Nong et al. 
Year : 
1999) 
2000| 
2001 
2002 
2003 
2004 ı 
2005 
2006 
2007 
2008 
2009 
2010 
2011 
2012 
2013 
2014 
2015 
2016 
2017 
2018 
2019 
200 
400} 
X 
a 
5 
“ t 
800 
5 
1000l. 
12001 
3 qag0) 
1600) 
00 L 
2000 '- 
0 
40 60 80 
Cumulative percentage of profiles (%) 
ZU 
10u 
FIGURE 7 | Cumulative percentage of Argo profiles that reached a given 
Dressure (in 50-dbar intervals from 0 to 2,000 dbar) in each year from 1999 to 
2019, For example, the percentage of profiles that sampled to at least 1,600 
dbar increased from onlv 10% in 1999 to 80% in 2019 
early float models were only capable of profiling from shallower 
pressures (1,000, 1,200, or 1,500 dbar), particularly in the tropical 
oceans due to limited buoyancy generation and battery energy. 
As the buoyancy issue was resolved and floats became capable 
of profiling from as deep as 2,000 dbar, some floats adopted 
a scheme of only sampling from 2,000 dbar every 3rd or 4th 
profile (and from shallower pressures for other profiles) to save 
battery energy. Float lifetimes are ultimately limited by battery 
exhaustion, and many Argo floats were originally deployed 
with alkaline batteries. Over time, the use of lithium batteries 
became more common, and most floats now use them. Lithium 
batteries have more than twice the energy density of alkalines, 
hence considerably extending float lifetimes. Presently, with the 
transition from alkaline to lithium batteries, and the increased 
capacity of the buoyancy engines, modern Argo floats operate to 
pressures of 2,000 dbar on nearly every profile. Some new models 
of floats may have suflicient battery energy for 10 years of 10-day 
cycling to 2,000 dbar. 
In recent years, float models have been developed that are 
capable of 4,000 dbar and 6,000 dbar operations over a duration 
of several years (Roemmich et al., 2019). Improvements have also 
been made to the accuracy and stability of the CTDs used on 
these deep floats, so that they are suitable for sampling the small 
temperature and salinity signals of the abyssal oceans. In 2016, 
rontiers in Marine Science | www.frontiersin.or. 
Argo Data 1999-2019 
the first Argo profiles deeper than 2,000 dbar became available. 
These technological developments have enabled the formulation 
of the Deep-Argo program, which will extend Argo’s sampling 
pressure ranges to full-ocean depths except over the deepest 
abyssal plains and trenches. 
For most profiling floats, the CTD pump is normally 
switched off at around 5 dbar during ascent, in order to avoid 
contamination of the conductivity cell from the ingestion of 
material on the sea surface. Some floats now profile to within 2 
dbar of the sea surface, as the pressure sensors have become more 
accurate and thus the risk of surfacing with the CTD pump on 
is much less. Several float types also continue to sample up to 
the sea surface with the pump off, or carry auxiliary modules for 
high-resolution near-surface sampling. 
Geographical Coverage 
The geographical coverage of Argo has expanded over the past 
20 years largely due to the use of Iridium communications 
(Figure 3). ARGOS floats deployed in the equatorial region 
(5°S—5°N) tended to disperse poleward via Ekman transport 
while on the sea surface. Similarly, ARGOS floats in marginal 
seas or in near-coastal regions tend to have a higher probability of 
grounding because their prolonged surface times expose them to 
more cross-bathymetric wind transport. With their short surface 
times, Iridium floats are subjected to less Ekman divergence 
and wind transport, and therefore tend to disperse less in the 
equatorial and near-coastal regions. 
The use of Iridium has also enabled the geographical coverage 
of year-round Argo data to expand to the seasonal sea ice zones 
in the Southern Ocean (Wilson et al., 2019) and in the Arctic 
Ocean (Smith et al., 2019). Early attempts to sample the ice- 
covered polar oceans showed high instrument mortality rates, 
either because of crushing between ice floes at the sea surface or 
hitting the bottom of the ice packs during ascent. However, the 
inclusion of a robust ice avoidance algorithm in the float software 
has enabled floats to operate more successfully in the seasonal 
sea ice zones without additional hardware requirements. The ice 
avoidance algorithm (originally called the ice sensing algorithm, 
[SA) was first developed at the Alfred Wegener Institute (AWTI) 
for Polar and Marine Research, and was based on the assumption 
that the likelihood of the presence of sea ice was related to the 
temperature of the water column below (Klatt et al., 2007). In 
practice, the algorithm computes the median temperature of the 
near-surface mixed layer between depths Zi and Z, as the float 
ascends. If the median temperature is less than a prescribed 
threshold T,ef, the presence of sea ice is assumed, and the float will 
abort its ascent to the sea surface, store the profile data onboard, 
and descend to park pressure to begin its next cycle. For floats 
that use the Iridium satellite system for data communication, 
under-ice profiles that are collected and stored onboard the floats 
during winter are transmitted when surface conditions become 
ice-free during early summer (Riser et al., 2018). 
The ice avoidance algorithm was first implemented by AWI on 
floats in the Weddell Sea in 2002 (Klatt et al., 2007). The first AWI 
algorithm set Zı = 50 m, Z2 = 20 m, and T,ef = —1.79°C. In 2007, 
the University of Washington began deploying floats around 
the Antarctic continent with a version of the AWI algorithm 
Qanteambear 2020 1 Valııme 7 | Article Z01