In dankelley/oceanglider: Analysis of Oceanographic Glider Data

knitr::opts_chunk$set(collapse = TRUE, comment = "#>")

Abstract. This vignette explains the basics of using the oceglider package on Slocum data. The main difference from SeaExplorer files is in the layout of the objects, with SeaExplorer basically considering glider and payload in separate streams. But normal use, with plotting, subsetting, etc., does not require the user to consider the details of the layout.

For this vignette to be suitable for CRAN submission, it must not download files to the user's working space. We avoid that by setting aside space in the next line.

fileNC <- tempfile(fileext = ".nc") # to hold a binary (NetCDF) file

Note that this fileNC file will be deleted at the end of this vignette. Obviously, in actual practice, the analyst is likely to store the downloaded file in a more permanent location, and not to delete it when finished a particular task.

Some Common Steps with Slocum Datasets

We will use a dataset from CPROOF (Reference 1).

library(oceglider)
url <- paste0(
    "https://cproof.uvic.ca/gliderdata/deployments/",
    "dfo-bumblebee998/dfo-bumblebee998-20221207/",
    "L0-timeseries/",
    "dfo-bumblebee998-20221207_delayed.nc"
)
file <- gsub(".*/", "", url)
# 45.8 MB file: download takes 0.4s user, 0.9s system but 11s elapsed
if (!file.exists(file)) {
    system.time(curl::curl_download(url = url, destfile = file, mode = "wb"))
}
g <- read.glider.netcdf(file) # under 0.2s

It always makes sense to get a summary, to see what data are contained in the object.

summary(g)

Next, make a map to see where the sampling was done.

plot(g, which = "map")

This map comports with the one on Reference 2, except we have proper scaling of the axes, because oceglider computes an aspect ratio that will preserve shapes at the centre of the plot.

Next, examine the pressure-time trajectory.

plot(g, which = "p")

Note that there is a large region of whitespace on the right of the plot panel. This is because the dataset has time values in that interval, but all the pressure values are NA. The subset() function could be used to trim the dataset, but that is not necessarily wise because there could be other valuable data (even just longitude and latitude are valuable, in case the hydrographic sensors provide nothing usable).

We can show temperature on a similar plot, by switching from lines to points, and colour-coding by temperature.

plot(g, which = "p", type = "p", pch = 20, cex = 0.2, colorby = "temperature")

Notice that the colour palette ranges up to red, although only a small region of the graph has that colour. but a quick glance at the graph does not reveal many points that are of sufficient temperature to be indicated with that color. The methods illustrated in the SeaExplorer vignette can be employed here, but for brevity we will just use percentile trimming to get a diagram that at least provides an initial view of the data.

plot(g,
    which = "p", type = "p", pch = 20, cex = 0.2, colorby = "temperature",
    colorbylim = quantile(g[["temperature"]], c(0.01, 0.99), na.rm = TRUE)
)

Readers might agree that this version looks better, displaying more features than in the first plot.

Salinity is often a problem, but we'll do the same thing, first plotting all the data.

plot(g, which = "p", type = "p", pch = 20, cex = 0.2, colorby = "salinity")

And, second, trimming.

plot(g,
    which = "p", type = "p", pch = 20, cex = 0.2, colorby = "salinity",
    colorbylim = quantile(g[["salinity"]], c(0.01, 0.99), na.rm = TRUE)
)

Again, we see improvement.

At this stage, we have some confidence that data are being read properly and that plotting, extraction with [[ and some other basic features work. Readers are directed to the SeaExplorer vignette for more ideas on things that might be worth pursuing.

As noted at the outset, CRAN rules dictate that the downloaded material be deleted. In actual analysis, though, it is unlikely that such material would be deleted, so don't just copy this line!

unlink(fileNC) # where the binary (NetCDF) file went

References

1. “dfo-bumblebee998-2022-12-07 Time series Delayed” Accessed September 5, 2024. https://cproof.uvic.ca/gliderdata/deployments/dfo-bumblebee998/dfo-bumblebee998-20221207/

dankelley/oceanglider documentation built on June 8, 2025, 4:20 a.m.