The sensitivity of p CO 2 reconstructions to sampling scales across a Southern Ocean sub-domain: a semi-idealized ocean sampling simulation approach

[摘要] The Southern Ocean is a complex system yet is sparselysampled in both space and time. These factors raise questions about theconfidence in present sampling strategies and associated machine learning(ML) reconstructions. Previous studies have not yielded a clearunderstanding of the origin of uncertainties and biases for thereconstructions of the partial pressure of carbon dioxide( p CO 2 ) at the surface ocean( p CO 2 ocean ). We examine these questions througha series of semi-idealized observing system simulation experiments (OSSEs)using a high-resolution ( ±  10 km) coupled physical and biogeochemicalmodel (NEMO-PISCES, Nucleus for European Modelling of the Ocean, Pelagic Interactions Scheme for Carbon and Ecosystem Studies). Here we choose 1 year of the model sub-domain of 10 ∘ of latitude (40–50 ∘  S) by 20 ∘ of longitude (10 ∘  W–10 ∘  E). Thisdomain is crossed by the sub-Antarctic front and thus includes both thesub-Antarctic zone and the polar frontal zone in the south-east Atlantic Ocean,which are the two most sampled sub-regions of the Southern Ocean. We showthat while this sub-domain is small relative to the Southern Ocean scales,it is representative of the scales of variability we aim to examine. TheOSSEs simulated the observational scales of p CO 2 ocean in ways that are comparable toexisting ocean CO 2 observing platforms (ships, Wave Gliders,carbon floats, Saildrones) in terms of their temporal sampling scales andnot necessarily their spatial ones. The p CO 2 reconstructionswere carried out using a two-member ensemble approach that consisted of two machinelearning (ML) methods, (1) the feed-forward neural network and (2) thegradient boosting machines. The baseline data were from the ship-basedsimulations mimicking ship-based observations from the Surface OceanCO 2 Atlas (SOCAT). For each of the sampling-scale scenarios, we appliedthe two-member ensemble method to reconstruct the full sub-domain p CO 2 ocean . The reconstruction skill was thenassessed through a statistical comparison of reconstructed p CO 2 ocean and the model domain mean. The analysisshows that uncertainties and biases for p CO 2 ocean reconstructions are very sensitive toboth the spatial and the temporal scales of p CO 2 sampling in themodel domain. The four key findings from our investigation are as follows: (1) improving ML-based p CO 2 reconstructions in the SouthernOcean requires simultaneous high-resolution observations ( 3  d)of the seasonal cycle of the meridional gradients of p CO 2 ocean ; (2) Saildrones stand out as theoptimal platforms to simultaneously address these requirements; (3) Wave Gliders with hourly/daily resolution in pseudo-mooring mode improve oncarbon floats (10 d period), which suggests that sampling aliases from the10 d sampling period might have a greater negative impact on theiruncertainties, biases, and reconstruction means; and (4) the presentseasonal sampling biases (towards summer) in SOCAT data in the SouthernOcean may be behind a significant winter bias in the reconstructed seasonalcycle of p CO 2 ocean .

[学科分类] 大气科学