I A E A ANAL YS IS OF THE CANADIAN BENTHIC V i DATABAS E e n n a Claire Della - Vedova (magelis company) 8 Jacqueline Garnier – Laplace (IRS N) s e p t e m b e EMRAS II Meeting, Vienna, Austria 06-09 S eptember 2010 r 2 0 1 0 1 1
RECALL OF THE CONTEXT I A The Fisheries Act of Canada requires the Minister to protect fish E populations, their habitat and their resources (Edible fish tissue). A V i Benthos is seen as food item for fish and must be protected or not e affected by releases of deleterious substances n n a Benthic database from Uranium mines and mill operation : 8 s - Benthos is identified at the family level but most e generally at the gender/ species level (presence/ absence – p t number of individuals ? ) e m - This is coupled with measurements of radiological and b e non radiological contaminants in water and sediment r - The study areas are mostly located in S askatchewan and 2 Ontario, Canada 0 1 0 2
WHAT HAD ALREADY BEEN DONE I A Univariate approach (contaminant by contaminant) : E A V i e n n a 8 s e p t S creening Level Concentration (S LC) method was used to derive Lowest e Effect Level (LEL) and S evere Effect Level (S EL) concentration for 9 metals m b and 3 radionuclides e r 2 0 1 0 3
WHAT HAD ALREADY BEEN DONE I A S creening Level Concentration approach (univariate, i.e contaminant E by contaminant) : A For a given specie (Acarina) All the SSLC for a given contaminant V Frequence distribution of contaminant [] Frequence distribution of SSLC i 100% 100% e 95% 90% n n a Site 35 8 s e 5% Site 12 p 0% [Ura] [Ura] 0% t e LEL SEL SSLC m b If the sediment contamination is below the Lowest effect level (LEL), e environmental risk is likely to be minimal and overruled r If the sediment contamination is above the S evere effects level, there is concern 2 about environmental risks 0 1 0 4
WHAT HAVE WE DONE ? I A Multivariate and global approach : E A We tried to demonstrate whether changes in species diversity of the benthos community may be explained by changes in contaminants V i concentrations in sediment. e n n a Plan : 8 s e - Methods used: matrix building, statistical analyses (R software, package {vegan}) p t - Obtained results e m - Conclusion b e - Future work r 2 0 1 0 5
METHOD - DATABAS ES I A Original database – sent by the CNS C (S teve Mihok) : E A V i e n n a 2270 lines, 1 line per Taxa and Station 8 s e p t Species matrix (presence / absence) Contaminants matrix [ ] sediment ) e ……………. ……………. Specie1 Specie209 Cont1 Cont12 m ……………. ……………. b Sta 1 Station 1 0 1 [ ] [ ] e ……. ……. ……………. ……………. [ ] r [ ] 1 0 [ ] Sta 196 Sta 196 [ ] ……………. ……………. 1 1 2 0 1 0 6
METHOD – S PECIES MATRIX I A Features of the species matrix (presence / absence) E A V i e n n a 8 s e p t e • 196 stations, 209 benthic species (Ablabesmyia – Zapada) m b • 96% of values equal to “0” e r 2 0 1 0 7
METHOD – CONTAMINANT MATRIX I A Features of the contaminants matrix ( [] sediment ) - 1 E A V i e n n a 8 s • 196 stations, 12 contaminants (uranium, arsenic, molybdenum, nickel, e lead, selenium,radium 226 , lead 210 , polonium 210 ) p t e • 27% of NA (empty cells) but with big difference between contaminants : m b -uranium : 2.04%, arsenic : 15.82%, molybdenum : 52.55%, nickel :0%, e r lead : 5.1%, selenium : 35.71%, radium 226 : 0.51%, lead 210 : 44.9%, polonium 210 : 50%, copper : 51%, vanadium : 15.82%, chrome : 62.24% 2 0 1 0 8
METHOD – S PECIES MATRIX I A Features of the contaminants matrix ( [] sediment ) - 2 E A V i e n n a 8 s • Finally, only 31 stations out of 196 (i.e 16%) have measurements for all of e the 12 contaminants We can consider two datasets p t - a first one built-on with the species and chemical matrix corresponding e m of the 31 sites without NA values (in chemical matrix) = "complete data" b - a second one built on with the species and chemical matrix of all the e r 196 sites = "all data" 2 0 1 0 9
METHOD – S TATIS TICAL ANALYS IS I A What we decided to do : E A 1. Investigate the contaminants which influence the distribution of the species V by means of ordination methods classically used in this situation but which i can be applied only to datasets containing no missing data, so to our "compete e data" set : n n a) constrained ordination method (Redundancy Analysis - RDA) and a b) unconstrained ordination method (Principal Components Analysis-PCA) 8 s with vectors fitting approach e p t 2. And to develop a method allowing to bring to light the contaminants which e influence the distribution of the benthos, even when the dataset contains m b missing data e r a) use the developed method with "all data" set 2 b) Use the developed method with "complete data" set 0 1 0 10
RDA – Recall1 I Redundancy analysis (RDA) A E A 1. Redundancy analysis (RDA) is an asymmetric form of canonical analysis V where a response matrix (here : species) has to be explained by an i e explanatory matrix ( here : contaminants). n n a Y 8 X s (sites * species) (sites * cont) e p t e m b e r 2 0 1 0 11
RDA – Recall2 I Redundancy analysis (RDA) A E A 2. Redundancy analysis (RDA) is a constrained ordination method "Ordination is the arrangement of units of some order. (…) For the purpose of analysis, V i ecologists therefore project the multidimensional scatters diagram onto bivariate graph e whose axes are known to be of particular interest. The axes of these graphs are chosen n to represent a large fraction of the variability of the multidimensionnal data matrix in a n space with reduced dimensionality relative to the original dataset (Legendre et al a 1998)". 8 s RDA Y X e Ordination triplot p (sites * species) (sites * cont) t e m 3. RDA is constrained ordination : it means that only the variation that can b be explained by the use of environmental variables, or constraints is e r displayed. 4. RDA is based on Euclidean distances. 2 0 1 0 12
RDA – Recall3 I A Redundancy analysis (RDA) E A RDA can be seen as an extension of Principal Components analysis (PCA) because the canonical ordination vectors are linear combinations of the V response variable Y . i e n RDA may be understood as a 2-step process : n 1. Regress each variable in Y on all variables in X and compute the fitted values a 8 Fitted values from the Y X s multiple regression e Ŷ =X[X'X] -1 X'Y (sites * species) (sites * cont) p t 2. Carry out a PCA of the matrix of the fitted values to obtain the eigen values and e eigen vectors. Two ordinations are obtained : m b ordination in the space e Fitted values from the of variable Y PCA U = matrix of r multiple regression eigen vectors Ŷ =X[X'X] -1 X'Y 2 ordination in the space 0 of variable X 1 0 13
RDA with "complete data" set I 1 st step = use of hellinger transformation for the species matrix A E A Raw data Hellinger Y=Transformed data X = transformation* V (sites X species) (sites X species) (sites X contaminants) i e n Ordination plot RDA n a 8 Representation of elements : s - sites : black number e - species : red names p - contaminants : blue arrows t e m b e r 2 0 1 0 14
RDA with "complete data" set I A Rules of interpretation : E A 1. The orthogonal proj ection of an obj ect (site) on a quantitative V explanatory variable (contaminant) i e approaches the value of this variable n in the obj ect (site). n a 2. The angle between a response variable (benthic species) and an 8 explanatory variable (contaminant) in s e the biplot reflects their correlation. p Those between 2 explanatory t variables has no meaning ; the same e m between two response variables b (species). e r 2 0 1 0 15
Recommend
More recommend