Several ecological indices were calculated at the replicate level using the OTOT dataset (n = 7,305). For indices calculations see http://rpubs.com/javivi77/index_calculator.

Traditional indices included:

Several indices based on ecological groups (EG) were calculated, using the original AZTI EG complemented either with Keeley (EG_K_AZTI_AB) or Robertson (EG_R_K_AZTI_AB) EGs and Anna’s taxa matching, including:

For the BQI sensitivity values were calculated for taxa that occurred in >5% of the samples and had N>50 individuals per sample. Below are the sensitivity values for each taxa.

Exploration of metadata variables and time

Time

## # A tibble: 3 × 1
##                                                       cesym
##                                                       <chr>
## 1 environmentcanterburyccc avonheathcote heathcote 2007 all
## 2                  environmentsouthland newriver e 2012 all
## 3                  environmentsouthland newriver f 2012 all

Vegetation

Most data come from unvegetated sites. There is no apparent differnce in abundance and AMBI due to vegetation.

Tides

No evident effect of tide. Most samples has no tidal height recorded.

Grain size

No effect evident effect of grain size method and mud relationship.

Typology

Samples are mostly from estuaries of typology7A, 8 and 9. If typology is to be incooporated as a categorical predictor data would need to be truncated to these three clases.

7A. Tidal lagoon permanently open 8. Shallow drowned valley 9. Deep drowned valley

Taxonomist

Coast

Nitrogen type

Geographic distrution of samples

Environmental variables

Correlations among environmental variables

##        mud TOC  AFDW  Cu   Cr   Zn     Ni     Pb     Cd     TN    TP
## mud   1.00 0.6  0.72 0.5 0.31 0.31  0.261  0.238 -0.068  0.660 0.621
## TOC   0.63 1.0  0.77 0.7 0.24 0.37  0.181  0.149  0.212  0.877 0.680
## AFDW  0.72 0.8  1.00 0.4 0.31 0.32  0.054  0.146 -0.059  0.706 0.450
## Cu    0.52 0.7  0.42 1.0 0.44 0.66  0.210  0.619  0.106  0.567 0.737
## Cr    0.31 0.2  0.31 0.4 1.00 0.41  0.741  0.160  0.065  0.244 0.501
## Zn    0.31 0.4  0.32 0.7 0.41 1.00  0.025  0.670  0.219  0.419 0.669
## Ni    0.26 0.2  0.05 0.2 0.74 0.02  1.000 -0.009 -0.051  0.080 0.286
## Pb    0.24 0.1  0.15 0.6 0.16 0.67 -0.009  1.000  0.092  0.388 0.556
## Cd   -0.07 0.2 -0.06 0.1 0.06 0.22 -0.051  0.092  1.000 -0.008 0.008
## TN    0.66 0.9  0.71 0.6 0.24 0.42  0.080  0.388 -0.008  1.000 0.607
## TP    0.62 0.7  0.45 0.7 0.50 0.67  0.286  0.556  0.008  0.607 1.000

PCA metals and nutrients

RUN PCA to reduce the correlated metals and nutrient variables. 

## Importance of components:
##                           Comp.1     Comp.2     Comp.3
## Standard deviation     3.0509801 0.90738380 0.72109111
## Proportion of Variance 0.8738882 0.07729638 0.04881546
## Cumulative Proportion  0.8738882 0.95118454 1.00000000

Could remove a couple of Pb max values as outliers.

## Importance of components:
##                            Comp.1     Comp.2
## Standard deviation     15.2653494 4.45008502
## Proportion of Variance  0.9216749 0.07832509
## Cumulative Proportion   0.9216749 1.00000000

EG based indices

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##    0.00   35.71   46.15   47.50   57.14  100.00      13
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   0.000   9.091  14.290  14.960  20.000 100.000      13

Average indices data at the level of cesym

Plot indices vs. environmental variables

Other indices

Tradicional indices

Map of selected indices

Multiple stressors models

Models summary and validation

## Source: local data frame [5 x 12]
## Groups: index [5]
## 
##    index r.squared adj.r.squared     sigma statistic      p.value    df
##    <chr>     <dbl>         <dbl>     <dbl>     <dbl>        <dbl> <int>
## 1   AMBI 0.2878322     0.2766346 0.8163589  25.70479 8.767908e-22     6
## 2 AMBI_S 0.4027984     0.3914949 0.6417415  35.63483 6.928344e-33     7
## 3 BENTIX 0.2059659     0.1883765 1.5194404  11.70969 2.968629e-13     8
## 4 M_AMBI 0.2865573     0.2753397 0.1410693  25.54521 1.157626e-21     6
## 5 MEDOCC 0.2543139     0.2402000 1.0642740  18.01865 5.420494e-18     7
## # ... with 5 more variables: logLik <dbl>, AIC <dbl>, BIC <dbl>,
## #   deviance <dbl>, df.residual <int>

Coefficient plots for multiple stressors

Boosted regression trees

##                 var   rel.inf
## mud             mud 85.109646
## metals       metals 12.515589
## nutrients nutrients  2.374765

GAM

## 
## Family: gaussian 
## Link function: identity 
## 
## Formula:
## AMBI_S ~ metals + nutrients + mud + s(mud, metals) + s(metals, 
##     nutrients) + s(mud, nutrients)
## 
## Parametric coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  0.972056   0.074676  13.017   <2e-16 ***
## metals      -0.037769   0.014727  -2.565   0.0108 *  
## nutrients   -0.001924   0.009111  -0.211   0.8329    
## mud          0.068558   0.004231  16.205   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Approximate significance of smooth terms:
##                        edf Ref.df      F  p-value    
## s(mud,metals)        5.052  7.125 15.783  < 2e-16 ***
## s(metals,nutrients)  3.927  5.871  0.741     0.51    
## s(mud,nutrients)    13.507 27.000  1.547 4.24e-08 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Rank: 85/88
## R-sq.(adj) =  0.468   Deviance explained =   51%
## GCV = 0.39078  Scale est. = 0.3596    n = 324