Validation of a hybrid coordinate ocean model for the ...€¦ · ity. Anderson et al6 modelled the SC during the southwest monsoon in a 16-level model and McCreary and Kundu 7 used

25

AUTHORSrsquo BIOGRAPHIESMary Swapna George is a PhD student at the Mohn-SverdrupCenter for Global Ocean Studies and OperationalOceanographyNansen Environmental and Remote SensingCenterThe main aim of the PhD is to develop and validate anumerical ocean model for the Indian Ocean She obtained herMasterrsquos degree from Cochin University of Science andTechnology India in 2001

Dr Laurent Bertino is deputy leader of the Mohn-SverdrupCenter for Global Ocean Studies and OperationalOceanographyNansen Environmental and Remote SensingCenter a position he has held since 2004 Prior to joining theNansen Center he obtained a PhD in geostatistics applied todata assimilation from Ecole des Mines de Paris

Prof Ola M Johannessen is the founding director and theChairman of the Board of the Nansen Environmental and RemoteSensing Center He is an oceanographer and has a wide interest inocean studies from the Arctic Ocean to the Indian Ocean

Dr Annette Samuelsen is a scientist at the Mohn-SverdrupCenter for Global Ocean Studies and OperationalOceanographyNansen Environmental and Remote SensingCenter She obtained a PhD in physical oceanography fromFlorida State University in 2005 and has worked at the NansenCenter since then

INTRODUCTION

The Indian Ocean is the only ocean bounded by landin the north and is divided by the Indian subconti-nent into two main basins in the north the ArabianSea (AS) and the Bay of Bengal (BoB) A major

impact of this geography is the intense forcing on the north-ern Indian Ocean by the annually reversing monsoon windsThis reversal of the seasonal winds also leads to the reversalin the circulation patterns and changes in the hydrographicfeatures of the surface waters in the northern parts of theregion The effect is more prominent in the AS especially

Validation of a hybrid coordinateocean model for the Indian Ocean

MS George Dr L Bertino Prof OM Johannessen and Dr A SamuelsenMohn-Sverdrup Center for Global Ocean Studies and Operational Oceanography

Nansen Environmental and Remote Sensing Center Bergen Norway

An eddy-permitting HYbrid Coordinate Ocean Model (HYCOM) configured for the IndianOcean has been validated using both in-situ and satellite observations The present workfocuses on a detailed study of the modelrsquos capability to simulate the major surface andsubsurface variables realistically Weekly data from the model for eight years from 1994 to2001 are used for the evaluation of the surface dataThe model simulation of the circulationpatterns in the Indian Ocean for both the monsoon seasons and the transition periodsmatches well with the observations Comparisons between model and satellite observationsfor the sea surface temperature (SST) patterns and its temporal evolution showed that themodel produces realistic SSTsThe sea level anomalies (SLA) from the model compared withthose from the altimeter data confirmed that the model is in good agreement with theobserved SLA A detailed comparison of results from the daily data of the model with theArgo profiles for the years from 2002 to 2004 showed that the model has a diffuse thermo-cline with warming in the subsurface waters but overall the model simulates the subsurfacetemperature and salinity patterns wellThe validation of the model indicates that the modelresults are satisfactory and that with improvements in some of the model configurations itcan be implemented in an operational forecasting system for the Indian Ocean

Validation of a hybrid coordinate ocean model for the Indian Ocean

Volume 3 No 2 2010 Journal of Operational Oceanography

Swapna_JOO_Augqxd 81310 516 PM Page 25

along its western coastal zones Thus in the earlier timesmost of the observational cruises as well as modelling studieswere focused on the monsoon seasons and the associatedreversal of Somali Current (SC) along the coast of Somaliaand Oman in the western part of the AS12

The International Indian Ocean Expedition (IIOE) from1964ndash66 carried out the first extensive study of the IndianOcean and the resulting hydrographic atlas3 started a series ofcomprehensive studies on the Indian Ocean circulationSchott and McCreary2 give an extensive review on the mon-soon circulation of the Indian Ocean based on observationstheory and modelling studies

Numerical modelling is a powerful tool to study theoceans especially in regions like the Indian Ocean where theobservational data are sparse Even though modelling studiesof the Indian Ocean are limited in number compared to otherparts of the worldrsquos oceans there has been major progress inthe past few years The SC the coastal currents along the eastand west coasts of India and the associated hydrographic fea-tures have received much attention in the past and presentMany of the earlier modelling studies of the SC implementedreduced gravity models1 4 5 to study its dynamics and variabil-ity Anderson et al6 modelled the SC during the southwestmonsoon in a 16-level model and McCreary and Kundu7 useda 212-layer model to analyse the characteristics of the SC dur-ing southwest monsoon All these modelling studies aided inunderstanding the effects of local as well as remote forcingin driving the SC during summer monsoon and the impor-tance of a slanted boundary in the models for proper gyre for-mation in the western AS

The anticyclonic eddy formations in the south eastern AS(SEAS) during northeast monsoon were analysed in detail byBruce et al89 employing a three layer reduced gravity modelforced by the Hellerman amp Rosenstein10 wind climatology Anassociated high in sea surface is developed during the north-east monsoon and a low is formed during southwest monsoonperiod The formation mechanisms of these high and lowwere examined in detail by Shankar and Shetye11 using theirreduced gravity model for the northern Indian OceanMcCreary et al12 published a comprehensive study of thedynamics thermodynamics and mixed-layer process of thecirculation of Indian Ocean simulated in a 212-layer thermo-dynamic numerical model The influences of forcing by localand remote along shore winds on the coastal currents werediscussed in this study The dynamics of the cyclonic (anticy-clonic) gyre formation east of Sri Lanka during summer (win-ter) monsoon termed as Sri Lankan Dome (Bay of BengalDome) are discussed by Vinayachandran amp Yamagata13

In recent years many of the major oceanographic featuresin the Indian Ocean area have been investigated using variousmodels The barrier layer (BL) formations in the IndianOcean have been modelled by Masson et al14 15 They used acoupled model to evaluate the effect of BL on the sea surfacetemperature (SST) and on the monsoon onset in the SEAS14

and in another study studied the BL formation in the easternequatorial Indian Ocean (EIO)15 The effect of salinity on thegeneration of the BL in the SEAS is analysed employing anocean general circulation model by Durand et al16 Numerousstudies have been done about Indian Ocean warm pool and inone recent study the mechanisms and formation of Arabian

Sea mini warm pool is discussed by Kurian andVinayachandran17 Most of the earlier models were based onreduced gravity model structure while as many of the recentmodels are based on versions of Geophysical Fluid DynamicsLaboratory (GFDL) Modular Ocean Model (MOM)

Haugen et al18 implemented the Miami IsopycnicCoordinate Model (MICOM) to examine the variability of theIndian Ocean in response to the monsoons and to study theinter-annual signals in the Indian Ocean They also discussedthe seasonal circulation and coastal upwelling along the south-west Indian coast using a high resolution nested model for theregion19 As a continuation of the Nansen Centerrsquos study usingMICOM in the Indian Ocean18 19 the HYbrid Coordinate OceanModel (HYCOM) has been implemented which allows a high-er vertical resolution in the mixed layer and hence give animproved representation of the surface layers20 contrasted toMICOM Most of the previous models developed for the IndianOcean have coarser horizontal resolutions Here the model hasbeen configured with higher resolution (model details in thenext section) in the northern Indian Ocean which is the focusarea of the authorsrsquo research The final goal is to implement aforecasting system for the Indian Ocean however at present nodata assimilation is used this is planned as the next step

Even though the models have been much improved acomplete study on the validation of the models used in theIndian Ocean are less common Haugen et al18 validatedMICOM for the Indian Ocean Kurian and Vinayachandran17

presents the results of comparisons of their model (based onMOM) with observations in the study about the AS miniwarm pool region Model results needed to be evaluated andvalidated before utilising them for analysing or addressingany scientific problem

Thus in this study the authors present the preliminaryresults from the validation of the model using in-situ and satel-lite observations To validate the model for surface featuresthe climatology of model results for eight years (1994ndash2001)are used The surface currents are compared with observationsbased on previously published results The surface tempera-tures and anomalies of sea level are compared with satelliteobservations The inter-annual variability of the surface tem-peratures and the anomalies of sea level are also compared andthe differences between the model and observations have beenquantified For the evaluation of model in the subsurface thetemperature and salinity for three years (2002ndash2004) are com-pared with the Argo profiles A detailed description of themodel configuration is given in the next section followed bythe satellite and in-situ data used for the model evaluation Theresults of validation are then discussed

MODEL DESCRIPTIONHYCOM is a hybrid coordinate model based on MICOM andis able to interchange between different vertical coordinateschemes It is a primitive equation general circulation modelwith vertical coordinates that remain isopycnic in the openstratified ocean In the weakly stratified upper ocean theisopycnal vertical coordinates smoothly transfer to z-coordi-nates This hybrid formulation gives a better emulation of thesurface mixed layer and the coastal shelf regions20 comparedto MICOM The terrain following coordinates are not used in

Journal of Operational Oceanography Volume 3 No 2 201026



the present configuration The model for the Indian Ocean isset up with horizontal resolution ranging from 14 to 42km forthe entire domain (Fig 1) The conformal mapping tool21 isused to enhance the horizontal resolution in the northernIndian Ocean Thus the resolution in the northern part of theIndian Ocean is 14 to 26km which is sufficient to resolve thelarger mesoscale features

Towards the southern part of the domain the resolution isgradually decreasing The model uses 30 vertical hybrid lay-ers The vertical mixing scheme used is the K-ProfileParameterisation (KPP) scheme of Large et al22 The topogra-phy used by the model is interpolated from the GeneralBathymetric Chart of the Oceans (GEBCO) one minuteresolution dataset (httpwwwgebconet) The model isinitialised using Generalised Digital Environmental Model(GDEM) climatology23 On the open boundaries relaxation oftemperature and salinities to the GDEM climatology isapplied For the surface temperature and salinity the model isrelaxed using 50 days time scale The model was spun up foreight years using climatological monthly means of atmos-pheric data from European Centre for Medium RangeWeather Forecasts (ECMWF) Re-Analysis (ERA 40) databefore transition to synoptic forcing

Following the spin up a 13-year model run was carriedout during 1992ndash2004 using the synoptic forcing The syn-

optic atmospheric fields such as winds surface temperatureand surface dew point temperature are taken from the ERA40 year reanalysis for the years from 1994 up to 2001 andlater the operational analysis from the European Centre forMedium Range Weather Forecasts (ECMWF) are used Theprecipitation from the climatological dataset of Legates andWillmott24 is used The Indonesian ThroughFlow (ITF)transfers warm saline waters from Pacific to the IndianOcean and has an impact on the Indian Ocean circulation andthe surface temperature and salinity distributions in theeastern EIO25

The ITF transport to the Indian Ocean alters with season andfluctuates between 4 to 12Sv25 Gordon26 reports a total transportof 8ndash14Sv and hence an average of ~10Sv of ITF (with interan-nual modulation by ENSO) through the different passages in theIndonesian seas to the Indian Ocean Lately Wijffles et al27

estimated the mean ITF transport to be 89 plusmn 17Sv based on 20years of XBT data Hence in the present model the ITF has beenadded as a constant 10Sv barotropic flux between Indonesia andAustralia The flux is maintained constant with depth Themodel does not include tides

From the 13-year model run the results from the years1994ndash2004 are presented here Weekly averages werearchived from 1994ndash2001 From 2002 daily files were storedfor comparison with Argo float data (which are available

27Volume 3 No 2 2010 Journal of Operational Oceanography


Fig 1The model domain with the horizontalresolution in kmThe boxes marked are thedifferent regions chosen for the study theArabian Sea (AS) in black the Bay of Bengal(BoB) in red and the Equatorial IndianOcean (EIO) in blue


from 2002 onwards) For some of the analysis the results areaveraged over the three major water bodies in the northernIndian Ocean the AS BoB and the EIO (Fig 1) The AS boxis from 56ordmEndash75ordmE and 8ordmNndash21ordmN the BoB from 80ordmEndash100ordmEand 8ordmNndash21ordmN and the EIO from 50ordmEndash95ordmE and 7ordmSndash7ordmN(Fig 1)

OBSERVATIONAL DATAFor validating the model several observational datasets wereused The results discussed in this paper are validation of themodel surface currents the comparisons of sea surfacetemperature (SST) from the model with that derived fromsatellites sea level anomaly (SLA) comparisons with altime-ter data and the comparisons of temperature and salinitystructure of the 1000m water column with the data obtainedfrom the Argo profiling floats

The dataset used for the SST comparison is the NOAAOptimum Interpolation (OI) SST V2 (version 2) data(httpwwwcdcnoaagov) The dataset renders weekly OI SSTanalysis produced on a one-degree grid The analysis uses bothin-situ and satellite derived SSTs with the satellite data adjust-ed for biases28 The weekly data are centred on Wednesday andare available from 1989 to the present The gridded SLA dataare obtained from the multimission altimeter products ofSsaltoDuacs system (httpwwwavisooceanobscom) Thedataset is the merged gridded sea surface heights computed withrespect to a seven-year mean from the multiple altimeter mis-sions of TopexPoseidon ERS-12 + Jason-1 Envisat Themaps of SLA available for weekly intervals are used in thisstudy The gridded data are provided in delayed mode with ahigh horizontal resolution of 13 deg on a Mercator grid Thedata are available for the time period of October 1992 to thepresent From both datasets the SST and SLA data from 1994to 2001 are used for the comparisons

Argo is a global drifting array of temperature-salinityprofiling floats which started operating in 2000 As of now3000 floats have been deployed in the worldrsquos oceans and inthe Indian Ocean more than 600 floats have been deployedsince the end of 2002 The Argo programme is an interna-tional collaboration and is a part of the Global ClimateObserving SystemGlobal Ocean Observing System(GCOSGOOS) The Argo floats are designed to collect highquality temperature and salinity profiles of the upper 2000mof the worldrsquos oceans Thus the Argo data makes it feasibleto carry out a three-dimensional validation of model resultsThe Argo dataset used for the present study are the griddeddata available from the Indian National Centre for OceanInformation Services (INCOIS) live access server(httplasincoisgovinlasgetUIdo)29 The data from years2002 to 2004 are used for the daily comparisons

STATISTICAL ANALYSISIn addition to the qualitative analysis some quantitativeanalysis has also been carried out for the validation The CostFunction (CF)30 is a non-dimensional value which quantifiesthe difference between the model values and satellite datathus indicating the goodness of fit between the two datasetsIt is given by

where N is the total number of observations n is the nth com-parison D is observed value M is the model value and

is the standard deviation of the observations denotes themean of observations The performance criteria used here isCF lt1 very good 1ndash2 good 2ndash3 reasonable gt3 poor The CFvalues for the SST and SLA are calculated for the northernIndian Ocean for the present study

VALIDATION RESULTS AND DISCUSSIONSurface circulationIn the first part of the validation the main focus is on the surfacecurrent patterns in response to the monsoon system Over thenorthern Indian Ocean winds blow from southwest duringMayndashSeptember and from northeast during NovemberndashFebruary In this paper we refer to MayndashSeptember asSouthWest (SW) period or summer monsoon and NovemberndashFebruary as NorthEast (NE) period or winter monsoon Thewinds are stronger during the summer monsoon than the wintermonsoon The transition periods with weak winds are in themonths of MarchndashApril and October This seasonal reversal ofwinds has great influence on the circulation pattern of the IndianOcean in particular the northern Indian Ocean Fig 2 shows theclimatology (averaged for the eight years from 1994ndash2001) ofthe surface circulation simulated by the model for the months ofJanuary (NE period) April (transition period) July (SW period)and October (transition period)

The South Equatorial Current (SEC) is the westwardflow present within the latitudes 12ndash25ordmS throughout theyear2 The observations from ship drifts31 and buoy data32

reports the speed of SEC as 04ms and 03ms respectivelyThe model simulates the SEC throughout the year within10ordmS to 15ordmS (Fig 2) The flow peaks during the summermonsoon time (Fig 2c) with the speeds reaching up to04ms which is in agreement with the observations East ofMadagascar it splits into two one branch flowing south-wards as the Southeast Madagascar Current (SEMC) Theother branch flows northwards forming the NortheastMadagascar Current (NEMC) that feeds into the EastAfrican Coastal Current (EACC)2

This branching of the SEC is clearly observed in the cir-culation patterns simulated by the model The East AfricanCoastal Current (EACC) flows northward along the Africancoast The model simulates the EACC along the coast ofAfrica (Fig 2c) During the summer monsoon along with theSEC it feeds water into the SC2 In the western part of theIndian Ocean the strong northward flowing SC is simulatedduring the summer monsoon with the monthly averagedspeeds of 17ms During the peak summer monsoon a partof the SC turns offshore at around 4ordmS and the Southern Gyre(SG) is formed while the other part continues northward andforms another gyre the Great Whirl (GW)2 Both these gyres

D

σ D nn

N

ND D= minus( )

=sum1 2

1

CF =minus

=sum1

1N

D Mn n

Dn

N

σ




are present in the model when the data for individual yearsare evaluated However in the climatology of eight years (Fig2c) these gyres are smoothed The southward flowing WestIndian Coastal Current (WICC)33 is simulated in the modelfrom May to September (Fig 2c) The model simulates theSummer Monsoon Current (SMC) flowing eastwards (Fig 2c)between 10ordmN and the equator The southwards flowingWICC also feed into the SMC In the BoB (Bay of Bengal)the East Indian Coastal Current (EICC) flows northeastwardfrom February with a fully developed phase duringMarchndashApril and later reverses its direction after the with-drawal of summer monsoon34 The poleward flow of EICC ispresent in the model from MarchndashAugust (Fig 2b amp c)

During the winter monsoon the flow along the Somali coastreverses its direction and forms a gyre with the SouthEquatorial Counter Current (SECC) (Fig 2a) The WICC flowsnorthwards during the winter monsoon35 The EACC at around4ordmS meets with the southward flowing SC and flows towardseast as the SECC The EICC in the BoB changes direction dur-ing this time and flow southwestward from November toJanuary34 (Fig 2a) The modelled Winter Monsoon Current(WMC) is situated between 8ordmN and the equator as the west-ward flowing current south of Sri Lanka A part of the flowsupplies into the WICC The WMC flows westwards and joinsthe southward flowing SC (Fig 2a)

During the transition seasons from April to June (Fig 2b)and October to December (Fig 2d) the eastward flowing

strong surface jets ndash the Wyrtki Equatorial Jet (EJ) ndash is wellsimulated The jet is formed due to the semiannual eastwardswind along the equator36 The strongest part of the jet isbetween 70ordmE and 90ordmE with averaged speeds up to 06msIn general it is concluded that the simulation of the surfacecirculation compares well with the observed surface circula-tion pattern

Sea surface temperaturesComparisons with SST (Fig 3) show that the model simulatesthe SST patterns for the month of January (Fig 3a e i) wellIn general the colder waters in the northeast AS and in thenorthern BoB are reproduced with approximately the sametemperatures During April (Fig 3b f j) the warm waters thatspread throughout the AS are clearly seen in the model Thetropical Indian Ocean surface waters are warm throughout theyear The building up of Indian Ocean Warm Pool (IOWP) inthe southwest AS from the month of February to May37 and itscollapse after the onset of summer monsoon in April and Julyare simulated (Fig 3b amp c) The region with surface tempera-tures exceeding 30ordmC in the southeast AS off the west coastof India is the warmest region ndash mini warm pool ndash during themonths of April and May17 38 39 This can be also seen in themodelled SST pattern (Fig 3b)

The coastal areas of the AS especially the Arabian coastare the major zones of upwelling during the summer mon-soon5 40 Here upwelling is observed to extend up to 400km



Fig 2The surface currents simulated by HYCOM (average of eight years 1994ndash2001)The results for both the monsoon months(January and July) and the transition months (April and October) are shownThe speeds in ms are denoted in the colour barEvery 6th vector in X and Y direction is plotted


offshore and run parallel to the coast The upwelling replacesthe warm surface waters with relatively colder subsurfacewater and the upwelling zones associated with lower SSTsalong the western coast of the AS can also be seen in themodel and satellite images (Fig 3c g) During July the colderupwelling waters are found in the model (Fig 3c) and the minicold pool which appears during the summer monsoon seasonnear the southern tip of India41 and its intrusion into the BoBis clearly visible in the model but not observed in the satellitemeasurements (Fig 3g)

The intrusion of warmer waters along the eastern equato-rial ocean during July is not as apparent in the model DuringOctober much of the equatorial waters are warmer than themodel by 1degC (Fig 3d amp h) The model is generally colder inthe equatorial region especially in the region that is influ-enced by the ITF waters (Fig 3indashl) The temperature differ-ences are typically around 1degC In the northern part of AS themodel is warmer than the satellite by up to 15degC In generalthe modelled SST patterns in the Indian Ocean region are ingood agreement with those of the satellite measurements

A time series of the temperature evolution (Fig 4) wasmade for the AS and BoB and the EIO The differencesbetween the observed and modelled SST for the three differ-ent regions are shown in Fig 4d For this both the model dataand the satellite data were averaged over the AS (56ordmEndash75E

8ordmNndash21ordmN) the BoB (80ordmEndash100ordmE 8ordmNndash21ordmN) and the EIO(50ordmEndash95ordmE 7ordmSndash7ordmN) (Figs 1 amp 4) for the time period fromJanuary 1994 to December 2001 The temporal variations ofthe SSTs in the model for the AS region match with that of thesatellite derived SSTs however with slightly higher tempera-tures in particular during the summer monsoon The meandifference between the satellite observations and the model is ndash05degC The SST plots for the BoB region show that themodel is in good agreement with the satellite SSTsThroughout the eight years the model is able to simulate theseasonal pattern of SST clearly but again with the largestdifference during the summer monsoon after 1997 For thisregion the mean difference between the satellite data andmodel is around +01degC The SST patterns for the EIO showthat the model performs well during 1994ndash1997 after whichthe differences increase up to 06degC and the mean differencebetween the observations and the model is +045degC Theincrease in the differences both in the BoB and equatorialregion is prominent after the extreme El Nintildeo event and thedipole mode42 during the 1997ndash1998 period There was ananomalous Indian Ocean warming43 44 reported during thistime Comparatively higher differences between observationsand model results could be explained by the fact that the ITFis constant in the model where as in reality the ITF hasseasonal variations The inter-annual variability of the ITF is



Fig 3 Comparison of SST (degC) average of the eight years (1994ndash2001) for the months of January April July and October SSTfrom the model (top) satellite (middle) and the difference between the results (satellite minus model) (bottom) are shown


correlated to the ENSO45 and Indian Ocean Dipole (IOD)46

the former was very strong in 1998 In general the model is slightly warmer than the satellite

measurements which is seen in Fig 3i and k particularly in thenorthernmost part of AS The BoB shows better agreementwith observations than the AS In both regions the modelreproduces the seasonal patterns well Fig 4b shows a consid-erable change in the pattern for the BoB after the El Nintildeo event

and the IOD of 1997ndash1998 Before 1997 the model is warmerthan the satellite measurements during the monsoon time andafter the monsoon of 1997 the model is colder than the obser-vations The results from the EIO region show that the modelis slightly colder than the satellite observations In this regionthere is also a sudden increase in the differences in the yearsfollowing 1997 In general the model agrees with the observedSST in all the three regions for the eight year simulations



Fig 5 Statistics of model performance for SST (a) the mean of differences (degC) between observations and the model (satelliteminus model) (b) standard deviations of the differences (degC) and (c) cost functionThe values are averaged for the eight yearsfrom 1994ndash2001

Fig 4Time series plots ofthe SSTs (degC) for theperiod 1994ndash2001 fromHYCOM (red) and thesatellite (black) for theregions of (a) ArabianSea (b) Bay of Bengaland (c) Equatorial IndianOceanThe difference(satellite minus model)for the three regions isshown in (d)


Fig 6 Comparison of SLA (cm) average of eight years (1994ndash2001) from HYCOM (top) and altimeter measurements (middle)Blue denotes negative anomaly and orange denotes positive anomalyThe contour interval is 5 cmThe figures at the bottomshow the difference in SLA between observations and the model (satellite minus model)



Along with the above comparisons a quantitative analysisof the SST is also carried out by calculating the CF for thenorthern Indian Ocean The mean difference between theobservation and the simulated values and the standard devia-tion of these differences are also calculated Fig 5 shows theresults The values are averaged for the eight years (1994 to2001) of the weekly simulation

The modelled surface temperatures are warmer in the ASthan the satellite SST specially towards the northern coast(Fig 5a) In the BoB and in the EIO region the mean differ-ence in temperature is less than 05degC Towards the easternside of the EIO the differences reach up to 1degC In general themean temperature differences do not exceed 1degC except forthe northern coastal regions of the AS Standard deviationsfrom the mean (Fig 5b) shows that for most parts of the north-ern Indian Ocean the deviations are less than 05degC Towardsthe western part of the AS there is much more variabilitycompared to the rest of the Indian Ocean This could be attrib-uted to the seasonal changes in SST along the coast wherethe seasonal upwelling of colder waters occur during the sum-mer monsoon40 and also to the mesoscale activities in thearea Higher standard deviations are also seen in the southwest coast of India (Fig 5b) which is also a region ofupwelling19 47 during the monsoon time The CF values in thenorthern Indian Ocean (Fig 5c) are less than one (CF lt 1 is

very good) for a major part of the region Only a part of theEIO has CF values higher than one reaching up to 16 (CFbetween 1ndash2 is good) This could be from the slight offset inthe temperatures of the region from 1997 as seen in the timeseries (Fig 4c) The CF values indicate that the model per-formance is very good giving CF values of less than one formost of the northern Indian Ocean The values also indicatethat the BoB SSTs show the least differences from the obser-vations

Sea level anomaliesThe sea level anomalies (SLA) computed from the modelhave been compared to the altimeter data (Fig 6) The resultsare shown for the two monsoon seasons January (winter Fig6ac) and July (summer Fig 6bd) The monthly means areaveraged from eight years of model results Since the modelis not eddy resolving the altimeter data are smoothed usingthe boxcar smoothing method in the Ferret software Theradius of the moving average is two degrees During Januarythere is low sea level near the Arabian coast associated withthe upwelling33 which is clearly present in the model TheSLA in the SEAS has a characteristic high during the wintermonsoon which is known as the Laccadive high11

This high sea level pattern is simulated well by themodel with comparable amplitudes to the altimeter data




The high sea level along 12degS of the eastern Indian Ocean isalso seen clearly in the model The SLA along the centralBoB in the model is not as high as in the satellite measure-ments During summer monsoon there is a low sea levelpattern observed in the southeastern AS11 This low sea levelis also reproduced well by the model (Fig 6b) The high sealevel associated with the summer monsoon (July) in the

African coast is also present in the model In generalHYCOM is able to simulate both the spatial and temporalsea level variations well and with comparable amplitude tothe altimeter observations

Time series plots for the SLA of the three regions of ASBoB and EIO and their differences are shown in Fig 7 Theseasonal pattern of the SLA in the AS is clearly simulated by

Fig 8 Statistics of model performance for SLA (a) the mean of differences (cm) between observations and the model (satelliteminus model) (b) standard deviations of the differences (cm) and (c) cost functionThe values are averaged for the eight yearsfrom 1994ndash2001

Fig 7Time series plots ofSLA (cm) for the periodof 1994ndash2001 fromHYCOM (red) andaltimeter measurements(black) for the regions of(a) Arabian Sea (b) Bayof Bengal and (c) theEquatorial Indian OceanThe differences in SLA(satellite minus model)for the three regions areshown in (d)


the model as shown by the altimeter However the simulatedSLA values for the BoB region (Fig 7b) show deviation fromthat of the altimeter even though the model simulates the highand low sea level patterns correctly These differences aremost prominent towards the end of 1997 and beginning of1998 coinciding with the El Nintildeo of 1997 After 1998 the dif-ferences are much less and the simulated sea levels are com-parable in amplitudes to that of altimeter In the EIO (Fig 7c)the model is able to simulate the seasonal variations of SLAclearly and the patterns match well throughout the time seriesThe differences between the simulated and altimeter sea lev-els for the three regions are shown in Fig 7d The BoB showsthe maximum variation in general Even there the differencesdo not go beyond 6cm The EIO SLA shows fewer differencesbetween the model and the altimeter observations The modelalso reproduces the AS SLAs well

The statistical analyses for the SLA are presented in Fig8 which shows the mean differences standard deviationsand CF averaged over the eight years from 1994 to 2001 forthe northern Indian Ocean The mean differences betweenthe observations and the model SLA (Fig 8a) remain lessthan 015cm which shows that the model could simulaterealistic SLA for all the three regions Since the satellite datado not have measurements near to the coast the statisticalcalculations also lack results close to the coast However themean differences for the whole region show that the modelresults are in agreement with the altimeter data The stan-dard deviations from the mean (Fig 8b) have values below8cm for most of the Indian Ocean except in two regions ndashnear to the western part of the AS and in the BoB near to theIndian coast

The long term comparison of temporal evolution of SLA(Fig 7) shows that the model simulates SLA that matches thealtimeter observations So the variations from the mean thatis seen in the averaged standard deviations (Fig 8b) must bearising from the seasonal mesoscale activities and eddy for-mations occurring in these particular regions The CF valuescalculated to test the goodness of fit between the observationsand model SLA are shown in Fig 8c The CF values for thewhole northern Indian Ocean are less than one thus indicat-ing that the model is very good in simulating the SLA realis-tically

Validation with Argo float dataIn parts of the worldrsquos oceans where only the surface datafrom the satellites and very limited hydrographic data fromcruises are available the Argo floats provide informationwith relatively high spatial and temporal resolution In theIndian Ocean there are around 600 floats deployed until nowMost of the deployments in the Indian Ocean started in 2002The data for three years from year 2002 until 2004 are usedfor the validation The Argo dataset provides daily data forthe Indian Ocean so for comparison studies from the year2002 the model stored daily averages

The number of floats deployed in the Indian Ocean wasvery few in the beginning of 2002 especially in the BoB Butfrom 2003 onwards the deployment of floats increased andhence more data became available The Argo dataset usedhere is from the live access server of INCOIS which givesthe gridded Argo float data produced by objective analysis29

The dataset provided by INCOIS has values in the IndianOcean in the gridded form (with one degree spatial resolu-tion) with temperature and salinities down to 1000m depthThe data for three years from 2002 to 2004 are used for theanalysis For the analysis the differences are calculated forthe three different geographic areas shown in Fig 1 Themodel data are remapped in vertical using cubic spline inter-polation and from this data the model temperatures andsalinities at the location of the floats are extracted and the dif-ferences between the Argo and model data are calculated atthe grid points of the float data From these differences themean errors and the root mean square of errors (RMSE) arecalculated for temperature and salinity and the results aver-aged over the three years from 2002 to 2004 are presented inFig 9

The temperature difference patterns at the surface levelindicate that at the surface the model produces similar tem-peratures as those measured by the floats in all the threeregions (Fig 9abc) The deviations from the observationsare more pronounced in the AS (Fig 9a) The mean differ-ences for the AS show that the model is slightly warmer inthe surface waters In the subsurface the model simulatesmuch warmer water compared than observed values Thedifference in the subsurface waters reaches up to 6degC TheRMSE are also at their maximum in the subsurface withvalues reaching up to 15degC This subsurface warming is inthe depth range of 100m to 300m Below 400m the modelsimulates slightly cooler water temperatures than measuredtemperatures but the differences are less than 2degC at most ofthe depth levels

The BoB (Fig 9b) agree better with the observations thanthe two other regions The mean differences and RMSE alsoshow that the model agrees well with the Argo data for thesurface waters of the area The subsurface warming is presentup to an extent in this area too with the model showing amean 3ndash5degC increase in temperature in the same depth rangeof 100m to 300m The mean differences below 400m are lessthan 2degC and shows that the model temperatures do not varymuch from the measured values

The differences between the model and the Argo temper-atures are close to zero in the EIO for the surface waters (Fig 9c) For the waters in the deeper levels (below 400m)the temperatures from the model are close to those meas-ured by the floats whereas in the subsurface the tempera-tures simulated by the model are warmer than the measure-ments The main feature that stands out in the comparisonsis the subsurface warming But this problem is not just spe-cific to this particular model set-up Most of the numericalmodels have problems in simulating a sharp thermocline48

The simulation of diffuse thermoclines are seen in othermodels too49

The HYCOM model used by Winther et al50 for the NorthSea and Skagerrak region to test the skills of the model incoastal shelf areas reports a diffuse thermocline Lee at al51

also reports a similar weakly stratified and warmer thermo-clines in their simulations The experiments performed withthe present model by changing target densities did not have aprominent effect on the diffuse thermocline formation It isassumed this could be more of an algorithmic problemtreat-ment of hybrid coordinate layers A more recent version of




HYCOM has improved behaviour at the base of the mixedlayer and will be used in further studies

The comparisons of model salinities with Argo float dataare also done for the upper 1000m water column of the threegeographic regions selected in the northern Indian Ocean (Fig 9def) In the AS (Fig 9d) the surface salinity differencesare around 02 psu with model simulating fresher waters Thesubsurface waters in the model are more saline than the meas-ured data with the differences reaching up to 03 psu Below300m the model produces fresher waters again but the differ-

ences do not go beyond 03 psu The RMSE values showmore deviations in the surface waters The RMSE values arealso less than +03psu

The mean salinity differences in the BoB (Fig 9e) are lowbelow 400m The surface mean errors also show that themodel simulates the surface salinities close to observationsfor the region However for the subsurface level the model ismuch fresher compared to the other two regions The meandifferences reach up to 09psu at 100m The RMSE valuesshow larger variations in the surface waters compared to the



Fig 9The mean error (black) and root mean square of errors (red) from the comparison with Argo float data (Argo minusmodel)The top panel shows the comparisons of temperature (degC) for (a) Arabian Sea (b) Bay of Bengal and (c) EquatorialIndian OceanThe bottom panel shows the comparisons of salinities (psu) for (d) Arabian Sea (e) Bay of Bengal and (f)Equatorial Indian Ocean All the plots are for the years 2002ndash2004


subsurface values which could be attributed to the freshwa-ter input fluctuations to the region This could be improved byproviding more realistic river inputs in the next version Themean differences between the model and Argo salinity valuesin deeper water levels are lower than 01psu with not muchdeviation from the mean values The EIO salinities (Fig 9f)from the model agree well with the salinities from the Argofloats almost throughout the water column with mean differ-ences below 02psu except around 100m where it reachesaround 03psu The deviations from the mean calculated forthe region are also small

SUMMARY AND CONCLUSIONSIn this study we have validated a HYCOM model for theIndian Ocean region An extensive comparison of modelresults with in-situ and satellite observations has been con-ducted and the results presented Weekly data from eightyears (1994ndash2001) are validated for the surface features ofcurrents SSTs and SLA For the next three years(2002ndash2004) the model is compared with the Argo float datato test how the model produces the temperature-salinity struc-ture in the upper 1000m water column

The weekly surface currents from the eight year run arecompared with the known circulation features of the area gath-ered from previously published results The model simulatesthe surface current in the study region remarkably well It isable to produce the major surface current patterns with realis-tic speeds The spatial comparison of SST patterns for theeight-year averages and its temporal evaluation during thistime for the entire region shows that the model is able to pro-duce accurate SSTs for the northern Indian Ocean The differ-ences between the model and observations after 1997 could bebecause the ITF is given as a constant flux into the model TheITF is correlated with the ENSO and IOD and hence thechanges in the flow during the anomalous events could nothave been simulated in the model as it is kept constant

This could be taken care of in the next version by using aseasonal cycle of the ITF rather than the mean value or bynesting a validated global model The mean error standarddeviation and CF are calculated to quantify the model per-formance The mean error is around 1degC in most parts of thenorthern Indian Ocean except for the northern coastal regionof AS The maximum deviations from the mean are associat-ed with the upwelling regions in the Indian Ocean The CFvalues for SST remain less than one for a major part of thenorthern Indian Ocean which shows a very good level of per-formance by the model

The SLA comparisons also give satisfactory results withthe model reproducing the major sea surface height featuresand their temporal variability The temporal evolution of SLAfor the eight years also shows that the model compares wellwith the observations especially in the AS and EIO regionThe mean differences between observations and the modelresults are lesser than 015cm The standard deviations calcu-lated shows that the model has more variability in the regionsnear to the western coast of the Arabian Sea (AS) near to theSomali coast and in the western coast of the Bay of Bengal(BoB) which are regions of seasonal eddy formations TheCF values of SLA remain less than one for the entire Indian

Ocean indicating that the model produces the SLA remark-ably well

The validation with the Argo float dataset has been carriedout for the three regions (AS BoB EIO) for the years 2002to 2004 The differences between observations and model(mean error) and the RMSE values were calculated at com-mon depths down to 1000m The results show that in all thethree areas of the Indian Ocean the model is able to reproducethe surface temperatures and salinities realistically BoBshows the maximum salinity differences and RMSE com-pared to AS and EIO This will be taken care of by introduc-ing more realistic river fluxes in the next version

In the subsurface waters the model shows considerabledifferences in temperatures between the observations and themodel especially in the thermocline region The model iswarmer than the observations here The warming however isnot just a problem in this present configuration but a com-mon problem in numerical ocean models as stated in differentmodelling studies48 49 Using a more advanced vertical inter-polation might bring an improvement in the new versionFurther studies with sensitivity experiments should be doneto verify this Elsewhere in the intermediate and deep watersthe model produces the temperature and salinity pattern thatis very much similar to that measured by the Argo floats

It is concluded from the validation results that the modelgives a good comparison with the in-situ and satellite dataThe model is developed with the objective of making a fore-casting system for the Indian Ocean It is concluded that withthe suggested improvements included the model can furtherbe used to study the major oceanographic features of theIndian Ocean and can be developed into a forecasting tool forthe region

ACKNOWLEDGEMENTSThe first author acknowledges a donation from Trond MohnCO Frank Mohn AS for financing her ongoing PhD studiesA grant of CPU time from the Norwegian Supercomputingproject NOTUR has been used The Ferret software is usedfor the analysis of the datasets and for preparing the graphicsThe remote sensing products used are SST from PhysicalSciences Division of NOAAESRL merged SLA producedby SsaltoDuacs distributed by Aviso with support fromCNES and the gridded Argo dataset provided byINCOISLAS

REFERENCES1 Schott FA 1983 Monsoon response of the Somali

Current and associated upwelling Progress in Oceanography12(3) 357ndash381

2 Schott FA and McCreary JP 2001 The monsoon cir-culation of the Indian Ocean Progress in Oceanography51(1) 1ndash123

3 Wyrtki K 1971 Oceanographic atlas of theInternational Indian Ocean Expedition National ScienceFoundation Publication Washington DC 531pp

4 Hurlburt HE and Thompson JD 1976 A numericalmodel of the Somali Current Journal of PhysicalOceanography 6(5) 646ndash664




5 Luther ME and OrsquoBrien JJ 1985 A model of theseasonal circulation in the Arabian Sea forced by observedwinds Progress in Oceanography 14 353ndash385

6 Anderson D Carrington D Corry R and Gordon C1991 Modelling the variability of the Somali CurrentJournal of Marine Research 49(4) 659ndash696

7 McCreary JP and Kundu PK 1988 A numericalinvestigation of the Somali Current during the southwestmonsoon Journal of Marine Research 46(1) 25ndash58

8 Bruce JG Johnson DR and Kindle JC 1994Evidence for eddy formation in the eastern Arabian Sea dur-ing the northeast monsoon Journal of Geophysical Research99(C4) 7651ndash7664

9 Bruce JG Kindle JC Kantha LH Kerling JL andBailey JF 1998 Recent observation and modeling in theArabian Sea Laccadive High region Journal of GeophysicalResearch 103(C4) 7593ndash7600

10 Hellerman S and Rosenstein M 1983 Normalmonthly wind stress over the world ocean with error esti-mates Journal of Physical Oceanography 13(7) 1093ndash1104

11 Shankar D and Shetye SR 1997 On the dynamics ofthe Lakshadweep high and low in the southeastern ArabianSea Journal of Geophysical Research 102(C6)12551ndash12562

12 McCreary JP Kundu P and Molinari RL 1993 Anumerical investigation of dynamics thermodynamics andmixed-layer processes in the Indian Ocean Progress inOceanography 31 181ndash244

13 Vinayachandran PN and Yamagata T 1998Monsoon response of the sea around Sri Lanka generation ofthermal domes and anticyclonic vortices Journal of PhysicalOceanography 28 1946ndash1960

14 Masson S Luo JJ Madec G Vialard J Durand FGualdi S Guilyardi E Behera S Delecluse P Navarra A andYamagata T 2005 Impact of barrier layer on winter-springvariability of the southeastern Arabian Sea GeophysicalResearch Letters 32 L07703 doi1010292004GL021980

15 Masson S Delecluse P Boulanger J and Menkes C2002 A model study of the seasonal variability and formationmechanisms of the barrier layer in the eastern equatorialIndian Ocean Journal of Geophysical Research 107(C12)8017 doi1010292001JC000832

16 Durand F Shankar D DeBoyer Montegut C ShenoiSSC Blanke B and Madec G 2007 Modeling the barrier-layer formation in the south-eastern Arabian Sea Journal ofClimate 20(10) 2109ndash2120

17 Kurian J and Vinayachandran PN 2007Mechanisms of formation of the Arabian Sea mini warm poolin a high-resolution Ocean General Circulation ModelJournal of Geophysical Research 112 C05009doi1010292006JC003631

18 Haugen VE Johannessen OM and Evensen G2002a Indian Ocean Validation of the Miami IsopycnicCoordinate Ocean Model and ENSO events during1958ndash1998 Journal of Geophysical Research 107(C5) 3043doi1010292000JC000330

19 Haugen VE Johannessen OM and Evensen G2002b Mesoscale modeling study of the oceanographic con-ditions off the southwest coast of India Journal of EarthSystem Science 111(3) 321ndash337

20 Bleck R 2002 An oceanic general circulationmodel framed in hybrid isopycnic-Cartesian coordinatesOcean Modelling 37 55ndash88

21 Bentsen M Evensen G Drange H and JenkinsAD 1999 Coordinate transformation on a sphere usingconformal mapping Monthly Weather Review 1272733ndash2740

22 Large WG McWilliams JC and Doney SC 1994Oceanic vertical mixing A review and a model with a nonlo-cal boundary layer parameterization Review of Geophysics32(4) 363ndash403

23 Teague WJ Carron M and Hogan PJ 1990 A com-parison between the Generalized Digital EnvironmentalModel and Levitus climatologies Journal of GeophysicalResearch 95(C5) 7167ndash7183

24 Legates D and Willmott C 1990 Mean seasonaland spatial variability in gauge-corrected global precipita-tion Journal of Climatology 10 111ndash127

25 Gordon AL Ma S Olson DB Hacker P Ffield ATalley LD Wilson D and Baringer M 1997 Advection anddiffusion of Indonesianthroughflow water within the IndianOcean South Equatorial Current Geophysical ResearchLetters 24 2573ndash2576

26 Gordon AL 2005 Oceanography of the Indonesianseas and their throughflow Oceanography 18 14ndash27

27 Wijffels SE Meyers G and Godfrey JS 2008 A20-yr average of the Indonesian Throughflow Regionalcurrents and the interbasin exchange Journal of PhysicalOceanography 38(9) 1965ndash1978

28 Reynolds RW Rayner NA Smith TM Stokes DCand Wang W 2002 An improved in situ and satellite SSTanalysis for climate Journal of Climate 15(13)1609ndash1625

29 Udaya Bhaskar TVS Ravichandran M andDevender R 2007 An operational Objective Analysis systemat INCOIS for generation of Argo value added productsIndian National Centre for Ocean InformationServicesTechinical Report No INCOISMOG-TR-207

30 OSPAR Commission 1998 Report of the modellingworkshop on eutrophication issues OSPAR Report 86 DenHaag The Netherlands

31 Rao RR Molinari RL and Festa JF 1989Evolution of the climatological near-surface thermalstructure of the tropical Indian Ocean 1 Description ofmean monthly mixed layer depth and sea surfacetemperature surface current and surface meteorologicalfields Journal of Geophysical Research 94(C8)10801ndash10815

32 Molinari RL Olson D and Reverdin G 1990Surface current distributions in the tropical India Oceanderived from compilations of surface bouy trajectoriesJournal of Geophysical Research 95(C5) 7217ndash7238

33 Shankar D Vinayachandran PN Unnikrishnan ASand Shetye SR 2002 The monsoon currents in the northIndian Ocean Progress in Oceanography 52(1) 63ndash120

34 Shetye SR Shankar D Shenoi SSCVinayachandran PN Sundar D Michael GS and NampoothiriG 1996 Hydrography and circulation in the western Bay ofBengal during the northeast monsoon Journal ofGeophysical Research 101(C6) 14011ndash14025




35 Shetye SR Gouveia AD Shenoi SSC Michael GSSundar D Almeida AM and Santanam K 1991 The coastalcurrent off western India during the northeast monsoon DeepSea Research Part A 38(12) 1517ndash1529

36 Wyrtki K 1973 An equatorial jet in the IndianOcean Science 181 262ndash264

37 Shenoi SSC Shankar D and Shetye SR 1999 Onthe sea surface temperature high in the Lakshadweep Seabefore the onset of the southwest monsoon Journal ofGeophysical Research 104(C7) 15703ndash15712

38 Joseph PV 1990 Warm pool over the Indian Oceanand monsoon onset Tropical Ocean-Atmosphere Newsletter53 1ndash5

39 Rao RR and Sivakumar R 1999 On the possiblemechanisms of the evolution of a mini-warm pool during thepre-summer monsoon season and the onset vortex in thesoutheastern Arabian Sea Quarterly Journal of the RoyalMeteorological Society 125(555) 787ndash809

40 Fischer AS Weller RA Rudnick DL Eriksen CCLee CM Brink KH Fox CA and Leben RR 2002 Mesoscaleeddies coastal upwelling and the upper-ocean heat budgetin the Arabian Sea Deep Sea Research Part II 49(12)2231ndash2264

41 Rao RR Girish Kumar MS Ravichandran MSamala BK and Sreedevi N 2006 Observed mini-cold pooloff the southern tip of India and its intrusion into the southcentral Bay of Bengal during summer monsoon seasonGeophysical Research Letters 33 L06607 doi1010292005GL025382

42 Saji NN Goswami BN Vinayachandran PN andYamagata T 1999 A dipole mode in the tropical InidanOcean Nature 401 360ndash363

43 Yu L and Rienecker MM 2000 Indian Oceanwarming of 1997ndash1998 Journal of Geophysical Research105(C7) 16923ndash16939

44 Murtugudde R McCreary JP and Busalacchi AJ2000 Oceanic processes associated with anomalous eventsin the Indian Ocean with relevance to 1997ndash1998 Journal ofGeophysical Research 105(C2) 3295ndash3306

45 England MH and Huang F 2005 On the interannualvariability of the Indonesian Throughflow and its linkagewith ENSO Journal of Climate 18(9) 1435ndash1444

46 Lan J Hong J and Wang Y 2009 Relationship of theinterannual variability of the Indonesian Throughflow withthe IOD over the tropical Indian Ocean Theoretical andApplied Climatology 97 75ndash79

47 Johannessen OM Subbaraju G and Blindheim J1987 Seasonal variations of the oceanographic conditionsoff the southwest coast of India during 1971ndash1975 FiskeridirSkr Ser Havunders 18 247ndash261

48 Wilson SG 2000 How ocean vertical mixing andaccumulation of warm surface water influence thelsquoSharpnessrsquo of the equatorial thermocline Journal of Climate13(20) 3638ndash3656

49 Griffies SM Adcroft AJ Banks H Boumlning CWChassignet EP Danabasoglu G Danilov S Deleersnijder EDrange H England M Fox-Kemper B Gerdes RGnanadesikan A Greatbatch RJ Hallberg RW Hanert EHarrison MJ Legg S Little CM Madec G Marsland SJNikurashin M Pirani A Simmons HL Schroumlter J SamuelsBL Treguier A Toggweiler JR Tsujino H Vallis GK andWhite L 2009 Problems and prospects in large-scale oceancirculation models OceanObsrsquo09 Community White Paper

50 Winther NG and Evensen G 2006 A HybridCoordinate Ocean Model for shelf sea simulation OceanModelling 13 221ndash237

51 Lee SK Enfield DB and Wang C 2005 Ocean gener-al circulation model sensitivity experiments on the annual cycleof western hemisphere warm pool Journal of GeophysicalResearch 110 C09004 doi1010292004JC002640




along its western coastal zones Thus in the earlier timesmost of the observational cruises as well as modelling studieswere focused on the monsoon seasons and the associatedreversal of Somali Current (SC) along the coast of Somaliaand Oman in the western part of the AS12

The International Indian Ocean Expedition (IIOE) from1964ndash66 carried out the first extensive study of the IndianOcean and the resulting hydrographic atlas3 started a series ofcomprehensive studies on the Indian Ocean circulationSchott and McCreary2 give an extensive review on the mon-soon circulation of the Indian Ocean based on observationstheory and modelling studies

Numerical modelling is a powerful tool to study theoceans especially in regions like the Indian Ocean where theobservational data are sparse Even though modelling studiesof the Indian Ocean are limited in number compared to otherparts of the worldrsquos oceans there has been major progress inthe past few years The SC the coastal currents along the eastand west coasts of India and the associated hydrographic fea-tures have received much attention in the past and presentMany of the earlier modelling studies of the SC implementedreduced gravity models1 4 5 to study its dynamics and variabil-ity Anderson et al6 modelled the SC during the southwestmonsoon in a 16-level model and McCreary and Kundu7 useda 212-layer model to analyse the characteristics of the SC dur-ing southwest monsoon All these modelling studies aided inunderstanding the effects of local as well as remote forcingin driving the SC during summer monsoon and the impor-tance of a slanted boundary in the models for proper gyre for-mation in the western AS

The anticyclonic eddy formations in the south eastern AS(SEAS) during northeast monsoon were analysed in detail byBruce et al89 employing a three layer reduced gravity modelforced by the Hellerman amp Rosenstein10 wind climatology Anassociated high in sea surface is developed during the north-east monsoon and a low is formed during southwest monsoonperiod The formation mechanisms of these high and lowwere examined in detail by Shankar and Shetye11 using theirreduced gravity model for the northern Indian OceanMcCreary et al12 published a comprehensive study of thedynamics thermodynamics and mixed-layer process of thecirculation of Indian Ocean simulated in a 212-layer thermo-dynamic numerical model The influences of forcing by localand remote along shore winds on the coastal currents werediscussed in this study The dynamics of the cyclonic (anticy-clonic) gyre formation east of Sri Lanka during summer (win-ter) monsoon termed as Sri Lankan Dome (Bay of BengalDome) are discussed by Vinayachandran amp Yamagata13

In recent years many of the major oceanographic featuresin the Indian Ocean area have been investigated using variousmodels The barrier layer (BL) formations in the IndianOcean have been modelled by Masson et al14 15 They used acoupled model to evaluate the effect of BL on the sea surfacetemperature (SST) and on the monsoon onset in the SEAS14

and in another study studied the BL formation in the easternequatorial Indian Ocean (EIO)15 The effect of salinity on thegeneration of the BL in the SEAS is analysed employing anocean general circulation model by Durand et al16 Numerousstudies have been done about Indian Ocean warm pool and inone recent study the mechanisms and formation of Arabian

Sea mini warm pool is discussed by Kurian andVinayachandran17 Most of the earlier models were based onreduced gravity model structure while as many of the recentmodels are based on versions of Geophysical Fluid DynamicsLaboratory (GFDL) Modular Ocean Model (MOM)

Haugen et al18 implemented the Miami IsopycnicCoordinate Model (MICOM) to examine the variability of theIndian Ocean in response to the monsoons and to study theinter-annual signals in the Indian Ocean They also discussedthe seasonal circulation and coastal upwelling along the south-west Indian coast using a high resolution nested model for theregion19 As a continuation of the Nansen Centerrsquos study usingMICOM in the Indian Ocean18 19 the HYbrid Coordinate OceanModel (HYCOM) has been implemented which allows a high-er vertical resolution in the mixed layer and hence give animproved representation of the surface layers20 contrasted toMICOM Most of the previous models developed for the IndianOcean have coarser horizontal resolutions Here the model hasbeen configured with higher resolution (model details in thenext section) in the northern Indian Ocean which is the focusarea of the authorsrsquo research The final goal is to implement aforecasting system for the Indian Ocean however at present nodata assimilation is used this is planned as the next step

Even though the models have been much improved acomplete study on the validation of the models used in theIndian Ocean are less common Haugen et al18 validatedMICOM for the Indian Ocean Kurian and Vinayachandran17

presents the results of comparisons of their model (based onMOM) with observations in the study about the AS miniwarm pool region Model results needed to be evaluated andvalidated before utilising them for analysing or addressingany scientific problem

Thus in this study the authors present the preliminaryresults from the validation of the model using in-situ and satel-lite observations To validate the model for surface featuresthe climatology of model results for eight years (1994ndash2001)are used The surface currents are compared with observationsbased on previously published results The surface tempera-tures and anomalies of sea level are compared with satelliteobservations The inter-annual variability of the surface tem-peratures and the anomalies of sea level are also compared andthe differences between the model and observations have beenquantified For the evaluation of model in the subsurface thetemperature and salinity for three years (2002ndash2004) are com-pared with the Argo profiles A detailed description of themodel configuration is given in the next section followed bythe satellite and in-situ data used for the model evaluation Theresults of validation are then discussed

MODEL DESCRIPTIONHYCOM is a hybrid coordinate model based on MICOM andis able to interchange between different vertical coordinateschemes It is a primitive equation general circulation modelwith vertical coordinates that remain isopycnic in the openstratified ocean In the weakly stratified upper ocean theisopycnal vertical coordinates smoothly transfer to z-coordi-nates This hybrid formulation gives a better emulation of thesurface mixed layer and the coastal shelf regions20 comparedto MICOM The terrain following coordinates are not used in


























D

σ D nn

N

ND D= minus( )

=sum1 2

1

CF =minus

=sum1

1N

D Mn n

Dn

N

σ




































































































































































D

σ D nn

N

ND D= minus( )

=sum1 2

1

CF =minus

=sum1

1N

D Mn n

Dn

N

σ

























































































































































D

σ D nn

N

ND D= minus( )

=sum1 2

1

CF =minus

=sum1

1N

D Mn n

Dn

N

σ


































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































Documents

Validation of a hybrid coordinate ocean model for the ...€¦ · ity. Anderson et al6 modelled the SC during the southwest monsoon in a 16-level model and McCreary and Kundu 7 used