Lets not kid ourselves: there is no climate model complicated enough to explain this type of climate change with a great degree of certainty.
Agreed. The day computer models can accurately and correctly predict temperature to the nearest tenth of a dregree, exact wind speed, direction and change, and how much it will rain in the next week will they come close to accurately predicting climate change. Until then............
well here how it works according to the lecturers notes from last yr!!!!!........
global circulation models dice up the world into grid size boxes and the more boxes the more accurate, the choice of number of boxes is limited to computer power hence the longer the simulation the fewe grid boxes we can have
weather models run for a max of 10 or so days and climate models run for >100 years... that is why climate models have fewer grid boxes
the most advanced global weather forecast model has more than 75000000 grid boxes with a 20km horizontal resolution and 90 levels in vertical
one way of getting higher resotuion is to only run the model for a limited area. this requires the model fields to be prescribed at the boundary of that rea, usually from a gloval model simulation. this only works for 2-3 days after this the influnece of what happens outside the area becomes too large and the benefits of this internal high resolution are lost.
to determine the initial state for the model we employ a technique that optimally blends a short model forecast (6-12hrs) with observations (surface stations, ship observations, drifiting buoys, radiondes, proilers, aircraft, satellites) taken in that time interval...satellite data is most important in the southern hemisphere.
CHAOS
deterministic chaos is a feature of many nonlinear systems, including the atmosphere. it does not mean that things are random, but rather that if the initial value given to the equations is slightly different, the solutions can diverge significantly. this is called sensitivity to the initial conditions. each of the divergent solutions are still fully deterministic.
we dont know everything about the present day climate to certaintty!! chaos is therefore inevitable
Ensemble prediction
the initial conditions for our forecast are uncertain. due to the chatoc nature of the atmosphere, small erros in the initial state can lead to large erros in the forecast. techniques have been developed to estimate the uncertaintes in the initial state. using these estimates the model is run not just once but 50-100 times using different possible inital states. this is called ensemble predictions....(which is not done in australia yet...but places like japan...)
HERE IS MORE INFO...
physical models are dependant on our understanding of the physics...
statistical models... are based on observations tather than an understanding of the physics, which is useful when we don't understand the physics completely or can't solve the complete physics in a computer , eg. seasonal prediction..
often hard to isolate the statistics of one system since in nature feedbacks are always operating
global circulation models take into account atmosphere, ocean land, vegetation...
GCM is solved........- space and time give 4 dimensions x,y,z,t - sometimes use different dimension models to solve different problems eg. 1D energy budget models, 2D,3D,4D are used to attack different problems.
computational limitations, - like 1D radiative equilibrium model, equations need to be calculated at each grid point... the constraints are that supercomputers have limited power and you need results in a reasonable amount of time...say your supercomputer can do X computations per second. we require an answer in 1 week. then you can only afford around X * 6 *10^5 computations.... so we need more computer power!
more physics means more accurate predictions? more physics= more feedbacks, more feedbacks= more avenues for error... we know the physics is not perfect in these models yet we keep adding more processes to these models. this seriously undermines our ability to understand what is happening in these models eg the rapidly dieing amazon perhaps..
some known problems... Sub grid scale processes.. turbulend, clouds, evaporation etc... CLOUDS in nature are not square blocks!!!!
parameterisation. statiscal or resuced physical model of a processes. much of the variablity amongst different models are due to parameterisations...
how can we trust these models?
ensemble model runs, run the model many times..
intercomparison, compare different models
compare to past and present climate, try simulate past and present climate to which we have observations... does this mean we can trust them? probably not entirely i reckon
ensembles run a model many times with slightly different conditions, this give a range of solutions that define the chaos in the system!!! this gives you confidence in the forecast
there is much less certainty in precipitation projections than temperature!
comparison to past and present climate... these comprations offer the model makers an opportunity to tune the model parameterisations to match observations...
GCMS are more accurate on a global scale than smaller scales such as cities..
GCMS are useful so long as we recognize which statistcal they predict well.... not tornadoes!
So anyway yeah lots of processes to deal with in climate or weather models and there is always uncertainty!!!!!!!!
Home
Help
Search
Login
