Read a Seabird CTD File

  columns = NULL,
  station = NULL,
  deploymentType = "unknown",
  btl = FALSE,
  monitor = FALSE,
  encoding = "latin1",
  debug = getOption("oceDebug"),



either a connection or a character value naming a file. For read.ctd.sbe() and read.ctd.woce(), this may be a wildcard (e.g. "*.cnv" or "*.csv") in which case the return value is a vector containing CTD objects created by reading the files from list.files() with pattern set to the specified wildcard pattern.


an optional list that can be used to convert unrecognized data names to resultant variable names. This is used only by read.ctd.sbe() and read.ctd.odf(). For example, if a data file named salinity as "SAL", then using

d <- read.ctd(f, columns=list(

would assign the "SAL" column to the salinity entry in the data slot of the CTD object returned by the read.* function.


optional character string containing an identifying name or number for the station. This can be useful if the routine cannot determine the name automatically, or if another name is preferred.


optional missing-value flag; data matching this value will be set to NA upon reading. If this is provided, then it overrules any missing-value flag found in the data. For Seabird (.cnv) files, there is usually no need to set missingValue, because it can be inferred from the header (typically as -9.990e-29). Set missingValue=NULL to turn off missing-value detection, even in .cnv files that contain missing-value codes in their headers. If missingValue is not specified, then an attempt is made to infer such a value from the data, by testing whether salinity and/or temperature has a minimum that is under -8 in value; this should catch common values in files, without false positives. A warning will be issued in this case, and a note inserted in the processing log of the return value.


character string indicating the type of deployment. Use "unknown" if this is not known, "profile" for a profile (in which the data were acquired during a downcast, while the device was lowered into the water column, perhaps also including an upcast; "moored" if the device is installed on a fixed mooring, "thermosalinograph" (or "tsg") if the device is mounted on a moving vessel, to record near-surface properties, or "towyo" if the device is repeatedly lowered and raised.


a logical value, with TRUE indicating that this is a .BTL file and FALSE (the default) indicating a .CNV file. Note that if btl is TRUE, the data column names are taken directly from the file (without e.g. translating to "Sal00" to "salinity". Also, the "avg" and "sdev" columns are blended together, with all the latter named as in the file, but with "_sdev" appended.


boolean, set to TRUE to provide an indication of progress. This is useful if filename is a wildcard.


a character value that indicates the encoding to be used for this data file, if it is textual. The default value for most functions is "latin1", which seems to be suitable for files containing text written in English and French.


an integer specifying whether debugging information is to be printed during the processing. This is a general parameter that is used by many oce functions. Generally, setting debug=0 turns off the printing, while higher values suggest that more information be printed.


if provided, the action item to be stored in the log. This is typically only provided for internal calls; the default that it provides is better for normal calls by a user.


additional arguments, passed to called routines.


This function returns a ctd object.


This function reads files stored in Seabird .cnv format. Note that these files can contain multiple sensors for a given field. For example, the file might contain a column named t090C for one temperature sensor and t190C for a second. The first will be denoted temperature in the data slot of the return value, and the second will be denoted temperature1. This means that the first sensor will be used in any future processing that accesses temperature. This is for convenience of processing, and it does not pose a limitation, because the data from the second sensor are also available as e.g. x[["temperature1"]], where x is the name of the returned value. For the details of the mapping from .cnv names to ctd names, see cnvName2oceName().

The names of the elements in the data slot of the returned value depend on the file type, as signalled by the btl argument. For the default case of .cnv files, the original data names as stored in file are stored within the metadata slot as dataNamesOriginal, and are displayed with summary alongside the numerical summary; see the Appendix VI of reference 2 for the meanings of these names (in the "Short Name" column of the table spanning pages 161 through 172). However, for the case of .btl files, the column names are as described in the documentation entry for the btl argument.

A note on hand-entered headers

CNV files may have a section that contains human-entered information. This is detected by read.ctd.sbe() as lines that begin with two asterisks. Decoding this information can be tricky, because humans have many ways of writing things.

For example, consider the date item in the metadata slot of the returned value. read.ctd.sbe() infers this value in one of two ways. First, if there is a header line staring with

* NMEA UTC (Time) =

then that value is decoded and used for date. This header line, preceded by a single asterisk, is not human-entered, and so there is reason to hope for a uniform format that can be handled by read.ctd.sbe(). However, if there is no NMEA header line, then read.ctd.sbe() will look for a line starting with

** Date:

which was human-entered. This is the second choice, because humans write dates in a bewildering variety of ways, and as.POSIXct(), which read.ctd.sbe uses to parse the date, cannot handle them all. If there is a problem, read.ctd.sbe() issues a warning and stores NA in date.

A similar error-detection procedure is used for human-entered location data, which appear in lines starting with either

** Longitude:


** Latitude:

which often take forms that read.ctd.sbe() cannot parse.

It is important to note that, even if no warnings are issued, there is a reasonably high chance that human-entered data will be scanned incorrectly. (Did the operator remember to indicate the hemisphere? Does 123.456 indicate decimal degrees, or 123 degrees plus 45.6 minutes? Is hemisphere indicated by sign or by letter, and, if the latter, where does it appear?)

In deep-sea work, a ship might steam for 6 hours between CTD stations, so the ship-time cost of each CTD file can be several thousand dollars. Surely it is not unreasonable for an analyst to take a minute to glance at the CNV file, to ascertain whether read.ctd.sbe() inferred correct values.

oceSetMetadata() is helpful for correcting problems with individual files, but if many files are systematically problematic, say for a whole cruise or perhaps even for a whole institution, then it might sense to set up a wrapper function to correct deficiencies in the CNV files. As an example, the following handles dates specified in a particular nonstandard way.

read.ctd.sbe.wrapper <- function(cnv)
    lines <- readLines(cnv)
    # Change month-day-year to year-month-day, so as.POSIXct() can parse it.
    lines <- gsub("^\\*\\* Date: (.*)-(.*)-(.*)", "** Date: \\3-\\1-\\2", lines)

A note on sampling times

Until November of 2018, there was a possibility for great confusion in the storage of the time entries within the data slot, because read.ctd.sbe renamed each of the ten variants of time (see reference 2 for a list) as "time" in the data slot of the returned value. For CTD profiles, this was perhaps not a great problem, but it could lead to significant confusion for moored data. Therefore, a change to read.ctd.sbe was made, so that it would Seabird times, using the start_time entry in the CNV file header (which is stored as startTime in the object metadata slot), along with specific time columns as follows (and as documented, with uneven clarity, in the SBE Seasoft data processing manual, revision 7.26.8, Appendix VI):

timeSseconds elapsed since start_time
timeMminutes elapsed since start_time
timeHhours elapsed since start_time
timeJJulian days since the start of the year of the first observation
timeNNMEA-based time, in seconds past Jan 1, 1970
timeQNMEA-based time, in seconds past Jan 1, 2000
timeKNMEA-based time, in seconds past Jan 1, 2000
timeJV2as timeJ
timeSCPas timeJ
timeYcomputer time, in seconds past Jan 1, 1970

NOTE: not all of these times have been tested properly, and so users are asked to report incorrect times, so that read.ctd.sbe can be improved.

A note on scales

The user might encounter data files with a variety of scales for temperature and salinity. Oce keeps track of these scales in the units it sets up for the stored variables. For example, if A is a CTD object, then A[["temperatureUnit"]]$scale is a character string that will indicate the scale. Modern-day data will have "ITS-90" for that scale, and old data may have "IPTS-68". The point of saving the scale in this way is so that the various formulas that deal with water properties can account for the scale, e.g. converting from numerical values saved on the "IPTS-68" scale to the newer scale, using T90fromT68() before doing calculations that are expressed in terms of the "ITS-90" scale. This is taken care of by retrieving temperatures with the accessor function, e.g. writing A[["temperature"]] will either retrieve the stored values (if the scale is ITS-90) or converted values (if the scale is IPTS-68). Even though this procedure should work, users who really care about the details of their data are well-advised to do a couple of tests after examining the first data line of their data file in an editor. Note that reading a file that contains IPTS-68 temperatures produces a warning.


  1. The Sea-Bird SBE 19plus profiler is described at Some more information is given in the Sea-Bird data-processing manual (next item).

  2. A SBE data processing manual was once at, but as of summer 2018, this no longer seems to be provided by SeaBird. A web search will turn up copies of the manual that have been put online by various research groups and data-archiving agencies. As of 2018-07-05, the latest version was named SBEDataProcessing_7.26.4.pdf and had release date 12/08/2017, and this was the reference version used in coding oce.


Dan Kelley and Clark Richards


f <- system.file("extdata", "ctd.cnv", package="oce")
d <- read.ctd(f)
#> Warning: startTime (1903-10-15 11:38:38) is < 1950, suggesting a turn-of-the-century problem in this cnv file
#> Warning: this CNV file has temperature in the IPTS-68 scale and this is stored in the object, but note that [["temperature"]] and the sw* functions convert the numbers to ITS-90 values