Can we Constrain Geographical Variability in the Biological Carbon Pump's Transfer Efficiency from Observations?

This repository contains the MATLAB scripts and datasets used for the data analysis and figure generation in the study:

"Can We Constrain Geographical Variability in the Biological Carbon Pump's Transfer Efficiency from Observations?"

A. Rufas¹, S. Khatiwala¹, K. M. Bisson^2,3, A. P. Martin⁴, H. A. Bouman¹

¹Department of Earth Sciences, University of Oxford, Oxford, UK ²Ocean Biology and Biogeochemistry Program, NASA Headquarters, Earth Science Division, Washington, D.C., USA ³Department of Botany and Plant Pathology, Oregon State University, Corvallis, OR, USA ⁴National Oceanography Centre, Southampton, UK

Find the pre-print in the Earth and Space Science Open Archive.

Requirements

To use the content of this repository, ensure you have the following.

• MATLAB version R2021a or later installed.
• Third-party functions downloaded from MATLAB's File Exchange: worstcase, MCErrorPropagation, swtest.m, FMINSEARCHBND, m_map, brewermap, longhurst_v4_2010, subaxis and plotBarStackGroups. Once downloaded, please place the functions in the ./resources/external/ directory.
• MATLAB toolboxes: the Optimization Toolbox, necessary to run worstcase, the Statistics and Machine Learning Toolbox, necessary to run MCErrorPropagation and swtest.m, and the Curve Fitting Toolbox, necessary to run fittype.

Repository Structure

• ./code/: contains the MATLAB scripts for analysing and visualising data (provided, see "MATLAB Scripts" section).
• ./data/
  • ./raw/: raw data compiled for this study (partly provided, see "Obtaining Raw Data" section).
  • ./interim/: input processed data for MATLAB scripts (provided, see "Obtaining Interim Data" section).
  • ./processed/: intermediate or input data used by the MATLAB scripts for further processing (provided).
• ./resources/
  • ./external/: third-party resources for plotting and functions (see "Requirements" section).
  • ./internal/: custom MATLAB functions generated specifically for calculating (provided).
• ./figures/: figures generated from processed data (provided).

Obtaining Raw Data

• Particulate organic carbon (POC) flux measurements from sediment traps and radionuclides, a compilation that we made for this study. The file, dataset_s0_trap_and_radionuclide_compilation.xlsx, is not available within this repository as it contains data owned by other authors that are not in a preservation repository. References for constructing this dataset are provided in the Supporting Information of our paper (Tables S1-S6), and a processed version of the dataset is available in the folder ./data/processed/pocflux_compilation.mat.
• Particle concentration measurements from the Underwater Vision Profiler 5 (UVP5) downloaded from the EcoPart repository hosted by IFREMER. This data are stored in the subfolder ./data/raw/UVP5/ and also in a processed format in ./data/processed/UVP5/pocflux_bisson_45sc.mat.

Obtaining Interim Data

• World Ocean Atlas 2023 annual climatology for temperature. This dataset, which I downloaded from the NCEI NOAA website, is stored in the file temp_annual_woa23.mat. It is required for running the Marsay et al. (2015) algorithm (refer to Text S5 in the Supporting Information).
• The VHRR Pathfinder v.5.0 global 4 km monthly climatology (1985–2001) for sea surface temperature (SST). This dataset, which I downloaded from the NCEI NOAA website, is stored in the file sst_pathfinder_v5.mat. It is required for running the Henson et al. (2012) algorithm (refer to Text S5 in the Supporting Information).
• Carr 2002 net primary production (NPP) monthly climatology. I constructed this climatology using the Carr 2002 algorithm for NPP, which incorporated inputs from (i) the SeaWiFS 9 km monthly climatology (1997-2010) for chlorophyll concentration and photosynthetically available radiation (PAR₀), available for download from the NASA website, and (ii) the AVHRR Pathfinder v.5.0 global 4 km monthly climatology (1985–2001) for SST. The resulting file, npp_carr2002_seawifs_pathfinder.mat, is essential for running the Henson et al. (2012) algorithm (refer to Text S5 in the Supporting Information).
• Euphotic layer depth (z_eu) monthly climatology. I constructed this climatology using two data products from the E.U. Copernicus Marine Service (CMEMS): the diffuse attenuation coefficient at 490 nm (k_d(490)) from the Copernicus-GlobColour merged product based on satellite observations, and the mixed layer depth (MLD) from the physics reanalysis product. I have used the 0.1% light depth for the calculation, as recommended by Buesseler et al. (2020). The file, zeu_calculated_kdcmems_mldcmems_pointonepercentpar0.mat, is necessary to run the functions for fitting Martin's b and the remineralisation length scale coefficient (z*).

The scripts that I used to read the .nc files downloaded from the web repositories and generate the .mat files presented above are available in this related repository.

MATLAB Scripts

The following scripts have been run in this order to analyse the data and reproduce the figures in our paper.

Num	Script name	Script action
1	plotGlobalMapWithTimeseriesStations.m	Creates Figure 1
2	processPocFluxFromTrapAndRadCompilation.m	Processes Dataset S0
3	plotPocFluxFromTrapAndRadCompilation.m	Creates Figure 2, S1 and S2
4	processPocFluxFits.m	Calculates b for various scenarios using Dataset S0 and generates Dataset S1
5	plotPocFluxFits.m	Creates Figure S3
6	processPocFluxFromUvp.m	Processes the UVP5 dataset downloaded from Ecopart
7	plotPocFluxFromUvp.m	Creates Figure S5
8	findAndPlotUvpVsCompilationPocFluxMatchups.m	Creates Figure 3
9	processBcpMetrics.m	Calculates b, z* and Teff using the best method determined after script 7 and generates Dataset S2
10	plotBcpMetrics.m	Creates Figure 4 and Figure S6
11	examineEffectSampleSize.m	Analyses Dataset S0 to examine effect of sample size in the calculations of the relative error of b and z*

The following are helper functions used by the scripts above:

Num	Script name	Script action
12	calculateBcpMetricsFromTrapAndRadCompilation.m	Steps to calculate and propagate error for b, z* and Teff from Dataset S0; called by scripts 4 and 9
13	calculateBcpMetricsFromUvp.m	Steps to calculate and propagate error for b, z* and Teff from the UVP5 dataset; called by script 9
14	calculateBcpMetricsFromHenson2012.m	Steps to calculate and propagate error for b, z* and Teff using the algorithm of Henson et al. (2012); called by script 9
15	calculateBcpMetricsFromMarsay2015.m	Steps to calculate and propagate error for b, z* and Teff using the algorithm of Marsay et al. (2015); called by script 9
16	propagateErrorWithMCforMartinbAndZstar.m	Monte Carlo error propagation algorithm for b and z* fits; called by scripts 12 and 13
17	propagateErrorWithMCforTeff.m	Monte Carlo error propagation algorithm for Teff; called by scripts 12 and 13
18	solveMartinbAndZstar.m	Algorithm to solve b and z*; called by scripts 11 and 16
19	samplePocFluxWithLH.m	Latin Hypercube sampling algorithm for POC flux profiles; called by scripts 11 and 16
20	binPocFluxDataAtRegularDepthIntervals.m	Low-level function, bins POC flux data by depth intervals; called by scripts 11 and 12
21	extractDataFromZrefToZmeso.m	Low-level function, extracts POC flux data from the reference depth to the base of the mesopelagic zone; called by scripts 5, 11, 12 and 13
22	extractDataFromZrefToEnd.m	Low-level function, extracts all POC flux data below the reference depth; called by script 5
23	constructFilenameFitMetrics.m	Low-level function, constructs file names for b fits for different scenarios; called by scripts 5, 6 and 16

Reproducibility

Our scripts showcase the application of Monte Carlo-based techniques for error propagation across diverse datasets of biological carbon pump (BCP) mesopelagic transfer efficiency metrics. Specifically, the scripts are hard-wired to handle data from our six designated study sites:

• the Hawaii Ocean Time-series (HOT) station ALOHA (HOT/ALOHA), in the subtropical NE Pacific (22.45ºN, 158ºW);
• the Bermuda Atlantic Time-Series/Oceanic Flux Program joint site (BATS/OFP), in the subtropical NW Atlantic (31.6ºN, 64.2ºW);
• the US JGOFS Equatorial Pacific process study experimental site (EqPac), in the central equatorial Pacific upwelling system (–2 to 2ºN, 140ºW);
• the Porcupine Abyssal Plain time-Series Observatory (PAP-SO), in the subpolar NE Atlantic (49.0ºN, 16.5ºW);
• Ocean Station Papa (OSP), in the HNLC region of the subpolar NE Pacific (50ºN, 145ºW), and
• the Long-Term Ecological Research (LTER) observatory HAUSGARTEN, in the polar Atlantic-Arctic boundary (79ºN, 4.0ºE).

Acknowledgments

This work was completed as part of my PhD project at the University of Oxford under the NERC large grant COMICS (Controls over Ocean Mesopelagic Interior Carbon Storage, NE/M020835/2). I also acknowledge funding from the University of Oxford's Covid-19 Scholarship Extension Fund and Oxford's Wolfson College Covid-19 Hardship Fund.