MIME regional assessment of status and trends in BDE concentrations in sediment

Overview

The 2017 MIME assessment describes the trends and status of contaminant concentrations in biota and sediment at monitoring stations in the OSPAR area. Assessments are made for a large number of time series, each of a single contaminant in a single species (for biota) at a single monitoring station. This document is one of a series that synthesises the results of the individual time series to assess status and trends at the MIME regional level. In particular, it considers BDE concentrations in sediment, where the BDEs are BDE28, BDE47, BDE99, BD100, BD153, BD154, BD209. For simplicity, the term ‘region’ is used throughout to describe MIME regions. OSPAR regions are always referred to as such.

The following sections:

  • summarise the individual time series results
  • describe the methods used to combine the individual time series results to estimate regional trends and status
  • present the estimates of regional trends
  • present the estimates of regional status


Summary of individual time series results

A time series of BDE concentrations is assessed for status if:

  • there is at least one year with data in the period 2010 to 2015
  • there are at least three years of data over the whole time series
  • a parametric model can be fitted to the data and used to estimate the mean concentration in the final monitoring year (or, occasionally, if a non-parametric test of status is applied)

The conditions are more stringent for trends. Specifically, a time series is assessed for trends if:

  • there is at least one year with data in the period 2010 to 2015
  • there are at least five years of data over the whole time series
  • a parametric model can be fitted to the data and used to estimate the trend in mean concentrations

Note that all trend assessments for individual time series and most status assessments are based on the fit of a parametric model. This is important because only the parametric results are passed into the regional assessments in the following sections. Only 0 of 257 BDE time series were assessed for status using a non-parametric test.

The first tab below shows all the monitoring stations where there are individual time series assessments of trend or status for BDEs. The magenta circles are stations where there is a trend assessment for at least one BDE, and the cyan cirles are stations where there are only status assessments.

The next set of tabs show the trend and status assessments for each BDE in turn. The colours have the following meaning:

  • black: there are no assessment criteria available

And the shapes have the following meaning:

  • downward triangle: the mean concentration is significantly decreasing (p < 0.05)
  • circle: there is no significant change in mean concentration (p > 0.05)
  • upward triangle: the mean concentration is significantly increasing (p < 0.05)

The final set of tabs give:

  • the number of time series with upwards, downwards or no trend in each MiME and OSPAR region
  • the same information presented as proportions


Stations


BDE28


BDE47


BDE99


BD100


BD153


BD154


BD209


Trend summary (numbers)

Number of time series with upwards, downwards or no trend by MIME and OSPAR region

OSPAR region MIME region status BDE28 BDE47 BDE99 BD100 BD153 BD154 total
2 Northern North Sea upward trend 0 0 0 0 0 0 0
no trend 0 5 0 0 0 0 5
downward trend 0 0 0 0 0 0 0
total upward trend 0 0 0 0 0 0 0
no trend 0 5 0 0 0 0 5
downward trend 0 0 0 0 0 0 0
3 Irish and Scottish West Coast upward trend 0 0 0 0 0 0 0
no trend 0 2 0 0 0 0 2
downward trend 0 0 0 0 0 0 0
Irish Sea upward trend 0 0 0 0 0 0 0
no trend 2 3 4 2 4 1 16
downward trend 0 2 1 0 0 1 4
total upward trend 0 0 0 0 0 0 0
no trend 2 5 4 2 4 1 18
downward trend 0 2 1 0 0 1 4


Trend summary (proportions)

Proportion of time series with upwards, downwards or no trend by MIME and OSPAR region

OSPAR region MIME region status BDE28 BDE47 BDE99 BD100 BD153 BD154 total
2 Northern North Sea upward trend 0 0
no trend 100 100
downward trend 0 0
total upward trend 0 0
no trend 100 100
downward trend 0 0
3 Irish and Scottish West Coast upward trend 0 0
no trend 100 100
downward trend 0 0
Irish Sea upward trend 0 0 0 0 0 0 0
no trend 100 60 80 100 100 50 80
downward trend 0 40 20 0 0 50 20
total upward trend 0 0 0 0 0 0 0
no trend 100 71 80 100 100 50 82
downward trend 0 29 20 0 0 50 18


MIME regional assessment - methods

Tabulating the number of time series with an upward or downward trend provides a quick summary of the individual time series results. However, it does not provide an objective regional assessment of trends. This section describes how the individual time series results can by synthesised in a meta-analysis to assess trends at the regional level.

Trends

For a regional trend assessment, the trend in each time series is summarised by the estimated change in log concentration over the last twenty years (or shorter if the time series doesn’t extend that far back). Regional trends are then estimated by fitting the following linear mixed model by restricted maximum likelihood:

  • response: trend (yearly change in log concentration)
  • fixed model: region
  • random model: BDE + region.BDE + station + trend estimation variation + residual variation

The fixed model means that a trend is estimated for each region. Trends in the different BDEs are assumed to be broadly similar, because inputs and control measures are likely to affect all BDEs in a similar way. Each regional trend can thus be interpreted as the common trend in BDE concentrations in that region (i.e. the trend averaged across all BDEs).

The random model has five terms:

  • BDE allows for variation in trend between BDEs common to all regions
  • region.BDE allows for additional variation in trend between BDEs within regions
  • station allows for variation in trend between stations common to all BDEs
  • trend estimation variation is the variance of the trend estimates from the individual time series analysis, assumed known and fixed
  • residual variation is the variation that cannot be explained by any of the fixed effects or the other random effects

The meta-analysis is restricted to time series from monitoring stations that are classified as ‘representative’. ‘Baseline’ stations (those with near pristine conditions or only subject to very diffuse inputs) are omitted because any trends there will likely be caused by different processes to those at representative stations. Similarly ‘impacted’ stations (those close to a point source) are omitted because trends there will likely reflect changes due to the point source, and will be ‘unrepresentative’ at the regional level.

The meta-analysis is further restricted to:

  • regions with at least three trend stations with good geographic spread. Three stations is considered the minimum required to provide some sort of evidence base at the regional level
  • BDEs with trend information for at least three stations (over all regions): BDEs are treated as random effects in the meta-anlaysis, so information about trends in one BDE can be ‘borrowed’ from the trend information in the other BDEs

Status

Unfortunately, no assessment criteria have yet been adopted for BDEs in sediment, so no formal assessment of status is possible. However, an informal ranking of regions can be made by comparing the regional mean concentrations. This is considered in the final section of this document which looks at concentration profiles.

Presentation

The results of the regional assessment of trends are back-transformed for presentation. The estimated regional trend is then interpreted as the percentage yearly change in concentration.


Regional assessment - availability of time series

This section provides more detail on the number and geographic spread of the time series available for the meta-analyses. The tabs show the stations with parametric trend and status assessments for each OSPAR region in turn, having excluded trend assessments at baseline and impacted stations. The magenta circles are stations where there is a trend assessment for at least one BDE, and the cyan cirles are stations where there are only status assessments.

Note that, although there are no asssessment criteria available, the cyan circles show where there are time series with estimates of the mean concentration in the final montitoring year. These can be used to construct concentration profiles at the regional level, and to compare mean concentrations across regions.

The following regions have an insufficient number or time series or spread of stations and are excluded from the regional assessment:

  • trend: Irish and Scottish West Coast
  • status: Channel, Iberian Sea


OSPAR region 1


OSPAR region 2


OSPAR region 3


OSPAR region 4


Regional trends

The first 4 tabs show:

  • Regional trends: estimates of the regional trend (averaged over BDEs) with pointwise 95% confidence limits
  • Regional trends by BDE: regional trend estimates for each BDE with pointwise 95% confidence limits plotted by BDE
  • Regional trends by region: regional trend estimates for each BDE with pointwise 95% confidence limits plotted by biogeographic region
  • Individual time series: the trend estimates from the individual time series on which the regional assessment is based. The estimates from regions with insufficient geographic spread (and excluded from the meta-analysis) are also shown.
  • The symbols in all these plots have the following interpretation:

    • downward triangle: the mean concentration is significantly decreasing (p < 0.05)
    • circle: there is no change in mean concentration (p > 0.05)
    • upward triangle: the mean concentration is significantly increasing (p < 0.05)

    The final 2 tabs show the estimates of the regional trend (averaged over BDEs) and the regional trend by BDE.


    Regional trends


    Regional trends by BDE


    Regional trends by region


    Individual time series


    Estimates by region

    This table shows the estimates of the regional trend:

    • trend: the estimated mean yearly change in log concentration across the region (multiplied by 100 for presentation)
    • se: the corresponding standard error
    • lower, upper: the corresponding pointwise 95% confidence limits
    • % yearly change: the estimated mean % yearly change in concentration across the region
    • %yc lower, %yc upper: the corresponding pointwise 95% confidence limits


    MIME region trend se lower upper % yearly change %yc lower %yc upper
    Northern North Sea -6.71 5.10 -16.69 3.28 -6.49 -15.37 3.34
    Irish Sea -13.69 3.96 -21.45 -5.93 -12.80 -19.31 -5.76


    Estimates by region and BDE

    This table shows the estimates of the regional trend by determinand:

    • trend: the estimated mean yearly change in log concentration across the region (multiplied by 100 for presentation)
    • se: the corresponding standard error
    • lower, upper: the corresponding pointwise 95% confidence limits
    • % yearly change: the estimated mean % yearly change in concentration across the region
    • %yc lower, %yc upper: the corresponding pointwise 95% confidence limits


    MIME region BDE trend se lower upper % yearly change %yc lower %yc upper
    Northern North Sea BDE47 -5.59 4.08 -13.59 2.42 -5.43 -12.71 2.45
    BDE99 -7.82 5.81 -19.21 3.56 -7.53 -17.48 3.63
    Irish Sea BDE47 -12.45 3.70 -19.71 -5.19 -11.71 -17.89 -5.06
    BDE99 -14.93 3.94 -22.67 -7.20 -13.87 -20.28 -6.95


    BDE profiles

    The profile of concentrations across BDEs can sometimes be useful for examining sources, and for comparing concentrations across regions where assessment criteria are not available. Profiles were examined by fitting the following linear mixed model by restricted maximum likelihood:

    • response: mean log concentration
    • fixed model: region
    • random model: BDE + region.BDE + station + estimation variation + residual variation

    where, for each time series, the mean log concentration is the estimated mean log concentration in the final monitoring year and the estimation variance is the variance associated with this estimate.

    The estimated BDE concentration profiles by region are shown below with pointwise 95% confidence limits. The units are ug/kg dw.