uncertainty as to effects mattias alveteg, harald sverdrup
TRANSCRIPT
Uncertainty as to effects
Mattias Alveteg, Harald Sverdrup
Acknowledge uncertainties
Specifying typesof uncertainties
Quantitativeestimates
Understandingpolicy relevance
Unc. managementin decision process Funding
+
++
+
+
+
+ +
+
Uncertainty in singular datavalues/events,”thin air” sensitivity analysis
Assessing uncertainty in structurally complex feedbacksystems of uncertain components
Paradigm shift
Uncertainty vs Sensitivity
• A sensitivity assessment is relatively simple
• An uncertainty assessment demands detailed knowledge on input uncertainty
Uncertainty assessment
• Estimating uncertainty range– replicates, spatial heterogeneity, literature, etc.
• Estimating uncertainty distribution– lower pedigree => thicker tails (rectangular?)– bounded/non bounded distributions?
• Estimating interdependencies– ensure consistent input, additional submodels
needed?, move system boundary?
Site specific unc.
-400 -300 -200 -100 0 1000
20
40
60
80
100CDF for EMEP 2021: 2010 deposition
Exceedance mmolcm-2year-1
Per
cent
Changing perspective
• Create one set of input for all sites taking interdependencies into account
• Calculate critical load for all sites => One CDF for the region• Repeat over and over again=>Several independent CDFs for the region• Calculate confidence intervals for the
different percentiles of the region CDF
Site specific unc.
-400 -300 -200 -100 0 1000
20
40
60
80
100CDF for EMEP 2021: 2010 deposition
Exceedance mmolcm-2year-1
Per
cent
(15 7
sit
es in
the
grid
)
Site/percentile specific unc.
-400 -300 -200 -100 0 1000
20
40
60
80
100CDF for EMEP 2021: 2010 deposition
Exceedance mmolcm-2year-1
Per
cent
Comments
• Shifting our view from sites to percentiles– makes us more certain about the excedance in
the region– we loose all information on where within the
grid CL is exceeded
Number of sites?
-400 -300 -200 -100 0 1000
20
40
60
80
100CDF for EMEP 2021: 2010 deposition
Exceedance mmolcm-2year-1
Per
cent
Fewer sites = larger uncertainty
0
50
100
150
200
250
300
0 20 40 60 80 100 120 140 160
95%-ile exceedance uncertainty (EMEP 2021)
App
rox.
unc
erta
inty
ran
ge
Number of sites
Fewer sites = less representativity
-250
-200
-150
-100
-50
0
50
0 20 40 60 80 100 120 140 160
95%-ile exceedance uncertainty (EMEP 2021)
App
rox.
ran
ge o
f po
ssib
le m
edia
ns
Number of sites
Thus:
• Increasing the number of sites up to 40-80 sites per grid reduces uncertainty and increases representativity
• Increasing the number of sites increases the 95%-ile exceedance
• Taking uncertainty into account increases the exceedance
Dynamic uncertainties
• A flexible tool, UNSAFE, has been developed– creates input to dynamic soil chem. models– effective sampling (Latin Hypercube)– option to specify explicit distributions– handles input databases of different quality– is being tested in Switzerland
Future work or end of the road?
• Most of our work so far has not been funded
• Uncertainty assessment requires substantial funding
• No funding => No future work
Where do the uncertainties come from ?
Territorialrepresentativtity Scaling
problems
Much
Little
DepositionN-immobilization
Limit valueWeathering
UptakeFlow rate
DecompositionNitrificationDenitrification
Indicator choiceCatchmentproperties
Organic acids
Limit value N