|
NMOC Quarterly Summary April-June 2002
Synoptic Intercomparison of Global Models
Analyses & Numerical Prediction |
About Products |
Map/Image/Chart Archives
In this section I present a classification of errors into broad synoptic
categories with the aim of identifying systematic biases in the model predictions that may
assist forecasters in the use of the model output. This study looks at 3 day MSLP predictions
from the 1200 UTC global model predictions.
For the period April to June 2002 daily forecast error charts (i.e. forecast
MSLP minus the verifying analysis field from the same NWP system) were produced for each of the
models. For the Australian region (loosely defined as between 10° and 40°S from 100° to
140°E and between 10° and 45°S from 140° to 160°E) I assigned each case of forecast
error exceeding 4 hPa in magnitude to a synoptic pattern category as listed in the table below.
The categorisation is subjective and can clearly be ambiguous. Use of the system’s own analysis
is another weakness due to the strong influence of the model on the analysis field in data sparse
areas and the poor representation of the intensity of deep systems, particularlt tropical cyclones.
The table below lists the error patterns used in the classification and the number of cases
identified in each category for each global NWP system. A graphical intercomparison for the most
significant categories is also shown.
The US and ECMWF again had the most "error free" cases (i.e. cases with forecast errors
less than 4 hPa over the area of interest). In the same quarter in 2001 ECMWF was clearly
ahead on this broad measure. Possible factors behind the improvement in the performance
of the US predictions over the past year were discussed in the previous edition.
Troughs in easterly flow or cut-off low pressure systems
In the April-June quarter the synoptic pattern showed many cut-off lows moving
or developing over the region. While the models often have errors due to the
complex structure of many such systems, the models also displayed the systematic
tendencies repeatedly noted for this type of pattern, namely:
- A clear trend to link these systems too much into the higher latitude
westerly flow rather than maintain them as cut-off systems (i.e.
category 3A rather than 3B)
- A bias toward moving such systems too rapidly to the east or south-east
(i.e. category 6A over 6B). There were also cases where the systems were
located several degrees too far to the south, particularly for GASP, UK and US.
In the ECMWF predictions, these cut-off systems tended to be too deep, while
the US showed a bias to under-prediction.
Other general biases
Systematic biases mentioned in previous editions were again shown this quarter,
particularly:
- Ridging and high pressure systems tended to be too strong, most markedly
in the UK predictions. In the few cases where errors were caused by
displacement, the highs were located too far to the west, again most
noticeably in the UK predictions.
- Except for ECMWF there was a tendency to under-predict cyclogenesis,
rather than to develop spurious lows, particularly in GASP and JMA.
- Troughs in the westerlies were on average under-predicted in intensity
by all models, and in UK and GASP any errors associated with speed of
movement were strongly biased toward slow movement of the trough.
Specific problems
Despite the time of year, UK still showed problems with the over-development
of tropical lows. This bias had been evident in the previous quarter when
more active tropical systems could be expected, but was not evident in the
same quarter in 2001.
Although this is not the main problem season GASP again showed cases in
April and May of over-developing the characteristic trough over Western
Australia. Severe cases occurred in the forecasts for 11 and 12 April
with over-intensification of the trough by nearly 10 hPa, resulting
in a spurious synoptic pattern unique to GASP.
Particular cases
Some cases worth mentioning are:
- 15-18 April. A cut-off low developed west of Western Australia
and moved southeastwards over the Bight. The initial development
and subsequent evolution were generally poorly handled in the
72 hour predictions from all models. The predicitions showed
the characteristic tendencies noted above for this type of situation.
- May. The development of a deep low (about 965 hPa central
pressure) centred south of Western Australia was well predicted
by all the models, although all located the centre about 5
degrees too far south, resulting in large MSLP errors.
- The formation of a tropical low in the far north-west on 15 May
was missed by all the models, and the development of a subtropical
low at 28S, 168E on 27 June out of a tropical trough was missed
by all the models except UK.
- 12-16 June. A complex cut-off low moved steadily from the eastern
Indian Ocean, across the Bight and over the Tasman Sea. This
unusual migratory cut-off low was reasonably well-handled by the
models, apart from the final stage when a centre over the Tasman
Sea developed rapidly to become the primary centre. The ECMWF
predictions were best in showing this rapid reorganisation of the
system but still had errors exceeding 10 hPa.
Table. Number of occasions during April-June 2002 when error occurred in each synoptic
patters category in the three day MSLP prediction from each of the global models.
| Error pattern |
GASP |
ECMWF |
JMA |
UKMO |
USA |
|---|
|
1 No significant errors in pattern |
4 |
15 |
4 |
3 |
16 |
|
3 Low in easterlies or cut-off low
too linked to westerlies
too much in easterlies or cut-off |
9 1 |
11 0 |
12 0 |
3 0 |
6 0 |
|
4 Trough in easterlies or cut-off low:
too deep
too weak
details of structure
|
13 11 6 |
14 8 13 |
8 6 7 |
11 10 6 |
2 9 7 |
|
5 Trough in easterlies over WA:
too deep
too weak
|
5 1 |
1 1 |
1 0 |
0 2 |
0 0 |
|
6 Trough in easterlies or cut-off:
too mobile (i.e. too far east) too slow (i.e. too far west) shifted latitudinally
|
17 6 6 |
13 5 1 |
21 6 2 |
14 9 5 |
19 3 8 |
|
7 Ridging or high pressure system:
too strong too weak too slow
|
17 9 3 |
6 0 0 |
17 7 3 |
28 1 6 |
11 1 0 |
|
8A Spurious low or marked over-development
8B Missed development of a low or trough |
9 15 |
5 2 |
3 13 |
7 9 |
2 5 |
|
10 Trough in westerlies basically correct in structure but:
too strong too weak details of shape
|
8 21 3 |
7 14 10 |
14 22 3 |
13 17 5 |
8 19 0 |
|
12 Trough in westerlies:
too mobile too slow
|
4 16 |
9 5 |
12 9 |
4 14 |
6 5 |
|
13 Tropical lows (oceanic):
over-developed missed or too weak
|
3 1 |
1 4 |
1 1 |
10 2 |
1 0 |
Number of errors in the 3 day predictions from the global models
grouped into the synoptic pattern categories listed in the table above.
| 
