ATMOSPHERIC AND OCEANIC SCIENCE LETTERS, 2012, VOL. 5, NO. 4, 273−279

Indian SST Biases in a Flexible Regional Ocean Atmosphere Land System (FROALS) Model HAN Zhen-Yu1, 2, ZHOU Tian-Jun1, and ZOU Li-Wei1 1State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of At- mospheric Physics, Chinese Academy of Sciences, Beijing 100029, China 2Graduate University of the Chinese Academy of Sciences, Beijing 100049, China Received 23 March 2012; revised 19 April 2012; accepted 20 April 2012; published 16 July 2012

Abstract The authors examine the Indian Ocean interaction. Recent studies have shown that coupled re- surface temperature (SST) biases simulated by a Flexible gional climate models (CRCMs) generally improve the Regional Ocean Atmosphere Land System (FROALS) simulation (Seo et al., 2008; Ratnam et al., 2009). A model. The regional coupled model exhibits pronounced Flexible Regional Ocean Atmosphere Land System cold SST biases in a large portion of the Indian Ocean (FROALS) model was established at the State Key Labo- warm pool. Negative biases in the net surface heat fluxes ratory of Numerical Modeling for Atmospheric Sciences are evident in the model, leading to the cold biases of the and Geophysical Fluid Dynamics, Institute of Atmos- SST. Further analysis indicates that the negative biases in pheric Physics (LASG, IAP) (Zou and Zhou, 2011, 2012). the net surface heat fluxes are mainly contributed by the An evaluation of the model performance over the western biases of sensible heat and latent heat flux. Near-surface North Pacific found that the SST shows cold biases, meteorological variables that could contribute to the SST which partly stem from the overestimation of convection biases are also examined. It is found that the biases of frequency by the (Zou and Zhou, sensible heat and latent heat flux are caused by the colder 2011). However, the performance of this model over the and dryer near-surface air in the model. IO is unknown. The main motivation of the current study Keywords: Indian Ocean, SST biases, FROALS, evalua- is to assess the performance of the FROALS model over tion the IO. We attempt to address the following questions: (1) Citation: Han, Z.-Y., T.-J. Zhou, and L.-W. Zou, 2012: How well does the FROALS model capture the clima- Indian Ocean SST biases in a Flexible Regional Ocean tological features of the IO SST? (2) What are the major Atmosphere Land System (FROALS) model, Atmos. sources of the SST biases? Oceanic Sci. Lett., 5, 273–279. The remainder of the paper is organized as follows. A brief description of the FROALS model, its experimental 1 Introduction design, and the datasets employed in this study are de- scribed in section 2. Section 3 compares the simulation The Indian Ocean (IO) plays a crucial role in the results from FROALS with the observational data. Finally, Asian-Australian (A-A) monsoon variability due to the conclusions are given in the last section. importance of both the land-sea thermal contrast in con- trolling the strength of A-A monsoon and the sea surface 2 Model, experiment, and data description temperature (SST)-convection relationship. It is believed that the interannual variability of Indian monsoon activity 2.1 Model description depends heavily on air-sea interactions that take place The atmospheric component of FROALS is Regional during the travel of the monsoon current across the ocean version 3 (RegCM3), which was devel- (Shukla, 1975; Webster et al., 1999; Meehl and Arblaster, oped at the Abdus Salam International Centre for Theo- 2002). The significant impacts of the IO on East Asian retical Physics (ICTP) (Pal et al., 2007). The oceanic and western North Pacific monsoon variability have also component is the Princeton Ocean Model version 2000 been identified in recent years (Li et al., 2008; Xie et al., (POM2K), which was developed at Princeton University 2009; Wu et al., 2010). (see http://www.aos.princeton.edu/WWWPUBLIC/htdocs. Partly due to their increased resolution and better oro- pom/ for details). The RegCM3 and POM2K models are graphic representation, regional climate models (RCMs) coupled through the Ocean Atmosphere Sea Ice Soil 3.0 have played active roles in regional climate studies. (OASIS3.0) coupler (Valcke, 2006). Further details about RCMs generally show better performance than AGCMs this model can be found in Zou and Zhou (2011, 2012). over the IO (e.g., Dash et al., 2006; Dobler and Ahrens, Different from the original version (Zou and Zhou, 2010; Mukhopadhyay et al., 2010; Polanski et al., 2010; 2011, 2012), a radiation method is used along the open Lucas-Picher et al., 2011). However, many RCMs still boundaries of the ocean to allow for stable, long-term show low skill in simulating precipitation over the oce- integrations. We require that the normal barotropic veloci- anic regions, partly due to the absence of two-way air-sea ties satisfy the following radiation condition, according to Flather (1976): Corresponding author: ZHOU Tian-Jun, [email protected] Ub = Uobs + (c/H)(ηb−1 − ηobs), (1) 274 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS VOL. 5 where subscripts b and b−1 denote the boundary point and Therefore, we use the mean temperature of the upper 5 m the interior point closest to b, respectively; obs denotes (T5m) to represent the SST for both model outputs and the observation; U denotes the normal barotropic velocity; observations. The observed SST used in the evaluation is 1/2 c = (gH) , the local shallow water wave speed; H de- the monthly mean T5m from SODA. 2) Because there are notes the water depth; and η denotes the surface elevation. significant uncertainties in the surface heat flux and radia- Because the fresh water flux is not included in the cou- tion estimates, two products of the monthly mean surface pling process, to reduce unrealistic changes in the turbulent fluxes, near-surface meteorological variables mixed layer, a weak relaxation of salinity to the monthly (Objectively Analyzed air-sea Fluxes (OAFlux) (Yu and in the upper five layers of the model is ap- Weller, 2007) and Goddard Satellite-Based Surface Tur- plied. bulent Fluxes (GSSTF) 2c (Chou et al., 2003)) and sur- face radiation (International Satellite Cloud Climatology 2.2 Experimental design and observational data Project-Flux Data (ISCCP-FD) (Zhang et al., 2004) and In this study, the FROALS model was set up over a Surface Radiation Budget (SRB) 3.0 (Gupta et al., 2006)) broad IO domain from 20°S to 40°N, 40°E to 120°E. The are used to evaluate the model outputs. Tibet Plateau is included to avoid steep orography along Here, we deduct the shortwave penetration into the the boundary regions. The horizontal resolution of the subsurface (deeper than 5 m) for both model outputs and RegCM3 is 60 km, and there are 18 vertical levels. observations. The deduction factor is 0.296, according to RegCM3 is initialized on 1 January 1995 using the Na- a simple subsurface heating parameterization (Paulson tional Center for Environmental Prediction-National Cen- and Simpson, 1977), ter for Atmospheric Research (NCEP-NCAR) reanalysis SW =SW (RR⋅+− e−−5/aa12 (1 )e 5/ ) , (2) dataset (Kalnay et al., 1996). The boundary conditions of pen 0 the RegCM3 are derived from NCEP-NCAR reanalysis where R (=0.62) is a separation constant, and a1 (=0.6 m) and updated every six hours. The horizontal resolution of and a2 (=20.0 m) are the attenuation length scales; SWpen the POM2K is 0.5°, and there are 30 vertical levels (see denotes the shortwave penetration into the subsurface; Table 1). The level thickness varies non-uniformly, with a and SW0 denotes the shortwave radiation at the surface. higher resolution in the upper ocean and a lower resolu- Then, the ensemble observation of surface turbulent tion in the deep ocean. The minimum and maximum fluxes (surface radiations) is the arithmetic average value depths for the model are 5 m and 5000 m, respectively. of OAFlux and GSSTF (ISCCP and SRB). Finally, these The initial conditions of the POM2K are derived from observed fields and model outputs are all remapped onto a Simple Ocean Data Assimilation (SODA) 2.1.4 monthly uniform 0.5×0.5 horizontal grid. data (Carton and Giese, 2008). The climatological and surface elevation from SODA are also em- 3 Results ployed for the oceanic boundary conditions. The air-sea coupling frequency is three hours. After a spin-up of three The spatial distribution of annual mean SST biases in years, a consecutive 10-year simulation from 1998 to the IO from FROALS along with the SST from SODA 2007 is performed. In the following analysis, the model’s are shown in Figs. 1a−b. The observed SST is character- monthly climatology is constructed based on the 10-year ized by a warm pool (>28°C) to the north of 10°S, except model outputs. for the northwestern Arabian Sea, and shows a strong To assess the model performance, the following obser- negative meridional gradient to the south of 10°S. Cold vational products are used: 1) In ocean model outputs, the biases in the SST are evident in the board regions of the SST is defined as the mean temperature over the first warm pool, with the greatest values in the Bay of Bengal layer of several meters thick, which may not be consistent (BoB), where the biases are larger than 1.5°C. Warm bi- with some observational products (Donlon et al., 2002). ases are found over the western Arabian Sea, the western

Table 1 Values of sigma coordinates used in the ocean model. Layer Sigma coordinate Layer Sigma coordinate Layer Sigma coordinate 1 0.0000 11 –0.1739 21 –0.6087 2 –0.0007 12 –0.2174 22 –0.6522 3 –0.0014 13 –0.2609 23 –0.6957 4 –0.0027 14 –0.3043 24 –0.7391 5 –0.0054 15 –0.3478 25 –0.7826 6 –0.0109 16 –0.3913 26 –0.8261 7 –0.0217 17 –0.4348 27 –0.8696 8 –0.0435 18 –0.4783 28 –0.9130 9 –0.0870 19 –0.5217 29 –0.9565 10 –0.1304 20 –0.5652 30 –1.0000 NO. 4 HAN ET AL.: INDIAN OCEAN SST BIASES IN FROALS 275

Figure 1 (a) Annual mean SST from SODA, (b) difference between FROALS and SODA (units: °C), (c) annual mean net surface heat flux from ensemble observation (see text for details), and difference between FROALS and ensemble observation (units: W m−2) for (d) net surface heat flux, (e) shortwave radiation, (f) latent heat flux, (g) longwave radiation, and (h) sensible heat flux. Rectangles: NIO (the northern Indian Ocean) and EIO (the equatorial Indian Ocean) denote 8°N–15°N, 65°E–97°E and 5°S–3°N, 60°E–95°E. boundary near the equator, and the southern Tropical In- in Figs. 1c−d. Based on the ensemble observation, which dian Ocean. is defined as the sum of ensemble surface turbulence heat The SST biases are possibly related to the errors in the fluxes and radiation, the Qnet is positive (warming the surface heat fluxes. The annual mean surface net heat flux ocean, positive downward) over most of the ocean basin, (Qnet) observed and the biases in the FROALS are shown with a magnitude of several tens of watts per square meter. 276 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS VOL. 5

There are strong maxima at the western boundary near the simulated values are systematically approximately 2°C equator, with lesser maxima off Somalia and Sumatra, colder throughout the year (Fig. 2a). In the EIO, the pic- while heat losses are found in the northern Arabian Sea. ture is quite different: the seasonal cycle of SST is domi- The FROALS has negative Qnet biases (i.e., insufficiently nated by the annual signal, but the FROALS features a warming the ocean) in almost the entire region. Positive strong semiannual pattern. The SST in the EIO also suf- biases are only found in small regions near the coastline fers significant cold biases over all months, with an aver- in the model. The nearly universal negative biases suggest age of 0.5°C to 1°C (Fig. 3a). A similar analysis of Qnet that the underestimation of Qnet is a major cause of the shows that the cold SST biases are associated with the cold SST biases. model’s tendency to produce less Qnet year-round (not The biases in each component of Qnet are further ex- shown). To determine the main bias sources for the sea- amined to identify which component contributes the most sonal cycle of SST, the seasonal evolutions of the clima- to the negative biases. Figures 1e−h display the biases of tological SW, LH, and SH are displayed in Figs. 2−3. surface shortwave radiation (SW), latent heat flux (LH), In the NIO, cooling by the LH is strongest in winter longwave radiation (LW), and sensible heat flux (SH), (Fig. 2c), mostly due to the cold and dry northeasterly respectively. The biases of LH and SH are relative to the winds (Figs. 4a−c). The spring is marked by peak SW ensemble mean of OAFlux and GSSTF, while the biases (Fig. 2b), resulting in the strongest net surface heating of of SW and LW are relative to the ensemble mean of the year (not shown). The enhanced latent heat loss in ISCCP and SRB. Positive biases of LW are found in most summer by southwesterly winds (Fig. 4a) again favors of the IO warm pool, except part of the eastern equatorial cooling. These features are all well simulated, while the Indian Ocean, indicating that the cold SST biases there model has approximately 20 W m–2 LH cooling biases are not caused by the LW biases. The major components from April to September (Fig. 2c) and less than 10 W m–2 contributing to the negative net surface heat flux biases SW cooling biases from July to September (Fig. 2b). The are SW, LH, and SH. SH mostly varies between 0 and −10 W m–2, with mini- The seasonal variations of SST simulated by the mum loss during July to August. The simulated SH indi- FROALS are also evaluated. For brevity, we choose two cates a clear semiannual signal with maximum loss in regions with significant cold biases: the northern Indian both June and December; it contributes 10−15 W m–2 Ocean (NIO: 8°N−15°N, 65°E−97°E) and the equatorial negative biases throughout the year on the same order of Indian Ocean (EIO: 5°S−3°N, 60°E−95°E), shown in Fig. magnitude as the absolute value of SH (Fig. 2d). 1b. In the observation, the NIO shows a bimodal SST In the EIO, the SW exhibits a bimodality with one ad- distribution. This seasonality of the SST in the NIO is ditional significant peak in the autumn, and the SST is reasonably simulated by the FROALS, although the cooled via LH by strong winds (Fig. 4d) in the summer.

Figure 2 The seasonal cycle (upper panel) and the biases (lower panel) of the (a) SST (units: °C), (b) shortwave radiation (units: W m−2), (c) latent heat flux (units: W m−2), and (d) sensible heat flux (units: W m−2) averaged over the northern Indian Ocean. NO. 4 HAN ET AL.: INDIAN OCEAN SST BIASES IN FROALS 277

Figure 3 Same as Fig. 2 but for the equatorial Indian Ocean.

FROALS simulation exhibits SW warming biases from over the IO by performing a consecutive 10-year simula- December to April with a magnitude of 10–20 W m–2 (Fig. tion from 1998 to 2007. The results show that the cold 3b). The LH values are close to those from the observa- biases of SST are evident in the simulation, covering most tions, but the phase of the seasonal cycle in the FROALS of the IO warm pool. The value of cold SST biases can is less satisfactory (Fig. 3c). The seasonal variations of exceed 1.5°C in the BoB. The seasonal variations of SST SH are well simulated, showing maximum loss during in the NIO show a semiannual cycle, whereas those in the summer, but there are still negative biases of approxi- EIO show an annual cycle. Only the former can be rea- mately 6 W m−2 throughout the year (Fig. 3d). sonably simulated by FROALS. The cold biases of SST The cooling biases in LH and SH are systematic and persist throughout the year in both regions. The cold SST can be found in different regions and months. Therefore, biases are primarily caused by errors in Qnet. Among the the errors in LH and SH appear to be the primary sources four components of Qnet, LH and SH together contribute –2 of cold SST biases in most regions. These negative biases more than 10 (30) W m negative biases in the EIO (NIO) –2 of LH and SH may further stem from the errors of year round, and SW only contributes less than 10 W m near-surface meteorological variables and SST. The equa- negative biases from July to September. In contrast, the tions for LH and SH are biases of LW are positive, and thus, they do not cause the negative biases in Q . Therefore, the errors in LH and LH~ws(qa−qs(SST)), net SH are the major sources of the cold SST biases. Analysis SH~ws(ta−SST), (3) of near-surface meteorological variables shows that the where qa is surface air humidity; qs is air saturated humid- negative biases of LH and SH are further due to the colder ity; ta is air temperature; and ws is wind speed. The cold errors in SST favor a lower release of both LH and SH. and drier near-surface air in the model. The biases of near-surface meteorological variables aver- Warm bias errors are found in the SST over the western aged over the two regions are shown in Fig. 4. The wind Arabian Sea, the western boundary near the equator, and speed in the model is less than that from either OAFlux or southern Tropical Indian Ocean. The pattern of warm er- GSSTF (Figs. 4a and 4d), which tends to induce smaller rors over the western Indian Ocean is similar to that of the state-of-the-art CGCMs (Bollasina and Nigam, 2009) and absolute values of both LH and SH. The colder and dryer the CRCM (Seo et al., 2008) in summer. Their results sea surface air in the model (Figs. 4b and 4e; Figs. 4c and show that a damped Somali jet in the models leads to a 4f) may result in higher absolute values of the surface large warm bias in the upwelled water. In FROALS, the turbulence fluxes. large negative biases in Qnet cannot lead to warm SST biases over these regions. Thus, the errors in the upper 4 Conclusions and discussion ocean processes may also contribute to the SST biases, We assessed the performance of the FROALS model which deserve further study. 278 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS VOL. 5

Figure 4 The seasonal cycle of the (a, d) wind speed (units: m s−1), (b, e) air temperature (units: °C), and (c, f) air humidity (units: g kg−1) averaged over (a−c) the northern Indian Ocean and (d−f) the equatorial Indian Ocean.

Acknowledgements. This study was supported by the National Gupta, S., P. Stackhouse Jr, S. Cox, et al., 2006: Surface radiation High Technology Research and Development Program of China budget project completes 22-year data set, GEWEX News, 16, (863 Program, Grant No. 2010AA012304). 12–13. Kalnay, E., M. Kanamitsu, R. Kistler, et al., 1996: The NCEP/NCAR References 40-year reanalysis project, Bull. Amer. Meteor. Soc., 77, 437–471. Li, S., J. Lu, G. Huang, et al., 2008: Tropical Indian Ocean basin Bollasina, M., and S. Nigam, 2009: Indian ocean SST, evaporation, warming and East Asian summer monsoon: A multiple AGCM and precipitation during the south Asian summer monsoon in study, J. Climate, 21, 6080–6088. IPCC-AR4 coupled simulations, Climate Dyn., 33, 1017–1032. Lucas-Picher, P., J. H. Christensen, F. Saeed, et al., 2011: Can re- Carton, J. A., and B. S. Giese, 2008: A reanalysis of ocean climate gional climate models represent the Indian monsoon? J. Hydro- using simple ocean data assimilation (SODA), Mon. Wea. Rev., meteor., 12, 849–868. 136, 2999–3017. Meehl, G. A., and J. M. Arblaster, 2002: Indian monsoon GCM Chou, S.-H., E. Nelkin, J. Ardizzone, et al., 2003: Surface turbulent sensitivity experiments testing tropospheric biennial oscillation heat and momentum fluxes over global based on the transition conditions, J. Climate, 15, 923–944. Goddard satellite retrievals, version 2 (GSSTF2), J. Climate, 16, Mukhopadhyay, P., S. Taraphdar, B. N. Goswami, et al., 2010: 3256–3273. Indian summer monsoon precipitation climatology in a high- Dash, S. K., M. S. Shekhar, and G. P. Singh, 2006: Simulation of resolution regional climate model: Impacts of convective param- Indian summer monsoon circulation and rainfall using RegCM3, eterization on systematic biases, Wea. Forecasting, 25, 369–387. Theor. Appl. Climatol., 86, 161–172. Pal, J. S., F. Giorgi, X. Bi, et al., 2007: Regional climate modeling Dobler, A., and B. Ahrens, 2010: Analysis of the Indian summer for the developing world: The ICTP RegCM3 and REGCNET, monsoon system in the regional climate model COSMO-CLM, J. Bull. Amer. Meteor. Soc., 88, 1395–1409. Geophys. Res., 115, D16101, doi:10.1029/2009jd013497. Paulson, C. A., and J. J. Simpson, 1977: Irradiance measurements in Donlon, C. J., P. J. Minnett, C. Gentemann, et al., 2002: Toward the upper ocean, J. Phys. Oceanogr., 7, 952–956. improved validation of satellite sea surface skin temperature Polanski, S., A. Rinke, and K. Dethloff, 2010: Validation of the measurements for climate research, J. Climate, 15, 353–369. HIRHAM-simulated Indian summer monsoon circulation, Adv. Flather, R., 1976: A tidal model of the northwest European conti- Meteor., 2010, doi:10.1155/2010/415632. nental shelf, Memo. Soc. Roy. Sci. Liege, 6, 141–164. Ratnam, J., F. Giorgi, A. Kaginalkar, et al., 2009: Simulation of the NO. 4 HAN ET AL.: INDIAN OCEAN SST BIASES IN FROALS 279

Indian monsoon using the RegcCM3-ROMS regional coupled effect on Indo-western Pacific climate during the summer fol- model, Climate Dyn., 33, 119–139. lowing El Niño, J. Climate, 22, 730–747. Seo, H., R. Murtugudde, M. Jochum, et al., 2008: Modeling of Yu, L., and R. A. Weller, 2007: Objectively analyzed air-sea heat mesoscale coupled ocean-atmosphere interaction and its feed- fluxes for the global ice-free oceans (1981–2005), Bull. Amer. back to ocean in the western Arabian Sea, Ocean Model., 25, Meteor. Soc., 88, 527–539. 120–131. Zhang, Y., W. B. Rossow, A. A. Lacis, et al., 2004: Calculation of Shukla, J., 1975: Effect of Arabian sea-surface temperature anomaly radiative fluxes from the surface to top of atmosphere based on on Indian summer monsoon: A numerical experiment with the ISCCP and other global data sets: Refinements of the radiative GFDL model, J. Atmos. Sci., 32, 503–511. transfer model and the input data, J. Geophys. Res., 109, D19105, Valcke, S., 2006: OASIS3 User Guide (prism_2-5), PRISM Support doi:10.1029/2003jd004457. Initiative Report No. 3, CERFACS, 64pp. Zou, L., and T. Zhou, 2011: Sensitivity of a regional ocean-atmos- Webster, P. J., A. M. Moore, J. P. Loschnigg, et al., 1999: Coupled phere coupled model to convection parameterization over west- ocean-atmosphere dynamics in the Indian Ocean during 1997–98, ern North Pacific, J. Geophys. Res., 116, D18106, doi:10.1029/ Nature, 401, 356–360. 2011jd015844. Wu, B., T. Li, and T. Zhou, 2010: Relative contributions of the Zou, L., and T. Zhou, 2012: Development and evaluation of a re- Indian Ocean and local SST anomalies to the maintenance of the gional ocean-atmosphere coupled model with focus on the west- western North Pacific anomalous anticyclone during the El Niño ern North Pacific summer monsoon simulation: Impacts of dif- decaying summer, J. Climate, 23, 2974–2986. ferent atmospheric components, Sci. China Ser. D-Earth Sci., 55, Xie, S.-P., K. Hu, J. Hafner, et al., 2009: Indian Ocean capacitor 802–815, doi:10.1007/s11430-011-4281-3.