Multivariate Analysis of Trace Elements from Coral Cores Sheena Weller MARS6300 4/24/18
Dataset Timeseries of trace elements for three coral cores Samples Measured as depth downcore (cm) Different sample sizes Variables 21 trace elements Li/Ca, B/Ca, Mg/Ca, Al/Ca, Mn/Ca, Cu/Ca, Zn/Ca, Sr/Ca, Cd/Ca, Ba/Ca, Pb/Ca, U/Ca, Na/Ca, Fe/Ca, V/Ca, Co/Ca, Ni/Ca, Rb/Ca, Mo/Ca, Sn/Ca, Sb/Ca
What trace elements explain the most shared variability? 9 1 2 For all three cores combined For each core separately “The goal of this project was to …” determine the temporal and spatial patterns of shared variability in all trace elements measured, between cores and for each core separately Predictions: H0: H0 = There is no difference between trace elements HA = There is a difference between trace elements
Original Data Matrix - Combination of Three Cores 81 samples (depth downcore) 21 trace element to Ca ratios Quantitative data Core 1: 37 Samples 21 Elements Core 2: 18 Samples Core 9: 26 Samples
Summary Analysis Transformation: Relativization: log (x + 1) general No Outliers, no empty cells, no deleted cells General / sum I propose a log (x+1) transformation first since we have values ranging from 0 – 1500 (used log (x+1) so there are no negative data) Then a general relativization was preformed so that all variables have equal weight The three highest concentrations found across all three cores are U/Ca, B/Ca, and V/Ca Lowest concentrations found across all three cores Mn/Ca, Sn/Ca, and Co/Ca
New Matrix after Transformation and Relativization Ordination Method chosen NMDS Non parametric data
NMDS Analysis Autopilot / medium (50 runs) 2-D solution In order to determine the temporal and spatial patterns of shared variability in all trace elements measured, a non-metric multidimensional scaling (NMDS) analysis was run on elements of the three cores together. P- values Stress reduction rule Final stress 12.184 = according to clarks rule of thumb this data corresponds tp a usable picture Autopilot / medium (50 runs) 2-D solution Final stress: 12.184
NMS Ordination Plot Results Most Variance Explained (tau) Axis 1 Fe, Mn, Cu, Co, Al, Sn, Ba, Cd Axis 2 Cu, Co, Sn, Al Axis 1 explains 78% of the variance and is described by variability in Fe, Mn, Cu, Co, Al, Sn, Cd, Ba Axis 1 and axis 2 are completely independent from each other Second axis explains the difference, more so then the first one, changing more on the second 0-14
Inferences Axis 1 - Fe, Al, Mn, Ba, Cu, Co, Sn, Cd Indicators Al, Mn, Ba, Mg, and Fe: Reflect changes in Sedimentation that are directly related to development Ba, Mn: River discharge, precipitation Axis 2 - Cu, Co, Cd and Sn Indicators of heavy metal Directly related to increasing industrial activity, population, and tourism
Separate NMS Analysis of each Core
Correlations with Main Matrix Core 1 Core 2 Core 9 2-D Axis 1: Al, Mn, Zn, V, Fe, Ba, Pb, U, Mo Axis 2: Co, Cu, Sn, Na, Cd, Fe, Zn 3-D Axis 1: Al, Mn, Zn, Cd, Pb, Fe, V, Sb Axis 2: Cu, V, Co, Mo, Sn Axis 3: Pb, Fe, Mo 2-D Axis 1: Al, Mn, Zn, Fe, V, Co Axis 2: Zn, Fe, Sn, Sb 001 002 009 : all have Al, Mn, Fe (Ba, Cd, Co) 001 and 002 show more similarities – makes sense since there from the same location Have higher concentrations of heavy metals for the resuspended area 001/002 Combined Axis 1 : Fe, Mn, Cu, Co, Al, Sn, Ba, Cd Axis 2: Cu, Co, Sn, Al
Results Core 1 has the highest trace element concentrations, followed by core 2 and then core 9 More similarities between core 1 and core 2 Core 9 has fewer heavy trace elements incorporated Most variability for all three cores explained by Al, Mn, and then Fe Closest to urbanization makes sense since there from the same location, experience similar environmental conditions
Next Step - Reanalysis Follow up with ISA MRPP Non-parametric data H0: There is no difference between trace elements of the tree coral cores Grouping variable: cores Follow up with ISA MRPP is a non para testing the hypothesis of no differences between two or more groups Grouping variable will be the three cores Likely to see that there is a significant difference between the cores Could still do this: Do NMDS on important elements only (Al, Fe, Sn, Mn -> those with high loadings only (top 8-10 elements of axes) Do NMDS on temperature/salinity trace elements