Principal Components Analysis, Factor Analysis and Trend Correlations of Twenty-Eight Years of Water Quality Data of Deer Creek Reservoir, Utah

• Main Author: Gonzalez, Nicolas Alejandro
• Format: Others
• Published: BYU ScholarsArchive 2012
• Subjects: PCA; factor analysis; time trends; nutrient loading; nonparametric; Deer Creek; Civil and Environmental Engineering
• Online Access: https://scholarsarchive.byu.edu/etd/3309; https://scholarsarchive.byu.edu/cgi/viewcontent.cgi?article=4308&context=etd

I evaluated twenty-eight years (1980-2007) of spatial-temporal water quality data from Deer Creek Reservoir in Utah. The data came from three sampling points representing the lotic, transitional and lentic zones. The data included measurements of climatological, hydrological and water quality conditions at four depths; Surface, Above Thermocline, Below Thermocline and Bottom. The time frame spanned dates before and after the completion of the Jordanelle Reservoir (1987-1992), approximately fourteen miles upstream of Deer Creek. I compared temporal groupings and found that a traditional month distribution following standard seasons was not effective in characterizing the measured conditions; I developed a more representative seasonal grouping by performing a Tukey-Kramer multiple comparisons adjustment and a Bonferronian correction of the Student's t comparison. Based on these analyses, I determined the best groupings were Cold (December - April), Semi-Cold (May and November), Semi-Warm (June and October), Warm (July and September) and Transition (August). I performed principal component analysis (PCA) and factor analysis (FA) to determine principal parameters associated with the variability of the water quality of the reservoir. These parameters confirmed our seasonal groups showing the Cold, Transition and Warm seasons as distinct groups. The PCA and FA showed that the variables that drive most of the variability in the reservoir are specific conductivity and variables related with temperature. The PCA and FA showed that the reservoir is highly variable. The first 3 principal components and rotated factors explained a cumulative 59% and 47%, respectively of the variability in Deer Creek. Both parametric and nonparametric approaches provided similar correlations but the evaluations that included censored data (nutrients) were considerably different with the nonparametric approach being preferred.