Based on ten years (1998–2007) of satellite ocean color data we analyze the spatiotemporal patterns in the seasonal Madagascar plankton bloom with respect to the advection of the recently discovered Southern Indian Ocean Countercurrent (SICC). In maps of Finite-time Lyapunov Exponents (FTLE) and Finite-Time Zonal Drift (FTZD) computed from altimetry derived velocities we observe a narrow zonal jet that starts at ∼25°S at the southern tip of Madagascar, an important upwelling region, and extends to the east further than the largest plankton blooms (∼2500 km). In bloom years, the jet coincides with large parts of the northern boundary of the plankton bloom, acting as a barrier to meridional transport. Our findings suggest that advection is an important and so far underestimated mechanism for the eastward propagation and the extent of the plankton bloom. This supports the hypothesis of a single nutrient source south of Madagascar.
 The Madagascar plankton bloom is one of the largest dendroid blooms in the world oceans. It develops in summer at ∼25°S south east of Madagascar in zonal direction, reaching from the coast of Madagascar at ∼47°E up to ∼70°E into the South Indian Ocean. The largest extent is reached in February or March in most years, but a strong interannual variability exists (for an overview over bloom years see Raj et al. ). Several studies have addressed the mechanisms controlling the initiation and the propagation of the plankton bloom. In its first description Longhurst  proposes mixed layer deepening as the nutrient providing process and finds some agreement between central (peripheral) upwelling in cyclonic (anticyclonic) eddies and the chlorophyll pattern in the bloom in 1999. Srokosz et al.  focus on the eastward propagation of the plankton bloom and report its assumedly direction against the mean flow, obtained by tracking Eulerian features in the Sea Surface Height (SSH) field. To justify the expansion of the bloom to the east despite a supposed lack of eastward transport Srokosz et al. give a general explanation based on a high eddy diffusivity in a reaction-advection-diffusion system.Uz  contradicts the theory of local upwelling, as the bloom occurs in a shallow and stably stratified surface layer, and introduces the hypothesis of remote nutrient supply by river runoff at the Madagascar coast. Lévy et al. propose the well-known upwelling south of Madagascar [Machu et al., 2002; Ho et al., 2004; DiMarco et al., 2000] followed by eastward transport by a possible retroflection of the South East Madagascar Current [Quartly et al., 2006] as an explanation for the bloom. However, the details of the transport mechanisms remain unclear. Raj et al.  explore a variety of longtime data sets and find upwelling along the south Madagascar coast, precipitation along east Madagascar, and mesoscale eddies as likely key factors influencing the bloom.
 The goal of this letter is to investigate the relation between the SICC and the remarkable eastward extension of the Madagascar plankton bloom. While many oceanographic studies analyze mesoscale transport by focusing on Eulerian aspects of the circulation, we address transport, using Lagrangian methods, that take into account the temporal evolution of the flow, and present the first Lagrangian description of the SICC jet. Of particular importance is the notion of Lagrangian Coherent Structures (LCS) introduced by Haller and Yuan to refer to key material surfaces (i.e., transport barriers) that are responsible for shaping global mixing patterns in the flow. We show that the SICC generates a fast coherent eastward transport and acts as a weakly-deforming LCS that widely inhibits meridional transport, both processes controlling the spreading of the plankton bloom.
2. Data and Methods
 We analyze the Madagascar plankton bloom based on standard global maps of chlorophyll concentration observed by the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and provided by NASA's ocean color data site (http://oceancolor.gsfc.nasa.gov). We use the 8-day composite level 3 product with a spatial resolution of 1/12° ≈ 9 km and temporal resolution of 8 days.
 Plankton evolution is often described by complex population dynamics models with several constituents of the food chain (e.g., nutrients, phytoplankton, zooplankton). A large variety of such models exist [e.g., Oschlies and Garçon, 1999] which include terms and parameters that generally are difficult to determine. Here, we concentrate on passive advection, a key process involved in the evolution of spatiotemporal plankton patterns [Abraham, 1998].
 In particular, we consider advection by ocean currents derived geostrophically from altimetric SSH maps. AVISO provides weekly SSH anomalies at a spatial resolution of 1/4° with respect to a 7-year (1993–1999) mean. These are added to the mean dynamic topography of the ocean surface based on altimetry, in-situ measurements, and a geoid model [Rio et al., 2005]. In a longitude–latitude spherical coordinate system (λ, θ) the position of a fluid particle on the ocean's surface is thus assumed to evolve according to
where η(λ, θ, t) is SSH, R the Earth's mean radius, g gravity and f(θ) := 2Ωsinθ the Coriolis parameter with Ω the Earth's mean angular velocity. Here particle trajectories are computed by integrating (1)–(2) using a Runge–Kutta scheme and a tricubic interpolation in time and space. A fine grid of particles with an initial separation of 1/16° is advected to obtain fields of the Lagrangian quantities introduced below. The finite integration time is chosen to be τ = 12 weeks, a typical time scale for the development of the bloom and the propagation of the jet.
 Finite-Time Lyapunov Exponents (FTLE), a measure of stretching about fluid trajectories, are used here to estimate LCS as in many previous works [Peacock and Dabiri, 2010, and references therein]. FTLE are computed in forward and backward time direction as derived by Beron-Vera et al. . Weakly deforming LCS, indicating meridional transport barriers, are unveiled by identifying regions where ridges of the forward and backward FTLE field do not excessively transversely intersect one another and FTLE values are relatively small [Beron-Vera et al., 2010; Boffetta et al., 2001]. These weakly deforming LCS are associated with the cores of long meandering jets and thus will be referred to as jet-like LCS, in contrast to more entangled hyperbolic LCS, the locally strongest attracting or repelling material curves. Both types of LCS inhibit cross transport such that a plankton bloom evolving on one side of an LCS cannot be spread to the other side. More precisely, the strict barrier property of a jet-like LCS is limited by the coherence of the jet, i.e., the stability of its spatial structure under perturbations of the adjacent eddies. Real zonal jets are not always necessarily as coherent as ideal jets in model flows [Beron-Vera et al., 2010]. Thus, real zonal jets, while indeed providing a fast zonal transport, might behave as strong though imperfect barriers with part of the water entraining and detraining along their path leading to partial cross-mixing.
 To further quantify the eastward transport, we introduce the Finite-Time Zonal Drift (FTZD) defined as the zonal distance between the final and initial position (λ0, θ0) of a backward advected particle
with t = t0 + τ and τ < 0. It simply denotes the eastward directed zonal distance the particle has covered in the time τ.
 In order to demonstrate the ability of the SICC to transport water masses to the east from potential nutrient sources at the coast of Madagascar [Lévy et al., 2007; Uz, 2007], we advect a passive continuous tracer concentration (zeroth-order plankton model). We use a simple semi-Lagrangian advection scheme [Durran, 2010] without additional diffusion. The tracer is initialized as a Gaussian blob with a standard deviation of ∼150 km at the south east tip of Madagascar where upwelling has been reported to take place [DiMarco et al., 2000; Machu et al., 2002; Ho et al., 2004].
3. Results and Discussion
 In Figure 1we extract the location of zonal jets and quantify the zonal transport of these jets in order to demonstrate their impact on the extensive plankton bloom in 1999 that exhibits a pronounced eastward propagation. We estimate two jet-like LCS (highlighted with a red line) on 17 February 1999 from a map of the added forward and backward FTLE fields (Figure 1a). Along these potential transport barriers high gradients of tracers can be expected, and indeed, they coincide with large parts of the boundaries of the plankton bloom when compared to the spatial distribution of chlorophyll concentration in Figure 1c. In some regions chlorophyll patches can be found across the jet-like LCS where the jets are leaking. This leakage can be identified with hyperbolic LCS of eddies adjacent to the jet (Figure 1a), and thus can also be explained by advective transport.
 Additionally to the geometric information of transport barriers revealed by the jet-like LCS, the new simple Lagrangian metric FTZD(3) in Figure 1ballows for a spatially resolved quantification of the zonal transport east of Madagascar. This is especially useful to address a possible mechanism for the fast eastward propagation of the Madagascar plankton bloom. As expected, the elongated regions (in red) with the highest eastward zonal transport match the jet's core lines represented by the jet-like LCS. Both bands of high zonal drift are bounded piecewise by backward FTLE ridges, indicating that the transport bands are well separated from adjacent water masses, and also revealing that the jets are composed of several parts rather than being an uninterrupted structure of 2000 km length. Particle eastward excursions along these jets can be as long as ∼1200 km in 12 weeks, a distance comparable to a typical zonal extent of the Madagascar plankton bloom. With the underlying advection time of 12 weeks the jets are persistent transport bands at this time scale.
 The origin of the eastward spreading Madagascar plankton bloom was proposed to be located south of Madagascar [Uz, 2007; Lévy et al., 2007; Raj et al., 2010]. We consider this hypothesis and advect a passive tracer evolving from an initial Gaussian blob at the coast of Madagascar. In Figure 1dtwo filaments of high tracer concentration extend to the east corresponding to the two present jets, while the associated jet-like LCS shape the boundaries of the passive tracer. In the background, contours of the instantaneous SSH field are plotted for comparison. Although eddies, defined as regions enclosed by SSH contours [Chelton et al., 2011], drift westwards in this domain, the jets meander around them in eastward direction (Animation S1 in the auxiliary material). This is clear evidence for a distinct eastward transport, supporting the above mentioned hypothesis. In order to obtain a maximum similarity in extent and form between the tracer pattern and the real widespread plankton bloom on 17 February 1999 (Figure 2c), the tracer has to be released on 4 November 1998, about two months before first traces of the bloom can be seen. This early release of the passive tracer may be justified with the seasonality of possible nutrient sources close to the release point, being river runoff at the beginning of the rain season or upwelling [Uz, 2007; Raj et al., 2010].
 As a result, based on the above Lagrangian analysis of the flow conditions for the plankton bloom of 1999, we suggest two ways of impact of the zonal jets on the plankton bloom. First, a zonal jet represents a fast transport band that favors the eastward propagation of the plankton bloom. Second, a jet-like LCS, coherent over a sufficiently long time, prevents meridional transport. In the following, we study both effects of a zonal jet in detail and for all bloom years.
Figure 2shows the spatial chlorophyll distribution of the Madagascar plankton bloom with nearly maximal extent respectively in the bloom years 2000, 2002, 2004 and 2006, overlaid with the jet-like LCS extracted as inFigure 1a. In four out of five bloom years (1999, 2000, 2004, and 2006) persistent coherent jets are present along ∼25°S and the corresponding jet-like LCS largely mark the northern boundary of the plankton bloom. The jet-like LCS are embedded in a tangle of attracting hyperbolic LCS, estimated as ridges in the backward FTLE field (Figure 2a). In 2002, an important part of the plankton bloom is located north of the jet-like LCS. This is due to a strong perturbation of the jet at the time of the bloom, so that mixing across the jet occurs. In 2006, we observe the special case of a second jet-like LCS in the south, as also observed in 1999 (Figure 1c). In both years, the jet-like LCS coincides with the southern boundary of the chlorophyll data and likely represents an additional spatial confinement of the plankton bloom. Generally, in the other bloom years, the southern boundaries are also shaped by LCS, but rather by more entangled attracting hyperbolic LCS (white lines inFigure 2a) than by jet-like LCS.
 In Figure 3we quantify the eastward propagation of the plankton bloom and the eastward transport of the jet by means of space-time Hovmöller diagrams for several bloom years. SeaWiFS chlorophyll concentration data and the passive tracer are both averaged meridionally in the same narrow band along the principal axis of the bloom (box in inset inFigure 3b) leading to a one-dimensional time-dependent zonal concentration signal. InFigure 3a Hovmöller diagrams of the bloom in the years 1999, 2002, and 2004 show a clear eastward propagation with a similar front velocity of about 25°/17 weeks ≈ 2500 km/17 weeks ≈ 0.25 m/s, in contrast to the westward velocity of SSH features of typically 0.05 m/s. Srokosz et al.  measure the front velocity of the bloom in 1999 as ∼0.12 m/s in a Hovmöller diagram which we find underestimates the velocity by a factor of 2. In 2000 (not shown) the bloom expands similarly as in 2002. Only in 2006 a clear propagation of the bloom cannot be observed, it appears rather instantaneously. The passive tracer (Figure 3b) expands to the east along the jet in all years with a front velocity of about 0.14 m/s. This is consistent with the integrated jet velocity of 10°/12 weeks ≈ 1000 km/12 weeks ≈ 0.14 m/s estimated from the zonal transport of particles in Figure 1b, as can be expected due to the same underlying geostrophic velocity data. Note the formation of tracer filaments along the jet in the insets of Figure 3b that reveal a fast eastward expansion.
 The analysis of the passive advected tracer in Figure 3b shows that in all bloom years the jet provides a significant constant transport to the east, the direction of propagation of the Madagascar plankton bloom (Figure 3a). However, we find that the magnitudes of the front velocities of the tracer and the bloom deviate. This basically indicates that the spreading of the bloom cannot be understood as a purely passive advective process. We hypothesize that nutrients are most likely transported by the jet first, and then a reaction is triggered by another still unknown, possibly seasonal mechanism, while both processes are not necessarily well separated in time. It is also important to note that we did not address possible local vertical upwelling of nutrients along the jet [Christian et al., 2004] that can also accelerate the front velocity and could further explain plankton growth in regions across the jet.
 Two fundamental characteristics of the Madagascar plankton bloom, its eastward propagation from the south tip of Madagascar and its confinement within a narrow zonal band with low chlorophyll values in adjacent regions, can be linked to the presence of zonal jets in the South Indian Ocean Countercurrent (SICC). We focus here on the role of horizontal advection of these jets and therefore do not take into account any reaction dynamics or subgrid motion. Based on geostrophic velocities derived from altimetry data and using Lagrangian methods we suggest two basic mechanisms for the impact of the SICC jet on the Madagascar plankton bloom: first, it provides a fast and persistent eastward transport, and second, its zonal jet-like LCS represent a transport barrier in meridional direction shaping the boundary of the bloom. We find that the jet alone cannot be the only reason for the plankton bloom, as the eastward jet persists throughout the years not matching the seasonality and interanual variability of the bloom (Animation S2 in theauxiliary material). Additionally, we observe that the plankton front propagates with a higher velocity than the pure transport velocity of the jet. However, the resulting persistent eastward transport caused by the jet yield support for the hypothesis of a nutrient source at the south tip of Madagascar as the origin of the plankton bloom [Uz, 2007; Lévy et al., 2007; Raj et al., 2010].
 Finally, we conclude that the recently discovered SICC causes a to date overlooked, significant rapid transport to the east at the exact location of the Madagascar plankton bloom and should not be ignored in future studies of this phenomenon. Similar plankton patterns with sharp zonal meandering boundaries in the Atlantic North Equatorial Countercurrent off Brazil and in the Pacific North Equatorial Countercurrent off Indonesia [Christian et al., 2004] suggest that the results presented here may be important in other regions of the ocean.
 The altimeter products used in this work are produced by SSALTO/DUACS and distributed by AVISO, with support from CNES (http://www.aviso.oceanobs.com/duacs). The Rio05 product is produced by CLS Space Oceanography Division and is distributed by AVISO, with support from CNES (http://www.aviso.oceanobs.com). FJBV and MJO were supported by NSF grant CMG0825547 and OCE-0648284, and NASA grant NNX10AE99G. FH and AVK receive funding from FPU AP-2009-3550 and AP-2009-0713. VPM is supported by grant PGDIT09MDS009CT of the Xunta de Galicia.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.