The Arctic sea ice is an important indicator of the progress of global warming and climate change. Prediction of Arctic sea ice concentration has been investigated by many disciplines and predictions have been made using a variety of methods. Deep learning (DL) using large training datasets, also known as deep neural network, is a fast-growing area in machine learning that promises improved results when compared to traditional neural network methods. Arctic sea ice data, gathered since 1978 by passive microwave sensors, may be an appropriate input for training DL models. In this study, a large Arctic sea ice dataset was employed to train a deep neural network and this was then used to predict Arctic sea ice concentration, without incorporating any physical data. We compared the results of our methods quantitatively and qualitatively to results obtained using a traditional autoregressive (AR) model, and to a compilation of results from the Sea Ice Prediction Network, collected using a diverse set of approaches. Our DL-based prediction methods outperformed the AR model and yielded results comparable to those obtained with other models.

Changes in Arctic sea ice affect atmospheric circulation, ocean current, and polar ecosystems. There have been unprecedented decreases in the amount of Arctic sea ice due to global warming. In this study, a novel 1-month sea ice concentration (SIC) prediction model is proposed, with eight predictors using a deep-learning approach, convolutional neural networks (CNNs). This monthly SIC prediction model based on CNNs is shown to perform better predictions (mean absolute error – MAE – of 2.28 %, anomaly correlation coefficient – ACC – of 0.98, root-mean-square error – RMSE – of 5.76 %, normalized RMSE – nRMSE – of 16.15 %, and NSE – Nash–Sutcliffe efficiency – of 0.97) than a random-forest-based (RF-based) model (MAE of 2.45 %, ACC of 0.98, RMSE of 6.61 %, nRMSE of 18.64 %, and NSE of 0.96) and the persistence model based on the monthly trend (MAE of 4.31 %, ACC of 0.95, RMSE of 10.54 %, nRMSE of 29.17 %, and NSE of 0.89) through hindcast validations. The spatio-temporal analysis also confirmed the superiority of the CNN model. The CNN model showed good SIC prediction results in extreme cases that recorded unforeseen sea ice plummets in 2007 and 2012 with RMSEs of less than 5.0 %. This study also examined the importance of the input variables through a sensitivity analysis. In both the CNN and RF models, the variables of past SICs were identified as the most sensitive factor in predicting SICs. For both models, the SIC-related variables generally contributed more to predict SICs over ice-covered areas, while other meteorological and oceanographic variables were more sensitive to the prediction of SICs in marginal ice zones. The proposed 1-month SIC prediction model provides valuable information which can be used in various applications, such as Arctic shipping-route planning, management of the fishing industry, and long-term sea ice forecasting and dynamics.

A new Sentinel-1 image-based sea ice classification algorithm using a machine-learning-based model trained in a semi-automated manner is proposed to support daily ice charting. Previous studies mostly rely on manual work in selecting training and validation data. We show that the readily available ice charts from the operational ice services can reduce the amount of manual work in preparation of large amounts of training/testing data. Furthermore, they can feed highly reliable data to the trainer by indirectly exploiting the best ability of the sea ice experts working at the operational ice services. The proposed scheme has two phases: training and operational. Both phases start from the removal of thermal, scalloping, and textural noise from Sentinel-1 data and calculation of grey level co-occurrence matrix and Haralick texture features in a sliding window. In the training phase, the weekly ice charts are reprojected into the SAR image geometry. A random forest classifier is trained with the texture features on input and labels from the rasterized ice charts on output. Then, the trained classifier is directly applied to the texture features from Sentinel-1 images operationally. Test results from the two datasets spanning winter (January–March) and summer (June–August) seasons acquired over the Fram Strait and the Barents Sea showed that the classifier is capable of retrieving three generalized cover types (open water, mixed first-year ice, old ice) with overall accuracies of 87 % and 67 % in winter and summer seasons, respectively. For the summer season, the classifier failed in distinguishing mixed first-year ice from old ice with accuracy of only 12 %; however, it performed rather like an ice–water discriminator with high accuracy of 98 % as the misclassification between the mixed first-year ice and old ice was between them. The accuracy for five cover types (open water, new ice, young ice, first-year ice, old ice) in the winter season was 60 %. The errors are attributed both to incorrect manual classification on the ice charts and to the semi-automated algorithm. Finally, we demonstrate the potential for near-real-time service of the ice map using daily mosaicked Sentinel-1 images.

Since the start of the industrial revolution, human activities have caused a rapid increase in atmospheric carbon dioxide (CO₂) concentrations, which have, in turn, had an impact on climate leading to global warming and ocean acidification. Various approaches have been proposed to reduce atmospheric CO₂. The Martin (or iron) hypothesis suggests that ocean iron fertilization (OIF) could be an effective method for stimulating oceanic carbon sequestration through the biological pump in iron-limited, high-nutrient, low-chlorophyll (HNLC) regions. To test the Martin hypothesis, 13 artificial OIF (aOIF) experiments have been performed since 1990 in HNLC regions. These aOIF field experiments have demonstrated that primary production (PP) can be significantly enhanced by the artificial addition of iron. However, except in the Southern Ocean (SO) European Iron Fertilization Experiment (EIFEX), no significant change in the effectiveness of aOIF (i.e., the amount of iron-induced carbon export flux below the winter mixed layer depth, MLD) has been detected. These results, including possible side effects, have been debated amongst those who support and oppose aOIF experimentation, and many questions concerning the effectiveness of scientific aOIF, environmental side effects, and international aOIF law frameworks remain. In the context of increasing global and political concerns associated with climate change, it is valuable to examine the validity and usefulness of the aOIF experiments. Furthermore, it is logical to carry out such experiments because they allow one to study how plankton-based ecosystems work by providing insight into mechanisms operating in real time and under in situ conditions. To maximize the effectiveness of aOIF experiments under international aOIF regulations in the future, we therefore suggest a design that incorporates several components. (1) Experiments conducted in the center of an eddy structure when grazing pressure is low and silicate levels are high (e.g., in the SO south of the polar front during early summer). (2) Shipboard observations extending over a minimum of ∼40 days, with multiple iron injections (at least two or three iron infusions of ∼2000 kg with an interval of ∼10–15 days to fertilize a patch of 300 km² and obtain a ∼2 nM concentration). (3) Tracing of the iron-fertilized patch using both physical (e.g., a drifting buoy) and biogeochemical (e.g., sulfur hexafluoride, photosynthetic quantum efficiency, and partial pressure of CO₂) tracers. (4) Employment of neutrally buoyant sediment traps (NBST) and application of the water-column-derived thorium-234 (²³⁴Th) method at two depths (i.e., just below the in situ MLD and at the winter MLD), with autonomous profilers equipped with an underwater video profiler (UVP) and a transmissometer. (5) Monitoring of side effects on marine/ocean ecosystems, including production of climate-relevant gases (e.g., nitrous oxide, N₂O; dimethyl sulfide, DMS; and halogenated volatile organic compounds, HVOCs), decline in oxygen inventory, and development of toxic algae blooms, with optical-sensor-equipped autonomous moored profilers and/or autonomous benthic vehicles. Lastly, we introduce the scientific aOIF experimental design guidelines for a future Korean Iron Fertilization Experiment in the Southern Ocean (KIFES).