Critical load functions: effects of
uncertainties in biogeochemical and
species responses and species choice
Ed Rowe, Simon Smart, Susan Jarvis, Pete Henrys,
Chris Evans & Jane Hall
Outline
1. The MADOC-MultiMOVE model chain
2. Generating biodiversity-based critical load functions
3. Progress towards a UK response to the “Call for Data”
4. Uncertainties: in biogeochemistry; transfer functions; niches; and evaluation
Acknowledgments
This work was funded by the UK Department for the Environment, Food and
Rural Affairs (DEFRA) under the National Focal Centre project, AQ0826.
UK dynamic modelling chain
Biogeochemistry
N&S
deposition
Other drivers
e.g. climate
Plant ecology
Rhynchospora
alba
Rhynchospora alba
Vegetation & soil
biogeochemistry: MADOC
VSD
N14C
Transfer functions
DYDOC
pH, N/C, mineral N, biomass
Other drivers
e.g. climate
Floristic response:
MultiMOVE
Habitat suitability
for ~1300 UK plant &
lichen species
MADOC: Rowe et al. 2014 Env Poll 184:271-282
N14C: Tipping et al. 2012 Ecol Mod 247: 11-26
MultiMOVE: Henrys et al. 2015 New J Bot 5: 89-100
Trait-means:
Fertility, Alkalinity,
Height
Evaluation
Biodiversity:
HQI
Selecting a biodiversity metric
Species-richness
Simpson Diversity
max Scarcity
+ves minus −ves
The best indicator of
‘overall habitat quality’
across habitats was n
positive indicator
species. We define
HQI = habitat-suitability
for locally-occurring
positive indicator
species
Positive indicators
Negative indicators
Sub-shrub cover
Similarity to reference Infertility (−1×Ellenberg N)
Only heathland data shown
Rowe et al. (submitted, yet
again…) PLOS-ONE
Ranking according
to metric
Ranking according
to specialists
More N deposition  decrease in HQI
cf. slopes of all responses
Slope of
Biodiversity Metric
vs. N deposition plot
HQ (relative to pristine)
Response of HQI to N deposition
Ndep (eq ha-1 yr-1)
average of range studied
“…the TF came to the conclusion that a common biodiversity indicator such as habitat
suitability indicator would be useful in addition to indicators that meet specific parties’
requirements. These indicators will be calculated using lists of species characteristic of
EUNIS habitats.”
Chair’s report, CCE/TF-ICP-M&M workshop, Rome, 2014
Plots: Jaap Slootweg
‘Biodiversity-based’ CL functions
•
Determine a critical threshold for HQI, corresponding to ecosystem damage or
unfavourable habitat condition
•
… by calculating HQI under N deposition at the empirical critical load, with zero
non-marine S deposition, from 1980-2100
•
Calculate the combinations of S and N deposition that give HQI = HQIcrit
•
Simplify the response function into the form requested in the Call for Data
Hypothetical site
Low
HQI
S dep.
S dep.CL
Smax
CLSmin
High
HQI
CLNmin
CLempN
N deposition
N deposition
CLNmax
Calculating the locations of the two nodes
1. Calibrate MADOC to typical values of pH and N/C for the soil / habitat combination
2. Calculate HQI at 10 x 10 combinations of N and S: (0-180 % CLmaxS) x (0-180 % CLempN)
3. Interpolate the ‘contour’ where HQI = HQIcrit
4. Fit the two nodes by least-squares minimisation
E1.7 dry acid grassland, Cadair Idris
S deposition
Low N & S
 High HQI
N deposition
High N & S
 Low HQI
Progress
2016 defining CL functions for 445 ‘Natura2000’ sites (SACs)
EUNIS Habitat
n sites
D1
bog
126
E1.7 dry acid grassland 43
E3.52 wet acid grassland 30
EUNIS Habitat
n sites
F4.11 wet heath
F4.2 dry heath
119
127
2017 simulate all UK 1 km2 squares with acid / N-sensitive habitats
Inverpolly
Dry heath
Dartmoor
Dry heath
North Harris
Dry acid grassland
Snowdonia
Dry acid grassland
How to summarise these results?
•
The Critical Load functions are site-specific
•
Usually HQI is highest at zero N & S, and declines with more N or more S …
•
… although not in all cases
•
HQI always has the critical value at [zero S, CLempN]
•
Responses to acidity are more variable
Inverpolly
Dry heath
Dartmoor
Dry heath
North Harris
Dry acid grassland
Snowdonia
Dry acid grassland
Sites vary in acid-sensitivity
Inchnadamph (264828)
Sensitive: damaged
by only 20% of CLmaxS
(at zero N deposition)
River Camel (861207)
Insensitive: damaged
only by 186% of CLmaxS
(at zero N deposition)
Dry heath SACs
CLSmax / CLmaxS (%)
0-50
50-100
100-200
200-300
Uncertainties in biogeochemistry
In which habitats does N limit productivity?
Peak biomass
•
Rowe et al. 2016 STOTEN
N/C ratio
•
doi: 10.1016/j.scitotenv.2016.03.066
Can we accurately predict responses of the MADOC outputs that are used to
indicate eutrophication?
soil N/C
Deposition
Climate
available N
MADOC
canopy height
MultiMOVE
Uncertainties in transfer functions
Issues:
• Which variables, interactions and forms (quadratic etc.)
should be used as predictors?
 Susan Jarvis is collating data and re-fitting the transfer
functions
• Modified Ellenberg/Hill scores are for the UK
 Could use EIVE (Jürgen Dengler et al. 2016) scores
Ellenberg N ~fertility
Basing niche models on trait-means (plant height, Ellenberg N, etc.)
• allows the use of big training datasets
• avoids method variation in abiotic measurements, especially of available N
• does not introduce “bias” – Wamelink et al. 2004 JVS 6: 847-851 rather shows that r2
increases when many intercepts and slopes are fitted
• Indicator-scores are mainly a European phenomenon
 Simon Smart is investigating mean leaf dry matter content
as a productivity indicator
Deposition
Climate
MADOC
MultiMOVE
Ellenberg R ~alkalinity
Uncertainties in species niche models
Uncertainties are reduced by calculating
habitat-suitability as the mean of values
fitted using different statistical models
(GLM, GAMS, MARS)
Rhynchospora alba
Rather than rescaling as “proportion of
maximum”, we use the Real et al. method
to correct for prevalence in the training
dataset  miraculous comparability
Botanists are currently inspecting visually
the model for each species in relation to
each axis (funded by Botanical Society of
the British Isles)
Deposition
Climate
MADOC
MultiMOVE
Uncertainties in evaluation
Use of HQI = habitat-suitability for locally-occurring positive indicator species as a
metric of overall habitat quality is agreed (for now)
Which list of positive indicator species should we use?
• ‘Common Standards Monitoring’ lists contain ambiguities, and aren’t for EUNIS classes
• ‘BSBI/JNCC indicator-species’ lists do not include bryophytes (& not for EUNIS)
• ‘Bioscore’ species include many non-UK and non-distinctive species
Is it appropriate to calculate HQIcrit as the value
at [zero S, empirical CLN] ?
S dep.
N dep.
Deposition
Climate
MADOC
MultiMOVE
CLempN
Conclusions
1. The MADOC-MultiMOVE model chain has advantages for UK simulations:
• MADOC is well-suited to simulating organic soils where DOC strongly affects pH;
• MultiMOVE was trained on UK species-occurrence data.
2. The UK NFC will (if funded) submit CLbiodiv functions for acid- or N-sensitive habitats in
all 1 km2 gridcells where they occur, in response to the current Call for Data.
3. Many biogeochemical uncertainties urgently need to be constrained, in particular
• in which habitats, and where, does productivity still respond to N deposition?
• how does management intensity affect the response of canopy height to N deposition?
4. Uncertainties in [abiotic indicators  trait-means] functions are being reduced.
5. Uncertainty in species niche models is comparatively low.
6. Uncertainty in evaluation has been greatly reduced by the decision to use a habitatsuitability-based indicator, but choice of indicator species remains an issue.