Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
126 lines (101 sloc) 4.97 KB
* Introduction Friederike and Institute
* What does attribution science try to do
Context: - does the climate change at all?
- is this change anthropogenic
- is a particular (extreme) weather event due to this change
It does not answer the second question (even though we all expect the answer is "yes"
* What is extreme weather
- criteria: Heat, Wind, Rain, Cold
- how do you define thresholds?
- Do you just look at meteorological numbers (temp, rain/time)
or also on when/where it occurs and the economic or health impact it had?
* Does global warming make extreme weather worse, or does it create more extreme
weather instances. Or are the two the same :) ?
* How does it work in principle
- Location-specific probabilities and then contrasting with what actually occurred
- Location specific because global warming effects the circulation.
(which "distributes" effects unequally)
- Key idea: distinguish thermodynamic (global, slow) and dynamical (circulation)
causes because they are quite different.
- basically statistical
- Non-linearity: eg once a threshold is crossed, precipitation remains stable.
- You don't really attribute a particular event, it's still probabilitic for a class
("xyz likely that this kind of event based on climate change").
You can't attribute ONE PARTICULAR event. Correct?
- Class of event vs. a particular event. What difference does this make?
Can't any event generalized?
* Is attribution the same as predicting?
- predicting as in "it will happen again every x years
- or predicing as in "when we see this pattern in the atmo developing,
event will likely happen in the new y days"
* Steps
- Step 1: What happened?
- Step 2: Event definition
- Step 3: Model evaluation
- Step 4: Estimate likelihoods
- Step 5: Interpret and synthesize
- Step 6: Communication
* Several different methodologies:
- Risk based approach. Run "current" climate model against counterfactual "natural" one.
. Seems like the obvious approach
. How do you know how a counterfactual world would have looked?
remove aerosols and co2. What to do about oceans sst?
. remove only one factor should give a more specific cause?
- Boulder approach: how is it different?
. Tries to disentangle the specific causes of a particular event.
. Critic: you remove the thing of which you *claim* it is the problem.
Isn't that self-fulfilling?
- Circulation-based approaches
- Storyline approaches
* Evolution of climate models
- Climate models had to evolve to contain circulations
(is this only the atmosphere, or also oceans? Gulf stream?)
- what are "coupled" models?
- how do you simplify? If you run 1000s of ensembles, you can't be full detail
* The role of ensemble simulations
- What are these?
- What do you vary? How do you know what to vary?
- How many ensembles do you typically run
(I read somewhere of 17,000 real, 117,000 counterfactual)
- Ensembles vs. Re-analysis
* Instead of using a hypothetical world (antropogenic removed)
we can also just go back 50 years ... less "hypothetical". True?
* Problem of scale
// - There's still averaging - otherwise it would be weather.
- How do you represent small-scale extreme events (huge thunderstorms) if
the climate models don't go down to this scale?
- What is a "regional climate model"?
- If you run local models, how do take care of the boundary conditions?
- Generally: how do you combine local and global models that run
at different scales?
* How do you get from a model run result to a probability?
How do you quantify uncertainty?
* Do the different methodologies agree if applied to the same event?
* What makes an event "easy" to attribute?
Are the different kinds (heat, rain, etc) differently complex to attribute?
* How can you validate the "predictions" or attributions/probabilities
of your attributions?
- What's your equivalent of 3 sigma.
- What about built-in biases?
- How do you correct?
Example: Krymsk, southwest Russia, near the Black Sea, caused by catastrophic rains
in July 2012, was no exception (Fig. 1). Writing in Nature Geoscience,
Meredith et al.1 now show that — exceptionally for such a localized event — it is
possible to attribute this flood to climate change: they present convection-resolving
simulations that suggest that incremental warming of sea surface temperature (SST)
in the Black Sea over the past few decades has led to an abrupt amplification of
convective precipitation.
-> ran a particular model. How do you ensure it is not self-fulfilling?
* Extreme weather event vs. long-term general trend:
the latter's attribution is trivial with climate models. isn't it?
* What is Weather@Home and how does it help?
* A little bit of history
- How old is the field?
- Key enablers
* The future
- What is still missing from the approach
- what can be improved?
- How?
You can’t perform that action at this time.