Prof. Gao Huiwang’s team of OUC’s Key Laboratory of Marine Environment and Ecology (Chinese Ministry of Education) published a paper entitled “Aerosol Iron Solubility Specification in the Global Marine Atmosphere with Machine Learning” in Environmental Science & Technology, a leading academic journal in environmental science. The paper integrates data of the global marine atmosphere and develops a deep learning neural network (DLNN) model for predicting aerosol iron solubility, which is a crucial progress in accurately assessing the amount of aerosol iron deposition available to marine organisms worldwide.

The study reproduces observation results from most seas and margin areas with the Pearson correlation coefficients (r) as large as 0.73−0.97. Compared with those of the Integrated Massively Parallel Atmospheric Chemical Transport (IMPACT) model, the results of DLNN model are more consistent with observations in the global marine atmosphere. In particular, in the Southern Ocean, a key oceanic area affected by atmospheric iron deposition, fog samples in DLNN models contribute to addressing the challenge that the high soluble iron (Fe) observed in the area has been significantly underestimated by the IMPACT model. Incorporating the DLNN model into global models is expected to largely improve the accuracy of predicting air-to-sea soluble Fe input on the global scale. An anonymous reviewer believes that the study “will have important implications to global biogeochemical cycles and will interest not only atmospheric chemistry scientists but also ocean biogeochemists”.