Opal and calcium carbonate are thought to regulate the biological pump’s transfer of organic carbon to the deep ocean. A global sediment trap database exhibits large regional variations in the organic carbon flux associated with opal flux. These variations are well-explained by upper ocean silica concentrations, with high opal \textquoteleft ballasting’ in the silica-deplete tropical Atlantic Ocean, and low ballasting in the silica-rich Southern Ocean. A plausible, testable hypothesis is that opal ballasting is due to mineral protection, and varies because diatoms grow thicker frustules where silica concentrations are higher, protecting less organic carbon per unit opal. These patterns do not emerge in an advanced ocean biogeochemical model when opal ballasting is represented using a single global parameterization for diatoms, indicating the need for additional parameterization of the dependence of diatoms traits on silica concentration to capture the links between elemental cycles and future changes in the biological pump.

The biological carbon pump is a key controller of how much carbon is stored within the global ocean. This pathway is influenced by food web interactions between zooplankton and their prey. In global biogeochemical models, Holling Type functional responses are frequently used to represent grazing interactions. How these responses are parameterised greatly influences biomass and subsequent carbon export estimates. The half-saturation constant, or k value, is central to the Holling functional response. Empirical studies show k can vary over three orders of magnitude, however, this variation is poorly represented in global models. This study derives zooplankton grazing dynamics from remote sensing products of phytoplankton biomass, resulting in global distribution maps of the grazing parameter k. The impact of these spatially varying k values on model skill and carbon export flux estimates is then considered. This study finds large spatial variation in k values across the global ocean, with distinct distributions for micro- and mesozooplankton. High half-saturation constants, which drive slower grazing, are generally associated with areas of high productivity. Grazing rate parameterisation is found to be critical in reproducing satellite-derived distributions of nanophytoplankton biomass, highlighting the importance of top-down drivers for this size class. Spatially varying grazing dynamics decrease mean total carbon export by >17% compared to globally homogeneous dynamics, with increases in faecal pellet export and decreases in export from algal aggregates. This study highlights the importance of grazing dynamics to both community structure and carbon export, with implications for modelling marine carbon sequestration under future climate scenarios.

Slow climate feedbacks are currently underrepresented in model assessments of climate sensitivity and their magnitudes are still poorly constrained. We combine a recently published record of Earth’s Energy Imbalance (EEI) with existing reconstructions of temperature, atmospheric composition, and sea level to estimate both the magnitude and timescale of the ice sheet-albedo feedback since the Last Glacial Maximum. This facilitates the first opportunity to quantify this feedback over the most recent deglaciation using a purely proxy data-driven approach, without the need for simulated reconstructions. We find the ice sheet-albedo feedback to be amplifying, reaching an equilibrium magnitude of 0.55 Wm-2K-1, with a 66% confidence interval of 0.45 - 0.63 Wm-2K-1. The timescale to equilibrium is estimated as 3.6Kyrs (66% confidence: 1.9 - 5.5Kyrs). These results provide new evidence for the timescale and magnitude of the amplifying ice sheet-albedo feedback that will continue to drive anthropogenic warming for millennia to come.

Climate feedbacks determine how much surface temperatures will eventually warm to balance anthropogenic radiative forcing, but remain difficult to constrain. The climate feedback due to some process X is defined as the partial derivative of outgoing radiation at the top of the atmosphere with respect to surface temperature following a change in X, λX=-∂Rout/TS|X, with total climate feedback a summation from all processes, λtotal=∑λX. Standard approaches evaluate climate feedbacks from finite temporal changes in surface temperatures and outgoing radiation, following observed or simulated perturbations to climate state. However, this introduces significant linear combination error (λtotal≠∑λX) when the applied perturbation is large enough to achieve a good signal-to-noise ratio. This study presents a new semi-empirical evaluation of non-cloud climate feedbacks, constrained instead by spatial variation in outgoing radiation and climate state. First, we observationally constrain functional relations for outgoing radiation over ocean and land in terms of surface temperature, pressure, relative humidity, the height of the tropopause, fractional clound amount and latitude. Then, these functional relations are differentiated with respect to surface temperature to calculate the climate feedbacks for infinitesimal perturbation, eliminating linear combination error at high signal-to-noise ratio. We find, when combined with a recent cloud feedback estimate, a present-day total climate feedback of -0.99 (-0.75 to -1.22 at 66% range) Wm-2K-1. Our method is independent of temporal variation approaches to evaluate climate feedback allowing Bayesian combination to further reduce uncertainty.

Marine dissolved organic carbon (DOC) is a major carbon reservoir influencing climate, but is poorly quantified. The lack of a comprehensive DOC climatology hinders model validation, estimation of the modern DOC inventory, and understanding of DOC’s role in the carbon cycle and climate. To address this problem, we used boosted regression trees to relate a compilation of DOC observations to different environmental climatologies, and extrapolated these inferred relationships to the entire ocean to compute annual layer-wise DOC climatologies with uncertainties. Prediction performance was satisfactory, with R² values within 0.6 - 0.8 for all layers and prediction error comparable to within-pixel measurement variability. DOC was mainly predicted by dissolved oxygen in the bathypelagic layer, and by nutrients in other layers. We estimate the total oceanic DOC inventory to be around 690 Pg C. Our results exemplify that machine learning is a powerful tool for constructing climatologies from limited observations.

Climate energy balance models (EBMs) – simple energy-balance-based models of climate change – are widely used. The simplest “linear” EBM is deficient in capturing the behavior of complex climate models, so a “two-layer” model with an additional degree of complexity, i.e. two vertical layers, is typically used. Other additional degrees of complexity are equally plausible as well, however, and different approaches to add a degree complexity have not been compared quantitatively. Here we compare four types of EBMs - two-layer, order-two (temperature-dependent feedback), two-region (in space), and two-timescale (fast and slow climate responses) - specifically, their ability to capture historical temperature change and simulated temperature changes in abrupt (4x) and gradual (1%-ramp) forcing scenarios. The two-region model outperforms the others. The two-region model’s best-fit parameters to historical temperatures are also more physically plausible than the next-best-fitting model, the two-layer model. We therefore conclude that the two-region model is the preferred climate EBM.

Products derived from remote sensing reflectances ($R_{rs}(\lambda)$), e.g. chlorophyll, phytoplankton carbon, euphotic depth, or particle size, are widely used in oceanography. Problematically, $R_{rs}(\lambda)$ may have fewer degrees of freedom (DoF) than measured wavebands or derived products. A global sea surface hyperspectral $R_{rs}(\lambda)$ dataset has DoF=4. MODIS-like multispectral equivalent data also have DoF=4, while their SeaWiFS equivalent has DoF=3. Both multispectral-equivalent datasets predict individual hyperspectral wavelengths’ $R_{rs}(\lambda)$ within nominal uncertainties. Remotely sensed climatological multispectral $R_{rs}(\lambda)$ have DoF=2, as information is lost by atmospheric correction, shifting to larger spatiotemporal scales, and/or more open-ocean measurements, but suites of $R_{rs}(\lambda)$-derived products have DoF=1. These results suggest that remote sensing products based on existing satellites’ $R_{rs}(\lambda)$ are not independent and should not be treated as such, that existing $R_{rs}(\lambda)$ measurements hold unutilized information, and that future multi- or especially hyper-spectral algorithms must rigorously consider correlations between $R_{rs}(\lambda)$ wavebands.

The shoreline development index – the ratio of a lake’s shore length to the circumference of a circle with the lake’s area – is a core metric of lake morphometry used in Earth and planetary sciences. In this paper, we demonstrate that the shoreline development index is scale-dependent and cannot be used to compare lakes with different areas. We show that large lakes will have higher shoreline development index measurements than smaller lakes of the same characteristic shape, even when mapped at the same scale. Specifically, the shoreline development index increases by about 14% for each doubling of lake area. These results call into question previously reported patterns of lake shape. We provide several suggestions to improve the application of this index, including a bias-corrected formulation for comparing lakes with different surface areas.

Turbulent mixing induced by breaking internal waves is key to the ocean circulation and global tracer budgets. While the classic marginal shear instability of Richardson number ∼ 1/4 has been considered as potentially relevant to turbulence wave breaking, its relevance to energetic zones where tides, winds, and buoyancy gradients excite non-linearly interacting processes has been suspect. We show that shear instability is indeed relevant in the ocean interior and propose an alternative generalized marginal stability criterion, based on the ratio of Ozmidov and Thorpe turbulence scales, which not only applies to the ocean interior, but remains relevant within turbulent boundary layers. This allows for accurate quantification of the transition from downwelling to upwelling zones in a recently emerged paradigm of ocean circulation. Our results help climate models more accurately calculate the mixing-driven deep ocean circulation and fluxes of tracers in the ocean interior.

It is well established that small scale cross-density (diapycnal) turbulent mixing induced by breaking of overturns in the interior of the ocean plays a significant role in sustaining the deep ocean circulation and in regulation of tracer budgets such as those of heat, carbon and nutrients. There has been significant progress in the fluid mechanical understanding of the physics of breaking internal waves. Connection of the microphysics of such turbulence to the larger scale dynamics, however, is significantly underdeveloped. We offer a hybrid theoretical-statistical approach, informed by observations, to make such a link. By doing so, we define a bulk flux coefficient, $\Gamma_B$, which represents the partitioning of energy available to an ‘ocean box’ (such as a grid cell of a coarse resolution climate model), from winds, tides, and other sources, into mixing and dissipation. $\Gamma_B$ depends on both the statistical distribution of turbulent patches and the flux coefficient associated with individual patches, $\Gamma_i$. We rely on recent parameterizations of ~$\Gamma_i$~ and the seeming universal characteristics of statistics of turbulent patches, to infer $\Gamma_B$, which is the essential quantity for representation of turbulent diffusivity in climate models. By applying our approach to climatology and global tidal estimates, we show that on a basin scale, energetic mixing zones exhibit moderately efficient mixing that induces significant vertical density fluxes, while quiet zones (with small background turbulence levels), although highly efficient in mixing, exhibit minimal vertical fluxes. The transition between the less energetic to more energetic zones marks regions of intense upwelling and downwelling of deep waters. We suggest that such upwelling and downwelling may be stronger than previously estimated, which in turn has direct implications for the closure of the deep branch of the ocean meridional overturning circulation as well as for the associated tracer budgets.

The ocean’s “biological pump” significantly modulates atmospheric carbon dioxide levels. However, the complexity and variability of processes involved introduces uncertainty in interpretation of transient observations and future climate projections. Much work has focused on “parametric uncertainty”, particularly determining the exponent(s) of a power-law relationship of sinking particle flux with depth. Varying this relationship’s functional form introduces additional “structural uncertainty”. We use an ocean biogeochemistry model substituting six alternative remineralization profiles fit to a reference power-law curve, to characterize structural uncertainty, which, in atmospheric pCO2 terms, is roughly 50% of the parametric uncertainty associated with varying the power-law exponent within its plausible global range, and similar to uncertainty associated with regional variation in power-law exponents. The substantial contribution of structural uncertainty to total uncertainty highlights the need to improve characterization of biological pump processes, and compare the performance of different profiles within Earth System Models to obtain better constrained climate projections.

Plankton play an important role in marine food webs, in biogeochemical cycling, and in Earth’s climate; yet observations are sparse, and predictions of how they might respond to climate change vary. Correlative species distribution models (SDM’s) have been applied to predicting biogeography based on relationships to observed environmental variables. To investigate sources of uncertainty, we use a correlative SDM to predict the plankton biogeography of a 21st Century marine ecosystem model (Darwin). Darwin output is sampled to mimic historical ocean observations, and the SDM is trained using generalised additive models. We find that predictive skill varies across test cases, and between functional groups, with errors that are more attributable to spatiotemporal sampling bias than sample size. End-of-century predictions are poor, limited by changes in target-predictor relationships over time. Our findings illustrate the fundamental challenges faced by empirical models in using limited observational data to predict complex, dynamic systems.