Comparison of freshwater nutrient boundary values Geoff Phillips 1 & Jo-Anne Pitt 2 1 University of Stirling & University College London 2 Environment Agency

Good response from all countries for freshwater Draft report circulated for comment to nutrient experts April 2015 Several changes in allocation of MS types to broad typology Corrections to boundary values Final draft report now produced Relatively wide variation of boundary values More variation for rivers than lakes More variation for N than P

Comparison of phosphorus Simple comparison for lakes, all countries use a mean for total P except ES (75 th percentile) For rivers 5 countries use 90 th percentiles (halved value for comparison), 4 countries only use soluble P. Three countries (FR,SE,UK) use water body specific rather than type specific values (mean of type used for comparison

How to make comparisons Country Different types Intercalibration type Most comparable water bodies, but fewer countries contributing Broad type Less comparable water bodies, more countries contributing Average difference by country Allow for type differences

Comparison of G/M phosphorus boundaries, lakes & rivers Values lower in lakes than in rivers majority lake values < 100ug/l 50% lake values < 50 ug/l Range of values smaller in lakes than in rivers Fewer different boundary values for rivers, despite more types More single values applied to all types Boundary setting more variable in rivers than in lakes Lakes Rivers Note river P boundary values for ES are as PO4 not P, to be corrected in report

P comparison by intercalibration type RiversLakes Small range where few countries contribute to type Rivers range 50 to >100µg/l Lakes range 30 to <100µg/l

P comparison by broad type Rivers Lakes Similar conclusions using Broad types. Range rivers µg/l, lakes 30-50µg/l

Comparison by country Calculate: 1.average boundary value by broad type 2.national difference for each broad type 3.average difference by country Lakes most differences ± 30 ug/l Rivers differences > ± 50 ug/l

Comparison by method Lakes Rivers Approximate order of resulting boundary values in lakes & rivers 1.Modelling, 2.Regression, 3.Distribution of P in classified WBs 4.Distribution in all WBs 5.Expert Judgement Objective methods directly related to ecological status produce lower boundary values

Nitrogen Fewer countries with boundary values More widespread use of upper percentiles in rivers (orange shading) Values in lakes < rivers Fewer different values in use Several countries report values for nitrate derived from drinking water standards Lakes Rivers

Comparison nitrogen by country Lakes Rivers Similar to results for phosphorus Lakes ± 0.5mgN/l Rivers ± 1.5mgN/l

Conclusions Different values for N & P used for boundary setting across Europe Differences in water body types Different methods used to establish boundary values Boundaries for lakes more similar than those for rivers Boundaries for P more similar than those for N How similar should values be ? What are the most appropriate methods to establish boundary values Discuss these issues at Berlin workshop Evaluation of pressure response relationships to inform these discussions

Comparison of boundary values with pressure response relationships from intercalibration Work currently in progress Pressure response was part of the IC process, but few technical reports contained sufficient information to allow prediction of boundary values Analysis of data from CBGIG and NGIG lakes Phytoplankton Macrophytes Phytobenthos (XGIG) Invertebrates (CBGIG) Analysis of limited new data sets for rivers Report and discuss results at Berlin workshop – November 2015

Initial results Regression relationships between IC Common Metrics and nutrients Better relationships for lakes than rivers Better relationships for TP than for TN reflect the variation in the reported boundary values Explored the use of type I and type II regression and compared with categorical approaches Ordinary Least Squares & Reduced Major Axis regression Box plots of TP or TN concentration in WBs classified by common metric Minimisation of mismatch between nutrient and biological class

Example results high alkalinity shallow lakes IC type L-CB1 Linear models Univariate and multivariate (N + P) Uncertainty (25 th -75 th quantiles of residuals) used to estimate possible range of boundary values. Box plots, average upper and lower adjacent quartiles

Class mismatch Example results IC type L-CB1 Plot % WBs where TP Not Good & Bio is Good TP Good & Bio not Good Point of intersection is minimum mis-match Potential G/M boundary value

Table 4.3 Summary of predicted total phosphorus boundary values for high alkalinity lakes Range of predicted values for different BQEs 1.Most likely boundary 2.Best regression model 3.Possible range (including uncertainty) Different BQEs provide similar range boundary values For LCB1 lakes range most likely boundary lower than range reported by MS Possible range similar to range reported by MS

Initial conclusions & topics for discussion in Berlin Relationships for lakes are relatively strong (R2 >0.36) Type II regression or multivariate (N + P) regressions provide best models Categorical methods produce similar values to regression models Range MS boundary values > range of the most likely boundary values < range of possible boundary values (within prediction error bars) How should uncertainty be interpreted when setting boundary values ? CIS guidance on classification “levels need to be established so that they are no more or less stringent than required by the WFD and hence do not cause water bodies to be wrongly downgraded to moderate status”.. What is the purpose of the supporting element boundary value?