Journal Search Engine
Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 1225-6692(Print)
ISSN : 2287-4518(Online)
Journal of the Korean earth science society Vol.42 No.4 pp.401-411

Assessment of New High-resolution Regional Climatology in the East/Japan Sea

Jae-Ho Lee, You-Soon Chang*
Department of Earth Science Education, Kongju National University, Kongju 32588, Korea
*Corresponding author: Tel: +82-41-850-8292, Fax: +82-41-850-8299
July 30, 2021 August 21, 2021 August 23, 2021


This study provides comprehensive assessment results for the most recent high-resolution regional climatology in the East/Japan Sea by comparing with the various existing climatologies. This new high-resolution climatology is generated based on the Optimal Interpolation (OI) method with individual profiles from the World Ocean Database and gridded World Ocean Atlas provided by the National Centers for Environmental Information (NCEI). It was generated from the recent previous study which had a primary focus to solve the abnormal horizontal gradient problem appearing in the other high-resolution climatology version of NCEI. This study showed that this new OI field simulates well the mesoscale features including closed-curve temperature spatial distribution associated with eddy formation. Quantitative spatial variability was compared to the other four different climatologies and significant variability at 160 km was presented through a wavelet spectrum analysis. In addition, the general improvement of the new OI field except for warm bias in the coastal area was confirmed from the comparison with serial observation data provided by the National Fisheries Research and Development Institute’s Korean Oceanic Data Center.



    The three-dimensional oceanic field is used as an initial/boundary condition and verification source of numerical ocean model. It also has known that the accuracy of the initial and boundary data has a great influence on the ocean/atmospheric prediction performance from several days up to several decades later (Smith et al., 2007). They are two types of three-dimensional ocean fields, one is the reanalysis data from the data assimilation system and the other is the objective analyzed product. While reanalysis data based on numerical model require complex processes such as observation data collection, quality control, model configuration, and assimilation algorithm application, studies on the development of objective analyzed fields without relying on numerical models have actively carried out with relatively simple algorithms and recent increased observations (Hosoda et al., 2008;Chang et al., 2009;Roemmich and Gilson, 2009;Carnes et al., 2010;Chang and Shin, 2012;Gouretski, 2019).

    To observe global ocean status in real-time, the Argo project successfully started providing real-time data on temperature and salinity in the 21st century. Therefore, with the increase of ocean observation data, an objective analysis field by using only observation data is continuously being developed by various institutions. The climatological mean field for the three-dimensional temperature and salinity structures was first created in 1982 by the Levitus group of National Centers for Environmental Information (NCEI) (Levitus, 1982). The global ocean climatological mean field developed under the name of the World Ocean Atlas (WOA) was continuously updated in 1994 (NOAA, 1994; WOA94), 1998 (NOAA, 1998; WOA98), 2001 (Levitus, 2002; WOA01), 2005 (Levitus, 2006; WOA05), 2009 (Levitus, 2009; WOA09), 2013 (Levitus, 2013; WOA13), and 2018 (Boyer et al., 2018; WOA18) as observational data were steadily acquired. Prior products to WOA13 had a horizontal resolution of 1° grid at the 33 vertical standard depth levels, but WOA13 and WOA18 increased the vertical resolution to 102 vertical standard depth levels and additionally provided a horizontal resolution of 0.25° grid. Recently, highresolution regional climatology mean fields with 0.1° grid have been produced for 9 major regions (Southwest North Atlantic, Northwest Atlantic, GIN Seas, Northern North Pacific, Northeast Pacific, Nordic Sea, Arctic Ocean, Gulf of Mexico, and East Asian Seas). Not only NCEI but also the U.S. Naval Research Laboratory (NRL) has created data under the name of Generalized Digital Environmental Model (GDEM) with 0.25° resolution and is continuously updating GDEM2 (Teague et al., 1990), GDEM3 (Carnes, 2009) and GDEM4 (Carnes et al., 2010) versions. The Integrated Climate Data Center of Hamburg University (ICDC) also provides the World Ocean Experiment Argo Global Hydrographic Climatology (hereafter WAGHC) with 0.25° resolution generated on isobaric and isopycnal surfaces (Gouretski, 2018, 2019). However, there are various errors in the production of such a climatology mean field because there are differences in spatial resolution, available data, quality control method of observation data, applied objective analysis algorithm, and corresponding de-correlation length scale.

    Chang and Shin (2012) developed high-resolution objectively analysis fields in the southwestern East/ Japan Sea, and compared it with various climatology mean fields (WOA01q, GDEM3, WOA05, WOA09). They confirmed that each data showed very different characteristics through the annual mean temperature and salinity distribution at 100 m depth. Especially, WOA05 and WOA09 with 1° resolution did not simulate the oceanic conditions of the coastal area of the East/Japan Sea and the northern coastal area of Kyushu. WOA01q and GDEM3 with 0.25° resolution also showed that there was a large difference in the horizontal distribution of temperature and salinity. Chang et al. (2014) compared the monthly mean upper ocean temperature and salinity variability from 2008 to 2011 for five types of objective analyzed fields (EN3 (Ensemble version 3 provided by the Met Office Hadley Center), GFDL (Geophysical Fluid Dynamics Laboratory), IPRC (International Pacific Center), JAMSTEC (Japan Agency for Marine-Earth Science and Technology), and SIO (Scripps Institution of Oceanography)). As a result, it was confirmed that there was a significant deviation between the data according to the type of background field used in the objective analysis process, the radius of influence, and the difference in the algorithm. Chang and Shin (2014) also found a problem with the vertical gradient in the high-resolution East Asian Seas Regional Climatology (EASRC) developed by NCEI. In particular, this problem appeared near the bottom areas where the topographic change was steep and observation data was scarce, and the problem was solved through the vertical gradient correction method. In this regard, recently, Lee and Chang (2021) calculated geostrophic current in the East/Japan Sea, which is a part of the East Asian Seas, using EASRC. As a result, it was confirmed that strong zonal current flows appeared at a 1° latitude interval between 38 and 41°N. So, they calculated the temperature difference in the meridional direction (dT/dy), and found that an abnormal horizontal temperature gradient appeared at 1° interval.

    They developed a high-resolution regional climatology with 0.1° resolution in which the horizontal gradient problem was solved by applying a wider (211 km) radius circle through the optimal interpolation (OI) method (hereafter new OI). However, Lee and Chang (2021) only compared the high-resolution climatology developed by the OI method with the EASRC in relation to solving the horizontal gradient problem, but did not provide comprehensive assessment result with other climatologies.

    Therefore, by comparing with the existing climatologies including EASRC in the East/Japan Sea, this study verified the high-resolution climatology (new OI) that solved the horizontal gradient problem developed by Lee and Chang (2021). Main results of 10 m and 100 m depth in February and August are presented, respectively. The following section describes the utilized data sets and OI method by Lee and Chang (2021) in detail. Section 3 presents intercomparison between the new OI and existing climatologies and comparison with serial observations, respectively. Section 4 provides a summary and discussion of the results.

    Data and Methods


    The new OI was produced using the World Ocean Database 2013 (hereafter WOD13) and WOA13 with 1° resolution as the latest version of EASRC. All data were collected from NCEI. In the WOD13, we used almost all instrument types (Ocean Station Data (OSD), Mechanical Bathythermographs (MBT), Expendable Bathythermograph Data (XBT), High-resolution Conductivity-Temperature-Depth (CTD), Drifting Buoy Data (DRB), Moored Buoy Data (MRB), Profiling Floats Data (PFL), Undulating Ocean Recorder Data (UOR), Glider Data (GLD), https://www.ncei.noaa. gov/products/world-ocean-database) after Quality Control (QC) procedure based on the NCEI technical report (Boyer and Levitus, 1994) and previous research (Chang et al., 2009). The QC process consists of the QC flag check, duplication check for pressure and cycle, stability check for temperature and salinity, density inversion check.

    Total 1,443,820 profiles from WOD13 standard depth data for the observation period from 1955 to 2012 are used. Among them, 1.9% (27,278 profiles) were removed by the QC process, and final 1,416,542 were used as observation source. New OI also used the monthly mean of WOA13 with 1° resolution in 24 vertical levels from the surface to a depth of 1500 m as background data to interpolate even in areas where observation data is scarce (Lee and Chang, 2021).

    For the comparison with existing climatologies, we used EASRC, WAGHC, WOA13q, and WOA13 at 10 m and 100 m depth in February and August. In order to provide verification results with serial observation, we used two different serial observation datasets. One is the 102, 103, 104, 105, 106, and 107 lines in the southwestern East/Japan Sea operated by the National Fisheries Research and Development Institute’s Korean Oceanic Data Center (NIFS/KODC). The other serial datasets are from PM and G lines in the central and eastern parts of the East/Japan Sea maintained by the Japan Metrological Agency (JMA).


    Lee and Chang (2021) produced a gridded temperature field at each depth using the OI method. An objective estimate of the temperature at each grid point is calculated as follows.

    T ( S ) o b j = d + ω ( d d )

    where d represents a set of observed WOD13 and gridded WOA13 temperature profiles within the influence radius established on the grid point being interpolated and <d> denotes the simple arithmetic mean value of the set d within the influence radius.

    Each historical WOD13 data (d) can be separated into a true signal (s) and some random noise (η), and the signal and noise variance of the data can be approximated from the relationship d = s + η (Fukumori and Wunsch, 1991). The signal variance is approximated by s = ( 1 / N ) i ( d i < d > ) 2 , where N is the number of data points, the noise variance is estimated from the data difference of neighboring stations, it is calculated as η = ( 1 / ( 2 N ) ) i ( d i d j ) 2 , where dj is the data point that has the shortest distance from di. Further ω is the weighting matrix, expressed as ω = C d g [ C d d + I < η 2 > ] 1 (McIntosh, 1990), where I denotes the identity matrix, Cdg denotes the datagrid covariance matrix, and Cdd denotes the data-data covariance matrix which is expressed as:

    C d g i ( x , y ) = s 2 exp { [ D x i , g 2 / L x 2 + D y i , g 2 / L y 2 ] }

    C d d i , j ( x , y ) = s 2 exp { [ D x i , j 2 / L x 2 + D y i , j 2 / L y 2 ] }

    The covariance is a function of the spatial length scale (Lx, Ly), where Dx and Dy are the spatial distances between the observed data (subscripted with i) and the grid point (subscripted with g) in zonal and meridional directions, respectively.

    As followed by Lee and Chang (2021), this study applied an influence radius of 211 by 211 km through a sensitivity experiment. Figure 1 shows a total number of data within the influence radius of 211 by 211 km on every grid point. We can see that enough data more than 100 profiles are used except for very small part of the coastal areas. Particularly, in the central East/Japan Sea and along the northeastern direction of the Tsushima Warm Current, a maximum of 2,900 profiles at 10 m depth and more than 1,000 profiles at 100 m depth were used within the influence radius of 211 by 211 km, respectively. As expected, summer has more observational data than winter season, but it does not significantly affect the results because of the plenty of data in the winter season as well.


    Intercomparison with existing climatologies

    We compared the spatial variability between new OI and existing climatologies (EASRC, WAGHC, WOA13q, and WOA13). Figure 2 shows the temperature distribution from four different climatologies around the East/Japan Sea at 100 m depth in February. The top panel shows the study area, the entire of the East/ Japan Sea, and the bottom panel represents the southwestern East/Japan Sea where a high-resolution climatology has previously been produced (Chang and Shin, 2012). All climatologies simulate well the general large-scale temperature distribution in the East/ Japan Sea (Fig. 2(a~e)). However, various small-scale features, such as meandering patterns along the isotherm, are evident only in the new OI, EASRC, and WAGHC. Specifically, in the southwestern part of the East/Japan Sea (Fig. 2(f~j)), the new OI shows relatively low temperatures along the eastern coast of Korea. The EASRC and WAGHC also reveal similar patterns. However, this pattern along the coastal area has been smoothed out in the WOA13q and WOA13. A notable difference between the new OI and other climatologies is that it presents a closed curve in a relatively low-temperature region near the eastern coast of North Korea. This closed curve shape is important in the formation of cold eddies, while other climatologies have depicted no similar pattern.

    One of the advantages of high-resolution climatology is its capacity to reproduce various meso-scale features. Therefore, we performed a wavelet analysis to compare the spatial variability of meso-scale features from new OI and other climatologies. By presenting spectral density function, we can check the spatial variability quantitatively (Fig. 3). Wavelet transform usually analyzes spectral characteristics over a period of time. However, in this study, we used it to analyze the characteristics of space by applying distance instead of time. The southwestern (northeastern) part of the East/ Japan Sea was defined as the first (last) data position and the number of data positions was increased in the meridional direction. For quantitative comparison, we simply re-grided the three climatologies (WAGHC, WOA13q, and WOA13) to a 0.1° grid, and extracted only points that match the two high-resolution climatologies (new OI and EASRC). Also, in order to focus on the meso-scale features, we extracted the temperature anomaly by 230 km high-pass filtering.

    Figure 3a shows the spatial-averaged spectral density function of the temperature anomaly at 10m depth in February from the new OI with a 95% confidence level. In the spatial-averaged spectral density function, a peak has appeared at a diameter of approximately 160 km which satisfies a 95% confidence level. For the EASRC and WAGHC, peaks appeared at a diameter of approximately 160 km, similar to the new OI (Fig. 3(b) and (c)). However, the magnitude of EASRC (0.18) and WAGHC (0.23) was weaker than that of new OI (0.37) in the 95% confidence level. In particular, EASRC showed weaker spatial variability than WAGHC despite of high resolution. This is considered to be related to the horizontal gradient problem of EASRC as suggested by Lee and Chang (2021). WOA13q and WOA13 did not exhibit peaks in the spatial averaged spectral density function around 160 km within the 95% confidence level. This is because the temperature distribution is smoother than in other climatologies. Therefore, we can confirmed that the new OI well simulates the spatial variability in the meso-scale batter than other climatologies.

    Comparison with serial observation data

    Based on the serial hydrographic lines that have been observed for a long time in the East/Japan Sea, we performed validation on the new OI and four different existing climatologies. One of the serial observation datasets is the observation line provided by the National Fisheries Research and Development Institute’s Korean Oceanic Data Center (NIFS/KODC), and verification was performed through the 102, 103, 104, 105, 106, and 107 lines operating in the southwestern East/Japan Sea. The others have been verified using the PM and G lines that have been performed observations by the Japan Metrological Agency (JMA) for the eastern East/Japan Sea. They have been simply averaged at each station from 1983 to 2010 for NIFS/KODC data and 1997 to 2010 for JMA data at 10 m depth and 100 m depth in February and August, respectively. For quantitative comparison and verification, the five climatologies including the new OI were re-grid to a 0.01° grid, and the grid points matched with the observation points were extracted.

    Figure 4 shows the spatial distribution of temperature bias at 10 m depth in February. Compared with the observation data, four climatologies (new OI: −0.44°C, EASRC: −0.61°C, WOA13q: −0.90°C, WOA13: −0.65 °C) except WAGHC (0.12°C) show cold bias. These results are related to the averaged period of the data used for each climatology. The EASRC, WOA13q, and WOA13 including new OI were produced based on WOD13, and because they contain many observations much earlier than 1983 and 1997, so they reflect less of the recent sea warming signal in the East/Japan Sea. However, WAGHC is a global climatology produced based on WOD data from 1985 to 2016 (Gouretski, 2018), it reflects the recent sea warming signal relatively larger than the other four climatologies. It should be noted that WOA13 with an original 1° resolution is somewhat inaccurate in comparison and verification with other climatologies because it does not include grid data near the coast (Fig. 4(e)).

    The root mean square error (RMSE) of the new OI was calculated to be 0.67°C. The RMSE of WAGHC, EARSC, WOA13q, and WOA13 was 0.59, 0.82, 1.02, and 0.98°C, respectively. Thus, it was shown that the new OI was relatively similar to the observed data than the EASRC, WOA13q, and WOA13 produced based on the WOD data of a similar averaged period.

    Figure 5 shows the spatial distribution of temperature bias at 100 m depth in February. As with the result of at 10 m depth, cold bias was shown in the other climatologies except for WAGHC. The averaged bias of the new OI was −0.12°C, which was lower than that of other climatologies (EASRC: −0.47°C, WAGHC: 0.14°C, WOA13q: −0.86°C, WOA13: −0.60°C). Also, the RMSE of the new OI was 0.84°C, which was similar to the observation data compared to other climatologies. Compared to the result for 10 m depth, there showed a warm bias on the coastal area in other climatologies except for WOA13. WOA13 provided no data on the coastal area of Korea because of the resolution. This warm bias in the most high resolution climatologies seems to be related to a limitation of horizontal interpolation algorithm without vertical gradient correction around steep topography regions as suggested by Chang and Shin (2014).

    The temperature bias of 10 m depth in August was warmer than the observation data in all climatologies (Fig. 6). Such a warm bias was large in the coastal area of Korea and the PM and G lines provided by JMA. The warm bias of the coastal area of Korea is generally related to a horizontal interpolation problem as shown in Fig. 5. Another reason is the inconsistency of observational time. The PM and G lines are mainly observed in summer between late July and early August. However, in 2003 and 2008, vessel observations are carried out in mid-June, and the temperature was 3~4°C lower than in other years at all observation points. Therefore, the average observed temperature was lower than that of other climatologies, and a strong warm bias appeared. The bias and RMSE of the new OI were calculated to be 0.40 and 1.10°C, which were lower than the EASRC (bias: 0.45°C, RMSE: 1.13°C), respectively. The bias of WOA13q was 0.09°C, which was calculated as the lowest value among all climatologies, but the RMSE of WOA13q (1.21°C) was calculated to be higher than that of the new OI (1.10°C) due to the large bias. The bias and RMSE of WOA13 were 0.23 and 0.90°C, which were calculated to be lower than relatively high resolution climatologies, but it is because that WOA13 does not provide grid data near the coastal area due to its relatively lower resolution as other climatologies.

    At 100 m depth in August, the bias of the new OI was calculated to be 0.52°C and the RMSE was 1.19°C, which was higher than that of other climatologies (Fig. 7). In addition, the climatologies except for WOA13 that has no data in the coastal area due to low resolution showed a large difference from the observed data in the coastal area. As previously mentioned, this significant difference could be related to a limitation of horizontal interpolation algorithm around coastal area. Therefore, other algorithm such as the isopycnal interpolation method preventing artificial water mass suggested by Gouretski (2018, 2019) can be consider to improve high-resolution climatology in the following further studies.

    Summary and Discussion

    This study verified the high-resolution regional climatology (new OI) in the East/Japan Sea developed by Lee and Chang (2021). This new OI is a highresolution objective analysis field covering all area of East/Japan Sea and solves the horizontal gradient problem shown in EASRC provided by NCEI. The influence radius of a circle of 211 km was applied, and at least 100 profiles were used at all grid points except the coastal area. By comparing the existing climatologies (EASRC, WAGHC, WOA13q, and WOA13), this study shows that this high-resolution climatology reproduces the various meso-scale features in the East/Japan Sea. In particular, the closed curve in a relatively low-temperature region near the eastern coastal area of North Korea is not well simulated in the existing climatologies, while the new OI successfully simulates these features. These results also could be quantified through wavelet analysis. Most climatologies except for WOAs (WOA13q, WOA13) showed significant variability at 160 km. WOA13q and WOA13 showed no spatial variability because the horizontal smoothing effect was large. However, WAGHC showed a relatively complex spatial distribution despite the same resolution as WOA13q. Even the EASRC has a higher spatial resolution than WAGHC, the spatial variability of EASRC was relatively weaker than that of WAGHC in relation to the horizontal gradient problem. The new OI showed the strongest spatial variability compared to the existing climatologies by solving the horizontal gradient problem. Thus, the new OI showed higher performance than other climatologies with high resolution for spatial distribution and variability.

    We confirmed the accuracy of the data for the climatologies including the new OI using serial observation lines that have been observed for a long time in the East/Japan Sea. Except for 100 m depth in August, WAGHC showed a positive bias, unlike other climatologies. This is because the warming signal was reflected using relatively recent observation data than other climatologies. WOA13 has a lower RMSE than WOA13q excluding 100 m depth in August, but this is because that WOA13 does not include results for the coastal area due to its low resolution.

    WOA13q showed the largest difference from observation data among all climatologies, and new OI appeared the smallest except for 100 m depth in August. It was confirmed that the new OI developed by Lee and Chang (2021) has a spatial distribution similar to that of the observation data as a whole compared with other climatologies including EASRC with the same resolution. Therefore, it was confirmed that the new OI is an improved objective analysis field compared to other climatologies including the EASRC.

    However, all climatologies including new OI had a large warm bias in the coastal area. Since they were produced by the horizontal interpolation method, there are limitations in areas with strong vertical gradients such as coastal areas. Therefore, in relation to this problem, it is possible to produce improved climatology by applying the potential vorticity term, vertical gradient correction, and the isopycnal interpolation methods to reflect topographic effect especially for the coastal areas. In addition, from this study, the analysis of vertical section was excluded. It is necessary to produce OI fields at more vertical levels in the further studies, which makes it possible to compare the distribution of major currents and water masses such as Tsushima Warm Current and East Sea Intermediate Water with serial observation data.


    We appreciate anonymous reviewers for helpful comments of this paper. This research was supported by the National Research Foundation of Korea (2019R1A2C1008490) and Chungcheong Sea Grant Program founded by Korean Ministry of Oceans and Fisheries.



    Number of data within the influence radius (211 km by 211 km) on every grid point being calculated to the objective estimate value.


    Temperature distribution of the East/Japan Sea (EJS) at 100 m depth in February from the (a) new OI, (b) East Asian Seas Regional Climatology (EASRC), (c) World Ocean Experiment Argo Global Hydrographic Climatology (WAGHC), (d) World Ocean Atlas 2013 with 0.25 o resolution (WOA13q), and (e) World Ocean Atlas 2013 with 1° resolution (WOA13). The bottom panels are an enlarged in the southwestern part of the East/Japan Sea (dashed box, SWES) from the top panels, respectively.


    Spatial-averaged wavelet power spectrum for temperature anomaly at 10 m depth in February using Morlet wavelet with 95% significant level (dotted line). The same information with except for (b) EASRC, (c) WAGHC, (d) WOA13q and (e) WOA13.


    Spatial distribution of temperature bias at 10 m depth in February for the cases of (a) new OI-OBS, (b) EASRC-OBS, (c) WAGHC-OBS, (d) WOA13q-OBS, and (e) WOA13-OBS.


    The same as Fig. 4 except at 100 m depth for February.


    The same as Fig. 4 except at 10 m depth for August.


    The same as Fig. 4 except at 100 m depth for August.



    1. Boyer, T. P. , Garcia, H. E. , Locarnini, R. A. , Zweng, M. M. , Mishonov, A. V. , Reagan, J. R. , Weathers, K. A. , Baranova, O. K. , Seidov, D. , Smolyar, I. V. ,2018, World Ocean Atlas 2018. [indicate subset used]. NOAA National Centers for Environmental Information.
    2. Boyer, T. P. , Levitus, S. ,1994, Quality Control and Processing of Historical Temperature, Salinity and Oxygen Data. NOAA Technical Report No. 81, National Oceanographic Data Center, 65.
    3. Carnes, M. R. ,2009, Description and evaluation of GDEM-V 3.0. Naval Research Lab, memorandum 7330-09-9165, 21 p.
    4. Carnes, M. R. , Helber, W. R. , Barron, C. N. , and Dastugue, J. M. ,2010, Validation test report for GDEM4. Naval Research Lab, A343035, 66 p.
    5. Chang, Y.-S. , Rosati, A. J. , Zhang, S. , and Harrison, M. J. ,2009, Objective analysis of monthly temperature and salinity for the world ocean in the 21st century: Comparison with World Ocean Atlas and application to assimilation validation, J. Geophys. RES., 114, C02014,
    6. Chang, Y.-S. , and Shin, H.-R. ,2012, Objective analysis of monthly temperature and salinity around the southwestern East Sea (Japan Sea) on a 0.1o grid, Cont. Shelf Res., 45, 54-64,
    7. Chang, Y.-S. , and Shin, H.-R. ,2014, Vertical Gradient Correction for the Oceanographic Atlas of the East Asian Seas. J. Geophys. Res. Oceans 119, 5546-5554,
    8. Fukumori, I. , Wunsch, C. ,1991, Efficient representation of the North Atlantic hydrographic and chemical distributions, Progress in Oceanography, 27, 111-195,
    9. Gouretski, V. ,2018, World ocean circulation experiment- Argo global hydrographic climatology, Ocean Sci. 14, 1127-1146,
    10. Gouretski, V. ,2019, A new global hydrographic climatology, Atmospheric and Oceanic Science Letters,
    11. Hosoda, S. , Ohira, T. , Nakamura, T. ,2008. A monthly mean dataset of global oceanic temperature and salinity derived from Argo float observations. JAMSTEC Rep. Res. Dev. vol. 8, 47-59.
    12. Lee, J.-H. , and Chang, Y.-S. ,2021, Development of highresolution regional climatology in the East/Japan Sea with a primary focus on meridional temperature gradient correction, Front. Earth Sci. 9:680881,
    13. Levitus, S. ,1982, Climatological Atlas of the World Ocean, US. Government Printing Office, NOAA professional paper 13, 190 p.
    14. Levitus, S. (Ed),2002, World Ocean Atlas 2001, vol. 1-6, NOAA Atlas NESDIS, vol. 49-54, NOAA, Silver Spring, MD.
    15. Levitus, S. (Ed),2006, World Ocean Atlas 2005, vol. 1-4, NOAA Atlas NESDIS, vol. 61-64, NOAA, Silver Spring, MD.
    16. Levitus, S. (Ed),2009, World Ocean Atlas 2009, vol. 1-4, NOAA Atlas NESDIS, vol. 68-71, NOAA, Silver Spring, MD.
    17. Levitus, S. (Ed),2013, World Ocean Atlas 2013, vol. 1-4, NOAA Atlas NESDIS, vol. 73-76, NOAA, Silver Spring, MD.
    18. McIntosh, P. ,1990, Oceanographic data interpolation: Objective analysis and splines. Journal of Geophysical Research 95, 13529-13541,
    19. NOAA,1994, World Ocean Atlas 1994, vol. 1-5, NOAA Atlas NESDIS, vol. 1-5, NOAA, Silver Spring, MD.
    20. NOAA,1998, World Ocean Atlas 1998, vol. 1-12, NOAA Atlas NESDIS, vol. 27-38, NOAA, Silver Spring, MD.
    21. Roemmich, D. , Gilson, J. ,2009. The 2004-2008 mean and annual cycle of temperature, salinity and steric height in the global ocean from Argo Program. Prog. Oceanogr. 82, 81-100,
    22. Smith, D. M. , S. Cusack, A. W. Colman, C. K. Folland, G. R. Harris, J. M. Murphy,2007, Improved Surface Temperature Prediction for the Coming Decade from a Global Climate Model, Science, 317 (5839): 796-799.
    23. Teague, W. J. , Carron, M. J. , and Hogan, P. J. ,1990, A comparison between the generalized digital environmental model and levitus climatologies. Journal of Geophysical Research, 95, 7167-7183.