Are zooplankton the key to climate models?

The Ocean carbon sink

At our current rate, about 10 PgC are released into the atmosphere every year, but only about 4 PgC actually stick around¹. The rest is quickly absorbed by either the terrestrial biosphere or the ocean. While plants on land currently uptake slightly more carbon and exhibit substantially more inter-annual variability than the ocean, the ocean sink controls climate on time scales of centuries to millenia. 

This is because the ocean holds ~60 times more carbon than the atmosphere and ~20 times more than plants and soils on land. Thus, even small changes in the ocean’s carbon reservoir can lead to big changes in atmosphere’s temperature, and because the timescales for ocean circulation are 1000s of years, those changes last a long time. But what controls the size of the ocean carbon sink?

If the ocean was a well-mixed box, the preindustrial atmosphere would have held roughly twice as much carbon², making it up to 5°C hotter. CITE. However, because there is a vertical gradient in dissolved inorganic carbon (DIC), the partial pressure of CO₂ at the surface is less than at depth. This allows the ocean to uptake more CO₂ from the atmosphere and sequester it at depth, keeping our climate cool and hospitable. 

Two processes are responsible for maintaining the DIC gradient: the solubility pump and the biological pump. The solubility pump is a physical mechanism driven by the fact that cold water preferentially holds more carbon (via higher solubility) and typically advects deeper (via deep water formation).  The biological pump is an ecological mechanism by with phytoplankton transport DIC to depth by consuming CO₂ in the surface, sinking, and decomposing at depth. The biological pump is thought to increase the surface-depth DIC gradient by about twice as much as the solubility pump and account for ~10% of the total ocean DIC inventory². 

Assuming that the mass of the ocean and atmosphere don't change, this means that the marine biological pump is the predominate control on changes to climate over 100-1000 timescales. This is consistent with extensive paleo-oceaographic records and preempted  John Martin's infamous quote: "Give me a half tanker of iron, and I will give you an ice age." 

Thus, in order to understand medium to long-term future climate projections, not to mention the durability of any biological-based marine carbon dioxide removal, we must be able to accurately model the marine biological carbon cycle and associated biological pump. 

Right for different reasons: Grazing pressure is large source of uncertainty

Unfortunately, even though most Earth system model (ESMs) can admirably recreate the present-day ocean, they can’t agree on the magnitude or direction (!) of future changes to net primary production and export production (the biological pump) under projected emissions scenarios. This, of course, is because the marine biogeochemical (BGC) models used in ESMs are notoriously over-parameterized and under-constrained⁴. That is, they can be tuned to match past-day observation, but can do so for different reasons.

In turn, once perturbed, different mechanisms will lead to different outcomes.

To help diagnoses which mechanisms are the least well-understood, or at least which ones exhibited the most variability between prominent models, we compared  14 major components in the marine carbon cycle, ranging from bulk carbon fluxes, to standing stocks, to specific rates terms.  We quantified the inter-model variability in each component and found the largest source of uncertainty across models was grazing pressure, or in other words the phytoplankton specific loss rate to zooplankton grazing. 

Specific grazing rates drive the uncertainty in grazing pressure

From a phytoplankton perspective grazing pressure is determine from two components: the amount of zooplankton and the specific speed at which they graze. Thus, a large population of slow grazing mesozooplankton could exert the same grazing pressure and a small population of rapidly-grazing microzooplankton or large filter-feeders. 

We found that there was much more uncertainty in the specific grazing rate than in the size of the zooplankton population and or in the rate at which zooplankton die.  Thus, the problem - or at least largest disagreement between BGC models - appears to be less about the much maligned closure term (i.e. zooplankton mortality) and more about the grazing formulation that prescribes how zooplankton specific grazing rates respond to changes in the type and abundance of prey. 

Variability in the grazing formulation is dominated it's size more than its shape

Within the grazing formulation, the structure of the food web, degree of trophic complexity, sensitivity to temperature, capacity for prey refuge and active vs. passive switching can all exert a strong top-down control on NPP and subsequent carbon cycling. While these qualitative differences are clearly important, less attention has been paid to large differences in the 'magnitude' of the grazing formulation.

This may be because differences in the structure of the food web, value of parameters, shape of the functional response, and determination of prey preferences collectively exert a multivariate often competing, influence on emergent grazing dynamics, making it difficult to discern how fast zooplankton, in a general sense, are assumed to graze by the full set of equations.

To clarify the combined influence of the multivariate response, we introduced a new diagnostic metric, the Prescribed Grazing Index (PGI), to approximate the magnitude of grazing rates prescribed by the innate properties of a given grazing formulation. The PGI is defined as the specific rate at which zooplankton would graze on the standardised, global-median, observed plankton community.

We found that the PGI varies hugely, increasing 70-fold(!) from the slowest to faster models, and largely explains inter-model differences in emergent grazing dynamics at global, zonal, and seasonal scales.  In turn, models with substantial qualitative differences in their grazing formulation and food-web structure but similar PGIs can yield similar zonal distributions of grazing pressure. 

This highlights how far CMIP6 models remain from converging on a first-order magnitude of zooplankton-specific grazing rates, leading to very large uncertainty in the representation of grazing pressure throughout the global ocean.

Influence of uncertainty in grazing pressure

In short, large differences in the PGI, and thus grazing pressure, appear to be tuned out in terms of historical NPP primarily via a faster nutrient uptake, but models with similar NPP can have very different standing stocks of phytoplankton, zooplankton, total biomass and zooplankton to phytoplankton ratios. 

We ran a sensitivity experiment in global ESM to determine the effect of simulating the same NPP, but with either fast or slow turnover of the phytoplankton population. Despite only increase the PGI by 5% of the full range used in IPCC models, we saw 58% more efficient carbon transfer up the food web and 35% more efficient carbon sequestration in the fast-turnover experiment. 

From a climate perspective, this translates to an extra 2 PgC yr⁻¹ sequestered, or double the maximum theoretical potential of Southern Ocean Iron Fertilization!  Thus, potentially unnoticed uncertainties in the parameter values describing grazing could dominate the entire signal produced by basin scale marine CO₂ removal.

Moving forward

References

1. ​​Friedlingstein, P. et al. Global Carbon Budget 2022. Earth Syst. Sci. Data. 14, 4811–4900 (2022).
2. DeVries, T. The Ocean Carbon Cycle. Annu. Rev. Environ. Resour. 47:317–41 (2022)
3. Sigman, D. and Boyle, E. Glacial/interglacial variations in atmospheric carbon dioxide. Nature. 407, 859–869 (2000) 
4. Ward, B. A. et al. When is a biogeochemical model too complex? Objective model reduction and selection for North Atlantic time-series sites. Prog. Oceanogr. 116, 49–65 (2013)
5. Rohr, T. Southern Ocean iron fertilization: an argument against commercialization but for continued research amidst lingering uncertainty. J. Sci. Policy Governance 15 (2019).