The Global Observing System Peter Bauer and colleagues European Centre for Medium-Range Weather Forecasts

The Global Observing System Peter Bauer and colleagues European Centre for Medium-Range Weather Forecasts

NWP, conventional and satellite observations General impact assessment of current observing system Data monitoring Future observations and observation usage Special Applications: Climate & Chemistry Concluding remarks

Atmospheric model Wave model ECMWF forecasting systems Seasonal Forecasts Medium-Range Forecasts (Deterministic and EPS) Monthly Forecasts Atmospheric model Wave model Ocean model Real Time Ocean Analysis ~8 hours Delayed Ocean Analysis ~12 days

Data assimilation system (4D-Var) • The observations are used to correct errors in the short forecast from the previous analysis time. • Every 12 hours we assimilate 4 – 8,000,000 observations to correct the 100,000,000 variables that define the model’s virtual atmosphere. • This is done by a careful 4-dimensional interpolation in space and time of the available observations; this operation takes as much computer power as the 10-day forecast.

Satellite observing system Data types: Data volume:

Example of conventional data coverage

Example of 6-hourly satellite data coverage LEO Sounders LEO Imagers Scatterometers GEO imagers GPS Radio Occultation Satellite Winds (AMVs) 9 April 2010 00 UTC

What types of satellites are used in NWP? Advantages Disadvantages GEO - large regional coverage - no global coverage by single satellite - very high temporal resolution - moderate spatial resolution (VIS/IR) > short-range forecasting/nowcasting > 5-10 km for VIS/IR > feature-tracking (motion vectors) > much worse for MW > tracking of diurnal cycle (convection) LEO- global coverage with single satellite - low temporal resolution - high spatial resolution >best for NWP!

Observation numbers per cycle EXP-HI EXP EXP-SV EXP-CLI EXP-RND Average radiance data count per analysis from period 08/12/2008-28/02/2009:

Data Assimilation – Incremental 4D-Var T799L91 T95L91 T159L91 T255L91 T799L91 (Trémolet 2004)

Data Assimilation – Radiances Transfer of information between radiances and control variables Control Variable / state vector Forecast model State at time i Radiative transfer Radiance observations Wind and mass, humidity Clear, cloud and rain Dynamics, moist physics Wind and mass, humidity, Clear, cloud and rain including scattering Clear sky Clear sky clouds and rain

What is the observation operator? Example 1: Radiosonde profile of T H = spatial interpolation Example 2: Clear-sky radiance observation H = spatial interpolation + clear-sky radiative transfer Example 3: Cloud/rain radiance observation H = spatial interpolation + moist physical parameterizations + multiple scattering radiative transfer MVIRI Model SSM/I Model

Combined impact of all satellite data • EUCOS Observing System Experiments (OSEs): • 2007 ECMWF forecasting system, • winter & summer season, • different baseline systems: • no satellite data (NOSAT), • NOSAT + AMVs, • NOSAT + 1 AMSU-A, • general impact of satellites, • impact of individual systems, • all conventional observations. •  500 hPa geopotential height anomaly correlation 3/4 day 3 days

Impact of microwave sounder data in NWP: Met Office OSEs • 2003 OSEs: • N-15,-16 and -17 AMSU • N-16 & N-17 HIRS • AMVs • Scatterometer winds • SSM/I ocean surface wind speed • Conventional observations • 2007 OSEs: • N-16, N-18, MetOp-2 AMSU • SSMIS • AIRS & IASI • Scatterometer winds • AMVs • SSM/I ocean surface wind speed • Conventional observations (W. Bell)

Sensitivity of analysis increments to observations • 2007 GMAO/GSI system, 1.875o, 64 levels, 6-hour window; • J from analysis increments; August 2004. temperature zonal wind North-Pacific North Pacific temperature zonal wind US US satellite conventional total (Zhu & Gelaro 2008)

State at analysis time Forecast sensitivity: max. 48 hours State at time i State at initial time NWP model Sensitivity of cost to observations Sensitivity of cost to change at initial time AD of forecast model Cost function J Analysis Advanced diagnostics Data assimilation: max. 12 hours State at time i Observation operator Observation simulations State at initial time NWP model Sensitivity of cost to change in state at time i Sensitivity of cost to change at initial time AD of forecast model AD of observation operator Cost function J Observations

Advanced diagnostics Relative FC error reduction per system The forecast sensitivity (Cardinali, 2009, QJRMS, 135, 239-250) denotes the sensitivity of a forecast error metric (dry energy norm at 24 or 48-hour range) to the observations. The forecast sensitivity is determined by the sensitivity of the forecast error to the initial state, the innovation vector, and the Kalman gain. Relative FC error reduction per observation (C. Cardinali)

Advanced diagnostics – MW sounder denial 3 AMSU-A, 2 MHS vs 1 AMSU-A, 0 MHS Forecast error reduction [%] (C. Cardinali)

Advanced diagnostics – MW imager denial No MW-imagers Control Forecast error reduction [%] (C. Cardinali)

Data monitoring – time series Time evolution of statistics over predefined areas/surfaces/flags (M. Dahoui)

Data monitoring – overview plots Time evolution of statistics for several channels Useful for quick and routine verifications Can not be used for high spectral resolution sounders RTTOV version upgrade (M. Dahoui)

Email-alert Email alert: Soft limits (mean ± 5 stdev being checked, calculated from past statistics over a period of 20 days, ending 2 days earlier) Hard limits (fixed) Data monitoring – automated warnings http://www.ecmwf.int/products/forecasts/satellite_check/ Selected statistics are checked against an expected range. E.g., global mean bias correction for GOES-12 (in blue): (M. Dahoui & N. Bormann)

Data monitoring – automated warnings (M. Dahoui & N. Bormann)

Satellite data monitoring Data monitoring – automated warnings (M. Dahoui & N. Bormann)

New data availabilities • 2010: • Oceansat-2 (Scatterometer: surface wind vector) • DMSP F-18 SSMIS (MW T:, q-sounding, clouds and precipitation) • SMOS (MW: soil moisture) • Megha Tropiques MADRAS/SAPHIR (MW: q-sounding, clouds and precipitation) • FY-3A IRAS/MWTS/MWHS/MWRI (IR/MW: T, q-sounding, clouds and precipitation) • GOSAT FTS (Advanced IR: T, q, trace gas sounding) • 2011: • NPP (Advanced IR: T, q-sounding) • ADM (Doppler-lidar: Atmospheric wind vector) • 2012 and beyond: • More advanced IR sounders in polar (Metop, NPOESS) and geostationary orbits (MTG, GOES) for general sounding • More active instruments (wind, clouds, precipitation)

Cloudsat/CALIPSO data monitoring (J.-J. Morcrette)

ECMWF usage of SMOS data • Global monitoring: • Development of model forward operator (emissivity model) • Data pre-processing (HDF2BUFR → ODB/IFS) • Implementation of passive monitoring system, diagnostics, quality control • Data assimilation study: • Impact of SMOS constrained soil moisture on medium-range forecasts H-pol H-pol V-pol H-pol 22 January 2010 00 UTC; 1st background departure monitoring (no q/c)

Soil moisture from ASCAT data FG departure in m3/m3 (January 2010) FG departure bias vs ASCAT incidence angle Histograms of FG departures (P. de Rosnay)

Active instruments: ESA’s ADM ESA ADM AEOLUS Doppler Lidar for wind vector observation Pressure (hPa) Control+ADM Control Control-sondes ECMWF is responsible for the development of the level 2 processor and will exploit the data as soon as available. Simulated DWL data adds value at all altitudes and well into longer-range forecasts. Zonal wind forecast error (m/s)

Areas of instability: Eady index • Eady-index as a proxy for baroclinic instability in the atmosphere • difference between seasons is rather strong; • year-to-year variability has significant seasonal dependence as well.

01-07/01/2009 Average SV RND CLI Data coverage 14/12/2008 00 UTC data density AMSU-A channel 9 EXP-HI: EXP: EXP-SV: EXP-CLI: EXP-RND:

Forecast impact: z500 – D08JF09 JAS08 D08JF09

1957 2002 METEOSAT reprocessed cloud motion winds TOMS/ SBUV 1982 1988 Aircraft data 1979 1973 Conventional surface and upper-air observations NCAR/NCEP, ECMWF, JMA, US Navy, Twerle, GATE, FGGE, TOGA, TAO, COADS, … 1973 1979 VTPR 1987 HIRS/ MSU/ SSU Cloud motion winds Buoy data 1991 SSM/I 1995 ERS-1 ERS-2 1998 AMSU ECMWF Reanalysis Observations used in ERA-Interim: • ERA-Interim is current ECMWF reanalysis project following ERA-15 & 40. • 2006 model cycle, 4D-Var, variational bias-correction, more data (rain assimilation, GPSRO); 1989-1998 period available, 1998-2005 period finished, real-time in 2009. 1989 The ERA-40 observing system: • ERA-40 observations until August 2002 • ECMWF operational data after August 2002 • Reprocessed altimeter wave-height data from ERS • Humidity information from SSM/I rain-affected radiance data • Reprocessed METEOSAT AMV wind data • Reprocessed ozone profiles from GOME • Reprocessed GPSRO data from CHAMP ERA-Interim

Reanalysis as inter-calibration tool Global mean bias corrections produced in ERA-Interim (MSU Channel 2): Recorded warm-target temperatures, NOAA-14: (Grody et al. 2004) • Variations in warm target • are due to orbital drift • VarBC is able to correct • the resulting calibration • errors (D. Dee)

Combining NWP with CTM models and data assimilation systems EC FP-6/7 projects GEMS/MACC (coordinated by ECMWF) towards GMES Atmospheric Service

Concluding remarks • At ECMWF, 95% of the actively assimilated data originates from satellites (90% is assimilated as radiances and only 5% as derived products and 5% from conventional products). • Impact experiments demonstrate the crucial role of conventional observations! • Ingredients for successful data implementation: - early data access after launch: (1) fast monitoring of data quality – feedback to space agencies, (2) early testing of data impact in NWP data assimilation systems. - near real-time data access to maximize operational use. optimal return of investment by global user community (example: METOP). • Currently most important NWP instruments at ECMWF:- advanced infrared sounders (temperature, moisture),- microwave sounders and imagers (temperature, moisture, clouds, precipitation),- GPS transmitters/receivers (temperature),- IR imagers/sounders in geostationary orbits (moisture, clouds, wind),- scatterometers (near surface wind speed, wave height), altimeters (height anomaly),- UV/VIS/IR spectrometers (trace gases, temperature).

Concluding remarks • Future challenges with respect to observations:- active instruments – radar, lidar (wind, aerosols, clouds, precipitation, water vapour),- advanced imagers – synthetic aperture radiometers (soil moisture). • Future challenges with respect to data assimilation:- model resolution upgrades also affect data assimilation resolution,- more intelligent data thinning using ensemble methods (B) and forecast error growth metrics,- assimilation of cloud/precipitation-affected data will require revised control variable, background error statistics. • Future upgrades to data monitoring:- more sophisticated data co-location tools to compare performance between data from different sensors,- more advanced automated warning system.

The Global Observing System Peter Bauer and colleagues European Centre for Medium-Range Weather Forecasts