Despite recent interest in understanding long-term trends in ocean acidity, natural variations of carbon chemistry on short timescales are still poorly understood. Unfortunately, historical observations of the oceanic CO2 system are relatively few in number. Such data are particularly scarce along the highly productive Canadian Pacific coast. However, hydrographic data such as temperature, salinity, oxygen and nutrients have been collected regularly in this region. I developed a fully cross-validated statistical model to predict the aragonite saturation state (Ωarag), a biologically relevant measure of the carbonate system. Different sensitivity tests were performed to assess the robustness of the statistical modelling skill to different model structures. In particular, this study found that in situ temperature and O2 used together were strong predictors of Ωarag. The carbon data used to build this statistical model came from five hydrographic surveys along the Pacific coast of Canada (in July 1998, August 2004, late May 2007, February 2010 and early August 2010) that contain direct measurements of CO2 system parameters. Only data from a depth range of 0-750 m were used, as data from below 750 m showed biases due to calcium carbonate dissolution. Although processes such as solar warming and gas exchange occur in the surface and could possibly introduce biases in the model, I show that these surface data can be included. The ability of the statistical models to compute robust estimates of Ωarag was assessed by exploring the generalizability of the model through cross-validation procedures using different partitions of the data. By predicting lnΩarag rather than Ωarag directly, I obtained a strong and robust predictive relationship. This MLR model form yielded a high value in the squared correlation coefficient between predicted and observed values (0.96) and a low percentage in erroneous prediction of undersaturated conditions (3.1%). This relationship was found to be insensitive to changes in spatial domain or interannual variability in the data. These results suggest that the model can be used to estimate the distribution of Ωarag along the outer west coast of Canada when basic hydrographic data on temperature and O2 are available. Predictions of Ωarag from historical observations (1980-2009) in this region reveal that the saturation horizon (Ωarag=1) tended to be more stable in winter and spring and highly variable and occasionally shallow in summer and fall during and following the upwelling season. Undersaturation with respect to aragonite was more likely to occur at shallower depths over the shelf relative to adjacent offshore waters likely as a result of upwelling. The Ωarag saturation horizon tended to be more variable in depth on the shelf compared to offshore waters. The saturation horizon tended to occur at deeper depths over the Queen Charlotte Sound (QCS) shelf and be more stable with respect to the west coast of Vancouver island (WCVI). Thus, the WCVI may experience adverse effects of ocean acidification more acutely than QCS. The use of this approach may provide insight into natural variability and the key controls of Ωarag in future studies at a low cost. However, this predictive model cannot hind-cast data to evaluate the presence of the anthropogenic signal. / Graduate
| Identifer | oai:union.ndltd.org:uvic.ca/oai:dspace.library.uvic.ca:1828/4405 |
| Date | 03 January 2013 |
| Creators | Lara Espinosa, Alejandra |
| Contributors | Ianson, Debby C., Hamme, Roberta Claire |
| Source Sets | University of Victoria |
| Language | English, English |
| Detected Language | English |
| Type | Thesis |
| Rights | Available to the World Wide Web |
Page generated in 0.0015 seconds