A global monthly field of seawater pH over 3 decades: a machine learning approach

• Main Authors: Zhong, Guorong, Li, Xuegang, Song, Jinming, Qu, Baoxiao, Wang, Fan, Wang, Yanjun, Zhang, Bin, Cheng, Lijing, Ma, Jun, Yuan, Huamao, Duan, Liqin, Li, Ning, Wang, Qidong, Xing, Jianwei, Dai, Jiajia
• Format: Text
• Language: English
• Published: 2024
• Subjects: Ocean acidification
• Online Access: https://doi.org/10.5194/essd-2024-151; https://essd.copernicus.org/preprints/essd-2024-151/

The continuous uptake of anthropogenic CO 2 by the ocean leads to ocean acidification, which is an ongoing threat to the marine ecosystem. The ocean acidification rate was globally documented in the surface ocean but limited below the surface. Here, we present a monthly four-dimensional 1°×1° gridded product of global seawater pH, derived from a machine learning algorithm trained on pH observations at total scale and in-situ temperature from the Global Ocean Data Analysis Project (GLODAP). The constructed pH product covers the years 1992–2020 and depths from the surface to 2 km on 41 levels. Three types of machine learning algorithms were used in the pH product construction, including self-organizing map neural networks for region dividing, a stepwise algorithm for predictor selection, and feed-forward neural networks (FFNN) for non-linear relationship regression. The performance of the machine learning algorithm was validated using real observations by a cross validation method, where four repeating iterations were carried out with 25 % varied observations for each evaluation and 75 % for training. The constructed pH product is evaluated through comparisons to time series observations and the GLODAP pH climatology. The overall root mean square error between the FFNN constructed pH and the GLODAP measurements is 0.028, ranging from 0.044 in the surface to 0.013 at 2000 m. The pH product is distributed through the data repository of the Marine Science Data Center of the Chinese Academy of Sciences at http://dx.doi.org/10.12157/IOCAS.20230720.001 (Zhong et al., 2023).