Observation of the ocean is crucial to the studies of ocean dynamics, climate change, and biogeochemical cycle. However, current oceanic observations are patently insufficient, because the in situ observations are of difficulty and high cost while the satellite remote-sensed measurements are mainly the sea surface data. To make up for the shortage of ocean interior data and make full use of the abundant satellite data, here we develop a data-driven deep learning model to estimate ocean subsurface and interior variables from satellite-observed sea surface data. Exclusively and simply using satellite data, three-dimensional ocean temperature and salinity fields are successfully reconstructed, which are at 26 level depths from 0 to 2,000 m. We further design a scheme to increase the horizontal resolution from 1° to 1/4°, which is higher than the Argo gridded data. Estimations from our model are accurate, reliable, and stable for a wide range of research areas and periods. Dynamic height fields that are derived from the estimated temperature and salinity, as well as the associated ocean geostrophic flows, are also calculated and analyzed, which indicates the potentials of our model for reconstructing the ocean circulation fields as well. This study enriches oceanic observations with respect to vertical dimension and horizontal resolution, which can largely make up for the paucity of the subsurface and deep ocean observation, both before and during Argo era. This work also provides some new foundations for and insights into geoscience and climate change fields.
Three-dimensional temperature and salinity anomaly fields are estimated exclusively from satellite data through a deep learning approach
Estimated data are at 26 depths (0–2,000 m) and horizontal resolution has been improved to 1/4°, higher than Argo gridded data
Our model is accurate and reliable for a wide range of research area and period; dynamic fields from the estimated data are also analyzed
Plain Language Summary
Ocean data are important for ocean science and climate change studies; however, oceanic observations are difficult and costly and thus are still very sparse in space and time. Satellites have been providing plentiful oceanic observations, but these data are only at the sea surface. To fully utilize the copious satellite data and to make up for the shortage of ocean interior data. Here, we establish a deep learning model to connect the surface ocean with the subsurface and deep oceans, through which, subsurface and deep ocean temperature and salinity data are estimated from the surface data observed by satellites. We also design a new scheme to improve the horizontal resolution of the obtained data. The results show that our model successfully reconstructed the three-dimensional field data of temperature and salinity. Our model could facilitate ocean science studies by largely enriching the ocean data availability.
Data Availability Statement
The authors thank OISST data from the NOAA for sea surface temperature data (https://www.ncei.noaa.gov/data/sea-surface-temperature-optimum-interpolation/v2.1/access/avhrr/), CCMP (version 2) for SSW vector analysis product (https://data.remss.com/ccmp/v02.0/), SMOS (data set: CEC-Locean L3 Debiased v2) for SSS data (ftp://ext-catds-cecos-locean:[email protected]/), Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO) for sea level data, available at Copernicus Europe's eyes on Earth (https://resources.marine.copernicus.eu/?option=com_csw&view=details&product_id=SEALEVEL_GLO_PHY_L4_REP_OBSERVATIONS_008_047), and EN4 (data set: version EN4.2.1, .g10.) for Argo profiles data (https://www.metoffice.gov.uk/hadobs/en4/download-en4-2-1.html). The authors also thank teh IPRC (http://apdrc.soest.hawaii.edu/projects/Argo/data/gridded/On_standard_levels/index-1.html) and the ISAS15 (https://www.seanoe.org/data/00412/52367/) for Argo data of monthly temperature and salinity gridded fields. Argo data were collected and made freely available by the International Argo Program and the national programs that contribute to it, the Argo Program is part of the Global Ocean Observing System.
|2021JC017605-sup-0001-Supporting Information SI-S01.docx2.4 MB||Supporting Information S1|
Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
- 2017). A review of ocean/sea subsurface water temperature studies from remote. sensing and non-remote sensing methods, 9(12), 936. https://doi.org/10.3390/w9120936
- 2004). Estimation of ocean subsurface thermal structure from surface parameters: A neural network approach. Geophysical Research Letters, 31(20). https://doi.org/10.1029/2004gl021192
- Argo. (2020). Argo float data and metadata from Global Data Assembly Centre (Argo GDAC). https://doi.org/10.17882/42182
- 2011). A cross-calibrated, multiplatform ocean surface wind velocity product for meteorological and oceanographic applications. Bulletin of the American Meteorological Society, 92(2), 157–174. https://doi.org/10.1175/2010bams2946.1
- 1996). A multiyear global surface wind velocity dataset using SSM/I wind observations. Bulletin of the American Meteorological Society, 77(5), 869–882. https://doi.org/10.1175/1520-0477(1996)077<0869:AMGSWV>2.0.CO;2
- 2020). Improved estimation of proxy sea surface temperature in the arctic. Journal of Atmospheric and Oceanic Technology, 37(2), 341–349. https://doi.org/10.1175/jtech-d-19-0177.1
- 2020). DINCAE 1.0: A convolutional neural network with error estimates to reconstruct sea surface temperature satellite observations. Geoscientific Model Development, 13(3), 1609–1622. https://doi.org/10.5194/gmd-13-1609-2020
- 2019). Applications of deep learning to ocean data inference and subgrid parameterization. Journal of Advances in Modeling Earth Systems, 11(1), 376–399. https://doi.org/10.1029/2018ms001472
- 2020). SMOS SSS L3 maps generated by CATDS CEC LOCEAN. debias V5.0. https://doi.org/10.17882/52804
- 2018). New SMOS sea surface salinity with reduced systematic errors and improved variability. Remote Sensing of Environment, 214, 115–134. https://doi.org/10.1016/j.rse.2018.05.022
- 2005). Slowing of the Atlantic meridional overturning circulation at 25 degrees N. Nature, 438(7068), 655–657. https://doi.org/10.1038/nature04385
- 2014). Varying planetary heat sink led to global-warming slowdown and acceleration. Science, 345(6199), 897–903. https://doi.org/10.1126/science.1254937
- 2000). Determination of vertical thermal structure from sea surface temperature. Journal of Atmospheric and Oceanic Technology, 17(7), 971–979. https://doi.org/10.1175/1520-0426(2000)017<0971:dovtsf>2.0.co;2
- 2007). Temporal variability of the Atlantic meridional overturning circulation at 26.5 degrees N. Science, 317(5840), 935–938. https://doi.org/10.1126/science.1141304
- 2003). Emerging ocean observations for interdisciplinary data assimilation systems. Journal of Marine Systems, 40–41, 5–48. https://doi.org/10.1016/s0924-7963(03)00011-3
- 2014). Learning a deep convolutional network for image super-resolution. Paper presented at the Computer Vision – ECCV 2014, Cham.
- 2016). Image super-resolution using deep convolutional networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(2), 295–307. https://doi.org/10.1109/tpami.2015.2439281
- 2000). Multivariate projection of ocean surface data onto subsurface sections. Geophysical Research Letters, 27(6), 755–757. https://doi.org/10.1029/1999gl010451
- 1983). Algorithms for the computation of fundamental properties of seawater. UNESCO Technical Papers in Marine Sciences (Vol. 44, pp. 1–54). UNESCO.
- 1995). A note on the barotropic response of sea level to time-dependent wind forcing. Journal of Geophysical Research, 100(C12). https://doi.org/10.1029/95jc02259
- 2012). ISAS-tool version 6: Method and configuration. Retrieved from https://archimer.ifremer.fr/doc/00115/22583/
- 2013). EN4: Quality controlled ocean temperature and salinity profiles and monthly objective analyses with uncertainty estimates. Journal of Geophysical Research: Oceans, 118(12), 6704–6716. https://doi.org/10.1002/2013JC009067
- 2016). Deep learning. MIT Press.
- 2012). High resolution 3-D temperature and salinity fields derived from in situ and satellite observations. Ocean Science, 8(5), 845–857. https://doi.org/10.5194/os-8-845-2012
- 2019). Deep learning based single image super-resolution: A survey. International Journal of Automation and Computing, 16(4), 413–426. https://doi.org/10.1007/s11633-019-1183-x
- 2019). Deep learning for multi-year ENSO forecasts. Nature, 573(7775), 568–572. https://doi.org/10.1038/s41586-019-1559-7
- 2019). A convolutional neural network using surface data to predict subsurface temperatures in the Pacific Ocean. IEEE Access, 7, 172816–172829. https://doi.org/10.1109/access.2019.2955957
- 2021). New observations from the SWIM radar on-board CFOSAT: Instrument validation and ocean wave measurement assessment. IEEE Transactions on Geoscience and Remote Sensing, 59(1), 5–26. https://doi.org/10.1109/tgrs.2020.2994372
- 2018). Statistical machine learning methods and remote sensing for sustainable development goals: A review. Remote Sensing, 10(9). https://doi.org/10.3390/rs10091365
- 2019). Indian Ocean warming can strengthen the Atlantic meridional overturning circulation. Nature Climate Change, 9(10), 747–751. https://doi.org/10.1038/s41558-019-0566-x
- 2021). Improvements of the daily optimum interpolation sea surface temperature (DOISST) version 2.1. Journal of Climate, 34(8), 2923–2939. https://doi.org/10.1175/jcli-d-20-0166.1
- 2009). Ocean circulation: Wind-driven and thermohaline processes. Cambridge University Press.
- 2001). Wind stress over the ocean. Cambridge University Press.
- 2014). Subsurface and deeper ocean remote sensing from satellites: An overview and new results. Progress in Oceanography, 122, 1–9. https://doi.org/10.1016/j.pocean.2013.11.010
- 2017). ISAS-15 temperature and salinity gridded fields. https://doi.org/10.17882/52367
- 2013). Recent global-warming hiatus tied to equatorial Pacific surface cooling. Nature, 501(7467), 403–407. https://doi.org/10.1038/nature12534
- 2012). ImageNet classification with deep convolutional neural networks. Neural Information Processing Systems, 25.
- 2016). Machine learning in geosciences and remote sensing. Geoscience Frontiers, 7(1), 3–10. https://doi.org/10.1016/j.gsf.2015.07.003
- 2014). Retrieving density and velocity fields of the ocean's interior from surface data. Journal of Geophysical Research: Oceans, 119(12), 8512–8529. https://doi.org/10.1002/2014jc010221
- 2019). Reconstructing the ocean interior from high-resolution sea surface information. Journal of Physical Oceanography, 49(12), 3245–3262. https://doi.org/10.1175/jpo-d-19-0118.1
- 2019). Filling the gaps of missing data in the merged VIIRS SNPP/NOAA-20 ocean color product using the DINEOF method. Remote Sensing, 11(2). https://doi.org/10.3390/rs11020178
- 2019). A sea change in our view of overturning in the subpolar North Atlantic. Science, 363(6426), 516. https://doi.org/10.1126/science.aau6592
- 2019). Subsurface temperature estimation from remote sensing data using a clustering-neural network method. Remote Sensing of Environment, 229, 213–222. https://doi.org/10.1016/j.rse.2019.04.009
- 2016). Submesoscale currents in the ocean. Proceedings of the Royal Society A: Mathematical, Physical & Engineering Sciences, 472(2189), 20160117. https://doi.org/10.1098/rspa.2016.0117
- 2020). Variability of the shallow overturning circulation in the Indian Ocean. Journal of Geophysical Research: Oceans, 125(2). https://doi.org/10.1029/2019jc015651
- 1980). A new high pressure equation of state for seawater. Deep-Sea Research, Part A: Oceanographic Research Papers, 27(3–4), 255–264. https://doi.org/10.1016/0198-0149(80)90016-3
- 1950). On the wind-driven ocean circulation. Journal of Meteorology, 7(2), 80–93. https://doi.org/10.1175/1520-0469(1950)007<0080:OTWDOC>2.0.CO;2
- 2019). On the future of Argo: A global, full-depth, multi-disciplinary array. Frontiers in Marine Science, 6(439). https://doi.org/10.3389/fmars.2019.00439
- 2009). The Argo program: Observing the global oceans with profiling floats. Oceanography, 22(2), 34–43. https://doi.org/10.5670/oceanog.2009.36
- 2014). Very deep convolutional networks for large-scale image recognition. arXiv 1409.1556.
- 1988). Coefficients for sea surface wind stress, heat flux, and wind profiles as a function of wind speed and temperature. Journal of Geophysical Research: Oceans, 93(C12), 15467–15472. https://doi.org/10.1029/JC093iC12p15467
- 2008). Introduction to physical oceanography (Vol. 65).
- 2018). Retrieving temperature anomaly in the global subsurface and deeper ocean from satellite observations. Journal of Geophysical Research: Oceans, 123(1), 399–410. https://doi.org/10.1002/2017jc013631
- 2015). Estimation of subsurface temperature anomaly in the Indian Ocean during recent global surface warming hiatus from satellite measurements: A support vector machine approach. Remote Sensing of Environment, 160, 63–71. https://doi.org/10.1016/j.rse.2015.01.001
- 2019). Estimating subsurface thermohaline structure of the global ocean using surface remote sensing observations. Remote Sensing, 11, 1598. https://doi.org/10.3390/rs11131598
- 2011a). Physical properties of seawater. Chap. 3 in Descriptive physical oceanography ( 6th ed., pp. 29–65). Academic Press. https://doi.org/10.1016/b978-0-7506-4552-2.10003-4
- 2011b). Typical distributions of water characteristics. Chap. 4 in Descriptive physical oceanography ( 6th ed., pp. 67–110). Academic Press. https://doi.org/10.1016/b978-0-7506-4552-2.10004-6
- 2011c). Mass, salt, and heat budgets and wind forcing. Chap. 5 in Descriptive physical oceanography ( 6th ed., pp. 111–145). Academic Press. https://doi.org/10.1016/b978-0-7506-4552-2.10005-8
- 2011d). Data analysis concepts and observational methods. Chap. 6 in Descriptive physical oceanography ( 6th ed., pp. 147–186). Academic Press. https://doi.org/10.1016/b978-0-7506-4552-2.10006-x
- 2011e). Dynamical processes for descriptive ocean circulation. Chap. 7 in Descriptive physical oceanography ( 6th ed., pp. 187–221). Academic Press. https://doi.org/10.1016/b978-0-7506-4552-2.10007-1
- 2015). Has there been a hiatus? Science, 349(6249), 691. https://doi.org/10.1126/science.aac9225
- 2010). Tracking Earth's energy. Science, 328(5976), 316–317. https://doi.org/10.1126/science.1187272
- 2013). An apparent hiatus in global warming? Earth's Future, 1(1), 19–32. https://doi.org/10.1002/2013ef000165
- 1990). The mean annual cycle in global ocean wind stress. Journal of Physical Oceanography, 20(11), 1742–1760. https://doi.org/10.1175/1520-0485(1990)020<1742:tmacig>2.0.co;2
- 2013). Reconstructing the ocean's interior from surface data. Journal of Physical Oceanography, 43(8), 1611–1626. https://doi.org/10.1175/jpo-d-12-0204.1
- 2012). Estimation of subsurface temperature anomaly in the North Atlantic using a self-organizing map neural network. Journal of Atmospheric and Oceanic Technology, 29(11), 1675–1688. https://doi.org/10.1175/jtech-d-12-00013.1
- 2017). What caused the global surface warming hiatus of 1998–2013? Current Climate Change Reports, 3(2), 128–140. https://doi.org/10.1007/s40641-017-0063-0
- 2016). The global warming hiatus: Slowdown or redistribution? Earth's Future, 4(11), 472–482. https://doi.org/10.1002/2016ef000417
- 2006). A new study of the Mediterranean outflow, air–sea interactions, and Meddies using multisensor data. Journal of Physical Oceanography, 36(4), 691–710. https://doi.org/10.1175/JPO2873.1
- 1992). Three-dimensional analytical model for the mixed layer depth. Journal of Geophysical Research, 97(C12). https://doi.org/10.1029/92jc01833
- 1991). An analytical model for remote sensing determination of the mixed layer depth. Deep-Sea Research Part A. Oceanographic Research Papers, 38(3), 267–287. https://doi.org/10.1016/0198-0149(91)90068-q
- 2019). Deep learning for single image super-resolution: A brief review. IEEE Transactions on Multimedia, 21(12), 3106–3121. https://doi.org/10.1109/tmm.2019.2919431