Ecological Indicators for the Pelagic Zone of the Ocean from Remote Sensing

Remote Sensing of Environment 112 (2008) 3426–3436

Contents lists available at ScienceDirect

Remote Sensing of Environment

journal homepage: www.elsevier.com/locate/rse

Ecological indicators for the pelagic zone of the ocean from remote sensing

Trevor Platt a,⁎, Shubha Sathyendranath b a Bedford Institute of Oceanography, Box 1006, Dartmouth, Nova Scotia, B2Y 4A2, Canada b Plymouth Marine Laboratory, Prospect Place, The Hoe, Plymouth PL1 3DH, United Kingdom

ARTICLE INFO ABSTRACT

Article history: It is generally accepted that responsible stewardship of the ocean implies ecosystem-based management. A Received 30 March 2007 requirement then arises for ecosystem indicators that can be applied in serial fashion with a view to Received in revised form 4 October 2007 detection of ecosystem change in response to environmental perturbations such as climate change or Accepted 14 October 2007 overﬁshing. The status of ecological indicators for the pelagic ecosystem is reviewed. The desirable properties of such indicators are listed and it is pointed out that remote sensing (ocean colour, supplemented by sea- Keywords: surface temperature) is an important aid to achieving them. Some ecological indicators that can be developed Ecological indicators Ecosystem-based management from remotely-sensed data on ocean colour are tabulated. They deal with the seasonal cycle of phytoplankton Remote sensing biomass, production and loss terms, annual production, new production, ratio of production to respiration, Ocean colour spatial variances in phytoplankton biomass and production, spatial distribution of phytoplankton functional Fisheries types, delineation of ecological provinces and phytoplankton size structure. Ocean ecosystems Crown Copyright © 2008 Published by Elsevier Inc. All rights reserved.

1. Preamble 2. Some existing indicators and their limitations

In the stewardship of the ocean, it is now generally accepted that What measurable properties of the pelagic ecosystem have been procedures for management and continuing oversight should have an suggested as potential ecological indicators? Rice (2003) has sum- ecosystem basis (Garcia & Cochrane, 2005). In other words, it is marised the variety of ecological indicators that have been proposed understood that management ought not to be motivated by only one or that are already in use, and has written at length on factors arising or a few narrow goals (which may in fact be mutually antagonistic), for in their implementation. They number “in the hundreds”. They can be example the yield or abundance of particular exploited fish stocks. organised into rather few classes. Rather, it is believed that the ecological context should be considered Indicator species, of which an exotic example is the canary in the in some holistic way, such that, in the example just mentioned, the mine shaft, may reveal something about the environment, but integrity of the ecosystem in which the exploited fish stocks are necessarily cannot be expected to convey information on the entire embedded will not be jeopardised, as it might otherwise be, by ecosystem (Spellerberg, 1991). Similar limitations would apply to concentrating attention on only a small part of the whole. attributes of the population of a particular species (especially of The mandate then is to think broadly and think ecologically. Given valued, exploited populations), such as its size distribution. The this point of departure, the next requirement is to identify connection between population structure of one species and the characteristics of the ecosystem that capture the imperative to quan- integrity of the ecosystem as a whole remains to be demonstrated. tify its somewhat elusive properties such as health, vigour or Indeed, such indices are merely extensions of an approach, already resilience. In particular, we need to identify or develop ecosystem found wanting, based on dynamics of single populations. Because they metrics that, if applied in a serial manner, would enable us to detect are not properties of the ecosystem-at-large, they have only restricted whether the ecosystem is modified in any significant way by, for value as ecological indicators. example, the suite of processes we describe collectively as climate Indicators based on the relative abundance of species in a change or by heavy fishing. We refer to these desired metrics as community (evenness, richness, diversity) are numerous (Ludwig & ecological indicators. In this paper, we consider the use of remote Reynolds, 1988; Legendre & Legendre, 1998). Apart from technical sensing in the development and application of ecological indicators limitations on interpretation, these methods have the disadvantage for the pelagic marine ecosystem. that implementation requires identification of species and enumeration of individuals. Generally, such activities are difficult, time- consuming and costly (Bundy et al., 2005). Ordination methods (Gauch, 1982; Legendre & Legendre, 1998) likewise suffer from fi ⁎ Corresponding author. dif culties in interpretation of results and are costly to implement, E-mail address: [email protected] (T. Platt). given that they also require identification and enumeration.

0034-4257/$ – see front matter. Crown Copyright © 2008 Published by Elsevier Inc. All rights reserved. doi:10.1016/j.rse.2007.10.016 T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436 3427

Another class of indicators seeks to capture community properties is worthwhile to consider what properties ideal indicators might have. without identification of species. An example is the community size They should represent some well-understood and widely-accepted spectrum, for which a body of theory already exists outside the subject ecosystem property that can be quantified unambiguously in standard of ecological indicators (Platt & Denman, 1977, 1978; Silvert & Platt, units. They should be measurable rapidly, at low cost, with a repeat 1978, 1980). Implementation does not require the expertise to identify frequency compatible with the intrinsic time scale implied in the taxa. Sizing of the organisms can be done in part by automatic measurement and also with the time scale relevant to the applications methods, but for some size intervals, individual sizing with the light envisaged. Ideally, the same indicator would be suitable for a variety of microscope may still be necessary. Implementation is not trivial applications, implying that indicators of choice would be measurable (Blanchard et al., 2005; Jennings & Dulvy, 2005; Stobberup et al., at a variety of scales. They should be capable of implementation in 2005). When the variables are suitably transformed, the spectrum can many locations using the same objective methodology, to allow be linearised, the slope and intercept having ecological interpreta- comparisons between ecosystems. tions. For example, the intercept is an indirect measure of ecosystem These are stringent conditions indeed, and it would be impossible productivity. But as Rice (2003) points out with justification, if the goal to meet them using conventional sampling platforms. However, they were to index ecosystem production, it would be more expedient to are not difficult to meet using remote sensing, provided a suitable measure it directly than to infer it from a size spectrum. metric can be defined. Remote sensing has the potential to provide A final class of indicators is based on quantitative attributes of the data with high spatial resolution (1 km or less) at high repeat community as revealed by input of observations to a model or models frequency (1 day). In developing indices from remotely-sensed of the marine ecosystem. For example, a compartmental model of the imagery, all spatial structure can be preserved (data, and indicators pelagic food web could be used to infer ecological fluxes between derived therefrom are georeferenced). Methods that fail to document compartments from measurements of the biomasses in the compart- and conserve spatial structure will be intrinsically inferior to those ments according to the inverse method (Vézina & Platt, 1988). Any set that do not. At the same time, the possibility always remains of taking of fluxes, or the aggregate of fluxes, may be used as ecological averages, weighted or unweighted, over any spatial domain of interest. indicators (Moloney et al., 2005). In another approach, mass-balance Potentially, coverage is global, so that any degree of spatial aggrega- criteria are used to infer particular ecological fluxes that may tion is possible: the oceanographic context can be shown for any otherwise be exceedingly difficult to measure (Christensen & Pauly, chosen region. Of course, data collected by ships provide a valuable 1992; Christensen & Walters, 2004). These approaches have two major and complementary means to enhance the information collected by limitations. First, the results will depend on the structure (food-web remote sensing. topology) of the ecosystem model in use (Cury et al., 2005). As an Turning to the choice of a suitable indicator, or suite of indicators, example, a model that failed to include a microbial loop would yield a we note that one of the most useful things to know about any quite different mass balance, and a different set of fluxes, than one in ecosystem is the autotrophic biomass. This is precisely the deliverable which a microbial loop were included. Because the same set of from visible-range spectroradiometry of the sea (ocean colour), a observations can be made to yield different conclusions depending on method to produce spatial fields of autotrophic biomass, indexed as the underlying model, the results may be contentious, and inacces- concentration of chlorophyll. Moreover, the biomass fields can be sible to those not expert in ecosystem modelling. The second difficulty converted to fields of primary production using methods based on the is that implementation requires enormous effort of sampling, analysis first principles of plant physiology. Thus, we have two important and interpretation just to make a single realisation of the model. It ecosystem properties that can be surveyed at the scales of time and may be prohibitively expensive to repeat them. space required for ecological indicators at very little cost compared Given that the rationale for development of ecological indicators is with the costs of conventional surveys. Such biological fields can be to detect temporal modifications in the structure and function of the complemented by sea-surface temperature fields (an important ecosystem occurring under perturbations, either natural or anthro- property of the environment), with the same temporal and spatial pogenic, it follows that the indicators should be computed at intervals resolution as for chlorophyll, also collected by remote sensing. so that possible differences may be revealed. The requirement for Recalling the goal of detecting ecosystem change, we note that a serial measurements raises another issue, that of resource require- fundamental application of remotely-sensed data on ocean colour is ment. As we have seen, many of the candidate indicators are the construction of time series. To effect this, all available (potentially exceedingly difficult, and costly, to measure. Hence the frequency in daily) imagery within a chosen period (say, one week) is combined to time at which they could be measured may be compromised by the produce a single image representative of conditions during that cumulative cost of repeated measurement. If indicators are too period. The resulting image is referred to as a composite image, and expensive to implement sufficiently often to document change, and the period over which it is constructed will be the temporal resolution in a timely enough manner to be useful, they have little merit as of the time series of composites. Creation of composites reduces the operational metrics. incidence of missing data points resulting from cloud cover or from Similarly, the spatial representativeness of observations may also any other difficulty in image processing. be compromised by the costs involved in making the measurement. The advantages of a time series are twofold. On the one hand it Fixed-point measurements from ships offer no peripheral vision. If permits the temporal development of ecological processes to be only one location is used for all the sampling, it will be a difficult realised and quantified: the seasonal dynamics are accessible. On the problem to know the spatial extent that this station can be taken to other it permits the comparison of conditions, or dynamics, between represent. And if it is indeed representative of a particular ecological years (Fig. 1). Of course, the extent to which the dynamics are revealed province (sensu Platt & Sathyendranath, 1988, 1999; Longhurst, 1998), depends on the temporal resolution of the series. But for the it will be impossible to know how the resultant indicators are autotrophic biomass and the primary production, as developed from influenced by conditions in neighbouring provinces (the oceano- ocean colour, a resolution of one week in the composite images can graphic context). usually be achieved, and this is quite sufficient to elucidate the seasonal dynamics. Access to the seasonal dynamics opens the way to 3. Desirable properties of ecological indicators: potential role of other choices for ecological indicators. If we characterise the dynamics remote sensing in some quantitative manner, we can base further ecological indicators on the phase relationships contained therein. These are In view of the difficulties surrounding the development and objectively-determined quantities that can be expressed in standard routine application of ecological indicators as conceived at present, it units with a resolution of 1 km and 1 week. As computed from a time 3428 T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436

Fig. 1. Remotely-sensed chlorophyll fields for Northwest Atlantic Ocean for second half of May, series of images for six-year period 1998 to 2003. This image and all subsequent satellite images are based on data from the SeaWiFS (Sea Wide Field of View Sensor). Pixels with missing data appear in white. series, they reveal the interannual variability in the phase relations of two choices. The threshold may be defined in absolute units [M L− 3], in the seasonal cycle of phytoplankton. which case it would be vulnerable to variations between sensors, or in An important technical issue relating to remote sensing is the relative terms (relative to the height of the peak itself), in which case it merging of data from different missions to achieve a seamless time would not. series of data when the data stream from one sensor is replaced by Notwithstanding the many advantages remote sensing provides in that from another. It is a problem that has yet to be solved: one cannot the development of ecological indicators, the applications to date have always be sure that the absolute values of an observable from two data been very few. For example, in the proceedings of a recent symposium streams have the same meaning (but see IOCCG, 2004). Does this on the subject (Daan, 2005), including some forty contributions, only invalidate the potential of remotely-sensed data as a basis for one featured remote sensing (Polovina & Howell, 2005). In our view, ecological indicators? The answer is in the negative because, by remotely-sensed data have been under-utilised in this field. Using choosing the indices carefully, they can be made independent of examples from the northwest Atlantic Ocean, we aim to show in this absolute calibration. For example, the date of the spring chlorophyll paper how one stream of remotely-sensed data, that of visible spectral maximum and the duration of the bloom meet this criterion. For the radiometry (ocean colour), can contribute to the provision of initiation of the bloom, we need to specify a threshold, and we have ecological indicators for the pelagic ocean. T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436 3429

Fig. 2. Schematic diagram showing development of a time series for chlorophyll concentration (biomass) at a given pixel from a sequence of satellite images.

In proposing the ecological indicators from remote sensing, we maximum amplitude, the time of initiation, the time of the peak and take into account the ideal characteristics of ecological indicators; the the duration of the bloom (Platt et al., 2003). Time of initiation is attributes, advantages and limitations of satellite data; and potential decided according to a threshold criterion. The threshold may be set in applications envisaged. The approach used is pragmatic, relying on absolute terms (chlorophyll concentration) or in relative terms (some what is feasible from currently-available satellite data, and the goal is fraction of the maximum amplitude). The important point is that all to explore ways in which satellite data can be applied to complement these characteristics of the bloom can be assigned in an objective indicators derived from in situ data. manner, for every pixel, with all spatial structure preserved (Fig. 4). Their significance can be understood easily by a non-specialist. 4. Ecological indicators from visible spectral radiometry In the northwest Atlantic Ocean, we have found substantial variation between years in the characteristics of the spring bloom. The first requirement is to develop a time series of the principal For example, the timing of the peak has a variability of about six weeks observable, chlorophyll concentration, for the area of interest. We take or more (Fig. 5). Because many fish and invertebrates reproduce chlorophyll concentration to be a first-order index of phytoplankton around the time of the bloom, variations in bloom timing may affect biomass (Platt & Sathyendranath, 1988). It is the best index of biomass their recruitment. We have found evidence that this may be so for a to use for calculation of primary production. The data may be used at ground fish (Platt et al., 2003) and an invertebrate (Koeller et al., 2007; the spatial resolution provided (nominally 1 km at time of writing) or Fuentes-Yaco et al., 2007). aggregated over larger scales. All the available imagery is collected and data are combined to make composite images at a suitable temporal 4.2. Indicators describing the seasonal dynamics of phytoplankton scale. Here a compromise has to be made between the increased risk production of having empty pixels (due to cloud cover or other problems in image analysis) when the time scale is shorter and the higher chance of In principle, chlorophyll concentration can be used to estimate aliasing the dynamics when the time scale is longer. For the northwest phytoplankton production, given data on solar irradiance. Chlorophyll Atlantic Ocean, we have found that one-week composites are suitable. fields obtained by remote sensing can therefore be exploited (Fig. 6)to The set of composite images constitutes the time series. In fact this set yield fields of primary production (Platt & Sathyendranath, 1988, 1993, represents many time series, one for each pixel in the images (Fig. 2). It is important to emphasise that all spatial relationships are preserved between the elemental time series for particular spatial points. At all times we can maintain all spatial structure available in the data. The derived indicators can therefore be considered as spatialised (or georeferenced), in the sense used by Fréon et al. (2005) and Babcock et al. (2005).

4.1. Indicators describing the seasonal dynamics of phytoplankton biomass

The spring bloom of phytoplankton is the most important event in the trophic calendar of the pelagic systems. It is the strongest signal in the seasonal variation of chlorophyll concentration. It can be Fig. 3. Objective characterisation of the properties of spring phytoplankton bloom at a characterised with respect to amplitude, phase and width using the given pixel from a time series of chlorophyll concentration. Deﬁning characteristics are chlorophyll time series. Speciﬁcally (Fig. 3), we can identify the amplitude, initiation, timing of maximum and duration. 3430 T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436

Fig. 4. Climatology of spring bloom characteristics for Northwest Atlantic Ocean. Note that, for exceptional areas with consistently high biomass, such as Georges Bank, the apparent duration approaches twelve months, indicating that the conventional deﬁnition of the spring bloom may not be applicable there. Bathymetry (200 m contour) is shown as a white line. T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436 3431

4.3. Indicators describing other community processes

Given the current (chlorophyll) biomass field and an estimate of the phytoplankton production field for a given composite image, we can anticipate the biomass field in the next composite image. In general, the predicted biomass field will indicate higher biomass than that observed in the next composite image, because loss terms (Platt et al., 1991) will have been ignored. In fact, the difference between the observed and predicted biomass fields (possibly corrected for advection) is a first-order estimate of the generalised loss field for phytoplankton. This is an important point because loss terms for phytoplankton (including grazing, respiration, sinking and death by pathogens) are extremely difficult to estimate in any other way. As input to coupled ocean-ecosystem models, an independent estimate of these loss terms is very useful. Another application for such information is in solving the Sverdrup (1953) equation to forecast the initiation of phytoplankton blooms.

4.4. Indicators related to variance in the chlorophyll ﬁelds

In ﬁsheries science, there is a growing belief (Berkeley et al., 2004) that only a small proportion of individuals in a spawning stock contribute to recruitment of the next cohort. Spawners that are successful in this respect are the ones spawning in the most favourable place at the most favourable time (both conditions subject to changes between years). According to Hedgecock (1994a,b),ina given year, most spawning ﬁsh are unsuccessful in producing offspring

Fig. 5. An example of an anomaly ﬁeld for a spring bloom characteristic, in this case the timing of the chlorophyll peak anomaly for 1998. Positive anomaly means that the bloom peaked later than the climatological average.

1999; Platt et al., 2008-this issue) from which time series can be developed as before. Similar indices for the spring bloom can be developed from primary production as they were for chlorophyll. But here there is another important consideration. The dimensions of primary production [M L− 2 T− 1] include time in the denominator: it is a rate measurement. Hence, its magnitude can be integrated through time to give further useful results. For example, we can integrate through the year to recover annual production, a significant ecosystem property that is otherwise estimated indirectly, and with less confidence, for example as the intercept on a community size spectrum. We can also integrate through the life of the bloom to find the total phytoplankton production during the bloom (Fig. 7). In fact, this quantity is a lower bound on the annual new production (part of primary production that is supported by nitrogen supplied from outside the surface layer), under the assumption that all the nitrate in the surface layer at the end of the winter would be assimilated into phytoplankton tissue by the end of the bloom. In the context of renewable resources from the sea, the new production (Platt et al., 1992) represents the nitrogen equivalent that can be removed from the pelagic zone in one year without jeopardising the integrity of the ecosystem (roughly equivalent to the production required to support exploitation in a fishery operating at maximum capacity). Given a lower bound on the new production, we can also find a lower bound on the f-ratio averaged at the annual scale (the export ratio), which is the new production divided by the total primary production, and which has been shown to be related to the ratio of production to Fig. 6. An example of the biweekly composite for phytoplankton production (second biomass (Quiñones & Platt, 1991). half of May, climatology). 3432 T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436

4.5. Indicators related to community structure

Phytoplankton biomass and production, as estimated by remote sensing, may be thought of as indices that address the question “How much?”. The other issue is: “What kind?”. As well as quantifying biomass and its turnover, we are also interested in qualitative matters such as the gross community structure. One reason is that, in the context of research on the ocean carbon cycle and climate change, the conventional models of the ocean ecosystem, in which all phytoplankton are grouped under a single state variable, are being replaced by models that take into account the so-called phytoplankton functional types (PFT). Here it is recognised that different phytoplankton taxa may have different biogeochemical functions that should be considered if ecosystem models are to be improved (Le Quéré et al., 2005). Hence the requirement to develop methods for identifying major taxa by remote sensing. For example, the diatom group is unusual in having a high requirement for silicon, and is an important agent for export of carbon from the surface to the deep sea. Sathyendranath et al. (2004) have developed an algorithm to assess whether the phytoplankton community at a particular location, as seen in remotely-sensed imagery, is dominated by diatoms (Fig. 9). This is a very useful information for ecosystem analysis and for biogeochemical calculations. Alvain et al. (2005) have also published a procedure to index functional groups using remotely-sensed data. There is general interest to improve our collective capacity in this regard (Nair et al., 2008-this issue). The outlook is positive that we should soon be able to produce time series showing the community Fig. 7. Phytoplankton production integrated over the duration of the spring bloom. The structure of phytoplankton as another ecological indicator. production field in this image is based on climatology of blooms, and the magnitudes therefore represent upper bounds on integral production (longest possible integration There is also evidence that trophic status of a pelagic ecosystem (as time because the procedure tends to broaden the apparent duration). In a particular indexed by its chlorophyll biomass) may be indicative of the size year, the production values would be smaller. This bias is more pronounced in the area structure of the resident phytoplankton (Chisholm, 1992; Bouman fl under the in uence of the Gulf Stream, where the seasonal dynamics are less et al., 2005; Platt et al., 2005). Thus, more eutrophic systems tend to pronounced, and where the temporal cycle may be aliased by advection. support more larger cells than do oligotrophic ones. The implication is that information on the size structure of the phytoplankton is latent in that survive because the spawning is not optimised, neither spatially the biomass fields. Furthermore, the absorption characteristics of nor temporally, to the environmental conditions most favourable for larval survival. The recruited cohort is then the achievement of only the small fraction of adults whose spawning is so optimised (compare Cushing, 1969, 1975, 1990). A corollary is that recruitment would not be even over the extent of the nursery grounds, emphasising the importance of the spatial variability of the environment, as well as its temporal evolution. For any given region, for example the presumed nursery ground for a particular species or group of species, it is simple to quantify the spatial variability in the biomass and the production of phytoplankton in a given season. For a given average biomass or production in a particular area, its value for recruitment may depend on whether it is distributed evenly or is highly aggregated. But the variability can be quantified objectively and reported as a property of the ecosystem. Related to the issue of spatial variability in conditions favourable for larval survival and the differential contribution of spawners in a given year to recruitment of the next cohort is the significance of age distribution in the spawning stock. It is found that older and larger fish produce more viable offspring and thus have a higher chance than younger fish to contribute to the next cohort (Longhurst, 2002). Berkeley et al. (2004) conclude that the best way to conserve a broad age structure (which is of course vulnerable to predation by man) in marine populations is interconnected networks of marine protected areas. The question then arises: Where should the protected areas be located? Here also, the chlorophyll time series can be very useful. For example, we may want to know the areas where phytoplankton biomass (or production) of a certain magnitude could be most consistently found (Fig. 8). Or areas where the biomass (or production) was least variable. These and similarly phrased questions can be Fig. 8. Distribution of areas of consistently high chlorophyll concentration. A given answered easily from a chlorophyll time series and used as input to colour indicates that the chlorophyll concentration exceeds the corresponding value decisions on locating future protected areas. shown on the colour bar more than 75% of the time. T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436 3433

Fig. 9. Probability that the phytoplankton community is dominated by diatoms, for images taken (a) in the spring (second half of April, 2002); and (b) in the summer (second half of August, in the same year). phytoplankton can also be derived from data on ocean colour (IOCCG, following the assumption that large cells dominate the community at 2006), and the absorption spectrum is also known to contain high chlorophyll concentrations and small cells dominate at low information on size structure. Sathyendranath et al. (2004) developed concentrations. Devred et al. (2006) veriﬁed this model in a variety of a two-component model of phytoplankton absorption in the ocean oceanic regimes, and were able to retrieve for any location the

Fig. 10. Ecological partitions of the Northwest Atlantic Ocean, (a) static partition according to Longhurst (1998); (b) dynamic partition according to Devred et al. (2007); and (c) static partition according to the Northwest Atlantic Fisheries Organization (see Halliday & Pinhorn, 1990). 3434 T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436 proportions of large and small (picoplankton and nanoplankton) cells this case visible spectral radiometry (ocean colour), has not been in the community. Here again, visible radiometry is able to supply attempted before. information on another aspect of community structure. The quantities so determined, proposed here as ecological indicators of the pelagic ocean, are collected in Table 1. Taken together, 4.6. Indicators related to the large-scale regional structure they represent a body of information, unobtainable by any other method, useful to assist ecosystem-based management of marine The final application we consider concerns the decomposition of resources. The original motivation for the work arose from assess- the region of interest into functional subregions for further analysis ments of the value of biological oceanography for fisheries manage- (Platt & Sathyendranath, 1999). This is referred to as partition into a ment (Platt & Sathyendranath, 1996; Platt et al., 2007). For example, suite of biogeochemical provinces (Platt & Sathyendranath, 1988). It after the collapse of major fisheries in the northwest Atlantic Ocean has been used, for example, for calculation of ocean biogeochemical towards the end of the last century, biological oceanography could fluxes at the basin and global scales (Longhurst et al., 1995; Longhurst, contribute little to debates on causality, lacking detailed information 2006; Sathyendranath et al., 1995) and for the calculation using at the appropriate scales of time and space. A prescription was set out remotely-sensed data of new production on and around Georges Bank (Platt & Sathyendranath, 1996) for application of ocean-colour data to (Sathyendranath et al., 1991). A detailed partition has been imple- the general question of ecosystem variability, emphasising the value of mented by Longhurst (1998). The partition delineates regions with time series of such data. If the indices listed in Table 1 had been common physical forcing in so far as phytoplankton production is available at the time of the fishery collapses, it is reasonable to concerned: abrupt changes in forcing occur across the boundaries. In suppose that the debates on causality would have been much more the idealised case, these boundaries are fixed (even rectilinear), but in informed about the possible role of ecosystem variation. reality they are elastic, varying for example with season. For depicting Recent results support this belief (Platt et al., 2007). For example, it the instantaneous (or the average) boundaries, the most useful aid is has been shown for a thirty-year time series of larval survival of the remotely-sensed data on ocean colour (Platt & Sathyendranath, 1999; haddock (Melanogrammus aeglefinus) on the continental shelf of Nova Platt et al., 2005; Devred et al., 2007). Such a picture of the ecological Scotia, that the two years of excellent survival corresponded to years structure of a region of interest is invaluable for management (Fig. 10). of exceptionally early spring phytoplankton blooms (Platt et al., 2003). For example, it can be used as a template for calculation of regional Moreover, throughout the availability of ocean-colour data, the larval ecological fluxes, where the ecophysiological (= biogeochemical) rate survival (normalised to spawning biomass) is higher when the bloom parameters vary with province. Another approach to partition of the is earlier than the climatological average. This amounted to an same area is reviewed in Halliday and Pinhorn (1990). operational test of the Hjort–Cushing Match–Mismatch Hypothesis On the subject of marine protected areas, Berkeley et al. (2004) (Hjort, 1914; Cushing, 1974, 1990). Again, it has been shown for the suggest that they be placed in different habitats within biogeogra- shrimp (Pandalus borealis) on the Newfoundland–Labrador Shelf that phical and oceanographic regions. The partition into provinces the size of the young in particular years is correlated with the timing provides the ideal template to identify candidate regions. of the bloom and also with the amplitude of the chlorophyll maximum (Koeller et al., 2007; Fuentes-Yaco et al., 2007). In these works, we 5. Discussion have found it advantageous (Platt et al., 2003) to present the indicators for particular years as anomalies or differences from the As illustrated above, remote sensing has great capability and climatological averages. potential for providing many indicators of the pelagic ecosystem in a Some of the indicators we propose can be seen as superior and cost- consistent, simple and cost-effective manner. But remote sensing is effective alternatives to ones already in common use. For example, the not without its own problems: consistent cloud cover, for example, primary production is here estimated directly, and with great spatial can introduce gaps in the time-series data, and merging data streams detail, rather than indirectly as the intercept of a biomass spectrum. from different sensors is technically difficult (IOCCG, 2004). Remote Moreover, the high spatial resolution enables assessment of the spatial sensing cannot therefore be considered to be a universal panacea for variances in the fields of biomass and production, now believed to be of all observational needs. In general, combining satellite observations substantial importance in the control of recruitment. with well-designed in situ observation is highly desirable because of Again, the delineation of ecological provinces provides a template complementarity of the two types of data: in situ observations are able on which to map information about (observations on) the pelagic to provide fine-scale, vertically-resolved data, whereas satellite data ecosystem, a richer intellectual scaffolding than, for example, the yield high-resolution horizontal maps; and not all variables amenable orthogonal grids on which fishery statistics are often mapped. It is a to in situ measurements are accessible to satellite observations. In situ basis for many calculations of ecological fluxes over the region of observations also help fine-tune satellite algorithms for regional applications. In fact, the global network of ecological observations of the marine ecosystem called ChloroGIN (Chlorophyll Global Inte- Table 1 grated Network) initiated by the Group on Earth Observations (GEO), The ecological indicators proposed here for the pelagic ocean, all developed from remotely-sensed spectral radiometry in the visible (ocean colour) which has the goal of integrating in situ and satellite observations at the global scale, is a step in the right direction. Indicator Label Dimensions

Given the richness in the ecological indicators to be derived from Initiation of spring bloom bi [T] − 3 remote sensing, it is surprising that so little use of them has been Amplitude of spring bloom ba [M L ] made in this context before now. In the paper mentioned earlier, Timing of spring maximum bt [T] Duration of spring bloom bd [T] Polovina and Howell (2005) used data on sea-surface topography to − 2 Total production in spring bloom bp [M L ] fi − 2 derive empirical orthogonal functions (an ordination method, dif cult Annual phytoplankton production PY [M L ] − − to understand for non-specialists) at the basin scale; ocean-colour Generalised phytoplankton loss rate L [M L 3 T 1] − 3 data to compile descriptive chlorophyll time series and to document Integrated phytoplankton loss LT [M L ] Annual-scale f-ratio f Dimensionless the movement of a major oceanographic boundary; and sea-surface 2 2 − 6 Spatial variance in biomass field σB [M L ] 2 2 − 4 temperature data for assimilation into numerical models. These Spatial variance in production field σP [M L ] applications were complementary to existing research programs in Phytoplankton functional types NA NA fisheries oceanography. To our knowledge, a compilation of the Delineation of biogeochemical provinces NA NA possible ecological indicators to be derived from remote sensing, in Phytoplankton size structure s Dimensionless T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436 3435

a contribution to the NCEO and Oceans 2025 programmes of NERC. Remotely-sensed data courtesy of NASA and Orbimage.

References

Alvain, S., Moulin, C., Dandonneau, Y., & Bréon, F. M. (2005). Remote sensing of phytoplankton groups in case 1 waters from global SeaWiFS imagery. Deep-Sea Research I, 52, 1989−2004. Babcock, E. A., Pikitch, E. K., McAllister, M. K., Apostolaki, P., & Santora, C. (2005). A perspective on the use of spatialized indicators for ecosystem-based fishery management through spatial zoning. ICES Journal of Marine Science, 62(3), 469−476. Berkeley, S. A., Hixon, M. A., Larson, R. J., & Love, M. S. (2004). Fisheries sustainability via protection of age structure and spatial distribution of fish populations. Fisheries, 29(8), 23−32. Blanchard, J. L., Dulvy, N. K., Jennings, S., Ellis, J. R., Pinnegar, J. K., Tidd, A., & Kell, L. T. (2005). Do climate and fishing influence size-based indicators of Celtic Sea fish community structure? ICES Journal of Marine Science, 62(3), 405−411. Bouman, H., Platt, T., Sathyendranath, S., & Stuart, V. (2005). Dependence of light- saturated photosynthesis on temperature and community structure. Deep-Sea Research I, 52, 1284−1299. Bundy, A., Fanning, P., & Zwanenburg, K. C. T. (2005). Balancing exploitation and conservation of the Eastern Scotian shelf ecosystem: application of a 4d ecosystem exploitation index. ICES Journal of Marine Science, 62(3), 503−510. Chisholm, S. W. (1992). Phytoplankton size. In P. G. Falkowski, & A. D. Woodhead (Eds.), Primary productivity and biogeochemical cycles in the sea (pp. 213−237). New York: Plenum Press. Christensen, V., & Pauly, D. (1992). Ecopath-II — A software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling, 61(3–4), 169−185. Christensen, V., & Walters, C. J. (2004). Ecopath with Ecosim: Methods, capabilities and limitations. Ecological Modelling, 172(2–4), 109−139. Cury, P. M., Shannon, L. J., Roux, J. P., Daskalov, G. M., Jarre, A., Moloney, C. L., & Pauly, D. (2005). Trophodynamic indicators for an ecosystem approach to fisheries. ICES Journal of Marine Science, 62(3), 430−442. Cushing, D. H. (1969). The fluctuation of year classes and the regulation of fisheries. Fiskeridirektoratets skrifter, Serie HavundersOkelser, 15,368−379. Cushing, D. H. (1974). The natural regulation of fish populations. In F. R. Harden Jones Fig. 11. The anomaly of chlorophyll concentration for the second half of July 2006, (Ed.), Sea fisheries research (pp. 399−412). London: Elek Science. revealing an intense episodic bloom in the central Labrador Sea. Cushing, D. H. (1975). Marine ecology and fisheries. Cambridge: Cambridge University Press. Cushing, D. H. (1990). Plankton production and year-class strength in fish populations: An update of the Match/Mismatch hypothesis. Advances in Marine Biology, 26, 249−294. interest. The taxonomic manifestation of this underlying structure is Daan, N. (2005). An afterthought: ecosystem metrics and pressure indicators. ICES Journal of Marine Science, 62(3), 612−613. revealed in the spatial distribution of phytoplankton functional types Devred, E., Sathyendranath, S., Stuart, V., Maass, H., Ulloa, O., & Platt, T. (2006). A two- and community size structure. component model of phytoplankton absorption in the open ocean: Theory and Looking to the future, application of the enormously rich ocean- applications. Journal of Geophysical Research, 111,C03011.doi:10.1029/2005JC002880. colour data stream to routine collection of these objective indicators Devred, E., Sathyendranath, S., & Platt, T. (2007). Delineation of ecological provinces using ocean colour radiometry. Marine Ecological Progress Series. should provide an invaluable set of information on variability in the Fréon, P., Drapeau, L., David, J. H. M., Moreno, A. F., Leslie, R. W., Oosthuizen, W. H., pelagic ecosystem within and between years, recognising the Shannon, L. J., & van der Lingen, C. D. (2005). Spatialized ecosystem indicators in the − fundamental importance of the biophysical systems that underlie Southern Benguela. ICES Journal of Marine Science, 62(3), 459 468. Fuentes-Yaco, C., Koeller, P. A., Sathyendranath, S., & Platt, T. (2007). Shrimp (Pandalus fisheries (Scandol et al., 2005). It is as general intelligence, rather than borealis) growth and timing of the spring phytoplankton bloom on the Newfound- as a way to set thresholds used to trigger managerial response, that land–Labrador Shelf. Fisheries Oceanography, 16(2), 116−129. doi:10.1111/j.1365- the main value of these indicators lies. We do not pretend that 2419.2006.00402.x. Garcia, S. M., & Cochrane, K. L. (2005). Ecosystem approach to fisheries: A review of indicators derived from remote sensing can replace all others. Rather, implementation guidelines. ICES Journal of Marine Science, 62(3), 311−318. we advocate their use as timely, georeferenced supplements, highly Gauch, H. G. (1982). Multivariate analysis in community ecology. In E. Beck, H. Birks, & E. resolved in time and space, comprehensible by non-specialists, to Connor (Eds.), Cambridge studies in ecology (pp. 298). New York: Cambridge University Press. enrich the information base upon which ecosystem-based manage- Halliday, R., & Pinhorn, A. (1990). The delimitation of fishing areas in the Northwest ment is conducted. In principle, they can be issued in real time and Atlantic. Journal of Northwest Altantic Fishery Science, 10 (December), 1−51. afford a priceless view of the oceanographic context in which to place Hedgecock, D. (1994). Does variance in reproductive success limit effective population sizes of marine organisms? In A. R. Beaumont (Ed.), Genetics and evolution of aquatic data collected by ships. Anomaly representations can give immediate organisms (pp. 122−134). London: Chapman & Hall. warning in cases where the ecosystem dynamics depart strongly from Hedgecock, D. (1994). Temporal and spatial genetic-structure of marine animal the climatological mean (Fig. 11). Doubts about representativeness of populations in the California Current. California Cooperative Oceanic Fisheries Investigations Reports, 35,73−81. observations (Rice, 2003) become irrelevant because all the spatial Hjort, J. (1914). Fluctuations in the great fisheries of Northern Europe, viewed in the structure in the original fields is preserved and the data can be light of biological research. Cons. Int. Explor. Mer, 20,1−228. averaged over any spatial domain according to requirements. IOCCG (2004). Guide to the creation and use of ocean-colour, level-3, binned data products. In D. Antoine (Ed.), Reports of the International Ocean-Colour Coordinating Group, No. 4 Dartmouth, Canada: IOCCG. Acknowledgements IOCCG (2006). Remote Sensing of Inherent Optical Properties: Fundamentals, Tests of Algorithms, and Applications. In Z. -P. Lee (Ed.), Reports of the International Ocean- We thank Carla Caverhill, George White, Emmanuel Devred, Cesar Colour Coordinating Group, No. 5 Dartmouth, Canada: IOCCG. Jennings, S., & Dulvy, N. K. (2005). Reference points and reference directions for size-based Fuentes-Yaco and Heidi Maass for their help in the preparation of the indicators of community structure. ICES Journal of Marine Science, 62(3), 397−404. illustrations. We thank Alida Bundy and Venetia Stuart for helpful Koeller, P. A., Fuentes-Yaco, C., & Platt, T. (2007). Decreasing shrimp (Pandalus borealis) comments on the manuscript. This work was supported by the Canadian sizes off Newfoundland and Labrador — Environment or fishing? Fisheries Oceanography, 16(3), 105−115. doi:10.1111/j.1365-2419.2006.00403.x. Space Agency (GRIP program) and by Discovery Grants toTP and SS from Le Quéré, C., Harrison, S. P., Prentice, I. C., Buitenhuis, E. T., Aumont, O., Bopp, L., Claustre, the Natural Sciences and Engineering Research Council of Canada. This is H., da Cunha, L. C., Geider, R., Giraud, X., Klaas, C., Kohfeld, K. E., Legendre, L., 3436 T. Platt, S. Sathyendranath / Remote Sensing of Environment 112 (2008) 3426–3436

Manizza, M., Platt, T., Rivkin, R. B., Sathyendranath, S., Uitz, J., Watson, A. J., & Wolf- Platt, T., Sathyendranath, S., & Fuentes-Yaco, C. (2007). Biological oceanography and Gladrow, D. (2005). Ecosystem dynamics based on plankton functional types for fisheries management: Perspective after ten years. ICES Journal of Marine Science, global ocean biogeochemistry models. Global Change Biology, 11,2016−2040. 64(5), 863−869. Legendre, P.,& Legendre, L. (1998). Numerical ecology, 2nd ed. Amsterdam: Elsevier Science. Platt, T., Sathyendranath, S., Forget, M. -H., White, G. N., Caverhill, C., Bouman, H., Longhurst, A. (1998). Ecological geography of the sea. San Diego: Academic Press. Devred, E., & Son, S. (2008). Operational estimation of primary production at large Longhurst, A. (2002). Murphy's Law revisited: Longevity as a factor in recruitment to geographical scales. Remote Sensing of Environment, 112, 3437−3448 (this issue). fish populations. Fisheries Research, 56,125−131. doi:10.1016/j.rse.2007.11.018. Longhurst, A. (2006). Ecological geography of the sea, 2nd ed. Amsterdam: Elsevier. Polovina, J. J., & Howell, E. A. (2005). Ecosystem indicators derived from satellite Longhurst, A., Sathyendranath, S., Platt, T., & Caverhill, C. (1995). An estimate of global remotely sensed oceanographic data for the North Pacific. ICES Journal of Marine primary production in the ocean from satellite radiometer data. Journal of Plankton Science, 62(3), 319−327. Research, 17(6), 1245−1271. Quiñones, R. A., & Platt, T. (1991). The relationship between the f-ratio and the P: Ratio Ludwig, J. A., & Reynolds, J. F. (1988). Statistical ecology: A primer in methods and in the pelagic ecosystem. Limnology and Oceanography, 36(1), 211−213. computing. New York: John Wiley and Sons. Rice, J. (2003). Environmental health indicators. Ocean and Coastal Management, 46(3–4), Moloney,C.L.,Jarre,A.,Arancibia,H.,Bozec,Y.M.,Neira,S.,Roux,J.P.,&Shannon,L.J.(2005). 235−259. Comparing the Benguela and Humboldt marine upwelling ecosystems with indicators Sathyendranath, S., Longhurst, A., Caverhill, C. M., & Platt, T. (1995). Regionally and derived from inter-calibrated models. ICES Journal of Marine Science, 62(3), 493−502. seasonally differentiated primary production in the North Atlantic. Deep-Sea Nair, A., Sathyendranath, S., Platt, T., Morales, J., Stuart, V., Forget, M. -H., Devred, E., & Research I, 42(10), 1773−1802. Bouman, H. (2008). Remote sensing of phytoplankton functional types.Remote Sathyendranath, S., Platt, T., Horne, E. P. W., Harrison, W. G., Ulloa, O., Outerbridge, R., & Sensing of Environment, 112, 3366−3375 (this issue). doi:10.1016/j.rse.2008.01.021. Hoepffner, N. (1991). Estimation of new production in the ocean by compound Platt, T., Bird, D. F., & Sathyendranath, S. (1991). Critical depth and marine primary remote sensing. Nature, 353,129−133. production. Proceedings of the Royal Society of London. Series B, Biological Sciences, Sathyendranath, S., Watts, L., Devred, E., Platt, T., Caverhill, C., & Maass, H. (2004). 246, 205−217. Discrimination of diatoms from other phytoplankton using ocean-colour data. Platt, T., Bouman, H., Devred, E., Fuentes-Yaco, C., & Sathyendranath, S. (2005). Physical Marine Ecology Progress Series, 272,59−68. forcing and phytoplankton distributions. Scientia Marina, 69(1), 55−73. Scandol, J. P., Holloway, M. G., Gibbs, P. J., & Astles, K. L. (2005). Ecosystem-based Platt, T., & Denman, K. L. (1977). Organisation in the pelagic ecosystem. Helgolander fisheries management: An Australian perspective. Aquatic Living Resources, 18(3), Wissenschaftliche Meeresuntersuchungen, 30,575−581. 261−273. Platt, T., & Denman, K. L. (1978). The structure of pelagic marine ecosystems. Rapport Et Silvert, W., & Platt, T. (1978). Energy flux in the pelagic ecosystem: A time-dependent ProceÂs-Verbaux Des ReÂunions - Conseil Permanent International Pour L'exploration equation. Limnology and Oceanography, 23(4), 813−816. De La Mer, 173,60−65. Silvert, W., & Platt, T. (1980). Dynamic energy-flow model of the particle size Platt, T., Fuentes-Yaco, C., & Frank, K. T. (2003). Spring algal bloom and larval fish distribution in pelagic ecosystems. In W. C. Kerfoot (Ed.), Evolution and ecology of survival. Nature, 423(22 May, 2003), 398−399. zooplankton communities (pp. 754−763). New Hampshire: The University Press of Platt, T., Jauhari, P., & Sathyendranath, S. (1992). The importance and measurement of New England. new production. In P. G. Falkowski, & A. D. Woodhead (Eds.), Primary productivity Spellerberg, I. F. (1991). Monitoring ecological change. Cambridge: Cambridge University and biogeochemical cycles in the sea (pp. 273−284). New York: Plenum Press. Press. Platt, T., & Sathyendranath, S. (1988). Oceanic primary production: Estimation by Stobberup, K. A., Inejih, C. A. O., Traoré, S., Monteiro, C., Amorim, P., & Erzini, K. (2005). remote sensing at local and regional scales. Science, 241,1613−1620. Analysis of size spectra off Northwest Africa: A useful indicator in tropical areas? Platt, T., & Sathyendranath, S. (1993). Estimators of primary production for interpreta- ICES Journal of Marine Science, 62(3), 424−429. tion of remotely sensed data on ocean color. Journal of Geophysical Research, 98(C8), Sverdrup, H. U. (1953). On conditions for the vernal blooming of phytoplankton. Journal 14,561−14,576. Du Conseil, 18(3), 287−295. Platt, T., & Sathyendranath, S. (1996). Biological oceanography and fisheries manage- Vézina, A. F., & Platt, T. (1988). Food web dynamics in the ocean. I. Best-estimates of flow ment. International Council for the Exploration of the Sea, CM 0,3. networks using inverse methods. Marine Ecology Progress Series, 42, 269−287. Platt, T., & Sathyendranath, S. (1999). Spatial structure of pelagic ecosystem processes in the global ocean. Ecosystems, 2, 384−394.