1. Relationships for Broad & Intercalibration Types Geoff Phillips
Analysis of data sets
Relationships for Broad & Intercalibration Types
Geoff Phillips

2. Aim of analysis Data Issues
Provide a guide to the range of nutrient boundary values for each of the “Broad” river and lake types
Data
Intercalibration data sets (IC)
EEA State of Environment data set (SoE)
Issues
IC data sets limited number of countries in each type
Many IC data sets did not contain variables needed
Nutrient concentrations
Common metric EQR
Information needed to normalise national EQR values
SoE data difficult to link to Broad Types
Inadequate environmental data
National type codes did not match lookup tables to Broad Types
Contradictions between IC and Broad Type allocations
Relatively few countries with data

3. Outcome - Lakes
Analysis of N & P using majority of available IC data (phytoplankton, phytobenthos, macrophytes) and presented in 1st report.
broad types 3 & 4, lowland calcareous lakes
L-CB1, L-CB2
broad L-N2a, type 2 lowland siliceous lakes
L-N2b, L-N1
broad types 5 & 9 organic & siliceous
L-N3a, L-N8a, L-N6a

4. Example for lowland calcareous lakes BT3
Lines show range of boundary values from pressure response models
Dots are MS boundary values

5. Outcome - Lakes
Analysis of N & P using majority of available IC data (phytoplankton, phytobenthos, macrophytes) and presented in 1st report.
broad types 3 & 4, lowland calcareous lakes
L-CB1, L-CB2
broad L-N2a, type 2 lowland siliceous lakes
L-N2b, L-N1
broad types 5 & 9 organic & siliceous
L-N3a, L-N8a, L-N6a
Added values for
broad type 8 mid-altitude calcareous
L-AL3, L-AL4
Extended analysis using EC GIG phytoplankton data
Data for MGIG made available, not analysed

6. Conclusions Lakes Reasonable coverage except for
Large lakes
Mediterranean lakes
Highland lakes
Generally reasonably good relationships
Except EC GIG phytoplankton
SoE data made available via MARS, but does not fill the gaps
Issues with linking data to broad types
Above lake types not well represented in data set

7. EC GIG data L-EC1 (high alkalinity very shallow)
Adding EC GIG data to other Broad Type 4 lakes increased the scatter substantially
L-EC1 data R2 = 0.15

8. EC GIG data L-EC1 (high alkalinity very shallow)
Explanation may be the role of nitrogen
Outliers had very low N:P ratio
(N limited lakes?)
How to deal with this when establishing boundary values ?
N:P

9. Quantile regression Fit a line to the lower 25th quantile
Intersection with biological G/M boundary provides a precautionary TP boundary at which 75% of WBs are at Good or better status.

10. Fit lines to lakes based on N:P ratio

11. Multivariate analysis
Produces relatively high TP boundary values ?
Is it appropriate to combine analysis from such a broad geographic region ?
ECGIG data no good relationship with TP

12. How to deal with N & P boundary values ?
If lake N limited does it need a P boundary ?
If lake P limited does it need an N boundary ?
Possible solution
Use a combination rule for the “nutrient” supporting element – Best of Either P or N
Might not work if need to protect downstream water bodies

13. River data 1st report found few significant relationships
Asked for additional GIG data
Med GIG macrophyte & phytobenthos
EC GIG macrophyte & phytobenthos
CB GIG macrophyte & phytobenthos
Obtained EEA SoE data for rivers
Linked to Broad Types (Thank you MARS!)

14. MGIG Macrophyte data
Very poor relationships, clear country differences

15. MGIG Phytobenthos Slightly better relationships
Highest R for Sol P
Unclear if nutrient data are spot samples or summary means
Summary R2 values for relationships
ICM EQR v TP by IC Type p<0.001
ICM EQR v Sol P by IC Type p<0.001
ICM EQR v TN by IC Type p<0.001
ICM EQR v NO3-N by IC Type p<0.001

16. EC GIG macrophyte data Summary R2 values for relationships
ICM EQR v TP by IC Type p<0.001
ICM EQR v Sol P by IC Type p=0.002
ICM EQR v NO3-N by IC Type p=0.313

17. EC GIG Phytobenthos Summary R2 values for relationships
National normalised EQR v soluble P by IC Type p<0.001
National normalised EQR v nitrate N by IC Type p<0.001

18. EEA SoE data Potential to use these data, but Dominated by FR
Issues with Broad Type allocation
R2 likely to be relatively low

19. Use of quantile regression
“Wedge” shaped relationship, typical of multi-pressure response
Upper surface defines upper limit of nutrient compatible with status.
Suggestion in 1st report that this approach might be useful (borderline response)
EQR
But
Difficult to define
Similar to use of an upper uncertainty band of simple regression
Non-precautionary value
Why use it in this case but not for clear linear relationships ?
Scatter may be due to influence of other pressures
Soluble P
EC GIG Phytobenthos relationship for R-E4

20. How to take boundary setting rivers forward ?
Are relationships for rivers weaker because of multiple pressures ?
Are the available data sets inadequate ?
Nutrient data spot values
Poor spatial temporal link to biological data
Further work
Current GIG and SoE data sets might reveal better relationships but uncertainty in the data remain high.
National EQR values need normalising, require the MP and PB boundary EQRs
Generally no other pressure metrics
Can we make strong recommendations for boundary setting for rivers at this point ?