D6.4: Report and technical documentation for high-frequency logging in subdomains

nextGEMS Next Generation Earth Modelling Systems Deliverable title Report and technical documentation for high-frequency logging in subdomains Deliverable number D32 (Del. Rel. No. D6.4) Revision 0 Status Final Planned delivery date 31/08/2023 Actual date of issue 30/08/2023 Nature of deliverable Report Lead partner MPI-M Dissemination level Public The research leading to these results has received funding from the European Union’s Horizon 2020 Research and Innovation programmes under grant agreement No 101003470 About this document Based on the requirements obtained from the assessment of observational data from the Tropical Atlantic, high-frequency output of selected variables will be provided for specific analyses to be compared in observations and simulations. Work package in charge: WP6 Lead author: Swantje Bastin Other contributing authors: Johann Jungclaus, Arjun Kumar Internal reviewer(s): 1st reviewer: Markus Jochum 2nd reviewer: Divya Praturi Contacts: [email protected] Visit us on: www.nextgems-h2020.eu Disclaimer: This material reflects only the authors view and the Commission is not responsible for any use that may be made of the information it contains. 2 Contents Contents Contents 1 Executive summary 4 2 Project objectives 4 3 Detailed report on the deliverable 7 3.1 Modelexperiments................................... 7 3.2 FESOM ......................................... 7 3.3 ICON-O......................................... 7 3.4 Common settings and experiment descriptions . . . . . . . . . . . . . . . . . . . 8 3.5 Data selection and publication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 3.5.1 WDCC: entire tropical Atlantic . . . . . . . . . . . . . . . . . . . . . . . . 11 3.5.2 Zenodo: selected time series . . . . . . . . . . . . . . . . . . . . . . . . . 11 4 References (Bibliography) 13 5 Dissemination and uptake 15 5.1 Uptake by the targeted audience . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 5.2 This is how we are going to ensure the uptake of the deliverables by the targeted audience ........................................ 15 6 Deliverable timeliness 15 7 Changes made and/or difficulties encountered, if any 15 8 Sustainability 16 8.1 Links built with other deliverables, WPs, and synergies created with other projects 16 9 Full track of dissemination activities 16 10 Full track of publications and IP 17 3 2 PROJECT OBJECTIVES 1 Executive summary NextGEMS task 6.5 carried out dedicated sensitivity experiments with the two ocean models used in NextGEMS (ICON-O and FESOM). This report describes technical details of the model set-ups and the handling of high-frequency data for analyses of mixed-layer processes in the Tropical Atlantic. We provide high-frequency model output used in Task 6.5 for testing different vertical mixing schemes and associated parameter settings. This 3-hourly model output is made publicly available for a selected subdomain of particular interest: the tropical Atlantic, where many observations are available to validate the models. The entire high-frequency tropical Atlantic output from the upper 200m on a 0.1◦×0.1◦latitude-longitude grid is published at the World Data Center for Climate (WDCC), in a long-term tape archive, where it is available under a CC BY 4.0 license (https://doi.org/10.26050/WDCC/nextGEMS_WP6oc). Because the WDCC tape archive has a large latency and data retrieval is thus slow, we additionally provide a subset of the tropical Atlantic output on Zenodo, where data access is fast but data size is limited. For the latter dataset, time series from the models are provided for selected locations where observational data are available during the modelled time period (January 2014 - December 2015), also under a CC BY 4.0 license (https://doi.org/10.5281/zenodo.8225706). 2 Project objectives The following table lists the overall objectives of nextGEMS. Objectives that the current report contributes to are flagged with yes. Specific objectives of the project Contribution of this deliverable? O1 To develop two SR-ESMs for applications (i) by demonstrating their capacity to more realistically represent the coupled (land-oceanatmosphere) climate system, also through an ability to better leverage observations; yes (ii) by performing the first global multi-decadal (30 y) SR-ESM based climate projections, testing for out-of-sample climate trajectories, i.e., surprises, and thereby giving a new perspective on uncertainty; and no (iii) by expanding their scope to begin more physically coupling ‘Earth-system’ processes, including the carbon cycle and the atmospheric aerosol no O2 4 2 PROJECT OBJECTIVES Specific objectives of the project Contribution of this deliverable? To use SR-ESMs to test emerging and long-standing hypotheses underpinning our understanding of climate change: (i) that convective organization contributes importantly to Earth’s energy budget and the strength of cloud feedbacks; no (ii) that a more explicit representation of cloudaerosol interactions mutes aerosol-radiative forcing; no (iii) that 2 km to 200 km scale atmospheric and oceanic circulations are of leading order importance for air-sea fluxes in the tropics, thereby influencing not only the mean tropical and midlatitude climate, but also its variability, including extremes; yes (iv) that storm-scale variability in weather systems and of the land surface strongly influence extratropical climate and extremes, for instance by conditioning circulation regimes, like blocking; no (v) that capturing landscape variability globally greatly improves the realism with which regional climate can be simulated; and no (vi) that storm-scale variability through its impact on hydrological extremes affects the carbon budget, with associated implications for the global carbon (emissions) stock-take. no O3 To build new, more integrated, communities of ESM users by: (i) exploiting the necessity of developing SR-ESMs around a centralized infrastructure to create development and analysis paradigms that can more directly involve a broader and more distributed scientific community; and yes (ii) by exploiting the affinity between what people experience and what SR-ESMs simulate (i.e., events, in addition to statistics) to more directly involve non-scientific users in model development, thereby fostering Knowledge Coproduction. yes 5 2 PROJECT OBJECTIVES This deliverable directly contributes to the achievement of specific objectives indicated in the description of the Work Package: Objectives of WP6 Relevance in this deliverable? To identify and evaluate key ocean and marine atmosphere features in the SR-ESM simulations of all three development cycles. (O1) yes To validate the marine planetary boundary layer representation of the SR-ESM simulations. (O1) no To provide ocean initial conditions for the coupled SRESM simulations. (O1) yes To provide developments towards an ocean mixed layer scheme adequate for SR-ESMs. (O1) yes To create numerical tools that generate output streams useful to the planetary boundary layer, ocean mixed-layer and fisheries communities. (O1,O3) yes 6 3 DETAILED REPORT ON THE DELIVERABLE 3 Detailed report on the deliverable As part of Task 6.5, we carried out a number of ocean-only sensitivity experiments with FESOM and ICON-O. The runs differ in their vertical mixing scheme and its parameter settings. In most of the experiments, the same mixing scheme and the same settings were used in FESOM and ICON. We then compared the different runs to each other as well as to observations, to find the best settings for the coupled nextGEMS configurations. In the report on Deliverable 6.5 and in the related publication (Bastin et al.,2024), the results of these analyses are described in detail. Here, we provide a technical description of how we saved the relevant data for the analyses and how we made high-frequency output of the model runs publicly available, for a selected subdomain of particular interest, i.e. the tropical Atlantic. 3.1 Model experiments The model versions used in this study are compatible with the ones used in nextGEMS cycle 3. The FESOM model is documented in Danilov et al. (2017); Scholz et al. (2019,2022), and ICONO in Korn et al. (2022). All model runs were done at approximately 10km horizontal resolution in the domain of interest (ICON-O: globally, FESOM: tropical Atlantic, otherwise coarser). This is slightly coarser than the coupled nextGEMS runs, but in order to do so many sensitivity runs and not overstretch our computing resources, 10km was the finest resolution that was feasible. We did test runs with 40km and 5km resolution as well, and the change e.g. in mean tropical Atlantic mixed layer depth between 40 and 10km was much more substantial than between 10 and 5km. This suggested to us that doing the ocean vertical mixing sensitivity study at 10km resolution for the 5km nextGEMS coupled runs was reasonable. 3.2 FESOM FESOM2 is a global unstructured-mesh ocean model developed at the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research (AWI) in Bremerhaven (Danilov et al., 2017). It is formulated on a triangular mesh, utilizes a finite-volume dynamical core and Arbitrary Lagrangian–Eulerian (ALE) vertical coordinates (Scholz et al.,2019). The model’s unstructured nature enables different types of local mesh refinements (e.g. one that follows local sea surface height variability. For this study we use a FESOM2 mesh that has 50 km resolution over most of the globe, except for the equatorial Atlantic, where it is set to 13 km resolution.FESOM uses az∗vertical coordinate, where the total change in SSH is distributed equally over all layers, except the layer that touches the bottom. 3.3 ICON-O ICON-O (Korn et al.,2022) is the ocean component of the Icosahedral Nonhydrostatic Weather and Climate Model (ICON) in its Earth System Model configuration. It is developed at the Max Planck Institute for Meteorology (MPI-M) in Hamburg. The ocean component of ICON-ESM, ICON-O, solves the hydrostatic (the “Nonhydrostatic” in the name only refers to the atmospheric component) Boussinesq equations with a free surface. These equations are solved on a triangular horizontal grid, which is generated by dividing the spherical domain into an icosahedron 7 3.4 Common settings and experiment descriptions3 DETAILED REPORT ON THE DELIVERABLE and subsequent division of the 12 icosahedron parts into triangles. In this study, an ICON grid with globally approximately uniform horizontal resolution of about 10km is used. While the ICON and FESOM horizontal grids are different, we argue that the two models are comparable, as we only compare results from our region of interest, i.e. the tropical Atlantic. In the vertical, a z∗ coordinate is used where model levels follow the free surface. For details on the numerics or other specifics about ICON-O, see Korn et al. (2022). 3.4 Common settings and experiment descriptions To meet the specific demands for Task 6.5, we decided on settings that would be shared by both models. Although we tried to homogenise the model settings between FESOM and ICONO that are directly connected to the representation of vertical mixing, we left the rest of the model settings as they are commonly used in the FESOM and ICON-O communities at our institutes, i.e. partly different between the two models. This is intended and part of the reason why we do this study, to see how much of the variation between the different vertical mixing settings is model specific and how much happens similarly in both models. We implemented a common vertical axis with 128 vertical levels for both models, with thicknesses ranging from 2 m near the surface to about 200 m near the seafloor. We also agreed to use the same vertical mixing schemes (TKE and KPP), as well as the same parameter settings. To make sure that our implementations of the TKE and KPP schemes are comparable, we use the versions provided by the CVMix (Community ocean Vertical Mixing) project, which has developed a library of standardised vertical mixing parameterisations to be used in ocean models. In the TKE scheme, we vary the ckparameter (see table 3). For the KPP scheme, we run the models once with the default setting of Ricrit = 0.3, and once with a reduced value of Ricrit = 0.27. We do this because the best value for the critical bulk Richardson number is resolution dependent (e.g. Large et al.,1994). The resolution dependence follows from the bulk Richardson number being an approximation of the exact gradient Richardson number due to the finite thickness of the model levels. As the thickness decreases, the bulk Richardson number converges on the gradient Richardson number. Similarly, the critical bulk Richardson number chosen should converge on the critical gradient Richardson number of 0.25 with decreasing thickness. All model runs with their parameter settings are listed in Table 3. We force all model runs with hourly ERA5 reanalysis (Hersbach et al.,2020), and run them from the end of the 5-year spinup with adjusted mixing parameter settings for two years (2014 and 2015). Of these, we analyse only the second year when the upper ocean has sufficiently adjusted to the changed mixing settings. The year 2015 was chosen because a particularly strong Near-Inertial Wave (NIW) mixing event occured in that year and was observed during a RV Meteor cruise in the tropical North Atlantic, as described and analysed by Hummels et al. (2020). We compare this unique set of observations against our models to assess whether they can reproduce deep reaching NIW mixing events like the one in 2015. One notable difference between ICON-O and FESOM is the parameterisation of the surface fluxes, for which different bulk formulae are used for the ERA5 forcing. To investigate the effect of the bulk formulae, we did an additional run with ICON-O using the standard FESOM bulk formulae. The default bulk formulae in ICON-O for the ERA5 forcing are those of Kara et al. (2002) over ocean and sea ice, with water vapor pressure and 2 m specific humidity calculated using the (modified) equations from Buck (1981) and longwave radiation calculated using Berliand (1952). FESOM instead uses bulk formulae calculated according to Large and Yeager (2009) over the ocean and with constant bulk exchange coefficients over sea ice, as described in Tsu8 3.4 Common settings and experiment descriptions3 DETAILED REPORT ON THE DELIVERABLE jino et al. (2018). These are also implemented in ICON-O, but usually only used together with JRA55-do forcing (Tsujino et al.,2018). The model runs were started on 1 January 2014, and run until 31 December 2015. This time period was chosen because of strong Near-Inertial-Wave (NIW) mixing events that were observed in 2015 (e.g. Hummels et al.,2020), and that we wanted to use for comparison and to assess the different mixing schemes’ and models’ ability to represent the turbulent mixing caused by strong NIWs. The results of this analysis have been submitted to JGR Oceans (Mrozowska et al.,subm.). All model runs were forced with hourly ERA5 data (Hersbach et al.,2020). However, the bulk formulae that are by default used in FESOM and ICON-O together with ERA5 forcing are different. In ICON-O, the formulae from Kara et al. (2002) are used over ocean and sea ice, with the calculation of water vapour pressure and specific humidity following Buck (1981) and the longwave radiation being calculated as in Berliand (1952). In FESOM, bulk formulae from Large and Yeager (2009) are used over the ocean and from Tsujino et al. (2018) over sea ice. All variables are available at 3-hourly intervals, to enable comparison to high-frequency observational data for the analysis of short-term processes, such as the diurnal cycle or near-inertial waves. The output variables that we provide include temperature, salinity, density, velocities, shear, buoyancy frequency, Richardson Number, turbulent kinetic energy, vertical diffusivity and viscosity (3D for the upper 200m), as well as 2D fields like surface fluxes (heat, momentum, freshwater) and mixed layer depth. A list of the model runs from which tropical Atlantic high-frequency output is available is given in Table 3. The two vertical mixing schemes that we tested are two of the most widely used vertical mixing schemes in global ocean models. Both of them are based on the classical eddy diffusivity concept, parameterising the small-scale vertical turbulent fluxes of a variable as the product of a coefficient called eddy diffusivity and the large-scale vertical gradient of the variable in question. In the TKE (turbulent kinetic energy) scheme (Gaspar et al.,1990), the eddy diffusivity is determined as a function of the turbulent kinetic energy, for which an additional prognostic equation is included in the model. In the KPP (K-profile parameterisation) scheme (Large et al.,1994), the eddy diffusivity is instead given as a specified vertical profile in the surface ocean boundary layer. The boundary layer depth is determined as the depth below which the bulk Richardson Number becomes larger than a critical value Ricrit.Ricrit is usually set to 0.3 but should ideally approach the critical gradient Richardson Number of 0.25 with increasingly fine vertical resolution. For KPP, we vary the value of Ricrit in the sensitivity runs. For TKE, we vary the value of the ckparameter (cf. Gaspar et al.,1990, Equation 10). Additionally, we did a number of runs only with ICON-O, changing different aspects of the TKE scheme. One of these has an extension of the TKE scheme: the parameterisation of Langmuir turbulence (Axell,2002). Langmuir turbulence, which is generated through the interaction of wind-driven surface currents and wind-generated surface waves, is responsible for additional turbulent energy input into the upper ocean, and it has been shown to be important over much of the global ocean area (e.g. Belcher et al.,2012). Since we do not simulate surface waves with ICON-O, the effect of the Langmuir turbulence is missing if it is not parameterised. Two runs were done with increased minimum background vertical mixing, one by increasing the minimum background TKE from 10−6J/kg to 10−5J/kg (in this case the diffusivity and viscosity are then still dependent on the stability of the water column), and one by setting a minimum background value directly for the vertical diffusivity (10−5m2/s) and viscosity (10−4m2/s) as it is also done 9 9 FULL TRACK OF DISSEMINATION ACTIVITIES than anticipated, and it is still not running due to technical problems. We therefore decided to do an additional test run with an extension of the TKE scheme instead: the parameterisation of Langmuir turbulence (Axell,2002), as described above. The run with additional Langmuir parameterisation (only ICON-O) is included in the published data sets. Although we did run FESOM with different vertical (and horizontal) resolutions, as proposed in the nextGEMS Grant Agreement for Task 6.5, we decided not to publish these runs but instead do additional test runs with FESOM using the same high vertical resolution as ICON-O and testing the same vertical mixing schemes and parameters. These runs have been described here and included in the published dataset. Doing these coordinated sensitivity experiments with the two models instead of only ICON-O as proposed in the Grant Agreement enabled us to see which changes happen systematically due to changes in the mixing scheme parameters, and which are model dependent. For more details see the report on Deliverable 6.5. 8 Sustainability 8.1 Links built with other deliverables, WPs, and synergies created with other projects This deliverable (D6.4) describes how we made high-frequency model output of the ocean-only sensitivity experiments publicly available for a selected subdomain of particular interest, i.e. the tropical Atlantic. A more detailed description of the results of the vertical mixing sensitivity study can be found in the Deliverable 6.5 report which belongs to the same work package. The ocean-only test runs with the different vertical mixing scheme settings served as a starting point for test runs with the coupled model, and for ICON finally led to changes in the vertical mixing settings from the coupled Cycle 2 to the Cycle 3 run (Work Package 2 ”Model Validation and Development”). Although we found less impact of the vertical mixing settings on model biases than we expected, small improvements could be achieved in the coupled ICON run by changing the vertical mixing settings based on our findings from the ocean-only sensitivity runs. In particular, we had problems in the coupled Cycle 1 run with a large bias in the equatorial Atlantic zonal SST gradient and a large associated westerly trade wind bias. This was already improved slightly in the coupled Cycle 2 run by nudging the SST to reanalysis during the last year of the ocean spinup, but the problem still persisted. With the ocean-only vertical mixing sensitivity runs, we could show that the equatorial zonal SST gradient is (very slightly) improved when the ckparameter in the TKE scheme is reduced to 0.1. Changing this in the coupled ICON model led to a further improvement in the coupled SST/trade wind bias in the tropical Atlantic in Cycle 3. 9 Full track of dissemination activities • Cycle 1 Hackathon 19. – 22. October 2021 in Berlin • ”Science meets Art” - NextCalendArt 2022 • Office Hour with IBERDROLA on solar energy in January 2022 16 10 FULL TRACK OF PUBLICATIONS AND IP • UCM Student Hackathon, 01. – 03. April 2022 near Madrid • Cycle 2 Hackathon 26. June – 02. July 2022 in Vienna • Storms & Ocean summer slackathon 31. June – 06. July 2022 at the Bornö Institute for ocean and climate studies, Sweden • ”Science meets Art” - NextCalendArt 2023 • Mini-hackathon of Storms & Land at Wageningen University 04. – 06. April 2023 • Cycle 3 Hackathon 29. May – 02. June 2023 in Madrid 10 Full track of publications and IP Publications within the project can be found on our website: https://nextgems-h2020.eu/ publications/ 17