Climate modelling

Question 1

Why is there limits to weather forecasting?

Answer 1

The dynamics of atmosphere and ocean are partially chaotic.
Initial conditions are critical – the further from the present moment the simulation evolves, the greater the error of the forecast.
Climate models are less subject to the initial conditions problem. They generate [daily] weather over many years (not just a few days) and their output can be summarised as climate statistics.

Question 2

Start simple…….

Make process models of different components, e.g.

Answer 2

Make process models of different components, e.g.
Energy exchange- radiation
Land-surface

First test: Can the model reproduce observations? If so, what does it simulate with a perturbation? * it’s mean ought to be approximately the same as that of the real world

Then the model can be used to see what happens when you change things
*Changing something in the system

Question 3

INFORMATION TO CONSTRUCT A MODEL

Answer 3

Radiation fluxes
Dynamics of fluids
Moisture fluxes and processes
Surface processes
Surface-atmosphere energy exchanges

Question 4

An ocean (OGCM) model needs to deal with (for example)

Answer 4

Energy exchange among levels 
Salt transport 
Upwelling 
Horizontal and vertical circulation 
Sea ice

Question 5

All models must be provided with…

Answer 5

Starting values. Initial conditions affect subsequent output (“butterfly effect”)

Question 6

Solutions:

Answer 6

long ‘spin-up’ – especially climate models
or
ensemble runs – many runs with different random starting conditions – weather forecasting models and climate models

Question 7

Specification

Answer 7

Variable is prescribed– cannot be derived interactively by processes in the model
Problem – a reduction of interactive effects including feedbacks (e.g. ice may not be able to melt)
Models now have interactive oceans, and usually interactive vegetation, but other features may still be specified (eg soil properties)

Question 8

Data assimilation

Answer 8

Used particularly in weather forecasting but also in palaeoclimate modelling.
Start model, run forward, assess model output and compare with observed data. Repeat.
Can do this for daily weather, hurricane prediction, or use palaeodata on long time scales

Question 9

Parameterization

Answer 9

Some processes happen at scales far smaller than the grid size processes represented with simplified (and thus probably inaccurate) equations example: many cloud-formation processes
Parameterizations can be evaluated using perturbed physics ensembles

Question 10

Models differ in their output even with the same forcing (such as future emissions scenario)

Answer 10

Different packages each have a slightly different depiction of physics.
Easier to modify a current package than to write a new one, so as models develop they tend to diverge
Give models same task, then diagnose why they differ in their performances

Question 11

How fast do the components of the climate system respond to forcing?

Answer 11

“Fast” feedbacks and equilibration
Water vapour, clouds – ‘instantaneous’ (hours, days)
Snow, sea-ice – seasonally
Vegetation/carbon changes – minutes/hours, seasonally

“Slower” feedbacks and equilibration
Vegetation (drought, treeline shifts) – decades/centuries
Ocean carbon cycle, conveyor belt, other terrestrial changes (e.g. glaciers) – centuries

Question 12

HOW DOES A MODELLING EXPERIMENT GET DONE?

Answer 12

It is a controlled experiment
Baseline run (control: usually present day) – no perturbations
Experimental run – perturbation
Compare control and experiment

DEALING WITH TIME
Change is either executed instantaneously (2 equilibria: before and after)
OR change is executed gradually over many time steps (transient)

Question 13

COMPARING EQUILIBRIUM AND TRANSIENT RUNS Equilibrium runs

Answer 13

– cheaper

• exploratory
• often overestimate change in temperature compared with a transient run with the same overall forcing
• used for palaeoclimate experiments, eg the Eocene, Snowball Earth

Question 14

COMPARING EQUILIBRIUM AND TRANSIENT RUNS

Transient runs

Answer 14

• more expensive

- more realistic – and provide time series output - used for 21st century warming projections

Question 15

SOURCES OF UNCERTAINTY IN MODELLING EXPERIMENTS

Answer 15

1. Perturbation a complex set of responses
   The particular construction of a model (e.g. the specific parameterizations it uses) may make it more or less sensitive to a perturbation
   (Hence model ensembles – as in the IPCC)
2. Some feedbacks take time to kick in and may increase the initial forcing – but how much? Some of these are not yet incorporated in models (e.g. permafrost thaw).
3. Future scenarios (e.g. GHG emissions values that are input to models) form a wide range of possible futures

Question 16

Key Points

Answer 16

• Models allow testing of hypotheses about the climate system, a highly complex system.
• Necessary simplifications lead to inaccuracies
• Model design attempts to reduce impact of this
• Each model is unique: they may share some packages but no two model codes are identical, thus models do not simulate exactly the same values for a given experiment
• Different components of the Earth System respond to perturbations at different rates.
• Some responses are important in short-term weather
  forecasting, others in climate change projections