CHASE: an integrated regional assessment of the status of contaminants and biological effects in biota and sediment

Overview

This report integrates the regional status assessments of contaminants and biological effects in biota and sediment from the 2022 OSPAR CEMP Assessment. The integration is based on the CHASE approach (Andersen et al., 2010). This was first developed for the HELCOM Holistic Assessment in 2010 with the aim of assessing the ‘overall status’ of the marine environment in relation to hazardous substances. Essentially, CHASE takes a set of indicators expressed as the ratio of concentrations to environmental thresholds, adds them, and divides by the square root of the number of indicators. The resulting values are known as CHASE contaminant scores. CHASE scores less than 1 indicate ‘good’ status and above 1 indicate ‘poor’ status.

For the 2022 OSPAR CEMP Assessment, the CHASE scores for biota are based on contaminant ratios for:

The CHASE scores for sediment are based on contaminant ratios for:

Separate CHASE scores are calculated for biota and sediment. These values are then combined.


Methods

CHASE contaminant scores

Suppose there are n indicators and, for any given region, \(y_i\) is the contaminant ratio for indicator \(i\). For example, for biota, \(y_i\) could be the ratio of the mean mercury concentration in the final monitoring year to the Quality Standard secondary poisoning, or the ratio of the mean polyaromatic hydrocarbon concentration to the Environmental Assement Criterion. The CHASE contaminant score (CS) is then defined as:

\[\text{CS} = \frac 1 {\sqrt n} \sum_{i=1}^n y_i\] The square root is used so that the impact of indicators with high ratios (e.g. above one) cannot be masked by indicators with low ratios. Other ways of combining the contaminant ratios have been investigated, but the method above has proved the most robust (Andersen et al, 2016).


Filling in missing ratios

In practice, some regions have an incomplete set of contaminant ratios. Ignoring these missing ratios can make it hard to compare the CHASE scores across regions. One option would be to only consider regions with a full set of ratios, but this is overly restrictive. An alternative, adopted here, is to use a statistical model to ‘fill-in’ the missing contaminant ratios using the observed ratios. Regions which require half or more of their ratios to be filled-in are discarded at this stage because they do not have a strong enough evidence base to construct a meaningful CHASE score.

The fill-in procedure is based on an additive model of the log contaminant ratios with region and indicator as fixed effects. The ratios are log transformed because they were originally estimated on this scale and because it leads to reasonably homogenous variances. The missing ratios are filled in using the predictions from this model. The idea behind the model is that a missing ratio will be predicted to be low / high if the ratios for the same indicator tend to be low / high in other regions. Further the missing ratio will be adjusted lower / higher depending on whether the observed ratios in the region tend to be low / high compared to other regions. CHASE scores for regions with missing ratios tend to have wider confidence limits (see below) and their interpretation should be moderated accordingly.

Formally, the fill-in model is a linear mixed model with normally distributed errors and:

  • response: log contaminant ratio
  • fixed model: region + indicator
  • random model: estimation variance + residual variation

The estimation variance is the variance of the contaminant ratio estimates from the indicator regional assessments, assumed known and fixed. The residual variation is the variation that cannot be explained by any of the fixed effects or the other random effects.

Some modifications are required for the contaminant ratios for VDS in biota. These are transformed to the square root scale (rather than the log scale) since they were originally estimated on this scale. However, this sometimes results in inadmissible negative predictions of the missing square root contaminant ratios. Since these negative predictions reflect very low values of VDS, a pragmatic solution is to replace the corresponding missing VDS contaminant ratios by one quarter of the lowest observed VDS contaminant ratio. To illustrate (see the results later), the VDS fill-in for the Barents Sea is one quarter of the VDS contaminant ratio for the Southern North Sea.


Confidence limits

Confidence limits on the CHASE scores are calculated by simulating from the posterior distribution of the indicator contaminant ratios. (The log contaminant ratios are estimated in the regional assessments and their posterior distribution is assumed to be Gaussian with variance equal to the covariance of the parameter estimates.) Each simulation consists of the following stages:

  • for each indicator, the contaminant ratios for each region are simulated from their posterior distribution
  • missing ratios are filled-in
  • the CHASE contaminant score iss calculated for each region

Confidence limits are then obtained from the simulated distributions of the regional CHASE scores.


Interpretation

Environmental status is based on the point estimates of the CHASE scores, with the confidence limits used to qualify these assessments where appropriate. This is in contrast to the indicator regional assessments where good status is based on a precautionary test using an upper confidence limit on the contaminant ratios. The justification for using the point estimate is that the square root in the denominator of the formula for calculating the CHASE scores is itself precautionary.


Biota

The results are presented in three stages:

  • contaminant ratios for each indicator and region
  • CHASE scores with confidence limits for each region
  • trends in the CHASE scores


Contaminant ratios

The first tab below shows the contaminant ratios for each indicator. They have been back-transformed from the log scale (contaminants) and square root scale (VDS).

The second tab shows the contaminant ratios with the missing values filled-in. The Gulf of Cadiz and the Norwegian Sea drop out of the assessment because they have too many missing ratios.

The third tab plots the contaminant ratios, with the filled-in values shown in red.

Observed ratios
region HG PAH np_PCB CB118 PBDE VDS
Barents Sea 1.5533 0.0394 0.5210 0.0032
Greenland-Scotland ridge 1.2247 0.0053 0.0123 0.1702 0.0036
Norwegian Sea 2.3833 0.0860 1.3725
Norwegian Trench 3.8727 0.0451 0.6626 0.0081
Northern North Sea 5.4743 0.0461 0.0640 0.9505 0.0088 0.1815
Skagerrak and Kattegat 2.9039 0.0245 0.0666 0.9703 0.0039 0.7136
Southern North Sea 4.4890 0.0331 0.1139 1.8224 0.0069 0.0076
Channel 3.6704 0.0208 0.0892 1.8018 0.0025 0.1317
Irish and Scottish West Coast 2.1132 0.0153 0.0095 0.1238 0.0081 0.0211
Irish Sea 3.1443 0.0461 0.0864 1.0350 0.0130 0.0470
Celtic Sea 2.1987 0.0172 0.0297 0.4578 0.0031 0.1151
Northern Bay of Biscay 4.0098 0.0187 0.0545 0.7242 0.0014 0.0924
Iberian Sea 2.4482 0.0209 0.0489 1.0738 0.0056 1.3864
Gulf of Cadiz 1.2963



Filled-in ratios
region HG PAH np_PCB CB118 PBDE VDS
Barents Sea 1.5533 0.0145 0.0394 0.5210 0.0032 0.0019
Greenland-Scotland ridge 1.2247 0.0053 0.0123 0.1702 0.0036 0.0019
Norwegian Trench 3.8727 0.0239 0.0451 0.6626 0.0081 0.1779
Northern North Sea 5.4743 0.0461 0.0640 0.9505 0.0088 0.1815
Skagerrak and Kattegat 2.9039 0.0245 0.0666 0.9703 0.0039 0.7136
Southern North Sea 4.4890 0.0331 0.1139 1.8224 0.0069 0.0076
Channel 3.6704 0.0208 0.0892 1.8018 0.0025 0.1317
Irish and Scottish West Coast 2.1132 0.0153 0.0095 0.1238 0.0081 0.0211
Irish Sea 3.1443 0.0461 0.0864 1.0350 0.0130 0.0470
Celtic Sea 2.1987 0.0172 0.0297 0.4578 0.0031 0.1151
Northern Bay of Biscay 4.0098 0.0187 0.0545 0.7242 0.0014 0.0924
Iberian Sea 2.4482 0.0209 0.0489 1.0738 0.0056 1.3864



Ratio plot



CHASE scores

The first tab plots the CHASE scores with pointwise two-sided 95% confidence limits. The scores have been coloured green if the point estimate is less than 1 and red otherwise. Scores less than 1 indicate good status. However, if the confidence limits also overlap 1, then this interpretation should be treated with some caution.

The second tab tabulates the CHASE scores and the 2.5, 5, 95 and 97.5 percentiles of their distribution. (The pointwise two-sided 95% confidence limits are the 2.5 and 97.5 percentiles.)

Scores



Estimates
region CHASE 2.5% 5% 95% 97.5%
Barents Sea 0.87 0.61 0.64 1.25 1.34
Greenland-Scotland ridge 0.58 0.38 0.41 0.86 0.90
Norwegian Trench 1.96 1.34 1.45 2.78 3.06
Northern North Sea 2.75 1.96 2.09 3.70 3.92
Skagerrak and Kattegat 1.91 1.48 1.54 2.46 2.61
Southern North Sea 2.64 1.98 2.06 3.38 3.56
Channel 2.33 1.75 1.82 3.03 3.16
Irish and Scottish West Coast 0.94 0.65 0.69 1.36 1.45
Irish Sea 1.78 1.37 1.43 2.34 2.44
Celtic Sea 1.15 0.84 0.88 1.57 1.67
Northern Bay of Biscay 2.00 1.42 1.51 2.79 2.91
Iberian Sea 2.03 1.63 1.70 2.52 2.65



Sediment

Contaminant ratios

The first tab shows the contaminant ratios for each indicator (back-transformed from the log scale).

The second tab shows the contaminant ratios with the missing values filled-in. The Celtic Sea drops out of the assessment because it has too many missing ratios.

The third tab plots the contaminant ratios, with the filled-in values shown in red.

Observed ratios
region CD HG PB PAH np_PCB CB118 PBDE
Northern North Sea 0.1220 0.4870 0.6830 0.0520 0.0260 0.2780 0.0010
Southern North Sea 0.3090 1.1480 1.3440 0.1060 0.0650 0.7720 0.0020
Channel 0.3640 1.1000 0.9950 0.0940 0.1350 1.9510 0.0000
Irish and Scottish West Coast 0.1160 0.2750 0.4720 0.0530 0.0180 0.1710 0.0010
Irish Sea 0.1610 0.8780 1.0120 0.1940 0.0870 0.7850 0.0020
Celtic Sea 0.2450 0.6450 0.8650
Iberian Sea 0.0860 0.6090 0.6030 0.0660 0.0290 0.3070



Filled-in ratios
region CD HG PB PAH np_PCB CB118 PBDE
Northern North Sea 0.1220 0.4870 0.6830 0.0520 0.0260 0.2780 0.0010
Southern North Sea 0.3090 1.1480 1.3440 0.1060 0.0650 0.7720 0.0020
Channel 0.3640 1.1000 0.9950 0.0940 0.1350 1.9510 0.0000
Irish and Scottish West Coast 0.1160 0.2750 0.4720 0.0530 0.0180 0.1710 0.0010
Irish Sea 0.1610 0.8780 1.0120 0.1940 0.0870 0.7850 0.0020
Iberian Sea 0.0860 0.6090 0.6030 0.0660 0.0290 0.3070 0.0010



Ratio plot



CHASE scores

The first tab plots the CHASE scores with pointwise two-sided 95% confidence limits. The scores have been coloured green if the point estimate is less than 1 and red otherwise. Scores less than 1 indicate good status.

The second tab tabulates the CHASE scores and the 2.5, 5, 95 and 97.5 percentiles of their distribution.

Scores



Estimates
region CHASE 2.5% 5% 95% 97.5%
Southern North Sea 1.42 1.30 1.32 1.54 1.55
Northern North Sea 0.62 0.53 0.55 0.73 0.76
Irish Sea 1.18 0.99 1.03 1.41 1.45
Irish and Scottish West Coast 0.42 0.32 0.33 0.55 0.59
Iberian Sea 0.64 0.57 0.58 0.72 0.73
Channel 1.75 1.55 1.59 1.97 2.02



Biota and sediment combined

CHASE scores for biota and sediment combined were calculated using the thirteen indicators (six for biota and seven for sediment). Trends were not investigated (a step too far!).

The first tab plots the CHASE scores. These have been coloured green if the point estimate is less than 1 and red otherwise. Scores less than 1 indicate good status.

The second tab tabulates the CHASE scores and the 2.5, 5, 95 and 97.5 percentiles of their distribution. The original CHASE scores for biota and sediment are also shown for comparison.

Scores



Estimates

region biota sediment CHASE 2.5% 5% 95% 97.5%
Northern North Sea 2.75 0.62 2.38 1.84 1.92 3.06 3.23
Southern North Sea 2.64 1.42 2.87 2.40 2.45 3.40 3.53
Channel 2.33 1.75 2.89 2.46 2.52 3.42 3.51
Irish and Scottish West Coast 0.94 0.42 0.96 0.75 0.78 1.27 1.35
Irish Sea 1.78 1.18 2.10 1.78 1.83 2.54 2.63
Iberian Sea 2.03 0.64 1.89 1.61 1.67 2.26 2.33



References

Andersen JH, Gustavson K, Korpinen S, Murray C, Boalt E, Brzozowska A, Garnaga G, Haarich M, Köngäs P, Mannio J, et al. 2010. Integrated assessment and classification of “hazardous substances status”. HELCOM, red. I Hazardous substances in the Baltic Sea: An integrated thematic assessment of hazardous substances in the Baltic Sea. Helsinki Commission. s. 11-16. (Baltic Sea Environment Proceedings; Nr. 120B).

Andersen JH, Murray C, Larsen MM, Green N, Høgåsen T, Dahlgren E, Garnaga-Budré G, Gustavson K, Haarich M, Kallenbach EMF, et al. 2016. Environmental Monitoring and Assessment. 188:Article 115 https://doi.org/10.1007/s10661-016-5121-x.