New paper out on how climate change will impact marine ecosystems and the ocean carbon sink

I’ve got a new paper out at EGU’s Earth System Dynamics this week, looking at the impact of climate change on the biological pump and ocean carbon sink, and in particular the role of ecological complexity in Earth system models in resolving non-linear climate-biosphere feedbacks. This is the first paper out of my recently-completed postdoc at Stockholm Resilience Centre on climate-biosphere feedbacks and tipping points, with a couple more to be submitted soon.

Following up on yesterday’s twitter thread explainer, here’s a blog version explaining what we did and why:

Pump down the Carbon

The oceans act as a massive carbon sink, taking up around a quarter of human CO₂ emissions so far. This CO₂ dissolves in to the surface ocean (making it more acidic in the process) before being transported to the deep ocean where it stays for hundreds of years (the “solubility pump”). Some of this dissolved CO₂ is used by photosynthesising plankton in the surface to make organic matter, which when they poo or die (or are eaten by zooplankton who then do likewise) produces “particulate organic carbon” (POC) which sinks through the ocean as “marine snow”.

As POC sinks it’s mostly consumed by microbes, who respire the organic matter and re-release the carbon & nutrients in dissolved form (known as “remineralisation”). The overall effect is transporting carbon & nutrients from surface to deep waters, i.e. the “biological pump”. Together the solubility and biological pumps transport carbon from the surface to deep ocean, allowing more CO₂ to dissolve in the surface and storing exported carbon in deep waters for hundreds of years before it is eventually mixed back to the surface and re-released.

Rising atmospheric CO₂ means more carbon is now dissolving in surface waters and being exported to deep waters, and so the ocean is acting as a net carbon sink (at least for a few hundred years before that deep water mixes back up).

Schematic illustrating the impact of warming on the soft tissue biological pump. On the left-side, under cooler preindustrial conditions cGEnIE’s surface layer remains fairly well mixed with the deep ocean (large green arrow from deep to surface layers), returning dissolved nutrients and carbon (DNut & DIC) from the remineralisation of exported POC (red arrow from POC to DIC & DNut), while some POC is remineralised partly within the surface layer. On the right-side, warming leads to a shift to dominance by smaller plankton as well as stratification leading to less mixing between the shallow and deep ocean, while shoaling of the remineralisation depth leads to greater recycling of nutrients and carbon close to the surface layer, combining to result in an overall reduction in POC export and sedimentation and an overall increase in the residence time of nutrients and carbon in the ocean
Schematic illustrating the impact of warming on the soft tissue biological pump. On the left-side, under cooler preindustrial conditions the surface layer remains fairly well mixed with the deep ocean (large green arrow from deep to surface layers), returning dissolved nutrients and carbon (DNut & DIC) from the remineralisation of exported POC (red arrow from POC to DIC & DNut), while some POC is remineralised partly within the surface layer. On the right-side, warming leads to a shift to dominance by smaller plankton as well as stratification leading to less mixing between the shallow and deep ocean, while the average remineralisation depth getting shallower leads to greater recycling of nutrients and carbon close to the surface layer, combining to result in an overall reduction in POC export.

But the ocean is also warming up, and this changes the strength of the two pumps. Warmer water holds less dissolved CO₂, limiting the solubility pump. Warming also makes it harder for surface waters to mix with colder deep waters, causing the ocean to stratify and less nutrients to be returned to the surface – conditions that favour smaller phytoplankton. Warming speeds up metabolic rates too, and as respiration increases faster than photosynthesis this means that sinking POC is remineralised – and the carbon and nutrients it contains released – closer to the surface, increasing nutrient recycling but also CO₂ in the surface ocean.

Overall this means we expect both the solubility and biological pumps to weaken with climate change, gradually reducing the capacity of the current ocean carbon sink and the negative climate feedback it provides. However, due to computational limits most Earth system models used to project the future ocean carbon sink don’t resolve key relevant ecological processes such as the effect of warming on remineralisation, plankton size shifts, or plankton adapting to lower nutrient availability.

Ask the eco-Genie

In this study we use ecoGEnIE, a recently developed version of a simpler Earth system model featuring both remineralisation that increases with temperature (“temperature-dependent remineralisation”) and multiple sizes of plankton that can use nutrients flexibly depending on availability (“trait-based ecology”). This allows the effects of ecological dynamics on the biological pump and ocean carbon sink in response to climate change to emerge.

We separate out these effects by turning on temperature-dependent remineralisation (TDR) and trait-based ecology (ECO) (instead of the default simpler FPR & BIO settings respectively) both separately and together, and running ecoGEnIE with future emission scenarios (based on the IPCC’s RCP scenarios, from low [RCP2.6], moderate [RCP4.5], high [RCP6.0], to very high [RCP8.5] emissions) until the year 2500.

Graph showing ecoGEnIE simulation results for global POC export flux under different configurations and forcing scenarios. Results for RCP4.5 and RCP8.5 are shown for each of the configurations (BIO+FPR, BIO+TDR, ECO+FPR, ECO+TDR), and the baseline POC export and the 21st century are marked by the horizontal and vertical dotted lines respectively. Under default BIO+FPR settings global POC export falls by ~7% by 2500 under RCP4.5 (~20% under RCP8.5); BIO+TDR instead leads to ~5% increase in POC export by 2500 under RCP4.5 (~22% under RCP8.5); ECO+FPR leads to ~9% less export by 2500 under RCP4.5 (~25% under RCP8.5); and ECO+TDR leads to ~3% less POC export by 2500 under RCP4.5 (and ~2% more under RCP8.5).
Graph showing ecoGEnIE simulation results for global POC export flux under different configurations and forcing scenarios. As time goes from left to right, going above the zero line means more POC is sinking from the surface ocean around the world, while going below the zero line means less POC is sinking from the surface ocean. Adding TDR (blue) leads to more sinking POC with warming than default (black), while adding ECO (yellow) leads to less sinking POC with warming (and adding both [pink] gives a smaller decline in sinking POC than default).

We find that turning on just temperature-dependent remineralisation (TDR) increases cumulative POC export relative to default runs (+∼1.3 %) as a result of increased nutrient recycling from remineralisation occurring closer to the surface with warming, whereas turning on just trait-based ecology (ECO) decreases cumulative POC export (−∼0.9 %) by enabling a shift to smaller plankton which produce less sinking POC.

EcoGEnIE POC export maps for default calibrations of BIO+FPR, BIO+TDR, ECO+FPR, and ECO+TDR, showing baseline export patterns (left) and the change in POC export by 2100 relative to the 1765 preindustrial baseline as a result of RCP4.5 (right). Baseline export is highest in high-latitude waters and along the equator in all configurations, but is higher in the ECO configurations and slightly lower in the TDR configurations. In BIO+FPR export declines in low and mid-latitude waters and increases in high-latitude waters by 2100; adding TDR reduces the low and mid-latitude decline, while adding ECO increases the decline in these areas. Plot created with Panoply, available from NASA Goddard Space Flight Center.
EcoGEnIE POC export maps for default calibration model runs, showing baseline export patterns (left) and the change in POC export by 2100 relative to the 1765 pre-industrial baseline as a result of RCP4.5 (right). Darker colours on the left indicate areas where more POC sinks from the surface ocean (i.e. a stronger biological pump). On the right, blue areas show where sinking POC decreases with warming, while red areas show where it increases. In general, adding TDR means a smaller decline in sinking POC in non-polar oceans, while adding ECO means a greater decline.

In contrast, interactions with complex surface carbonate chemistry and ocean acidification cause opposite responses for the ocean carbon sink in both cases: activating temperature-dependent remineralisation (TDR) leads to a smaller sink relative to default runs (−∼1.0 %), whereas activating trait-based ecology (ECO) leads to a larger relative sink (+∼0.2 %).

Graphs showing ecoGEnIE simulation results for the absolute cumulative ocean carbon sink and the cumulative ocean carbon sink relative to BIO+FPR under different configurations and forcing scenarios. Results for RCP4.5 and RCP8.5 are shown for each of the default calibration configurations (BIO+FPR, BIO+TDR, ECO+FPR, ECO+TDR). Adding TDR reduces the cumulative ocean carbon sink (-20 GtC by 2500 under RCP4.5, -40GtC under RCP8.5), adding ECO temporarily increases the sink (~0 GtC RCP4.5, +5GtC RCP8.5), and ECO+TDR results in an overall decrease in the sink (-15 GtC RCP4.5, -40 GtC RCP8.5).
Graphs showing ecoGEnIE simulation results for the absolute cumulative ocean carbon sink and the cumulative ocean carbon sink relative to BIO+FPR under different configurations and forcing scenarios. The bottom plot better shows the differences between the configurations – from the left-to-right, going above the zero-line (which represents the default model without the new features) means the ocean carbon sink is bigger in the new model configuration than the default configuration, while going below the zero-line means it’s smaller than the default configuration. In general adding ECO (yellow) leads to a bigger ocean carbon sink with warming, while adding TDR (blue) or combining ECO & TDR (pink) leads to a smaller sink.with warming

Down the sink

Combining both temperature-dependent remineralisation (TDR) and trait-based ecology (ECO) results in an overall strengthening of POC export (+∼0.1 %) and an overall reduction in the ocean carbon sink (−∼0.7 %) relative to default runs. Around 6 gigatonnes less carbon is taken up by the ocean in the 21st century as a result – a bit under 1 year of current human emissions.

This isn’t a huge difference, but is still more than current Earth system models project. There are also other important ecological processes not yet in the model (e.g. separated plankton shell types, ballasting, low resolution) that future work will need to look at to refine these estimates.

These results illustrate though the degree to which ecological dynamics & biodiversity modulate biological pump strength, and indicate that incorporating ecological complexity in Earth system models allows them to more fully resolve non-linear climate–biosphere feedbacks.

With thanks to co-authors Sarah Cornell, Katherine Richardson, and Johan Rockström, and the ERC-funded Earth Resilience in the Anthropocene (ERA) project for supporting this work during my postdoc at SRC!

New Paper Out: Reduced Carbon Cycle Resilience across the PETM

Much like buses, after a waiting a while for new papers to be published two have come along in short succession. This time though we’re back in the palaeoclimate domain, with a paper based on my work on a ReCoVER-funded Early Career Research award hosted at Ocean & Earth Science at Southampton which applied ‘early warning signal’ methodology to Cenozoic palaeoclimate records. It’s now available open access from Climates of the Past, and as with other papers I’ll summarise it here on my blog as well.

CP cover enlarged
My first EGU journals paper , and a pleasant public review process!

The setting of this paper is the Palaeocene-Eocene Thermal Maximum (i.e. the PETM), which is a natural case of carbon cycle disruption and linked rapid global warming that happened about 56 million years (My) ago. The triggers of this event are still being investigated, but palaeorecords point to the release of several thousand gigatonnes of carbon being released over a few thousand years driving ~5oC of global warming. As a comparison to today, this is a similar amount of carbon as humans are likely to emit from fossil fuel burning but over ~10 times the time, making it a partial but limited analogue to current climate change. The PETM was then followed by several smaller ‘hyperthermal’ events on a regular timescale into the warm Eocene.

As with many other big ancient climate shifts, the PETM was preceded by more gradual changes before a rapid shift, which has led many to hypothesise that it involved some sort of ‘tipping point’ (i.e. when gradual changes can eventually lead to a sudden shift in a system after reaching a critical threshold – see climatetippingpoints.info for more info!) that led to lots of carbon from parts of the Earth system like methane hydrates or peat being suddenly released. Alternatively, the PETM also coincided with a time of mass volcanism associated with the opening of the North Atlantic (of which Iceland is now the distant hangover of), and so could have been directly triggered by volcanic eruptions without any sort of tipping point involved.

Theory suggests that tipping points are often preceded by small but detectable ‘early warning signals’ (EWS), which can be found using statistical analysis of data. After an early proliferation of EWS techniques a few years ago though researchers have found them to have important limitations, with data quality being a big constraint and a propensity for false or missed alarms. Despite this, using multiple EWS indicators of different types along with strong statistical significance testing can still give us a pretty good idea of changes in a system’s overall resilience, with increasing variability and system ‘memory’ indicating the weakening of the system’s stabilising negative feedbacks and therefore a greater risk of being disturbed.

In our study we put this to the test by analysing some good quality long palaeorecords covering 5 My before the PETM and ~2 My after in order to look for any significant changes in carbon-climate system resilience that might help explain the origins of the PETM. We found consistent evidence from several different methods of a gradual destabilisation of the geological carbon cycle in the ~2 My before the PETM, and long-lasting carbon-climate system instability in the aftermath. This period coincides with the North Atlantic volcanism, leading us to suggest that these eruptions helped to gradually destabilise the carbon cycle by suppressing organic carbon burial (in particular either the marine biological pump or peat on land) as the result of volcanism-driven warming .

However, although this could mean the PETM itself was a tipping point resulting from this destabilisation it cannot solidly prove it, and we find no evidence of a tipping point in just the climate system either. Despite this, a decline in carbon cycle resilience would’ve still made it easier for the PETM to occur and last longer than it would’ve been otherwise, as weaker negative feedbacks would slow down the carbon cycle’s recovery to pre-PETM conditions. We also find evidence that the subsequent hyperthermal was preceded by slightly different dynamics than the PETM itself, which fits with the hypothesis that the PETM required an extra “push” from say volcanism but that the later events were more traditional tipping points.

To find out more, the full article is open access and free to read for all, and direct questions are welcome. Future follow-up work include a similar analysis of the Cretaceous/Palaeogene boundary and the Deccan Traps (paper TBC), and other Cenozoic climate shifts as more long and high-resolution records become available. Thanks also go to EPSRC/ReCoVER for funding the initial project, OES at Uni. Southampton for hosting the project back in Summer 2016, SRC where I did the final revisions/reanalyses, and Stockholm University for funding the open access publication.

New paper out on the extent of Sustainable Intensification in England

I’ve had a paper accepted over at Science of the Total Environment called “To what extent has sustainable intensification in England been achieved?” – click here for free access to the final published version (open until 18/19/2018), and after that you can find a free pdf of the accepted post-print manuscript over on my publications page. This is the final product of the first postdoc I took on at Southampton’s Geography department in 2015-16, so it’s good to see it finally out there! It’s also my first paper branching out into Sustainability Science & Socio-Ecological Systems from my Geological/Earth System roots. Here’s a blog summary:

The background to the paper is the drive for the Sustainable Intensification of Agriculture (SI) – the goal being to produce more food with a smaller environmental footprint, as agriculture remains one of the biggest global drivers environmental degradation while global population is projected to substantially grow by mid-century. This has been a hot topic over the last couple of decades in agricultural & environmental sciences, and there’s evidence for some progress in some places. However, the process of SI itself is not sufficient if the environmental footprint is still too big to be truly sustainable – we need to make sure that SI leads to global and regional agroecosystems reaching a ‘safe operating space‘ in which no ecosystem service is degraded beyond acceptable planetary/regional boundaries. In other words, making intensification a bit more sustainable or efficient isn’t enough if the environment that agriculture depends on is still being degraded in some way.

In this study we used publicly available data to determine the progress of SI in England and two sub-regions (mostly arable Eastern England & mostly pastoral South-Western England), using it as a case study of a rich, developed country with some signs of SI occurring. We use the data to produce metrics and indices of various important ecosystem services important for keeping the English agroecosystem in a safe operating space, such as river contamination by agricultural nutrients (water quality), atmospheric emissions (air quality), the farmland bird index (proxy for wider biodiversity), and soil erosion, as well as how much food is produced (wheat yield, and meat & dairy) and socio-economic context (e.g. farm income, subsidies, labour, etc.). [It’s worth noting here that even in a rich country there are some things we just couldn’t get enough data for, making this only a partial analysis]. From this we looked at their relative trends, performed some statistical analyses to explore these trends, and then built a simple system dynamics model of the agroecosystem and make some future projections.

Here’s a key summary figure of our main results (from the graphical abstract to the paper):

Armstrong McKay et al - Graphical Abstract
Graphical Abstract from Armstrong McKay et al. (2018) showing some key intensification trends in the English agroecosystem between 1950 and 2013.

The trends above (shown as z scores, i.e. relative changes rather than the absolute values) can be split into two main phases. Prior to the mid 1990s, Wheat Yield increases as a result of conventional Agricultural Intensification driven by increased fertiliser use, with livestock population and outputs also increasing. This results in greater food self-sufficiency (i.e. less food imported), but also leads to increased environmental degradation (the thick brown line) primarily due to pollution and falling farmland biodiversity (the dot-dashed green line). However, after the mid 1990s fertiliser use decouples from yield and starts to fall (while yield stays stable), which along with falling livestock population (mostly due to the 2001 foot-and-mouth disease outbreak) leads to a fall in pollution and therefore in overall degradation.

This seems like good news, and it partially is (especially for river-water and air quality, the biggest improvers), but this fall in overall degradation masks the fact that farmland biodiversity fails to recover at all. The parallel decline in self-sufficiency also means more food is being imported to the UK, and indicates that some of the UK’s agricultural degradation is simply being ‘offshored’ to other countries to deal with instead. These trends were further analysed and confirmed by the statistical analyses and modelling, and can also be plotted against GDP per capita to show a partial ‘Environmental Kuznets Curve’ (a reduction in degradation relative to yield beyond a certain level of prosperity – implying that prosperity leads to less degradation – but critically not for biodiversity, and could also be the result of using prosperity to pay others to degrade instead). Another interesting outcome from the analysis is the despite there being some evidence of ‘land sparing’ – i.e. intensification allowing the transfer of marginal agricultural land to conservation areas, here primarily from rough grazing areas – this has not stopped biodiversity decline so far despite being mooted as a key conservation strategy (see the paper for thoughts as to why).

Overall, these trends indicate that SI has indeed begun in England, but with major negative trade-offs on biodiversity and offshored degradation. These trade-offs undermine the general SI trend, and so must be dealt with before the English agroecosystem can reach a safe, or just, operating space. In future there’ll also be the challenges of climate change (which is likely to reduce yields and biodiversity overall) and post-Brexit subsidy changes (which might affect the financial viability of farming), which will require even more SI to counteract. What this future SI will look like is hard to say, but it’ll certainly require new approaches to subsidies and agri-environment schemes that better encourage both biodiversity restoration and reduced offshoring of degradation.

For more details, go read the paper and feel free to ask me questions!