21. Computer Models Flashcards
To study the ocean, we use theory, observations, and:
numerical models
_______ can observe from ocean processes almost everywhere every few days. But they can only observe some processes, and only near the sfc
_____ can measure more variables, and deeper into the water, but measurements are sparse
___ ____ provide the only useful, global view of ocean currents
satellites
ships
numerical models
What is a model?
(according to Entry 4 and Entry 12 of the Webster Dictionary!)
Entry 4= a usually miniature representation of something
Entry 12= a system of postulates, data, and inferences presented as a mathematical description of an entity or state of affairs
any kind of representation of the real world=
a model
Numerical model=
a set of equations and parameters that describe the time and spatial evolution of physical processes that occur within the earth system (oceans here)
Most models are what type of equation?
differential equations
Give the 3 main elements of a model
- governing equations
- initial conditions (temp, salinity, & velocity fields)
- Boundary conditions (atmosphere above, basin geometry, bottom topography)
Describe discrete vs continuous models
Continuous: dt and dx are very very small (graph is a solid, smooth line)
Discrete: chunky (bar graph)
Analogy: continuous= water running from a tap as a steady stream, while discrete= random drips, no steady stream
What are 3 air-sea interactions that a 3D ocean model would consider?
water vapor
heat
momentum (winds)
Ocean processes need to be parameterized for 2 main reasons:
- we don’t want to spend the computational resources to directly treat/ resolve them as they’re too small and complex
- we don’t completely understand the process
In low resolution models, what is the main issue?
meso-scale eddies
In high resolution models, what are 2 main issues?
- internal wave mixing
- flow-topography interactions
List 3-5 ocean processes that are commonly parameterized
- mesoscale eddy effects
- coastal processes
- ice-ocean mixing
- friction
- surface mixed layer processes
Calibration=
adjustment of independent variable (parameters) in order to match observations
Describe what a parameterization is
= a model within a model
Things the models do not resolve, we have to provide them those things
eg. coefficient of diffusion is a constant given to the model
Give 4 ‘requirements’ that a valid model should meet
- contain no coding errors
- have sound metrics
- give the same results on different computers
- agree with analytic solutions (where available)
- This is tricky for oceans: analytical solutions do not exist. & comparing with observed data is not ideal
Give up to 4 problems with comparing models with observed data for ‘validation’
- limited data coverage
- time variability
- spatial scale of model vs data (hard to collect observed data)
- many unresolved processes
Give up to 6 uses of models
- hypothesis testing
- sensitivity analysis
- what-if scenarios
- estimation of hard-to-measure quantities
- extrapolation
- forecast
Ocean general circulation models are based on a series of coupled partial differential equations. The full set involves __ (how many).
7
Parameterizations and results of ocean general circulation models depend on what?
grid resolution
What are the main 8 processes considered in climate models?
- atmosphere
- land sfc
- ocean & sea ice
- aerosols
- carbon cycle
- dynamic vegetation
- atmospheric chemistry
- land ice
T/F
The complexity of climate models has increased a lot over the last few decades
true
Use FAR, SAR, TAR, and AR4 as examples of improvements in model resolution
FAR (1990)= 500km resolution, not good
SAR (1996)= 250km resolution, better
TAR (2001)= 180km resolution: much better
AR4 (2007)= 110km resolution (can clearly see topography)