Addressing Climate Change Impacts on Streamflow in the ... - MDPI

water Article

Addressing Climate Change Impacts on Streamflow in the Jinsha River Basin Based on CMIP5 Climate Models Jun Yin 1 , Zhe Yuan 2, *, Denghua Yan 3, *, Zhiyong Yang 3 and Yongqiang Wang 2 1 2 3

*

Faculty of Resources and Environmental Science, Hubei University, Wuhan 430062, China; [email protected] Changjiang River Scientific Research Institute, Changjiang Water Resources Commission of the Ministry of Water Resources of China, Wuhan 430010, China; [email protected] State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research, Beijing 100038, China; [email protected] Correspondence: [email protected] (Z.Y.); [email protected] (D.Y.); Tel.: +86-137-1656-5927 (Z.Y.); +86-135-0103-8825 (D.Y.)





Received: 7 May 2018; Accepted: 6 July 2018; Published: 10 July 2018

Abstract: Projecting future changes of streamflow in the Jinsha River Basin (JRB) is important for the planning and management of the west route of South-to-North Water Transfer Project (SNWTP). This paper presented an analysis of the implications of CMIP5 climate models on the future streamflow in the JRB, using SWAT model. Results show that: (1) In the JRB, a 10% precipitation decrease might result in a streamflow increase of 15 to 18% and a 1 ◦ C increase in temperature might results in a 2 to 5% decrease in streamflow; (2) GFDL-ESM2M and NORESM1-M showed considerable skill in representing the observed precipitation and temperature, which can be chosen to analyze the changes in streamflow in the future; (3) The precipitation and temperature were projected to increase by 0.8 to 5.0% and 1.31 to 1.87 ◦ C. The streamflow was projected to decrease by 4.1 to 14.3% in the upper JRB. It was excepted to change by −4.6 to 8.1% in the middle and lower JRB (MLJRB). The changes of low streamflow in the MLJRB were −5.8 to 7.4%. Therefore, the potential impact of climate on streamflow will have little effect on the planning and management of the west route of SNWTP. Keywords: climate change; streamflow; CMIP5 climate models; Jinsha River Basin

1. Introduction Global surface temperature increased by 0.85 ◦ C in the 20th century and this trend has been even more obvious over the past 30 years [1]. Because of warmer climate, the atmospheric moisture content will be higher than before, which will further affect the earth's hydrological cycle [2,3]. Streamflow is an important factor in environmental, agricultural and economical applications [4]. Therefore, investigating changes about streamflow under future climate conditions is important to the discussion of climate change effects. Predicting the impacts of climate change on streamflow is mainly based on hydrological modeling driven by outputs from general circulation models (GCMs). Githui et al. have assessed the potential future climatic changes on the streamflow in western Kenya using SWAT model and MAGICC and Scenario Generator [5]. The predicted streamflow increase in this region was about 6 to 115%. Chien et al. have found that the annual streamflow was projected to decrease up to 41.1% to 61.3% in agricultural watersheds of the Midwestern United States for the second half of the 21th century [6]. Eisner et al. have detected an increase trend of high-flows and/or a decrease trend of low-flows in 11 representative large river basins (e.g., Rhine and Tagus in Europe, upper Amazon in South America, upper Mississippi in North America) for the twenty-first century [7]. Yuan et al. Water 2018, 10, 910; doi:10.3390/w10070910

www.mdpi.com/journal/water

Water 2018, 10, 910

2 of 19

have projected the streamflow in typical regions of China by a water balance model and multiple climate change scenarios [8]. The results showed that streamflow was expected to decrease by 12 to 13% in the northern part of China, while those in the southern part would be likely to decrease by 7 to 10%. However, the performance of GCMs remain variable because that the realization and structure of GCMs are different [9,10]. Some studies analyze the climate change impact based on multiple GCM-scenario combinations without GCMs selection. Thus, uncertainty may exist among the projected results. Filtering out the GCMs that under perform in comparison to the others will help to predict changes in future climate and their impacts reasonably. There is a wide range of metrics for evaluating the GCMs performance, such as means and standard deviations, comparisons of seasonal or diurnal fluctuations, skill in reproducing probability density functions (PDFs) of specific climate variables [11]. The study by Perkins et al. has demonstrated that the PDF-based assessment can test the ability of climate models to simulate entire distributions of climate variables (e.g., precipitation, temperature) [12]. It can offer a tougher test than only comparisons of the mean and variance. The Jinsha River Basin (JRB) is located in the upstream Yangtze River. It is an important water source region in the west route of South-to-North Water Transfer Project (SNWTP). The climate change may have potential impact on streamflow in the JRB. It is needed to discuss whether this impact will have effect on the planning and management of the west route of SNWTP. Thus, this research did a case study in the JRB. Relatively optimal GCMs in this study area were selected with PDF-based assessment, and future changes of streamflow in the context of climate change were projected. The sections of this paper are organized as follows: the study area, data sets and methodologies are introduced in Section 2; Section 3 shows the performance of hydrological model, sensitivity of streamflow to climate change, comparison of GCM simulations with observations, changes in precipitation, temperature and streamflow for 2020 to 2050; The conclusions are summarized in Section 4. 2. Materials and Methods 2.1. Study Area The Jinsha River Basin (JRB, 90◦ 300 –105◦ 150 E and 24◦ 360 –35◦ 440 N) is watershed of the upper Yibin city, covering an area of 473,200 km2 shared by four provinces and autonomous region (namely Qinghai, Tibet, Yunnan, Sichuan and Guizhou), is about 26% of the total drainage area of the Yangtze River Basin [8,13]. The Jinsha River is about 3464 km in length and originates from the peak of east Geladan Snowy Mountain in the Tanggula Mountains, flowing through Western Sichuan Plateau, Hengduan Mountains and Yunnan-Guizhou Plateau to the mountain area of Southwest Sichuan [14]. The Yalong River is the longest tributary of the JRB, with 1187 km in length (Figure 1). The mean annual temperature is 10.2 ◦ C in the JRB. It increases progressively from northwest to southeast [13]. The general distribution of precipitation in JRB increases gradually from upstream to downstream. The annual average precipitation in the upper, middle and lower reaches is 350 mm, 600 mm and 1000 mm respectively [15]. The wet season lasts from June to October when 75 to 85% of total yearly rainfall happens. The average annual runoff is about 152 billion m3 . The JRB acts as an important water source in the west route of South-to-North Water Transfer Project. A total of 1.5 billion m3 of water will be transferred from the upper reaches of the Yalong River to the northern part of China in the first phase every year. The JRB also contributes to irrigation, water supply, flood control, wood drift, and tourism. Overall, the JRB plays a very important role in regional and national economic development.

10, 910 Water 2018, 10, xWater FOR2018, PEER REVIEW

3 of 19

3 of 19

Figure 1. Locationof of Jinsha River Basin. Figure 1. Location Jinsha River Basin.

2.2. A Brief Review of ArcSWAT Model

The JRB acts as an important water source in the west route of South-to-North Water Transfer The ArcGIS10.2.2 interface of SWAT (ver.2012) was used in this study for water cycle simulation. Project. A total of 1.5 billion m3 of water will be transferred from the upper reaches of the Yalong It is a watershed-scale, process-based, continuous-time, and semi-distributed hydrological model, River to the northern of for China in the phase everychange year.on The also contributes which can part be used predicting thefirst impacts of climate theJRB water hydrology of river to irrigation, [16]. The waterwood balancedrift, equation which governs Overall, the hydrological components water supply,basins flood control, and tourism. the JRB playsofa SWAT verymodel important role in is given as: regional and national economic development.  t  SWt i = SW0 + ∑ Rdayi − Qsur f i − Eai − Wseepi − Q gwi (1) i =1 2.2. A Brief Review of ArcSWAT Model where, SW is soil water content at time t (mm H O), SW is initial soil water content (mm H O), t is 2

ti

0

2

simulation period (days), Rdayi is the amount of precipitation on the i-th day (mm H2 O), Qsurfi is the The ArcGIS10.2.2 interface of SWAT (ver.2012) was used in this study for water cycle simulation. amount of surface runoff on the i-th day (mm H2 O), Eai is the amount of evapotranspiration on the i-th It is a watershed-scale, andzone semi-distributed model, day (mm H2 O),process-based, Wseepi is the amountcontinuous-time, of water entering the vadose from the soil profile hydrological on the i-th (mm H2predicting O) and Qgwi is the the amount of baseflow on thechange i-th day (mm H2 O). which can be day used for impacts of climate on the water hydrology of river basins The JRB is comprised of 181 sub-basins which were divided into 579 Hydrological Response [16]. The water balance equation which governs the hydrological components of SWAT model is Units (HRUs) that satisfactorily represent watershed’s heterogeneity. In defining HRUs, 20% and 10%

given as:

t

(

SWt i = SW0 +  Rdayi − Qsurfi − Eai − Wseepi − Qgwi i =1

)

(1)

Water 2018, 10, 910

4 of 19

threshold values of land use and soil area have been considered to ignore minor land uses and soil Water 2018, 10, x FOR PEER REVIEW types in each sub basin so as to avoid unnecessary large number of HRUs (Figure 2).

4 of 19

Figure 2. The catchments divided byby SWAT. Figure 2. The catchments divided SWAT. 2

The coefficient of determination thelinear linear regression equation (R ) and Nash-Sutcliffe The coefficient of determination ofofthe regression equation (R2) the and the Nash-Sutcliffe efficiency coefficient (ENS ) are chosen to assess the model’s feasibility in the study area [17]. efficiency coefficient (ENS) are chosen to assess the model’s feasibility in the study area [17]. The The calculation formulas are as follows: calculation formulas are as follows:   2

n

∑ Rsim-i − Rsim n




Robs-i − Robs




 R = ni=1
2 n R
2 R − R − R sim i sim obs i obs  ∑ Robs-i − Robs ∑ Rsim-i − Rsim    2 ii= i =1 R = n =11 2 n  Robs -i − R∑nobs(Rsim-i− RRobs-isim)-2i − Rsim 2

(

(

i =1 ENS

= 1−

i =1 n

)(

) ( i =1

− Robs ∑ R ni =1 obs-i

)

2

(2)

)

(2) 2

(3)


2

R −R ) is(simulated depth of runoff(mm); R sim -i

2

obs -i

Robs-i is observed depth of runoff (mm); Rsim-i obs and Rsim are i =1 = 1 − the mean value of observed andEsimulated depth of runoff respectively(mm); n is the number of NS n 2 2 observed value. Models perform better with higher R and greater ENS . When ENs ≥ 0.75, it means Robs -i − Robs the simulation is good; when 0.36 < ENS < 0.75, the simulation is basically good; when ENS ≤ 0.36, i =1 the simulation is poor [18].

(

)

(3)

Robs-i is observed depth of runoff (mm); Rsim-i is simulated depth of runoff(mm); Robs and Rsim are the mean value of observed and simulated depth of runoff respectively(mm); n is the number of observed value. Models perform better with higher R2 and greater ENS. When ENs ≥ 0.75, it means the simulation is good; when 0.36 < ENS < 0.75, the simulation is basically good; when ENS ≤ 0.36, the simulation is poor [18].

Water 2018, 10, 910

5 of 19

2.3. Data 2.3.1. Observed Hydro-Meteorological Data In this study, as the SWAT was used in water balance mode, the meteorological parameters which were considered to force the model were daily precipitation and the daily minimum and maximum Water 2018, 10, x FOR PEER REVIEW 5 of 19 temperature. The daily meteorological data during 1956 to 2013 from 196 meteorological stations in and around the JRB were collected for this study. The observations used to evaluate the AR5 Climate Models’ simulated monthly precipitation The observations used to evaluate the AR5 Climate Models’ simulated monthly precipitation and temperature are gridded monthly temperature and precipitation data prepared by the National and temperature are gridded monthly temperature and precipitation data prepared by the National Meteorological Information Center (NMIC) of the China Meteorological Administration (CMA) [19]. Meteorological Information Center (NMIC) of the China Meteorological Administration (CMA) [19]. In In this study, 208 describethe theareas areasofofthe theJRB JRB(Figure (Figure this study, 208grid gridboxes boxesare are selected selected to to describe 3).3).

Figure 3. 3. Grid Grid boxes in Figure in the theJRB. JRB.

Streamflow collectedfrom from four hydrological stations in different ofduring the JRB Streamflowdata data was was collected four hydrological stations in different parts ofparts the JRB during the period from 1957 It towas 2012. It to was used the to runoff analyze the and runoff trend and sensitivity the period from 1957 to 2012. used analyze trend sensitivity of streamflow to of streamflow to climate calibrate validate the SWAT. Theof basic information of the climate change, and tochange, calibrateand andtovalidate theand SWAT. The basic information the four hydrological four hydrological is listed in Table 1. stations is listed stations in Table 1. Table1.1.Basic Basicinformation information of Table of the the 44 hydrological hydrologicalstations. stations.

Hydrological Hydrological Station Station Zhimenda Zhimenda Shigu Shigu Panzhihua Panzhihua Huatan Huatan

River River

Lon. (E° (E◦) Lon.

Lat. (N° (N◦ ) Lat.

Tongtian River Tongtian River Jinsha River River Jinsha Jinsha River Jinsha River Jinsha River Jinsha River

97.24 97.24 99.96 99.96 101.72 101.72 102.88 102.88

33.01 33.01 26.87 26.87 26.58 26.58 26.88 26.88

Catchment Area Catchment Area 33 km22) (10 (10 km ) 159.8 159.8 234.1 234.1 280.9 280.9 447.7 447.7

Area Percent Area Percent (%) (%) 32.4 32.4 47.4 47.4 56.9 56.9 90.7 90.7

Data Period Data Period 1957–2012 1957–2012 1956–2010 1956–2010 1966–2000 1966–2000 1977–2000 1977–2000

2.3.2. Future Climate Scenarios The Inter-Sectoral Impact Model Intercomparison Project (ISI-MIP) [20] provides continuous daily-series meteorological data derived from GFDL-ESM2M, HADGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM and NORESM1-M (Table 2), including historical scenario and future emission

Water 2018, 10, 910

6 of 19

2.3.2. Future Climate Scenarios The Inter-Sectoral Impact Model Intercomparison Project (ISI-MIP) [20] provides continuous daily-series meteorological data derived from GFDL-ESM2M, HADGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM and NORESM1-M (Table 2), including historical scenario and future emission scenarios RCP 2.6, RCP4.5, RCP6.0 and RCP8.5. The data was bias-corrected and resampled by the ISI-MIP. It covers the period from 1960 to 2099 on a horizontal grid with 0.5◦ × 0.5◦ resolution [21]. The statistical characteristic of simulated data and observed data are consistent with each other. The spatial resolution of data can meet the requirements of hydrological simulation in the JRB. In this study, we used the three mitigation emissions scenarios (RCP2.6, RCP4.5, RCP8.5) and the period 2011–2050. Water 2018, 10, x FOR PEER REVIEW

6 of 19

Table 2. List of general circulation models (GCMs) used in this study.

Table 2. List of general circulation models (GCMs) used in this study. Centre

Country

Name

Centre Country Name GeophysicalFluid FluidDynamics DynamicsLaboratory Laboratory (GFDL) (GFDL) United States GFDL-ESM2M Geophysical United States GFDL-ESM2M Hadley Centrefor forClimate Climate Prediction Prediction and Met Office Kingdom HADGEM2-ES Hadley Centre andResearch, Research, Met Office United United Kingdom HADGEM2-ES L’Institut France IPSL-CM5A-LR L’InstitutPierre-Simon Pierre-SimonLaplace Laplace (IPSL) (IPSL) France IPSL-CM5A-LR Technology, ResearchInstitute, Institute, and Technology,Atmosphere Atmosphere and and Ocean Ocean Research Japan MIROC-ESM-CHEM Japan MIROC-ESM-CHEM National Institute for Environmental Studies and National Institute for Environmental Studies Norwegian Climate Norway NORESM1-M Norwegian ClimateCentre Centre Norway NORESM1-M

2.3.3. Digital Elevation Model, Land Use/Land Cover Map and Soil Map 2.3.3. Digital Elevation Model, Land Use/Land Cover Map and Soil Map

The spatial resolution resolutionofof9090mmwas wasobtained obtained from TheDigital DigitalElevation ElevationModel Model (DEM) (DEM) with with aa spatial from thethe CGIAR-CSI GeoPortal [22]. Land use/land cover map for 1980 and 2000 was obtained from the Data CGIAR-CSI GeoPortal [22]. Land use/land cover map for 1980 and 2000 was obtained from the Data Center for ChineseAcademy AcademyofofSciences Sciences (Figure and land Center forResources Resourcesand andEnvironmental Environmental Sciences, Sciences, Chinese (Figure 4),4), and land use properties were directly from the SWAT model database. Soil map of JRB at a scale of 1:1,000,000 use properties were directly from the SWAT model database. Soil map of JRB at a scale of 1:1,000,000 was procured ChineseAcademy AcademyofofSciences Sciences [23] (Figure was procuredfrom fromInstitute Institute of of Soil Soil Science, Science, Chinese [23] (Figure 5). 5).

Figure of JRB JRB in in1980 1980(a) (a)and and2000 2000(b). (b). Figure4. 4. The The Landuse Landuse of

Water 2018, 10, 910 Water 2018, 10, x FOR PEER REVIEW

7 of 19

7 of 19

Figure5.5.Soil Soildistribution distribution in in JRB. JRB. Figure

2.4.2.4. Sensitivity Analysis of of Runoff totoClimate Sensitivity Analysis Runoff ClimateChange Change In this study, definedthe thesensitivity sensitivityof of streamflow streamflow to climate In this study, wewedefined climate change changeas asthe theproportional proportional change of simulated streamflowbybycomparing comparingtheir their values values with hypothetical change of simulated streamflow hypotheticalclimatic climaticscenarios scenariostoto their values with observed climatic data [24]. We consider the proportional change in precipitation their values with observed climatic data [24]. We consider the proportional change in precipitation mean temperature the factorscontained containedby byclimate climate change. change. Then Then the andand mean temperature asas the factors the sensitivity sensitivityofofstreamflow streamflow to climate change could be defined as to climate change could be defined as ∆T f (f (PP ++∆P, P, TT + + T))−− ff( P, ( PT,)T×) 100% 100% f ( P, T ) f ( P, T )

(∆P,  ( δP , T∆T ) =) =

(4)

(4)

where, P and T are precipitation and temperature; ∆P and ∆T are the hypothetical changes of

where, P and Tand are precipitation and temperature; ΔP and ΔT are the hypothetical precipitation temperature; f (P,T) is the relation function between streamflow and climaticchanges scenario;of precipitation temperature; f(P,T) is the relationchange. function between and climatic δ(∆P,∆T) isand the sensitivity of streamflow to climate Based on thestreamflow mean precipitation and scenario; δ(ΔP,ΔT) is the sensitivity of streamflow to climate change. Based on the mean precipitation temperature in the period 1961–1990 (baseline), precipitation is assumed to change by +30%, +20%, ◦ C, and+10%, temperature in −the period (baseline), precipitation is assumed to change by+2+30%, 0%, −10%, 20% and −1961–1990 30% in the future; temperature is assumed to change by +3 ◦ C, ◦ C, −2 ◦ C and +20%, 0%, −20% and−−30% thefuture future; +1 ◦+10%, C, 0 ◦ C, −1−10%, 3 ◦ C ininthe [8].temperature is assumed to change by +3 °C, +2 °C, +1 °C, 0 °C, −1 °C, −2 °C and −3 °C in the future [8]. 2.5. Selection of Global Climate Model

2.5. Selection of Global Climatethe Model This study evaluated coupled climate models and selected the relatively optimal GCMs for hydro-meteorological variability projection. Themodels evaluation focused 4 regionsoptimal of the JRB for the This study evaluated the coupled climate and was selected theon relatively GCMs for daily simulation of precipitation and temperature. It was based on probability density functions (PDFs). hydro-meteorological variability projection. The evaluation was focused on 4 regions of the JRB for We defined an alternative metric to describe the similarity between two PDFs. This metric calculates the daily simulation of precipitation and temperature. It was based on probability density functions the cumulative minimum value of two distributions of each binned value, thereby measuring the (PDFs). We defined an alternative metric to describe the similarity between two PDFs. This metric common area between two PDFs. The skill score (SS) can range between 0 and 1. SS is close to 1 when calculates the cumulative minimum value of two distributions of each binned value, thereby the model simulates the observed conditions perfectly (Figure 6a) and it is close to 0 when the overlap measuring the common area between two PDFs. The skill score (SS) can range between 0 and 1. SS is is negligible (Figure 6b) [12].

close to 1 when the model simulates the observed conditions perfectly (Figure 6a) and it is close to 0 when the overlap is negligible (Figure 6b) [12].

Water Water2018, 2018,10, 10,910 x FOR PEER REVIEW

observed

0.450

88 of 19 19

(a)

GCM

(b)

GCM

0.400

0.350

0.350

0.300

Probability

Probability

observed

0.450

0.400

Skill score

0.250

0.200

0.150

0.300

0.250

0.200

0.150

0.100

0.100

0.050

0.050

Skill score

0.000

0.000 5

5

15

15

Variable

Variable

Figure 6. 6. Diagrams Diagrams of of GCM GCM vs vs observed observed PDF PDF illustrating illustrating the the total total skill skill score score in in (a) (a) aa near-perfect near-perfect skill skill Figure scoreand and(b) (b)aavery verypoor poorskill skillscore. score. score

Expressed formally, Expressed formally, nn

SS = = ∑ min n) ) SS , Fo  min((FsFsn ,n Fo n

(5) (5)

1

1

where n is the number of bins, Fsn is the frequency of values in a given bin from the model, and Fon is where n is the number of bins, Fsn is the frequency of values in a given bin from the model, and Fon the frequency of values in a given bin from the observed data. SS was the summation of the minimum is the frequency of values in a given bin from the observed data. SS was the summation of the frequency values over all bins. minimum frequency values over all bins. In this study, the monthly precipitation was fitted by a mixture function which was shown in In this study, the monthly precipitation was fitted by a mixture function which was shown in Equation (6). Equation (6). G ( x ) = (1 − p) H ( x ) + pF ( x ) (6)

G( x) = (1 − p) H ( x) + pF ( x) (6) where, x represents a rainfall amount. G(x) is the mixture function. p is the probability of no rainfall. H(x) is step function, awhen x > amount. 0, H(x) = G(x) 1, else = 0. F(x)function. is gammapcumulative probability where, x represents rainfall isH(x) the mixture is the probability of nofunction, rainfall. the is shown in Equation H(x)PDF is step function, when x > (7). 0, H(x) = 1, else H(x) = 0. F(x) is gamma cumulative probability function, ( x/β)α−1 e − x/β the PDF is shown in Equation (7). (7) f (x) = βΓ(α)  −1

( )

−x



x e β is called the scale parameter. Γ(α) is where, f (x) is the PDF. α is called the shape parameter  and (7) gamma function. f ( x) =  ( function. ) The monthly average temperature was fitted by beta The PDF was shown in Equation (8). where, f(x) is the PDF. α is called the shape the scale parameter. Γ(α) is 1 parameter andp−β1 is calledq− 1 f ( x ) = ( x − a ) · ( b − x ) (8) gamma function. p + q +1 B( p, q)(b − a) The monthly average temperature was fitted by beta function. The PDF was shown in Equation where, x represents monthly average temperature. f (x) is the PDF. a, b, p and q are the parameters for (8). beta function(B(p,q)). 1 1 Driven by the baselinef daily projected ( x) =climatology andpthe ( x − a) p −future  (b −daily x) q −1 climate data sets coming (8) + q +1 B( pthe , q)( b − a) SWAT model can output the simulated daily from the relatively optimal GCMs, calibrated streamflow in both baseline (1961 to 1990) and future period (2020 to 2050). Then the changes of where, x represents monthly average temperature. f(x) is the PDF. a, b, p and q are the parameters for future streamflow can be estimated. beta function(B(p,q)). Driven byDiscussion the baseline daily climatology and the projected future daily climate data sets coming 3. Results and from the relatively optimal GCMs, the calibrated SWAT model can output the simulated daily 3.1. Calibration and Verification streamflow in both baseline (1961 to 1990) and future period (2020 to 2050). Then the changes of future streamflow can be estimated. Table 3 summarized the performance of the SWAT in the four catchments. The E and R2 were NS

higher than 0.8 in all catchments. Figure 7 showed the simulated and observed monthly streamflow in the calibration period and verification period. The results denoted good performance of the SWAT, although there were some differences between observation and simulation. Overall, the calibration and verification accuracy of the SWAT was acceptable for monthly streamflow simulation in the JRB.

3.1. Calibration and Verification Table 3 summarized the performance of the SWAT in the four catchments. The ENS and R2 were higher than 0.8 in all catchments. Figure 7 showed the simulated and observed monthly streamflow in the calibration period and verification period. The results denoted good performance of the SWAT, Water 2018, 10, 910 9 of 19 although there were some differences between observation and simulation. Overall, the calibration and verification accuracy of the SWAT was acceptable for monthly streamflow simulation in the JRB. Table 3. Performance of the SWAT for monthly streamflow simulation in the Zhimenda, Shigu, Panzhihua and Huatan Table 3. Performance ofcatchment. the SWAT for monthly streamflow simulation in the Zhimenda, Shigu, Panzhihua and Huatan catchment.

Catchment Zhimenda Shigu

Panzhihua

Panzhihua

Huatan

Huatan

Calibration period Calibration Verificationperiod period

Year

ENS

1991–2012

0.88

1991–2010

0.80

1991–2000

0.88

1991–2000

0.93

1959–1990 0.80 0.84 1959–1990 0.88 0.80 1991–2012 0.91

Verification period Calibration period Calibration Verificationperiod period Verification period Calibration period Calibration Verificationperiod period Verification period Calibration period Calibration Verificationperiod period Verification period

Precipitation

R2

ENS

Year

Period

Zhimenda Shigu

Period

1959–1990 0.82 0.91 1959–1990 0.80 0.82 1991–2010 0.93 1966–1990 0.85 0.92 1966–1990 0.88 0.85 1991–2000 0.93 1977–1990 0.90 0.95 1977–1990 0.93 0.90 1991–2000 0.96

Observation

R2 0.84 0.91 0.91 0.93 0.92 0.93 0.95 0.96

Simulation

Streamflow(m³/s)

4000

0

3000

100 Verification period

Calibration period

2000

200 1000

0

(a) Precipitation(mm)

Catchment

300 1959

1963

1967

1971

1975

1979

1983

1987

1991

1995

1999

2003

2007

2011

Year Observation

Simulation

Streamflow(m³/s)

0

7500

100 Verification period

Calibration period

5000

200

2500

300

Water 2018,0 10, x FOR PEER REVIEW 1959

1963

1967

1971

1975

1979

1983

1987

1991

1995

1999

2003

2007

(b) Precipitation(mm)

Precipitation 10000

400 10 of 19

Year Observation

Simulation

Streamflow(m³/s)

15000

0

100

10000 Verification period

Calibration period

200

5000

300

0

(c) Precipitation(mm)

Precipitation

400 1966

1970

1974

1978

1982

1986

1990

1994

1998

Year Observation

Figure 7. Cont.

Simulation 0

15000

10000 5000

100 Verification period

Calibration period

200

o data

Streamflow(m³/s)

20000

300

(d) Precipitation(mm)

Precipitation 25000

100 Verification period

Calibration period

Water 2018, 10, 910 1966

1970

1974

1978

1982

5000

1990

200

1994

1998

Year Precipitation

300

Observation

400

10 of 19

Simulation

0

25000

400 1966

1970

1974

1978

1982

1986

1990

1994

0

(d)

1998

Observation

Simulation 0

15000

No data

Streamflow(m³/s)

100

15000

5000

5000

1977

1980

1983

1986

0 1977

1980

1983

200

No data

0

Verification period

Calibration period

10000

100

200

Verification period

Calibration period 20000

10000

(d)

1986

300

1989

1992

1995

300

Precipitation(mm)

Precipitation 25000

Precipitation(mm)

Year

20000

Streamflow(m³/s)

1986

300

Precipita

0

10000

0

200

(c)

Simulation

Precipitation(mm)

Observation

15000

5000

Streamflow(m³/s)

Streamfl

Precipitation

400

1998 400

Year 1989

1992

1995

1998

Year

Figure andobserved observed monthly discharge during the calibration period and the Figure7.7. Simulated Simulated and monthly discharge during the calibration period and the verification Figure 7. Simulated and observed monthly discharge during the calibration period and the verification period: (a) Zhimenda catchment; (b) Shigu catchment; (c) Panzhihua catchment; (d) period: (a) Zhimenda catchment; (b) Shigu catchment; (c) Panzhihua catchment; (d) Huatan catchment. verification period: (a) Zhimenda catchment; (b) Shigu catchment; (c) Panzhihua catchment; (d) Huatan catchment. Huatan catchment. 3.2. Sensitivity of Streamflow to Climate Change 3.2. Sensitivity of Streamflow to Climate Change 3.2. Sensitivity of Streamflow to Climate Change

With the output of the SWAT using hypothetical climate scenarios, the sensitivity of runoff With the output of the SWAT using hypothetical climate scenarios, the sensitivity of runoff to to With climate was estimated (5).Then Then response of streamflow to changes of to thechange output ofwas the SWATby using hypothetical climate scenarios, the sensitivity of climate change estimated by Equation Equation (5). thethe response of streamflow to changes of runoff precipitation and cancan bebe analyzed Figures 8 and 9).general, In general, a rise in of climate change wastemperature estimated by Equation (5).(shown Then the response of to in changes precipitation and temperature analyzed (shown ininFigures 8 and 9). streamflow In a rise temperature would be a negative impact in streamflow due to higher evaporation and an increase in temperature would be a negative impact in streamflow due to higher evaporation and an increase in in precipitation and temperature can be analyzed (shown in Figures 8 and 9). In general, a rise precipitation would certainly lead to higher streamflow. precipitation would leadimpact to higher streamflow. due to higher evaporation and an increase in temperature would becertainly a negative in streamflow

precipitation would certainly lead to higher streamflow. 80

80

(a)

80

(b)

80

40

40

ΔQ(%) ΔQ(%)

ΔQ(%)

40

ΔQ(%)

(a)

0 -2°C -1°C 0°C +1°C +2°C

-40

0 -80

Water 2018, 10, x FOR PEER REVIEW

-20

-40

-10

0 ΔP(%)

10 -2°C

(c)

-80

0

0 ΔP(%)

10

-20

-10

0 ΔP(%)

10

0 ΔP(%)

11 of 19

10

20 -2°C -1°C 0°C +1°C +2°C

(d)

-80 40 -20

-2°C -1°C 0°C +1°C +2°C

-80

-10

80

20

-40

-20

0 -40

ΔQ(%)

ΔQ(%)

-10

-2°C -1°C 0°C +1°C +2°C

-80 20

40

-20

0 40

-40

-1°C 0°C +1°C +2°C

80

(b)

-10

0

0 ΔP(%)

20

-2°C -1°C 0°C +1°C +2°C

-40 -80 -20

20

10

-10

0 ΔP(%)

10

20

Figure 8. Sensitivity on mean streamflow due to mean precipitation change in JRB: Zhimenda (a);

Figure 8. Sensitivity on mean streamflow due to mean precipitation change in JRB: Zhimenda (a); Shigu (b); Panzhihua (c); and Huatan (d). Shigu (b); Panzhihua (c); and Huatan (d). 80

80

ΔQ(%)

ΔQ(%)

Figure 8 showed(a) the relation between ∆P (the changes(b) in precipitation) and ∆Q (the changes in streamflow). It can be seen that a 10% precipitation increase might result in a streamflow increase of 15 40 40 to 18%. The responses of streamflow to temperature change were illustrated in Figure 9. From the figure 0 -40 -80

+20% +10% 0% -10% -20%

0 -40 -80

+20% +10% 0% -10% -20%

40 ΔQ(%)

ΔQ(%)

40

0

Water 2018, 10, 910

-2°C -1°C 0°C +1°C +2°C temperature

-40

0 -2°C -1°C 0°C +1°C +2°C streamflow.

-40

11 of 19

◦ we can find that will result Among -80 in a 2 to 5% decrease in -80a 1 C increase in -20 scenario -10 will0 increase 10 by 30%, 20 while the -20 -10 0the precipitation 10 20 in wettest 26 hypothetical climate scenarios, ΔP(%) ΔP(%) temperature will decrease by 3 ◦ C. In addition, the precipitation in driest situation will decrease by Figure 8. Sensitivity onwill mean streamflow to Under mean precipitation change in JRB:streamflow Zhimenda (a);would be 30%, while the temperature increase by due 3 ◦ C. the wettest scenario, Shigu (b); (c); while and Huatan (d). change by −35 to −33% under the driest scenario. likely to change byPanzhihua 39 to 49%, it might

80

80

(b)

(a) 40 ΔQ(%)

ΔQ(%)

40

0 -40

-40

+20% +10% 0% -10% -20%

-80 -2

-1

0 ΔT(°C)

0

-80

1

-2

2

-1

0 ΔT(°C)

1

2

1

2

80

80

(d)

(c) 40 ΔQ(%)

40 ΔQ(%)

+20% +10% 0% -10% -20%

0 -40

-40

+20% +10% 0% -10% -20%

-80 -2

-1

0 ΔT(°C)

0 +20% +10% 0% -10% -20%

-80 1

2

-2

-1

0 ΔT(°C)

Figure 9. Sensitivity mean streamflowdue due to to mean mean temperature in JRB: Zhimenda (a); (a); Figure 9. Sensitivity on on mean streamflow temperaturechange change in JRB: Zhimenda Shigu (b); Panzhihua (c); and Huatan (d). Shigu (b); Panzhihua (c); and Huatan (d).

Figure 8 showed the relation between ΔP (the changes in precipitation) and ΔQ (the changes in

According Chiew, with higher runoff will be less sensitive streamflow).toItthe canstudy be seenofthat a 10%basins precipitation increase mightcoefficients result in a streamflow increase of to climatic changes [25]. If the runoff coefficient is more than 0.2 (high runoff coefficient), 15 to 18%. The responses of streamflow to temperature change were illustrated in Figure 9.10% Fromdecrease the figure we can findresult that a 1in°Caincrease in temperature will result in a 2 to 5% decrease in streamflow. in precipitation may 10 to 20% decrease in streamflow. While the same 10% decrease Among 26 hypothetical climate scenarios, the precipitation in wettest scenario will increase 30%, in precipitation may lead to a 40% decrease in streamflow in a basin with a low runoffby coefficient. 3 while the temperature will decrease by 3 °C. In addition, the precipitation in driest situation In Jinsha River Basin, the annual average runoff is 143 billion m (319.4mm) and the annual will average decrease by 30%, while the temperature will increase by 3 °C. Under the wettest scenario, streamflow precipitation is 592 mm. Therefore, the runoff coefficient in the JRB is about 0.54. It means that this basin should be less sensitive to climatic changes. Our results on the JRB are in accordance with this rule. 3.3. Assessment of Skill Score from GCMs’ Outputs and Changes in Precipitation and Temperature for 2020 to 2050 with the Output of the SWAT Using Hypothetical Climate Scenarios, the Sensitivity of Runoff to Climate Change Figures 10 and 11 showed the ensemble monthly skill score for each model averaged over all 4 regions defined in Figure 1. Because the outputs from five GCMs in this study have been bias-corrected by ISI-MIP, the SSs for precipitation and temperature simulation were not low. The SSs averaged over the JRB were more than 0.46. However, there still exists some difference among GCMs or variables. The averaged SSs over the JRB and all 12 months for precipitation were between 0.63 and 0.68. However, these averaged SSs for temperature were between 0.50 to 0.63 and the SSs for 3 of the 5 models were less than 0.6. These models were HadGEM2-ES, IPSL-CM5A-LR and MIROC-ESM-CHEM. Based on

corrected by ISI-MIP, the SSs for precipitation and temperature simulation were not low. The SSs averaged over the JRB were more than 0.46. However, there still exists some difference among GCMs or variables. The averaged SSs over the JRB and all 12 months for precipitation were between 0.63 and 0.68. However, these averaged SSs for temperature were between 0.50 to 0.63 and the SSs for 3 Water 2018, 10, 910 of 19 of the 5 models were less than 0.6. These models were HadGEM2-ES, IPSL-CM5A-LR and 12 MIROCESM-CHEM. Based on the above analysis, we considered GFDL-ESM2M and NORESM1-M as the relatively optimal GCMs in simulating the precipitation and temperature in theoptimal JRB. The outputs the above analysis, we considered GFDL-ESM2M and NORESM1-M as the relatively GCMs in came from these two GCMs were chosen forin the futurecame climate change andGCMs its effects simulating the precipitation and temperature theanalysis JRB. Theofoutputs from these two were on streamflow. chosen for the analysis of future climate change and its effects on streamflow. (a) D

(b) D

(c) D

(d) D

(e) D

N

N

N

N

N

O

O

O

O

O

S

S

S

S

S

A

A

A

A

A

J

J

J

J

J

J

J

J

J

J

M

M

M

M

M

A

A

A

A

A

M

M

M

M

M

F

F

F

F

F

J

J

J

J

J

0

0.3 0.6 0.9 SS for precipitation

0

0.3 0.6 0.9 SS for precipitation

0

JRB

0.3 0.6 0.9 SS for precipitation

YLRB

LJRB

0

MJRB

0.3 0.6 0.9 SS for precipitation

0

0.3 0.6 0.9 SS for precipitation

UJRB

Figure 10. score Skill for score for precipitation simulation: GFDL-ESM2M (a); HADGEM2-ES (b); Figure 10. Skill precipitation simulation: GFDL-ESM2M (a); HADGEM2-ES (b); IPSL-CM5AIPSL-CM5A-LR (c); MIROC-ESM-CHEM (d); NORESM1-M (e). Water LR 2018, 10,MIROC-ESM-CHEM x FOR PEER REVIEW (c); (d); NORESM1-M (e). (a)

(b)

13 of 19

(c)

(d)

(e)

D

D

D

D

D

N

N

N

N

N

O

O

O

O

O

S

S

S

S

S

A

A

A

A

A

J

J

J

J

J

J

J

J

J

J

M

M

M

M

M

A

A

A

A

A

M

M

M

M

M

F

F

F

F

F

J

J 0

0.3 0.6 0.9 SS for temperature

J 0

0.3 0.6 0.9 SS for temperature

JRB

J 0

0.3 0.6 0.9 SS for temperature

YLRB

LJRB

MJRB

J 0

0.3 0.6 0.9 SS for temperature

0

0.3 0.6 0.9 SS for temperature

UJRB

Figure Skill for score for temperature simulation: GFDL-ESM2M (a); HADGEM2-ES (b); Figure 11. 11. Skill score temperature simulation: GFDL-ESM2M (a); HADGEM2-ES (b); IPSL-CM5AIPSL-CM5A-LR (c); MIROC-ESM-CHEM (d); NORESM1-M (e). LR (c); MIROC-ESM-CHEM (d); NORESM1-M (e).

The precipitationand andtemperature temperatureare are the the two main model. The changes The precipitation main factors factorsdriving drivingSWAT SWAT model. The changes in precipitation and temperaturewill willdirectly directlyaffect affect the the change change in in precipitation and temperature in streamflow streamflowsimulated simulatedbybySWAT. SWAT. Thus, we would firstly analyze changes in precipitation and temperature for the baseline (1961 to to Thus, we would firstly analyze changes in precipitation and temperature for the baseline (1961 1990) theprojection projection period 2050). To eliminate the impacts of systematic deviationdeviation in GCMs in 1990) totothe period(2020 (2020to to 2050). To eliminate the impacts of systematic on precipitation and temperature projection in the future, relative change between simulated value GCMs on precipitation and temperature projection in the future, relative change between simulated in projection period and that in baseline was chosen to reveal the evolution trend. To reduce the value in projection period and that in baseline was chosen to reveal the evolution trend. To reduce from GCMs, we only used the relatively optimal GCMs (GFDL-ESM2M and NORESM1-M) theuncertainty uncertainty from GCMs, we only used the relatively optimal GCMs (GFDL-ESM2M and to analysis the changes in precipitation and temperature in the future period. NORESM1-M) to analysis the changes in precipitation and temperature in the future period. Figures 12 and 13 showed the relative changes in mean annual precipitation simulated by the relatively optimal GCMs between RCP2.5, RCP4.5 and RCP8.5 when compared with that of the baseline. The spatial pattern was different in GFDL-ESM2M and NORESM1-M. GFDL-ESM2M suggested that the precipitation would increase in the upper and middle Jinsha River Basin (UMJRB), but decrease in the lower basin (LJRB). Future precipitation changes in the UJRB, MJRB, YLRB and

ptimal GCMs between RCP2.5, RCP4.5 and RCP8.5 when compared with e spatial pattern was different in GFDL-ESM2M and NORESM1-M. GF hat the precipitation Water 2018, 10, 910 would increase in the upper and middle Jinsha 13 of 19 River Bas e in the lower basin (LJRB). Future precipitation changes in the UJRB, MJRB Figures 12 and 13 showed the relative changes in mean annual precipitation simulated by the projected relatively to beoptimal 4.3 GCMs to 6.7%, 1.1 RCP4.5 to 5.7%, 0.6 4.2% −6.9 to −1.5%, r between RCP2.5, and RCP8.5 whento compared withand that of the baseline. The spatial pattern was different in GFDL-ESM2M and NORESM1-M. GFDL-ESM2M suggested that he spatial characteristics of changes inmiddle annual precipitation by N the precipitation would increase in the upper and Jinsha River Basin (UMJRB), butpredicted decrease in the lower basin (LJRB). Future precipitation changes in the UJRB, MJRB, YLRB and LJRB were nt from GFDL-ESM2M. Large increases in precipitation covered the LJRB thr projected to be 4.3 to 6.7%, 1.1 to 5.7%, 0.6 to 4.2% and −6.9 to −1.5%, respectively. However, spatial characteristics ofwas changeslikely in annual precipitation predictedby by NORESM1-M were different eriod. Thetheprecipitation to increase 2.8 to 8.1% in this area. from GFDL-ESM2M. Large increases in precipitation covered the LJRB throughout the projection The precipitation was likely increase0.8 by 2.8to to 8.1% in this area. However, it would change ge slightlyperiod. in the whole JRB bytoonly 5.0%. slightly in the whole JRB by only 0.8 to 5.0%.

Water 2018, 10, x FOR PEER REVIEWFigure 12. Changes in mean precipitation for 2021 to 2050.

14 of 19

Changes in precipitation (%)

Figure 12. Changes in mean precipitation for 2021 to 2050. 15

(a)

UJRB

MJRB

LJRB

YLRB

JRB

10 5

0 -5 -10 GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M

RCP2.6

RCP4.5

RCP8.5

Figure 13. Changes in mean precipitation for 2021 to 2050 in different regions.

Figure 13. Changes in mean precipitation for 2021 to 2050 in different regions.

Models have a variable performance in simulating precipitation and temperature. Limited by the complexity of climate models and existing methods, only one or several indices (e.g., skill score) can be used to evaluate the simulation effect. So the SS in UJRB is higher than the SS in the other

Changes in precipitation (

UJRB MJRB in LJRBsimulating YLRB JRB (a) Models have a variable performance precipitation and temperature. Limite 10 omplexity of climate models and existing methods, only one or several indices (e.g., skill s 5 be used to evaluate the 0simulation effect. So the SS in UJRB is higher than the SS in the o 2018, to 10, 910 14 of 19 a certain cli -5 ons, but it isWater hard explain from the aspects of climate model structure why el has better simulation-10ofGFDL-ESM2M precipitation or temperature in some certain places. Further rese NorESM1-M GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M Models have a variable performance in simulating precipitation and temperature. Limited by the RCP2.6 RCP4.5 RCP8.5 be done to improve this issue. complexity of climate models and existing methods, only one or several indices (e.g., skill score) can be used15 to evaluate the simulation effect. So the SS in UJRB isin higher than themean SS in thetemperature other regions, Figures 14 and illustrated thein absolute change annual in the fu Figure 13. Changes mean precipitation for 2021 to 2050 in different regions. but it is hard to explain from the aspects of climate model structure why a certain climate model has od. The spatial almostor temperature the samein in GCM-scenario bu betterpattern simulation was of precipitation somemultiple certain places. Further research will combinations, be done Models have a variable performance in simulating precipitation and temperature. Limited by to improve this issue. nitude varied considerably. Extreme warming was more widespread in NORESM1-M th the complexity of climate models and existing methods, only one or several indices (e.g., skill score) Figures 14 and 15 illustrated the absolute change in annual mean temperature in the future period. L-ESM2M, under RCP4.5. warming wasthan projected in other the UMJRB can particularly be topattern evaluate simulation Solarger the SS in UJRBcombinations, is higher in the Theused spatial wasthe almost the sameeffect. inThe multiple GCM-scenario butthe theSS magnitude but it is hard explain from the aspects of climateinto model structure a certain varied considerably. Extreme warming was more widespread NORESM1-M than in GFDL-ESM2M, ual mean regions, temperature overto the UMJRB was predicted increase bywhy more thanclimate 1.77 °C in RC model has betterunder simulation precipitation or temperature in some certain places. Furthermean research particularly RCP4.5.of The larger warming was projected in the UMJRB. The annual ario. In the the would increase by 1.31 to 1.38 °C for RCP2.6, 1. willfuture be done period, to improve this temperature issue. temperature over the UMJRB was predicted to increase by more than 1.77 ◦ C in RCP8.5 scenario. In the future period, theto temperature increase by change 1.31 to 1.38 C for RCP2.6, to 1.75 ◦ C for Figures 14 and 151.87 illustrated theRCP8.5. absolute in ◦annual mean1.38 temperature inRCP4.5 the future °C for RCP4.5 and 1.72 °Cwould for

◦

and The 1.72 to 1.87 C for RCP8.5. period. spatial pattern was almost the same in multiple GCM-scenario combinations, but the magnitude varied considerably. Extreme warming was more widespread in NORESM1-M than in GFDL-ESM2M, particularly under RCP4.5. The larger warming was projected in the UMJRB. The annual mean temperature over the UMJRB was predicted to increase by more than 1.77 °C in RCP8.5 scenario. In the future period, the temperature would increase by 1.31 to 1.38 °C for RCP2.6, 1.38 to 1.75 °C for RCP4.5 and 1.72 to 1.87 °C for RCP8.5.

Figure 14.14.Changes temperaturefor for2021 2021 2050. Figure Changesin inmean mean temperature to to 2050.

2.0 1.5 1.0

Changes in temperature (°C)

Changes in temperature (°C)

Figure 14. Changes in mean temperature for 2021 to 2050. (b)

2.0

(b)

UJRB

MJRB

LJRB

YLRB

JRB

1.5

UJRB

MJRB

LJRB

YLRB

JRB

1.0

0.5 0.0 GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M

RCP2.6

0.5

RCP4.5

RCP8.5

Figure Changesinintemperature temperature for regions. Figure 15.15.Changes for 2021 2021toto2050 2050inindifferent different regions.

0.0 GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M GFDL-ESM2M NorESM1-M

RCP2.6

RCP4.5

RCP8.5

Figure 15. Changes in temperature for 2021 to 2050 in different regions.

Water 2018, 10, x FOR PEER REVIEW

15 of 19

3.4. Changes in Streamflow for 2020 to 2050 Water 2018, 10, 910

15 of 19

Using GFDL-ESM2M and NORESM1-M-projected daily precipitation and temperature under the three RCP scenarios, the SWAT model was run on the JRB from 1961 to 2050. Because it is hard Changes in Streamflow for 2020 to 2050 to project3.4. future human activities (e.g., land use), the anthropogenic effects were not taken into GFDL-ESM2M and NORESM1-M-projected dailyAll precipitation and of temperature under considerationUsing in projection of future streamflow changes. parameters the SWAT model were the three RCP scenarios, the SWAT model was run on the JRB from 1961 to 2050. Because it is hard kept with no change. With the outputs from the SWAT model, the changes in streamflow for period to project future human activities (e.g., land use), the anthropogenic effects were not taken into 2020 to 2050 were statistically Results showed that different combinations consideration in projectionanalyzed. of future streamflow changes. All parameters of GCM-scenario the SWAT model were were associated different changes in streamflow, resulting in in opposite conclusions. Taking kept withwith no change. With the outputs from the SWAT model, the changes streamflow for period 2020 to 2050 were statistically analyzed. Results projected showed that4.2% different GCM-scenario combinations Huatan Station as an example, GFDL-ESM2M streamflow decrease for the future were associated with different changes in streamflow, resulting in opposite conclusions. Taking Huatan period under RCP4.5, while NORESM1-M estimated 6.0% increase in streamflow for the same period Station as an example, GFDL-ESM2M projected 4.2% streamflow decrease for the future period under and climate scenario. However, estimated projected trends from the different GCMs were same under RCP4.5, while NORESM1-M 6.0% increase in streamflow for the same period and the climate RCP2.6 and RCP8.5. The projected streamflow Zhimenda to under decrease byand 4.1RCP8.5. to 5.2% under scenario. However, trends in from the differentwas GCMsprojected were the same RCP2.6 The6.6 streamflow Zhimenda was projected to decrease in by 4.1 5.2% under RCP2.6 and 6.6 to 8.1% RCP2.6 and to 8.1%inunder RCP8.5. The streamflow thetoMLJRB (Shigu, Panzhihua and Huatan) under RCP8.5. The streamflow in the MLJRB (Shigu, Panzhihua and Huatan) was likely to increase by was likely to increase by 2.4 to 8.1% under RCP2.6, but was excepted to increase by 0.3 to 4.5% in the 2.4 to 8.1% under RCP2.6, but was excepted to increase by 0.3 to 4.5% in the same region under RCP8.5 same region under (Figure 16). RCP8.5 (Figure 16). Changes in sreamflow (%)

15 Zhimenda

Shigu

Panzhihua

Huatan

10 5 0 -5 -10 -15 GFDL-ESM2M

NorESM1-M

RCP2.6

GFDL-ESM2M

NorESM1-M

RCP4.5

GFDL-ESM2M

NorESM1-M

RCP8.5

Figure 16. Changes in mean streamflow for2021 2021 to 2050 in different Figure 16. Changes in mean streamflow for to 2050 in different stations.stations. To compare the frequency of annual mean streamflow streamflow between baseline and future P-III To compare the frequency of annual mean between baseline andperiod, future period, P-III distributions were illustrated comparison (Figure (Figure 17). thethe frequency is lessis less than frequencyfrequency distributions were illustrated forfor comparison 17).When When frequency than 10%, RCP2.6 and RCP4.5 scenarios suggested a rather similar increasing trend in the MLJRB 10%, RCP2.6 and RCP4.5 scenarios suggested a rather similar increasing trend in the MLJRB (Shigu, (Shigu, Panzhihua and Huatan). The opposite trend was found in the same region under 8.5 scenario. PanzhihuaToand Huatan). The wasthree found in the same region under 8.5and scenario. To further investigate the opposite changes in trend streamflow, different frequencies (P = 90%, P = 50% P = 10%) were (Table As shown inthree Table 4, the low streamflow (P =(P 90%, Q90 ), median further investigate thechosen changes in4). streamflow, different frequencies = 90%, P = 50% and P = streamflow (P = 50%, Q ) and higher streamflow (P = 10%, Q ) in Zhimenda were projected 10%) were chosen (Table 4).50As shown in Table 4, the low10 streamflow (P = 90%, Qto90), median decrease in all the climate change scenarios. The largest percentage decreases in these three kinds streamflow (P = 50%, Q50) and higher streamflow (P = 10%, Q10) in Zhimenda were projected to of streamflow were mainly found for NORESM1-M. In Shigu station, nearly all the climate change decrease in all theshowed climatethat change The likely largest percentage decreases in these large three kinds of scenarios the Q90scenarios. and Q50 would to decrease, but there was a relatively uncertainty in the projection Q10NORESM1-M. . One of the six scenarios projected that the Q10 would increase by streamflow were mainly found of for In Shigu station, nearly all the climate change 5%. Two of the scenarios projected that the Q would decrease by 5%. The rest scenarios projected scenarios showed that the Q90 and Q50 would10 likely to decrease, but there was a relatively large no significant change. The Q90 will increase slightly in Panzhihua and Huatan under RCP2.6 while uncertainty in the projection of Q10. One of the six scenarios projected that the Q10 would increase by may decrease under RCP4.5 and RCP8.5. The Q10 will increase in the same stations under RCP2.6. 5%. Two of the scenarios projected that the Q 10 would decrease by 5%. The rest scenarios projected However, the changes in Q10 under RCP4.5 and RCP8.5 were not obvious. no significant change. The Q90 will increase slightly in Panzhihua and Huatan under RCP2.6 while may decrease under RCP4.5 and RCP8.5. The Q10 will increase in the same stations under RCP2.6. However, the changes in Q10 under RCP4.5 and RCP8.5 were not obvious.

Water 2018, 10, 910

16 of 19

Water 2018, 10, x FOR PEER REVIEW

16 of 19

Figure Figure 17. PIII distribions ofofannual streamflow Zhimenda 17. frequency PIII frequency distribions annual mean mean streamflow in in Zhimenda (a,b),(a,b), ShiguShigu (c,d), (c,d), Panzhihua (e,f) Huatan and Huatan (g,h) underdifferent different climate scenarios. Panzhihua (e,f) and (g,h) under climatechange change scenarios. Table 4. Percentage changes (%) in streamflow under different scenarios. Frequency RCP2.6 P = 90%

RCP4.5 RCP8.5 RCP2.6

P = 50%

RCP4.5

Scenario GFDL-ESM2M NORESM1-M GFDL-ESM2M NORESM1-M GFDL-ESM2M NORESM1-M GFDL-ESM2M NORESM1-M GFDL-ESM2M NORESM1-M

Zhimenda −3.0 −17.4 −10.8 −23.5 −10.6 −16.0 −5.8 −14.1 −9.8 −23.2

Shigu −1.0 −6.0 −7.0 −11.2 −5.4 −13.4 2.4 −3.3 −2.3 −8.6

Panzhihua 1.6 2.5 −6.0 −3.1 −2.1 −6.8 3.8 2.2 −3.3 −0.8

Huatan 0.0 6.6 −8.1 2.1 −4.3 −4.2 3.2 4.8 −4.7 4.0

Water 2018, 10, 910

17 of 19

Table 4. Percentage changes (%) in streamflow under different scenarios. Frequency

P = 90%

P = 50%

P = 10%

Scenario

Zhimenda

Shigu

Panzhihua

Huatan

RCP2.6

GFDL-ESM2M NORESM1-M

−3.0 −17.4

−1.0 −6.0

1.6 2.5

0.0 6.6

RCP4.5

GFDL-ESM2M NORESM1-M

−10.8 −23.5

−7.0 −11.2

−6.0 −3.1

−8.1 2.1

RCP8.5

GFDL-ESM2M NORESM1-M

−10.6 −16.0

−5.4 −13.4

−2.1 −6.8

−4.3 −4.2

RCP2.6

GFDL-ESM2M NORESM1-M

−5.8 −14.1

2.4 −3.3

3.8 2.2

3.2 4.8

RCP4.5

GFDL-ESM2M NORESM1-M

−9.8 −23.2

−2.3 −8.6

−3.3 −0.8

−4.7 4.0

RCP8.5

GFDL-ESM2M NORESM1-M

−8.6 −18.2

−2.7 −10.5

−2.2 −4.3

−4.1 −2.1

RCP2.6

GFDL-ESM2M NORESM1-M

−6.2 −11.9

5.7 −0.5

7.4 4.1

6.7 7.3

RCP4.5

GFDL-ESM2M NORESM1-M

−6.1 −23.1

1.2 −5.6

0.7 1.4

−0.8 5.7

RCP8.5

GFDL-ESM2M NORESM1-M

−5.7 −14.5

−0.6 −5.8

−1.4 −1.7

−4.8 −0.4

4. Discussion and Conclusions Based on five CMIP5 CMs, this study discussed the changes in projected streamflow of the JRB under RCP2.6, RCP4.5 and RCP8.5 scenarios. The SWAT model was employed to simulate monthly streamflow, using observed meteorological data from NMIC and the ISI-MIP climate dataset. Observed monthly streamflow at four stations in the JRB was used for calibration and validation of the SWAT model. The model performed well in simulating monthly streamflow in the JRB, with ENS and R2 higher than 0.8 in the four catchments. This demonstrated that calibrated model can be used to estimate the impact of climate change on future streamflow in the JRB. The study of Chiew revealed that the sensitivity of a river basin to changes in precipitation and temperature was strongly related to the runoff coefficient [24]. The basin with higher runoff coefficient is less sensitivity to climate change. The JRB is in this category. In this study, area, a 10% precipitation decrease might result in a streamflow increase of 15 to 18% and a 1 ◦ C increase in temperature might results in a 2 to 5% decrease in streamflow. The skill of the selected five GCMs was assessed using a skill score (SS) based on the overlap between the observed and simulated PDFs (grid by grid) in the baseline (1961 to 1990). In general, GFDL-ESM2M and NORESM1-M showed considerable skill in representing the observed PDFs of monthly precipitation and temperature. The averaged SSs over the JRB and all 12 months for precipitation and temperature were more than 0.6. Therefore, GFDL-ESM2M and NORESM1-M can be chosen as the relatively optimal GCMs to analysis the changes in precipitation and temperature and their effects on streamflow in the future period. In the future period, the precipitation was projected to increase slightly in the whole JRB by only 0.8 to 5.0%. While the temperature was likely to increase significantly. It would increase by 1.31 to 1.38 ◦ C for RCP2.6, 1.38 to 1.75 ◦ C for RCP4.5 and 1.72 to 1.87 ◦ C for RCP8.5. The projected changes in mean annual streamflow were generally similar under RCP2.6 and RCP8.5. The streamflow was projected to decrease in the UJRB (Zhimenda) while it was excepted to increase in the MLJRB (Shigu, Panzhihua and Huatan). However, all the changes (increase or decrease) were less than 10%. The projections of future mean annual streamflow carried high uncertainty under RCP4.5. The different GCMs were associated with different changes in mean annual streamflow, resulting in

Water 2018, 10, 910

18 of 19

opposite conclusions. When considering the results from three future scenarios of the two relatively optimal GCMs, the range of relative changes in the UJRB and the MLJRB was 4.1 to 14.3% and −4.6 to 8.1%, respectively. The low streamflow, median streamflow and higher streamflow were projected to change more obvious in the Zhimenda and Shigu than other stations. These three kinds of streamflow in Zhimenda were projected to decrease in all the climate change scenarios. In Shigu station, nearly all the climate change scenarios showed that the low streamflow and median streamflow would likely to decrease, but there was a relatively large uncertainty in the projection of higher streamflow. In the MLJRB, the changes of low streamflow for 2020 to 2050 were between −5.8 to 7.4%. Therefore, the potential impact of climate on streamflow will have little effect on the planning and management of the west route of SNWTP. This research chose SWAT model to do hydrological simulation. Hydrologic model is likely to cause less error (or uncertainty) than the climate projections. However, compared with physics-based hydrological models, transferring the calibration of model parameters under historical conditions to futures with different climatic characteristics raises questions of validity [26,27]. In future research, different kinds of hydrological models coupled with more GCMs will be added to do uncertainty analysis. Author Contributions: J.Y., Z.Y. (Zhe Yuan) and Z.Y. (Zhiyong Yang) worked together in forming ideas of this paper; J.Y. and Z.Y. (Zhe Yuan) worked together in calculating and writing of this manuscript; Y.W. carried out the calculation and analysis of runoff in the study area; D.Y. and Z.Y. (Zhiyong Yang) provided supervision during the whole process. Funding: This research was funded by [National Key Research and Development Project] grant number [2017YFC1502404]; [the Open Research Fund of State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin(China Institute of Water Resources and Hydropower Research)] grant number [IWHR-SKL-KF201804]; [National Natural Science Foundation of China] grant number [51779013, 51709008]; [National Public Research Institutes for Basic R&D Operating Expenses Special Project] grant number [CKSF2017029]. Conflicts of Interest: The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

References 1. 2. 3.

4. 5. 6. 7.

8. 9. 10.

IPCC. Climate Change 2013: The Physical Science Basis; Cambridge University Press: Cambridge, UK, 2013. Arnell, N.W. Climate change and global water resources. Glob. Environ. Chang. 1999, 9, 31–49. [CrossRef] Huntington, J.; Mcgwire, K.; Morton, C.; Snyder, K.; Peterson, S.; Erickson, T.; Niswonger, R.; Carroll, R.; Smith, G.; Allen, R. Assessing the role of climate and resource management on groundwater dependent ecosystem changes in arid environments with the Landsat archive. Remote Sens. Environ. 2016, 185, 186–197. [CrossRef] Thodsen, H. The influence of climate change on stream flow in Danish rivers. J. Hydrol. 2007, 333, 226–238. [CrossRef] Githui, F.; Gitau, W.; Mutua, F.; Bauwens, W. Climate change impact on SWAT simulated streamflow in western Kenya. Int. J. Climatol. 2009, 29, 1823–1834. [CrossRef] Chien, H.; Yeh, P.J.F.; Knouft, J.H. Modeling the potential impacts of climate change on streamflow in agricultural watersheds of the Midwestern United States. J. Hydrol. 2013, 491, 73–88. [CrossRef] Eisner, S.; Flörke, M.; Chamorro, A.; Daggupati, P.; Donnelly, C.; Huang, J.; Hundecha, Y.; Koch, H.; Kalugin, A.; Krylenko, I.; et al. An ensemble analysis of climate change impacts on streamflow seasonality across 11 large river basins. Clim. Chang. 2017, 141, 401–417. [CrossRef] Yuan, Z.; Yan, D.; Yang, Z.; Yin, J.; Zhang, C.; Yuan, Y. Projection of surface water resources in the context of climate change in typical regions of China. Hydrol. Sci. J. 2017, 62, 283–293. [CrossRef] Bollasina, M.; Nigam, S. Indian Ocean SST, evaporation, and precipitation during the South Asian summer monsoon in IPCC-AR4 coupled simulations. Clim. Dyn. 2009, 33, 1017–1032. [CrossRef] Anandhi, A.; Nanjundiah, R.S. Performance evaluation of AR4 Climate Models in simulating daily precipitation over the Indian region using skill scores. Theor. Appl. Climatol. 2015, 119, 551–566. [CrossRef]

Water 2018, 10, 910

11. 12.

13. 14.

15. 16. 17. 18. 19. 20. 21.

22. 23. 24. 25. 26.

27.

19 of 19

Yin, J.; Yan, D.; Yang, Z.; Yuan, Z.; Yuan, Y.; Zhang, C. Projection of extreme precipitation in the context of climate change in Huang-Huai-Hai region, China. J. Earth Syst. Sci. 2016, 125, 417–429. [CrossRef] Perkins, S.E.; Pitman, A.J.; Holbrook, N.J.; McAneney, J. Evaluation of the AR4 climate models’ simulated daily maximum temperature, minimum temperature, and precipitation over Australia using probability density functions. J. Clim. 2007, 20, 4356–4376. [CrossRef] Wang, H. The Future Situation of Water Resources in China and its Management Requirements. World Environ. 2011, 2, 16–17. Liu, H.; Lan, H.; Liu, Y.; Zhou, Y. Characteristics of spatial distribution of debris flow and the effect of their sediment yield in main downstream of Jinsha River, China. Environ. Earth Sci. 2011, 64, 1653–1666. [CrossRef] Chen, Y.; Wang, W.; Wang, G.; Wang, S. Characteristics analysis on temperature and precipitation variation in Jinsha River Basin. Plateau Mt. Meteorol. Res. 2010, 30, 51–56. Uniyal, B.; Jha, M.K.; Verma, A.K. Assessing climate change impact on water balance components of a river basin using SWAT model. Water Resour. Manag. 2015, 29, 4767–4785. [CrossRef] Nash, J.E.; Sutclife, J.V. River low forecasting through conceptual models—Part I: A discussion of principles. J. Hydrol. 1970, 10, 282–290. [CrossRef] Motovilov, Y.G.; Gottschalk, L.; Engeland, K.; Rodhe, A. Validation of a distributed hydrological model against spatial observations. Agric. For. Meteorol. 1999, 98, 257–277. [CrossRef] NMIC (National Meteorological Information Center). Assessment Report of China’s Ground Precipitation 0.5◦ × 0.5◦ Gridded Dataset (V2.0); National Meteorological Information Center of China: Beijing, China, 2012. The Inter-Sectoral Impact Model Intercomparison Project. Available online: https://isimip.org/ (accessed on 5 June 2018). Warszawski, L.; Frieler, K.; Huber, V.; Piontek, F.; Serdeczny, O.; Schewe, J. The Inter-Sectoral Impact Model Intercomparison Project (ISI–MIP): Project framework. Proc. Natl. Acad. Sci. USA 2014, 111, 3228–3232. [CrossRef] [PubMed] The CGIAR Consortium for Spatial Information. Available online: http://srtm.csi.cgiar.org/ (accessed on 5 June 2018). China Soil Information Service Platform. Available online: http://www.soil.csdb.cn (accessed on 5 June 2018). Bao, Z.; Zhang, J.; Liu, J.; Wang, G.; Yan, X.; Wang, X.; Zhang, L. Sensitivity of hydrological variables to climate change in the Haihe River basin, China. Hydrol. Process. 2012, 26, 2294–2306. [CrossRef] Chiew, F.H.S. Estimation of rainfall elasticity of streamflow in Australia. Hydrol. Sci. J. 2006, 51, 613–625. [CrossRef] Coron, L.; Andréassian, V.; Perrin, C.; Lerat, J.; Vaze, J.; Bourqui, M.; Hendrickx, F. Crash testing hydrological models in contrasted climate conditions: An experiment on 216 Australian catchments. Water Resour. Res. 2012, 48, 213–223. [CrossRef] Guerreiro, S.B.; Birkinshaw, S.; Kilsby, C.; Fowler, H.J.; Lewis, E. Dry getting drier—The future of transnational river basins in Iberia. J. Hydrol. Reg. Stud. 2017, 12, 238–252. [CrossRef] © 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).