Week 8 Flashcards
What two satelittes are used to look at clouds through a laser?
CALIPSo and LIDOR
What was the orbit train?
a series of satilettes orbiting in a train - originally they were 12 seconds apart
this enabled an in depth analysis of processes in the atmosphere which could be used to constrain or improve processes in climate models
What is the idea of equations and models?
Climate models solve sets of equations by stepping forward in time
Can you give some examples of natural processes that are difficult to include in climate models?
Roseby waves - information propogating around the atmosphere
El Nino - 100s and 1,000s of km and time scale of years
all of these get resolved by other processes that happen at smaller and faster time scales - but not explicitly due to the fact that they often happen at much smaller scales
What does uncoupling mean in models and can you give ma an example?
individual components of a climate model can be run separately
the other components become the boundary conditions
eg. running the atmospheric model only, with the prescribed SST as a lower boundary condition
Why would you uncouple the climate model?
it can be expensive to run the full model
eg. running an ocean model is expensive if you are just looking at the atmosphere, you can just fix the ocean at a certain ermpature
what are some of the advatnages of running a single component?
- increased resolution
- easier to control resemblance to reality
- useful for physical understanding of the processes
What is required for a regional climate model?
boudary conditions on the side walls
- grid boxes don’t have to be the same size globally
this requires a high resolution model over the bit you care about - can run at a v low resolution everywhere else
typically use a global climate model run
what is the advantage of regional climate models?
increased resolution and ability to control forcing
how are small scale processes delt with in models?
not all process key to the cliamte can be captured using climate models
- cloud processes occur on v small spatial and temporal scales that cant be resolved
eg. surface drag, radiative fluxes and convection
in these cases climate models are limited by resolution and subgrid-scale processes need to be parameterised
what is a physical parameter
simplifications of unresolved physics
eg. % of the grid box covered by cloud - high relative humidity - high cloud cover but what is the relationships?
Why are parameters important?
parameterised processes can have a large effect on global climate - feedbacks etc.
important to develop parameters that are useful and good in terms of global climate - but a major problem is trying to determine what a good equation or parameter is
Why might climate models have errors?
- parameterisation is not perfect- there is uncertainty in how to represent certain aspect of the climate
- not high enough resolution
numerical errors - drift in temperatures when there shouldn’t be
What are some of the key things to look at when evaluating climate models?
Mean climate - should be the right climatology
trends - models should correctly reproduce climate trends of the 20th cent
Natural variability - models should be able to simulate natural modes of variability
spatial patterns - models should get not only the global quantities right, but also their geographical patterns
what is the multi-model mean
the mean results from all the models
what was the global warming hiatus?
In the early 2000s the observations seem to indicate a slowdown of warming
the hiatus has since ended - rapid increase in global temperature since 2010 in observations
most likely it was caused by natural variability
what are some of the boundary conditions that a model requires
SST
sea ice location
Co2 emissions
need an initial state to start from - need to be a valid state of the atmosphere
What is a model bias?
model errors relative to observations
Why is the mean bias of all models small?
idea of averaging - compensating errors among models
the average surface temperature of models vary significantly - however the MMM is quite a good estimate
Is it important that most models cannot get the right temperature?
not really as it is more important that they get the same trend
as long as the change is correct it doesn’t really matter what the temrpature it is - being at a particular temperature wont significantly impact the response to this change.
Why are tempeature anomonlies better to look at ?
temp anomollies are better observations corelated over long distances when average temperature is harder to estimate
difficult to get a good value for average temperature of earth - how it varies from average is easier to estimate
In surface temperature what does the pattern for MM bias look like? Why?
Too hot on the western cost of south America and south Africa and eastern coast of America
due to clouds - regions of low lying couds are coarse in models - they are low and thin thus difficult to model and simulate
- causes the ocean to absorb more sunlight causing errors in ocean circulation - upwelling and cooling currents
What are the MMM biases for precipitation? what is the consequence of this?
model have too much rainfall south of the equator
there are two inter-tropical convergence zone (ITCZ) double rainy band that doesn’t exist in observations giving you two peaks in precipitation that don’t exist
What drives the bias in MMM precipitation patterns?
- changes in temperature in the southern ocean
- how the winds move around and within the equator
- parameterization issues
key - there is no one cause for it
also more generally there is model uncertainty - feedback loops are not well understood and depend on parameterization of processes - different mean temperatures can have huge impacts on feedback loops
When is the min of sea ice?
September
What are two other natural pheomena which the MMM gets pretty well?
- arctic sea ice depsire large biases in models with some far too much or too little sea ice
- ocean heat transport
Why can enso be used to test models?
models can be evaluated based on natural variability
a good model should have realistic variability
- climate models are generally able to simulate enso however there are some biases in SST - leads biases in ENSo impacts like rainfall and temperature
Why does the IPCC publish simple, unweighted MMM?
the underlying rationale for validating modles against observations is ….the smaller the model bias, the better the model
but it is unclear whether smaller bias translates to a more accurate climate projects
this is why the IPCC publishes simple, unweighted MMMs
Why might a model that accurate simulates today not be good for future predictions?
might have tweaked values for certain aspects like solar radiation to get temp correct
might distort future predictions
Why are we not that bothered about bias for future model projections?
there is no obvious relationship between historical temperatures and climate sensitivity in models
it is unlikely that reducing biases in global temperature would reduce the range of climate sensitivity in models
How do we objectively evaluate model climate performance?
objective, quantitive evaluation is useful as it allows a comparison with each model and can be used to assess how changes in the model code affect the model performance
- pattern correlation
- root mean-square error (RMSE)
What is pattern correlation?
the correlation between two patterns
typically you would compare models with some reference observation dataset
high correlation means that the model can reproduce the observed pattern well
What does TAS stand for?
surface temperature
observed pretty well - models get it pretty accurate
What is RLUT?
outgoing longwave radiation
upwelling at the TOA - how much and where is it escaping from - can be observed from satellites - models get it pretty accurate
What is PR?
precipitation
harder to predict - temp is well correlated over reasonable difference but there is a high uncertainty
What is SW CRE?
shortwave cloud-radiative effect
how much clouds reflect sunlight back into space - observations are reasonable and models are not doing a great job
somewhat correct but could be doing a lot better
What is the comparison between CMIP3 and CMIP 5?
CMIP 5 models tend to perform better than CMIP3
models are getting better but there are still issues
large spread between models
What is RMSE?
root mean-square error
a measure of the size of the verticle bars
large bars = large error = higher RSME
if the models were perfect the dots would lie on the line of model and observations
normally on a scatter plot - look how they deviate vertically
How does the MMM perform in terms of RSME in CMPI5 models?
it is better than any single model
What is model tuning?
climate model parameterisation use many parameters whose values are uncertain
cloud parameters
surface roughness parameters
ocean mixing parameters
–> because these values are uncertain, model developers optimise these parameters to reduce model biases in historical climate
how can you improve model parameters?
using observations - use an observational parameter and determine the plausible range of values for the observational range
In models what is typically tunned?
climatologies of temperautre, radiative flux, clouds
spatial patters of some variables
most models are not tuned for a given climate sensitivity or historical trend but they can be
What are examples of good constraints to tell if a model behaviour is working?
use constraints like sea ice extent
What is an example of an emergent constraint?
eg. snow-albedo feedback - the percent albedo change per degree of warming
when temperature increases, snow cover decreases and albedo goes down
positive radiative feedback therefore negative numbers
seasonal variation is critical for this - idea that it can be measured and we can correlate this value to changes in climate - we can use this rule to evaluate model projections
Why is seasonal variation importat?
seasonal variation is a very good predictor of cliamte change feedback
models with large seasonal feedback have a large climate feedback etc.
the seasonal feedback can be estimated with good confidence
this allows us to rule out many of the model projections and to provide a most likely value for a true feedback like albedo in the future
What is detection and attribution?
detection = demonstrating that cliamte has changed in some defined statistical sense, without providing a reason for that change
attribution = establishing that the likely cause for the detected change with some defined level of confidence
important for both weather and climate - extreme events linked to climate change and the attribution of this to anthropogenic forces
What do you mean when you talk about the fingerprint of human activities?
human-induced ghg leave specific fingerprints on climate, distinguishable from the effects of natural variability, volcanoes and the sun or other forcing
why are human fingerprints useful to look at?
they are patterns in space and time
important for attributing cc to human activities
can help us determine what these fingerprints should look like
the pattern is the fingerprint is much more important than the mag of the signal - we know that climate models may get the mag wrong but they should get the pattern in space and time correct
What are the three key forcings that demonstrate the change of global temperature
the fingerprints of volcanoes, solar activity and human activity (CO2)
cannot explain current temperatures without solar forcing
they all have very different fingerprints
Can you describe the GHG fingerprint?
warming in the lower atmosphere but cooling in the stratosphere/upper atmosphere
observations show this fingerprint
How does anthropogenic fingerprints differ from natural and volcanic fingerprints?
NV= gives you a warmish enviroment more high altutde than in the poles
VF= depends on whether there has been an eruption - in 1979-2012 it gives you a warming pattern as there were more volcanoes earl on and then cooler because of volcanoes as they disappear it warms
Are there any other places other than the atmosphere where the human fingerprint is visible?
the oceans - ocean salinity
increase in atlantic and reduction in the pacific, has to do with evaporation and transpiration
How do you include small scale processes in a climate model?
these processes are simiplified - parameterised.
this is a significant source of uncertainty - what is the equation for a cloud
what constraints can we place on climate model behaviour?
we can match models to observations -eg. energy fluxes. emergent constraints also allow us to understand behaviour under climate change
how do we attribute climate change to different causes?
models can uncover the fingerprints or warming from different sources
running models with only natural forcings can provide a control earth for comparison
