Rationale, and continuing research beyond funding

Katherine

This paper is the final output (or swan song if you will) for a project at Cardiff University that started in 2016 – 7 years before the publication of this paper. In fact, we had been talking about this paper in 2018, whilst preparing the keynote work of that project, that looked to reconstruct the vertical gradient of δ¹³C (the stable isotope of carbon) in the upper 500m of the ocean water column, and reconstruct depth-habitats of multiple foraminifera species from 15 million years ago to the present. The fixation of dissolved carbon dioxide into organic carbon during photosynthesis preferentially uses up lighter carbon (¹²C) leaving the surface ocean enriched in the heavier carbon (¹³C). When this organic carbon is broken down as it sinks through the water column, it releases ¹²C back into the water column. Our paper showed that the ¹³C-depth gradient changed across the past 15 million years, consistent with the hypotheses that warmer temperatures in past climates sped up the degradation of the organic matter as it sank. This leading to less carbon being stored in the deep ocean in warmer climates. This has implications for us because this may happen as the climate warms leading to the oceans adding more CO₂ to the atmosphere creating a positive feedback on warming. However, we decided not to include future projections in the project's keynote work – deciding rather that a final additional paper could focus more specifically on the effect of ocean warming on future Twilight Zone life.

By the time work started on this paper, all of the four fixed-term researchers involved in this paper were working on different contracts than the original funding. I, lead author, was now employed on an entirely different project for Exeter University, which meant getting to the final draft was a slow process. Ultimately, this paper was worth the effort but it shows how post-doctoral funding through grants can often be at odds with the nature of scientific work.

The scientific challenge of connecting data from 50 million years ago to the future

Jamie

Our previous papers that we cite in this study  give the picture of how the δ¹³C gradient in the upper ocean changed across the past 15 to 56 million years. This encompasses a long-term cooling trend in Earth’s climate from a greenhouse climate where atmospheric CO₂ was much higher than today and temperatures were much warmer. Looking back at those past climates gives us a real-world example of what happened in the Twilight Zone in the type of warm climate we might be currently heading towards. We set out to make this qualitative comparison more of a quantitative projection. To do that we derived a statistical fit between sinking organic matter fluxes which feeds Twilight Zone organisms, and the abundance of microfossils in the Twilight Zone. We could then apply this to the future, if we could project what the sinking fluxes might be. We needed a numerical model of the Earth System that could reproduce the δ¹³C data from past climates, that we could then run for the modern climate and into the future (with anthropogenic emissions) to indicate future organic matter fluxes. This formed one of the biggest scientific challenges of the paper.

The model we usually run for questions about past climates is an Earth System Model of Intermediate Complexity (EMIC) called cGENIE. EMICs are a class of climate or Earth System models that typically have a reduced spatial resolution (the world is split up into large grid-boxes) and/or some simplifications in the representation of physical processes. This is needed because it saves computational power that can be used to include other processes in the model, such as ocean biogeochemistry, and it also allows the model to simulate longer periods of time and give us essential information on deep ocean circulation. To get an estimate of δ¹³C for past climates, we had to prescribe inputs like continental configuration and atmospheric CO₂ and then perform a “spin up” - the process of running the model from initial conditions with the prescribed inputs to a point where everything in the model is in a dynamic balance. For our past climate experiments this requires about 10,000 model years, taking about 1-2 days in real time on a linux cluster. The low resolution of the model also suits the type of data we have for past climates which is often derived from deep-sea sediment cores which are not densely sampled in space and reveal longer time-scale features. This is very different to the models used to project into the future!

The models typically used to project into the future, such as those that underpin the IPCC reports, have a much higher spatial resolution (e.g., 1° lon/lat) than cGENIE (e.g.,10° lon/lat) and a more detailed representation of ocean physics because this is crucial to getting projections that can inform us of quantities like the average temperature over the UK in the next 50-100 years. However, these models take a lot longer to run because of this (1-10 model years per day on a supercomputer for the higher resolution model vs. 5000 model years per day on a standard machine for cGENIE). We had to choose whether to switch to a higher-resolution for detailed future projections but lose the consistency with our paleo-data or use cGENIE but not have the higher resolution. In the end we chose to keep the same model because we felt it was important to have the same tool with the same limitations applied consistently across past to future climates; “comparing apples with apples”.

In order to use cGENIE for future predictions, we had to convince ourselves (and the reviewers!) that the model was doing as good a job of the future projections. We compared the cGENIE future projection of ocean circulation, using Atlantic overturning and streamfunction, against published compilations of IPCC-class models. We also compared dissolved oxygen in the upper ocean as this is strongly dependent on ocean circulation and warming. Ultimately cGENIE did a reasonable job at reproducing the ensemble means of the IPCC-class models - or at least was not worse! This comparison exercise took more time and focus than running and analysing the future projections in cGENIE. But of course, having confidence in our future modelled climate was essential.

Working on the paper during the COVID pandemic

Katherine

The work on this paper was done largely during the COVID “Lock Down” – where we all worked from home, with the aim of reducing the spread and threat of Covid19. The authors were living in the UK, France, US and Norway at the time. We started with weekly zoom calls, and sketched out the manuscript all together, with individuals responsible for certain sections of the paper. Two of the authors lived a short walk from each other but only spoke over zoom on these calls. It was really the most team-effort writing I have ever done. Eventually the zoom calls got less frequent as the work came together. Although the pandemic had big impacts on researchers, it showed us that working as a team remotely could be very productive. The project that led to this work itself actually officially ended in 2019 (along with funding), but it seems this is often the way, at least in Earth Sciences: an interesting result leads to another interesting question that unfortunately pushes well past the end date of the project! At least with Zoom calls we could continue with the study, and that the model could be run on University servers remotely, and we had no need to get in to the lab, meant that “Lock Down” was not an unproductive time (sparing here a thought for the lecturers and professors  in our research team, who still had teaching responsibilities - they were just as busy, if not more so, than usual!).

We had initially wanted to include the results and write up in a special issue of a different journal but missed the deadline several times even though the deadline was moved to account for the effects of lock down, closed universities, and the need for many researchers to move all their teaching online. In the end, we submitted it to Nature Communications, and a mere 18 months later and 2 rounds of reviews and revisions, we were very happy to have the paper accepted. We understand the findings are quite shocking, ours is really a first attempt to understand the vulnerability of the Twilight Zone to fast warming. We do need more data and scientific understanding of this vast ocean region and the life in it, but looking to the past can already tell us the best thing to do is drastically reduce carbon emissions and keep warming to a minimum.