Use Of AMSU data in the UK Mesoscale Model Brett Candy Steven English, Richard Renshaw & Bruce Macpherson Satellite Applications & Data Assimilation, Met Office ITSC-X111, Sainte Adele, Canada © Crown copyright Talk Outline Background and Motivation Limited Area Models at the Met Office Data Usage in the Mesoscale model – Source of observations – Data screening Bias Correction Impact Assessment – Method – Some results ITSC X111 Slide 2 Future Work Background Contribution of ATOVS in global NWP is very important To date effort has focused on assimilating satellite data in global NWP – Some data types are currently precluded by timeliness Initial tests of assimilating radiance data in the UK Mes encouraging – Information retained in the short-range But…..objectives are different. – Key forecast parameters cloud cover, precip and surface temp ITSC X111 Slide 3 UK Mesoscale Model 1 Model Model Domain Domain (Grid (Grid resolution=12km) resolution=12km) Background: Full Resolution AMSUB Imagery 89 GHz ITSC X111 Slide 4 UK Mesoscale Model 2 Assimilation system: – incremental 3D-Var – assimilation window ±1½ hours – 2 hour data cutoff Observations: – radiosondes, air reps, wind profilers – land station reps, including visibility – satellite winds from Meteosat Additionally cloud cover and surface rainrate information is assimilated via a different route (i.e. outside of Var) ITSC X111 Slide 5 Data Acquisition Washington West Freugh Local Passes NESDIS HRPT 1b data AAPP AAPP 1d data on HIRS grid UK Mes NWP ITSC X111 Slide 6 Comparisons for quality monitoring 1d data on HIRS grid Global NWP ATOVS Data Use HIRS data not used – calibration problems associated with partial super swath AMSU data – Remapped to HIRS grid (allows use of same code as global) – AMSUB 183 GHz channels over sea only AMSU data screening – Liquid water test in AAPP → reject channels 4,5 & 20 – Ice test on 183 GHz channels → reject channels 19,20 – Rain test in AAPP → reject channels 4-8 & 18-20 Data Thinning – 1 observation every 40 km. More weight given to clear & microwave clear scenes. ITSC X111 Slide 7 Tuning AMSU Observation Errors NOAA17 NOAA16 Global R Matrix 4.5 4 AMSUB humidity errors 3.5 RMS (o-b) K 3 Reduced to 2K 2.5 2 1.5 1 0.5 0 4 5 6 7 8 9 10 AMSU channels ITSC X111 Slide 8 11 18 19 20 Mid-lat Cyclone Case Study AVHRR IR image AMSU data screening green: lwp yellow: precip red: AMSUB cirrus ITSC X111 Slide 9 Bias Correction 1 Airmass dependent predictors (Eyre, 1992) – problem in a LAM is to sample enough representative synoptic systems – could monitor departures over a year, assuming negligible instrument drift Current solution is to use global bias correction coefficients – assumes global and LAM NWP are unbiased – monitoring with sondes confirms this, at least for the troposphere ITSC X111 Slide 10 Bias Correction 2 AMSU channels Mean O-B Difference (K) over Mesoscale Domain Uncorrected Radiances Corrected Radiances ITSC X111 Slide 11 Strategy for Assessing Impact Case study for poor operational forecast. – Convection over S.W. Britain – Rain forecasts compared to radar Set of cases containing range of weather situations observed over UK. – Chosen by forecaster – Subjective verification from station reports of 6 hour precip, surface temp & cloud cover – NOAA15 & 16 assimilated Extended Trial. – Ran for 1 month – Avoids spin-up problems – Near Real Time to get operational boundary conditions – Forecasts assessed by forecaster – NOAA16 & 17 assimilated ITSC X111 Slide 12 Convective Event 1 Observations Used Channel 19 Situation: Warm moist air moving northwards, mixing with cooler air at higher latitudes ITSC X111 Slide 13 dark colours indicate Negative o-b model too dry Channel 20 Convective Event 2 Integrated Water Vapour Analysis AMSU No AMSU region > 30kgm-2-2 T+6 rainrate forecast Radar ITSC X111 Slide 14 AMSU No AMSU Verification of Case Studies 6 cases improved, 6 cases worsened due to inclusion of AMSU Worse case highlights difficulties of using sparse verification sites for reporting precipitation Hourly Precip, 0z 26th August 2001 T+6 Verifying Radar ITSC X111 Slide 15 Control +AMSU AMSU Impact on Cloud T+6 Cloud Cover AMSU Clear low medium high ITSC X111 Slide 16 No AMSU Validation: vis image Conclusions Operational in Mesoscale model from May 2003. – NRT trial positive for cloud & visibility. – Including a significant fog clearance case. Similar approach adopted for European model. Future Work – AMSUB at full resolution. » Issues for qc & bias correction. » Extend number of channels – Assimilation in regions of significant LWP. » Total humidity control variable. » 1D Var » 3D Var ITSC X111 Slide 17 Additional Slides ITSC X111 Slide 18 Local – Global BT Difference Channel 16 HIRS Channel 15 ~0.5 K AMSU Channel 5 ITSC X111 Slide 19 Channel 6 ~0.02 K Initialisation of the Mesoscale Model: Weights given to Var & MOPs data VAR incs T-3 T-2 T-1 T+0 T+1 T+2 T+3 Cloud incs T-3 T-2 T-1 T+0 T+1 T+2 T+3 RainRate incs from hourly fields T-3 ITSC X111 Slide 20 T-2 T-1 T+0 T+1 T+2 T+3