Wong et al. 
provided to users. The formatted and flagged data are passed 
an to the two Argo GDACSs in netCDF files, as well as inserted 
onto the Global Telecommunications System (GTS) in the Binary 
Universal Form for the Representation (BUFR) format. The 
3UFR format replaced the earlier TESAC code form in 2018, 
which did not allow the inclusion of quality flags. The GTS 
channel is mainly used by operational meteorological agencies. 
Available within < 24h of satellite transmission, the real-time 
data are used for operational weather and ocean forecasting, data 
assimilation, and other applications that require timely data that 
are not necessarily of final and highest possible quality. 
Delayed-Mode 
[n the delayed-mode process, data are subjected to visual 
examination by oceanographic experts and are re-flagged where 
necessary, as the real-time automatic procedures are not flawless. 
Float data can also be affected by sensor drift, but because 
retrieving floats for recalibration is rarely possible, statistical 
tools and climatological comparisons are used to adjust the data 
for sensor drift when needed. Determination of sensor drift 
requires accumulation of a relatively long time series. In Argo, 
the usual practice is to examine the profiles in delayed-mode 
initially about 12 months after they are collected, and then 
revisit several times as more data from the floats are obtained, 
until the floats become inactive. Thus, the most recent version 
of the global dataset should be used whenever possible to take 
advantage of these activities. The delayed-mode pathway aims 
to provide the highest-quality version of the data and includes 
realistic error estimates. Both the raw and adjusted versions of 
the data are retained, as well as comments on what adjustments 
have been made to the data. Delayed-mode quality data are 
suitable for use in scientific applications that require high 
accuracy, such as climate research. 
In order to enhance the real-time and delayed-mode pathways 
for detecting data errors, three additional independent global 
analyses have been added to the Argo data system. First, since 
2010, a satellite altimetry comparison is performed every 3 
months at CLS, France, in partnership with the French GDAC at 
Coriolis. For each float time series, the steric heights from Argo 
profiles are compared with independent and contemporaneous 
(Le., collocated in time and space) satellite altimetric height 
estimates (Guinehut et al., 2009). The comparison provides an 
overview of the behavior of the time series of the floats and can 
detect outliers in the float measurements, including those that 
may be affected by sensor drift or calibration offsets. Second, a 
statistical procedure for detecting outliers by exploiting mapping 
error residuals is performed daily at Coriolis (Gaillard et al., 
2009). This method detects float data that are not consistent with 
their neighbors in time and space. And third, since 2019, a daily 
MIN-MAX test (Gourrion et al., 2020) has been implemented at 
Coriolis to compare float profiles with a climatology of minimum 
and maximum values computed from Argo delayed-mode data 
and high-quality CTD data. This aids in the identification of 
sensor drift at an early stage. Results of these global analyses are 
sent to the DACs regularly, where the anomalies are flagged or 
adjusted by expert examination. 
rontiers in Marine Science | www.frontiersin.or 
Argo Data 1999-2019 
Since 2013, regional reanalysis of delayed-mode salinity data 
has been performed regularly at Coriolis. For each float that has 
been processed in delayed-mode, the OWC method (Owens and 
Wong, 2009; Cabanes et al., 2016) is run with four different sets of 
spatial and temporal decorrelation scales and the latest available 
reference dataset. If the salinity adjustments obtained from the 
four runs all differ significantly from the existing adjustment, 
then the salinity data from the float are re-examined and a new 
adjustment is suggested if necessary. This step has been proven 
to be effective in increasing consistency of delayed-mode salinity 
adjustments for floats in the North Atlantic Ocean. 
The final component of the Argo data system is a network 
monitoring system developed by JCOMMOPS. This was 
developed as a float tracking service to ensure compliance 
with Intergovernmental Oceanographic Commission (IOC) 
resolutions regarding Argo, and subsequently expanded. It 
monitors the status of data availability at the GDACs and 
provides Key Performance Indicators on the implementation of 
the data system. 
Extension of the Argo Data System 
The Argo data system has had to expand its capacity in response 
to the advent of new capabilities of the profiling floats. In 2014, 
the Argo data system underwent a major format change to 
manage mission changes due to two-way communications via 
[ridium, to better accommodate biogeochemical profiles, to cope 
with different vertical sampling schemes, and to store more 
metadata (Argo Data Management Team, 2019). A large effort 
was put into homogenizing the metadata and technical data files 
to facilitate comparisons of float and sensor models, tracking of 
the health of the array, and identifying of floats with potentially 
bad sensors by serial numbers. The trajectory data files were 
revamped to include more information about the events during 
a float mission cycle and the times associated with these events. 
The profile data files were re-formatted to allow multiple profiles 
from a single sampling cycle (instead of the traditional limit 
of one profile per cycle). The ability to store multiple profiles 
within one cycle has allowed the addition of biogeochemical data 
and other specialized data, such as the un-pumped temperature 
measurements, in the profile data files. 
The Argo Program presently consists of three elements: 
Core, Biogeochemical (BGC), and Deep (Figure 5). Core-Argo 
is concerned with the standard mission of sampling CTD data 
from 0 to 2,000 dbar every 10 days. Deep-Argo aims to sample 
temperature and salinity over the full ocean depth up to 6,000 
dbar. BGC-Argo is based on integrating new sensors onto 
standard float platforms to measure six BGC ocean variables: 
Chlorophyll fluorescence, particle backscatter, dissolved oxygen, 
nitrate, pH, and irradiance, in addition to temperature and 
salinity. While Deep-Argo profiles require some increase in data 
management effort in terms of data processing and new quality 
control procedures, the introduction of BGC float data into the 
Argo data system has generated multiplicative challenges due to 
their complexity (Bittig et al., 2019). To minimize the impact of 
adding BGC data to the Argo data streams, the CTD and BGC 
data are stored in two separate profile data files: a Core-profile 
file, which contains the CTD data, and a BGC-profile file, which 
Qanteambear 2020 1 Valııme 7 | Article Z01