EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Could an ocean climate atlas generated by a model compete with an observational one?

Georgy I. Shapiro, Jose M. Gonzalez-Ondina, Xavier Francis, Hyee S. Lim, and Ali Almehrezi
Georgy I. Shapiro et al.
  • University of Plymouth, School of Bilogical and Marine Sciences, Plymouth Ocean Forecasting Centre, Plymouth, Devon, United Kingdom of Great Britain and Northern Ireland (

Modern numerical ocean models have matured over the last decades and are able to provide accurate fore- and hind-cast of the ocean state. The most accurate data could be obtained from the reanalysis where the model run in a hindcast mode with assimilation of available observational data. An obvious benefit of model simulation is that it provides the spatial density and temporal resolution which cannot be achieved by in-situ observations or satellite derived measurements. It is not unusual that even a relatively small area of the ocean model can have in access of 100,000 nodes in the horizontal, each containing vertical profiles of temperature, salinity, velocity and other ocean parameters with a temporal resolution theoretically as high as a few minutes. Remotely sensed (satellite) observations of sea surface temperature can compete with the models in terms of spatial resolution, however they only produce data at the sea surface not the vertical profiles. On the other hand, in-situ observations have a benefit of being much more precise than model simulations. For instance a widely used CTD profiler SBE 911plus has accuracy of about 0.001 °C, which is not achievable by models.

In the creation of a climatic atlas the higher accuracy of individual profiles provided by in-situ measurements may become less beneficial. Assuming the normal distribution of data at each location, the standard error of the mean (SEM) is calculated as SE=S/SQRT(N), where S is the standard deviation of individual data points around the mean, and N is the number of data points. The climatic data are obtained by averaging a large number of individual data points, and here the benefit of having more data points may become a greater advantage than the accuracy of a single observation.  

In this study we have created an ocean climate atlas for the northern part of the Indian Ocean including the Red Sea and the Arabian Gulf using model generated data. The data were taken from Copernicus Marine Environment Monitoring Service (CMEMS) reanalysis product GLOBAL_REANALYSIS_PHY_001_030 with 1/12° horizontal resolution and 50 vertical levels for the period 1998 to 2017. The model component is the NEMO platform driven at the surface by ECMWF ERA-Interim reanalysis. The model assimilates along track altimeter data, satellite Sea Surface Temperature, as well as in-situ temperature and salinity vertical profiles where available. The monthly data from CMEMS were then averaged over 20 years to produce an atlas at the surface, 10, 20, 30, 75, 100, 125, 150, 200, 250, 300, 400, and 500 m depths.  The standard error of the mean has been calculated for each point and each depth level on the native grid (1/12 degree).

The atlas based on model simulations was compared with the latest version of the World Ocean Atlas (WOA)  2018 published by the NCEI.  WOA has objectively analysed climatological mean fields on a ¼  degree grid. The differences between the mean values and SEMs from observational and simulated atlases are analysed, and the potential causes of mismatch are discussed.

How to cite: Shapiro, G. I., Gonzalez-Ondina, J. M., Francis, X., Lim, H. S., and Almehrezi, A.: Could an ocean climate atlas generated by a model compete with an observational one?, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-1582,, 2019