Electronic Supplementary Material for Ecotoxicology Publication Of

Electronic Supplementary Material for Ecotoxicology Publication of:

Spatiotemporal Trends of Mercury in Walleye and Largemouth Bass from the Laurentian Great Lakes Region

Bruce A. Monson1a, David F. Staples2, Satyendra P. Bhavsar3, Thomas M. Holsen4, Candy S. Schrank5, Sara K. Moses6, Daryl J. McGoldrick7, Sean M. Backus7, and Kate A. Williams8

1Minnesota Pollution Control Agency, Saint Paul, Minnesota

2Minnesota Department of Natural Resources, Forest Lake, Minnesota

3Ontario Ministry of the Environment, Toronto, Ontario

4Clarkson University, Potsdam, New York

5Wisconsin Department of Natural Resources, Madison, Wisconsin

6Great Lakes Indian Fish and Wildlife Commission, Odanah, Wisconsin

7Environment Canada, Burlington, Ontario

8BioDiversity Research Institute, Gorham, Maine

aCorresponding author: , tel.651-757-2579, fax 651-297-8676

This supplements the summary in Table 4.The mixed model was based on the hierarchical structure of the data, known factors that likely affect sample Hg (e.g., fish size, waterbody type, season, and sample type), and our parameters of interest (namely annual trend, but also the random yearly deviations and lake-specific deviations from the average trend). Information criteria model selection was not performed because we were mainly interested in describing the temporal patterns in the data, not in evaluating potential factors correlated with the changes. Both random effects for year and trend are important because the former allows for within-year correlations in Hg levels, while the random lake*year factor allows each lake to have a unique trend (the fixed effect estimate is the ‘average’ trend among the sites). We used maximum likelihood (ML) estimates rather than restricted maximum likelihood (REML) because we wanted joint optimization over fixed and random effects, both of which were parameters of interest here. REML if often preferred because ML is biased low for variance estimates; however, the ML bias is negligible for large sample sizes. Given the size of the dataset and the large number of sites, the bias here is not likely an issue; we had essentially the same parameter estimates for all models whether fit by REML or ML.

The following is R code and output for the mixed model analyses of fish-Hg concentrations for largemouth bass and walleye from the MercNet and OMOE datasets.

R code and output for mixed model analyses of fish-Hg levels

MercNet Data

names(Data.MercNet) # Data file variables

# "HG_ppm" = measured Hg levels

# "YEAR" = Sample Year

# "ORG_COMNAME" = Common name for fish species

# "LNG_CM" = Fish Length

# "SEASON" = Season in which sample was taken

# "TIS_PART" = Tissue part (filet skin on, whole body, etc)

# "STAID" = Sample Station Identification code

# "TYPE" = Type of Habitat (Lake, River, etc)

Data.MercNet.wae = subset(Data.MercNet, ORG_COMNAME == “walleye”)

Data.MercNet.lmb = subset(Data.MercNet, ORG_COMNAME == “largemouth bass”)

#------

### Model of Walleye log(Hg) with YEAR centered at 1995

#------

MercNet.wae.log = lmer(log(HG_ppm) ~ YEAR + TIS_PART*LNG_CM + TYPE +

SEASON + (1|YEAR) + (YEAR+LNG_CM|STAID),

data = Data.MercNet.wae, REML=F)

summary(MercNet.wae.log)

#Linear mixed model fit by maximum likelihood

#Formula: log(HG_ppm) ~ YEAR + TIS_PART * LNG_CM + TYPE + SEASON +

# (1 | YEAR) + (YEAR + LNG_CM | STAID)

#Random effects:

# Groups Name VarianceStd.Dev.Corr

# STAID (Intercept) 0.317360.56356

# YEAR 0.000890.029770.140

# LNG_CM 0.000150.01234-0.322 -0.143

# YEAR (Intercept) 0.004840.06961

# Residual 0.105830.32531

#Number of obs: 21358, groups: STAID, 2202; YEAR, 39

#

#Fixed effects:

# EstimateStd. Errort value

#(Intercept) -1.172220.02200-53.27

#YEAR -0.007810.00191-4.08

#TIS_PARTfillet skin off 0.048790.02334 2.09

#TIS_PARTwhole body -0.147030.03478-4.23

#LNG_CM 0.044530.0005187.71

#TYPEGLK -0.862050.05767-14.95

#TYPEIMP -0.055650.04725-1.18

#TYPERIV -0.082390.04015-2.05

#SEASONFall -0.186290.01405-13.26

#SEASONSummer -0.086960.01520-5.72

#SEASONWinter -0.103650.04313-2.40

#TIS_PARTfillet skin off:LNG_CM -0.002850.00101-2.82

#TIS_PARTwhole body:LNG_CM -0.006970.00238-2.92

#

#Correlation of Fixed Effects:

# (Intr) YEAR TIS_PARTfso TIS_PARTwb LNG_CM TYPEGL TYPEIM TYPERI

#YEAR 0.079

#TIS_PARTfso -0.199 0.044

#TIS_PARTwhb -0.031 0.030 0.032

#LNG_CM -0.153 -0.029 0.058 0.048

#TYPEGLK -0.187 -0.034 0.050 -0.074 -0.029

#TYPEIMP -0.240 0.032 0.060 -0.011 -0.002 0.088

#TYPERIV -0.266 -0.007 0.072 -0.025 0.024 0.108 0.123

#SEASONFall -0.222 0.001 0.082 0.054 0.005 -0.050 -0.007 -0.050

#SEASONSummr -0.333 0.033 0.090 0.034 0.007 0.030 0.061 -0.018

#SEASONWintr -0.048 0.007 0.007 0.015 -0.004 -0.013 -0.007 -0.022

#TIS_PARTso: 0.018 -0.022 -0.183 -0.025 -0.394 0.025 0.011 0.007

#TIS_PARTb:L 0.027 -0.027 -0.017 -0.394 -0.118 -0.027 0.003 -0.001

# SEASONF SEASONS SEASONW TIS_PARTso:

#YEAR

#TIS_PARTfso

#TIS_PARTwhb

#LNG_CM

#TYPEGLK

#TYPEIMP

#TYPERIV

#SEASONFall

#SEASONSummr 0.433

#SEASONWintr 0.086 0.093

#TIS_PARTso: 0.007 0.014 0.008

#TIS_PARTb:L -0.003 -0.009 0.002 0.051

#------

### Model of MercNet Largemouth Bass log(Hg) with YEAR centered at 1995

#------

MercNet.lmb.log = lmer(log(HG_ppm) ~ YEAR + TIS_PART*LNG_CM + TYPE +

SEASON +(1|YEAR) + (YEAR+LNG_CM|STAID),

data = Data.MercNet.lmb,REML=F)

summary(MercNet.lmb.log)

#Linear mixed model fit by maximum likelihood

#Formula: log(HG_ppm) ~ YEAR + TIS_PART * LNG_CM + TYPE + SEASON +

# (1 | YEAR) + (YEAR + LNG_CM | STAID)

#Random effects:

# Groups Name VarianceStd.Dev.Corr

# STAID (Intercept) 0.287260.53596

# YEAR 0.001190.034440.049

# LNG_CM 0.000530.023100.197 -0.015

# YEAR (Intercept) 0.006960.08345

# Residual 0.078160.27958

#Number of obs: 7728, groups: STAID, 1959; YEAR, 34

#Fixed effects:

# EstimateStd.Errort value

#(Intercept) -0.851680.02782-30.62

#YEAR -0.006280.00274-2.29

#TIS_PARTfillet skin off 0.114040.087781.30

#TIS_PARTwhole body -0.414720.07532-5.51

#LNG_CM 0.052420.0010748.84

#TYPEGLK -0.329540.18311-1.80

#TYPEIMP -0.172940.03997-4.33

#TYPERIV -0.275630.03475-7.93

#SEASONFall -0.197170.02140-9.21

#SEASONSummer -0.042480.02142-1.98

#SEASONWinter -0.164360.16777-0.98

#TIS_PARTfillet skin off:LNG_CM 0.007630.008360.91

#TIS_PARTwhole body:LNG_CM -0.013950.00517-2.70

#

#Correlation of Fixed Effects:

# (Intr) YEAR TIS_PARTfso TIS_PARTwb LNG_CM TYPEGL TYPEIM TYPERI

#YEAR -0.062

#TIS_PARTfso -0.044 0.063

#TIS_PARTwhb -0.061 0.037 0.116

#LNG_CM 0.253 0.009 -0.061 -0.067

#TYPEGLK -0.057 -0.001 -0.020 -0.012 0.004

#TYPEIMP -0.305 -0.065 -0.040 0.001 0.007 0.052

#TYPERIV -0.311 -0.063 -0.016 -0.027 0.056 0.069 0.280

#SEASONFall -0.289 0.034 -0.011 0.044 0.013 -0.020 -0.041 -0.093

#SEASONSummr -0.253 0.049 -0.026 0.034 0.046 -0.044 -0.025 -0.178

#SEASONWintr -0.037 -0.001 -0.007 0.004 0.005 0.004 0.019 0.007

#TIS_PARTso: -0.024 -0.008 0.537 0.076 -0.091 0.012 0.029 0.003

#TIS_PARTb:L -0.045 -0.011 0.096 0.726 -0.135 -0.006 0.003 0.006

# SEASONF SEASONS SEASONW TIS_PARTso:

#YEAR

#TIS_PARTfso

#TIS_PARTwhb

#LNG_CM

#TYPEGLK

#TYPEIMP

#TYPERIV

#SEASONFall

#SEASONSummr 0.406

#SEASONWintr 0.027 0.032

#TIS_PARTso: -0.015 0.009 0.019

#TIS_PARTb:L 0.029 0.008 0.001 0.110

Ontario Data

names(Data.Ontario) # Data file variables

# "HG_ppm" = measured Hg levels

# "YEAR" = Sample Year

# "ORG_COMNAME" = Common name for fish species

# "LENGTH" = Fish Length

# "Season" = Season in which sample was taken

# "STAID" = Sample Station Identification code

# "Habitat" = Type of Habitat (Lake, River, etc)

Data.Ontario.wae = subset(Data.Ontario, ORG_COMNAME == “walleye”)

Data.Ontario.lmb = subset(Data.Ontario, ORG_COMNAME == “largemouth bass”)

#------

### Model of Ontario Walleye log(Hg) with YEAR centered at 1990

#------

Ontario.wae.log = lmer(log(HG_ppm) ~ YEAR + LENGTH + Season + Habitat +

(1|YEAR) + (YEAR+LENGTH|STAID),

data = Data.Ontario.wae, REML=F)

summary(Ontario.wae.log)

#Linear mixed model fit by maximum likelihood

#Formula: log(HG_ppm) ~ YEAR + LENGTH + Season + Habitat +

#(1 | YEAR) + (YEAR + LENGTH | STAID)

#Random effects:

# Groups Name Variance Std.Dev. Corr

# STAID (Intercept) 0.25229786 0.502293

# YEAR 0.00044187 0.021021 0.147

# LENGTH 0.00012558 0.011206 -0.273 -0.139

# YEAR (Intercept) 0.01609376 0.126861

# Residual 0.11393635 0.337545

#Number of obs: 28487, groups: STAID, 937; Yr, 39

#

#Fixed effects:

# Estimate Std. Error t value

#(Intercept) -0.7192258 0.0315267 -22.81

#YEAR -0.0069216 0.0022655 -3.06

#LENGTH 0.0447489 0.0004854 92.19

#SeasonFall -0.1029722 0.0118443 -8.69

#SeasonSummer -0.0529025 0.0108895 -4.86

#SeasonWinter -0.0957715 0.0158871 -6.03

#HabitatGreat Lake -0.7342368 0.0549227 -13.37

#HabitatRiver -0.1132689 0.0449500 -2.52

#

#Correlation of Fixed Effects:

# (Intr) YEAR LENGTH SsnFll SsnSmm SsnWnt HbttGL

#YEAR -0.036

#LENGTH -0.114 -0.041

#SeasonFall -0.207 -0.011 0.004

#SeasonSummr -0.233 -0.017 0.007 0.579

#SeasonWintr -0.220 0.114 0.005 0.401 0.412

#HabittGrtLk -0.270 -0.002 -0.012 0.000 0.000 0.038

#HabitatRivr -0.334 -0.018 0.014 0.013 0.006 0.025 0.190

#------

### Model of Ontario Largemouth Bass log(Hg) with YEAR centered at 1990

#------

Ontario.lmb.log = lmer(log(HG_ppm) ~ YEAR + LENGTH + Season + Habitat +

(1|YEAR) + (YEAR+LENGTH|STAID),

data = Data.Ontario.lmb, REML=F)

summary(Ontario.lmb.log)

#Linear mixed model fit by maximum likelihood

#Formula: LogHg ~ YEAR + LENGTH + Season + Habitat +

(1 | YEAR) + (YEAR + LENGTH | STAID)

#Random effects:

# Groups Name Variance Std.Dev. Corr

# STAID (Intercept) 0.16165885 0.402068

# YEAR 0.00048309 0.021979 0.059

# LENGTH 0.00016962 0.013024 -0.392 -0.531

# YEAR (Intercept) 0.01396099 0.118157

# Residual 0.10363661 0.321926

#Number of obs: 4701, groups: STAID, 316; Yr, 37

#

#Fixed effects:

# Estimate Std. Error t value

#(Intercept) -1.177548 0.041855 -28.13

#YEAR -0.008706 0.003023 -2.88

#LENGTH 0.059835 0.001183 50.56

#SeasonFall -0.097823 0.032848 -2.98

#SeasonSummer -0.037326 0.027225 -1.37

#SeasonWinter 0.002970 0.045215 0.07

#HabitatGreat Lake -0.233435 0.098195 -2.38

#HabitatRiver -0.111552 0.065565 -1.70

#

#Correlation of Fixed Effects:

# (Intr) YEAR LENGTH SsnFll SsnSmm SsnWnt HbttGL

#YEAR -0.142

#LENGTH -0.147 -0.187

#SeasonFall -0.316 0.051 0.009

#SeasonSummr -0.385 -0.012 0.033 0.522

#SeasonWintr -0.377 0.299 0.005 0.377 0.325

#HabittGrtLk -0.227 -0.056 -0.013 -0.008 0.017 0.022

#HabitatRivr -0.349 -0.032 0.035 -0.008 0.026 0.043 0.150