1 / 19

RO assimilation at the Meteorological Service of Canada

RO assimilation at the Meteorological Service of Canada. Josep M. Aparicio Godelieve Deblonde Aug 23/24, 2005. Overview. We have analyzed the properties of the RO data, its coherence vs MSC models, and explored the impact of a future operational assimilation

talisa
Download Presentation

RO assimilation at the Meteorological Service of Canada

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. RO assimilation at theMeteorological Service of Canada Josep M. Aparicio Godelieve Deblonde Aug 23/24, 2005

  2. Overview • We have analyzed the properties of the RO data, its coherence vs MSC models, and explored the impact of a future operational assimilation • Some assimilation cycles have been performed • Systematic biases present (data & model) • Consistently positive impact will depend on a good choice of bias description

  3. Comparison models • 2 models have been compared • GEM (Global Environmental Multiscale) • Standard version • Currently operational, now in 4DVar mode • 400x200x28 grid • Low lid (10 hPa) • Stratospheric extension • Undergoing tests • 400x200x80 grid • High lid (0.1 hPa) • Mesoscale extension • Undergoing tests • 800x600x58 grid • Low lid (10 hPa) • CMAM (Canadian Middle Atmosphere Model) • Research model • 96x48x72 grid • Very high lid (0.000575 hPa) • Total of 4 versions

  4. Objectives • Comparison models are different • Cover some range of horiz/vert resolutions • Different vertical coordinates • Terrain-following η • Hybrid • terrain-following (low altitude) • non-following (high altitude) • Several approaches for physical parameterizations • Study of statistical coherence of RO across models • Testing of the observation operator with realistic data and atmosphere properties against subtle effects & coding errors • Example of issues identified • Ellipsoidal vs spherical latitude • Geoid offset • Gravity dependence with height • Compressibility of WV • Non-gaussian statistics • Properties of the new info fed during assimilation

  5. Observation theoretical accuracy • The obs error is smaller than Obs-Model scatter. • 0.2-0.5% strato • 1% low tropo • The rest is model error: • Obs useful to assimilate Image by E.R.Kursinski

  6. Obs-Forecast (6h) I Low lid models: GEM-operational & GEM-mesoscale FRAC: fraction of profiles covering a given height

  7. Obs-Forecast (6h) II High lid models: GEM-stratospheric & CMAM (mid-atmosphere) FRAC: fraction of profiles covering a given height

  8. Obs-Forecast (6h) III • Each model has a characteristic bias signature • O-P seems to be dominated by model resolution (not data accuracy!) • Refractivity ok. No benefit from bending angle. • Mesoscale version seems the less biased & more accurate a-priori • Some common features (data biases?) • Negative low tropo bias • Negative tropopause bias • Negative low strato bias

  9. Obs-Forecast (6h) IV Against GEM mesoscale

  10. Obs-Forecast (6h) V Against GEM mesoscale

  11. Some thoughts • Low tropo bias appears in wide class of models • Two smaller biased areas around tropopause & low strato • Bias and STD seem related with WV

  12. Dependences • O-F distribution is not symmetric: • Positively skewed at very low saturation • Negatively skewed at very large saturation • Skewness concentrated at low tropo

  13. Skewness bias • In a skewed distribution • The mean is usually not at the point of maximum probability (=mode) • The assimilation process works by maximizing the probability of the analysis field (estimates the mode)

  14. Skewness bias estimation • If we assume: • underlying distribution gaussian • WV above saturation not observable • Mean-Mode offset

  15. CMC tests on Assimilation of ROVariational impact • 2 Runs of 3DVar: Control and RO • Control = current operational • Experiment performed over 1 month (Jan 2004). • Here: typical impact of 1 cycle of 6h (with RO obs wrt without RO obs) • Notably: • consistence in T increments in independent obs in near areas. • Larger corrections in southern hemisphere • Most corrections in ocean & unpopulated areas. • Obs in populated areas agree Example: Increments of Moisture at 1000hPa and Temperature at 500hPa wrt operational 3DVar, due to occultation data.

  16. Correlation anomaly I Control: Blue RO: Red

  17. Correlation anomaly II Control: Blue RO: Red

  18. Time Series Forecast Consistency: 48h FCST vs later Analysis • Higher Short FCST consistency: • Less bias • Less STD Control: Blue RO: Red

  19. Current status at MSC • Extensive O-P analysis against 4 models • Each model presents characteristic signatures ~0.5% • Common features: • Negative O-P bias in low tropo • Negative O-P bias above tropopause (sp. tropics) • Large O-P STD in tropical tropo • Small O-P STD around tropopause • Small O-P STD in polar strato • Data impact maximization: • Identify data bias • Optimized a-priori STD (situation-dependent STD) • Assimilated N • O-P still too big to benefit from bending angle • More obvious positive impacts observed • Generalized positive impacts require finer tuning • Large analysis of O-P statistics performed • WV saturation seems to produce some artifacts. Skewness bias seems to be present.

More Related