CONCAWE Experience with the CityDelta Toolkit

1
CONCAWE Experience with the City-Delta Toolkit

• Pete Roberts
• Shell Global Solutions

2
Concawe

• Oil industry - downstream refining fuels
  distribution
• technical organisation
• carries out studies to provide advice on
  environmental issues affecting the industry
• provides information about the activities of the
  industry
• provides information to industry about
  environmental issues
• CAFE and UN-ECE Concawe has working groups for
• modelling
• emissions
• techno-economics
• scenarios
• The groups do
• limited original work
• try to understand verify provide feedback

3
Concawe Experience with the City Delta Tool-kit

• Goal is modeling of atmospheric pollutants at
  scale less than 50 km a necessary input to
  effects assessment used in IAM ?
• Basic Questions
• Can the performance of models be summarized using
  the toolkit?
• Can the differences in performance of the models
  against data be summarized?
• Are differences in performance between models and
  data
• scale dependent
• significant?
• What is the geographic distribution same for
  all countries/cities?
• Do models predict all chemical species equally
  well?
• For different scenarios is the range in
  predictions of key indicators the same?
• What is the variation between future scenarios
  compared with variation against now and data ?
• Is there enough information to draw robust
  conclusions about modeling?
• If not where is the shortfall?

4
Making things tractable

• Key points
• Models are always approximations there is no
  perfect model and all models have strengths and
  weaknesses
• There is never enough data.
• Model predictions and measured data are
  fundamentally different quantities and neither
  can be corrected to a common spatial and
  temporal state of information.
• Need to rank models
• comparison with data offers a way.
• 3 categories
• A - models which predict a different time
  variation to measurements. (Time-series
  comparison, correlation coefficient)
• B - models consistently predict higher/lower
  values than measured data. (NMSE, BIAS-T, Scatter
  Plot)
• C - remaining models.

5
Milan
6
Paris
7
Berlin
8
London
9
What variability is there between models and
their predicted concentrations in urban areas

• Are all models different?
• Not really
• Is there consensus between some groups of models
• YES
• Is this related (correlated) with how the models
  are formulated this could include the scale for
  example.
• Unclear finer scale models can show more
  variability within a domain but quantitative
  metrics such as correlation co-efficient is not
  necessarily better.
• The metric used to compare models affects
  decision
• broad metrics (average concs, correlation
  co-efficients,) show models as more similar than
  extreme measures such as max, AOT-x etc.

10
Of the main pollutants what ranking can be made
on how well they are predicted

• i.e. is O3 better than NO2 - yes
• NO2 under predicted
• PM not tested -