1. Defining Reference Conditions Setting Class Boundaries
&
Setting Class Boundaries

2. A number of definitions:
The condition that is representative of a group of minimally disturbed sites organized by selected physical, chemical, and biological characteristics (Reynoldson et al. 1997).
Representing important aspects of 'natural' or pre-Columbian conditions and at the same time, politically palatable and reasonable (Hughes 1995).

3. WFD’s definition
Expected background (i.e. reference) conditions with no or minimal anthropogenic stress and satisfying the following criteria: (i) they should reflect totally, or nearly, undisturbed conditions for hydromorphological elements, general physico-chemical elements, and biological quality elements, (ii) concentrations of specific synthetic pollutants should be close to zero or below the limit of detection of the most advanced analytical techniques in general use, and (iii) concentrations of specific non-synthetic pollutants, should remain within the range normally associated with background levels (European Commission 2000).
75 words in Eus definition or 4xs the number of words of the previous 2 definitions.

4. A number of approaches:
Best judgement (e.g. theoretical)
Historical
Spatial (regional) reference sites
Predictive modelling
Paleoecological reconstruction
Curve fitting (stress trajectories)
Extrapolation of lab to field

5. Best professional judgement
convening expert panels to determine reference conditions
BUT...
judgement is a function of scientists’ expertise and quality of data supplied
may be biased “it was always better then…”
static measure

6. Historical may work for general statements about conditions BUT...
useful if sites have been periodically resampled
BUT...
data often incomplete
methods may not be comparable
frequency/timing of sampling may confound analysis
usually limited to a single community

7. Spatial reference sites
acceptable levels of disturbance must be ascertained
community variability represents the range of acceptable conditions
BUT...
ecoregions may not apply universally
habitat (ecotype) classification often needed
availability?

8. Predictive modelling
allows for natural environmental gradients (continua)
applicable to many metrics
site-specific resolution
BUT...
high initial costs
requires data, calibration and validation

9. Paleoecological reconstruction
often restricted to lakes
use of diatoms, pollen & chironomids
BUT...
high initial costs
indicators may (only) reflect changes in water quality
restricted to only a few organism groups

10. Curve fitting
plotting metric or index values against disturbance or natural variables to determine reference conditions through curve fitting
BUT...
outliers, uneven data and absence of data from minimally disturbed sites can distort models

11. Extrapolating lab results to field
relationships between test species and some stressors are known
BUT...
data not applicable to entire communities
unsuitable for systems disturbed by other stressors

13. Do different approaches differ?
What is the expected variability among methods commonly used to establish reference?

14. Lake Härsvatten size = 0.19 km2 Zmean =5.7m & Zmax = 26m
altitude = 130 m a.s.l.
catchment = 68% coniferous forest, 13% mire & 9% other water
oligotrophic, mean TP = 7 µg/L

15. What is the reference ”pH” of lake Härsvatten?

16. pH - expert opinion

17. pH - regional variability

18. pH - site-specific modelling

19. pH - paleoreconstruction

20. pH - paleoreconstruction

21. So, what is the reference ”pH” of lake Härsvatten?

22. What is the reference ”TP” of lake Mälaren?

23. What is the reference ”TP” of lake Mälaren?

24. Setting Class Boundaries

25. WFD - the normative classification can be summarized as:
high ≈ no or only minor deviations;
good ≈ low levels of disturbance, but deviate only slightly;
moderate ≈ moderate deviations and significant effects;
poor ≈ major biological alterations and substantial deviation;
bad ≈ severe biological alterations and large deviation.

26. WFD - stipulates 3 types of monitoring
Surveillance - status & long-term changes
Operational - systems at risk & ameliorative effects
Investigative - ascertain causes of systems failing to meet environmental objectives

27. Factors to be considered in setting class boundaries
number of classes - scientific or political
stressor - response relationship
symmetry or asymmetry of classes
upper and lower limits or anchors
variance within and among classes
errors…errors…errors...

28. Errors associated with classification - sample collection & processing
sampling variability (natural spatial heterogeneity and interoperator)
sample processing (sort & identification)
natural temporal variability
asymmetrical pollution effects

29. Example of stressor - response relationship

30. Distribution reference values & selection of upper and lower anchors

31. Choice of upper anchor
95% CI may result in higher frequency of type 1 errors
10th percentile may result in higher frequency of type 2 errors

32. Choice of lower anchor
zero value may result in higher frequency of type 2 errors
minimum value may result in higher frequency of type 1 errors

33. Probability of misbanding (PM)†
misclassification error increases markedly with band-width error!
†Taken from Clarke 2000

34. The WFD is only the first footprint, we still need to find the path!