A global monthly climatology of total alkalinity: a neural network approach

Abstract. Global climatologies of the seawater CO2 chemistry variables are necessary to assess the marine carbon cycle in depth. The seasonal variability should be adequately captured in them to properly address issues such as ocean acidification. Total alkalinity (AT) is one variable of the seawater CO2 chemistry system involved in ocean acidification and frequently measured during campaigns assessing the marine carbon cycle. We took advantage of the data product Global Ocean Data Analysis Project version 2 (GLODAPv2) to extract the relations between the drivers of the AT variability and this variable using a neural network to generate a monthly climatology. 99% of the GLODAPv2 dataset used was modelled by the network with a root-mean-squared error (RMSE) of 5.1 µmol kg-1. The validation carried out using independent datasets revealed the good generalization of the network. Five ocean time-series stations used as an independent test showed an acceptable RMSE in the range of 3.1-6.2 µmol kg-1. The successful modeling of the monthly variability of AT in the time-series makes our network a good candidate to generate a monthly climatology. It was obtained passing the climatologies of the World Ocean Atlas 2013 (WOA13) through the network. The spatiotemporal resolution of the climatology is determined by the one of WOA13: 1ºx1º in the horizontal, 102 depth levels (0-5500m) in the vertical, and 12 months. We offer the product as a service to the scientific community at the data repository of the Spanish National Research Council (CSIC; doi: http://dx.doi.org/10.20350/digitalCSIC/8564) with the purpose to contribute to a continuous improvement of the understanding of the global carbon cycle.