Modelling

Front 1

What is a model?

Back 1

A simplification/abstraction/analogy of reality.

Front 2

What are the different types of models?

Back 2

- Empirical models

- Theoretical models

Front 3

What is an empirical model?

Back 3

Empirical models are based on observations (data driven). They are good for prediction, but not for explanation.

Front 4

What is a theoretical model?

Back 4

Theoretical (or physically based) models are based on process representation (theory driven). They are often based on assumptions and need a lot of measurements to fully implement them. They are not always good for prediction but can have high explanatory power.

Front 5

Which type of model is best for prediction?

Back 5

Empirical models.

Front 6

Which type of model is best for explanation?

Back 6

Theoretical models.

Front 7

How can the predictive capability of theoretical models be improved?

Back 7

Calibration to observations.

Front 8

What are the objectives of a model?

Back 8

- To facilitate understanding by emulating unnecessary components

- To aid in decision making by simulating 'what if' scenarios

- To explain, control, and predict events on the basis of of past observations

Front 9

What is diagnostic modelling?

Back 9

It aims to understand/investigate important processes.

Front 10

Give an example of diagnostic modelling.

Back 10

Payne et al. (2014) - Understanding how quickly inland ice can respond to changes at the periphery.

Front 11

What is prognostic modelling?

Back 11

It is used to make future predictions/forecast and assess the likelihood of something occurring based of past reconstructions for example. Useful for long timescales that we can't observe over.

Front 12

Give an example of diagnostic modelling.

Back 12

Shepherd et al. (2001) - "Inland Thinning of Pine Island Glacier, West Antarctica"

They used satellite altimetry and interferometry. The grounded PIG thinned by up to 1.6 metres per year between 1992 and 1999. They found the thinning can't be explained by short-term variability in accumulation thus must result from glacier dynamics.

Front 13

Describe the study by Payne et al. (2004).

Back 13

"Recent dramatic thinning of largest West Antarctic ice stream triggered by oceans"

Used a numerical model of ice flow to determine whether an oceanic trigger can affect the deep interior of the ice stream on the appropriate timescales. Results show changes in PIG's ice shelf and/or ice plain can be transmitted rapidly upstream on decadal timescales. Therefore likely thinning is a response to recent changes in the oceanography of the Amundsen Sea.

Front 14

Give an example of regression modelling.

Back 14

Adalgeisdottir et al. (2003) - "A regression model for the mass-balance distribution of the Vatnajokull ice cap, Iceland"

They found that these variables could explain 96% of the spatial and temporal variance observed in the mass-balance data.

Front 15

What is numerical modelling?

Back 15

A simplification/abstraction of reality (in process and in space and time).

Front 16

Describe the Gladstone et al. (2012) prognostic modelling example.

Back 16

"Calibrated prediction of PIG retreat during 21st and 22nd centuries with a coupled flowline model"

This was a 1D flowline ice sheet model coupled to a box model for cavity circulation.

Front 17

Describe the Cornford et al. (2015) prognostic modelling example.

Back 17

"Century-scale simulations of the response of the West Antarctic Ice Sheet to a warming climate"

BISICLES adaptive mesh ice sheet model was used. One, two and three century simulations were carried out of the fast-slowing ice streams of the West Antarctic Ice Sheet.

Front 18

Name the steps in the modelling process.

Back 18

1. physical process

2. mathematical description

3. computer code

4. application

Front 19

Define constant.

Back 19

Quantity whose value does not vary in the target system (e.g. speed of light).

Front 20

Define parameter.

Back 20

Quantity whose value is constant in the case considered but may vary in different cases (e.g. total solar radiation at the top of Earth's atmosphere).

Front 21

Define variable.

Back 21

Quantity whose value may change freely in response to the functioning of the system (e.g. amount of precipitation).

Front 22

Define relation.

Back 22

Functional connection or correspondence between two or more system elements (e.g. rainfall, run-off and soil erosion).

Front 23

Define relationship.

Back 23

State of being related.

Front 24

Define process.

Back 24

Operation or event, operating over time (temporal process) or space (spacial process) or both, which changes a quantity in the target system (e.g. evapotranspiration).

Front 25

Define scale.

Back 25

Relative dimension, in space or time, over which processes operate and measurements are made (e.g. local, regional, global; diurnal, seasonal, annual).

Front 26

Define structure.

Back 26

Manner in which component parts of a system are organised.

Front 27

Define system.

Back 27

Set of related elements (e.g. constants, parameters and variables), the relations between them, the functions or processes that govern these relations and the structure by which they are organised (e.g. forest ecosystem, drainage basin, global carbon cycle, Earth's climate).

Front 28

State a process that occurs in the glacier system.

Back 28

Internal deformation.

Front 29

State a relation that exists within internal deformation.

Back 29

Ice velocity is related to the gravitational driving stress.

Front 30

Why do we not want to always represent systems as a whole?

Back 30

- It may be too difficult numerically (e.g. turbulence)

- We may only be interested in the large scale result

- May not have enough data across the model domain (e.g. infiltration rate)

Front 31

State the two types of parameters.

Back 31

Physical parameters and process parameters.

Front 32

Give examples of physical parameters.

Back 32

Cliff height, basin area and soil height.

Front 33

What are the two types of process parameters.

Back 33

Physical resemblance and non-physical resemblance.

Front 34

Give examples of process parameters.

Back 34

Infiltration rate, stomatal resistance and weathering rate.

Front 35

What is a physical resemblance parameter?

Back 35

Require field measurements (e.g. infiltration rate).

Front 36

What is a non-physical resemblance parameter?

Back 36

Requires optimisation to find the optimal parameter values (e.g. ice flow parameter).

Front 37

What is a sensitivity analysis?

Back 37

An investigation into how a model responds to changes in the parameter values.

Front 38

Why do parameters need calibration?

Back 38

Physical parameters may not have measurements for the whole domain. Non-physical parameters need to tune the parameter values to give best result.

Front 39

What is calibration?

Back 39

Measures of assessing model prediction (overlaps with validation).

Front 40

How is a model calibrated?

Back 40

Root Mean Square Error (RMSE) .

Front 41

What is the split sample approach?

Back 41

It only uses some of the observations for calibration and then it is compared with predictions.

Front 42

What is validation?

Back 42

This abstraction represents a complex reality in the simplest way that is adequate for the purpose of the modelling.

Front 43

What is the difference between validation and verification?

Back 43

Verification involves converting equations into computer code. Validation: is the model formulation itself fit for purpose.

Front 44

List the different types of verification/validation.

Back 44

- Analytical solutions

- Field data

- Comparison with other models

- Benchmarking

- Scale models

Front 45

Outline limitations to model validation.

Back 45

- It is often vague and requires subjective judgement

- There may be a lack of data if it was all used in calibration

- Limited available data may fit different models equally well

Front 46

Outline questions that need to be addressed in experimental design.

Back 46

- What is the research question?

- What scenarios do I need to set up to answer the question best?

Front 47

State a consideration that must be taken when using a pre-existing model.

Back 47

It may need calibrating/validating for the application.

Front 48

Why is understanding uncertainty important?

Back 48

People make decisions based on predictions/projections (policy makers, industry). Identifying a worst, best and most 'likely' outcome can be useful.

Front 49

List the sources of model uncertainty.

Back 49

- Parameter uncertainty

- Input data uncertainty

- Uncertainty in the forcing conditions

- Conceptual uncertainty

Front 50

What is parameter uncertainty?

Back 50

It comes from the model parameters that are inputs to the computer model (mathematical model) but whose exact values are unknown to experimentalists and cannot be controlled in physical experiments, or whose values cannot be exactly inferred by statistical methods.

Front 51

What are parameter 'ensemble' approaches?

Back 51

They are part calibration and part uncertainty. A model is run many times with different sets of parameter values. The predicted and observed responses are compared and each set of parameter values is assigned a likelihood of being a simulator of the system (i.e. may be 0 when it is not a good predictor). All simulations with a likelihood measure significantly greater than zero are retained for consideration.

Front 52

Give an example of a parameter 'ensemble' approach.

Back 52

GLUE (Generalised Likelihood Uncertainty Estimation) by Bevan and Binley, 1992.

Front 53

What is input data uncertainty?

Back 53

It refers to the fact that computers will unquestioningly process unintended, even nonsensical, input data ('garbage in') and produce undesired, often nonsensical, output ('garbage out').

Front 54

What is conceptual uncertainty?

Back 54

It links back to the validity of the model.

Front 55

How can different conceptual approaches be compared?

Back 55

Model intercomparison exercises.

Front 56

Give examples of model intercomparison exercises.

Back 56

- CMIP (Coupled Model Intercomparison Project)

- PMIP (Paleoclimate Modelling Intercomparison Project)

- EISMINT (European Ice Sheet Modelling INiTiative)