ARIOS: An Acidification Ocean Data Base in the Galician Upwelling Ecosystem

Since 1975, pH and alkalinity measurements have been carried out with ancillary variables (temperature, salinity, nutrients and O₂) within the framework of 22 national projects in the Rias and Galicia platform. The objectives have evolved over time with the aim of understanding and modeling the primary production of a highly productive and strongly seasonal ecosystem where upwelling plays a key role. With more than 16,000 pH data, the different projects have intermittently measured the pH in all Galician Rias, coastal zone and oceanic region, from the surface to the bottom, mainly in summer and spring, but also monitored monthly time series and more frequently for some years. Here we have joined and harmonized the 22 databases to study long-term trends, as well as the pH variability observed according to ancillary variables.

The 4300 surface pH measures (depth <7 m) correlate mainly with the biogeochemical variables (nutrients and O₂) together explaining more than 66% of the pH variance, and together with the temperature 77% of the variance of the [exCO₃⁼] (concentration of carbonate ion over its saturation value). Once the biogeochemical variability is removed, an acidification trend of -1.47 ± 0.09 ·10^-3 of pH yr^-1 is revealed (p-level <0.008 n = 4300). Seasonally, the variation of this trend shows a bias with trends of -2.13 ± 0.19 ·10^-3 of pH yr^-1 (p-level <0.05 n = 530) with spring data, while the fall trend obtained is -1.18 ± 0.15 ·10^-3 of pH yr^-1 (p-level<0.034 n = 1258). In the coastal zone (bottom depth <70 m) acidification is slightly higher [-1.61 ± 0.11 ·10^-3 of pH yr^-1, p-level<0.005 n = 3627] than in the open sea (bottom depth > 70 m) [-1.15 ± 0.11 ·10^-3 of pH yr^-1, p-level <0.01 n = 673]. The [exCO₃⁼] also shows a decreasing trends of [-0.50 ± 0.03 µM·yr^-1, p-level<0.005] and [-0.31 ± 0.04 µM·yr^-1, p-level<0.01] in the coastal region and open sea, respectively. In addition, through the numerical techniques of neural networks, the pH variability observed in these 22 chemical databases has been projected into a weekly recorded database of physicochemical variables (INTECMAR) from 1992 to 2018. Independent data showed how the seasonal variability derived from the neural network highly correlates with those of the measured pH data and allows having a robust weekly time series to assess the variability of the pH for almost three decades.