Multiple linear regression models for the estimation of PH and Aragonite saturation state in the Northwestern Gulf of Mexico




Jundt, EvaLynn

Journal Title

Journal ISSN

Volume Title




The ocean plays a vital role in making up 70% of the Earth’s surface, producing over half of oxygen globally, and absorbing approximately 30% of anthropogenic CO2 since the industrial revolution. Ocean acidification (OA) is a direct threat to many organisms living in the oceans across the globe, yet the state of carbonate chemistry and the rate of OA vary in different parts of the world’s oceans. Although current data suggest that the open Gulf of Mexico (GOM) surface waters have relatively high pH (> 8) and aragonite saturation state (ΩArag > 3), the GOM could still experience ecological impacts of OA. In addition, the combination of increasing atmospheric CO2, upwelling, and increasing terrestrial nutrient export may acidify the coastal waters even further. Acidifying ocean waters have decreasing ΩArag, posing serious threats to calcifying organisms, affecting their populations, growth patterns, and shell or skeletal density. The GOM is home to the northernmost tropical coral reefs around the contiguous United States as well as prominent shellfish industry. Historical water column carbonate measurements are scarce, so the progression of OA in the GOM is poorly understood. Research regarding OA in the GOM is needed to manage and protect these resources. In the literature, multiple linear regression (MLR) models have been created to fill data gaps in different ocean regions such as the Gulf of Alaska, the Southern Ocean, the Sea of Japan, and coasts of the northeastern and northwestern United States. Prior to this study, no statistical model existed for carbonate chemistry parameters (i.e., pH and ΩArag) in the GOM. By creating models built upon the relationships between commonly measured hydrographic properties (salinity, temperature, pressure, and dissolved oxygen (DO)) and pH as well as ΩArag, data gaps can be filled in areas that do not have sufficient sampling coverage. In this study, I created statistical models for the estimation of ΩArag and pH in the northwestern GOM (NWGOM) from latitudes 27.1-29.0˚N and longitudes 91.5-95.0˚W. The calibration data used in the models include depth, salinity, temperature, pressure, and DO collected from four cruises that took place in July 2007, July 2017, and April and August of 2021. The models predict ΩArag with R2≥0.98, RMSE ≤ 0.14 and pH with R2 ≥ 0.93, RMSE ≤ 0.02 for four different subsets of the data depending on depth (with and without removal of upper 20 m) and geographic location (with and without removal of stations to the east). The data used to create the models are also used to create contour plots that show variation of ΩArag and pH over the timeframe of the study from 2007 to 2021. Relatively low ΩArag (ΩArag ≤ 2) values are present in the depths ≥ 180 m. The depth range of the water column between ΩArag = 2-1.5 decreased over this period. The depths for ΩArag = 2 and ΩArag = 1.1 vary ±20 and ±50 m respectively, while the depth for ΩArag = 1.5 decreased 50 m from 2007 to 2021. Depth profiles for pH revealed consistent patterns over all four cruises with highest values over the shelf and upper 125 m, and minimum values around 500 m. The pH = 7.9 isopleth remained around 265 m for all cruises, while the pH = 8 isopleth showed fluctuation of ±10 m (from 2007 to 2021). On the shelf, the maximum and minimum pH values were 0.0356 and 0.0133 units lower in 2021 than in 2007, respectively. This resulted in the range of pH values experienced narrowing by 0.0223 and transitioning to lower pH values overall. These MLR models are valuable tools for reconstructing ΩArag and pH data where direct chemical observations are absent but hydrographic information is available. These models can be applied to the NWGOM within ±10 years of 2014, although observations of potential shifts in circulation, water mass composition, and anthropogenic CO2 should be monitored to improve or revise these models in the future.



aragonite, carbonate chemistry, modeling, multilinear regression, ocean acidification, pH



Attribution-NonCommercial-NoDerivatives 4.0 International