Dataset: California Current winter climate reconstruction - upwelling index
Deployment: Black_model

To investigate variability in upwelling with respect to season, we computed the monthly mean, standard deviation and coefficient of variation for each physical variable (Uw, SST, and UI).  Subsequently, we conducted a Principal Component Analysis (PCA) to explore shared patterns of variability among these time series, which were extracted and used as multivariate indices of the CCE.  Given that upwelling varies among locations, among season and among years (García-Reyes and Largier 2012; Thompson and others 2012), we arranged each physical data variable (Uw, SST and UI) into a three-dimensional matrix consisting of 12 locations x 12 months x 23 years; resulting components were labeled as PCUw, PCSST, and PCUI.  Each column was normalized (zero mean and variance equal to 1 standard deviation) before calculating the PCA.  Next we ran a PCA that combined Uw and SST data by arranging their data arrays into a single matrix with dimensions: 24 locations (12 Uw locations + 12 SST locations) x 12 months x 23 years.  Resulting PCs, labeled as PCenv, captured the dominant and sub-dominant seasonal "modes" or patterns in upwelling and their interannual variability.  PCs (scores) from the three physical variables (PCUw, PCSST, and PCUI) were compared to one another as well as to PCenv using Spearman ranked correlations.  PCs with Eigenvalues < 1 and explaining < 10% of the variability in the data set were dropped from further analysis (Jolliffe 2002). 

The 15 biological indicators included in the study were cross-correlated with one another to generally assess the extent to which they shared common patterns.  Subsequently, we conducted a PCA (resulting components labeled as PCbio) using the nine longest (1982-2006) biological indicators.  Biological indicators excluded from the PCA were correlated against the PCbio components as a measure of their agreement with dominant patterns in the longer datasets.  Those that were significant at the p < 0.05 level were retained.  To summarize physical-biological interactions, the scores of the environmental PCs (PCUw, PCSST, PCUI and PCenv), the biological indices, and biological PCs (PCbio) were compared using Spearman correlations.

Biological Principal Component Sources and Loading (PDF)

Environmental Principal Component Sources and Loading (PDF)

References, this data:

Reyes, M., W. J. Sydeman, S. A. Thompson, B. A. Black, R. R. Rykaczewski, J. A. Thayer, and S. J. Bograd. 2013. Integrated assessment of wind effects on Central California’s pelagic ecosystem. Ecosystems. DOI: 10.1007/s10021-013-9643-6.

Related References:

García-Reyes M, Largier JL. 2012. Seasonality of coastal upwelling off central and northern California: new insights, including temporal and spatial variability. J Geophys Res 117:C03028.

Jolliffe IT. 2002. Principal component analysis. 2nd ed. Springer series in statistics. ISBN 0-387-95442-2.

Large WG, Pond S. 1981. Open ocean momentum flux measurements in moderate to strong winds. J Phys Oceanogr 11:324-36.

Thompson SA, Sydeman WJ, Santora JA, Black BA, Suryan RM, Calambokidis J, Peterson WT, Bograd SJ. 2012. Linking predators to seasonality of upwelling: using food web indicators and path analysis to infer trophic connections. Prog Oceanogr 101:106–20.