This paper revisits Stommel's vision for a global glider network and the Argo design specification. A concept of floats with wings, so-called slow underwater gliders, is explored. An analysis of the energy or power consumption shows that, by operating gliders with half the vehicle volume at half the speed compared to present gliders, the energy requirements for long-duration missions can be met with available battery capacities. Simulation experiments of slow gliders are conducted using the horizontal current fields from an eddy-permitting ocean reanalysis product. By employing a semi-Lagrangian, streamwise navigation whereby the glider steers at right angles to ocean currents, we show that the concept is feasible. The simulated glider tracks demonstrate the potential for efficient coverage of key oceanographic features and variability.
In Stommel's (1989) vision for the year 2021, oceans would be monitored using instruments with wings. These robots would profile through the water column by changing their buoyancy in alternating vertical cycles of ascents and descents, with their wings providing the horizontal propulsion to “glide” through the oceans. It is now timely to look back and revisit this vision and assess its status.
Today, the oceans are indeed extensively monitored by buoyancy-driven robots
– albeit without wings. These instruments are called floats, and
approximately 4000 floats profile the oceans in the Argo programme (Roemmich
et al., 2009). Their contribution to the knowledge of the oceans is
substantial (Riser et al., 2016), yet juxtaposing the Argo programme and
Stommel's vision invites a curious investigation as to why the floats lack
wings. In that sense floats fall short of realizing his vision recognizing
that wings would also allow for dynamic positioning. The original Argo
design specification (Roemmich et al., 1999, p. 3) explicitly mentions the
possibility of a winged gliding float: a profiling float equipped with wings for dynamic
positioning during ascent and descent, offers further potential. This
`gliding' float will provide a similar number of
The stated potential has not yet been realized and further motivates the inquiry presented here.
Floats, without wings, are now a robust and mature technology developed
since the 1950s (Gould, 2005; Davis et al., 2001). Underwater gliders,
floats with wings (hereafter referred to as gliders), are a newer, more
complex development starting in the 1990s. The first successful glider
designs materialized in the early 2000s (Davis et al., 2002;
Jenkins et al., 2003) and have by now demonstrated their role as a reliable
and useful tool for oceanographic exploration of phenomena, e.g. boundary
currents where floats typically have short residence times (Rudnick, 2016;
Lee and Rudnick, 2018). As the glider technology matured, the enthusiasm has
gradually cooled (Rudnick, 2016). Gliders also fell short of realizing
Stommel's vision. For instance, Stommel envisioned a global glider effort,
compared to the current regional efforts; he envisioned endurance of years
compared to months; and he foresaw 1000 gliders compared to a few tens in
operation simultaneously. In one aspect current gliders do meet Stommel's
expectation: their horizontal velocity is indeed approximately
25 cm s
Stommel never seriously assessed the power requirements, suggesting instead that gliders could harvest energy from the ocean thermocline. This has proved less than practical and is also not a solution for the global ocean.
Current glider designs (Sherman et al., 2001; Eriksen et al., 2001; and Webb
et al., 2001) operate roughly according to the maxim “
Consider first the profiling vehicle (float or glider) of volume
However, due to ocean stratification additional pumping is necessary as the
vehicle rises to maintain the initial excess buoyancy
Equation (1) ignores the effect of hull thermal expansion since it is small compared to seawater thermal expansion (but not negligible depending on the hull material). An exact equation must include the full equation of state (EOS) of both seawater and the vehicle hull (but then becomes intractable). Furthermore, all terms and integrands must be weighted with the efficiency of the buoyancy engine of a particular vehicle.
We have also only stated the energy usage for the ascent part of the
dive–climb cycle, which is essentially the same for both floats and gliders.
Gliders generally operate in a symmetric mode in which the glider arrives at
the target profile depth with a negative buoyancy equal to that used for the
ascent (
In the following we will assume a pump efficiency as indicated in Fig. 1. This is similar to pump efficiencies reported by Davis et al. (1992) and Kobayashi et al. (2010).
Typical buoyancy engine efficiency (electric to
Further we will assume an aluminium hull with volumetric coefficient of
thermal expansion of
As an example, we calculate the energy consumed to ascend a tropical Atlantic profile from the World Ocean Atlas 2018 (Locarnini et al., 2018; Zweng et al., 2018) as shown in Fig. 2. Vehicle volume is set to 25 L.
Profiles of
Equation (1) approximates the energy consumption well but slightly overestimated compared to the full EOS formulation. The difference (compare blue and red lines in Fig. 2d) is primarily due to Eq. (1) neglecting the thermal expansion of the hull (which will assist the vehicle in reaching the surface).
The drag force acting on the vehicle may be expressed as (Khoury and
Gillett, 1999):
The power required, i.e. the product of force and velocity, is thus
For a winged vehicle, i.e. glider, lift is generated by the wings (Anderson, 2011; Thomas, 1999). It is known that wings are not efficient in flow with low speeds (low Reynolds numbers) (Schmitz, 1975; McMasters, 1974); however, Sunada et al. (2002) demonstrate that wings at low Reynolds numbers will perform adequately. At low speeds lift-to-drag ratios will be low (5–10) but sufficient for ocean profiling.
The generation of lift also causes so-called induced drag. In other words, the drag coefficient is also a function of the vehicle's angle relative to the direction of flow past the vehicle (the angle of attack). This effect is discussed in greater detail by Anderson (2011) and Thomas (1999) and is reasonably small here.
Categorized as flying vehicles, gliders (as discussed herein) operate in the regime of paper planes, small birds and large insects.
A hydrodynamic model is needed to calculate the vertical and horizontal
components of the vehicle velocity arising from the action of the drag and
lift forces. We define the hydrodynamic model in its abstract and implicit
form: Given expressions for vehicle drag and lift, and values for vehicle net
buoyancy and orientation (pitch angle), apply Newton's first law to solve
for the velocity and the angle of attack in conditions of steady planar
flight.
The vehicle will then glide through the water at an angle which is the sum
of pitch angle and angle of attack (
The angle of attack, however, deserves a comment in relation to lift and
drag. Lift is proportional to angle of attack until the vehicle stalls and
the production of lift reduces abruptly. In slow flight in particularly,
caution must be exercised not to exceed the stall angle of attack. Drag
resulting from the generation of lift, i.e. induced drag, is proportional to
In Fig. 3 we show the results from hydrodynamic models of two widely used gliders: the Seaglider (Eriksen et al., 2001; Frajka-Williams et al., 2011) and the Slocum glider (Webb et al., 2001; Merckelbach et al., 2010, 2019). The Seaglider has a relatively larger surface area and hence more drag. At higher velocities and buoyancies, however, the laminar flow profile of the Seaglider improves the performance relative to the Slocum glider. We also show the performance, in terms of velocity, of a hypothetical Slocum glider with half the volume and 20 % reduced drag. The reduction in volume is discussed in the next section. The 20 % lower drag is justified since Eq. (3) indicates a drag reduction of 37 % for a vehicle with half the volume. We deem 20 % drag reduction to be a conservative estimate and will account for induced drag and parasitic drag from appendages not represented in Eq. (3). The lift and size of the wings of this hypothetical glider are left unchanged, but it might be necessary to increase the size of the wings slightly, to compensate for the reduction in lift from a smaller hull (Merckelbach et al., 2010).
Performance as reflected in velocity (speed polars) of the Seaglider and the Slocum glider for three different net buoyancies of 25, 100 and 400 g. Also shown: a hypothetical Slocum glider with half volume and 20 % reduced drag. At low vertical speeds, the polars are cut off at an angle of attack of 5
Our discussion about the performance of a glider with smaller volume is
preliminary. A careful glider design should include simulations (Lidtke et
al., 2018), tank tests (Sherman et al., 2001), tank tests in combination
with simulations (Jagadeesh et al., 2009) and field tests where velocities
are measured (Eriksen et al., 2001; Merckelbach et al., 2019). We find the
extrapolation for the hypothetical glider toward a slower velocity and lower buoyancy to be safe and expect no significant Reynolds number effects,
neither on lift nor drag. For speeds of
We see in Fig. 3 that the modified Slocum glider with 80 % drag can
achieve the desired horizontal velocity of 13 cm s
The low excess buoyancy of 25 cm
All propulsive energy, pressure–volume work
Based on the definition of work we restate Eq. (1) with the drag expression,
Eq. (3), inserted and integrated over a linear path
The following considerations follow from this equation. The speed and
vehicle volume should be as small as possible. The factor
The distance
Equation (5) in itself indicates no optimum, and instead a viable low energy consumption must be sought. Considerations including so-called hotelling loads arising from the energy consumed by sensors may introduce optima (Graver, 2005, Sect. 7.2.1; Jenkins et al., 2003) but fall outside of the scope of this paper. For the vision presented here, power-hungry sensors must be avoided. It is doubtful whether a pumped conductivity and temperature system could be employed on a slow glider. Unpumped conductivity cells have been successfully used in gliders, and after appropriate corrections (Lueck and Picklo, 1990; Garau et al., 2011) they supply data of adequate quality. Such corrections will be challenging for a relatively slow flow past the sensor in a slow glider, but technically possible, provided an adequate sampling rate and flushing of the conductivity cell (Kim Martini, personal communication, 2019). Further calibrations and bias removal will also be possible against Argo floats and ship-based measurements. The user must carefully assess the accuracy needed for salinity against a trade-off from endurance.
A net buoyancy change of 50 cm Based on the current specification of
the Electrochem 3B0036 DD lithium primary cell:
Present gliders indeed look compact and crammed on the inside. Yet Eq. (5) clearly shows that volume drives energy consumption. As energy considerations are of prime importance, vehicle volume must come down. This would be achievable if the glider was designed with this consideration in mind from the start. This direction of development is necessary on the grounds of basic energy considerations. An example of a low-volume vehicle is the SOLO-II float, which has a volume of approximately 18 L – in its previous technological iteration, the SOLO-I float, it had a volume of 30 L (Owens et al., 2012). Reduction in volume seems possible. If glider volume could only be reduced to 30 L rather than 25 L, Eq. (5), being almost linear in volume, shows that volume and energy consumption would both increase by roughly 20 %.
The 2000 m hull of the vehicle must satisfy three requirements. It must be strong enough to withstand pressure, yet the compressibility should match that of seawater, and finally, it should offer the necessary payload volume for batteries, electronics and the buoyancy engine. This poses a real engineering challenge. Jenkins et al. (2003, Sect. 6.3) contains detailed considerations for an aluminium hull. However, it is likely that alternative composite materials must be considered for the hull (Osse and Eriksen, 2007; Webb, 2006).
In summary, we suggest that a slow glider (or float with wings) is feasible if the volume and speed are halved relative to present gliders.
As an example of a complete power budget we use a low-power and slow
Seaglider dive. The dive was conducted in the Iceland Sea by Seaglider sg564
on 5 November 2015 (dive number 227). The vehicle was diving with a buoyancy
of
Energy or power breakdown for low-power Seaglider dive to
1000 m. Dive buoyancy was only
The controller (processor) is the most power-hungry main component, with 37 % of the total energy expenditure (Table 1). This, however, is not because of complex control but rather due to the fact that the processor of the glider is severely outdated. The controller of both Seagliders and Slocums is based on a processor design from the 1980s (the Motorola 68000-series) in a 1990s package (the Persistor). Based on a conservative application of Moore's law, we estimate that the power consumption could be reduced by a factor of 4 for a modern processor.
Only 6 % of the total energy was expended on the conductivity, temperature and depth (CTD) sensor – a figure
that should arguably be increased in order to apply appropriate corrections
for free-flush conductivity cells. We would like to allocate savings from a
new controller to increasing the number of CTD samples possibly including an
In this paper, we are mainly concerned with the energy expended by the
buoyancy engine (Eqs. 1 and 5). Nevertheless, we allow for an additional 1 kJ 2000 m
Power budgets will be related to the vehicle volume as the displacement must
make up for the weight of batteries. If we allocate
As a basis for estimating the mission cost we use the current costs for a
core Argo float mission. The cost for the float itself is about USD 20 000
which approximately doubles when programme management costs are included
(Argo, 2019b). Basing the cost estimate on Argo float costs can be justified
for two reasons. The economy of scale for
In Table 2 we include the additional costs for various glider-specific items. A glider is inherently a more complex instrument than just a float with wings plus other components, and we also allow for costs associated with the increase in complexity of integrating the additional parts. Furthermore, we include a healthy profit of 50 % and development costs. While the relative distribution of profit, component costs and operation costs can be different, the overall cost estimate is deemed representative.
Cost estimate for a slow glider mission based on Argo float costs and Argo programme costs.
The simple budget in Table 2 indicates that a slow glider (winged float) mission would cost about 3 times as much as an Argo float mission (USD 40 000). This may or may not be deemed prohibitive depending on scientific potential and value of such an endeavour.
What missions would be possible with a glider travelling at only
13 cm s
In an environment where ocean current velocities typically exceed vehicle velocity, the navigation strategy must be adjusted, or else the vehicle is simply too slow for the normal navigational notions to be feasible. Specifically, navigation with traditional latitude and longitude waypoints along straight lines must be given up. Instead, we propose to navigate in Lagrangian streamwise coordinates.
The Lagrangian streamwise navigation is achieved when the glider steers at right angles to ocean currents and never attempts to compensate for these currents. Thus, it will be able to step into or out of any coherent current structure – be it an eddy, a front or a boundary current. Trajectories will be spirals and oblique winding lines and not linear transects along bathymetric gradients. The glider will operate in a semi-Lagrangian and semi-Eulerian mode. This, however, represents a significant upgrade to Lagrangian only floats.
We call the proposed method of navigation “Eulerian roaming”, where
Eulerian refers to the streamline traversing capability and roaming to the
Lagrangian drift. Colloquially one might be tempted to summarize the
Eulerian roaming with two common sayings or proverbs: “only dead fish
follow the flow” and “never oppose a stronger force – out-manoeuvre”.
Davis et al. (2009) summarize it as follows: in a strong adverse current, steer rapidly across the current while making up ground where the currents are weak or favorable.
Stommel (1989) notes the following: “Having to decide what heading to choose stimulated modellers and descriptive oceanographers to exercise their minds and their computers.” We will attempt to do so in the following.
In the simulation the gliders will attempt to navigate the reanalysed ocean
of Mercator GLORYS12 provided by the Copernicus Marine Environment
Monitoring Service (CMEMS). This reanalysis is based on the real-time global
ocean forecasting of CMEMS which is detailed by Lellouche et al. (2018). The
reanalysis is eddy permitting with a horizontal resolution of
A fourth-order Runge–Kutta method with adaptive time steps (RK45) is used to
integrate glider and ocean velocities to estimate the glider's position.
The maximum time step is 600 s, but this is reduced to 60 s at the surface or
near the bottom and otherwise adjusted automatically. The glider's
horizontal velocity is fixed at 13 cm s
The velocity fields from the reanalysis product are linearly interpolated in space and time. The glider is advected in a Euclidian flat-earth coordinate system but re-projected per dive or if glider displacement exceeds 25 km. We observe no artefacts arising from the numerical scheme, linear interpolation or spatial reference. The coarse bathymetry of the model with only 50 levels (steps increasing with depth) aggravates plunges steeper than the glider trajectory. When climbing bathymetry, the glider would occasionally fly into these plunges and get stuck, and in such cases the glider was jerked up 5 m at a time until the glider was clear of the bathymetry. This is not an issue for real gliders equipped with altimeters.
The drift at the surface, for about 5 to 10 min while communicating in between dives, is ignored. The results, however, are not sensitive to this.
The glider is steered according to the principles set out in Sect. 3.1. To
express a recipe for the Eulerian roaming navigation we formulate the
following pseudo-code or set of rules:
Traverse the ocean at If depth-average current is not available, steer along or opposite to the gradient of the local bathymetry. If neither (A) nor (B), steer to the nearest current feature as indicated by satellite altimetry (or in future, as appearing in operational ocean nowcasts and forecasts). If none of the above provide an informed heading, use an opportunistic heading deemed suitable for the mission in general.
The ordering of rules is not coincidental. They provide a hierarchy from the
simplest autonomous modes (A and B The glider could have a
bathymetric map installed to autonomously calculate the topographic gradient.
We suppose that an up-to-date and accurate map of sea surface heights (SSHs) is available, and to mimic this we use the SSH of the reanalysis as an input for the mode (C) above. As will be discussed in Sect. 4.5, we find this a reasonable assumption for the near future.
Occasionally the simulated glider visited ice-covered waters (eastern coast of Greenland), and we will here assume that the following under-ice navigation can be executed: head west under ice until the 500 m isobath, then turn back (without surfacing). This can be interpreted as an under-ice version of rule (B) above. Gliders today are equipped with ice-avoidance algorithms (Renfrew et al., 2019), which make similar scenarios applicable.
To test the slow glider, we first simulate a mission in the Nordic Seas where we attempt to visit known features and currents. The Nordic Seas are bounded by Norway, Greenland, the Greenland–Scotland Ridge in the south, and Fram Strait in the north.
Slow glider mission in the Nordic Seas. The glider mission starts
at the south-eastern corner, off Norway at 62.8
The mission, Fig. 4, starts off the west cape of Norway in the south at the
500 m isobath (62.8
The mission executed can be summarized as follows: visit the main features of the Nordic Seas (excluding the shallow Barents Sea). Due to the relatively modest currents encountered we find that we may “ferry” the glider around according to rule (D) (Sect. 3.3) in the central parts of the basins. Near boundaries we used rules (A) and (B), which often resulted in the same heading.
The glider performs 691 cycles. The energy consumption, using the technique
and values described in Sect. 2, is 2.9 MJ (or 2.3 kg of lithium primary
batteries). This is calculated by evaluating Eq. (1) using the established
operating point with an excess buoyancy of 25 cm
In order to test the slow glider in a more challenging, energetic environment, we visit the Gulf Stream.
This mission, Fig. 5, starts at the coast of Florida and Georgia (again at the
500 m isobath; 30
Slow glider in the Gulf Stream off the east coast of the USA. The glider starts at the coast of Florida at the 500 m isobath. Bathymetric contours drawn at 500 m intervals to 3000 m. The temperature at 200 m is also shown to represent the water mass distribution.
The Eulerian roaming through this energetic environment is realistic and has been successfully performed previously. Using Spray gliders, Todd et al. (2016) collected transects across the Loop Current in the Gulf of Mexico
and across the Gulf Stream between 35 and 41
Since our hypothetical glider ended up off Newfoundland, it was natural to continue the mission into the northern branch of the North Atlantic Current (NAC), and the continuation of the mission is shown in Fig. 6.
At 55
Slow glider mission continued into the North Atlantic Current. The glider ends at Iceland in the north-eastern corner of the map. Bathymetric contours are at 500 m intervals. The temperature at 200 m is also shown.
The Drake Passage between the South American and the Antarctic continents probably represents the world's most interesting choke point (or area) as the Antarctic Circumpolar Current (ACC) must pass through it, and we simulate a mission here as well.
The glider is launched off the tip of South America (67.8
Slow glider trajectory in the Drake Passage, launched off the tip of South America (NW corner). Bathymetric contours are at 500, 1500 and 3000 m. Mission ends at South Georgia Island.
After being advected out of the Drake Passage, the glider is capable of staying in the Scotia Sea to the east, where it executes a distorted butterfly before recovery at South Georgia Island after 638 cycles. Energy consumption is estimated at 3.1 MJ (2.5 kg of lithium primary batteries).
In the Nordic Seas, the slow roaming glider or winged float would significantly complement the Argo float array in the area. The slow glider is able to sample fronts, eddies and boundary currents as well as basin interiors, whereas Argo floats tend to be constrained within the 2000 m isobath of the basin where they were launched (Voet et al., 2010).
The mission exemplified in the Nordic Seas targets the observation of the circulation and water mass properties at key locations in the Nordic Seas. This region is a key component of the Atlantic Meridional Overturning Circulation (AMOC), in which warm waters flow northward near the surface and cold waters return equatorward at depth. The variability of the Atlantic water characteristics is of importance to the climate in western Europe, to weather and sea ice conditions, and to primary production and fish habitats. The Nordic Seas are an important area for water mass transformation (Mauritzen, 1996; Isachsen et al., 2007). The newly produced or transformed dense waters return southward between Iceland and Greenland through Denmark Strait and east of Iceland across the Greenland–Scotland Ridge, contributing to the lower limb of the AMOC. Transects worked by a slow glider will provide crucial observations in the Norwegian Atlantic Current at Svinøy (Høydalsvik et al., 2013), in the deep convection regions in the Greenland and Iceland seas (Brakstad et al., 2019; Våge et al., 2018), and in the Lofoten Basin, which is a hotspot for Atlantic water transformation (Bosse et al., 2018). The transect in Fram Strait will capture the properties and variability in the return Atlantic water along the Polar Front in the northern Nordic Seas (de Steur et al., 2014). Particularly the interior Greenland and Iceland seas and the east Greenland shelf are under-sampled, and the observations will be useful in understanding the role of wintertime open-ocean convection in the western basins of the Nordic Seas and the effect of an ice edge in retreat toward Greenland (Moore et al., 2015; Våge et al., 2018).
Similarly, slow glider observations from the Gulf Stream and the Drake Passage mission examples will advance characterization of mean pathways, mesoscale variability and energetics in climatologically important regions. Furthermore, the Eulerian roaming will allow sampling of snapshots of mesoscale eddies. In the Lofoten Basin, a similar navigation option was used to spiral in and out of the Lofoten Basin Eddy by instructing the glider to fly at a set angle from the measured depth-averaged current (Yu et al., 2017). Profiles collected from such missions will be useful in characterizing the coherent eddy structures, filaments along fronts and around mesoscale eddies (see Testor et al., 2019, and the references therein). An additional strength of glider observations is the ability to infer absolute geostrophic currents (e.g. Høydalsvik et al., 2013). The transects resulting from Eulerian roaming are different than and less regular compared to the sections occupied by ship-based surveys, typically normal to the isobath orientation. Strong current speed exceeding the speed of gliders will result in oblique sampling. However, the local streamwise coordinate system (Todd et al., 2016), applied to for instance the Gulf Stream and the Loop Current, is demonstrated to be a powerful approach to calculate volume transport rates and potential vorticity structures and to provide insight into processes governing flow instabilities.
The key assumption in using the local streamwise coordinate system for geostrophic current calculations along the glider trajectory is that flow is parallel to the depth-averaged current (DAC). When the depth-average current direction is not perpendicular to the transect segment of the glider path, a decomposition into cross-track and along-track components must be made. In these conditions, using the currents from the local streamwise coordinate system will be an error; however, the transport will remain relatively unaffected. In a recent study, Bosse and Fer (2019) reported geostrophic velocities associated with the Norwegian Atlantic Front Current along the Mohn Ridge, using Seaglider data, following Todd et al. (2016) and assuming DAC is aligned with the baroclinic surface jet. They also calculated the geostrophic velocities and transports using the traditional method, i.e. across a glider track line, and found that the peak velocities of the frontal jet were 10 %–20 % smaller but the volume transports were identical to within error estimates. The Eulerian roaming can thus be used to obtain representative volume transport estimates of relatively well-defined currents. We also note that the present 1000 m depth capability of gliders limits our ability to compose the geostrophic currents into barotropic and baroclinic components in water depths substantially deeper than 1000 m. A 2000 m range will allow us to more reliably approximate barotropic currents as the depth-averaged profile, up to depths of around 2000 m.
Regarding the navigation in general we note that it was easier and required less skill and intelligence in strong currents, when only a choice between left or right could be made (rule A); this often coincided with bathymetry (rule B). In weak currents, the navigational options increased, and we resorted to skilful and intelligent use of rules (C) and (D). The intelligence, however, does not seem to be very advanced as it essentially is an image processing task on the SSH image (albeit a vector image) for local steering decisions. Global steering decisions such as the general area to visit will require some oceanographic intelligence and are probably not suitable for automation.
The control and steering of a network or fleet of slow gliders should aim to optimize for some scientific objective, possibly in conjunction with other sensing platforms. Alvarez et al. (2007) have looked at synergies between floats and gliders to improve the reconstruction of the temperature field. Synergies also exists between a glider fleet and altimetry to map geostrophic currents (Alvarez et al., 2013). We suggest that future work should see the slow glider concept as part of a heterogenous suite of ocean sensing technologies. The topology of the network needs some consideration and one interesting option is to cluster the gliders in and near an oceanographic feature to explore it in greater detail. Some in situ experiments with glider fleets have been conducted (e.g. Leonard et al., 2010; Lermusiaux et al., 2017a). The problem of planning optimal paths for gliders is reviewed by Lermusiaux et al. (2017b).
While we have shown tracks of individual gliders, it should be clear that the impact of a slow roaming glider concept will increase when employed in large numbers. Also, the simulations here where a few gliders are hand-piloted does not show the full potential of the approach. Future simulations can include a large number of gliders to train artificial intelligence to perform the piloting.
The upcoming Surface Water and Ocean Topography (SWOT) altimetry mission (Fu and Ubelmann, 2014) will yield an unprecedented view of the ocean surface. Within the swath of the altimeter, approximately 120 km wide, we will see snapshots of oceanic mesoscale and sub-mesoscale structure and variability, albeit at a slower repeat cycle of 21 d. However, advances in processing (Ubelmann et al., 2015) will likely fill the temporal gaps in a dynamically meaningful way, leading to maps of SSH with high temporal resolution, and enable operational model capabilities and applications hereto unimagined (Bonaduce et al., 2018).
There was always a strong coupling among altimetry, models and observations of ocean interior (Le Traon, 2013). The Argo programme's name was chosen because of its affinity with the then upcoming altimetric mission of the Jason satellites. Argo was the ship of Jason and the argonauts in their quest for knowledge. As altimetry advances, it is necessary to ask whether Argo and our quest for knowledge should advance in parallel. While SWOT altimetry will yield a (sub-)mesoscale view of the ocean, Argo remains primarily a basin- and seasonal-scale technology. We propose that the slow glider concept, essentially the gliding float that the Argo design specification calls for, could add a mesoscale component to Argo. This natural development enhances the Argo as a component of the global ocean observation system and supplements the regular glider operations, which are at present regional and process-oriented (Testor et al., 2009, 2019; Liblik et al., 2016).
Since we propose to steer the glider using maps of SSH and model output, the proposed slow glider would also provide an even tighter integration of altimetry, models and observations of ocean interior.
Other developments in the Argo programme further suggest a progression in this direction. Floats are increasingly being equipped with more advanced sensor suites in the biogeochemical programme (Riser et al., 2016; Roemmich et al., 2019). A new ArgoMix component with turbulence sensors (thermistors and airfoil shear probes), is also under consideration to map the spatial and temporal patterns of ocean mixing. The capabilities of the sensors call for a more advanced vehicle navigating the mesoscale ocean as well.
In the example missions presented here the glider is launched and recovered near the coast. Logistical challenges aside, this opens up new participatory dimensions with coastal communities. It might also be judged as a more environment friendly alternative to the Argo floats, which are submitted to the ocean upon mission completion.
While the tracks of the slow glider (or winged float) presented in this
section clearly demonstrate oceanographic
We show that oceanographically useful and sensible trajectories are possible with a slow roaming glider. Looking back at the quote from the Argo design specification in the Introduction, one might say that the expectations for a gliding float were too high. The notion of “a fixed location or along a programmed track” is not feasible due to energy constraints limiting velocity, nor is the notion indispensable or necessary. Even though we here mostly explore the concept of Eulerian roaming navigation, the slower and smaller glider will be able to maintain station (virtual mooring) or follow well-defined section lines at sites where currents are weak.
The velocity of 25 cm s
Future work should firstly attempt to verify the concepts and findings presented here using existing gliders in the real ocean. The gliders should be operated at a lower speed than usual (refer to Fig. 3) and navigated as outlined in Sect. 3.1 and 3.3. Future work should also include observing system simulation experiments (e.g. L'Hévéder et al., 2013; Chapman and Sallée, 2017), whereby data assimilation from fleets of slow gliders demonstrates benefit and increased model skill in operational models. The piloting should also be automated, and work might be directed at developing artificial intelligence doing day-to-day piloting.
This paper demonstrates that slow gliders or Argo floats with wings are desirable and potentially feasible – the slow glide is on.
The data used in this paper are made freely available by CMEMS (CMEMS, 2018) and in the World Ocean Atlas 2018 (
EMB and IF wrote the paper with revisions and suggestions from KV and PMH. EMB performed the simulation experiments.
Ilker Fer is a member of the editorial board of
The glider team at the Geophysical Institute spurred many interesting
discussions about gliders and glider technology. The glider activities here were initially financed by the Research Council of Norway under grant no. 197316. We thank Lucas Merckelbach and Daniel Hayes, for their constructive review and criticism of our paper. Interested readers
should consult their detailed remarks and consider them supplemental in the
spirit of
Ketil Våge was supported by the Trond Mohn Foundation (grant no. BFS2016REK01).
This paper was edited by Oliver Zielinski and reviewed by Lucas Merckelbach and Daniel Hayes.