AMV usage in the Met Office NWP model

Use of AMVs in the Met Office model

Contents

Physical characteristics

Global Model (coupled atmosphere-ocean)

  • GC4 global coupled atmosphere-land-ocean-seaice
  • GA8 atmospheric configuration
    • Grid-point model: 2560×1920 grid points
    • Horizontal resolution: Forecast N1280 (~10 km in mid-latitudes)
    • Vertical resolution: 70 vertical levels, hybrid-eta configuration, lid ~80 km
    • Forecast range: out to T+168 (7 days)
  • GL9 land configuration
  • GO6 ocean (with GSI8 sea ice) configuration at ORCA025 tri-polar grid 1/4 deg resolution

UKV Model

  • Grid-point model: 622×810 grid points (inner), 950×1025 grid (total)
  • Horizontal resolution: variable stretching from ~1.5 km (outer domain) to ~4 km (inner)
  • Vertical resolution: 70 vertical levels, lid ~40 km (different levels from global)
  • Forecasts to T+120 at 03/15 UTC, T+54 or T+12 otherwise

Data assimilation method

Global Model

  • Type: Hybrid incremental 4D-Var, Errors of the day (EOTD) information provided by global ensemble (N640 ensemble differences reconfigured to modes at N320) every 6 hours. EOTD now uses the lagging-and-shifting technique with a one lag and two shifts configuration – see Fig 12 of Lorenc (2017)
  • Resolution: N320 L70 (~40 km, PF timestep = 10 mins)
  • Analysis times (T): 00, 06, 12, 18 Z
  • Time window: 6 hrs, T ± 3 hr
  • Time constraints (model runtime):
  • Main forecast run: 20 min before time window ends
  • Update run: 3 hr 15 min after time window ends

UKV Model

  • Type: Incremental 4D-Var (except for radar rainrate)
  • Resolution: ~4.5 km fixed resolution
  • Analysis times (T): 00, 01, 02,..,21,22,23 Z
  • Time window: 1 hour (centred on nominal analysis time i.e. 0830-0930 for 09 UTC run).
  • Time constraints (model runtime):
  • T+45 mins from analysis time (e.g. 0945 for 09 UTC run)
  • T+80 mins from analysis time for 11 UTC and 23 UTC

AMV types assimilated

Global Model

  • Meteosat-10 IR, cloudy WV 7.3, VIS and HRVIS
  • Meteosat-9 IR, cloudy WV 7.3, VIS and HRVIS
  • Himawari-9 IR, VIS, cloudy WV 6.7
  • GOES-16 IR, VIS and IR3.8
  • GOES-18 IR, VIS and IR3.8
  • NESDIS and direct broadcast Terra IR
  • NESDIS and direct broadcast Aqua IR, cloudy WV, clear sky WV
  • CIMSS and direct broadcast NOAA-18 IR
  • CIMSS and direct broadcast NOAA-19 IR
  • CIMSS Metop-B IR
  • EUMETSAT Metop-B IR (single satellite product)
  • EUMETSAT Metop-C IR (single satellite product)
  • CIMSS LeoGeo IR
  • NESDIS and direct broadcast Suomi NPP VIIRS IR
  • NESDIS NOAA-20 VIIRS IR

UKV Model

  • Meteosat-10 IR produced using the NWC SAF software over the wider UKV domain

Global quality control

Blacklisting in space

  • All geostationary winds beyond 68&deg satellite zenith angle
  • All winds reporting in height at or above 160 hPa and all extratropical (polewards of 20N/S) winds above 200 hPa
  • All VIS winds at 700 hPa and above
  • MSG WV, Himawari WV 6.7, and GOES WV below 400 hPa
  • All polar WV and CSWV winds below 600 hPa
  • All MSG IR, VIS and HRVIS winds below 600 hPa over land north of 0 degrees
  • All Himawari and GOES IR and visible winds below 600 hPa over land north of 20N
  • All polar winds below 600 hPa over land and sea ice
  • All polar winds below 400 hPa over Greenland and Antarctica
  • Low and mid level winds dependent on topography
  • All VIIRS and AVHRR polar winds at or above 350 hPa in SH
  • MSG IR below 600 hPa in the Mediterranean, Red Sea and Gulf
  • MSG VIS and IR, and GOES-16 IR below 600 hPa over sea near Guinea
  • All LeoGeo above 300 hPa
  • LeoGeo winds down-weighted in thinning to only use where no conventional AMV available
  • All obs polewards of 89 degrees

QI thresholds

QI1 – EUMETSAT QI with first guess check
QI2 – EUMETSAT QI without first guess check
RFF – Recursive Filter Function (CIMSS/NESDIS)

For information on how the quality indicators are formulated see Holmlund (1998, Weather
Forecasting 13 1093-1104) and
Hayden and Purser (1995, Journal of Applied Meteorology 34 3-15).

All thresholds are applied to QI2

  • MSG: 85
  • Himawari: 70
  • GOES: 50
  • All polar winds: 60
  • LeoGeo: 70

Thinning

  • All AMVs thinned in 200 km by 200 km by 100 hPa boxes.
  • Wind selected by lowest error for all geostationary winds
  • Wind selected by closest to centre of box for all polar winds.
  • Assimilated in 2-hour time slots

Background check

  • Background check: comparison with 6-hour forecast from previous model run (symmetric)

UKV quality control

Background check is applied and QI threshold of 80 for all winds.
Only one set of 15 min winds are considered in each hourly assimilation window.

Blacklisting in space

  • All geostationary winds beyond 68 deg satellite zenith angle
  • All observations at or above 200 hPa
  • Reject WV below 950 hpa
  • Low and mid level winds dependent on topography

Thinning

  • AMVs thinned to 25 km by 100 hPa
  • Wind selected by lowest error

Observation errors

Observation errors calculated individually for each wind using estimates of the error in vector, error in height and variation in the background wind column. For more information see AMV errors: a new approach in NWP (9IWW paper).

The mean profile and zonal plots in the table below give an indication of the typical observation errors used. Statistics calculated for June 2012 with QI2 > 80.

satellite NH TR SH Zonal
Meteosat-10


Himawari-8

GOES15

MODIS


AVHRR

 

References

Lorenc, A.C. (2017), Improving ensemble covariances in hybrid variational data assimilation without increasing ensemble size. Q.J.R. Meteorol. Soc., 143: 1062-1072. https://doi.org/10.1002/qj.2990