Field campaigns are instrumental in providing ground truth for understanding and modeling global ocean biogeochemical budgets. A survey however can only inspect a fraction of the global oceans, typically a region hundreds of kilometers wide for a temporal window of the order of (at most) several weeks. This spatiotemporal domain is also the one in which the mesoscale activity induces through horizontal stirring a strong variability in the biogeochemical tracers, with ephemeral, local contrasts which can easily mask the regional and seasonal gradients. Therefore, whenever local in situ measures are used to infer larger-scale budgets, one faces the challenge of identifying the mesoscale structuring effect, if not simply to filter it out. In the case of the KEOPS2 investigation of biogeochemical responses to natural iron fertilization, this problem was tackled by designing an adaptive sampling strategy based on regionally optimized multisatellite products analyzed in real time by specifically designed Lagrangian diagnostics. This strategy identified the different mesoscale and stirring structures present in the region and tracked the dynamical frontiers among them. It also enabled back trajectories for the ship-sampled stations to be estimated, providing important insights into the timing and pathways of iron supply, which were explored further using a model based on first-order iron removal. This context was essential for the interpretation of the field results. The mesoscale circulation-based strategy was also validated post-cruise by comparing the Lagrangian maps derived from satellites with the patterns of more than one hundred drifters, including some adaptively released during KEOPS2 and a subsequent research voyage. The KEOPS2 strategy was adapted to the specific biogeochemical characteristics of the region, but its principles are general and will be useful for future in situ biogeochemical surveys.
The role of iron as a key limiting micronutrient for large phytoplankton in
high nutrient–low chlorophyll (HNLC) waters was brought to prominence by
Martin (1990) and motivated a series of bottle incubation experiments.
Difficulties in unambiguously interpreting the results of these experiments
led to the design and the implementation of more ambitious field studies.
They were conducted in regions fertilized artificially or naturally with
iron. One of the most striking difference between these types of study is the
role played by the horizontal transport of iron. Most of the artificial iron
fertilization experiments were conducted using specific strategies aimed at
minimizing horizontal effects. Some artificial experiments have targeted
quasi-isolated eddy cores, so that the water patches which are trapped by the
mesoscale circulation are only marginally mixed with the environmental waters
on the time scale of the induced bloom. This was the case for the EIFEX,
LOHAFEX and SAGE experiments
(
However, these quasi-isolated or homogeneous regions are not typical or
representative of the vast majority of the ocean. In general, water parcels
are stirred by the mesoscale field in a non-local way, experiencing the
cumulative effects of several mesoscale structures from their iron enrichment
event to the moment of the phytoplanktonic bloom. Horizontal transport does
not only modulate the extension of a fertilized region with respect to its
sources (e.g.,
This paper first presents our efforts to use the prior and real-time satellite information to understand bloom dynamics, and thereby define an optimal sampling strategy. In doing so, we examine both Eulerian eddy fields and Lagrangian maps of water mass origins. The paper then validates these initial perspectives on the circulation against drifters released during the KEOPS2 project and a subsequent voyage. It provides an iron supply and removal budget based on the integration of satellite, ship-based and drifter information. It illustrates how the context provided by these circulation perspectives informed the interpretation of the observed ecological states (with reference to other works in this volume). Finally, it offers perspectives for biogeochemical process study planning.
Altimeter and ocean color products used in this study were specifically
produced for the Kerguelen region by Ssalto/Duacs and CLS with support from
CNES from mid-May 2011 to July 2012. Altimetry maps were generated from
Jason-1, Jason-2 and Envisat along-track observations with procedures
analogous to those used for AVISO data
In particular, a high resolution (
Maps of absolute dynamic topography at a resolution of
Geostrophic currents were computed from the maps of absolute dynamic
topography by solving the equation for geostrophic equilibrium
Composite maps of surface chlorophyll concentration with a resolution of
Maps of geostrophic and total velocities have been used to compute the
Okubo–Weiss parameter, OW (
In this study, mesoscale eddies have also been identified from satellite
velocities using the vector geometry-based eddy detection algorithm developed
by
Several approaches were used to identify the effect of
the eddy field on iron redistribution. Altimetry-based finite-size Lyapunov
exponents (FSLE,
Trajectories were derived by applying a Runge–Kutta fourth-order scheme with
a time step of 6 h. Velocity fields have been linearly interpolated in
both space and time. Initial and final separation distances were set to
0.05
In practice, the Lyapunov exponent computed backward in time provides the
exponential rate at which horizontal stirring has brought water parcels to a close distance
(
The same Runge–Kutta scheme used to compute the backward FSLE was applied to
retrieve additional diagnostics fundamental to the real-time analysis.
Firstly, the ability of mesoscale eddy cores to retain water parcels was
quantified by backward advection of particles released within regions with
negative OW values, and their duration within these regions was measured. In
effect this yields an estimate of how long ago a water parcel entered a given
eddy, yielding a retention value expressed in days. Similarly, backward
advection was used to identify water parcels originally above the Kerguelen
Plateau, defined as the region where ocean depths are shallower than 700 m.
This was achieved by finding the trajectories of virtual drifters which had
touched the 700 m isobath in the past, in analogy to the method developed
for the Crozet experiment (
More than 200 WOCE-SVP drifters were used to validate the
altimetry-based estimation of stirring patterns. 48 drifters have been
deployed according to an adaptive strategy during the KEOPS2 campaign and
24 during the MYCTO-3D campaign that took place in the Kerguelen region in
January 2014. Furthermore, we collected 120 trajectories measured between
2011 and 2014 in the same region from the Global Drifter Program (GDP –
The circulation of the Kerguelen region (see Fig.
Bathymetry and a sketch of the main current branches. The thick 500 m isobath identifies the Kerguelen and Heard plateaux. The 750 m isobath marks the shelf break.
Sea surface height and eddy contours (11 November 2011).
Some improvement in the outlook for understanding and predicting the SCHL
pattern emerges when altimetry is used to reconstruct the 30 day longitudinal
and latitudinal origin of surface waters (Fig.
Lagrangian and Eulerian diagnostics derived from satellite altimetry: (top left) total kinetic energy; (top right) Okubo–Weiss parameter; (bottom left) Lyapunov exponents (finite-size); (bottom right) retention parameter. These maps were generated in near-real time for guiding the adaptive sampling strategy of the KEOPS2 cruise.
Satellite chlorophyll maps:
Altimetry-derived 30 day horizontal origin. Left: longitude. Right: latitude.
Altimetry-derived estimation of the iron pathways from the plateau to the
open ocean:
These Lagrangian images provide the putative extension of the plume,
including the positions of fronts which create contrasts in terms of origin
from the shelf break and of time since last contact with the plateau. The
temporal information is especially relevant for biogeochemical
interpretations, as at a first approximation, the dissolved iron concentration
can be modeled as decaying exponentially in time due to scavenging
These maps were computed starting from winter 2011 and were then updated daily using near-real-time altimetry. When the ship headed to the region for measuring pre-bloom conditions, the altimetry-derived calculations were used to forecast the extension of the plume and refine the position of the stations.
Validation of Lagrangian diagnostics derived from altimetry with the same
quantity computed from SVP drifter trajectories.
The altimetry-based estimation of the plume was validated during the cruise with ocean color images of the bloom, and post-cruise by analyzing the trajectories of the drifters released during the KEOPS2 and a following cruise 2 years later (MYCTO-3D-MAP), as well as from historical trajectories of Lagrangian drifters that had crossed the region.
The first cloud-free chlorophyll image
(Fig.
Age from the plateau averaged in latitude along a band 48
The table provides the extent and the degree of overlap (congruence)
for the altimetry-based forecast – defined through the “water age”
diagnostic – and the plume visible from remote-sensed chlorophyll (Chl.). The
boundaries of the plume are defined, setting threshold values. The thresholds
for the forecasted plumes were chosen to match (within 10 %) the
extension of the chlorophyll plume with typical moderate and high values
(0.5 and 1
Table 1 provides the extension and degree of superposition (congruence) for
the altimetry-based forecast and the SCHL plume as seen in ocean color data.
In order to define the boundary of the SCHL plume, we used a threshold value
of 0.5
In order to validate the altimetry-based estimation of stirring patterns, and in particular the origin from, and time since leaving the apparently iron-rich Kerguelen shelf break, 48 SVP Lagrangian drifters were released during the KEOPS2 cruise. An ideal plan would have released the drifters on a regularly spaced linear array which would have followed the shelf break. As this would have consumed too much ship time, drifters have been released instead on transit from one station to another – when approaching the shelf break – or adaptively, when crossing key dynamical features like fronts. In order to palliate to this sub-optimal release scheme, a second opportunistic release experiment was performed by the MYCTO cruise during January–February 2014. This second experiment targeted particular regions which remained undersampled by the KEOPS2 scheme. Other 2012–2014 SVP historical drifter trajectories were also included in the analysis.
The time since having left the plateau and the latitude of departure from the
shelf break also were estimated for each SVP drifter trajectory in analogy
with the altimetry-based calculation. Results and comparison are shown in
Fig.
A great advantage stemming from the ability of the multisatellite maps to reveal the presence and locations of regions with contrasting origins and histories is that the small-scale variability in the age since leaving the plateau can be exploited to examine the large-scale contrasts in physical–biological coupling.
Based on the Lagrangian analysis, the chlorophyll plume can be zoned in terms of expected biogeochemical contrasts: a recirculation region; a jet, on the north flank of the polar front; a cold water tongue propagating northward along the eastern shelf break and unfertilized, HNLC waters (the “reference” case).
For KEOPS2, this perspective was developed in real time. The initial station
sampling – consisting of a north–south transect to capture latitudinal
variations, with an east–west transect to examine variations with distance
from the plateau – was modified opportunistically to examine the development
of a bloom within the “fast-lane” jet to the north of the polar front, and
to cover features evolving in a quasi-Langrangian time series within the
recirculation feature (
Assuming that the iron sources at the shelf break are homogenized by local
mixing process and that iron dynamics can be modeled by first-order removal,
the “age” field can be converted into an iron field (see Appendix). The model
has only one free parameter, the scavenging constant
We first estimated the removal constant for abiotic conditions
(
We then determined the removal constant
Having estimated
Biogeochemical field studies are becoming more and more interdisciplinary, so that an increasingly large number of parameters have to be collected and only a very limited number of stations can be occupied during the same day. Nowadays the big challenge of a biogeochemical campaign is how to address this trade-off between the biogeochemical analytical resolution and the spatiotemporal coverage with a limited number of multidisciplinary stations. In-cruise knowledge of the biogeochemical provinces present in a region is therefore essential information to avoid wasted efforts, for instance sampling the same conditions multiple times. In this regard, remote sensing is an unavoidable tool, because it is the only observation capable of snapshots of regional variability.
The time scales characteristic of a bloom (days to weeks) are also the ones of the (sub)mesoscale and in particular of horizontal stirring. This coupling can create very complex biogeochemical contrasts which evolve in time during the campaign itself. One obvious source of information for mesoscale transport is satellite altimetry. Our study shows that these data can be mapped into biogeochemical regions by dedicated diagnostics. The model we developed for the KEOPS2 cruise may be considered as an attempt to translate altimetric sea surface height (SSH) patterns into patterns of primary production, maximizing the spatiotemporal information. Indeed, the model provides a pre-condition to the bloom, but does not inform on the intensity (or it does in a qualitative way only), nor on the timing. It is important to notice that this ability to estimate the plume extension at high precision by such a simple model is that the biogeochemical drivers that control the onset of the bloom are relatively simple: high nutrient waters with only one (spring-time) limiting micronutrient; a fixed source of the limiting micronutrient constrained by the topography (the Kerguelen–Heard Plateau), possibly homogenized along the shelf by a boundary current, and hence stable in time and space; deep winter convection all over the open-ocean region, which dissipates possible nutrient vertical inhomogeneity in the upper layer of the water column. Moreover, the circulation in the region is well captured by altimetry, because transport is dominated there by a barotropic current with a strong geostrophic signal (the ACC). These conditions make the Kerguelen region an ideal large-scale laboratory for studying fertilization events occurring in the Southern Ocean HNLC systems.
It is tempting to
extend the altimetry-based model into the vertical for predicting the phytoplanktonic bloom more
quantitatively. Indeed, without explicitly accounting for vertical dynamics, one may be surprised by the possibility of
studying nutrient dynamics at all. However, in the Southern Ocean,
geostrophic horizontal velocities are typically strongly correlated with the
flow within the thick surface layer (of the order of 100 m, see
In terms of future biogeochemical campaigns, it may be instructive to ask which aspects were the most informative, which ones redundant and which ones should be improved. The most important information for planning the position of the stations was the possibility of sampling waters with contrasted iron concentrations. What would have happened to guiding the adaptive sampling strategy we used without the Lagrangian model? Considering (i) that iron is scavenged during its advection, (ii) that the main source of iron is the Kerguelen Plateau and (iii) that the mean circulation is eastward, it would have probably been tempting to choose the eastward distance from the plateau as a first-order proxy of iron scavenging. However, this choice would have been quite misleading due to the presence of mesoscale activity, which creates meandering “high-speed lanes” and recirculation features so that the iron-to-distance relation may become quite distorted locally. In particular, we showed that in the few hundreds of kilometers east of Kerguelen (which was the region in the range of the ship), the recirculation region discussed above creates a reservoir of “old” – hence iron-depleted – water in the vicinity of the source, and in turn a locally inverted iron-to-distance relation (i.e., iron locally increasing with distance, and not decreasing).
Although the Lagrangian diagnostic discussed here is specific to the Kerguelen situation, the possibility of analyzing multisatellite data with diagnostics based on mechanistic models is a general approach which looks to be an appealing companion of multiplatform campaigns. In this study we have performed part of the validation of the diagnostics after the cruise, because the drifters that we used were released during the campaign. In the case of a multiplatform campaign, a better and safer strategy would be to deploy some of the instruments (like Lagrangian drifters and profilers) before the cruise, in order to have better information on the physical drivers active in the region and to integrate in situ information with remote sensing. The possibility of releasing some instruments before the campaign would also avoid the scenario in which a diagnostic is first used for guiding a campaign, and then discovered to perform poorly.
Biogeochemical campaigns are under heavy pressure to resolve biogeochemical processes at higher precision and to include an increasingly larger number of coupling mechanisms. This need is changing our conceptual view of a biogeochemical process from an approximated zero-dimensional or one-dimensional water column to a spatial extended system with full three-dimensional physical drivers on one side, and complex top-down ecological controls on the other one. This paradigmatic change is blurring the boundaries between biogeochemistry, physics and ecology, pushing towards end-to-end studies. In this perspective, the use of Lagrangian tools for dynamically zoning a region, mapping the transport structures present in a region in near-real time will become even more critical for disentangling the temporal from the spatial biogeochemical variability, and for optimally choosing sampling stations in terms of their representativeness.
Along a trajectory, the iron content of a water parcel is not affected by
transport, because altimetry-derived velocities are almost divergence-free.
For a water parcel with coordinate
From this equation we see that the spatiotemporal variability of the field
We start by assuming that the iron concentration in the upper layer (0–150 m),
considered as the winter mixed layer in the plume over the season, is
entirely controlled by the off-plateau iron flux
We estimated the horizontal iron supply (denoted
Using the altimetric mean velocities and these
The idealized model Eq. (
the solution at equilibrium is
The amount of iron that is exported (
If
Thus the vertically exported flux during
The authors would like to thank the Marion Dufresne crew and B. Queguiner. The altimeter and color/temperature products for the Kerguelen area were produced by Ssalto/Duacs and CLS with support from CNES. The authors would also like to acknowledge AVISO/CLS, Météo-France and the Global Drifter Program/NOAA/AOML, Miami, Florida, and both the Drifter Operations Center and Data Assembly Center for arranging drifter deployments and data assembly, quality control and distribution of the data. A. Della Penna was supported by a conjoint Frontières du Vivant (Paris 7) and CSIRO-UTAS Quantitative Marine Science PhD scholarship. This work was supported by the French Research program of INSU-CNRS LEFE-CYBER (Les Envéloppes Fluides et l'Environnement–CYcles Biogéochimiques, Environnement et Ressources), the French ANR (Agence Nationale de la Recherche, SIMI-6 program, ANR-10-BLAN-0614 and ANR MYCTO-3D-Map), the French CNES (Centre National d’Etudes Spatiales), the French Polar Institute IPEV (Institut Polaire Paul-Emile Victor) and the NASA/CNES OSTST ALTIMECO project. Edited by: I. Obernosterer