Transcript

INTERNATIONAL JOURNAL OF CLIMATOLOGYInt. J. Climatol. (2012)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/joc.3509

Probabilistic downscaling of GCM scenarios over southernIndia

N. Vigaud,a* M. Vracb and Y. Caballeroc

a Service EAU, Bureau de Recherches Geologiques et Minieres (BRGM), Montpellier, Franceb Laboratoire des Sciences du Climat et de l’Environnement (LSCE-IPSL), CEA/CNRS/UVSQ, Gif-sur-Yvette, France

c Service Geologique Regional, Bureau de Recherches Geologiques et Minieres (BRGM), Montpellier, France

ABSTRACT: The cumulative distribution function transform (CDF-t) is used to downscale daily precipitation and surfacetemperatures from a set of Global climate model (GCM) climatic projections over southern India. To deal with the fullannual cycle, the approach has been applied by months, allowing downscaled projections for all seasons. First, CDF-t isvalidated over a historical period using observation from the Indian Meteorological Department (IMD). Resulting highresolution fields show substantial improvements compared to original GCM outputs in terms of distribution, seasonal cycleand monsoon means for arid, semi-arid and wetter regions of the subcontinent. Then, CDF-t is applied to GCM large-scalefields to project rainfall and surface temperature changes for the 21st century under the IPCC SRES A2 scenario. Theresults obtained show an increase of rainfall, mostly during the monsoon season, while winter precipitation is reduced, andsuggest a widespread warming especially in the winter and post-monsoon season. Copyright 2012 Royal MeteorologicalSociety

KEY WORDS downscaling; climatic scenarios; southern India

Received 14 June 2011; Revised 22 March 2012; Accepted 3 April 2012

1. Introduction

Associated with fast social and economic growth locally,climate changes are likely to seriously impact India. Asnoted by Kundzewicz et al. (2007), southern India isalready a water-stressed region. Climate changes havealready been observed over the subcontinent, whereincreases of 0.4–0.6 °C have occurred over the past cen-tury together with the annual mean temperature warmingmost pronounced during post-monsoon and winter peri-ods (Rupa Kumar et al., 2006; Bhattacharya, 2007). Interms of precipitation, Cruz et al. (2007) have observedfor the last decades that extreme summer monsoon rainsincrease over northwest India and the number of rainydays decrease along the east coast. While these projec-tions are subject to large uncertainties (Paeth et al., 2008),the potential impacts on water resources in India still needto be assessed depending on their location.

Global climate models (GCMs) are nowadays the onlytool at disposal to investigate future climate variability.However, GCM projections cannot be used directly forimpact studies due to the coarse resolution of GCM out-puts which are not suited for regional assessments (Wilbyet al., 2004). Therefore, downscaling methods have beendeveloped to go from large-scale data to local-scaledata. The dynamical approach consists of using regional

* Correspondence to: N. Vigaud, Service EAU, Bureau de RecherchesGeologiques et Minieres (BRGM), Montpellier, France.E-mail: [email protected]

climate models (RCMs) to resolve physical equationsof atmospheric regional dynamics (Wood et al., 2004).RCMs are, however, domain dependant and computation-ally expensive, which restricts their use for many applica-tions. The statistical approach, on the other hand, refers tostatistical relationships between large-scale GCM featuresand local-scale climatic variables (such as precipitation ortemperature, for instance). Statistical downscaling meth-ods (SDMs) are quite flexible and generally require lesscomputational costs. Such advantages make them partic-ularly attractive for regional impact studies. SDMs canbe classified into three major categories: transfer func-tions, weather typing and weather generators. Transferfunctions are based on direct quantitative relationshipsbetween predictand and predictors through regression-like methods (Prudhomme et al., 2002). Weather typingapproaches consist in the grouping (or clustering) ofatmospheric circulations in relation to local meteorolog-ical variables (Vrac et al., 2007), while weather genera-tors are stochastic models simulating local-scale variablesbased on their probability density function, whose param-eters depend on large-scale information (Hughes et al.,1999; Wilks and Wilby, 1999; Vrac and Naveau, 2007).Worthnotingly, a common assumption to all SDMs is thatthe physical relationships underlying the statistical rela-tionships identified over a historical period remain validfor the future climate scenarios to be downscaled.

Several SDMs have already been used for downscal-ing rainfall over India, such as relevance (Ghosh and

Copyright 2012 Royal Meteorological Society

N. VIGAUD et al.

Mujumdar, 2008; Mujumdar and Ghosh, 2008) and sup-port vector machine (Tripathi et al., 2006; Anandhi et al.,2008, 2009), fuzzy clustering (Ghosh and Mujumdar,2007; Mujumdar and Ghosh, 2008) and also conditionalrandom field distribution (Raje and Mujumdar, 2009). Inthese recent studies, methods are generally validated byexamining the quantiles, cumulative and probability dis-tribution functions (CDFs and PDFs) of daily rainfall aswell as dry/wet spells’ lengths at local scale. However,SDMs can also be used to model relationships betweenlarge-scale and local-scale statistical characteristics andcan be referred in this context as probabilistic downscal-ing methods (PDMs). The cumulative distribution func-tion transform (CDF-t) presented in Michelangeli et al.(2009) has the advantage of directly dealing with andproviding CDFs. This method is used in this paper todownscale GCM projections from the Intergovernmen-tal Panel on Climate Change Fourth Assessment Report(IPCC AR4) in order to investigate projected changesin both rainfall and surface temperatures over south-ern India. The purpose of this study is to documentthe use of CDF-t for providing regional precipitationand surface temperature changes as derived from severalmedium-term GCM projections under the Special Reporton Emission Scenarios (SRES published by the IPCC) A2scenario (horizons 2040–2060). Most of the recent statis-tical downscaling studies of future climate scenarios overIndia were restrained to the monsoon season (Tripathiet al., 2006; Mujumdar and Ghosh, 2008) or to specificwatersheds (Ghosh and Mujumdar, 2007; Anandhi et al.,2008, 2009). This paper aims to present results from theCDF-t probabilistic downscaling method applied not onlyto the June–September (JJAS) monsoon period alone butto the full annual cycle, and for a domain covering thewhole of southern India. In the next section, the dataused and the downscaling method are presented. Then,the CDF-t approach is validated on a historical periodin Section 3, prior to applying the method to downscalefuture scenarios from IPCC AR4 experiments in Section

4. Discussion and conclusions are then gathered at theend in Section 5.

2. Data and methods

2.1. GCMs data and observations

In order to investigate climate change impacts over south-ern India, several GCMs from the last IPCC AR4 exercisehave been used. Kripalani et al. (2007) identified a set ofseven GCMs as most reliable regarding Indian monsoonrainfall based on their representation of the mean sea-sonal cycle in phase and amplitude. Regarding the avail-ability of daily standard outputs from the Program forClimate Model Diagnosis and Intercomparison (PCMDI)database, five GCMs have been retained (see Table I).However, at the time of this study MIROCMR dailysurface temperatures for the SRES A2 scenario werenot available from the PCMDI archives. Consequently,only four GCMs will be used for the downscaling ofsurface temperatures. Except for CGCM3, which incor-porates heat and water fluxes adjustments, these new-generation GCMs do not use surface flux correction tomaintain a stable climate in their control runs. Moredetails about the model components can be found athttp://www.pcmdi.llnl.gov/ipcc/model-documentation/.

For the purpose of this study, simulated daily rainfalland surface temperatures have been considered from bothhistorical experiments (run 20cm3 for the 1971–1999period) and future projections under the greenhousegas emission scenario A2 (run A2 for the 2046–2065period). The A2 storyline, based on high population andregionally oriented economic growth with significant andwidespread decline in fertility (Nakicenovic et al., 2000),actually describes a very homogeneous world with slowereconomic and technical changes than other scenarios.

Local-scale observations from the Indian Meteorolog-ical Department (IMD) are also used in this study. IMDdaily rainfall is available from 1971 to 2005 on a half-degree grid and surface temperatures from 1969 to 2005

Table I. Climate models and their references participating in the IPCC AR4 experiments (adapted from Kripalani et al. (2007)).Abbreviated acronyms are used in the text to identify each GCM.

No. Originating group Country IPCC ID Abbreviation References

1 Canadian centre for climatemodelling

Canada CGCM3.1 (t47) CGCM3 Flato et al. (2000)

2 Meteo-France/Centre Nationalde Recherches Meteorologiques

France CNRM-CM3 CNRM3 Salas-Media et al.(2006)

3 Max Planck Institute forMeteorology

Germany ECHAM5/MPI-OM ECHAM5 Jungelaus et al.(2006)

4 Bjerknes Centre for ClimateResearch

Norway BCCR-BCM2.0 BCCR2 Furevik et al.(2003)

5 Centre for Climate SystemResearch (The University ofTokyo) National Institute forEnvironmental Studies andFrontier Research Centre forGlobal Change (JAMSTEC)

Japan MIROC3.2 (medres) MIROCMR K-1 ModelDevelopers (2004)

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

at one degree resolution. It leads to sets of 729 and210 grid points for rainfall and surface temperature,respectively, over a region comprised between 7.5 °N and20.5 °N and 70 °E and 83 °E to which we will refer in thefollowing as southern India. More details about the IMDhigh resolution daily gridded rainfall and surface temper-atures dataset can be found in Rajeevan and Bhate (2008)and Srivastava et al. (2008), respectively.

2.2. CDF-t method

The downscaling approach chosen here is the CDF-t (Michelangeli et al. (2009)) which can be seen asan extension of the quantile-matching method. CDF-toffers the advantage to directly deal with and provideCDFs. In its non-parametric form it does not make anyassumption on the shape or family of distribution andthus can be applied separately to both rainfall and surfacetemperatures in the context of this study. CDF-t hasalready been successfully used to downscale GCMs andreanalyse 10 m wind over France by Michelangeli et al.(2009).

Let Fh stand for the CDF of observed local dataat a given weather station (or IMD grid cell) over ahistorical time period h, Gh the CDF of GCM outputs bi-linearly interpolated at the station location for the sameperiod, Ff and Gf their equivalent for the future periodconsidered. The method is based on the assumption thatthere exists a transformation T translating the CDF ofa GCM variable (predictor) into the CDF representingthe local-scale climate variable (predictand) at the givenweather station, through the transformation T : [0, 1] →[0, 1]

T (Gh(x)) = Fh(x) (1)

Replacing x by G−1h (u) in Equation (1) with u ∈ [0, 1]

allows the following definition for the transform T :

T (u) = Fh(G−1h (u)) (2)

Assuming that this later relationship remains valid in thefuture (i.e. Ff = T (Gf )), the researched CDF is givenby:

Ff (x) = Fh(G−1h (Gf (x))) (3)

Following Michelangeli et al. (2009), the CDF-t is thendefined in two steps. First, estimates of (Fh, G−1

h , Gf )are non-parametrically modelled. Then, their combinationusing Equation (3) provides an estimate of Ff . Unlike theclassical quantile-matching approach which projects thesimulated future large-scale values on the historical CDFto compute and match quantiles, CDF-t takes into accountthe change in the large-scale CDF from the historical tothe future period. To downscale rainfall/surface tempera-tures over the full annual cycle, CDF-t is applied at eachgrid point to multiannual chronicles consisting of dailyrainfall/surface temperatures for each month of the cal-endar year (all January months, February months, etc.).The resulting local-scale daily chronicles of each month

are then rearranged to produce yearly estimates for theperiod chosen.

3. Calibration and validation over a historicalperiod

The CDF-t method is first calibrated and validated over aso-called historical period (1971–1999) which is cut intotwo contiguous temporal windows, 1971–1985 used forcalibration and 1986–1999 for which downscaled resultsare evaluated regarding local observations from IMD.The resulting downscaled CDFs are evaluated at eachgrid point against IMD data using Kolmogorov–Smirnov(KS) statistics providing an estimate of the maximumdifference between downscaled and observed local CDFs(Darling, 1957). The results for both rainfall and surfacetemperatures over southern India are presented. However,because rainfall has a discontinuity in zero, KS scoreshave also been computed after removal of days withno precipitation, this analysis is further discussed inSection 3.1. In addition, dedicated diagnostics are donefor three watersheds ranging from arid (Pandam Eru),semi-arid (Kudaliar) to wetter conditions (South Gundal).These locations are chosen to illustrate and discussthe performance of the method for different climaticenvironments.

3.1. Precipitation regimes

Figure 1 presents KS statistics between CDFs fromoriginal GCMs rainfall (bi-linearly interpolated on the0.5° IMD grid) as well as from downscaled GCMs dataand IMD observed precipitation for 1986–1999 withcalibration of the CDF-t over the 1971–1985 period.As mentioned previously, this KS test is performed foreach grid point; therefore, the box-plots displayed inFigure 1 represent the spatial dispersion of the KS scoresfor the whole of the South India domain. For bothperiods, the different GCMs downscaled rainfall dataare characterized by a spatial dispersion comparable tooriginal GCM outputs. For the full year (top panels)and for all GCMs, even if they stay above the levelof statistical significance (0.019 at 0.05 significancelevel, not plotted), KS scores are substantially improvedfor downscaled daily rainfall compared to raw GCMsdata. Similar results are found for the monsoon period.Maximum KS scores computed in Figure 1 actually occurat the discontinuity of rainfall in zero, and the aboveresults thus show that dry days are better represented indownscaled fields than in original GCM outputs whencompared to observation. This is consistent with thefact that dry days are generally rare in GCMs data.Consequently, similar KS scores have been plotted inFigure 2 after removal of days with no rainfall (zerovalues) in all precipitation dataset. For the whole year, theresults are very contrasted depending on GCMs. CDF-t seems to perform best in the case of ECHAM5 forwhich downscaled rainfall are closer to observation thanoriginal GCM outputs. While there is no improvement

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

Annual KS Test for CGCM3

CNRM3 CNRM3–CDFt

Annual KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

Annual KS Test for ECHAM5

BCCR2 BCCR2–CDFt

Annual KS Test for BCCR2

MIROCMR MIROCMR–CDFt

Annual KS Test for MIROCMR

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

JJAS KS Test for CGCM3

CNRM3 CNRM3–CDFt

JJAS KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

JJASl KS Test for ECHAM5

BCCR2 BCCR2–CDFt

JJAS KS Test for BCCR2

MIROCMR MIROCMR–CDFt

JJAS KS Test for MIROCMR

Figure 1. Kolmogorov–Smirnov statistics for original (left box-plots) and downscaled (right box-plots) GCMs rainfall when compared to IMDobservation data over the 1986–1999 period for the whole of southern India, with calibration of the CDF-t over 1971–1985. Statistical scoresfor the full year and the JJAS period are presented in the top and bottom panels, respectively. This figure is available in colour online at

wileyonlinelibrary.com/journal/joc

for CNRM3, the performance of the method is mitigatedin the case of BCCR2, MIROCMR and CGCM3. For theJJAS period however, downscaled rainfall using CDF-tare all closer to observation than original GCMs data,witnessing of a good performance of CDF-t over themonsoon season.

In addition, downscaled GCMs daily rainfall fields areevaluated in regards to the seasonal cycle and meanmonsoon precipitation rate over the validation period1986–1999. Mean rainfall seasonal cycles shown inFigure 3 for Pandam Eru, Kudaliar and South Gundallocations (referenced in Figure 4) suggest for all GCMs,substantial improvements for downscaled data comparedto large-scale outputs. The gain appears to be betterfor humid (South Gundal) and semi-arid (Kudaliar)regions than for arid areas (Pandam Eru), illustrating thevarying performance of the method depending on climaticenvironments.

The mean JJAS pattern from IMD observed rainfall(Figure 4 upper right panel) shows maximum rainfallrates over the northeastern regions and along the westcoast, while minimum values over central southern Indiaseparate these two regions. Clearly, such a structure isnot found accurately in any of the GCMs used in thisstudy. Nevertheless, these gradients are well reproducedin downscaled rainfall fields for all GCMs. The main

differences between downscaled GCMs data are charac-terized by the amplitudes of the above extremes duringthe monsoon season: for instance, greatest maximum val-ues over northeast India are found for CNRM3 down-scaled rainfall while central regions of southern India arethe least dry for BCCR2 resulting local-scale data.

Dry spells lengths PDFs at Pandam Eru, Kudaliar andSouth Gundal locations are shown for original and down-scaled GCMs together with IMD observation in Figure 5.The different climatic conditions over these watershedsare well represented from observed rainfall with increas-ing slopes from arid to wetter climate. Inflexion pointswithin short dry spells lengths suggest that most of thedry spells have a duration below five days but with avarying proportion over all watersheds, short dry spellsbeing less prevalent for arid and semi-arid regions (about60–70%) than for wetter climate (almost 80%). Suchdifferences are less clear for the original GCMs data(dashed coloured lines): large biases are found concern-ing both slopes and proportion of short and longer dryspells. Nevertheless, downscaled fields (thick colouredlines) systematically exhibit a better representation ofdry spells lengths compared to raw GCMs rainfall forarid (Pandam Eru) and semi-arid (Kudaliar) regions. Atthese locations out of the five GCMs dataset, ECHAM5appears to be the closest to observed dry spells lengths

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

Figure 2. Similar to Figure 1 but after removal of days with no rainfall. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

from IMD data, similarly for downscaled fields, down-scaled ECHAM5 rainfall provide the best results. Theremaining four GCMs display larger biases while sub-stantial gains in dry spells representation characterizetheir corresponding downscaled rainfall fields. For wet-ter climatic conditions (South Gundal), similar but lessefficient improvements are noticeable between originaland downscaled GCMs precipitation except in the case ofECHAM5 for which dry spells lengths PDF from originalGCM outputs seems closer to IMD observation than theresulting downscaled data. This could indicate a lesserperformance of the CDF-t method for wetter climaticregions in terms of dry spells lengths.

Finally, JJAS anomalies over the 1986–1999 periodfor original and downscaled GCMs rainfall at the dif-ferent watersheds (not shown) suggest that interannualvariability within the resulting local-scale data is drivenby the GCM outputs. Compared to IMD observations,no GCM seems able to reproduce the observed year toyear monsoon rainfall variability, and such is the caseregarding their corresponding downscaled fields.

3.2. Surface temperatures

Regarding annual fields, even though some scores are stillabove significance level (not plotted), KS statistics shownin Figure 6 (top panels) exhibit systematic improvementsfor downscaled GCMs surface temperatures when com-pared to observation. Interestingly, the spatial dispersionof the KS is much smaller for the resulting high resolu-tion fields than for original GCMs data. Concerning themonsoon period, no such gain is found for JJAS KS diag-nostics (Figure 6 bottom panels): except from CNRM3,statistical scores are similar to those obtained for originalGCMs data.

Mean surface temperatures seasonal cycle at PandamEru, Kudaliar and South Gundal locations are presentedin Figure 7 for the 1986–1999 period. For all watersheds,in dry or more humid conditions, downscaled fields aresystematically improved compared to original GCMssurface temperatures and are fitting closely the IMDobservations over this historical period.

All mean JJAS GCMs downscaled surface tempera-tures (Figure 8) display patterns close to what is observedfor IMD data while fields from original GCMs werepresenting substantial biases. In particular, the latitudi-nal gradient along the west coast which is subject tomarked discrepancies in large-scale GCM outputs, is wellrepresented in all downscaled data. The meridional gra-dient from western to eastern parts of the subcontinentis also better reproduced in resulting local-scale surfacetemperatures, with a more or less marked minimum overnortheastern regions, as found in IMD observations.

Mean JJAS surface temperatures differences betweenthe validation (1986–1999) and calibration (1971–1985)periods are plotted in Figure 9 for raw GCM outputs andIMD observations. Noteworthingly, no much differenceis found for IMD observations but also for CNRM3 datawhile other GCMs are characterized by more or lessmarked variations. Given that CDF-t downscaled fieldsare strongly driven by the evolution of the original GCMoutputs, this could explain the poor JJAS KS scores inFigure 6 for CGCM3, ECHAM5, and BCCR2 resultinglocal-scale data.

In addition, JJAS surface temperature anomalies at thedifferent watersheds locations for the 1986–1999 period(not shown) exhibit substantial discrepancies in termsof interannual variability for raw GCM outputs whencompared to observations. Here again, these differencesremain in the downscaled data.

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

J F M A M J J A S O N D0

2

4

6

8

10

12

14M

ean

Rai

nfal

l (m

m/d

ay)

Months

J F M A M J J A S O N D0

2

4

6

8

10

12

14

Months

J F M A M J J A S O N D0

2

4

6

8

10

12

14

Mea

n R

ainf

all (

mm

/day

)

Months

J F M A M J J A S O N D

Months

J F M A M J J A S O N D0

2

4

6

8

10

12

14

Months

(1986–1999) MeanGCM Rainfall @ Pandam Eru

(1986–1999) MeanGCM–CDFt Rainfall @ Pandam Eru

(1986–1999) MeanGCM Rainfall @ Kudaliar

(1986–1999) MeanGCM–CDFt Rainfall @ Kudaliar

IMD

CNRM

CGCM

ECHAM

BCCR

MIROC

IMD

CNRM CDF–t

CGCM CDF–t

ECHAM CDF–t

BCCR CDF–t

MIROC CDF–t

0

2

4

6

8

10

12

14

16

18

20

22

Mea

n R

ainf

all (

mm

/day

)

J F M A M J J A S O N D

Months

0

2

4

6

8

10

12

14

16

18

20

22

(1986–1999) MeanGCM Rainfall @ South Gundal

(1986–1999) MeanGCM–CDFt Rainfall @ South Gundal

Figure 3. Mean seasonal cycle (in mm d−1) at Pandam Eru (top panel), Kudaliar (middle panel) and South Gundal (bottom panel) locations fororiginal (left) and downscaled (right) GCMs rainfall together with IMD observation (thick blue line) over the 1986–1999 period, with calibration

of the CDF-t over 1971–1985. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

4. Application to projected GCM A2 scenarios

The CDF-t method is now applied to downscale large-scale precipitation and surface temperature projectionsfrom IPCC AR4 experiments under the greenhousegas emission scenario A2 over southern India. As forthe historical period, resulting high resolution fields

are generated by applying CDF-t by months on dailyGCM outputs. Regarding the availability of daily datasetfor the GCMs selected within the PCMDI archive,2046–2065 (hereafter A2 period) is the period chosenfor downscaling future A2 scenarios at medium term.The 1971–1999 period is used for calibration this timein order to take advantage of the longest period of

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

Figure 4. Mean June to September rainfall (in mm d−1) for original (top panels) and downscaled (bottom panels) GCMs data together with IMDobservation (top right) over the 1986–1999 period for the whole of southern India, with calibration of the CDF-t over 1971–1985. PandamEru (Pa), Kudaliar (Ku) and South Gundal (Gu) locations are referenced on the maps for indication. This figure is available in colour online at

wileyonlinelibrary.com/journal/joc

observation available from IMD. It is worth noting thatthe aim of this paper is not to fully investigate thepotential impacts of climate change over southern Indiabut rather to document the use of the CDF-t method forstatistical downscaling of a future climatic scenario overthe full year cycle.

4.1. Projected rainfall changes

Projected large-scale as well as downscaled GCM outputsare now compared to their corresponding 1971–1999fields using KS tests (Figure 10). For both the annualcycle and the monsoon period, systematic higher KSscores, are found for downscaled data. In the lightof Section 3.1 the medians (red lines) would indicatemore dry days for A2 scenario projections than for thecontemporary period. When computing KS after removalof days with no rainfall (zero values) very similar plotswere obtained (not shown), suggesting more pronouncedprojected climate changes in local-scale precipitationthan for raw GCM outputs. The results obtained alsoexhibit a greater spatial dispersion for downscaled fieldsrelatively to original large-scale GCM outputs, suggestingthat resulting changes linked to climate change are morecontrasted geographically at local-scale for all GCMs.

Mean seasonal rainfall changes are shown in Figure 11at Pandam Eru (left), Kudaliar (centre) and South Gundal(right) locations for all downscaled GCMs individually.In order to have a benchmark to which CDF-t downscaledfields could be compared, the differences of GCMs sea-sonal cycles between two 20 year periods have been com-puted to represent the climate change signal as depictedin the raw GCMs data. Hereafter, rainfall changes are cal-culated relative to the historical 1980–1999 period (XXperiod in the following).

The dispersion within the different GCMs projectedchanges illustrates clearly the need to consider a panelof GCMs for climate change impact studies over south-ern India. In terms of mean GCMs ensemble, the CDF-tmethod gives similar results to original GCM outputs:

most pronounced changes are found during the mon-soon season over all watersheds with maximum variationsduring May–June and August. Nevertheless, as inferredfrom JJAS KS scores discussed previously, changes inprecipitation between the A2 and XX periods seemto differ in magnitude from one location to another.For the arid Pandam Eru basin, maximum changes arejust below 1 mm d−1 in June and August (representingapproximately a 15% increase of the monthly mean pre-cipitated amount) with very small differences betweenCDF-t results and original GCM fields. At Kudaliar,raw GCMs data exhibit similar changes (below 1 mmd−1) while the CDF-t method shows variations of almost2 mm d−1 (about 50% increase) in June. For the wet-ter South Gundal location, differences increase betweenCDF-t downscaled fields and original GCM outputs.Maximum downscaled precipitation changes are rangingfrom about 1 mm d−1 (1.5 mm d−1 for raw GCM out-puts) in May–June to 1.5 mm d−1 (2.8 mm d−1 for orig-inal GCMs data) in August (about 10% to 15% increase).While differences in CDF-t projected and GCMs originalrainfall changes seem to increase from arid to wetter cli-matic regions of southern India, noteworthingly the dis-persion within downscaled GCMs climate change signalsalso becomes more pronounced. In addition, weak nega-tive changes are found during the dry season, most par-ticularly from October to April, from the CDF-t methodas well as from raw GCMs data. At Pandam Eru loca-tion, both the CDF-t approach and original GCMs rain-fall exhibit variations just below zero from October toDecember while no substantial changes seem to prevailfrom January to April. Similar results are found fromraw GCM outputs at Kudaliar while negative changesare slightly more pronounced for CDF-t estimates (below0.5 mm d−1 corresponding to a 50% decrease). Finally,more substantial negative changes are characteristic ofthe South Gundal basin: the CDF-t approach and originalGCMs data suggest most pronounced rainfall decreasesfrom October/November to December (up to 1 mm d−1,

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

Figure 5. Dry spells length PDFs (in %) at Pandam Eru (top left), Kudaliar (top right) and South Gundal (bottom) locations for original (dashedcoloured lines) and downscaled (thick coloured lines) GCMs data together with IMD observations (black thick line) over the 1986–1999 period,

with calibration of the CDF-t over 1971–1985. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

about 80% of precipitated amounts) respectively whilenegative variations from January to April are less substan-tial (maximum about 0.5 mm d−1 in March, about 50%decrease). Overall, these results corroborate findings fromprevious studies emphasizing enhanced precipitation oversouthern India, mostly during the monsoon season (RupaKumar et al., 2006; Tripathi et al., 2006; Kripalani et al.,2007; Anandhi et al., 2008).

Regarding JJAS rainfall means (Figure 12 two top pan-els), downscaled fields are more coherent with the known

climatology from IMD data for the historical period (seeFigure 4) than raw GCMs. In particular, maximum pre-cipitated amounts along the west coast and northeastregions of the subcontinent as well as the central pat-terns associated to more arid regions, are well representedin the resulting downscaled fields from all GCMs. Ofcourse, there is no benchmark to compare future pro-jections, nevertheless these regional variations are farmore realistic for downscaled rainfall than for the orig-inal GCM projections. Moreover, the spatial changes in

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

Annual KS Test for CGCM3

CNRM3 CNRM3–CDFt

Annual KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

Annual KS Test for ECHAM5

BCCR2 BCCR2–CDFt

Annual KS Test for BCCR2

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

JJAS KS Test for CGCM3

CNRM3 CNRM3–CDFt

JJAS KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

JJAS KS Test for ECHAM5

BCCR2 BCCR2–CDFt

JJAS KS Test for BCCR2

Figure 6. Kolmogorov–Smirnov statistics for original (left box-plots) and downscaled (right box-plots) GCMs surface temperatures whencompared to IMD observation data over the 1986–1999 period for the whole of southern India, with calibration of the CDF-t over 1971–1985.Statistical scores for the full year and the JJAS period are presented in the top and bottom panels, respectively. This figure is available in colour

online at wileyonlinelibrary.com/journal/joc

JJAS rainfall obtained for individual GCM downscaleddata (Figure 12 bottom panels) are systematically compa-rable to these from their corresponding raw GCM outputs.The amplitudes of these variations however appear to beenhanced for the resulting high resolution fields, corrob-orating findings from JJAS KS scores (Figure 10), withmost pronounced differences over western coastal areasand northeastern regions of the subcontinent.

Differences in dry spells lengths PDFs between the A2and XX periods are presented for the three watersheds inFigure 13. At all locations, maximum changes are foundfor short dry spells lengths (below 5 d) with enhancedvariations for the downscaled fields when compared toraw GCM outputs. At Pandam Eru and Kudaliar, themean GCM A2 projections in JJAS are characterized bya reduction of very short dry spells (below 2 d), whiledry spells with a duration above 2 d are likely to increase.The reverse is found for the wetter South Gundal basinwith enhanced very short dry spells occurrences and areduction of longer periods without rain (above two days)during the monsoon season. Interestingly similar resultsare found from mean raw GCM projections at Kudaliarand South Gundal locations with some differences interms of magnitude of these changes. However, original

GCMs data would rather suggest a slight increase ofvery short dry spells at Pandam Eru contrasting with thereduction suggested by the resulting high resolution fields

4.2. Projected surface temperatures

Surface temperature variations associated with theserainfall regime changes are examined here in order to givefurther description of A2 scenario projections consideredin this study. As mentioned in Section 2.1, due to theavailability of daily surface temperatures data for the2046–2065 period, only four GCMs will be used inthis part (MIROCMR daily surface temperatures beingunavailable from PCMDI archives at the time of thisstudy).

Similarly to Figure 10, KS diagnostics characterizesurface temperatures evolution for both the full annualcycle and the monsoon season for all GCMs (Figure 14).As shown by their respective median, a signal with asimilar amplitude is recovered from the original anddownscaled GCMs data between the contemporary periodand the projected A2 scenario, downscaled median valuesbeing a little lower than for raw GCMs except forECHAM5. A slightly higher spatial dispersion is foundfor CGCM3, ECHAM5 and BCCR2 downscaled fields

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

J F M A M J J A S O N D20

25

30

35

Mea

n S

urfa

ce T

empe

ratu

re (

 °

C)

15

20

25

30

40

35

15

20

25

30

40

35

Mea

n S

urfa

ce T

empe

ratu

re (

 °

C)

Mea

n S

urfa

ce T

empe

ratu

re (

 °

C)

Months

J F M A M J J A S O N D

Months

J F M A M J J A S O N D

Months

J F M A M J J A S O N D

Months

J F M A M J J A S O N D

Months

(1986–1999) Mean GCMSurface Temperature@ Pandam Eru

20

25

30

35

20

25

30

35

20

25

30

35

(1986–1999) Mean GCM–CDFtSurface Temperature@ Pandam Eru

(1986–1999) Mean GCMSurface Temperature@ Kudaliar

(1986–1999) Mean GCM–CDFtSurface Temperature@ Kudaliar

(1986–1999) Mean GCMSurface Temperature@ South Gundal

(1986–1999) Mean GCM–CDFtSurface Temperature@ South Gundal

J F M A M J J A S O N D

Months

IMD

CNRM

CGCM

ECHAM

BCCR

Figure 7. Mean seasonal cycle (in °C) at Pandam Eru (top panels), Kudaliar (middle panels) and South Gundal (bottom panels) locations fororiginal (left) and downscaled (right) GCMs surface temperatures together with IMD observation (thick blue line) over the 1986–1999 period,

with calibration of the CDF-t over 1971–1985. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

compared to GCM outputs, but overall the KS scores areroughly of the same order for both the full year and themonsoon season. Such results would suggest very similarevolutions between the A2 and XX periods for large-scaleGCMs and downscaled surface temperatures.

Mean surface temperatures seasonal differences fromthe XX to the A2 periods with the CDF-t method are

compared in Figure 15 with raw GCMs. Both CDF-t downscaled fields and raw GCMs data lead to quitesimilar conclusions in terms of GCMs ensemble meansat least. Despite the dispersion within all GCM projec-tions, GCMs ensemble mean surface temperature changesappear to be more pronounced over arid and semi-arid basins (Pandam Eru and Kudaliar respectively) than

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

Figure 8. Mean June to September surface temperatures (in °C) for original (top panels) and downscaled (bottom panels) GCMs data togetherwith IMD observation (top right) over the 1986–1999 period for the whole of southern India, with calibration of the CDF-t over 1971–1985.

This figure is available in colour online at wileyonlinelibrary.com/journal/joc

72°E 75°E 78°E 81°E 72°E 75°E 78°E 81°E 72°E 75°E 78°E 81°E 72°E 75°E 78°E 81°E72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

9°N

12°N

15°N

18°N

9°N

12°N

15°N

18°N

9°N

12°N

15°N

18°N

9°N

12°N

15°N

18°N

JJAS (86–99)minus (71–85) CGCM3

JJAS (86–99)minus (71–85) CNRM3

JJAS (86–99)minus (71–85) ECHAM5

JJAS (86–99)minus (71–85) BCCR2

JJAS (86–99)minus (71–85) IMD

0

0.5

1

1.5

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

Figure 9. Mean JJAS surface temperatures differences (in °C) between the validation (1986–1999) and calibration (1971–1985) periods for rawGCMs and IMD observations. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

for wetter environment (South Gundal). Nevertheless,maximum surface temperature changes for all water-sheds are found during the dry season, most particularlyin February–March (about 2.5 °C at Pandam Eru andKudaliar, and 2.2 °C at South Gundal). Smaller varia-tions are found during the monsoon season with min-imum changes in August (about 1.5 °C for all water-sheds). These findings are in agreement with other studiesemphasizing an increasing trend in mean annual surfacetemperatures with a more pronounced warming during thepost-monsoon and winter seasons (Bhattacharya, 2007).Moreover, it appears that arid and semi-arid areas arewhere the seasonal amplitude of surface temperaturesprojected changes would be maximum.

In Figure 16 are shown mean JJAS surface tempera-tures for original/downscaled (top panels/2nd line fromtop) GCM outputs under the A2 scenario (2046–2065)and their respective differences (3rd/4th lines) with orig-inal/downscaled GCMs data for the historical period(1980–1999). Again, there is no reference to whichdownscaled fields can be compared to for the projectedscenario. Nevertheless, in regards to the mean JJASclimatology from IMD observations over the contem-porary period (see Figure 8), the patterns found in allGCMs downscaled data is far more coherent than rawGCMs surface temperatures, in particular regarding the

latitudinal gradient along the west coast and the merid-ional gradient inland. Concerning surface temperaturechanges between the A2 and XX periods, similar pat-terns characterize both original and downscaled data.The amplitude of these projected changes are compa-rable between large-scale and downscaled data, exceptfor ECHAM5 for which downscaled changes are slightlygreater, agreeing with the findings from Figure 14 dis-cussed previously. ECHAM5 projections depict most pro-nounced changes (up to 3 °C) over northern regions ofthe subcontinent while the other GCMs suggest maxi-mum surface temperature increases (from 1 to 2 °C) forsouthern regions. Interestingly, for CGCM3 and BCCR2,regions of maximum surface temperatures changes inJJAS correspond approximately to areas of maximummonsoon rainfall increases. Recent studies have empha-sized, from multi-model projections, the intensificationof different pressure systems at play over the regionand enhanced moisture advection from the oceans inland(Kripalani et al., 2007). This could suggest a possibleincrease of local convection in these two GCM pro-jections. However, the possible linkages between rain-fall and surface temperatures changes for CNRM3 andECHAM5 are less clear from the short diagnostics donein this paper and deeper research is needed in order togive further elements of description relative to the pro-cesses at play.

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

Annual KS Test for CGCM3

CNRM3 CNRM3–CDFt

Annual KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

Annual KS Test for ECHAM5

BCCR2 BCCR2–CDFt

Annual KS Test for BCCR2

MIROCMR MIROCMR–CDFt

Annual KS Test for MIROCMR

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

JJAS KS Test for CGCM3

CNRM3 CNRM3–CDFt

JJAS KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

JJASl KS Test for ECHAM5

BCCR2 BCCR2–CDFt

JJAS KS Test for BCCR2

MIROCMR MIROCMR–CDFt

JJAS KS Test for MIROCMR

Figure 10. Kolmogorov–Smirnov statistics between original (left box-plots) and downscaled (right box-plots) GCMs rainfall for the A2 scenario(2046–2065) and the 1971–1999 calibration period for the whole of southern India. Scores for the full year and the JJAS period are presented

in the top and bottom panels, respectively. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

5. Discussion and conclusions

In order to downscale GCM projections over south-ern India for the whole annual cycle, the CDF-tmethod (Michelangeli et al., 2009) has been applied tomonthly chronicles of daily large-scale rainfall and sur-face temperatures. First, CDF-t has been validated onthe 1986–1999 period and compared to historical IMDobservations. In terms of KS statistics, resulting local-scale fields exhibit substantial improvements in com-parison to original GCM outputs regarding distributioncharacteristics but also mean seasonal cycle and monsoonmeans for both precipitation and surface temperatures.Then, the CDF-t method has been applied to GCMs cli-mate simulations of the 21st century under the SRES A2scenario. Resulting high resolution fields have been com-pared to original GCM outputs at different locations (arid,semi-arid and wetter environment) where both lead tosimilar conclusions. Concerning precipitation, the resultsshow a substantial increase of rainfall in particular duringthe monsoon season and for semi-arid and wetter climaticzones (from about 15 to 50%) while winter precipita-tion are generally reduced (maximum decrease of about50–80% for wetter climatic regions) in accordance withprevious findings (Rupa Kumar et al., 2006; Kripalaniet al., 2007; Raje and Mujumdar, 2009). These changes

are accompanied by increases in surface temperaturesmost pronounced during the post-monsoon (up to 2.5 °C)and winter season at all locations also agreeing withearlier studies (Rupa Kumar et al., 2006; Bhattacharya,2007).

This method was used to provide local-scale climaticvariables for impact studies (hydrological and socio-economical) at basin scale over southern India, a water-stressed region where the impacts of global changes aredue to increase significantly (Kundzewicz et al., 2007).Most statistical downscaling studies generally focus ona single season (for example JJAS) and downscaledprojections from different GCMs for the full year arerarely documented. The monthly approach chosen heregave better results than for the full year (not shown)and supports findings from other studies regarding theneed to seasonalize SDMs for better projected local-scaleestimates (Tripathi et al., 2006).

The same non-parametric approach was chosen in thisstudy to downscale separately rainfall and surface tem-peratures. However, in the case of rainfall, the ‘no- pre-cipitation’ occurrences (which are particularly importantduring the dry season) may need more developments.Days without precipitation are generally not adequatelyrepresented in GCM outputs, and this could affect the

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

J F M A M J J A S O N D

−3

−2

−1

0

1

2

3

4

5

6

−3

−2

−1

0

1

2

3

4

5

6

−3

−2

−1

0

1

2

3

4

5

6

Mea

n R

ainf

all C

hang

es (

mm

/day

)

Months

A2 minus XXGCM–CDFt @ Pandam Eru

J F M A M J J A S O N D

Months

A2 minus XXGCM–CDFt @ Kudaliar

J F M A M J J A S O N D

Months

A2 minus XXGCM–CDFt @ South Gundal

CNRM CDFt CGCM CDFt ECHAM CDFt BCCR CDFt

MIROCMR CDFt GCM CDFts mean GCM Deltas mean

Figure 11. Mean seasonal rainfall cycle changes (in mm d−1) at Pandam Eru (left panel), Kudaliar (middle panel) and South Gundal (right panel)locations between the A2 (2046–2065) and XX (1980–1999) periods seen by CDF-t compared to results from raw GCMs outputs. This figure

is available in colour online at wileyonlinelibrary.com/journal/joc

*Pa

*Ku

*Gu

A2 minus XX CDF–t A2 minus XX CDF–t A2 minus XX CDF–t A2 minus XX CDF–t A2 minus XX CDF–t

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

*Pa

*Ku

*Gu

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

*Pa

*Ku

*Gu

A2 minus XX GCM

*Pa

*Ku

*Gu

A2 minus XX GCM

*Pa

*Ku

*Gu

A2 minus XX GCM

*Pa

*Ku

*Gu

A2 minus XX GCM

*Pa

*Ku

*Gu

A2 minus XX GCM

−2

−1

0

1

2

Figure 12. Mean June to September original (top panels) and downscaled (second line) GCMs rainfall for the A2 scenario (2046–2065) oversouthern India as well as the original GCMs (third line) and CDF-t projected (bottom panels) changes (in mm d−1) when compared to the

historical XX period (1980–1999). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

CDF-t formulation used for downscaling rainfall. Asshown in Section 3.1, dry days are better representedin downscaled precipitation than in GCMs data whencompared to observation. Nevertheless, for all GCMs

CDF-t seems to perform better, in terms of dry days andprecipitated amount, during the monsoon season with few‘no-precipitation’ occurrences, than for the whole year.To address this issue, a next step would be to use more

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

Figure 13. Differences in dry spells length PDFs (in %) at Pandam Eru (left), Kudaliar (centre) and South Gundal (right) locations for original(dashed coloured lines) and downscaled (thick coloured lines) GCMs data between the A2 (2046–2065) and XX (1981–1999) 20 year periods.

This figure is available in colour online at wileyonlinelibrary.com/journal/joc

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

Annual KS Test for CGCM3

CNRM3 CNRM3–CDFt

Annual KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

Annual KS Test for ECHAM5

BCCR2 BCCR2–CDFt

Annual KS Test for BCCR2

CGCM3 CGCM3–CDFt0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

Kol

mog

orov

–Sm

irnov

JJAS KS Test for CGCM3

CNRM3 CNRM3–CDFt

JJAS KS Test for CNRM3

ECHAM5 ECHAM5–CDFt

JJAS KS Test for ECHAM5

BCCR2 BCCR2–CDFt

JJAS KS Test for BCCR2

Figure 14. Same as Figure 10 but for surface temperatures. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

relevant distribution models for precipitation through aparametric approach as it has been done with other SDMsusing Gamma or mixed distributions for example (Vracand Naveau, 2007).

Nevertheless, CDF-t proved to be an interesting andefficient statistical tool, offering substantial perspectivesin terms of downscaling and climate change impactstudies at local-scale, with the low computational cost

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

PROBABILISTIC DOWNSCALING OF GCM SCENARIOS OVER SOUTHERN INDIA

J F M A M J J A S O N D0

0.5

1

1.5

2

2.5

3

3.5

0

0.5

1

1.5

2

2.5

3

3.5

0

0.5

1

1.5

2

2.5

3

3.5

Mea

n S

urfa

ce T

empe

ratu

re C

hang

es (

 °

C)

Months

A2 minus XXJJAS GCM–CDFt @ Pandam Eru

J F M A M J J A S O N D

Months

A2 minus XXJJAS GCM–CDFt @ Kudaliar

J F M A M J J A S O N D

Months

A2 minus XXJJAS GCM–CDFt @ South Gundal

CNRM–CDFt CGCM–CDFt ECHAM–CDFt BCCR–CDFt

GCM–CDFt mean GCM–Deltas mean

Figure 15. Same as Figure 11 but for surface temperatures (in °C). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

*Pa

*Ku

*Gu

A2 minusXX JJAS CGCM3 CDF–t

*Pa

*Ku

*Gu

A2 minusXX JJAS CNRM3 CDF–t

*Pa

*Ku

*Gu

A2 minusXX JJAS ECHAM5 CDF–t

*Pa

*Ku

*Gu

A2 minusXX JJAS BCCR2 CDF–t

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

72°E 75°E 78°E 81°E

9°N

12°N

15°N

18°N

*Pa

*Ku

*Gu

A2 minusXX JJAS CGCM3

*Pa

*Ku

*Gu

A2 minusXX JJAS CNRM3

*Pa

*Ku

*Gu

A2 minusXX JJAS ECHAM5

*Pa

*Ku

*Gu

A2 minusXX JJAS BCCR2

0

1

2

3

Figure 16. Same as Figure 12 but for surface temperatures (in °C). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

and flexibility of this approach making it even moreattractive.

Finally, deeper research is needed to give furtherelements of description regarding future climatic pro-jections over southern India and the processes involved.

First the period of study regarding these projections islimited to 20 years and could be extended. It wouldalso be relevant to compare our results to projectedfields obtained through numerous statistical downscalingapproaches (other than the single delta method) in order

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)

N. VIGAUD et al.

to examine the uncertainties linked to the method itself.Likewise the use of ensemble runs for the different GCMswould help identify these uncertainties associated withthe GCM outputs to better characterize potential climaticchanges over the subcontinent. Given these limitations,the regional projections presented in this study shouldthus be considered with caution as the aim here is toprimarily illustrate the use and capability of CDF-t fordownscaling large-scale GCMs climate variables such asrainfall and surface temperatures over the Indian region.However, this paper corresponds to one more step in thatdirection helping to document medium-range future cli-matic scenarios (2040–2060), these horizons being cru-cial for local adaptation strategies.

Acknowledgements

This study has been supported by the Agence Nationalepour la Recherche (ANR) through the VMCS pro-gram (project SHIVA contract ANR-08-VULN-010-01)and the Bureau de Recherches Geologiques et Minieres(BRGM). The authors thank the SHIVA partners fortheir contribution. Mathieu Vrac was partially fundedby the GIS-REGYNA Project. IMD observations wereobtained from the Institute of Meteorology Depart-ment of India. GCM outputs from the IPCC AR4exercise were downloaded from the PCMDI server(http://www.pcmdi.llnl.gov/ipcc/model documentation/).The downscaling has been realized with the ‘CDF-t’ R package freely available on the CRAN website(http://cran.r–project.org/).

References

Anandhi A, Srivanas V, Kumar DN, Nanjundiah RS. 2009. Role ofpredictors in downscaling surface temperature to river basin for IPCCSRES scenarios using support vector machine. International Journalof Climatology 29: 583–603.

Anandhi A, Srivanas V, Nanjundiah RS, Kumar DN. 2008. Downscal-ing precipitation to river basin in India for IPCC SRES scenariosusing support vector machine. International Journal of Climatology28: 401–420.

Bhattacharya S. 2007. Lessons learnt for vulnerability and adaptationassessment from India’s first national communication, vol 7, BASICEU Project.

Cruz R, Harasawa M, Wu S, Anokhin Y, Punsalmaa B, Honda Y,Jafari M, Li C, Ninh NH. 2007. Asia Climate Change 2007: Impacts,Adaptation and Vulnerability. Contribution of Working Group II tothe Fourth Assessment. Cambridge University Press: UK, 469–506.

Darling D. 1957. The Kolmogorov-Smirnov, Cramer-von Mises tests.Annals of Mathematical Statistics 28: 823–838.

Flato GM, Boer GJ, Lee WG, McFarlane NA, Ramsden D, ReaderMC, Weaver AJ. 2000. The Canadian Centre for Climate Modelingand Analysis of Global Coupled Model and its climate, ClimateDynamics 16: 451–467.

Furevik T, Bentsen M, Drange H, Kindem IKT, Kvamsto NG,Sorteberg A. 2003. Description and evaluation of the Bergen ClimateModel: ARPEGE coupled with MICOM, Climate Dynamics 21:27–51.

Ghosh S, Mujumdar P. 2007. Nonparametric methods for modellingGCM scenario uncertainty in drought assessment. Water Research43: W07405, 19 pp., DOI: 10.1029/2006WR005351.

Ghosh S, Mujumdar P. 2008. Statistical downscaling of GCM

simulations to streamflow using relevance vector machine. Advancesin Water Resources 31: 132–146.

Hughes J, Guttorp P, Charles S. 1999. A non-homogeneous hiddenMarkov model for precipitation occurrence. Applied Statistics 48:15–30.

Jungclaus JH, Keenlyside N, Botzet M, Haak H, Luo JJ, Latif M,Marotzke J, Mikolajewicz U, Roeckner E. 2006. Ocean circulationand tropical variability in the coupled model ECHAM5/MPIOM,Journal of Climate, 19: 3952–3972.

Kripalani R, Oh J, Kulkarni A, Chaudhari H. 2007. South Asiansummer monsoon precipitation variability: coupled climate modelsimulations and projections under IPCC AR4. Theoretical andApplied Climatology 90: 133–159.

Kundzewicz Z, Mata L, Arnell N, D”ll P, Kabat P, Jim’nez B,Miller K, Oki T, Sen Z, Shiklomanov I. 2007. Freshwater Resourcesand their Management. Cambridge University Press: UK, 173–210.

Michelangeli P, Vrac M, Loukos H. 2009. Probabilistic downscal-ing approaches: application to wind cumulative distribution func-tions. Geophysical Research Letters 36: L11708, 6 pp., DOI:10.1029/2009GL038401.

Mujumdar P, Ghosh S. 2008. Modelling GCM and scenario uncer-tainties using a possibilistic approach: application to the MahanadiRiver. Water Resources Research 44: W06407, 15 pp., DOI: 10.1029/2007WR006137.

Nakicenovic N, Davidson O, Davis G, Grbler A, Kram T, RovereELL, Metz B, Morita T, Pepper W, Pitcher H, Sankovski A, ShuklaP, Swart R, Watson R, Dadi Z (eds). 2000. In Special Report onEmissions Scenarios: A Special Report of Working Group III of theIntergovernmental Panel on Climate Change. Cambridge UniversityPress: 599.

Paeth H, Scholten A, Friederichs P, Hense A. 2008. Uncertaintiesin climate change prediction: El Nino-Southern Oscillation andmonsoons. Global and Planetary Change 60: 265–288.

Prudhomme C, Reynard N, Crooks S. 2002. Downscaling of globalclimate models for flood frequency analysis: where are we now?Hydrological Processes 16: 1137–1150.

Raje D, Mujumdar P. 2009. A conditional field-based downscalingmethod for assessment of climate change impact on multisite dailyprecipitation in the Mahanadi basin. Water Resources Research 45:W10404, 20 pp., DOI: 10.1029/2008WR007487.

Rajeevan M, Bhate J. 2008. A high resolution gridded rainfall dataset(1971–2005) for mesoscale meteorological studies. vol 9, NCCResearch Report.

Rupa Kumar K, Sahai A, Kumar KK, Patwardhan S, Mishra P,Revadekar J, Kamala K, Pant G. 2006. High-resolution climatechange scenarios for India for the 21st century. Current Science90(3): 334–345.

Salas-Melia D, Chauvin F, Deque M, Douville H, Gueremy JF,Marquet P, Planton S, Royer JF, Tyteca S. 2006. Description andvalidation of the CNRM-CM3 global coupled model, ClimateDynamics (in press).

Srivastava A, Rajeevan M, Kshirsagar S. 2008. Development of a highresolution daily gridded temperature dataset (1969–2005) for theIndian region. vol 8, NCC Research Report.

Tripathi S, Srinivas V, Nanjundiah SR. 2006. Downscaling ofprecipitation for climate change scenarios: a support vector machineapproach. Journal of Hydrology 330: 621–640.

Vrac M, Naveau P. 2007. Stochastic downscaling of precipitation: fromdry events to heavy rainfalls. Water Resources Research 43: W07402,13 pp., DOI: 10.1029/2006WR005308.

Vrac M, Stein M, Hayhoe K. 2007. Statistical downscaling ofprecipitation through non-homogeneous stochastic weather typing.Climate Research 34: 169–184, DOI: 10.3354/cr00696.

Wilby R, Charles S, Zorita E, Timbal B, Whetton P, Mearns L. 2004.Guidelines for use of climate scenarios developed from statisticaldownscaling methods. Supporting Material of the IPCC, 1–27.

Wilks D, Wilby R. 1999. The weather generation game: a reviewof stochastic weather model. Progress in Physical Geography 23:329–357.

Wood A, Sridhar V, Lettenmaier D. 2004. Hydrologic implications ofdynamical and statistical approaches to downscaling climate modeloutputs. Climate Change 62: 189–216.

Copyright 2012 Royal Meteorological Society Int. J. Climatol. (2012)


Recommended