The research, led by Christopher Traill of the Institute for Marine and Antarctic Studies at the University of Tasmania, together with colleagues from CSIRO, the University of Liverpool and the Australian Antarctic Program Partnership, focused on the Subantarctic Zone south of Tasmania. This is the home of the Southern Ocean Time Series observatory, a long-running mooring program that has documented the region’s seasonal cycles of temperature, nutrients, iron and biological production. The team analysed a region spanning 45 to 49 degrees south and 140 to 150 degrees east, an area broadly representative of subantarctic waters between 90 and 145 degrees east.

The starting point for the study was a puzzle that has dogged Southern Ocean science for years. Five widely used satellite-based algorithms for estimating net primary production, the rate at which phytoplankton convert carbon dioxide into organic matter, disagree not only on how much production occurs but on which environmental factors drive it. In the study region, the VGPM and Eppley-VGPM algorithms linked high-production years to warmer sea surface temperatures and a southward shift of the subtropical front, the boundary that carries warm, salty subtropical water into the region. The CbPM, AbPM and CAFE algorithms instead associated productive years with shallower mixed layers and cooler conditions. These divergent signatures matter, because without knowing which algorithms are right, scientists cannot confidently detect long-term trends in Southern Ocean productivity from space.

To break this impasse, the researchers turned to a state-of-the-art biogeochemical ocean model called PISCES-Quota-Fe, coupled to the NEMO circulation model and run as a hindcast simulation from 1958 to 2022. The model explicitly resolves three phytoplankton types, two zooplankton groups, and a detailed iron cycle that has been tuned against global observations of dissolved iron. Crucially, unlike satellites, the model exposes every internal term: how strongly each phytoplankton group is limited by iron, nitrogen, light and temperature, how fast cells divide, and how much of the population is grazed away by predators. The team compared the five highest and five lowest production years in the simulation, testing whether the differences between them exceeded the background variability using a bootstrap resampling procedure repeated ten thousand times.

The model’s verdict was unambiguous: the most productive years were driven primarily by relief of iron limitation, the region’s master nutrient constraint. In those years, winter and spring mixed layers, the surface layer stirred by storms and cooling, reached deeper into the water column, entraining dissolved iron from below. This lifted the iron limitation on the phytoplankton community from July through November, allowing faster growth within the mixed layer. Nanophytoplankton, the dominant group, contributed most of the extra production. Later in summer, a second iron source kicked in: enhanced remineralisation, the microbial breakdown of organic matter, recycled iron back into the surface layer, sustaining the bloom as the mixed layer shoaled and light became more abundant.

But the deeper mixing did something unexpected as well, and this is where the study becomes genuinely surprising. Deeper mixing pushed phytoplankton biomass below the average mixed layer depth, diluting the food available to zooplankton grazers in the surface layer. Grazing loss rates in the mixed layer dropped significantly, along with zooplankton biomass. This relaxation of predation, a top-down control, primed the surface layer for explosive biomass accumulation once spring arrived and growth conditions improved. In effect, the productive years were shaped by a one-two punch: bottom-up relief of iron scarcity combined with a top-down release from grazing pressure, both triggered by the same physical event of deeper winter mixing.

This mechanism explains a decoupling that has long confused observers: in the model, years with higher total production did not simply have more phytoplankton biomass, and spring biomass changes did not track growth rates. Total biomass anomalies peaked in spring, while the production anomaly peaked in January, and average community growth rates actually fell in spring even as production rose, because cells were redistributed into deeper, darker water where light was scarce. For satellite algorithms, which must infer production from surface-visible biomass proxies such as chlorophyll, this decoupling is a serious problem. The study suggests that the disagreement among algorithms stems largely from how each one links biomass, rather than growth rates, to production, and from which environmental variables, temperature, mixed layer depth, absorption spectra, they fold into their equations.

The implications extend to the models used to project Earth’s climate future. When the team examined eight Earth system models from the CMIP6 archive, they found that all converged on increasing Southern Ocean production trends, consistent with declining iron limitation, but the mechanisms behind their year-to-year variability differed widely. Several models showed a strong bias toward sea surface temperature as the dominant driver, and the magnitude of production swings between the highest and lowest years ranged from 2 to 40 millimoles of carbon per square metre per day. Because the PISCES-Quota-Fe analysis shows that production responds to interacting nutrient, light and grazing processes rather than temperature alone, models that lean on thermal drivers may be getting the right trend for the wrong reasons, a known pitfall in climate projection evaluation.

The trend picture from observations is equally unsettled. Over the satellite record from 1998 to 2022, the VGPM algorithms showed significant positive trends in the study region while CbPM, AbPM and CAFE showed negative ones. The model itself produced no significant trend over that window but a significant increase over its full 1975 to 2022 span, with the trend direction sensitive to start and end dates. The authors argue that in-situ trends in this part of the Subantarctic Zone are very likely to reflect decreasing production, as the algorithms that best match field observations suggest, but they urge caution in interpreting any single trend estimate given these sensitivities.

The study closes with a clear prescription. Constraining the future of subantarctic productivity requires better observations of the processes satellites cannot see: zooplankton grazing rates, the vertical distribution of phytoplankton biomass through the water column, and the multi-resource limitation of growth by iron, nitrogen and light simultaneously. These are precisely the terms that models handle with the greatest uncertainty, and the ones that determine whether deepening or shoaling mixed layers in a warming ocean will strengthen or weaken this critical carbon sink. Until those measurements exist, both satellite products and climate models should be read with a healthy dose of mechanistic scepticism, because in the subantarctic ocean, what drives a bloom is far more complicated than what floats on the surface.

Subject of Research: Drivers of interannual variability and trends in net primary production in the subantarctic Southern Ocean

Article Title: Towards constraining the drivers of variability and trends in subantarctic productivity

Article References: Traill, C. D., Rohr, T. W., Shadwick, E. H., Buchanan, P. J., Tagliabue, A., & Bowie, A. R. (2026). Towards constraining the drivers of variability and trends in subantarctic productivity. Biogeosciences, 23(19), 6947-6977. https://doi.org/10.5194/bg-23-6947-2026

Image Credits: AI Generated

DOI: 10.5194/bg-23-6947-2026

Keywords: Southern Ocean, subantarctic, net primary production, phytoplankton, iron limitation, mixed layer depth, zooplankton grazing, remote sensing, biogeochemical model, PISCES-Quota-Fe, CMIP6, carbon cycle