AR6 WG1 Chap3 Figure 3.26 Ocean Heat Content

Figure number: Figure 3.26 From the IPCC Working Group I Contribution to the Sixth Assessment Report: Chapter 3

Description:

This is a time series of ocean heat content, with four panes. On the left, the figure shows the change (relative to 1971) in heat content of the total global ocean. The red area is the 5-95 percentile range of the CMIP6 ensemble and the grey area is the observational record.

On the right hand side, the three panes show the depth ranges 0-700m, 700m-2000m, and below 2000m.

The heat content was calculated using the TEOS-10 gsw (v3.4.0) python toolkit: https://teos-10.github.io/GSW-Python/

In all cases, the model temperature and salinity were de-drifted against the pi-control. From there, we ensured that the pressure was calculated correctly, the cell volume was available, we use absolute salinity & conservative temperature.

The multi model mean is calculated such that each modelled ensemble has the same weight. ie One-model, one vote.

From there the calculation of the "heat content" was:

1. The global ocean heat content is interpreted to be calculated as the volume integral of the product of in situ density, ρ, and potential enthalpy, h0 (with reference sea pressure of 0 dbar).

2. The in situ density is calculated using gsw_rho(SA,CT,p).
   Here the actual pressure at the target depth is used (i.e., the mass of the water).

3. The surface referenced enthalpy should be calculated using gsw_enthalpy(SA,CT,0). Note that here the 0 dbar pressure is critical to arrive at the surface value, which is the value of enthalpy and absolute salinity that is available for exchange with the atmosphere.

4. The product of the in situ density times the surface-referenced enthalpy is the relevant energy quantity: gsw_rho(SA,CT,p)*gsw_enthalpy(SA,CT,0)

5. For the anomalous energy, we calculate: gsw_rho(SA,CT,p)*gsw_enthalpy(SA,CT,0)-gsw_rho(,,p)*gsw_enthalpy(,,0).

6. Integrate the surface-referenced enthalpy times rho, i.e., the previous line gives the 3D integrand, and then we want to integrate it from the bottom up.

Author list:

• Lee de Mora, Plymouth Marine Laboratory, UK; ledm@pml.ac.uk
• Paul J. Durack, Lawrence Livermore National Laboratory, USA; durack1@llnl.gov
• Nathan Gillett, University of Victoria, Canada
• Krishna Achutarao, Indian Institute of Technology, Delhi, India
• Shayne McGregor, Monash University, Melbourne, Australia
• Rondrotiana Barimalala, University of Cape Town, South Africa
• Elizaveta Malinina-Rieger, Environment and Climate Change Canada
• Valeriu Predoi, University of Reading, UK
• Veronika Eyring, DLR, Germany

Publication sources:

• Durack, P., Gleckler, P., Landerer, F. et al. Quantifying underestimates of long-term upper-ocean warming. Nature Clim Change 4, 999–1005 (2014). https://doi.org/10.1038/nclimate2389

ESMValTool Branch:

• ESMValTool-AR6-OriginalCode-FinalFigures: ar6_chap_3_ocean_figures

ESMValCore Branch:

• ESMValCore-AR6-OriginalCode-FinalFigures: optimize_mem_annual_statistic_plus_amoc

Recipe & diagnostics:

Recipe used:

• recipes/recipe_ocean_heat_content_TSV_all.yml

Diagnostic used:

• ocean/diagnostic_chap3_ocean_heat_content.py

The OHC, Halo and SSS trends plots are all produced using tyhe same recipe and diagnostic. This is because they all require the same process to de-drift.

Expected image path:

This is the path of the image relative to the automatically generated ESMValTool output location:

• plots/diag_ohc/diagnostic/multimodel_ohc/ multimodel_ohc_range_10-90_large_full_1995.0-2014.0.png

Recipe generation tools:

Two scripts are included to populate this recipe:

• check_TSV.py
• recipe_filler.py

Check_TSV is a tool to generate the dataset list in the recipe_ocean_heat_content_TSV_all.yml recipe.

This tool is relatively complex, as it needs to find all possible cases where the following six datasets exist for a given model & ensemnle member:

• historical temperature (thetao)
• historical salinity (so)
• piControl temperature (thetao)
• piControl salinity (so)
• volcello: both historical and piControl for models where volume varies with time.
• volcello: piControl only for models where volume is fixed in time.

The tool checks that the data for all these 5 or 6 datasets must be available for the entire time range. In addition, the tool checks where the historical was branched from the piControl and adds the relevant picontrol years.

The recipe filler is an earlier and more general version of the check_TSV.py tool. It can be used to add whatever data is available into a recipe. I believe that a version of it was added to the ESMValTool master by Valeriu.

Ancillary figures and datasets:

In addition to the main figure, diagnostics may produce several figures and datasets along the way or several versions of the main figure.

The OHC diagnostic produces the OHC, SSS trends and Halosteric SLR figures. This code is particularly complex and several ancillary figures are produced along the way for each model and each ensemble member.

These figures include the following directories related to the de-drifting process:

• piControl:
  • maps showing the raw temperature and salinity data at the surface at the final time step of the PI control run.
• piTrend:
  • histograms showing the distribution of the de-drifting linear regression (slope & intersect)
• slope:
  • maps showing the slope over the surface for the entire PI control
• intersect:
  • maps showing the intersect over the surface for the entire PI control
• trend_intact:
  • maps showing the raw temperature and salinity data at the surface at the final time step of historical and hist-nat run
• detrended:
  • maps showing the dedrifted temperature and salinity data at the surface at the final time step of historical and hist-nat run.
• detrended_quad:
  • 4 pane figure showing the surface map for the historical detrended, trend-intact, the difference and the quotient.
• vw_timeseries:
  • time series figure showing the volume weighted mean for the detrended and trend intact.
• detrending_ts:
  • time series figure showing the global volume weighted mean (or total) temperature, salinity or OHC for the historical and piControl.
• multi_model_mean:
  • shows maps of the multi-model mean surface temperature and salinity at various points in time and specific time ranges.

The following directories contain figures related to the Ocean Heat Content calculation:

• detrending_ts:
  • time series figure showing the global volume weighted mean (or total) temperature, salinity or OHC for the historical and piControl.
• ohc_summary:
  • Single model ensemble version of the final figure, showing each volume range.
• OHC_full_instact and OHC_full_detrended:
  • map showing the full water column OHC for each ensemble member at various points in time.
• ohc_ts:
  • single model time series figure showing the time development of the detrended OHC.
• dynheight_ohcAtlantic & dynheight_ohcPacific:
  • map showing the dynamic height in the Atlantic and pacific regions (useful to heightlift the regional maps.)
• multimodel_ohc:
  • The full ocean heat content figure.

Additional datasets:

The observational data for this figure is take from the file:

• 210204_0908_DM-AR6FGDAssessmentTimeseriesOHC-v1.csv All columns are used in the final figure.

These files were downloaded directly from Paul Durack via the invite-only google drive page: https://drive.google.com/drive/folders/1VO2FehHCz1zJu8tLvp1dNPF2IURJudJN

Software description:

• ESMValTool environment file: IPCC_environments/development_ar6_chap_3_ocean_environment.yml
• pip file: IPCC_environments/development_ar6_chap_3_ocean_pip_environment.txt

Hardware description:

What machine was used: Jasmin When was this machine used? December 2020 to March 2021

Any further instructions:

While this code was written for the IPCC report, there are several limitations and potential sources of error. In this section, we document some potential problems.

This code uses shelve files, which are sometimes not portable between different versions of python.

We cannot guarantee that the auxiliary data will remain available indefinitely.

If the hatching is turned on in the Halosateric SLR figure, and the multi_model_agrement_with_* figures do not exist, then the code will try to create a new figure while another is unfinished. This will break matplotlib.

The dedrifting is calculated using the entire picontrol instead of limiting it to the period specific in the historical run. Other analyses have used shorter periods for their dedrifting range. This method was chosen due to the time constraints.

Other analyses have used polymetric dedrifting, to remove even more of the picontrol trend from the hisotircal run, where as we used a much linear regression straight line fit.

The DAMIP experiment has the flaw that the Omon wasn't required to contribute the cell volume. This means that the hist-nat datasets do not include any time-variying cell volume data. To maximize the data available, we assume that the hist-nat data can use the mean along of the time axis of the pre-industrial control data.

We have interchangeably used the terms de-drifting and de-trending, but the correct term for the process that we've applied is de-drifting. When something is marked as de-trended, it is actually dedrifted.

Additional information can be found in IPCC_README_files/IPCC_AR6_Chap3_Ocean_figures_notes_ledm_README.md