Exam 1 Flashcards

(31 cards)

1
Q

Describe the temperature distribution during the middle of the day in florida

A

Due to the Bermuda high temperature contours are shifted to western parts of the state

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Which models are global

A

GFS, ECMWF, GALWEM, HAFS-A?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Which models are regional

A

NAM, RAP, HRRR

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is the future for US modeling

A

The Unified Forecast System

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

How to deal with uncertainty

A

Perturbed (Moved) initial conditions which would lead to a range of outcome
Can also perturb the physics of a model

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What do ensembles attempt to do

A

Cover a range of forecasts

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What does the ensembles attempt do for us

A
  1. Allows us to estimate the predictability for an event
  2. Good argument among members usually implies high predictability
  3. Regardless of errors in initial conditions you can still get a valuable forecast
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Ensemble bias

A

the model is just systematically too cold or too wet

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Underdispersion

A

Estimate of initial conditions errors tend to be too small and the real atmosphere may behave in ways completely outside the realm of our physics equations. Ensembles don’t have enough spread

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Cone of uncertainty

A

Average of errors over the last 5 years that remains the same all year

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Machine Learning

A

Parameters are fit on a training data set in order to optimize a predefined loss function

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is the most common loss function for traditional linear regression

A

Residual summed squared error (RSS)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What are all ML essentially built on and is the simplest of models

A

Linear regression

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

In order to have a good forecast…

A
  1. Be consistent with users prior knowledge
  2. Have good quality (accuracy)
  3. Be valuable (provide benefit)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Neutral Networks

A

Are deep webs of information transfer.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Research modeling and its problems

A

-operational models, determine what improvements can be made
Problems: they are “tuned” and it’s difficult to do ‘true’ tests

17
Q

Research models

A

WRF, Cloud Model 1 CM1

18
Q

What can’t AI capture at the moment

A

Insintric predictabiltiy of the atmosphere: slightly different conditions may lead to the same “result” which is not physically possible

19
Q

What are the limits of NWP

A
  1. Equations can’t be solved exaclty
  2. Cannot measure the entire atmosphere
  3. Data has measurement errors
  4. Certain physical processes occur at scales below model resolution
  5. The true state of the atmosphere is never known
20
Q

Intrinsic Predictability

A

The ability to predict given nearly perfect representation of the dynamical system (by a forecast model) and nearly perfect initial boundary conditions, an inherent limit due to the chaotic nature of the atmosphere and cannot be extended by any means -> 2 weeks

21
Q

Practical Predictability

A

Same as forecast skill, the ability to predict given realistic uncertainties in both the forecast model and initial boundary conditions

22
Q

Anomaly correlation coefficient

A

Measures the correlation between forecast and actual observed deviations from climatology

23
Q

Errors in deterministic forecast accumulate _______

24
Q

Boundary conditions

A

How we handle the first and last grid points near our grid

25
Ways to handle grid points
-that them as constants -treat derivative as a constant -assume that grid repeats
26
For grid points that represent the entire earth…
Cyclic conditions can represent an entire latitude circle
27
True Error
The difference between the true value and the approximated value, can be positive or negatives
28
Relative error
Is the true error divided by the true value
29
NWP Sources of error
-truncation error: related to the transaction of the Taylor Series -Round-off error- computer cannot represent numbers with infinite accuracy
30
Truncation Error Behavior
Generally decreases with decreasing grid spacing
31
Order of accuracy
First order -> reducing delta x by half -> reduces error by 1/2 Second order -> reducing delta x by half -> reduces error by 1/4 Third Order -> reducing delta x by half -> reduces error by 1/8