regional climate and hydrological models. Nat Hazards 72(1):241–262 Tekle A, Tadele K (2014) Assessment of climate change impact on water availability of Bilate watershed, Ethiopian Rift Valley Basin. J Environ Earth Sci 4(15):148–156 USDA-SCS (United States Department of Agriculture – Soil Conservation Service) 1972 National engineering handbook, Section 4 Hydrology, Chapter 4–10, USDA-SCS, Washington, USA
Van Griensven A, Bauwens W (2003) Multiobjective autocalibration for semidistributed water quality models. Water Resour Res 39(12): 1348 Van Liew MW, Arnold JG, Bosch DD (2005) Problems and potential of autocalibrating a hydrologic model. Trans Am Soc Agric 48(3): 1025–1040 Viviroli D, Archer DR, Buytaert W et al (2011) Climate change and mountain water resources: overview and recommendations for research, management and policy. Hydrol Earth Syst Sci 15:471–504. doi:10.5194/hess-15-471-2011 Wilby RL, Dawson CW (2007) SDSM user manual- a decision support tool for the assessment of regional climate change impacts; Available at https://copublic.lboro.ac.uk/cocwd/SDSM/main.html. Accessed 20 Oct 2012 Wilby RL, Dawson CW (2013) The Statistical Down Scaling Model: insights from one decade of application. Int J Climatol 33:1707– 1719. doi:10.1002/joc.3544 Wilby RL, Dawson CW, Barrow EM (2002) SDSM—a decision support tool for the assessment of regional climate change impacts. Environ Model Softw 17:147–159 Wu F, Zhan J, Wang Z, Zhang Q (2014) Stream flow variation due to glacier melting and climate change in upstream Heihe River Basin, Northwest China. Phys Chem Earth. doi:10.1016/j.pce.2014.08.002 Xu H, Zhou B, Song Y (2011) Impacts of climate change on headstream runoff in the Tarim River Basin. Hydrol Res. doi:10.2166/nh.2010. 069 Yamada T, Fushimi H, Aryal R, Kadota T et al (1996) Report of avalanches accident at Pangka, Khumbu Region, Nepal in 1995. Japan Soc Snow Ice 58(2):145–155 Yang J, Reichert P, Abbaspour KC, Xia J, Yang H (2008) Comparing uncertainty analysis techniques for a SWAT application to the Chaohe Basin in China. J Hydrol 358:1–23