Ocean acidification, caused by the ongoing absorption of atmospheric CO₂, poses threats to marine ecosystems and biodiversity. Accurately assessing variations in seawater pH is crucial for evaluating biological responses to acidification and predicting the ocean's capacity for carbon sequestration.

However, global ocean acidification has not been thoroughly studied due to sparse observations of seawater pH and inconsistent spatial coverage, especially at depths below the ocean's surface.

To address these challenges, a research team from the Institute of Oceanology of the Chinese Academy of Sciences (IOCAS) utilized a Stepwise Feed-Forward Neural Network (Stepwise FFNN) algorithm to identify the predictors that yielded the lowest reconstruction errors for seawater pH. Additionally, they integrated observational data from the Global Ocean Data Analysis Project (GLODAP) to create a global monthly 3D gridded pH dataset spanning the past 30 years. 

"Our 3D gridded pH dataset extends to a depth of 2000 meters and improves in both accuracy and reliability," said Dr. ZHONG Guorong, the first author of the study.

By categorizing global oceans into biogeochemical provinces based on pH drivers, the researchers optimized the selection of environmental variables, which enhanced the dataset's accuracy. In addition, the use of cross-boundary optimal interpolation technology improved the accuracy of reconstructing marine chemical parameters.

Moreover, the pH dataset has been validated using a cross-validation method that reduces the risk of model overfitting, ensuring its reliability. The dataset is available to the public via the IOCAS Data Center, making it an essential resource for global climate modeling and marine conservation efforts.

This study was published in Earth System Science Data.

Vertical distribution of seawater pH from the 3D gridded dataset. (Image by IOCAS)