Craig Young introduced key concepts in larval biology and highlighted the importance of distinguishing mechanisms of dispersal versus genetic connectivity, which is influenced by many processes including patterns of dispersal. Neither of these parameters are addressed equally well by a single, currently-available method, because they operate on different time scales. Marine populations are connected through individual exchange and they persist, decline or increase because of survival and recruitment. Thus, larval dispersal is often a key parameter when marine metapopulations are examined to understand connectivity patterns and develop conservation management models (Puckett and Eggleston 2016). Metapopulation studies involve either demographic modelling approaches or the analysis of the migratory component linking source and sink populations (Neubert et al. 2006, Mullineaux et al. 2018). The latter is based on the concept of a larval pool, through which populations are connected. Yet, the main question to be answered is how do larval interactions within their dynamic environment contribute to species and ecosystem resilience. In order to answer this question, we need more data and information on the temporal trends of connectivity, recruitment and survival.

Currently, knowledge on deep-sea larvae is collected using different approaches. Genetic connectivity is estimated for some species by molecular methods (sequencing), while potential dispersal capabilities or trajectories are estimated by modelling (e.g., Yearsley and Sigwart 2011, Young et al. 2018). Cultures of selected species also provide data on some physiological, developmental and behavioural traits, which are useful for constraining models of larval dispersal. Lastly, discrete data on vertical distribution are provided from in situ plankton sampling, as well as from a few time series, which allow both ground-truthing of model predictions and insights into in situ distribution patterns of deep-sea larvae.

Ocean current models can help estimate larval trajectories, but they are based on underlying assumptions, especially for biological parameters. Biological information is often missing, and modelling is thus best used as a predictive tool to inform future sampling. Planktonic larval duration (PLD) is one of the most important biological parameters needed to estimate larval dispersal. PLD is however extremely difficult to measure, due to the difficulties in culturing deep-sea larvae, and the fact that many factors, both biological and environmental, introduce intraspecific variability in PLD (e.g. Larsson et al. 2014, Rouse et al. 2009): poecilogony, delay of metamorphosis, maternal yolk investment, temperature, food availability, larval budding, and pre-settlement metamorphosis. Extrapolation from shallow relatives often fails to accurately estimate deep-sea PLDs, mainly because physiological and embryological (larval) processes are generally slower in deep-sea than in shallow-water species (e.g., Hilário et al. 2015). Indirect estimations from spawning and settlement patterns are sometimes used to infer development times and PLDs, but such estimates require knowledge of reproductive patterns. In situ collection of larvae is mainly hindered by the enormous dilution factor in the deep ocean that makes larvae extremely rare in this environment. Large volume plankton pumps deployed near vents may collect several hundreds of larvae (Mullineaux et al. 2005, Mullineaux et al. 2013). However, in most deep-sea ecosystems, the number of collected larvae is often less than 10 specimens, representing only a few morphotypes, with different morphotypes at different depths. Moreover, sorting is time consuming, and identification requires significant expertise. Although various systems have been employed over the years to overcome this issue, all have their caveats (see further discussion below and also Theme 2).

Larval diversity per se tells little of ecological interest, since it is a poor and unpredictable reflection of adult diversity in the context of metapopulation ecology. Critical knowledge gaps on deep-sea connectivity are mainly in the temporal patterns of dispersal and settlement (Metaxas 2004, Mullineaux et al. 2010), the vertical trajectories of larval dispersal (Arellano et al. 2014, Yahagi et al. 2017) and the comparison of larval and juvenile stages to identify post-settlement process (Arellano and Young 2010). Assessment of spatial-temporal patterns is a huge task that can only be tackled by strategic, long-term sampling through the deployment of multiple devices including automated plankton samplers (Doherty and Butman 1990, Lewis and Heckl 1991, O’Hara 1984), low cost tube traps, and recruitment experiments (Baldrighi et al. 2017, Cunha et al. 2013, Cuvelier et al. 2014). Assessing both larval supply and settlement patterns would allow identifying post-settlement processes that are relevant for recruitment success.

Because metapopulations are connected by dispersal and dispersal potential varies with depth, important unanswered questions are: At what depth do larvae disperse? and How much time do larvae spend drifting at various depths? Together with PLD, ontogenetic vertical trajectories of larval dispersal are the most essential biological parameters in modeling connectivity (Young et al. 2012). Possible approaches are:

1. target larval sampling at defined depth horizons sampling larvae using precision large-volume plankton samplers (e.g. Sentry/SyPRID, Billings et al. 2017);
2. collect larvae and juveniles from the ocean floor and demersal water column using tube larval traps; and
3. trace the characteristics of water masses (¹⁸O tracer) where larvae develop using isotopic and geochemical markers (elemental fingerprinting, Génio et al. 2015). Such approaches contribute crucial information to overcome the lack of knowledge on ontogenetic vertical trajectories of larval dispersal – a persistent and perhaps the greatest barrier to understanding connectivity.

Discussions following the keynote talk focused on the various caveats of existing sampling equipment. Fabio De Leo questioned the reliability of sampling to accurately quantify densities of larvae, knowing that tools can be selective. It was acknowledged that the best tools to achieve really large sampling volumes currently are MOCNESS nets as well as the SyPRID system mounted on the Sentry AUV. However, MOCNESS nets are semi-quantitative, because mesh size may be too large to collect some taxa and estimates of sample location and volume are not always accurate. Stace Beaulieu and Dhugal Lindsay suggested some adaptations (fine mesh sizes and longer nets or double-net systems to separate nekton and plankton), but Craig Young argued that such adaptations may bring other operational issues. Anna Metaxas mentioned that data and results obtained by the SyPRID system need to published in order to understand its effectiveness in different environments and for different purposes. It was recognized that comparisons between sampling tools are difficult, and that results are better compared in terms of diversity patterns than densities.

Strategies to overcome the limitations set by the time-consuming process of sorting and identifying larvae were also debated. Jim Birch asked how humanpower could be replaced or aided by an automated system in this process. The consensual perspective of the participants about this was that automated solutions, although desirable, are not achievable in the near future mainly because current equipment such as the Continuous Plankton Recorder (CPR) and other video plankton recorders do not have sufficient resolution, and reference databases to assist both genetic and morphological identification are lacking. Stace Beaulieu and Dhugal Lindsay reminded participants of some existing platforms (e.g. EcoTaxa, Picheral et al. 2017), which are currently cataloguing predominantly holoplankton and could be expanded for meroplankton. In any case, the need to reinforce taxonomy training of young scientists is undeniable.

Florence Pradillon presented the summary of the pre-workshop responses to the questions concerning Theme 1.

Question 1.1. Why do we need to quantify deep-sea larval diversity and distributions?

Data on larval diversity and distribution can provide insights about the reproductive traits of some species, such as reproductive timing, seasonality, fecundity, and larval traits such as developmental mode, swimming ability, and buoyancy. These biological traits are important parameters for accurate modelling of larval trajectories, and thus for predictions of connectivity patterns. A better understanding of basic biological parameters will also allow for assessment of the effects of geographic and seascape settings, and provide more accurate predictions of species' dispersal ranges. Besides bridging gaps in the fundamental knowledge of the life history of deep-ocean species, this will help increase understanding of metacommunity dynamics and inform spatial planning strategies (e.g. MPA networks).

Deep-sea larval diversity and distribution can also inform patterns of larval supply and settlement potential. These parameters will help better evaluate the role of larvae in regulating diversity and distribution of deep-sea benthic fauna, increase mechanistic understanding of deep-sea benthic community dynamics and resilience, facilitate the development of realistic and spatially explicit population models, and provide informed scientific advice on the status of and threats to marine resources.

Larvae are not only a major determinant of benthic and pelagic faunal distribution through settlement; they may also have a role through trophic networks. Meroplankton may represent a yet-undocumented food source in the deep ocean, but also potentially a competitor for food in the planktonic community. However, the trophic contribution of propagules (gametes and larvae) to the carbon cycle is likely to be significant only on local patchy scales, since dilution effects probably reduce its relevance to large scales.

Lastly, the importance of quantifying and monitoring meroplankton in the deep ocean was recognized with regards to its likely sensitivity to global change. It is critical to understand how larvae in the deep ocean may respond to natural variation and multiple climate change stressors, and how this may influence the functional role of meroplankton. Larval distributions could even be used as biological indicators of climate-driven influence on deeper water layers, although the current lack of knowledge on deep-ocean larvae makes this a distant prospect.

Question 1.2. What are the key challenges and limitations?

The most relevant limitations identified by the participants can be grouped into the following categories: ecological and biological knowledge, reference databases and taxonomical expertise, observation design and technological limitations.

Ecological and biological knowledge

One of the major challenges to gain knowledge of larval diversity and distribution in the deep ocean is in their in situ observation and collection. Deep-sea larvae exhibit very low abundance in general, even near ‘biomass hotspots” such as hydrothermal vents (dilution factor - densities in the deep ocean are 100 to 10000 times lower than in coastal systems; Table 1). Tools allowing the screening of very large volumes of waters are required (in the order of tens to thousands of m³). Developing such tools is costly, in terms of energy, operational time…and few of them exist.

Table 1. Download as CSV 

Some examples of meroplankton abundance at different deep-sea environment.

Abundance (ind/m3)	Setting	Depth (m)	Instrument (mesh µm)	References
0.03-0.3	Abyssal plain, BBL (CCZ)	4050	Pump (63)	Kersten et al. 2017
0.25-18	Oceanic ridge axis near hydrothermal vents (Pacific)	2500	Pump (63)	Beaulieu et al. 2009, Mullineaux et al. 2005, Mullineaux et al. 2013
>1	Oceanic ridge axis >1 km from hydrothermal vents (Pacific)	2500	Pump (63)	Mullineaux et al. 2013
0.01-9	Seamounts, (western Pacific)	350-1600	Net (63)	Metaxas 2011
Up to 1000	Fjord (Arctic)	0-300	Net (180)	Michelsen et al. 2017

Describing larval diversity and distribution is not sufficient to assess dispersal trajectories and connectivity patterns. Sources of dispersing larvae, as well as the identification of settlement and recruitment mechanisms, are also needed to understand how benthic communities are established and vary temporally. The use of tracers such as elemental fingerprinting of larval shells may help identifying the source of some larvae, but these tools first require mapping the signature of potential sites of origin.

Another challenge is the lack of biological information on deep-sea species' life cycles (reproductive timing, larval biology and ecology). Ex situ experimentation with larval stages could provide information on larval physiology, behaviour and biology (e.g. Larsson et al. 2014, Strömberg and Larsson 2017), but remains an extremely challenging task due to the difficulties in culturing deep-sea animals (Jiang et al. 2016). In situ, large, and multi-scale experiments may be conducted but require the development of a coherent, coordinated and comprehensive network of observation stations/platforms.

Reference databases and taxonomical expertise

A major limitation to our knowledge of meroplankton is the identification of deep-sea larvae, most of which remain undescribed with no recognition of their adult counterpart. The use of genetic barcoding is promising as an accurate tool to link adult forms with their respective larval forms, but depends heavily on the availability of reference data which are still scarce for the deep sea (some data are available through disparate databases: e.g., BOLD, WORMS, EOL, OBIS). Although relevant work has been carried out on some taxa from specific deep-ocean environments (e.g. Mills et al. 2007), there is currently no global taxonomic database dedicated to deep-sea larval identification, and a large effort in sequencing and imaging remains necessary to develop such a platform. New imaging technologies (3D holographic images, fluorescence; Colin et al. 2017) are being developed and would allow automated (potentially in situ) recognition of larvae using supervised machine-learning methods. However, such tools require prior learning with annotated training sets and the development of computer-based classification tools for high-throughput identification.

Observation design and technological limitations

From an operational point of view, developing our knowledge on deep-sea meroplankton diversity and distribution would require global engagement among marine observing systems; integration of the deep ocean with the coastal component of the marine observing systems covering physical, chemical and biological variables; seascape representativeness, and observation of key locations, with criteria based on seascape variability, biodiversity hotspots, MPAs vs non-protected areas, major environmental boundaries, larval gateways, etc.

Technological development required for meroplankton studies in situ faces many challenges related to the remoteness and hostility of the environment. There is difficulty in maintaining equipment, sustaining, and implementing an evolving observation system. Challenges include longevity of equipment, pressure resistance, power communication, timing of sample collection, miniaturization, corrosion, drift detection, deployment, communication, and recovery costs. Sampling bias due to larval avoidance and selectivity must also be addressed when developing sampling tools /strategies.

Solutions to the technological challenges exist (as proved by developments presented in theme 2), although adaptations are still needed regarding sampling size. The main scientific questions must be defined to drive technological development. For example, many questions identified above need time-series data and could push for the development of carrousel-type multi-samplers. Nevertheless, effective sampling strategies and larval tracking will not be possible until we have a sense of timing and direction of larval release. Focusing research first at sites where a large amount of baseline reproductive data is available thus may be a good strategy.

Marina Cunha presented the synthesis of the pre-workshop responses to three questions concerning Theme 2.

Question 2.1. What research or new technologies do we need for deep-sea meroplankton observations?

The answers to this question were structured around the main research questions in meroplankton studies, the methodological approaches to tackle these questions, and their expected outputs Table 2. The need for accurate larval identification (integrated taxonomy), building up genetic and image reference libraries, mapping distributions (from fine to large scale) and combining ecological and genetic data to understand life histories was consensual among the workshop participants.

Table 2. Download as CSV 

Main methodological approaches for meroplankton research.

Approaches	Data and outputs	Research questions
Metabarcoding, eDNA/eRNA	Genetic databases	Taxonomy, biodiversity, distribution, biogeography
Gene expression		Larval development
Laser Capture Microdissection combined with molecular analysis	Larval gut contents	Trophic ecology
Synchrotron, micro-CT	Internal anatomy and physiology	Ontogeny, larval development
Elemental fingerprinting, stable isotope analyses	Individual biogeochemical signatures	Ontogenic migration, dispersal trajectories, source-sink dynamics
In situ and ex situ imaging	Image databases	Taxonomy, biodiversity
In situ imaging combined with automated identification	Faunal composition datasets for a given water mass	Taxonomy, biodiversity, distribution, biogeography
In situ imaging combined with environmental data	Ecological databases	Fine-scale distribution, ontogenic migration

New methodologies are being increasingly used to gain more knowledge on trophic ecology and internal structure of larvae. A combination of Laser Capture Microdissection (LCM) and DNA metabarcoding was successfuly applied to analyse the gut contents of wild-collected cephalopod paralarvae (Fernández-Álvarez et al. 2018). Synchrotron- and micro-CT can be used to observe internal structural (physiological or functional) change during ontogenic development (Chen et al. 2018). These methodologies become more powerful when combined with genetic (e.g. gene expression) and ecological data (e.g. larval, juvenile, adult distributions). Information on life-history traits is crucial to provide adequate data for modelling and advance our knowledge of connectivity.

Other concerns raised during the discussion included insufficient genetic resolution and the need to gain more accurate knowledge on larval sources. Also debated was the extent to which taxonomy is currently a limitation for meroplankton research and related topics, such as the usefulness of taxonomical vs functional information (e.g. ecosystem function, response to stressors), and the large-scale sequencing data quality issue.

Obtaining more larval samples and improving the spatial and temporal coverage of sampling is indisputably the fundamental step for further advancing meroplankton research. But quantification is definitively a major challenge. Lombard et al. 2019 carried out an extensive review on the technologies available to make globally quantitative ocean observations of particles in general and plankton in particular, detailing a variety of relevant technologies and measurements, including bulk measurements of plankton, pigment composition, uses of genomic, optical and acoustical methods, as well as analysis using particle counters, flow cytometers and quantitative imaging devices. However, there are many deep-ocean meroplankton-specific constraints that remain to be resolved.

The main technological challenges identified by the workshop participants in relation to sampling and sampler type are summarised in Table 3. The debate was focused around bulk water samples and individual organism samples and continued around some of the topics already addressed in the discussion following the keynote talk, such as automated plankton imaging and identification, operational autonomy and versatility of samplers, duration vs resolution of time-series. The biggest challenge is imposed by the dilution factor and the subsequent large volume needed for obtaining representative samples of deep-ocean meroplankton. Morane Clavel and Jim Birch suggested that one way to mitigate the dilution factor, while waiting for future technological breakthroughs, is to identify physical ‘pinch-points’ for sampling where larvae are moving through a physically constrained space (e.g. straits, oceanographic fronts).

Table 3. Download as CSV 

Main technological challenges associated with sampling for meroplankton research.

Sampler type	Technologies	Technical challenges
Bulk water samplers	general samplers (nets, pumps)	increase adaptability and versatility for operation with autonomous, land-or vessel-based control, or mounted on different underwater vehicles (ROV, manned submersibles, gliders); resolve spatial vs temporal resolution and coverage
	large volume pumps for sequential, long term sampling	resolve volume vs mesh size and long term vs temporal resolution constraints; improve specimen preservation
	autonomous sampling gear with geolocation, controlled speed and water flowmeter	control geolocation (both vertical and horizontal); adjust filtering speed to the size of target organisms
	multitraps with open-closure system	control/monitor from land
Sensors	ecogenomic sensors (automated, miniaturized in situ environmental nucleic acids samplers)	increase depth range, sample volume, and sampling duration; decrease cost; resolve lack of reference databases to feed supervised machine learning methodologies
Individual-based methodologies	in situ imaging	increase sample volume; resolve in situ optic constraints; enable simultaneous larval collection for sequencing
	automated identification	resolve lack of reference databases to feed supervised machine learning methodologies
	larvae-specific tools (e.g. polarization of mollusc shells, elemental fingerprinting, stable isotopes)	resolve lack of geolocated reference databases on target elements; improve resolution of source populations (refernce datasets with larva with known origin/trajectories

Scientists need to establish the minimum requirements for enabling sampling gear to effectively answer their research questions. This is an important step to overcome the technological challenges of meroplankton sampling - as Jim Birch stressed, communication between scientists and engineers must be improved so that engineers can comprehend the scientific goals and design the necessary fit-for-purpose equipment.

Question 2.2. What do we expect from an observing system for deep-sea meroplankton?

An observing system for deep-ocean meroplankton is expected to improve the temporal and spatial coverage and resolution of sampling. In the pelagic ecosystem, characterized by its high variability at various spatial and temporal scales (e.g. Game et al. 2009), the dynamics of meroplankton cannot be adequately monitored, or even captured, by isolated or opportunistic sampling efforts. The observing system should be sustained, i.e. provide long-term time series (10 years or more), and of high temporal resolution (ideally one week or less). The spatial resolution is most likely constrained by available funds, and therefore continuous fine-scale and high resolution data should target specific habitats and ecosystems (biodiversity hotspots) or focus on specific processes, such as source-sink dynamics. The observing system should provide not only qualitative but also quantitative data, which are the basis of the mechanistic understanding of ecological processes, enable the parameterization, validation, and predictive capacity of modelling efforts, and are also required for effective adaptive management of ecosystems. The detection of larval vertical distributions was considered a priority for which large-volume gear, such as precision pumps (SyPRID) and vertical multinet plankton samplers (e.g. VMPS-6000D with frame opening area of 1m² and 8 nets of 100 μm mesh size), which must be optimised.

The workshop participants established that the observing system must ideally have the capability to deliver relevant data for the following research fields:

• Biodiversity: quantitative data on community composition, abundance and biomass;
• Source-sink dynamics: data on fecundity, timing and synchronicity for reproductive output assessments and data on larval inputs and post-metamorphic processes for net recruitment assessments;
• Individual-based traits: size, life-history stage, and genetic data supported by accurate taxonomic identifications;
• Bio-physical modelling: environmental and biological Essential Ocean Variables.

It is foreseeable that the development of low-cost, automated, and miniaturized in situ environmental nucleic acid (eDNA/RNA) samplers (Gan et al. 2017), together with advances in metagenomics (Ruppert et al. 2019, Sinniger et al. 2016), new approaches to the processing of molecular data (Stat et al. 2017), and the trivialization of supervised machine-learning methodologies (Cordier et al. 2017) will be able to deliver large volumes of molecular data. Marina Cunha commented that while accurate taxonomic identifications will be always irreplaceable when investigating life-history traits and population processes, our traditional approach to biodiversity studies may have to change to deal with the expected volume of data, especially with regards to undescribed species. We may need to move away from a static approach to biodiversity based on the current “biological” species concept towards a more flexible approach based on evolutionary lineages and how they change over time. Rob Young added that increasing genetic data outputs will allow streamlining taxonomic (lineage) composition, detecting trends and more easily identifying changes over time in a semi-automated way. Functionality can also be used as an add-on, to identify whether or not a change in species composition is paralleled by a change in ecosystem functioning (or vice versa). Other discussion topics related to genetic data outputs involved several participants and were summarized by Rob Young as: i) data quality issues associated with sequencing choices; ii) the need to develop multi-gene methodologies to overcome the insufficient resolution of genetic data for tracking larval origins and for connectivity studies; and iii) the advantages of using RNA vs DNA, with RNA being a better proxy for the presence of living organisms and also more effective in revealing epigenetic changes and physiological responses to stress.

Several workshop participants raised the point that molecular identifications of deep-sea larvae are presently hindered by the limited amount of data on adult molecular sequences. Species-level identifications of larvae necessitate large reference databases of sequences from described species in order to match larvae to adults. Basic research to build up public databases would significantly enhance the utility of deep-sea larval observations. In the meantime, larvae can be identified morphologically. Many Classes and Orders of invertebrates have distinct larval forms (see Young et al. 2001), so larvae can be matched to these higher taxa.

Question 2.3. How can we develop an observing system for deep-ocean meroplankton?

The participants agreed on the following important steps for developing a sustained observation system for deep-ocean meroplankton:

• Identify priority scientific questions for meroplankton research and define a strategy with long-term goals but short-term actions that can be implemented relatively quickly.
• Communicate the scientific and societal value for sustained observations of meroplankton to secure support from potential stakeholders and funding entities. The involvement of stakeholders working across multiple sectors is fundamental for the successful implementation and continuity of integrated ocean observing systems. Mackenzie et al. (2019) identified convergence on common goals, effective communication, co-production of information and knowledge, and the need for innovation as the most important orverarching principles for stakeholder engagement.
• Develop autonomous robotic assets and other innovative meroplankton-specific sampling and observation technologies. For this, it is crucial to improve interdisciplinary collaboration and communication between scientists and engineers. Networked robotic systems for adaptive sampling (Zhang et al. 2019) and fixed observatories for continuous monitoring of the water column and seafloor (Aguzzi et al. 2019) are equipped with interoperable cutting-edge techonology and rely on technological innovation for the constant enhancement of their performance. 
• Establish synergistic collaborations with active ocean observation initiatives and infrastructures at national and international levels. For a successful implementation of sustained meroplankton observations, it is imperative to foster partnerships, converge on common goals, and gain access to the distributed infrastructures and a range on in situ elements under the auspices of the Global Ocean Observing System (GOOS; IOC 2019, Weller et al. 2019) and more specifically to their regional (e,g, EOOS) and deep ocean (DOOS) counterparts.

Two different but complementary schemes for sustained observing systems were envisaged: i) a global approach with a large spatial coverage network using simple, low-cost sampling methodologies (Fig. 2); ii) a more specialised approach at regional scales using more complex observing and sampling assets with high spatial and temporal resolutions (Fig. 3).

Figure 2.  

Sustained meroplankton observations - a basin-scale approach. Low-cost samplers for deep-ocean observations (colonization modules coupled with larval traps) deployed at moorings or mounted on landers of the EMSO-ERIC distributed observatories and covering different water masses (Project LO3CATED; Génio, Cunha and Young). NACW: North Atlantic Central Water; AIW: Antarctic Intermediate Water; MOW: Mediterranean Outflow Water; NADW: North Atlantic Deep Water; MABW: Modified Antarctic Bottom Water.

Figure 3.  

Sustained meroplankton observation systems. envisioned approach at regional scale. 

Morane Clavel presented the summary of the pre-workshop responses to the questions concerning Theme 3.

Question 3.1. What methodologies or approaches do we need for data integration?

Comprehensive ocean observing systems must be interoperable to enable studies across different science domains and observing regimes (Benedetti-Cecchi et al. 2018). However, the multi-dimensional nature of biological and ecological data results in multiple and heterogenous approaches for quantitative measurements and sampling techniques, which create difficulties for data management and interoperability. Currently, there is a pressing need to overcome these limitations and improve data dissemination according to the best practices and standards (Pearlman et al. 2019). Data integration requires that datasets share the same type of data, comply with well-established and sustainable file formats, and use common vocabulary and well-documented information. The outputs must ultimately address the needs of intermediate- and end-users (e.g., scientists, modellers, managers) and support advice to policy-makers. Moreover, data should be easily understandable by humans and interpretable by machines in compliance with the FAIR (Findable, Accessible, Interoperable, Reusable) data principles (Wilkinson et al. 2016).

The first step for data integration is probably the agreement on the critical variables to focus on. GOOS adopted the Framework for Ocean Observing (FOO, Task Team for the Integrated Framework for Sustained Ocean Observing 2012) as a foundational document setting the common guidelines that comprise elements from Inputs (e.g., essential ocean variables - EOVs) to Processes (observations and maintenance), to Outputs (data and products) (Tanhua et al. 2019). Eight biological EOVs were prioritized by GOOS, with two others emerging. More recently, the Biodiversity Observation Network from the Group on Earth Observations (GEO BON) identified 22 Essential Biodiversity Variables candidates organized in six classes (Benedetti-Cecchi et al. 2018). DOOS also recommended three biological EOVs (including "connectivity of species") that are relevant for the deep ocean (Levin et al. 2019). Although these sets of candidate variables aim to account for the complexity of marine ecosystems, including some relevant aspects for meroplankton research, their feasibility and consideration for operational needs may be somehow limited.

In the context of meroplankton observations, the workshop participants emphasised the importance of standardizing sampling protocols, ensuring the consistency of types of data and file formats, and implementing quality control procedures. A recurrent issue - meroplankton identification - was widely debated. The use of flow imaging microscopy with particle analyzers (Álvarez et al. 2011) and other integrated analysis systems for acquisition and classification of digital zooplankton images (Gorsky et al. 2010) may enable a faster, semi-automatic processing and identification of meroplankton samples. However, these approaches rely extremely on the existing reference databases (e.g., EcoTaxa - Picheral et al. 2017, World Wide Web of Plankton Image Curation - wwwpic Team 2020) to facilitate the specimens’ classification and minimize errors. The use of the World Register of Marine Species (WoRMS, WoRMS Editorial Board 2020) was recommended as the most up-to-date reference database for scientific names.

Another requirement fundamental to data integration is the existence of publishing platforms and repositories, which improve the interoperability of data (connections to related information). Biogeography online repositories such as OBIS (OBIS 2020) or GBIF (GBIF 2020), provide templates and dedicated publication tools that facilitate the data submission and ensures the maintenance and preservation of data, in an openly and freely accessible manner. The Darwin Core data format, used by OBIS and GBIF, was also recommended as the ideal vehicle for storage of meroplankton data. The Darwin Core facilitates data integration by offering a standard form for proper cataloguing and sharing of biological and environmental data (Wieczorek et al. 2012), and is flexible enough to allow the development of data standards for specific sub-disciplines. Whenever possible, it is desirable to link the biological occurrences to associated environmental data and, in the future, also add information on individual life history traits. In order to maximize data dissemination and use, and besides being open and freely accessible, data portals should provide pre-visualization tools that digest the information available. Yet, the expected large volume of data and the continuous need for accessibility and maintenance require long-term funding and a strong investment in informatics expertise.

Question 3.2. Who are the potential end-users?

A facilitated access to data products is a fundamental step of the Framework for Ocean Observing. The ultimate objective is to provide added value to the observation data by processing reusable information (“measure once/use many times”) for a broader group of intermediate and end-users in research, management, and policy making. The availability of ocean observation data in online repositories fosters its use beyond the scope of the primary studies, enabling a broader analysis of ecological, environmental and biological research questions, supporting scientific discovery and addressing societal issues related to global change Achieving this goal is a strong justification for supporting sustained data collection efforts (Levin et al. 2019, Task Team for the Integrated Framework for Sustained Ocean Observing 2012).

A better knowledge of deep-sea meroplankton is beneficial for multiple profiles of intermediate and end-users. In the scientific field, the potential users of meroplankton data include biologists and ecological modellers aiming to address fundamental research questions including connectivity and resilience to climate change and other natural and anthropogenic disturbances. Modellers have a pivotal function among the end-users because they process the data into presentable and relevant outputs to non-scientific communities (i.e., managers, policy makers, industrial companies). Deep-ocean meroplankton data products may have a particular interest for environmental and regulatory intergovernmental bodies (e.g. OSPAR commission, International Seabed Authority), managers, and policy makers by contributing to environmental impact assessments and supporting marine spatial planning. Meroplankton data may also be used by stakeholders associated with educational and outreach programs.

Understanding the extent to which populations are connected by larval dispersal is critical for the development of strategies to sustain biodiversity and preserve deep-sea ecosystems. A growing commitment to protect the ocean’s biodiversity in both national waters and areas beyond national jurisdiction requires the development of concepts for the observation and valuation of marine biodiversity and ecosystem services and their integration into conservation efforts, as well as development of scientific and technical solutions relevant to environmental impact assessments in marine areas.

Anna Metaxas first presented the criteria of the Convention on Biological Diversity (CBD) for the selection of Marine Protected Areas (MPAs) and the design of Networks of Marine Protected Areas (MPAn). MPAn are increasingly being designed and implemented for deep-ocean ecosystems. Some of the criteria provided by CBD require an understanding of processes in early life history. However, while "maximising connectivity" is considered a critical design element for all MPAn (WCPA/IUCN 2007), it ranks very low in the scale of criteria most often used by managers (Balbar and Metaxas 2019). Its implementation to date has been extremely limited and hindered by inconsistencies in the terminology and concepts considered in management, which favours landscape connectivity metrics, compared to scientific research, which uses predominantly demographic and genetic connectivity estimates and modelling approaches (Balbar and Metaxas 2019). Various techniques (e.g. network theory, Treml et al. 2008) exist to then use these estimates for the development of metrics (e.g. betweenness centrality) that can used for the design of MPAn. Overall, the data needs for conservation are similar to those needed to address fundamental ecological questions, but the data outputs differ between the two.

The discussion that followed Anna Metaxas' talk was focused on effective ways of measuring connectivity among MPAs. Although connectivity is being proposed to integrate the biological set of EOVs (Levin et al. 2019), the types of measurements, scalability and temporal variation need to be specified. The limited feasibility and associated operational issues may be responsible for the difficulty in applying this CBD criterion, often discounted in favour of socioeconomic criteria when selecting the areas for protection or resource exploration. The added value of an observation system for meroplankton research capable of providing science-based evidence to support the design of MPA networks is therefore unquestionable.

The up-to-date map of the main in-situ components of the Global Ocean Observing System, easily obtained online, shows a variety of assets including drifters, Go-ships, buoys, moorings and other mobile and fixed platforms that cover significant areas of the oceans. However, for the deep ocean both in the seafloor and the water column, the number and distribution of such elements are dramatically fewer (Levin et al. 2019). Deep-ocean observatories with the capability to sample environmental and biological variables on a systematic and regular basis are the keystone for long-term monitoring ecosystem changes and environmental trends. There are currently few examples of infrastructures with such capabilities. Two of these infrastructures were represented in the workshop: the EMSO-Azores Observatory, a component of the EMSO-ERIC distributed observatory, and the Neptune Observatory from Ocean Networks Canada. The invited talks by their representatives Jozée Sarrazin and Fabio De Leo, respectively, provided a detailed overview of the infrastructures, the equipment and types of measurements and observations available, as well as the science focus, data products and data access policy (more details in Suppl. material 1). The contributions of the observatories to the knowledge of deep-ocean ecosystems were illustrated with various examples of scientific outputs and pathways for future collaborations in the field of meroplankton research were outlined.

Following the course of the discussions from previous sessions, it was agreed that taking advantage of the synergies resulting from the collaboration with existing ocean observation infrastructures is the way forward and will support the vision of a sustained system for meroplankton observations. Fixed observatories are crucial elements of the envisaged regional approaches for obtaining high spatial-temporal resolution and long-term temporal datasets. They are also visited regularly for maintenance purposes, providing cruises of opportunity for sampling at the site or during transit to the observatory nodes. Both talks raised the interest of the workshop participants in relation to funding and visiting opportunities and the possibility of applying for add-on experimentation (e.g. larval traps, recruitment plates, colonization modules).