Academia.eduAcademia.edu

Figure 3 – uploaded by Henrik Vedel

See full PDFdownloadDownload figure
Further studies are ongoing with an enchained system with regard to tuning of error statistics and observation bias correction and observation thinning. Fig. 4. Day to day verification of specific humidity forecasts (unit: g/kg) for 850hPa against 10 stations over Spain-Portugal area (a), and verification of 12 h accumulated precipitation forecast by the Kuiper Skill Score using 115 SYNOP stations over Spain and Portugal ,both figures for 16-30 September 2012. Red curve is for CRL, green curve for STA and blue for VBC.

Figure 4 Further studies are ongoing with an enchained system with regard to tuning of error statistics and observation bias correction and observation thinning. Fig. 4. Day to day verification of specific humidity forecasts (unit: g/kg) for 850hPa against 10 stations over Spain-Portugal area (a), and verification of 12 h accumulated precipitation forecast by the Kuiper Skill Score using 115 SYNOP stations over Spain and Portugal ,both figures for 16-30 September 2012. Red curve is for CRL, green curve for STA and blue for VBC.

Related Figures (2)

To illustrate how a ZTD observation modifies the background state a single observation impact experiment has been carried out. The single observation was located in the inner part of Spain (40.5'N/3.1 W, red dot in Fig. 1), and the observed ZTD was 24 mm higher than the corresponding background value. The main impact is on the specific humidity field of the model state, but there is also a weak impact on other model variables. This influence is due to the ZTD GNSS observation operator and due to the background error statistics. The resulting horizontal and vertical spread of the ZTD observation on the specific humidity field is illustrated in Fig. 2. The current system give rather large weight to the GNSS ZTD observations and the observation minus analysis departure is 5 mm. Fig. 2. Impact of single GNSS ZTD observation 236 mm higher than the background value on model specific humidity field (unit: kg/kg times 10000). Both horizontal distribution of impact at the 850 hPa level (a) and vertical/horizontal distribution of impact along a cross- section from 40.5N/6W to 40.5:N/O'W (b) is shown. Red iso-lines are for analysis increments and blue for the background state.
To illustrate how a ZTD observation modifies the background state a single observation impact experiment has been carried out. The single observation was located in the inner part of Spain (40.5'N/3.1 W, red dot in Fig. 1), and the observed ZTD was 24 mm higher than the corresponding background value. The main impact is on the specific humidity field of the model state, but there is also a weak impact on other model variables. This influence is due to the ZTD GNSS observation operator and due to the background error statistics. The resulting horizontal and vertical spread of the ZTD observation on the specific humidity field is illustrated in Fig. 2. The current system give rather large weight to the GNSS ZTD observations and the observation minus analysis departure is 5 mm. Fig. 2. Impact of single GNSS ZTD observation 236 mm higher than the background value on model specific humidity field (unit: kg/kg times 10000). Both horizontal distribution of impact at the 850 hPa level (a) and vertical/horizontal distribution of impact along a cross- section from 40.5N/6W to 40.5:N/O'W (b) is shown. Red iso-lines are for analysis increments and blue for the background state.
The functionality of the variational bias correction is illustrated by a showing the time-evolution of observation and background counterparts for one particular station (37.6'N/0.7 W, blue dot in Fig. 1) during the spin-up phase, when GNSS ZTD observations enter the system in passive mode only. Fig 3. demonstrates that the VarBC removes bias of observation with a adaptively time of roughly 10 days. Fig. 3. Time-series from CABOIGE_ GNSS ZTD station. (a): Observation minus background departure before and after bias correction of observation, estimated observation bias to be corrected and bias corrected observation minus analysis departure. (b): Bias corrected observation and model state background equivalent. (unit: mm).
The functionality of the variational bias correction is illustrated by a showing the time-evolution of observation and background counterparts for one particular station (37.6'N/0.7 W, blue dot in Fig. 1) during the spin-up phase, when GNSS ZTD observations enter the system in passive mode only. Fig 3. demonstrates that the VarBC removes bias of observation with a adaptively time of roughly 10 days. Fig. 3. Time-series from CABOIGE_ GNSS ZTD station. (a): Observation minus background departure before and after bias correction of observation, estimated observation bias to be corrected and bias corrected observation minus analysis departure. (b): Bias corrected observation and model state background equivalent. (unit: mm).
Related topics:
GNSS applicationsForecast models
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved