readProfiles() and diagnostics should handle flags 1, 2 ,5, and 8

[j-harbin]

As shown in the screenshot below, the Argo User Manual [1] states (on page 79) that 'good' data are considered to be 1, 2, 5 or 8. Flags 9 (missing) are not used, and all other flags are considered 'bad' data.

[1] Carval, Thierry, Bob Keeley, Yasushi Takatsuki, Takashi Yoshida, Stephen Loch, Claudia Schmid, Roger Goldsmith et al. "Argo User's Manual, Version 3.2. December 29th 2015.[SUPERSEDED]." (2015).

[dankelley]

@j-harbin please note from https://github.com/dankelley/oce/issues/1704 that I plan to do this in 'oce' (given approval of coauthors) tomorrow, but 'll do it in argoFloats today.

[dankelley]

Actually, it turns out that oce is set up (intentionally) to prevent the user from changing default flag-mapping schemes using oce::initializeFlagSchemeInternal(). There is a way for argoFloats to over-ride that, but I do not think we should, because then a user would get different behaviour by using argoFloats::readProfiles("file.nc") than they would with oce::read.argo("file.nc") and that is not a good thing.

The only solution is that anyone wanting the solution to this issue must first update oce from github, using

devtools::install_github("dankelley/oce", ref="develop")

which is admittedly a bit of a hassle, but oce users do that all the time, so it should not be regarded as an unusual hurdle.

Once that's done, using

library(argoFloats)
argos <- readProfiles("~/data/argo/SR2902204_131.nc")
argo1 <- argos[[1]]
summary(argo1)

yields as below, the last line of which tells us that things flagged 0, 3, 4 and 9 will get set to NA, but that all other values will be as stored in the netcdf file.

* Processing Log

    - 2020-06-10 19:31:10.008 UTC: `create 'argo' object`
    - 2020-06-10 19:31:10.188 UTC: `read.argo(file="/Users/kelley/Dropbox/data/argo/SR2902204_131.nc")`
    - 2020-06-10 19:31:10.189 UTC: `initializeFlagScheme(object, name="argo", mapping=list(not_assessed=0,passed_all_tests=1,probably_good=2,probably_bad=3,bad=4,changed=5,not_used_6=6,not_used_7=7,estimated=8,missing=9)), default=c(0,3,4,9))`

Therefore, I think things will be ready for testing tomorrow.

[j-harbin]

As discussed in the vftf to @dankelley today, he will make argoFloats specifically catch these flags.

[dankelley]

I think this working now, in "develop" commit b0847fc17b0c51596c6d64556034897a89f0aaaf.

A test code is

library(argoFloats)
data(index)
s <- subset(index,1:5)
p <- getProfiles(s)
a <- readProfiles(p)
summary(a[[1]])

where you will note near the end of the summary that the oce flagScheme has been overridden. You can see the installed scheme with

str(a[[1]][["flagScheme"]])

which should give

List of 3
 $ name   : chr "argo"
 $ mapping:List of 10
  ..$ not_assessed    : num 0
  ..$ passed_all_tests: num 1
  ..$ probably_good   : num 2
  ..$ probably_bad    : num 3
  ..$ bad             : num 4
  ..$ changed         : num 5
  ..$ not_used_6      : num 6
  ..$ not_used_7      : num 7
  ..$ estimated       : num 8
  ..$ missing         : num 9
 $ default: num [1:4] 0 3 4 9

and I think this issue can be closed, if you get that. If you don't get that, please add a comment showing what you get, and leave the issue open.

Thanks.

[j-harbin]

I can confirm I also receive this. I will now close this issue.

[j-harbin]

Hi Team,

I realized I closed this issue prematurely. I have made diagnostics handle flags 1,2,5 and 8 in commit 2d79f942f84c4de94832ff72e437be8cc477ead7 in the develop branch.