Code review comments

[jens-daniel-mueller]

General comments

• [x] Use <- instead of = to assign objects
• [x] Define function arguments explicitly
• [x] Use visual editor
• [x] Avoid manually numbering sections, use # instead
• [x] Remove quite=TRUE in library()
• [x] Define root path at the beginning of script

loading_data.Rmd

• [ ] Difference between argument max_global_cache_age = Inf in argo_update_global() or in argo_set_cache_dir()
• [ ] Difference between argo_global_synthetic_prof() and its alternatives?
  • [ ] Why not use argo_global_bio_prof()?
• [x] Write data sets to .rds files: Revise columns that need to be written

count_observations.Rmd

• [x] Rename to coverage_time_series.Rmd
• [x] Avoid repetition of loading_data.Rmd
• [x] Focus on profile metadata
• [ ] Time series per region

maps_observations.Rmd

• [x] Rename to coverage_maps.Rmd

[jens-daniel-mueller]

List of main changes

• Modified the routine to calculate flag A time series. Same output, more elegant code ;-)
• Load a basinmask that I created from World Ocean Atlas in the framework of my eMLR work. (See: https://jens-daniel-mueller.github.io/emlr_obs_preprocessing/read_World_Ocean_Atlas_2018.html)
• Use basinmask to subset regions.
• For consistency with my eMLR code, I replaced variable names longitude and latitude by lon and lat; and converted longitudes to 20 - 380 °E. This was done for bgc_metadata.
• Load an object “map” that contains a basic ggplot worldmap. The map is consistent with the basinmask and lon-lat formatting, and I use it for plotting.
• Added time series per ocean basin.
• Added an example to save individual plots to file.

Some next tasks

• [ ] Read all new code and try to follow. Let me know if something isn’t clear.
• [ ] Apply lon-lat conversions to all created data sets, in addition to bgc_metadata.
• [ ] Rerun code with all years beyond 2015 (let me know if this causes trouble, we can then discuss how to chunk per year or so)
• [ ] Prepare plots for meeting with Niki.

[jens-daniel-mueller]

• [ ] Subsetting of argo profiles: I suggest to use only argo profiles with quality flag A for pH AND temp. The reasons are (1) that I want to avoid using incomplete profiles with good measurements only at some depths and (2) that we should use the exact same profiles for the pH and temp analysis to make sure we capture sampling biases in time and space similarly for both parameters. At a later stage, we can consider running the analysis for all available temp profiles (not only bgc argo) to assess the impact of reduced sampling density in the bgc argo fleet.

• [ ] Extreme detection: Calculate surface pH/temp climatologies on 1x1 grid. Remove the climatology from each absolute value (i.e. for each month x lon x lat you should have one climatological value that you substract from the 8 corresponding absolute values for each year). The regression should than be performed based on the offset values. For the regression, combine the offset values in a 2x2 grid but without averaging (i.e. use 32 data points for each regression). Finally, plotting the surface extremes and combination with the argo profiles should be done again on the 1x1 grid. Just to make sure: Do not average the climatology onto a 2x2 grid. Otherwise you would loose the information of latitudinal gradients which we want to remove by subtracting the climatology.

• [ ] Understanding of patterns in 4 dimensions (lat, lon, depth, month): As regional and temporal variability in the pH and temp profiles seems to be the major issue for the comparison of the H/N/L profiles, we need to better understand this variability. Can you open two new scripts, one each for pH/temp and visually explore this variability through a couple of plots from different perspectives? Just one example: Plot all profiles separately for each month and biome, and indicate lat or lon by colour. Adding other types of plots in which you change the parameters on the axis and colour scales should help us to gain a quantitative understanding of the variability of pH/temp in space and time.

• [ ] Removal of climatological patterns: We should certainly try to remove regional pH/temp patterns from the profiles before comparing the H/N/L profiles over an entire biome (or even larger region). For this, we require monthly resolved climatologies for pH and temp. As a first step, check the monthly pH climatology from Matt Mazloff.