There and back again, an organic carbon journey: mapping pathways and loops

Understanding and determining where organic carbon (OC) ends up in the ocean and how long it remains there is one of the most pressing tasks of our time, as the fate of OC in the ocean links to the climate system. To provide an additional tool to accomplish this and other related tasks, we map and conceptualize OC pathways in a qualitative model. The model is complementary to existing concepts of OC processes and pathways which are based mainly on quantifications and observations of current states and dominant processes. Our model, on the contrary, presents general pathway patterns and 5 embedded processes without focusing on dominant processes or pathways or omitting rare ones. By mapping, comparing, and condensing pathways and involved spatial scales, we define three remineralization and two recalcitrant dissolved organic carbon loops that close within the marine systems. Pathways that exit the marine system comprise inorganic atmospheric, OC atmospheric, and long-term sediment loops. With the defined loops and the embedded process options, the model is flexible and can be adapted to different systems, changing understanding or changing mechanisms. As such, it can help tracking pathway 10 changes and assessing the impact of human interventions on pathways, marine ecosystems, and the oceanic organic carbon cycle.

Optional path segments comprise processes that are not required to produce a certain outcome.
Sequence of critical path segments plus space: OC remineralization (SLS), DIC uptake by primary producers (SLS) A closed loop is a pathway pattern and is embedded in the OC cycle.

Closed loop: Surface remineralization loop
Open loop An open loop comprises all pathways that lead outside the marine system. Matter needs more time to reach the surface again and can escape remineralisers due to changing positions or its refractory or degraded character (Baker et al., 2017). In the upper sedimentary space (USS), remineralisers remineralize even highly 160 degraded matter as it remains in their vicinity longer than in the water column (Middelburg, 2019). The lower sedimentation space (LSS) is mostly abiotic and undisturbed and allows lithification processes. Users of the model can modify the spaces, e.g. by partitioning the water column space. However, any coastal system must be represented by at least two spaces (SLS and USS), and pelagic marine systems by at least three spaces (SLS, WCS and USS).
We now define five closed OC loops ( Figure 1 and Table 2) by their unique combinations of 1) their sequences of critical 165 path segments and 2) the involved spaces.
The first set of loops comprise three remineralization loops: a surface remineralization loop (SRL), a water column remineralization loop (WCRL), and an upper sediment remineralization loop (USRL) ( Table 2). All three loops include pathways on which OC is remineralized to DIC (path segment D), which is taken up by primary producers in the SLS (F). The path segments "OC position change" (A) and "DIC upward position change" (E) as well as the space in which the OC is remineralized 170 distinguish the remineralization loops. The WCRL includes pathways that lead to a downward position change of OC into the WCS, remineralization in the WCS, and an upward position change of DIC into the SLS, where it is taken up. An exemplary WCRL pathway involves OC uptake by zooplankton in the SLS, its migration into and respiration in the WCS, and the upward mixing of the resulting DIC into the SLS where it is taken up by primary producers. If zooplankton respiration occurs in the SLS, the pathway belongs to the SRL. We define the USRL analogous to the WCRL, but with remineralization taking place in 175 the USS.
The two path segments "Formation of rDOC" (B) and "rDOC conversion to more labile DOC" (C) in the SLS are part of the second set of closed loops, the rDOC loops ( Figure 1 and Table 2). The rDOC loops describe the change of labile OC to more recalcitrant forms, its persistence in the system, and its return to bioavailable forms in the SLS. We differentiate a short rDOC loop (SrDOCL), rDOC that accumulates in the surface waters on time scales of human life, and a long-term rDOC loop 180 (LrDOCL), rDOC that persists in the entire water column for magnitudes longer than human life scales. The short-term rDOC loop is defined by the "Formation of rDOC" (B) and "rDOC conversion to more labile DOC" (C) in the SLS, while the rDOC  long-term loop additionally comprises the path segment "OC position change" (A), with accumulation mostly or even entirely in the WCS (Figure 1). In contrast to the remineralization pathways, we do not explicitly consider a rDOC loop in the upper sediment, as the temporal scales of rDOC produced there or in the water column overlap to our knowledge. Therefore, the 185 long-term rDOC loop includes rDOC production in the USS alongside its transport to the WCS.
All loops include a continuum of processes that are optional and thus belong to non-critical path segments. Optional path segments do not change the outcome but can still alter the spatiotemporal scales of the pattern. For example, the SRL includes carbon that is remineralized and taken up by primary producers in the SLS. This definition ultimately means that OC, which is transported and processed below the SLS but returned to the SLS as OC to be remineralized and taken up, is also part of the 2.4 What process options do general pathway patterns comprise?
The general pathway patterns, e.g. closed loops, do not resolve processes or involved agents (organisms, OC species, etc.).
We add this information in this section and describe global process options embedded in each path segment ( Figure 2 and Table 3). This addition allows applying the model to smaller-scale ecosystems, relating existing concepts to our concept, 210 and demonstrates how to add different processes and agents to the model. Global in this context means that the process mechanisms are globally valid but that the frequency, extent, initialization, and agents driving these processes differ. We focus on non-anthropogenic processes and critical path segments.
Two of three remineralization loops include path segments related to the position change of OC or IC. Processes that belong to the path segments A and E are either biotic or abiotic and comprise sinking, diffusion, and advection as well as direct and 215 indirect biota-induced position change ( Figure 2 and Table 3).
Particulate matter that sinks from one space in the water column into another is mostly large or dense and/ or escapes consumption or dissolution in the upper space (De La Rocha, 2006). Burial by subsequent matter is the analogous process within the sediment-water interface and the sediment. Matter is buried and compacted by weight deposited above it and "sinks" as it loses volume. Sinking is always downward directed (Boyd et al., 2019) and restricted to POC. Gravitational induced 220 sinking (and burial) is thus part of each downward-pointing path segment A of POC ( Figure 2 and Table 3).
DOC and DIC can diffuse in every direction following large or small-scale gradients in the water column, water-sediment boundary layer, and pore-waters in the sediment. We use present-day decreasing DOC concentrations with water depth (Hansell, 2013), higher sedimentary DOC concentrations compared to overlying waters (Burdige et al., 1999), and higher DIC concentrations at the surface compared to the deep sea (Oka, 2020). Hence, DOC diffuses downwards in the water column and upwards 225 within and from the sediment since consumption cannot cope with production (Rowe and Deming, 2011) (path segments A, E in Figure 2). DIC diffuses upwards in A and E ( Figure 2).
Other physically induced position changes are related to water or sediment mass movements based on advection. These include large-scale upwelling and downwelling patterns, seasonal mixing, wind-induced turbulence and eddies, and storminduced resuspension. Advection is globally applicable although its direction, magnitude, and frequency vary. The advection- Indirect biota-induced position change comprises biogenic turbulence (Kunze et al., 2006;Huntley and Zhou, 2004), and induced drift, which describes the transport of substances that adhere to the bodies of swimming organisms (Katija and Dabiri, 2009). Indirect biota-induced position change in the sediment is related to inter alia bioturbation (Berke, 2010), associated with sediment reworking and resuspension, and bioirrigation (Kristensen et al., 2012), which leads to inflows of ocean water into the sediment. Indirect biota-induced position change works in all directions and is involved in all critical path segments A and 245 E for (r)DOC and POC in the water column and the sediment (Figure 2).
The next group of processes belongs to the path segment remineralization of OC (D). We define remineralization as the provision of DIC based on OC and restrict it to the spaces above the LSS, assuming that remineralization in the LSS is negligible.
Bacteria and archaea remineralize DOC in path segment D in every space above LSS (all path segments D of DOC, Figure   250 2), also under different oxygen conditions. The DOC is either of allochthonous origin (e.g. entering via riverine input (Dai et al., 2012)), or of autochthonous origin based on living or non-living POC. For instance, POC dissolves while sinking (Carlson and Hansell, 2015)), is fragmented by turbulence (Ruiz, 1997;Briggs et al., 2020), or photodissolved (Mayer et al., 2006). Consumers directly reduce the size of organic POC by sloppy feeding on living and non-living POC (e.g. zooplankton coprorhexy (Lampitt et al., 1990)), by producing small metabolites, and/ or by excreting DOC (Lampert, 1978). Indirectly, 255 consumers fragment non-living POC by swimming or moving (Dilling and Alldredge, 2000). Further, primary producer exudate DOC in the water column (e.g. under nutrient-limited conditions or through viral lysis (Azam and Malfatti, 2007)) and in the sediment (by macrophytes (Duarte and Cebrián, 1996)). Bacteria, for their part, hydrolyse POC to DOC (Smith et al., 1992) and additionally release DOC by viral lysis (Middelboe et al., 1996). The conversion from POC to DOC (arrows from POC to DOC, Figure 2) that occur before bacterial remineralization are optional path segments since not all OC needs to undergo one 260 of these changes to be remineralized.
Another form of remineralization is respiration by living organisms other than bacteria. Primary producers respire in the 265 photic SLS. Macrophytes additionally respire with their roots in the USS at night (Pedersen et al., 1995). Higher trophic levels, POC-consumers (e.g. zooplankton and fish) and non-bacterial DOC consumers (e.g. suspension-feeding sponges at the sediment-water interface (Wooster et al., 2019)), also remineralize by respiring. We, therefore, include remineralization by primary producers in path segments D in the SLS and USS and respiration by POC(DOC)-consumers in all spaces with aerobic conditions above the LSS (Figure 2).  Light-induced photoremineralization, the only physically induced remineralization, directly oxidizes DOC and POC to IC (Mopper and Kieber, 2002;Mayer et al., 2009) and works only in the SLS. We include this process into path segment D in the SLS.
Once OC is remineralized to DIC, this DIC is transported by the above-described processes of position change to the SLS (path segment E). Subsequently, primary producers take up the DIC for photosynthesis (path segment C) and close the 275 remineralization loops.
The rDOC loops include the formation of rDOC (B), the reconversion to DOC in the SLS (C), and, in case of the long-term loop, position change of OC (A). We present some of the involved abiotic and biotic processes, which have been reviewed e.g.
Biota supply rDOC via successive microbial processing, successive consumption by higher trophic levels (Jiao et al.,280 2010, 2011), the release of capsular material by bacteria (Stoderegger and Herndl, 1998), and other reactions of phytoplankton and bacteria described for the DOC production before. Processes of conversion of living and non-living POC to DOC, e.g. via dissolution, can result in highly diluted DOC that is not available for consumption and hence also recalcitrant ( 285 rDOC that stays in or returns to the SLS, via the position change processes described above (path segment A), can be converted back to more bioavailable forms by photodegradation/ photooxidation (path segment C in the SLS) (Kieber et al., 1989). We consider other removal processes, such as direct photooxidation from rDOC to DIC (Shen and Benner, 2018), sorption of rDOC into POC (Hansell et al., 2009), and hydrothermal removal mechanisms in hydrothermal vents or the Earth's crust (Lang et al., 2006), as optional path segments of either one of the closed remineralization or open loops. Once the rDOC 290 is converted in the SLS, the rDOC loops are closed.  Our qualitative concept of OC cycling is complementary to existing models of OC dynamics and processes in the ocean and resolves general pathway patterns and different process options without assessing the importance or rarity of these.
By not stopping at one process, such as respiration, but showing complete pathways to their end, we illustrate and emphasize the cycling nature of OC dynamics in the ocean. The decoupling of carbon pumping from return pathways in some previous concepts and their graphical representations seems to imply that increased transport of OC into the ocean interior always leads 300 to increased sequestration and storage of atmospheric carbon in the ocean. However, increased export of OC is not necessarily associated with increased carbon storage, which depends, among other things, on the ratio of regenerated and preformed nutrients and on the carbon that escapes the deep ocean (Gnanadesikan and Marinov, 2008). The export of carbon to the sediment and the deep oceans is part of the carbon processing, but not the whole story. Research on and the communication of the potential oceanic carbon sinks necessarily need to consider the return and exit pathways. Our model can help to communicate 305 and acknowledge the cycling nature of OC pathways.
Another add-on of our model is that we refrain from quantification or interpretation and do not indicate dominant pathways or omit rare ones, as our understanding of OC dynamics is constantly changing. Higher trophic levels were previously neglected and are now recognized as relevant to the carbon cycle. For example, large migratory species link to nutrient distribution and overall mixing (Roman and McCarthy, 2010), zooplankton significantly influence the carbon export (Steinberg and Landry,310 2017), and fishes and mammals contribute to the carbon cycle through various processes (Martin et al., 2021). In addition, some studies suggest that current models overestimate particle export by underestimating processes that lead to shifts in carbon pools, such as fragmentation (Baker et al., 2017). These ongoing new findings show how short-living and dynamic our understanding of processes and OC pathways is.
To account for these changes in understanding, our model provides a conceptual skeleton, an overarching concept that can be 315 brought to life by users. Processes, organisms, pathways, and loops can easily be added, changed, or deleted to accommodate new insights or specific systems. At the same time, existing concepts of OC cycling also fit into our model, e.g. the microbial carbon pump finds its reference in the rDOC loops and the lipid pump in pathways of the WCRL. Overall, this gives our model a high degree of flexibility.
This flexibility is necessary since not just our understanding but ocean systems around the globe are changing (Doney et al., 320 2012). Today's dominant pathway or process may not be dominant tomorrow. For example, the assumption that dense particles predominately sink (e.g. by Le Quere et al. (2005)) is only correct as long as their density-induced sinking is not prevented by default or to a large extent, e.g. by aggregation with microplastics that reduces density and thus sedimentation (Long et al., 2015). Concepts that implicitly or explicitly state that most dense particles sink and only or mainly consider pathways of OC sinking cannot resolve what might happen to the particle instead, for example, when microplastics reduce the sinking of these 325 particles.
In our model, density-induced sinking is only one option among many others. Dense particles can end up in different loops, e.g. in the SRL, staying in the SLS, or the WCRL if transported to deeper water layers by processes other than sinking. Besides these changes in sinking, there are numerous other examples of process changes that have the potential to alter OC cycle pathways, e.g. changes in the phytoplankton community (Vernet et al., 2017), changes in the DOC pool (Lønborg et al., 2020), 330 and depletion of higher trophic levels (Wilmers et al., 2012), to name a few. All of these changes might deconstruct and modify the pathways of particles. Therefore, the fate of a carbon particle in our model is open. And we present pathway options instead of a selection of dominant ones with a fixed destiny.
How OC moves through the system and where it ends affects the climate system (e.g. via alteration of the biological pump (Barange et al., 2017)), the ecosystem functioning (e.g. via changes in the benthic-pelagic coupling (Griffiths et al., 2017)) 335 but also human well-being and socio-ecological systems (via alterations of the trophic energy flows leading to less productive coastal systems (Ullah et al., 2018)) at all scales. Therefore, there is a societal need to understand today's systems and their possible changes due to anthropogenic interventions under different scenarios.
Our model can serve as a basis for such scenario considerations and assessments of interventions and management strategies in marine ecosystems. Deliberate interventions in OC pathways, like geoengineering, and interventions that incidentally change 340 OC pathways, e.g. fish management strategies, should both be tested for their impact on the different pathways in different systems. For illustration, changes in the food web through fisheries likely alter OC cycling. Our model helps to identify all pathways with involved fishes and invites us to ask: Do changes in fish stocks dry up carbon pathways? Are others taking over?
Are there changes in loops? Do these changes differ in different ecosystems? What are the implications for the food web and carbon storage?

345
Science and society need to find answers to these and many other questions to adapt system-tailored management options, assess the impacts of these options on OC pathways, and identify and evaluate nature-based solutions that are efficient and preserve ocean resources and health for future generations. While our model helps to address these questions and adjust our understanding of the mechanisms of the OC cycle, it needs complementation and extensions.
Quantifications of the magnitudes of carbon moving through different pathways in individual ecosystems and qualitative 350 modelling of scenario-based changes in loops and processes need to follow. In particular, to understand whether and how much pathways change due to anthropogenic interventions, our qualitative conceptual model is only a first step. In addition to these methodological next steps, advanced conceptual pathway models need to consider the interactions of alkalinity and solubility with the OC cycle to understand and predict how the OC cycle might change, also in terms of possible consequences for the climate system and the options and barriers that exist for society to alter carbon pathways and adapt to unintended or irreparable 355 changes. Marine ecosystems and the carbon cycle are changing. Just as important, our perception and approaches to address these changes are changing and must adapt even further. We offer a new conceptual model to accommodate these changes, structure them, and make them useful for further research.
Data availability. We will publish the base pathway model at PANGAEA during the review process. So far the material is attached as supplement.