Development of the scientific rationale and formulae for altering RIVPACS predicted indices for WFD Reference Condition
September 2006
Background to research

With the advent of the EU Water Framework Directive (WFD), the concept of the ‘reference condition’ (RC) against which assessments of biological degradation must be compared has become explicit within the legislative framework of the European Union (Council of the European Communities, 2000). It is therefore essential that member states can demonstrate that the biological datasets and models used to define RC meet the WFD criteria and set the same standards for all types of river site.

The selection of the RIVPACS (River InVertebrate Prediction And Classification System) reference sites and development of the RIVPACS methodology and software system for assessing the ecological status of UK rivers preceded the WFD. For several years it has been a concern that several, or even many, of these reference sites may not have been in WFD RC at the time of macroinvertebrate sampling, but merely represented the “best available” sites for each type of river site. In particular the predictions of expected fauna for some types of river site will be based on inadequate quality reference sites, leading to under-estimation of RC values for biotic indices and over-estimates of the RIVPACS observed to expected (O/E) ratios (termed Ecological Quality Indices (EQI)) and site quality for macroinvertebrates.

To help address this problem, the Centre for Ecology and Hydrology (CEH) collated and assessed the available pressure data on each site (SNIFFER project WFD46). Subsequently UK agency aquatic ecologists provided an assessment score (1-6) of the perceived WFD ecological status class of each reference site (1 =  top of high, 2 = middle of high, 3 = high/good boundary, 4 = middle of good, 5 = good/moderate boundary, 6 = worse).

In the current UK classification scheme, an EQI value of 1.0 is used to set the “high/good” boundary; a river site has to achieve biological index values equal to or in excess of those predicted by RIVPACS in order to be classified as the highest status. This means that, in effect, that roughly half of the RIVPACS references sites are assumed to be of “high” ecological status and roughly half of “good” status. The problem is that this assumption has been applied across the board for all types of river sites in the UK.

UK regulatory agency ecologists have been devising methods of adjusting the RIVPACS expected (E) values or EQI values of indices (ASPT and number of BMWP taxa (TAXA)) by determining the weighted average assessment score of the reference sites involved in the prediction for any particular test site. However, these approaches are, in effect, determined solely by the EQI values used to set the good/moderate boundary.

Objectives of research

To develop a robust defensible mechanism for adjusting the RIVPACS expected values of biotic indices for any specific test site according to the perceived ecological condition (at the time of sampling for RIVPACS) of the RIVPACS reference sites actively involved in the prediction for that test site. The resulting adjusted O/E values should then more evenly reflect the WFD concept of ecological status across all UK river types

To provide the necessary formulae for implementing this adjustment mechanism for the environmental regulatory agencies’ classification sites.
Key findings and recommendations

The research approach used was based on statistical modeling to assess and quantify the actual relationship between assessment score (1-6) and the observed values of TAXA and ASPT amongst reference sites within RIVPACS site types (TWINSPAN groups).

Negligible differences in average index values were found between reference sites with assessment scores of 1,2 or 3. Observed index values are lower relatively for reference sites with assessment scores of 4 and especially 5. (There were insufficient reference sites with scores of 6 (worse than good/moderate boundary) to estimate its adjustment factor and it is recommended that any such sites are treated as having assessments scores of 5 in the adjusted of test site expected values.)

Models were fitted separately to data for each RIVPACS module (GB Northern Ireland (NI), Scottish Islands (SI) and Scottish Highlands (SH)) and for samples based on each the seven possible combinations of one, two or three RIVPACS seasons. There was insufficient range of scores in the high quality SI and SH sites to determine any significant relationships. Therefore a recommended single UK-wide adjustment model was fitted using all UK reference sites and samples. Although actual effects of changes in site quality on index values may vary with site type, to achieve adequate precision and for simplicity, the best single overall estimates for the adjustment factors were  derived.

Models were also fitted based on grouping sites by WFD System A typology; some estimates of adjustment factors were greater, but overall model fits were poorer.

The recommended statistical model for estimating the adjustment factors was model M4 which estimates the average proportional increase or decrease in index values due to each level of site assessment score relative to sites with scores of 3 (“target” high/good boundary) within the same TWINSPAN group. Recommended adjustment factors are given in Table 30.

Procedures and formulae are derived to combine these score-specific adjustment factors with information on the proportion of reference sites in each site group with each assessment score to calculate the recommended adjustment factors for expected values for any test site based on its RIVPACS probability of belonging to each TWINSPAN group.

An EXCEL spreadsheet adjustment “calculator” with encoded formulae to automate this procedure for adjusting RIVPACS expected values of any UK RIVPACS test sites has been produced and is available as an project deliverable and output.

One advantage of the M4 type of model is that the approach could also be used in other European ecoregions where reference condition expected values of metrics for a test site are based on some average or percentile value of the reference sites in the same WFD  System A or B stream typology (Council of the European Union, 2000) (i.e. stream type replaces TWINSPAN group in the models)).

Keywords: Water Framework Directive (WFD), Reference Condition, RIVPACS, Adjusted expected

Copies of this report are available from the Foundation, in electronic format on CDRom at 20.00 + VAT or hard copy at 25.00, less 20% to FWR members.

N.B. The report is available for download from the SNIFFER Website