Abstract. The Southern Ocean is a complex system yet is sparsely sampled in both space and time. These factors raise questions about the confidence in present sampling strategies and associated machine learning (ML) reconstructions. Previous studies have not yielded a clear understanding of the origin of uncertainties and biases for the reconstructions of the partial pressure of carbon dioxide (pCO2) at the surface ocean (pCO2ocean). Here, we examine these questions by investigating the sensitivity of pCO2ocean reconstruction uncertainties and biases to a series of semi-idealized observing system simulation experiments (OSSEs) that simulate spatio-temporal sampling scales of surface ocean pCO2 in ways that are comparable to ocean CO2 observing platforms (Ship, Waveglider, Carbon-float, Saildrone). These experiments sampled a high spatial resolution (±10 km) coupled physical and biogeochemical model (NEMO-PISCES) within a sub-domain representative of the Sub-Antarctic and Polar Frontal Zones in the Southern Ocean. The reconstructions were done using a two-member ensemble approach that consisted of two machine learning (ML) methods, (1) the feed-forward neural network and (2) the gradient boosting machines. With the baseline observations being from the simulated ships mimicking observations from the Surface Ocean CO2 Atlas (SOCAT), we applied to each of the scale-sampling simulation scenarios the two-member ensemble method ML2, to reconstruct the full sub-domain pCO2ocean and assess the reconstruction skill through a statistical comparison of reconstructed pCO2ocean and model domain mean. The analysis shows that uncertainties and biases for pCO2ocean reconstructions are very sensitive to both the spatial and temporal scales of pCO2 sampling in the model domain. The four key findings from our investigation are the following: (1) improving ML-based pCO2 reconstructions in the Southern Ocean requires simultaneous high resolution observations of the meridional and the seasonal cycle (< 3 days) of pCO2ocean; (2) Saildrones stand out as the optimal platforms to simultaneously address these requirements; (3) Wavegliders with hourly/daily resolution in pseudo-mooring mode improve on Carbon-floats (10-day period), which suggests that sampling aliases from the low temporal frequency have a greater negative impact on their uncertainties, biases and reconstruction means; and (4) the present summer seasonal sampling biases in SOCAT data in the Southern Ocean may be behind a significant winter bias in the reconstructed seasonal cycle of pCO2ocean.

WG Waveglider nUSV new unmanned surface vehicle ML Machine Learning This supporting information document provides ancillary methodological details and results pertaining to (1) descriptions of the study domain and mode data variables including the motive of the selection of the experimental domain, the characteristics of the NEMO-PISCES model (BIOPERIANT12) data variables of interest and processing, the experimental setting and steps used in the 𝑝CO ! reconstruction; (2) descriptions of the ML regression methods; and (3) additional components on the results and discussion including the model training errors or in-sample uncertainties and biases, and the overall results of the SHIP experiment. Accompanying this supporting information text are four supplementary figures and four supplementary tables. S1 Descriptions of the study domain and mode data variables S1.1 Selection of the study domainMany studies, the seasonal cycle is known as the strongest mode of natural variability of carbon dioxide (CO2) and also the one that most strongly links climate and ocean ecosystems. The seasonal cycle characteristics are largely shaped by higher frequency intra-seasonal modes defining the response modes in physics and biogeochemistry components (Mongwe et al., 2016(Mongwe et al., , 2018. Therefore, the SOSCEx -an initiative of the SOCCO, a research programme led by the CSIR-was launched in 2013. The SOSCEx aimed to explore the nature and links in dynamics and scale sensitivities of atmospheric forcing, CO2 fluxes, and primary production, with a particular focus on the seasonal cycle mode as a test for the climate sensitivity of earth systems models in respect of the evolution of both atmospheric CO2 and ocean ecosystems in the 21st century (Swart et al., 2012;Monteiro et al., 2010Monteiro et al., , 2015. The novel aspect of the third phase (SOSCEx III, 2015-2018) of the project was the integration of a multi-platform approach that consisted of combining gliders, ships, floats, satellites and prognostic models in order to explore new questions about climate sensitivity of CO2 and ocean ecosystem dynamics and how these processes are parameterized in forced ocean models such as the high-resolution (±10km) forced NEMO-PISCES ocean model BIOPERIANT12.