1. Introduction
Extensive effort has been recently made to produce accurate sea surface temperature (SST) data products with an aim to provide a reliable boundary condition for numerical weather prediction (NWP). For example, international collaboration such as the Global Ocean Data Assimilation Experiment (GODAE) High-Resolution SST Pilot Project attempts to produce the optimized SST product by effectively merging various satellite and in situ SST data (Donlon et al. 2007).
Although satellite measurements of SST have enabled us to produce the global SST data, they still have many limitations. For example, proper information is not available either under the cloudy condition (e.g., satellites with infrared sensors) or for the diurnal variation (e.g., polar-orbiting satellites). Furthermore, they cannot provide information of subsurface temperature profiles.
One possibility to overcome these limitations is to take advantage of the atmosphere–ocean mixed layer coupled model that is able to predict SST continuously regardless of the weather condition. The predicted SST from the model can be used as an additional source of information to various satellite and in situ SST data for more effective SST production.
In the previous work, Noh et al. (2011b) developed an atmosphere–ocean mixed layer coupled model that is able to predict the diurnal variation of SST in good agreement with satellite and in situ data. To realize the strong diurnal warming of the skin SST, which appears under weak wind and strong solar radiation, the near-surface process in the ocean mixed layer model (Noh and Kim 1999; Noh et al. 2002) is modified.
Although there have been several previous attempts to develop an atmosphere–ocean mixed layer coupled model (Price et al. 1994; Haarsma et al. 2005; Woolnough et al. 2007; Davis et al. 2008; Duan et al. 2008; Lebeaupin Brossier and Drobinski 2009), the main purpose of those models have been usually to improve the prediction of hurricanes rather than to predict the diurnal variability of SST. Meanwhile, NWP results have been used recently to improve the accuracy of satellite-derived SST by optimal estimation (Merchant et al. 2009).
To make use of an atmosphere–ocean mixed layer coupled model for the production of improved SST data, it is essential to develop an effective data assimilation (DA) technique that merges the predicted and observed SST in an optimal way. One important aspect for the assimilation of satellite SST data is the fact that a satellite measures the skin temperature (Hepplewhite 1989; Wick et al. 2002; Notarstefano et al. 2006; Kawai and Wada 2007), whereas the ocean model usually predicts the mean temperature of the uppermost grid. Another important aspect is to reflect the diurnal variation of SST, since satellite SST data are often obtained at a certain time of a day. Finally, error associated with satellite measurements and the unavailability of data under the cloud condition must be taken into account.
Although satellite SST data are now widely used for DA into the ocean model (Artale et al. 1998; Brankart et al. 2003; Fan et al. 2004; Manda et al. 2005; Martin et al. 2007; Powell et al. 2008; Pimentel et al. 2008; Ba et al. 2010; Kurapov et al. 2011), only a few efforts have attempted so far to include the effects of the diurnal variation of SST or the effects of the difference between the skin and bulk SST (Pimentel et al. 2008).
In the present work, we illustrate the possibility of utilizing the atmosphere–ocean mixed layer coupled model to produce an improved SST product in combination with observed SST data. For this purpose we developed a new DA scheme, in which the daily mean and the diurnal variation of SST are corrected in two steps. The performance of the produced SST data in this way is evaluated by the comparison with other independent satellite and buoy SST data. Finally, the sensitivity and seasonal variation of weighting factors used in DA are examined.
2. Model and simulation
a. Atmosphere model
The Weather Research and Forecasting (WRF) Model version 3.1 is used for the atmosphere model (Skamarock et al. 2008). In the present work the model has 31 vertical levels.
The microphysical scheme is the WRF single-moment 3-class scheme (Hong et al. 2004). Convection is parameterized by the new Kain–Fritsch convective scheme (Kain 2004), and the turbulence scheme is the Yonsei University (YSU) planetary boundary layer scheme (Noh et al. 2003). The radiative scheme is the Rapid Radiative Transfer Model (RRTM) (Mlawer et al. 1997) for the longwave flux and the Dudhia parameterization (Dudhia 1989) for the shortwave flux.
b. Ocean mixed layer model
The ocean mixed layer model is based on the Noh model, which reproduces well the realistic upper-ocean structure (Noh and Kim 1999; Noh et al. 2002; Hasumi and Emori 2004; Duan et al. 2008; Rascle and Ardhuin 2009) and shows a good agreement with large-eddy simulation (LES) results (Noh et al. 2004, 2011a). It is a turbulence closure model using eddy diffusivity and viscosity, similar to the Mellor–Yamada model (Mellor and Yamada 1982), but it reproduces a uniform mixed layer, consistent with bulk models (e.g., Niiler and Kraus 1977), by taking into account the effects of wave breaking and Langmuir circulation.
Under the condition of weak wind and strong solar radiation, however, the downward heat transport from the sea surface is suppressed in the absence of turbulent mixing. In this case the diurnal warming of SST, ΔSST, which is usually less than 1 K, often reaches a few degrees, and a strong temperature gradient appears near the surface (Flament et al. 1994; Merchant et al. 2008). This condition is not usually considered in the mixed layer model, but it is important to predict SST, as it produces a large ΔSST.
In the present work we will regard Ts and T1 as the skin and bulk SST, respectively. According to the detailed definitions of SST proposed by Donlon et al. (2002), the bulk SST measured at 1-m depth is called the 1-m-depth SST. On the other hand, the bulk SST in the present work represents the mean temperature over the top 1 m.
For the penetration of solar radiation, the nine-band model proposed by Paulson and Simpson (1981) and Soloviev and Schlüssel (1996) with Jerlov’s water type II are used. Very high vertical resolution using 
c. Coupled model simulation
The coupled model has a 25-km horizontal resolution and the domain covers East Asia (11°–55°N, 110°–150°E). The initial condition of SST is given by the National Centers for Environmental Prediction (NCEP) Final Analysis (FNL) data (http://rda.ucar.edu/datasets/ds083.2), and the subsurface temperature gradient is obtained from the temperature profiles of the World Ocean Atlas 2005 (WOA05) climatological data (Locarnini et al. 2006). Salinity is assumed to be uniform throughout the whole depth with the climatological sea surface salinity from the WOA05 data. The start of the model at 1800 UTC [0300 local standard time (LST)] helps to spin up the ocean until sunrise. Initial and lateral conditions of the atmosphere model are also given by the NCEP FNL data. An exchange of fluxes at the sea surface between the atmosphere and ocean mixed layer models is carried out every 120 s. Integration is carried out each day during the period 4–14 June 2008. This period, which is chosen because of relatively calm weather and strong insolation before the start of monsoon season in this region, is the same as in Noh et al. (2011b).
Performance of the coupled model with regard to the diurnal variation of SST was verified by the comparison with buoy and satellite data and regression model results, including scatterplots, time series, and probability density functions of ΔSST (Noh et al. 2011b). One can also refer to the cases without DA (in Figs. 5–10) in section 5 for the evaluation of the coupled model performance.
3. Satellite and buoy data
Satellite data used for DA are from the Multifunctional Transport Satellite-1 Replacement (MTSAT-1R; http://www.jma.go.jp/jma/jma-eng/satellite), which is a geostationary satellite with an infrared sensor. The MTSAT-1R measures SST 44 times a day, but it does not provide data under the cloudy region. In addition the data from two polar-orbiting satellites, Aqua Advanced Microwave Scanning Radiometer–Earth Observing System (AMSR-E; http://www.ssmi.com/amsr/amsr_browse.html) and the National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (NOAA AVHRR; http://www.nodc.noaa.gov/SatelliteData/pathfinder4km/available.html) are also used for the verification of DA results (Table 1). The satellite data are processed using the multichannel sea surface temperature (MCSST) method (McClain et al. 1985) to estimate SST. For cloud detection, the algorithm using dynamic thresholds by Dybbroe et al. (2005) is applied.
Characteristics of satellites.
The drifting buoy SST data available from the Global Telecommunications System (GTS) by the Data Buoy Cooperation Panel (DBCP) of the World Meteorological Organization (WMO) are also used for the verification of the DA results (http://www.jcommops.org/dbcp/data/access.html). The SST data were measured at about 1–2-m depth with the irregular sampling period, so they can be regarded as the bulk SST. One can refer to Kim et al. (2011) for the detailed information on MTSAT-1R and buoy data.
The root-mean-square differences (RMSD) of the SST from MTSAT-1R, AVHRR, and AMSR-E in comparison with the drifting buoy data are 0.75, 0.62, and 0.90 K, representatively, in the East Asian region (National Institute of Meteorological Research 2009).
To complement the lack of satellite data under the cloudy region, Gaussian interpolation was carried out. We recovered the values in the cloud region using Gaussian function in space with an e-folding scale of 1°, which is equivalent to the length scale of mesoscale eddies in the midlatitude ocean (e.g., Carter and Robinson 1987). For example, the distributions of SST at 0400 LST 5 June 2008, before and after the Gaussian interpolation, are shown in Figs. 1a and 1b. Similarly, we interpolated the number of data per a grid Nd by using the same Gaussian function (Figs. 1c,d), which will be used later for DA. The corresponding figures for the daily mean SST are also shown in Fig. 2. According to the present DA method, which will be discussed in section 4, the contribution of SST data to DA from the region with a small value of Nd—for example, Nd < 1—is insignificant. The region with Nd < 1 is much smaller for the daily mean SST than for SST at 0400 LST.
Distributions of (a),(b) SST and (c),(d) number of data per grid Nd from the MTSAT-1R at 0400 LST 5 Jun 2008 (a),(c) before and (b),(d) after Gaussian interpolation.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
Distributions of (a),(b) SST and (c),(d) number of data per grid Nd from the MTSAT-1R on 5 Jun 2008 (a),(c) before and (b),(d) after Gaussian interpolation.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
4. Data assimilation method
a. Comparison of variance between simulation and satellite data
Evaluation of uncertainties of both satellite and simulated SST data is crucial for DA. The standard deviations from the daily mean SSTs are shown in Fig. 3 for both satellite and simulated SST data. The simulated SST is resampled to the time and location of the MTSAT-1R measurement. The satellite SST data show a much larger standard deviation than the simulated SST. It indicates that the satellite data have a higher level of noise, since the amplitude of the diurnal variation of SST is comparable between two cases (Noh et al. 2011b). The high-frequency noise in the satellite data is also clearly identified from the time series of SST (see Fig. 5 in section 5a).
Distributions of standard deviation (5 Jun 2008): (a) skin SST from coupled model (CTL) and (b) SST from satellite.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
The values of error covariance are given in Table 2, separately for the cases where the number of data in a grid per day Nd is larger (and smaller) than 10. Here, the values of R and Q indicate the error covariances of observation and simulation data, and Nd = 10 roughly represents the value at which the sign of R changes from positive to negative. Note that R and Q sometimes have negative values inevitably, contrary to its definition, because of the approximations used to derive the covariance matching method (see the appendix), and in this case the values are assumed to be zero (Fukumori et al. 1999; Hirose et al. 2007).
Error covariance estimates (June 2008): (a) daily mean and (b) diurnal variation.
It indicates that the daily mean of the satellite SST is more accurate than the simulated SST in the region where Nd is larger than 10. On the other hand, the diurnal variation of the satellite SST always has a larger error than the simulated SST, regardless of Nd. This analysis suggests that it is desirable to separate DA into the daily mean and the diurnal variation of SST, as strong and weak DAs are suitable for the former and the latter.
b. Strategy for data assimilation
Based on the different characteristics between the daily mean and diurnal variation of satellite SST, discussed above, we carry out the assimilation of satellite SST data in two steps. In the first step, DA is carried out to correct the daily mean SST from the difference of satellite and simulated SST. The correction of the daily mean SST is then applied to modify the initial condition for the restart of the model. In the second step, DA is carried out to the diurnal variation of the skin SST by nudging the anomalies of the modeled skin SST from the daily mean toward those of the satellite SST every 30 min. The first and second methods follow the basic concepts of the four-dimensional variational data assimilation (4DVAR) method and the Kalman filter method, respectively. The present method is a simple and effective way of correcting the simulated SST toward the observed SST, while requiring much less computational cost than other optimal-class methods.
Besides, the daily mean and the diurnal variation of SST are assimilated to the bulk and skin SST, respectively, because algorithms of satellite SST are usually tuned by using buoy data, but its diurnal variation reflects that of the skin SST (Stuart-Menteth and Robinson 2003; Kawai and Kawamura 2005). Satellite data were gridded to the model grid for DA.
c. Daily mean bias correction as the first DA
Figure 4 shows the difference between satellite and simulated SST and the correction of the daily mean SST 
Differences of SST: (a) 
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
d. Sequential SST anomaly correction as the second DA
5. Results
a. Effects of DA
Figure 5 compares the time series of SST at several locations on 5 June 2008. Figures 5a and 5b show the cases in which a good agreement is found between satellite (MTSAT-1R) and simulated SST (CTL), except for the high-frequency noise appearing in satellite data. On the other hand, Fig. 5c represents the case in which the daily mean SSTs are different, and Fig. 5d represents the case in which both the daily mean and the diurnal variation of SST are different. As expected, the first DA (DA-1) contributes significantly to adjust CTL to MTSAT-1R in Figs. 5c and 5d, and the second DA (DA-2) contributes to the further improvement in Fig. 5d. Meanwhile, it is observed that DA-2 tends to pick up the noise of MTSAT-1R in all cases.
Time series of SST from MTSAT-1R (black line), CTL (blue line), DA-1 (orange line), and DA-2 (red line) (5 Jun 2008): (a) 20°N, 130°E; (b) 25°N, 146°E; (c) 46°N, 145°E; and (d) 45°N, 145°E.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
Figure 6 shows the correlation coefficient and RMSD between the diurnal variation of SST from MTSAT-1R and from CTL, DA-1, and DA-2 on the same day. The domain-averaged values of the correlation coefficient, RMSD, and bias are also shown in Table 3. RMSD and the bias are improved mainly after DA-1, while the correlation is improved after DA-2. It implies that the model bias of SST is corrected by DA-1, and the diurnal variation of SST becomes more realistic after DA-2.
Distributions of (a),(c),(e) correlation coefficient and (b),(d),(f) RMSD of SST diurnal variation with respect to MTSAT-1R (5 Jun 2008): (a),(b) CTL, (c),(d) DA-1, and (e),(f) DA-2.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
The spatial mean correlation coefficient, RMSD, and bias of SST from CTL, DA-1, and DA-2 with respect to MTSAT-1R, obtained from the average over 11 days.
To assess the effects of DA, we compare the distributions of the daily mean SST and the diurnal amplitude of SST, defined by the SST difference between 0900 and 1500 LST, from MTSAT-1R, CTL, and DA-2 (Figs. 7 and 8). One can observe that many finescale structures from MTSAT-1R are realized after DA. Note that in the case of DA-1, the daily mean SST is similar to DA-2 and that the diurnal amplitude of SST is equivalent to CTL.
Distributions of daily mean SST: (a) MTSAT-1R, (b) MTSAT-1R after Gaussian interpolation, (c) CTL, and (d) DA-2 (5 Jun 2008).
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
Distributions of diurnal SST variation: (a) MTSAT-1R, (b) CTL, and (c) DA-2 (5 Jun 2008).
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
b. Assessment of data assimilation
To assess the performance of DA, the comparison with independent sources of measurements is necessary. For this purpose we compare the SST data with other satellite data (AVHRR, AMSR-E) and buoy data. Table 4 shows the spatial mean bias and RMSD with AVHRR, AMSR-E, and buoy data for the cases of MTSAT-1R, CTL, DA-1, and DA-2, obtained from the average over 11 days (4–14 June 2008). The SST data from the model and MTSAT-1R are interpolated to the time and location of each measurement (AVHRR, AMSR-E, and buoy data) for the calculation.
The spatial mean (a) bias and (b) RMSD of SST from CTL, DA-1, and DA-2 with respect to AVHRR, AMSR-E, and buoy data, obtained from the average over 11 days.
It is remarkable to find that the bias and RMSD of DA-1 are generally smaller than those of both MTSAT-1R and CTL, thus validating the contribution of DA. DA-2 is also found to contribute to the decrease of RMSD slightly. For example, DA-1 contributes to the decrease of RMSD to 81%, 81%, and 84% with respect to AVHRR, AMSR-E, and buoy data, respectively; and DA-2 contributes to the decrease of RMSD further to 80%, 79%, and 82%, respectively. Distributions of bias and RMSD of MTSAT-1R, CTL, and DA-2 with respect to AMSR-E (Fig. 9) and the time series of the domain average of these values during the period 4–14 June (Fig. 10) help us grasp the contribution of DA.
Distributions of (a),(c),(e) bias and (b),(d),(f) RMSD with respect to AMSR-E (5 Jun 2008): (a),(b) MTSAT-1R, (c),(d) CTL, and (e),(f) DA-2.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
Variations of the domain average of (a) bias and (b) RMSD with respect to AMSR-E during the period (4–14 Jun 2008): MTSAT (black line), CTL (blue line), and DA-2 (red line).
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
Note that the two-step approach in the present DA is based on the assumption that the daily mean SST of the MTSAT-1R data is more accurate than the model data, when 
The spatial mean (a) bias and (b) RMSD of the daily mean SST from MTSAT-1R, CTL, DA-1, and DA-2 with respect to the daily mean SST of buoy data for grid points with 
From this estimation one can expect that DA-1 decreases RMSD from both MTSAT-1R and CTL (Table 4), by filtering out the noise during the diurnal cycle in MTSAT-1R (Figs. 3, 5) and by adjusting the daily mean SST in CTL (Table 5).
c. Sensitivity to the weighting factor
The sensitivity to the values of w in (10) is discussed in this section. The satellite data include noisy signals. The DA result can become even closer to MTSAT-1R with the stronger w in (10), but it tends to pick up more noises rather than real signals. Figure 11 compares the diurnal variation from DA-2 with 
Time series of SST from MTSAT-1R (black line) and DA-2 with wmax = 0.1 (purple line), 0.35 (red line), and 0.5 (green line) (5 Jun 2008): (a) 20°N, 130°E; (b) 25°N, 146°E; (c) 46°N, 145°E; and (d) 45°N, 145°E.
Citation: Journal of Atmospheric and Oceanic Technology 30, 12; 10.1175/JTECH-D-12-00238.1
The spatial mean (a) bias and (b) RMSD of SST from DA-2 (wmax = 0.0, 0.1, 0.35, 0.5) with respect to AVHRR, AMSR-E, and buoy data (5 Jun 2008).
d. Seasonal variation of Nd and wmax
The present work suggest the optimal values for Nd and wmax, as Nd = 10 and wmax = 0.35, but it is important to estimate how the values of Nd and wmax vary depending on season. For this purpose we applied the covariance matching method for 2 days in January and July, representing the winter and monsoon seasons, respectively.
Tables 7 and 8 indicate that the value of Nd = 10 can be applied to other seasons as well. On the other hand, the estimation of 
Error covariance estimates (January 2008): (a) daily mean and (b) diurnal variation.
Error covariance estimates (July 2008): (a) daily mean and (b) diurnal variation.
6. Conclusions
The present work demonstrates that the atmosphere–ocean mixed layer coupled model can be applied to the production of SST data by combining with the assimilation of satellite SST data. The WRF and the Noh model are used for the atmosphere and the ocean mixed layer models, respectively.
A new DA scheme is developed, in which SST are corrected in two steps: the daily mean SST bias correction in the first DA and the sequential SST anomaly correction in the second DA. It is based on the estimation from the covariance matching method that the daily mean SST of satellite data are more accurate than the model data, when the number of data in a grid per day 
Statistical comparison of satellite (MTSAT-1R), model (CTL), and DA results with independent satellite and buoy data, such as correlation coefficient, RMSD, and bias, reveals that the results after DA are more accurate than both MTSAT-1R and CTL SST data. The present two-step DA scheme provides an efficient approach without heavy computational cost, which is ideal for the operational purpose. Furthermore, it is found that the daily mean SST from MTSAT-1R is more accurate than that of CTL, thus supporting the estimation from the covariance matching method in section 4a.
The sensitivity of the weighting factor wmax used in the second DA is examined, which confirms its estimation based on the covariance matching method. Meanwhile, the covariance matching method applied to January and July reveals that the value of wmax remains unchanged during summer, but it is smaller in winter, which implies that the second DA is less likely to contribute in winter.
Although the present work illustrates the approach of using the atmosphere–ocean mixed layer coupled model to improve the SST data successfully, further improvement is necessary for practical application in both model development and data assimilation technique. For example, elaboration of the DA scheme, such as the optimization of wmax depending on various regional and seasonal conditions, can help improve the accuracy of the produced SST. It may also be necessary to extend the coupled model so as to include the effects of the Ekman transport that is generated under the typhoon condition. Finally, the further improvement in the accuracy of the produced SST can be obtained by utilizing the up-to-date technology of producing the SST product based on the assimilation of various satellite and buoy data.
Acknowledgments
This study was supported by the Korea Meteorological Administration Research and Development Program under Grant CATER 2012-6090 and the National Research Foundation of Korea Grant funded by the Korean government (MEST) (Grant NRF-2009-C1AAA001-0093068). One of the authors is partially supported by the Japan Society for the Promotion of Science KAKEN Grant 22106002. We also express our gratitude to the National Institute of Meteorological Research (NIMR) in South Korea for providing us satellite and buoy data, D. H. Kim for helping with the code implementation, and Jihye Lee for analyzing buoy SST data. Calculations were performed by using the supercomputing resources of the KISTI.
APPENDIX
Evaluation of the Error Covariance Using the Covariance Matching Method
One way to evaluate the uncertainties of satellite and simulated data is the covariance matching method proposed by Fu et al. (1993), which has been widely used to estimate data error (Fukumori et al. 1999; Menemenlis and Chechelnitsky 2000; Hirose et al. 2007; Ueno and Nakamura 2013).
REFERENCES
Artale V., D’Ortenzio F. , Iudicone D. , Marulloa S. , Rupolo V. , and Santolerib R. , 1998: Assimilation of satellite AVHRR SST in an OGCM of the Mediterranean Sea: Data processing, new parametrizations, and physical results. Earth Surface Remote Sensing II, G. Cecchi and E. Zilioli, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 3496), doi:10.1117/12.332715.
Ba, S. O., Corpetti T. , Chapron B. , and Fablet R. , 2010: Variational data assimilation for missing data interpolation in SST images. Proc. 2010 IEEE International Geoscience and Remote Sensing Symp., IEEE, 264–267, doi:10.1109/IGARSS.2010.5649206.
Bellenger, H., and Duvel J.-P. , 2009: An analysis of tropical ocean diurnal warm layers. J. Climate, 22, 3629–3646.
Brankart, J.-M., Testut C.-E. , Brasseur P. , and Verron J. , 2003: Implementation of a multivariate data assimilation scheme for isopycnic coordinate ocean models: Application to a 1993–1996 hindcast of the North Atlantic Ocean circulation. J. Geophys. Res., 108, 3074, doi:10.1029/2001JC001198.
Carter, E. F., and Robinson A. R. , 1987: Analysis models for the estimation of oceanic fields. J. Atmos. Oceanic Technol., 4, 49–74.
Davis, C. A., and Coauthors, 2008: Prediction of landfalling hurricanes with the Advanced Hurricane WRF model. Mon. Wea. Rev., 136, 1990–2005.
Donlon, C. J., Minnett P. J. , Gentemann C. , Nightingale T. J. , Barton I. J. , Ward B. , and Murray M. J. , 2002: Toward improved validation of satellite sea surface skin temperature measurements for climate research. J. Climate, 15, 353–369.
Donlon, C. J., and Coauthors, 2007: The Global Ocean Data Assimilation Experiment High-Resolution Sea Surface Temperature Pilot Project. Bull. Amer. Meteor. Soc., 88, 1197–1213.
Duan, A., Sui C. , and Wu G. , 2008: Simulation of local air-sea interaction in the great warm pool and its influence on Asian monsoon. J. Geophys. Res., 113, D22105, doi:10.1029/2008JD010520.
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 3077–3107.
Dybbroe, A., Karllson K.-G. , and Thoss A. , 2005: NWCSAF AVHRR cloud detection and analysis using dynamic thresholds and radiative transfer modeling. Part I: Algorithm description. J. Appl. Meteor., 44, 39–54.
Fan, S., Oey L.-Y. , and Hamilton P. , 2004: Assimilation of drifter and satellite data in a model of the northeastern Gulf of Mexico. Cont. Shelf Res., 24, 1001–1013.
Flament, C. W., Firing J. , Sawyer M. , and Trefois C. , 1994: Amplitude and horizontal structure of a large diurnal sea surface warming event during the Coastal Ocean Dynamics Experiment. J. Phys. Oceanogr., 24, 124–139.
Fu, L.-L., Fukumori I. , and Miller R. N. , 1993: Fitting dynamic models to the Geosat sea level observations in the tropical Pacific Ocean. Part II: A linear, wind-driven model. J. Phys. Oceanogr., 23, 2162–2181.
Fukumori, I., Raghunath R. , Fu L.-L. , and Chao Y. , 1999: Assimilation of TOPEX/Poseidon altimeter data into a global ocean circulation model: How good are the results? J. Geophys. Res., 104 (C11), 25 647–25 665.
Haarsma, R. J., Campos E. J. D. , Hazeleger W. , Severijns C. , Piola A. , and Molteni F. , 2005: Dominant modes of variability in the South Atlantic: A study with a hierarchy of ocean–atmosphere models. J. Climate, 18, 1719–1735.
Hasumi, H., and Emori S. , Eds., 2004: Coupled GCM (MIROC) description. K-1 Tech. Rep. 1, Center for Climate System Research, University of Tokyo, 34 pp. [Available online at http://ccsr.aori.u-tokyo.ac.jp/~hasumi/miroc_description.pdf.]
Hepplewhite, C. L., 1989: Remote observation of the sea surface and atmosphere: The oceanic skin effect. Int. J. Remote Sens., 10, 801–810.
Hirose, N., Kawamura H. , Lee H. J. , and Yoon J.-H. , 2007: Sequential forecasting of the surface and subsurface conditions in the Japan Sea. J. Oceanogr., 63, 467–481.
Hong, S. Y., Dudhia J. , and Chen S.-H. , 2004: A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation. Mon. Wea. Rev., 132, 103–120.
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181.
Kawai, Y., and Kawamura H. , 2005: Spatial and temporal variations of model-derived diurnal amplitude of sea surface temperature in the western Pacific Ocean. J. Geophys. Res., 110, C08012, doi:10.1029/2004JC002652.
Kawai, Y., and Wada A. , 2007: Diurnal sea surface temperature variation and its impact on the atmosphere and ocean: A review. J. Oceanogr., 63, 721–744.
Kim, M.-J., Ou M.-L. , Sohn E.-H. , and Kim Y. , 2011: Characteristics of sea surface temperature retrieved from MTSAT-1R and in-situ data. Asia-Pac. J. Atmos. Sci., 47, 421–427.
Kurapov, A. L., Foley D. , Strub P. T. , Egbert G. D. , and Allen J. S. , 2011: Variational assimilation of satellite observations in a coastal ocean model off Oregon. J. Geophys. Res., 116, C05006, doi:10.1029/2010JC006909.
Lebeaupin Brossier, C., and Drobinski P. , 2009: Numerical high-resolution air-sea coupling over the Gulf of Lions during two tramontane/mistral events. J. Geophys. Res., 114, D10110, doi:10.1029/2008JD011601.
Locarnini, R. A., Mishonov A. V. , Antonov J. I. , Boyer T. P. , and Garcia H. E. , 2006: Temperature. Vol. 1, World Ocean Atlas 2005, NOAA Atlas NESDIS 61, 182 pp.
Manda, A., Hirose N. , and Yanagi T. , 2005: Feasible method for the assimilation of satellite-derived SST with an ocean circulation model. J. Atmos. Oceanic Technol., 22, 746–756.
Martin, M. J., Hines A. , and Bell M. J. , 2007: Data assimilation in the FOAM operational short-range ocean forecasting system: A description of the scheme and its impact. Quart. J. Roy. Meteor. Soc., 133, 981–995.
McClain, E. P., Pichel W. P. , and Walton C. C. , 1985: Comparative performance of AVHRR-based multichannel sea surface temperature. J. Geophys. Res., 90 (C6), 11 587–11 601.
Mellor, G. L., and Yamada T. , 1982: Development of a turbulent closure model for geophysical fluid problems. Rev. Geophys. Space Phys., 20, 851–875.
Menemenlis, D., and Chechelnitsky M. , 2000: Error estimates for an ocean general circulation model from altimeter and acoustic tomography data. Mon. Wea. Rev., 128, 763–778.
Merchant, C. J., Filipiak M. J. , Le Borgne P. , Roquet H. , Autret E. , Piollé J.-F. , and Lavender S. , 2008: Diurnal warm-layer events in the western Mediterranean and European shelf seas. Geophys. Res. Lett., 35, L04601, doi:10.1029/2007GL033071.
Merchant, C. J., Le Borgne P. , Roquet H. , and Marsouin A. , 2009: Sea surface temperature from a geostationary satellite by optimal estimation. Remote Sens. Environ., 113, 445–457.
Mlawer, E. J., Taubnam S. J. , Brown P. D. , Iacono M. J. , and Clough S. A. , 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated k-model for the longwave. J. Geophys. Res., 102 (D14), 16 663–16 682.
National Institute of Meteorological Research, 2009: National Institute of Meteorological Research annual report: Research for the meteorological observation technology and its application. NIMR, 215 pp.
Niiler, P. P., and Kraus E. B. , 1977: One-dimensional models of the upper ocean. Modeling and Prediction of the Upper Layers of the Ocean, E. B. Kraus, Ed., Pergamon Press, 143–172.
Noh, Y., and Kim H. J. , 1999: Simulation of temperature and turbulence structure of the oceanic boundary layer with the improved near-surface process. J. Geophys. Res., 104 (C7), 15 621–15 634.
Noh, Y., Jang C. J. , Yamagata T. , Chu P. C. , and Kim C.-H. , 2002: Simulation of more realistic upper-ocean process from an OGCM with a new ocean mixed layer model. J. Phys. Oceanogr., 32, 1284–1307.
Noh, Y., Cheon W. G. , Hong S.-Y. , and Raasch S. , 2003: Improvement of the K-profile model for the planetary boundary layer based on large eddy simulation data. Bound.-Layer Meteor., 107, 401–427.
Noh, Y., Min H. S. , and Raasch S. , 2004: Large eddy simulation of the ocean mixed layer: The effects of wave breaking and Langmuir circulation. J. Phys. Oceanogr., 34, 720–735.
Noh, Y., Goh G. , and Raasch S. , 2011a: Influence of Langmuir circulation on the deepening of the wind-mixed layer. J. Phys. Oceanogr., 41, 472–484.
Noh, Y., Lee E. , Kim D.-H. , Hong S.-Y. , Kim M.-J. , and Ou M.-L. , 2011b: Prediction of the diurnal warming of sea surface temperature using an atmosphere-ocean mixed layer coupled model. J. Geophys. Res., 116, C11023, doi:10.1029/2011JC006970.
Notarstefano, G., Mauri E. , and Poulain P.-M. , 2006: Near-surface thermal structure and surface diurnal warming in the Adriatic Sea using satellite and drifter data. Remote Sens. Environ., 101, 194–211.
Paulson, C. A., and Simpson J. J. , 1981: The temperature difference across the cool skin of the ocean. J. Geophys. Res., 86 (C11), 11 044–11 054.
Pimentel, S., Haines K. , and Nichols N. K. , 2008: The assimilation of satellite-derived sea surface temperatures into a diurnal cycle model. J. Geophys. Res., 113, C09013, doi:10.1029/2007JC004608.
Powell, B. S., Arango H. G. , Moore A. M. , Di Lorenzo E. , Milliff R. F. , and Foley D. , 2008: 4DVAR data assimilation in the Intra-Americas Sea with the Regional Ocean Modeling System (ROMS). Ocean Modell., 25, 173–188.
Price, J. F., Sanford T. , and Forristall G. Z. , 1994: Forced stage response to a moving hurricane. J. Phys. Oceanogr., 24, 233–260.
Rascle, N., and Ardhuin F. , 2009: Drift and mixing under the ocean surface revisited: Stratified conditions and model-data comparisons. J. Geophys. Res., 114, C02016, doi:10.1029/2007JC004466.
Skamarock, W. C., 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3_bw.pdf.]
Soloviev, A. V., and Schlüssel P. , 1996: Evolution of cool skin and direct air-sea gas transfer coefficient during daytime. Bound.-Layer Meteor., 77, 45–68.
Stuart-Menteth, A. C., and Robinson I. S. , 2003: A global study of diurnal warming using satellite-derived sea surface temperature. J. Geophys. Res., 108, 3155, doi:10.1029/2002JC001534.
Takaya, Y., Bildot J.-R. , Beljaars A. C. M. , and Janssen P. A. E. M. , 2010: Refinements to a prognostic scheme of skin sea surface temperature. J. Geophys. Res., 115, C06009, doi:10.1029/2009JC005985.
Ueno, G., and Nakamura N. , 2013: Iterative algorithm for maximum-likelihood estimation of the observation-error covariance matrix for ensemble-based filters. Quart. J. Roy. Meteor. Soc., doi:10.1002/qj.2134, in press.
Wick, G. A., Bates J. J. , and Scott D. J. , 2002: Satellite and skin-layer effects on the accuracy of sea surface temperature measurements from the GOES satellites. J. Atmos. Oceanic Technol., 19, 1834–1848.
Woolnough, S. J., Vitart F. , and Balmaseda M. A. , 2007: The role of the ocean in the Madden-Julian Oscillation: Implications for MJO prediction. Quart. J. Roy. Meteor. Soc., 133, 117–128.
Zeng, X., and Beljaars A. , 2005: A prognostic scheme of sea surface skin temperature for modeling and data assimilation. Geophys. Res. Lett., 32, L14605, doi:10.1029/2005GL023030.