The tubing is a method designed to classify ensemble forecasts. The aim of the classification is to condense the information coming from the 51 ensemble forecasts. In the traditional clustering, ensemble forecasts are grouped into a few clusters, each cluster being represented by its mean field. In the case of the tubing, there is only one cluster, the so-called central cluster grouping those forecasts that are similar to the ensemble mean. The other forecasts are grouped into a few tubes indicating different deviations from the ensemble mean that can be found in the ensemble. Each tube is represented by its extreme i.e. the forecast which is the most different from the ensemble mean in the direction of the tube.

How to use TUBE products?

- The central cluster mean shows the evolution which is the most likely to occur, on which can be based the most likely weather forecast, that is the "deterministic core" of the forecast. The central cluster mean field has a level of smoothing increasing with the time-step, allowing to give more details on day 3 than on day 8. This level of smoothing is day-to-day consistent for a given time-step (it depends on the internal variance of the central cluster which follows a slow seasonal trend, independent on the ensemble spread variations).

- The extremes of the tubes are not alternatives: they are extreme representatives of the tubes, almost caricatures, allowing to better visualise the different tendencies present in the ensemble, by contrast with the central cluster mean forecast. The significant tendencies, in terms of weather parameters over the area of interest for instance, can be used to describe

possible meteorological variants.

- The number of possible variants is a good indicator of the uncertainty of the deterministic

forecast based on the central cluster: if there is no significant variant the confidence is high; if there are one or two variants, the confidence is normal; if there are several variants, the confidence is low.

Reliability of TUBE products:

TUBE products are relatively reliable: the EPS is not perfect but its performance has dramatically improved since December 1996.

- The verification is not always indicated by the central cluster or one of the tubes: there is still a proportion of 10 to 20% of cases when the ensemble misses the verification.

- The verification is much more likely to be found in the central cluster than in a tube. The likelihood of the central cluster varies according to its population, but is generally grater than 50%, unless when the spread is especially large.

- The verification may be found in a tube, even if it contains only few forecasts. As a general rule a tube likelihood is around 10%, whatever its population.

For each tubing reference step (+96h, +144h, +168h, +192h, +240h), tubing products are generated over a 48-hour sequence finishing on the reference step (e.g. +48h/+72h/+96h for the 96h tubing) allowing a sequential view of the different tendencies. (In the case of the 168h tubing, the sequence is over 96 hours from +72h to +168h.)

Tube is the central cluster mean. This forecast is not exactly the same as the ensemble mean, but generally very similar. The remaining tubes (maximum 9) are the extreme members, sorted according to decreasing distance from the central cluster. Tubes are computed based on the 500 hPa geopotential at the reference step over each of the five geographical domains Europe, NW Europe, NE Europe, SW Europe and SE Europe. The results of these computations to determine the tubing structure are then applied to 500 hPa and 1000 hPa geopotential and 850 hPa and 500 hPa temperature.

* On pressure levels 1000/500 ** On pressure levels 850/500

Table 5.8 TUBE products

FIELD CODE	MARS ABBREV.	FIELD NAME	UNITS	NO. OF BITS/VAL.
129	Z	Geopotential*	m2s-2	16
130	T	Temperature**	K	12

TUBE requirements

The following command language parameters are specific for the TUBE requirements. Other parameters are listed under the general EPS requirements above.

TYPE

TUBE

[TU]

DOMAIN

General European Area

North West Europe

North East Europe

South West Europe

South East Europe

[G]

[A]

[B]

[C]

[D]

(75.0 -20.0 30.0 45.0)

(70.0 -27.5 40.0 10.0)

(72.5 0.0 50.0 45.0)

(57.5 -15.0 32.5 17.5)

(57.5 2.5 32.5 42.5)

NUMBER

OFF

LEVTYPE

PRESSURE LEVEL

[PL]

LEVELIST

1000/500

850/500

(PARAM=Z only)

(PARAM=T only)

GRID

a/b (latitude-longitude grid)

a= grid resolution (deg.) along a latitude line (i.e. East/West)

b= grid resolution (deg.) along a longitude line (i.e. North/South)

1.5/1.5 or multiples thereof;

REFERENCE

* (hours)

OFF (to switch off the reference);

gives product's reference time step (for the values see definition of the parameter STEP below);

STEP

gives tube time step(s);

48/72/96

96/120/144

72/96/120/144/168

144/168/192

192/216/240

(hours)

for the REFERENCE=96;

for the REFERENCE=144;

for the REFERENCE=168;

for the REFERENCE=192;

for the REFERENE=240;

PARAM

Field name abbreviation

NB ALL is not permitted.

Available parameters are listed in the Table 5.3.

GRIB section 1 definition for TUBE products

The set of local definitions for EPS Event probability fields:

Table 5.9 GRIB section 1 definition for TUBE products

Octet

Ksec1

element

Contents

ECMWF local GRIB use definition identifier

10 = EPS tubes

44-45

46-49

Class:

1 = Operations

Type:

24 = Tubes

Stream:

1035 = Ensemble forecasts

Version number or experiment identifier.

(4 ASCII characters, right justified)

Tube number (0=central cluster)

Total number of tubes (excluding central cluster)

Central cluster definition

1: radius =% of total variance

2: radius = predefined value

Indicator of parameter considered (see code table 2 in

section 1)

Indicator of type of level considered (see WMO code

table 3)

55-56

58-60

61-63

64-66

Northern latitude of the domain of tubing

Western longitude of the domain of tubing

Southern latitude of the domain of tubing

Eastern longitude of the domain of tubing

69-70

Number of tube to which the operational forecast

belongs(*)

0 = central cluster,

254 = does not belong to any tube

Number of tube to which the control forecast

belongs(*)

0 = central cluster,

254 = does not belong to any tube

Height/pressure of level considered (see WMO code

table 3)

71-72

73-74

75-76

Reference step considered (same units of time as fore-

cast timesteps)

Radius of central cluster (in units of parameter defined

in element 45)

Ensemble standard deviation (in units of parameter

defined in element 45)

77-78

80->(79+N)

59->(58+N)

Distance of the tube extreme to the ensemble mean (in units of parameter defined in element 45).

Not applicable if this is the central cluster, in which

case the value is set = 65535, i.e. a missing value.

Number of forecasts belonging to the tube or central

cluster, including the control forecast (N)

List of N ensemble forecast numbers (**)

Order is important, first on the list is the tube extreme

18-01-2011