bullet  EUMETSAT   bullet  SAFs   bullet   NWP SAF   bullet  ATOVS monitoring

Observation and forward model errors


The errors in kelvin currently being assumed in 1D-var are:
 
HIRS 1-10 0.40  0.40  0.40 0.40  0.40  0.40  0.40  0.40  ---  0.40
HIRS 11-20:  0.40  0.40  0.40 0.40  0.40  ---  ---  --- ---  ---
AMSU 1-10  ---  ---  --- 0.50 0.24  0.24  0.24  0.24 0.30  0.30
AMSU 11-20  0.35  0.45  0.55 0.95 ---  (3.00)  (4.00)  (2.00) (2.00)  (2.00)
MSU 1-4 ---  0.40  0.40 0.40
SSU 1-3 0.40  0.40  0.40
 
The errors in kelvin currently being assumed in 3D-var are:
 
HIRS  1-10  ---  0.50  0.50  0.50  0.50 0.50  1.22  6.32  --- 6.32
HIRS  11-20  3.00  4.00  ---  ---  0.50 --- ---  ---  ---  ---
AMSU  1-10 --- ---  ---  1.26  0.25 0.25 0.25  0.25  0.40  0.40
AMSU  11-20 0.50  0.95  ---  ---  --- (3.00) (4.00)  (2.00)  (2.00)  (2.00)
MSU  1-4 ---  0.40  0.40  0.40
SSU  1-3 --- ---  ---
 
Notes:
1. --- denotes where a channel is not used.
2. A value in brackets e.g. (3.00) denotes where a channel has only been used in experimental assimilation.
3. The combined observation and observation operator error covariance matrix is assumed to be diagonal.
4. The error variances assumed in 3D-var are higher than those assumed in 1D-var. This is partially to allow for horizontally correlated errors.
5. The errors assumed in 1D-var are based on observation minus background standard deviation.
6. In 1D-var moisture channels are often given high weight e.g. HIRS channels 11 and 12. However in 3D-var they are strongly downweighted to avoid excessively large analysis increments.