Ensemble Prediction

Question 1

What is the traditional method of making a weather forecast?

Answer 1

Take the best model available and run it until it loses its skill due to the growth of small errors in the initial conditions

Question 2

Skill is typically lost after ____ days or so, depending on the season

Answer 2

6

Question 3

An alternate method that produces forecasts with skill up to ____ days after the initial forecast uses what is called “____ _____”

Answer 3

15 days

“ensemble forecasting”

Question 4

Instead of using just one model run, many runs with slightly different ____ conditions are made

Answer 4

initial

Question 5

An average, or “____ _____” of the different forecasts is created

Answer 5

ensemble mean

Question 6

Why will the ensemble mean likely have more skill?

Answer 6

• It averages over the many possible initial states

- Smoothes the chaotic nature of climate

Question 7

Is now possible to forecast probabilities of different conditions because of the ____ ensemble of forecasts available

Answer 7

large

Question 8

Every day, at __Z, ___Z, __Z, and __Z global weather observations are collected, transmitted to major weather centers, and are used to produce a snapshot of the global atmosphere

Answer 8

00Z, 06Z, 12Z, and 18Z

Question 9

What does the “snapshot” of the global atmos (i.e. the model analysis or 0-hour forecast) include?

Answer 9

Winds, pressure, moisture, temperature, etc. at multiple vertical levels in the atmos and at grid intersections of about 100 km

Question 10

Why does an individual model run merely “ sample” one of many possible current and future states of the atmospheric circulation?

Answer 10

• Because of the uncertainty in the initial conditions

Question 11

What are some examples of uncertainty in initial conditions?

Answer 11

• instrument error
• spatial/temporal sampling error
• lack of data
• erroneous influences of the first guess background field

Question 12

How does NCEP (and other weather centers) make many forecast runs out of a model out to 15 days?

Answer 12

• Make initial conditions of the individual runs slightly perturbed

Question 13

What is the most skillful numerical forecast that can be made given the chaotic nature of the atmospheric circulation?

Answer 13

The average of all these model integrations

Question 14

Somewhere between ___ and ___ numerical model runs are executed daily using slightly altered initial conditions

Answer 14

12 and 30

Question 15

These slight alterations are known as “____”.

Answer 15

perturbations

Question 16

These 12 to 30 model runs using perturbed initial conditions are called the ____ ____, and theoretically represent the current _____ ____ of the distribution of atmos states expected in the atmos out to forecast times of 15 days

Answer 16

ensemble members

best estimate

Question 17

One can also use the members to estimate probabilities of certain events, such as ____ ____ or ____ ____ normal temperatures

Answer 17

much below

much above

Question 18

Such probabilities are dependent on the ____ in the forecasts of individual members

Answer 18

spread

Question 19

The ensemble prediction approach attempts to define the ____ ____ ____ of atmosphere variables

Answer 19

Probability Density Function (PDF)

Question 20

This forecast scatter between ensemble members ____ with forecast lead time and should eventually approach the ____ observed in everyday weather

Answer 20

increases

scatter

Question 21

The scatter between separate ____ ____ should approach the ____ _____ scatter of the observed weather

Answer 21

model forecasts

normal climatological

Question 22

What are the 5 steps of the NWP Process?

Answer 22

1) Gather Observations
2) Data Assimilation
3) Numerical Weather Predictions
4) Forecast Postprocessing
5) Issue forecasts and evaluate

Question 23

What are the 2 categories of problems with ensemble predictions?

Answer 23

1) Chaos

2) Model Error

Question 24

As forecast lead time ____ the ensemble mean maps get _____

Answer 24

increases

smoother

Question 25

This increasing ____ reflects the fact that there is more _____ in the predictions at ___ lead times or further into the future

Answer 25

smoothness
uncertainty
long lead times

Question 26

The “____ or _____” distribution of states at the initial time is becoming more ____ out as individual members sample different but plausible future solutions of the atmospheric circulation

Answer 26

tight or peaked

more spread out

Question 27

Even when there are only slight differences, it is these differences that will result in a very different forecast of the future atmospheric state (due to ____)

Answer 27

chaos

Question 28

As one goes further in to the future, the lines start showing more and more spread or scatter so that they look like ______

Answer 28

spaghetti

Question 29

Why is the ensemble mean a smoother damped wave pattern?

Answer 29

Troughs and ridges which are superimposed on one another from different ensemble members cancel each other out in the mean

Question 30

With ensemble means, information on individual storms is what?

Answer 30

Essentially lost so that one member may be predicting high pressure in a region while another is predicting low pressure

Question 31

_____ smoothes out these differences. However, there may still be useful information at these “long-lead” forecast times, especially for the ____ scales of motion

Answer 31

Averaging

larger scales of motion

Question 32

What do the two heavy green lines represent on spaghetti plots? And how often do they change?

Answer 32

The two climatological lines of 500 mb height that correspond to the red and blue forecast lines.

They change slightly every day and are based on about 30 years of global data

Question 33

By comparing the high latitude ___ line with the ___ lines and the lower latitude ___ line with the ___ line, one can determine how far removed from climatology a particular region is observed or predicted to be

Answer 33

green with blue

green with red

Question 34

This suggests that the ensemble mean is approaching ____

Answer 34

climatology

Question 35

Sometimes, even for predictions beyond ___ days, most of the predicted lines may remain ___ or ____ their respective climatological lines

Answer 35

10 days (240 hours)
remain above or below

(due to ocean SST departures due to El Nino)

Question 36

At the initial time the spread is small so that the distribution must be strongly ____
i.e.: there is relatively ____ confidence about the current state of the atmos

Answer 36

peaked

high confidence

Question 37

Any spread is due to the small _____ introduced in the initial conditions of individual ensemble members. Are generally within the range of observational errors

Answer 37

perturbations

Question 38

Early on in the forecast integration or cycle, many locations over the hemisphere are ____ removed from their climatological value

Answer 38

far

Question 39

As the forecast lead time ____, the departures from climatology get ____ and the spread gets ______

Answer 39

increases
smaller
larger

Question 40

This means that the probability distribution is becoming less ____ and at very ____ lead times it should approach the climatological distribution

Answer 40

peaked

long lead times

Question 41

Once this climatological distribution is reached, any predictive skill present in the initial conditions of the atmosphere has been ____ and the “forecast” would be __ ____ than using climatology

Answer 41

lost

no better

Question 42

Model errors and the manner whereby the initial conditions are perturbed produces a ____ distribution that may deviate from the ____ distribution

Answer 42

probability distribution

climatological distribution

Question 43

Individual ___ ____ can systematically contaminate all the ensemble forecasts run on that model, regardless of how the initial conditions are perturbed

Answer 43

model biases

Question 44

One solution is to generate ensembles from a ____ of different models.

Name one example

Answer 44

cluster

i.e.: the FSU Superensemble

Question 45

What is a big contributor to model error?

Answer 45

Parameterizations

Question 46

Much of the weather occurs at scales ____ than those resolved by the weather forecast model.

Answer 46

smaller

Question 47

Models must treat, or _____ the effects of the sub-gridscale on the resolved scale

Answer 47

parameterize

Question 48

What are five examples of parameterization?

Answer 48

1) land surface characteristics and processes
2) cloud microphysics
3) turbulent diffusion and interactions with surface
4) Orographic drag
5) radiative transfer

Question 49

What is a historical weather event which is very similar or closely parallel to the current event?

Answer 49

weather analog

Question 50

Even if NWP centers produce calibrated ensemble forecasts at ____ scale, much of the important weather happens at the _____

Answer 50

grid scale

sub-gridscale

Question 51

Can solve this problem by finding analogs from the ___, and then use ____ weather observations from these analogs

Answer 51

past

actual

Question 52

______ _______ help to compensate for some of the effects of chaos

Answer 52

ensemble forecasts

Question 53

ensemble forecast technology is ____; better probabilistic forecasts with each passing year

Answer 53

maturing

Question 54

Still can’t expect ____ ____ data to provide reliable weather input to drive extended local forecasts without some adjustment

Answer 54

raw ensemble