Ocean mesoscale structure-induced air-sea interaction in a high-resolution coupled model

LIN Pengfei,LIU Hailong,MA Jing and LIYiwen

aState Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics,Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing,China;

bCollege of Earth and Planetary Sciences,University of Chinese Academy of Sciences,Beijing,China;

cKey Laboratory of Meteorological Disaster of Ministry of Education,Nanjing University of Information Science and Technology,Nanjing,China

ABSTRACT Using the Community Earth System Model framework,the authors build a very-high-resolution quasi-global coupled model by coupling an eddy-resolving quasi-global ocean model with a high-resolution atmospheric model.The model is successfully run for six years under present climate conditions,and the simulations are evaluated against observational and reanalysis data.The model is capable of simulating large-scale oceanic and atmospheric circulation patterns,sea surface temperature(SST)fronts,oceanic eddy kinetic energy,and f i ne-scale structures of surface winds.The ocean mesoscale structure-induced air-sea interaction characteristics are explored in detail.The model can effectively reproduce positive correlations between SSTand surface wind stress induced by mesoscale structures through comparison with observations.The positive correlation is particularly signi f i cant over regions with strong oceanic fronts and eddies.However,the responses of wind stress to eddy-induced SST are weaker in the simulation than in the observations,although different magnitudes exist in different areas.Associated with weak wind responses,surface sensible heat f l ux responses to eddy-induced SST are underestimated slightly,while surface latent heat f l ux responses are overestimated because of the drier atmospheric boundary layers in the model.Both momentum mixing and pressure adjustment mechanisms play important roles in surface wind changes over oceanic fronts and eddies in the high-resolution model.

KEYWORDS Air-sea interaction;oceanic front;eddy;high-resolution simulation

1.Introduction

High-resolution observational data show positive correlationsbetween mesoscale sea surface temperatures(SSTs)and surface winds in areas where oceanic mesoscale activities(including oceanic frontsand eddies)are prominent,such asthose in the Kuroshio and itsextension(KE),the Gulf Stream(GS),in the Agulhas Return Current(ARC),and in other areas(e.g.,Xie et al.1998;Liu et al.2000;Chelton et al.2001;O'Neill,Chelton,and Esbensen 2003,O'Neill 2012).Such positive correlations result from mesoscale SST effects on the atmosphere(Xie 2004;Chelton et al.2004;Chelton and Xie 2010;Small et al.2008),which are distinct from negative correlations between basin-scale SST and surface winds due to SST responses to atmospheric forcing(Namias and Cayan 1981;Palmer and Sun 1985;Wallace,Smith,and Jiang 1990).Atmospheric changes due to oceanic mesoscale activities may further affect the ocean through turbulent momentum and heat f l ux.For brevity,the air-sea interaction on the scaleof oceanic frontsand eddiesisreferred to here as‘mesoscale air-sea interaction(MASI)'.

Mesoscale oceanic effects on the atmosphere can be found in high-resolution observations.Minobe et al.(2008)found that the SSTfront of the GScan affect not only the local atmospheric boundary layer,but also the entire troposphere.Frenger et al.(2013)found that oceanic eddies in the Southern Ocean(SO)can alter surface wind,cloud cover,and rainfall.The change in surface wind or rainfall can affect ocean eddy and other ocean dynamics in reverse(Gaube et al.2015).

Present ocean components in most global coupled general circulation models(CGCMs)have relatively low horizontal resolution(～100 km).Such ocean general circulation models(OGCMs)with low horizontal resolution cannot resolve oceanic mesoscale eddies.To resolve mesoscale oceanic structures,the horizontal resolution of the OGCM must be approximately 10 km or f i ner.Few studies on MASIuse a global CGCM with an eddy-resolving ocean component because huge computing resource and advanced technology for high performance computing are needed.Bryan et al.(2010)found that a CGCM with an eddy-resolving ocean component can capture MASI effectively.However,the responses of wind stress to mesoscale SST anomalies are weaker.There is a doubt about whether this phenomenon exists in other CGCMs.Thus,it is necessary to employ different CGCMs(with an eddy-resolving ocean component)to study MASI.

The purpose of the present study is to evaluate MASI in a high-resolution CGCM that we developed,through comparison with observations and reanalysis data.Different from the study of Bryan et al.(2010),who found and discussed the sensitivity of MASI to horizontal resolution,we focus on the simulation f i delity of MASI and discuss the underlying mechanisms involved.

The remainder of the paper is organized as follows.The models,experiments and observational datasets are introduced in Section 2.Section 3 provides both a basic evaluation of the model results and an analysis of the physical processes of MASIin relation to the observations.Section 4 is a summary of our f i ndings.

2.Models,experiments,and observational datasets

The quasi-global eddy-resolving LASG/IAP Climate System Ocean Model,version 2(LICOM2)was developed and reported by Yu,Liu,and Lin(2012).This eddy-resolving OGCM can simulate large-scale ocean circulations and oceanic mesoscale processes effectively(Yu,Liu,and Lin 2012).Version 7 of the Flux Coupler interface has been incorporated with LICOM2 over the past two years,and a fully coupled lowresolution model using LICOM2 and other components of the Community Earth System Model(CESM),version 1,have been run for 300 modelyears of testing(Lin et al.2016).The horizontal resolution of LICOM2 was increased to 10 km.Other components that we selected from CESM include the following:the f i nite volume dynamic core of the Community Atmosphere Model,version 4(CAM4),with a horizontal resolution of 0.23°×0.31°;the Community Land Model,version 4,with the same resolution as CAM4;version 4 of the Community Ice Code sea-ice model,with a horizontal resolution of 0.1°×0.1°,which is the same as that of LICOM2.Because the Arctic Ocean is not included in the OGCM,the SST and sea surface salinity poleward of 65°are assigned to the observations from the World Ocean Atlas 2005(Antonov et al.2006;Locarnini et al.2006).The high-resolution observational data are available from 2000,and external forcing values and greenhouse gas concentrations are set to the values of 2000 in the coupled experiment.A three-year integration of the stand-alone eddy-resolving ocean component was f i rst conducted as a spin-up run forced by observation data and starting from observed temperature and salinity.The results of the f i nal day of the spin-up integration were then used as the initial condition of the high-resolution coupled experiment.

The high-resolution coupled model was run for six model-years using the six-hour coupling frequency,and the last f i ve model-years of daily output were used for analysis in the study.For comparison,highresolution simulation,observation,and reanalysis data were interpolated to a 0.25°×0.25°grid via bilinear interpolation.In addition,a low-resolution coupled simulation with LICOM2 at 1°× 1°and CAM4 at 1.9°×2.5°(Lin et al.2016)was used to compare high-and low-resolution data.

To evaluate the high-resolution simulation,SST values from the Advanced Very High Resolution Radiometer and Advanced Microwave Scanning Radiometer(Reynolds et al.2007)for 2003-08 and sea level anomaly(Ducet,Traon,and Reverdin 2000)values from Archiving,Validation,and Interpretation of Satellite Oceanographic were used.Wind stress and wind speed values at 10 m were obtained from Quick Scatterometer(QuickSCAT Liu et al.2000).Latent and sensible heat f l ux(LHF and SHF respectively)data from the Japanese Ocean Flux based on Use of Remote Sensing Observations,version 2(Kubota and Tomita 2007),were also used.

The spatial high-pass f i lter approach was adopted to extract the MASI signal(Bryan et al.2010).The f i lter box was 18°in longitude and 6°in latitude.The Student's t-test was used for signi f i cance testing.The mesoscale wind-SST coupling strength was quanti f i ed by the slope of the least-squares linear f i tting line of the high-pass f i ltered wind stress and SST.The coupling strength magnitudes and their standard deviations were calculated for each 0.2°C SST interval.Before applying the spatial f i lter,the seasonal cycle was removed by subtracting the climatological multi-year monthly mean,to escape its effect(Chelton,Schlax and Sanelson 2007).This approach was also applied to the LHF and SHF data.

3.Results

3.1.Large-scale circulation and mesoscale eddies

The simulated SST describes the large-scale horizontal characteristics effectively,with only slight biases(Figure 1(a)and (b)).Compared with the lowresolution CGCM,the evident improvement is a decrease in warm SST biases along the oceanic eastern coast,due to superior resolving coastal upwelling in the high-resolution model(f i gure not shown).

Figure 1.(a,b)Mean SST(contours;interval:1°C),SST gradient and(c)their difference(f i lled;units:°C/100 km).(d,e)EKE(units:cm2 s-2)and(f)their difference.The observed(g,j)and simulated(h,k)wind stressdivergence(units:10-7 Nm-3)and wind stress curl(units:10-7 N m-3),and their difference(i,l).The wind stress curl is calculated without considering Coriolis parameter effects.The left-hand panels denote the observational results,the middle panels denote the high-resolution coupled model simulation results,and the right-hand panels denote the difference between simulation and observation.

The oceanic mesoscale structure is also reproduced well in the high-resolution model.These oceanic mesoscale features can be measured from horizontal SST gradients.The horizontal structures and magnitudesof SSTgradientsin the simulation are comparable with those in the observation,with a spatial correlation of 0.67(Figure 1(a)and(b),shaded).The strongest SST gradients can reach 4°C/100 km alongside strong currents such as the KE,GS,and ARC,and in places where tropical instability waves(TIWs)are located.The spatial correlation(0.68)of the SSTgradient in the KEand ARC is 0.7 and 0.68,respectively.The values are higher than that(0.61)in the GS,which may be due to the unrealistic GS path(Figure 1(c)).Around strong currents,strong horizontal SST gradients are often referred to as oceanic fronts.Over most of the world's oceans,the SST gradients are large,with a value of 0.5°C/100 km,except in the Indo-Paci f i c warm pool.The gradient is particularly large in the SO.These large gradients result from oceanic mesoscale features such as eddies and fronts.These gradients cannot be produced by the low-resolution coupled model,due to the overly coarse horizontal resolution of the OGCM.

The eddy kinetic energy(EKE)can also measure the activity of ocean mesoscale features,which is calculated via the sea level anomaly(Figure 1(d)and(e)).High observed EKE(>1000 cm2s-2)values are located around the KE,the GS,the western coast of the Indian Ocean,the ARC,and amidst strong currents in the SO.The simulated horizontal distribution of the EKEin the high-resolution model is similar to that in the observations,with spatial correlation of 0.72.The EKEmagnitudes are also comparable,but slightly higher magnitudes are found in some areas,such as the jet axis of the KEand ARCand near the equator.The EKEis underestimated south of the strong currents and in the open ocean(Figure 1(f)).

The wind stress is important to represent the air-sea interaction.The simulated wind stress can capture the observed one(not shown).The derived large-scale wind stress divergence and wind stress curl are simulated well by the high-resolution model(Figure 1(g-k)),such as in the Intertropical Convergence Zone(ITCZ)and the South Paci f i c Convergence Zone(SPCZ),and the negative(positive)curl in the subtropical high(subpolar gyre)of the Northern Hemisphere,but with large magnitudes(Figure 1(i)and(l)).The large magnitudes may be due to the higher resolution or the position deviation of the ITCZ or SPCZ.Small-scale structures are embedded in large-scale divergence and curl(Figure 1(g)-(k)).These small-scale structures are captured well in the high-resolution simulation(Figure 1(h)and(k)).The most evident small structures are located in the SO,KE,GS,and along the western coast of the Indian Ocean,while they have smaller magnitudes along the eastern coast and in the open ocean.These are closely related to oceanic fronts,mesoscale eddies,and island effects(e.g.,Hawaii),and are directly induced by mesoscale SST anomalies(Chelton et al.2004).

3.2.Mesoscale air-sea interactions and mechanisms

In this subsection,the KEis used as an example region to present MASIcharacteristics.High correspondences between(positive or negative)high-pass f i ltered SST anomalies and(positive or negative)wind stress magnitude anomalies are found over the KEin the observation and high-resolution simulation,particularly in the winter(Figure 2(a)and(b)).This indicates that changes in wind stress magnitudes are closely connected to mesoscale SST anomalies over the KE in winter.To quantify the responses of wind stress to mesoscale SST anomalies,their scatterplots and associated linear f i t lines are presented in Figure S1(a)and(b).Their linear correspondence indicates larger SST anomalies will lead to larger wind stress magnitude anomalies.The coupling strengths,which are measured by the slope of the linear f i t lines,over the KE,GS and ARC are provided in Table 1.Over the KE,the simulated coupling strength(0.007 Pa/°C)is almost equal to that of the observations.The simulated coupling strength valuesin the GSand ARCare 30%and 26%weaker than those in the observations,respectively.Therefore,simulated coupling strength values are smaller than those in the observations.Bryan et al.(2010)found the simulated values were underestimated by about 20%,50%,and 30%in the KE,GS,and ARC,respectively(Bryan et al.2010,Table 2).This indicates that the coupling strength values in our study are much closer to observation.The difference may mainly be due to the different eddy regimes using different ocean components.The weakness in coupling strength may also have related to inappropriate physical parameterizations of the boundary layer(Perlin et al.2014).

A signi f i cant relationship between the high-pass f i ltered SSTand wind stress magnitude anomalies can be found in observations.Positive temporal correlations are signi f i cant in regions where oceanic mesoscale processes are active,such as in the KE,GS,ARC,and TIWin the equatorial Paci f i c and Atlantic(Figure 3(a)).This denotes the existence of synchronous variations between mesoscale SST and wind stress magnitude values.The high-resolution coupled model can reproduce positive correlations presented in the observations(Figure 3(b)),while the correlations are generally negative between 30°S and 30°N,and signi f i cant positive correlations are sparse around strong eddy activities(including the KE,GS,ARC,and in the SO)in the low-resolution model(Figure 3(c)).These results indicate the eddy-induced air-sea interaction can be captured well,as the ocean model can resolve themesoscale eddies,which is consistent with the f i ndings of Bryan et al.(2010).

Table 1.Coupling strengths de f i ned by the least-squares f i tting slopes of mesoscale variables and SSTover the KE(20°-45°N,110°-160°E),the GS(30°-55°N,280°-350°E),and around the ARCin the SO(30°-60°S,0°-100°E).

Figure 3.(a)Correlation coefficients between high-pass f i ltered SST and surface wind stress magnitude values according to the observations,and(b)between high-pass f i ltered SSTand surface wind stress magnitude values according to the high-resolution and(c)low-resolution simulations.Thedotted areasare statistically signi f i cant according to the Student’s t-test at the 95%con f i dence level.

In the observation and simulation,signi f i cant positive correlations between high-pass f i ltered SST anomalies and LHFand SHFanomalies appear over the KEregion(Figure S1(c-f)).Here,the release of f l ux from ocean to atmosphere(upward)isde f i ned aspositive.Thereleaseof LHF(SHF)increases(decreases)from the ocean asthe SST value increases(decreases).These 1°CSSTanomalies can induce LHF anomalies of roughly 20 W m-2and SHF anomalies of roughly 10 W m-2over the KE region in the observation and simulation(Figure S1(c-f)).The SHF responses are slightly underestimated in t

he simulation.However,the LHFresponsesto SSTanomaliesin thehighresolution simulation are much larger than those in the observations,with values of 10%over the KE and GS region and 50%over the ARCregion(Table 1).Thislarger eddy-induced LHFresponse hasalso been found in other data(Ma et al.2015).According to the above analysis,small responsesof surface wind cannot lead to large LHF responses to SST anomalies.According to the bulk formula,LHFis proportional to air speci f i c humidity.We do f i nd that air in the lower layer is much drier in the highresolution simulation than that in the observation(not shown),and this may be caused by improper triggering conditions of the convection scheme(Wang and Zhang 2013).Therefore,overestimated LHFresponsesto mesoscale SSTanomaliesin the simulation may be attributable to biasesin speci f i c humidity.

Therearetwo primarymechanismsthat explain mesoscale surface wind responses:the momentum mixing mechanism(Wallace,Mitchell,and Deser 1989)and the pressure adjustment mechanism(Lindzen and Nigam 1987).When surface winds blow from the warm side to the cold side,a decrease in temperature increases the stability of the lower atmosphere and then weakens momentum mixing and decreases low-level wind speeds over cold-water areas.The convergence occurs along the wind direction.On theotherhand,wind stresscurlwillalso strengthen across the wind direction(Chelton et al.2001;O'Neill,Chelton and Esbersen 2003).Wind stress convergenceand wind stresscurlstrengthsarelinearly related to SST gradients of the along-wind and across-wind directions,respectively.These linear relationships are used to analyze whether MASIis due to the momentum mixing mechanism(Chelton et al.2004).By calculation,both in observation and simulation,these correlation coefficients arepositivealmost everywhere,and largevaluesarefound in the KE,GSARCand TIWregions;in the eastern coastal region of each basin and in the SO(Figures S2(a-d)).This suggests that the momentum mixing mechanism plays akeyroleover theseactive-eddyregions,both in observation and the high-resolution model.The simulated linear relationships have smaller values than those observed,and particularly in the ARC(Table 1).These small values can explain the underestimated coupling response of surface wind to mesoscale SSTanomalies.

For the pressure adjustment mechanism,SST anomalies can induce hydrostatic adjustment in the boundary layer,resulting in sea level pressure(SLP)adjustments and surface wind changes.Under such circumstances,wind divergence is proportional to the negative Laplacian of SLP(Lindzen and Nigam 1987;Minobe et al.2008).Therefore,the negative correlation between the Laplacian of SLP and Laplacian of SST can be used to diagnose pressure adjustment mechanisms.By calculation,over the active eddy regions(KE,GS,and ARC),the simulated negative correlations are obvious(Figure S2(e)and Table 1).This suggests that the pressure adjustment mechanism also plays a role over these active eddy regions.In the GS,the pressure adjustment is more important due to larger magnitudes than those of the KE and ARC.In the SO,the pressure adjustment plays a minor role compared with the vertical momentum mixing mechanism.

According to the above diagnostics,both the vertical momentum mixing and pressure adjustment mechanisms play key roles in surface wind speed responses to oceanic mesoscale SST over oceanic fronts and eddies(e.g.,the KE,GS,and ARC regions).It is worth noting that in some areas with high EKE,momentum mixing processes still play a role,though pressure adjustment effects are located in marginal areas such as the northwest coastal region of the Indian Ocean and along the eastern coast bordering the eastern Paci f i c.

4.Summary

In the present study,we created a high-resolution airsea coupled model with an eddy-resolving ocean model,LICOM2,based on CESM.First,the simulation capabilities of MASIare evaluated against observational data.It is found that the high-resolution coupled model can effectively reproduce MASI,i.e.,positive correlations between wind stress and mesoscale SST anomalies.Positive correlations are signi f i cant over areas where oceanic mesoscale phenomena are active.However,responses of wind stress magnitude to mesoscale SST anomalies are still weaker in the simulation than those of the observations.The different ocean components can affect the responses.Associated with weak strengths of wind stress,SHF responses to mesoscale SST are underestimated slightly,while LHF responses are overestimated due to the drier atmospheric boundary layer in the simulation.

Although different mechanisms were found in different regions,both momentum mixing and pressureadjustment mechanisms affect surface wind responses to mesoscale SSTanomaliesover oceanic fronts and mesoscaleeddies.Therolesof momentum mixing processesare more pronounced than pressure adjustments along the west coast north of the Indian Ocean and in the SO.This may be attributable to seasonal variations in large-scale wind patterns(Small,Xie,and Wang 2003).

Acknowledgments

Thanks for the suggestions and comments from the two reviewers.The main experiments were carried out on the TianHe 1 supercomputer.

Disclosure statement

No potential con f l ict of interest was reported by the authors.

Funding

This study was supported by the National Key R&D Program for Developing Basic Sciences [grant numbers 2016YFC1401401 and 2016YFC1401601]and the National Natural Science Foundation of China [grant numbers 41376026 and 41576025].