Downscaling GCM data for climate change impact assessments on rainfall

42 Pages·2016·3.92 MB·English

Hydrol.EarthSyst.Sci.Discuss.,doi:10.5194/hess-2015-499,2016 D
ManuscriptunderreviewforjournalHydrol.EarthSyst.Sci. isc HESSD
u
Published:15January2016 s
s
©Author(s)2016.CC-BY3.0License. io doi:10.5194/hess-2015-499
n
P
a
p
Thisdiscussionpaperis/hasbeenunderreviewforthejournalHydrologyandEarthSystem e Downscaling GCM
r
Sciences(HESS).PleaserefertothecorrespondingfinalpaperinHESSifavailable. data for climate
|
change impact
Downscaling GCM data for climate D assessments on
is
c
u rainfall
change impact assessments on rainfall: ss
io
n R.J.Dahmetal.
a practical application for the P
a
p
e
r
Brahmani-Baitarani river basin
TitlePage
|
Abstract Introduction
D
1 2 2 1 1
R. J. Dahm , U. K. Singh , M. Lal , M. Marchand , F. C. Sperna Weiland , is
S. K. Singh3, and M. P. Singh3 cu Conclusions References
s
s
io Tables Figures
1Deltares,Boussinesqweg1,P.O.Box177,2600MHDelft,theNetherlands n
2RMSI,A-8Sector16,NOIDA20130,India Pa
3CentralWaterCommission,R.K.Puram,NewDelhi110066,India pe J I
r
J I
Received:16November2015–Accepted:10December2015–Published:15January2016 |
Correspondenceto:R.J.Dahm([email protected]) D Back Close
is
c
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion. u FullScreen/Esc
s
s
io
n
Printer-friendlyVersion
P
a
p
e InteractiveDiscussion
r
1 |
Abstract D
is
c HESSD
u
The delta of the Brahmani-Baitarani river basin, located in the eastern part of India, s
s
frequentlyexperiencesseverefloods.Forfloodriskanalysisandwatersystemdesign, io doi:10.5194/hess-2015-499
n
insights in the possible future changes in extreme rainfall events caused by climate P
a
5 change are of major importance. There is a wide range of statistical and dynamical pe Downscaling GCM
r
downscaling and bias-correction methods available to generate local climate projec- data for climate
tions that also consider changes in rainfall extremes. Yet the applicability of these | change impact
methods highly depends on availability of meteorological observations at local level. D assessments on
is
In the developing countries data and model availability may be limited, either due to cu rainfall
s
10 the lack of actual existence of these data or because political data sensitivity hampers sio
open sharing. n R.J.Dahmetal.
P
We here present the climate change analysis we performed for the Brahmani- a
p
Baitarani river basin focusing on changes in four selected indices for rainfall extremes e
r
usingdatafromthreeperformance-basedselectedGCMsthatarepartofthe5thCou- TitlePage
|
pled Model Intercomparison Project (CMIP5). We apply and compare two widely used
15 Abstract Introduction
D
and easy to implement bias correction approaches. These methods were selected as is
bestsuitedduetotheabsenceofreliablelonghistoricmeteorologicaldata.Wepresent cu Conclusions References
s
the main changes – likely increases in monsoon rainfall especially in the Mountainous s
io Tables Figures
regions and a likely increase of the number of heavy rain days. In addition, we discuss n
P
thegapbetweenstate-of-the-artdownscalingtechniquesandtheactualoptionsoneis a
20 p J I
e
faced with in local scale climate change assessments. r
J I
|
1 Introduction D Back Close
is
c
u FullScreen/Esc
In the past extreme rainfall over India has resulted in landslides, flash floods, severe s
s
river floods, and crop damage that have had major impacts on society, the economy, ion
Printer-friendlyVersion
andtheenvironment(Goswamietal.,2006).Increasesinrainfallextremeshavequan- P
25 a
p
tifiableimpactsonintensity–duration–frequencyrelations(KaoandGanguly,2011)and e InteractiveDiscussion
r
2 |
are expected to enhance coastal and river flood risks and will, without adaptive mea- D
is
sures, substantially increase flood damages. Climate adaptation strategies for emer- c HESSD
u
s
gencyplanning,thedesignofengineeringstructures,reservoirmanagement,pollution s
io doi:10.5194/hess-2015-499
control, risk calculations rely on knowledge of the frequency of these extreme rainfall n
P
5 events (Guhathakurta et al., 2011; Goswami et al., 2006). a
p
Several studies already investigated the trends in rainfall extremes over India e Downscaling GCM
r
(Parthasarathy et al., 1993; Dash et al., 2009; Ramesh Kumar et al., 2009; Chaturvedi data for climate
|
et al., 2012 and many others). Goswami et al. (2006) found significant positive trends change impact
D
in the frequency and the magnitude of heavy rain events and a significant negative assessments on
is
c
10 trend in the frequency of moderate events over central India during the monsoon sea- u rainfall
s
sons from 1951 to 2000. Climate change is expected to further alter the intensity and s
io
frequency of extreme rainfalls (Chatuverdi et al., 2012; Frei et al., 2006; Kharin et al., n R.J.Dahmetal.
P
2007).Globallytheintensityofextremerainfallisprojectedtoincreaseeveninregions a
p
e
where mean rainfall decreases (Semenov and Bengtsson, 2002). Moreover, future cli- r
TitlePage
matestudiesbasedonclimatemodelsimulationssuggestthatgreenhousewarmingis
15 |
likely to intensify the monsoon rainfall over a broad region encompassing South Asia Abstract Introduction
D
(Laletal., 2000;May,2011;MeehlandArblaster, 2003;Rupakumaretal.,2006;Tren- is
c Conclusions References
berth,2011).However,preciseassessmentsoffuturechangesintheregionalmonsoon u
s
s
rainfallhaveremainedambiguousduetowidevariationsamongthemodelprojections io Tables Figures
n
(Kripalanietal.,2007;Sabadeetal.,2011;TurnerandAnnamalai,2012).Andthesim-
20 P
a
ulated rainfall response to global warming by climate models is actually accompanied p J I
e
by a weakening of the large-scale South-West (SW) monsoon flow (Kripalani et al., r
J I
2003; Krishnan et al., 2013; Ueda et al., 2006). Nonetheless, the recently released |
CMIP5 projections confirm significant reduction in return times of annual extremes of D Back Close
daily rainfall for the late 21st century in India (Ramesh and Goswami, 2014) and in- is
25 c
u FullScreen/Esc
creasesinseasonalmeanrainfallamounts(Menonetal.,2013).Inaddition,Chaturvedi s
s
etal.(2012)foundasteadyincreaseinthenumberofdayswithextremerainfallforthe io
n
Printer-friendlyVersion
period of 2060 and beyond. P
a
p
e InteractiveDiscussion
r
3 |
Within this paper we try to assess future climate induced changes in several rainfall D
is
indices of relevance to water resources and flood risk management in the Brahmani- c HESSD
u
s
Baitarani river basin. The basin is located in the eastern part of India. It experienced s
io doi:10.5194/hess-2015-499
severefloodsinrecentyearsof2001,2003,2006,2008,2011and2013(Government n
P
5 of Odisha, 2011; Government of Odisha, 2013). a
p
To enable a quantitative climate change analysis we use Global Climate Model e Downscaling GCM
r
(GCM) data. GCM rainfall data normally have biases from observations and needs data for climate
|
to be corrected to ensure its applicability at the local scale (Teutschbein and Seibert, change impact
D
2012; Sperna Weiland et al., 2012; Christensen et al., 2008). There is a wide range assessments on
is
c
10 ofstatisticalanddynamicaldownscalingandbias-correctionmethodsavailabletogen- u rainfall
s
erate local climate projections that also consider changes in rainfall extremes. Yet the s
io
applicability of these methods highly depends on availability of meteorological obser- n R.J.Dahmetal.
P
vations at local scale. In this study, data from a limited number of rain gauges with in- a
p
e
completetime-seriesandalowerresolutiongriddedrainfallproduct–theAPHRODITE r
TitlePage
dataset (Yatagai et al., 2012) – could be accessed and used. Due to this data qual-
15 |
ity issue we concentrate on two widely used and easy to implement bias correction Abstract Introduction
D
procedures – Linear Scaling (LS) and Delta Change (DC). We evaluate the influence is
c Conclusions References
of these procedures on resulting changes in rainfall indices. Instead of using a large u
s
s
ensemble of GCMs we focus on three models that nearly span the full range of an- io Tables Figures
n
nual mean rainfall projections available for India – analysing in more detail the locally
20 P
a
relevant biases within these three GCMs and the changes they project. p J I
e
This paper is organized as follows. Section 2 presents an overview of the selected r
J I
GCMs,referencedataset,andbiascorrectionmethods.Thisisfollowedbyashortde- |
scription of the impact analysis framework applied and the study area. In Sect. 3 we D Back Close
reportthemainfindingsofourstudyandpresenttherangeofrainfallindicesundercli- is
25 c
u FullScreen/Esc
mate change. Here, we are specifically concerned with an inter- and intra-comparison s
s
of the GCMs and downscaling methods. Next the limitations experienced for bias- io
n
Printer-friendlyVersion
correction techniques experienced in this study are discussed. Section 5 summarizes P
a
p
e InteractiveDiscussion
r
4 |
our main conclusions and provide outlook for future work and practical applications in D
is
the study area. c HESSD
u
s
s
io doi:10.5194/hess-2015-499
n
2 Methods P
a
p
e Downscaling GCM
The method adopted in this study for the analysis of climate change impact on four r
data for climate
5 rainfall indices is based on four steps described in subsequent sub-sections. |
change impact
D
1. GCMselectionbycomparingobserved(gaugeandgriddedtimeseries)andsim- assessments on
is
c
ulated (GCM) variability of monthly rainfall. u rainfall
s
s
2. A description of the selected reference rainfall dataset. ion R.J.Dahmetal.
P
a
3. BiascorrectionofdailyrainfalltimeseriesfromtheselectedGCMsforthebaseline p
e
and future projections using Linear Scaling and Delta Change method. r
10 TitlePage
|
4. Rainfallindicesanalysisforthebaselineandfutureprojectionsfocussingonwater
Abstract Introduction
D
resources and flood risk management in Brahmani-Baitarani river basin.
is
c Conclusions References
u
s
2.1 Global Circulation Model selection s
io Tables Figures
n
Data sets generated in simulations from three GCMs undertaken for the Fifth Coupled P
a
15 Model Inter-comparison Project (CMIP5 – Taylor et al., 2012; Knutti and Sedláček, pe J I
r
2013)havebeendownloadedwhicharetheHadleyCenterGlobalEnvironmentModel J I
|
2–EarthSystem(HadGEM2-ES),theGFDL-CM3,andtheMIROC-ESM(seeTable1).
Dataforthetimeslicesrepresentingperiods1961–1990(baseline),shorttermclimate D Back Close
is
change projection (2030–2059), and long term projection (2060–2099) were used. c
u FullScreen/Esc
s
20 Almost all CMIP5 models show good performance for surface air temperature simu- sio
lations averaged over South Asia and the Indian Sub-continent, with MIROC5 as one n
Printer-friendlyVersion
P
of the models closest to the observations. Yet for area-averaged annual and seasonal a
p
rainfallthemodelssignificantlydeviatefromobservationsoverIndia(Chaturvedietal., e InteractiveDiscussion
r
5 |
2012). Here GFDL-CM3 is one of the best performing models. In addition, HadGEM2- D
is
ESisselectedbecauseitisoneofthemostcommonlyusedGCMs.ForIndiathethree c HESSD
u
s
selectedGCMsnearlyspantheuncertaintybandforannualmeanrainfallthatwasob- s
io doi:10.5194/hess-2015-499
tainedwith18EarthSystemModels(ESMs)usedinavalidationexercise(Mondaland n
P
5 Mujumdar,2014;RameshkumarandGoswami,2014).InterpretingMenonetal.(2013), a
p
HadGEM2-ES could be regarded as “too dry”, GFDL-CM3 as “representative”, and e Downscaling GCM
r
MIROC-ESM as “quite wet”. data for climate
|
change impact
2.2 Representative Concentration Pathway (RCP) Selection D assessments on
is
c
u rainfall
The RCPs as considered in IPCC AR5 (Vuuren et al., 2011) are compatible with the s
s
10 full range of stabilization, mitigation, and baseline emission scenarios, and span a full ion R.J.Dahmetal.
range of socio-economic driving forces (Hibbard et al., 2011). In this study, we consid- P
a
p
ered RCP6.0 for the generation of the climate change projections as it follows a stabi- e
r
lizing CO concentration close to the median range of all four policy pathways and will TitlePage
2
|
give us a not too extreme indication of what might change. Most likely the spread and
Abstract Introduction
uncertainty in projections will increase when including additional RCPs and changes D
15 is
maybecomemorepronouncedwhenusingtheextremeRCP8.5(Hibbardetal.,2011). c Conclusions References
u
s
s
2.3 Observed rainfall ion Tables Figures
P
a
ObservedrainfallwasobtainedfromtheAPHRODITElong-termdailygriddedprecipita- p J I
e
r
tiondataset(versionV1101)forMonsoonAsia(Yatagaietal.,2012).TheAPHRODITE
J I
(Asian Precipitation Highly Resolved Observational Data Integration Towards Evalua- |
20
tionofWaterResources, Japan)isalong-term(1951–2007)continental-scaleproduct D Back Close
is
thatcontainsadensenetworkofdailyraingaugedataforAsiaincludingtheHimalayas, c
u FullScreen/Esc
South and Southeast Asia. The APHRODITE dataset consists of 0.25◦×0.25◦ reso- s
s
io
lution gridded daily precipitation derived from the Global Telecommunication System n
Printer-friendlyVersion
(GTS), precompiled datasets and APHRODITE’s individual data collection. It covers P
25 a
p
the period 1951–2007. The data is quality controlled and corrected for orographic ef- e InteractiveDiscussion
r
6 |
fects.It includesover2000stationsover Indiaandcapturesthe largescalefeaturesof D
is
monsoon rainfall over the Indian region well (Rajeevan and Bhate, 2008). Hereafter it c HESSD
u
s
is referred to as the climatological rainfall dataset. Within this paper we perform addi- s
io doi:10.5194/hess-2015-499
tional validation focusing on extremes using observations from three rain gauges that n
P
5 were obtained from the Central Water Commission of India (CWC). Locations of the a
p
raingaugesarelistedinTable2andshowninFig.1.Gaugeobservationscouldnotbe e Downscaling GCM
r
used as reference since they do not overlap with the full GCM baseline period. data for climate
|
change impact
2.4 Bias correction D assessments on
is
c
u rainfall
There is often a clear bias from observations in the statistics of variables produced by s
s
10 GCMs such as temperature and rainfall due to limitations in among others the incor- ion R.J.Dahmetal.
poration of local topography and non-stationary phenomena within the GCMs. GCM P
a
p
outputs can therefore often not be directly applied for impact studies at the catchment e
r
scale (Christensen et al., 2008; Kay et al., 2006; Kotlarski, 2005; Fan et al., 2010). TitlePage
|
Dynamic downscaling, statistical downscaling, and bias correction are the most com-
Abstract Introduction
monly used methods to generate locally applicable climate data (Bergstrom, 2001; D
15 is
Fowleretal.,2007;Schoofetal.,2009).Dynamicdownscalingincludesnestingofhigh c Conclusions References
u
s
resolution Regional Climate Models (RCMs) with GCM outputs at boundaries which s
io Tables Figures
ensuresconsistencybetweenclimatologicalvariables.However,theyarecomputation- n
P
ally expensive, their skill strongly depends on GCM boundary and for India there are a
p J I
no ready to use RCM scenarios for the RCP emission scenarios at hand. Statistical e
20 r
downscalingmodels,ontheotherhand,arebasedonstatisticalrelationshipsbetween J I
|
large-scale climate variables (predictors) and local-scale climate variables (predictant)
D Back Close
and hence require less computational time. Yet for the establishment of reliable sta- is
c
tistical relationships long historical observed records of both predictor and predictant u FullScreen/Esc
s
25 variables should be available. sio
Simpler bias correction procedures consisting of general transformation techniques n Printer-friendlyVersion
P
for adjusting the GCM output time series are often used. They assume a stationary a
p
e InteractiveDiscussion
bias between GCM output and observations for the baseline period and future climate r
7 |
(Teutschbein and Seibert, 2012a). With the more advanced Quantile Mapping method D
is
(Thrasher et al., 2012; Camici et al; 2014) changes in mean and extreme rainfall are c HESSD
u
s
corrected individually – yet this method requires reliable CDFs of observed rainfall s
io doi:10.5194/hess-2015-499
and from the comparison between the APHRODITE dataset and station observations n
P
5 (Rana et al., 2014) we know that large biases in rainfall extremes exist. a
p
Therefore,werestrictourselvestotwosimplemethods(1)theDeltaChangemethod e Downscaling GCM
r
and (2) the Linear Scaling method and applied these to the data extracted for our data for climate
|
region of interest from all three GCM outputs (Christensen et al., 2008; Chen et al., change impact
D
2012),realizingthatthesemethodsonlyconcentrateonthecorrectionofmonthlymean assessments on
is
10 rainfallamountsandcanthusaffectouranalysisofchangesinextremerainfallindices. cu rainfall
s
s
io
2.4.1 Delta change method n R.J.Dahmetal.
P
a
p
The Delta Change method transforms historical observations into future projections e
r
using monthly average correction factors that are derived from the GCM simulations TitlePage
|
forthebaselineandfutureclimate(Camicietal.,2014;Chenetal.,2013;Teutschbein
Abstract Introduction
et al., 2012b) according to: D
15 is
c Conclusions References
u
s
P_GCMfut s
∆P = m (1) io Tables Figures
m n
P_GCMBmL Pap J I
P_OBSfut=∆P P_OBSBL (2) er
m
J I
|
wherefutreferstofutureclimate,BLreferstobaselineclimate(i.e.,1961–1990),GCM
D Back Close
refers to GCM simulations and OBS refers to either observations (BL period) or future is
c
20 projections (fut), P refers to daily rainfall values and Pm refers to monthly long term us FullScreen/Esc
mean rainfall values. s
io
Thedisadvantageofthismethodisthatthefutureandbaselinescenariosdifferonly n Printer-friendlyVersion
P
in terms of their means and intensity while all other statistics of the data, such as the a
p
e InteractiveDiscussion
skewness and number of wet days, remain almost unchanged (Camici et al., 2014). r
8 |
Thiswillhampertheassessmentofchangeinextremerainfallfrequencyasthechange D
is
in return period is essentially a derivative of the multiple factors. c HESSD
u
s
s
2.4.2 Linear scaling method io doi:10.5194/hess-2015-499
n
P
a
Within the Linear Scaling method the long-term average monthly bias between the p
e Downscaling GCM
baseline (i.e., 1961–1990) GCM simulations and observations is derived and this bias r
5 data for climate
is applied to correct the future GCM simulations assuming a stationary bias over time, |
change impact
according to: D
assessments on
is
c
u rainfall
BIAS = P_OBSBmL (3) ssio
m n R.J.Dahmetal.
P_GCMBL P
m a
p
P_OBSfut=BIAS P_GCMfut (4) e
m r
TitlePage
|
where BIAS is the monthly average bias for the baseline climate and the correction
10 m Abstract Introduction
factorforthefutureGCMsimulations.Theadvantageofthismethodisthatthevariabil- D
is
ity of the corrected data, for both baseline and future climate is more consistent with c Conclusions References
u
s
the GCM data, which implies that changes in wet day frequencies and intensities can s
io Tables Figures
be derived, yet they are not individually corrected as all events are adjusted with the n
P
15 same monthly average correction factor (Teutschbein et al., 2012b). ap J I
e
r
2.5 Impact analysis framework J I
|
To study the impact of climate change on rainfall and associated fields like water re- D Back Close
is
sources and flood risk management, we selected the following four indices. c
u FullScreen/Esc
s
1. Mean annual and seasonal rainfall in order to explore possible effects of climate sio
n
change on water resources and flood risk management. Printer-friendlyVersion
20 P
a
p
e InteractiveDiscussion
r
9 |
2. Frequency of seasonal Wet Days (WD). According to the definition followed by D
is
the India Meteorological Department (IMD), a day with rainfall is considered a c HESSD
u
s
“wet day” if the rainfall equals or exceeds 2.5mm. s
io doi:10.5194/hess-2015-499
n
3. Seasonal frequency of days with light, moderate, and heavy rainfall (LRD, MRD, P
a
andHRDrespectively).Table3showstheIMDclassificationforrainfallintensities. p
5 e Downscaling GCM
We clustered all “heavy rain” classifications into one group to examine climate r
data for climate
change induced alterations in potential flood inducing events. |
change impact
D
4. One-day extreme rainfall return periods to investigate the potential impact of cli- is assessments on
c
mate change on rainfall intensities. In this study we applied the Gumbel extreme u rainfall
s
10 value type-1 distribution (Gumbel, 1941) to derive rainfall intensities for different sion R.J.Dahmetal.
return periods (2, 5, 10, 25, 30, 50, and 100years).
P
a
p
Table 4 shows the classification of seasons according to IMD. e
r
TitlePage
2.6 Study area |
Abstract Introduction
D
This study focusses on the Brahmani-Baitarani river basin located in the eastern part isc Conclusions References
u
of India. The Brahmani-Baitarani river basin and neighbouring Mahanadi river basin s
15 s
experienced serious floods in the years 2001, 2003, 2006, 2008, 2011 and 2013. Dur- io Tables Figures
n
ing September 2011 heavy rainfall together with high sea levels led to the flooding of P
a
large parts of the Delta. It affected about 3.4million people of which 45 people lost pe J I
r
their life (Government of Odisha, 2011). In 2013 cyclone Phailin created havoc along J I
|
the coastal districts of Odisha state. Due to storm surge up to 3.5m, large areas were
20
inundated. The Baitarani River, along with other rivers, experienced floods as a result D Back Close
is
oftorrentialdownpour.Nolessthan13millionpeoplewereaffectedofwhich44people c
u FullScreen/Esc
s
lost their life (Government of Odisha, 2013). s
io
The basin is an inter-state basin and spreads across the states of Chhattisgarh, n
Printer-friendlyVersion
P
25 Jharkhand, and Odisha. The elevation ranges from >750m in the north-western part ap
of the basin to approximately 10m in the delta region. The Baitarani River enters the e InteractiveDiscussion
r
10 |