| Experiment |
Observation errors for satellite winds at the Met
Office are routinely calculated from observation/background
differences. We assume that observation and background
are contributing equally to the overall difference.
For many years we have used errors roughly within
the range 1.5 m/s at the lowest levels to 5.5 m/s
at the highest. These figures are assumed errors
in each vector component.
For this experiment, we recomputed the errors based
on most recent data, doubled the values, then ran
the trial exactly as the current operational model,
but with the doubled observation errors. The
errors we used were as follows:
| hPa: |
1000 |
850 |
700 |
500 |
400 |
300 |
250 |
200 |
150 |
100 |
70 |
Component
error (m/s): |
3.6 |
2.8 |
4.0 |
4.8 |
6.2 |
6.2 |
5.6 |
5.8 |
6.6 |
11.8 |
11.8 |
|
Model and
assimilation
details |
The Met Office's Unified Model (UM) uses 3D variational
assimilation. It is a grid-point model running operationally
at 432(E-W) x 325(N-S). These trials were carried out
at 288 x 217 resolution due to computer resources.
The vertical resolution is 30 levels (hybrid-eta) |
| Run dates |
10 March - 18 April 2001; 21 June 2001 - 1 August
2001 |
| Highlights |
- Using a set of clearly defined forecast parameters
at different forecast ranges, and verifying against
observations, the averaged forecast RMS error decreased
in all latitude bands. NH: -0.4%, TR: -0.2%,
SH: -0.7% for Mar/Apr trial. NH: -0.2%, TR: -0.5%,
SH: -1.2% for Jun/Aug trial. It is noted
that the most positive impact is in the SH
- No forecast parameter was significantly degraded
- The T+6 forecast pulled closer to most other
observation types at validity time, but away from
satellite winds, as would be expected
- It was discovered that the impact of satellite
winds on tropical wind forecasts can be highly
variable, depending on the synoptic situation
|
| Publication |
NWP
SAF Technical Report NWPSAF_MO_TR_007 (pdf) |
Operational
consequences
and/or
further work |
- Due to the positive impact of this trial, the
Met Office's operational global model was changed
to use the new satellite wind observation errors
(Oct 2001)
- However, since the assimilation system should
operate well using errors as previously derived,
the consequences of these trials raises additional
questions concerning the overall quality control
of satellite wind observations. Further research
will be conducted to attempt to address these concerns,
particularly the signal in the tropics
|
