Up to Specific issues in real coupled models
Dear Sophie, We are using the last version of EC-Earth (3.2.0) coupling IFS to NEMO via OASIS3-MCT. The coupling for Kinetic energy, heat and mass from IFS to NEMO is done using a post-processing global conservation method implemented using OASIS. We have seen that this method is very expensive computationally, increasing the computational time of IFS up to 30%. We understand all these communications are needed to do the field coupling conservative. However, we wonder if the coupling efficiency can be improved: - First of all, Aurore Voldoire told us that they are using a couple frequency of 3 hours for CNRM-CM6 while EC-Earth 3.2.0 is using a coupling frequency of 45 min (every time step). Though we are discussing the coupling frequency with the technical working group of EC-Earth, we would like to know why three hours is normally used. Are you aware of any study related with the coupling frequency? - Secondly, we have seen that the three fields (kinetic energy, heat and mass) are coupled independently (using different OASIS put and get for each field), is it possible to do the coupling for the three fields at the same time? this would reduce the MPI communications. Thank you for your help, it will be very useful for us and the technical working group of EC-Earth. Best regards, Mario
Hi Mario, Yes, this is a problem. I think the real solution here is to allow the computation of the CONSERV to be done at a lower frequency than the coupling exchanges themselves. E.g. if you exchange a coupling field every 3 hours, there would be an internal mechanism that would accumulate the values of the coupling field before and after the interpolation for all exchanges during one day; the global communication to sum the field over the whole geographical domain, the computation of the difference of the sums before and after the interpolation would be done only once per day and the difference would then be distributed over all the coupling exchanges during the next day. Do you think this would make sense? If so, please, let me know how urgent it is for you to have this option implemented. Concerning more precisely your two questions : - Within SPECS (HR), we looked at the difference between coupling every 24h and every 3h, and this makes a big difference. I will ask to the oasis user list if some studies have been done on the impact of the coupling frequency. We will post the answers on the forum. - Yes, this functionality is supported, see "Support to couple multiple fields via a single communication." in the User Guide https://oasis.cerfacs.fr/wp-content/uploads/sites/114/2021/02/GLOBC-TR-oasis3mct_UserGuide3.0_052015.pdf, section B1.1.1. This is supported through colon delimited field lists in the namcouple, for example : ATMTAUX:ATMTAUY:ATMHFLUX TAUX:TAUY:HEATFLUX 1 3600 3 rstrt.nc EXPORTED in a single namcouple entry. Best regards, Sophie
Hi again, Below are some papers on the the impact of the coupling frequency: - Rimac, A., J.-S. von Storch, C. Eden, and H. Haak (2013), The influence of high-resolution wind stress field on the power input to near-inertial motions in the ocean, Geophys. Res. Lett., 40, doi:10.1002/grl.50929. - Bernie D. J., E. Guilyardi, G. Madec, J. M. Slingo, S. W. Woolnough and J. Cole (2008). Impact of resolving the diurnal cycle in an ocean-atmosphere GCM. Part 2: A diurnally coupled CGCM. Clim. Dyn., 31, 909-925 - Masson, S., P. Terray, G. Madec, J.-J. Luo, T. Yamagata and K. Takahashi, 2012 : Impact of intra-daily SST variability on ENSO characteristics in a coupled model. Climate Dynamics, 39, 3-4, 681-707, doi:10.1007/s00382-011-1247-2 - Terray P., K. Kamala, S. Masson, G. Madec, A. K. Sahai, Jing-Jia Luo, T. Yamagata, 2012: The role of the intra-daily SST variability in the Indian monsoon variability and monsoon-ENSO–IOD relationships in a global coupled model, Climate Dynamics, Volume 39, Issue 3-4 , pp 729-754 , doi:10.1007/s00382-011-1240-9
I finally got round to trying to collate at the things people have sent me about studying the effects of reducing coupling frequency in coupled models. There are no substantial reports and most of the info is in the form of incidental comments in reports largely dedicated to other matters or validation notes which are generally just lots and lots of plots of various model fields and summary tables of RMS errors without detailed text analysis or descriptions, so I've tried to interpret some of these things without having much detailed background knowledge of the tests that were run or what I'm looking at. Dan Copsey, in our GC2 model, (UM atmosphere NEMO ocean + CICE seaice) studied the effect of reducing the coupling exchange timestep for all fields from 3 hours to 1 hour. This model used OASIS3-MCT, NEMO vn3.5 and CICE vn5.0. The resulting summary of these tests indicates : 1) A reduction of wind stress in the Southern Ocean 2) Increased wind stress in the Equatorial Pacific region (regarded as an improvement to the model). 3) Colder 1.5 metre Arctic temperatures in the Atlantic sector. 4) Colder Northern Hemisphere SSTs and warmer Southern Ocean (degrading the model slightly). 5) Increased salinity in the Arctic and reduced salinity in the North Atlantic (degrading the model slightly). Earlier tests also using GC2 by Jose Rodriguez/Tim Johns also studied the effects of changing from 3 hours to 1 hour : 1) Reducing coupling frequency obviously improves the representation of the diurnal cycle of SST. 2) There are extensive validation comparisons between control and experiment runs. To my (non-expert) eye these generally appear to indicate that where an improvement in model results v observations occurs, it is most noticeable in ocean model fields, but improvement is by no means across the board. Ocean temperatures appear improved in most cases, though any salinity improvement is less obvious. 3) There was also a slight increase in the strength of Antarctic meridional overturning circulation. The file, https://oasis.cerfacs.fr/wp-content/uploads/sites/114/2021/08/amoc_1h_vs_3h_coupling.pdf, which was sent to me illustrates this, however Jose cautions that "these results are based on a 10-year simulation, which is probably not long-enough. So the result has to be taken with a grain of salt." A study of a high resolution coupled model (N512 - global ORCA 1/12th degree) mentions in passing that it found this set-up found a minor change in SST biases but also mentions the enhanced meridional overturning circulation compared with lower resolution models which it attributes to the high resolution of the ocean AND the increased coupling frequency. This report also suggests that the combination of higher resolutions and increased coupling frequency has a positive effect in regions of "strong air-sea interaction" where large improvements are seen in stratocumulus regions. Various groups plan to undertake studies relating to coupling frequency in the near future (e.g. in N96-eORCA1 coupled model to see if they can improve "water mass formation" in the North Atlantic). The wave model people plan to run tests later this year by reducing wave model coupling to the atmosphere and ocean to sub-hourly frequencies. Richard
