Use of Advanced Infrared Sounder Data in NWP

Use of Advanced Infrared Sounder Data in NWP models Roger Saunders (Met Office, U. K. ) n Why use advanced sounder data? n How do we use advanced sounders in NWP? n Ø Radiative transfer Ø Cloud detection Ø Bias tuning Ø Real time data monitoring Ø Data assimilation n Impacts of sounder data Workshop for Soundings from High Spectral Resolution Observations Demonstrate with AIRS data

Why Do We Need Satellite Data for NWP ? n Global coverage - main source of data over oceans and remote land areas. n Measurements closer to scale of models grids. n Has greater impact than radiosonde data on N. Hemisphere forecasts. n Model validation (using data not assimilated) used for assessing impact of changes made and errors of the model analyses/forecasts. Workshop for Soundings from High Spectral Resolution Observations

Global Coverage Plot: radiosondes Workshop for Soundings from High Spectral Resolution Observations

Global Coverage Plot: Aircraft Workshop for Soundings from High Spectral Resolution Observations

Global Coverage: Polar Satellite NOAA-15 NOAA-17 Workshop for Soundings from High Spectral Resolution Observations NOAA-16

IASI vs HIRS IASI channels HIRS channel Spectrum of infrared radiation from atmosphere HIRS 19 channels vs IASI 8461 channels Workshop for Soundings from High Spectral Resolution Observations

Expected Retrieval Performance IASI (METOP) HIRS(NOAA)) Workshop for Soundings from High Spectral Resolution Observations

Resolving Atmospheric Features Workshop for Soundings from High Spectral Resolution Observations

Met Office NWP Models Model formulation: Exact equations of motion in 3 D, non-hydrostatic effects included, semi-Langrangian scheme, hybrid-eta in height. Data Assimilation: 3 DVar, FGAT, 6 hourly cycle 3 hr cut-off with update runs for next cycle Provides model background from 6 hour forecast Workshop for Soundings from High Spectral Resolution Observations

Observations Required for NWP Workshop for Soundings from High Spectral Resolution Observations

Polar Satellites for NWP Workshop for Soundings from High Spectral Resolution Observations

IR Advanced sounders for NWP Name AIRS IASI Cr. IS GIFTS Instrument Grating FTS FTS Spectral range (cm‑ 1) 649 – 1135 Contiguous 650 – 1095 685 -1130 1217– 1613 645‑ 2760 1210 – 1750 1650 -2250 2169 – 2674 2155 – 2550 Unapodized spectral resolving power 1000 – 1400 2000 – 4000 900 – 1800 2000 Field of view (km) 13 x 7 12 14 4 Sampling density per 50 km square 9 4 9 144 Platform Aqua METOP NPOESS GIFTS Launch date May 2002 2005 (NPP) 2008 Workshop for Soundings from High Spectral Resolution Observations

How do we use advanced sounders in NWP? 1. Radiative transfer 2. Cloud detection 3. Bias tuning 4. Real time data monitoring 5. Data assimilation Workshop for Soundings from High Spectral Resolution Observations

Fast radiative transfer model used n RTTOV-7 developed by EUMETSAT NWP SAF – Line database: HITRAN-96 – Lb. L model GENLN 2 at 0. 001 cm-1 – Water vapour continuum: CKD 2. 1 – 43 L fixed pressure level parametrisation – T, q, O 3 and surface from NWP model – Masuda for sea surface emissivity, 0. 98 for land – Jacobians also computed essential for radiance assimilation Workshop for Soundings from High Spectral Resolution Observations

S. Dev Br. Temp difference (K) RT model validation Workshop for Soundings from High Spectral Resolution Observations

Gastropod k. CARTA RTTOV-7 model validation for AIRS Ozone jacobian Response to 10% change in ozone deg. K Workshop for Soundings from High Spectral Resolution Observations

Effect of bad Jacobians Workshop for Soundings from High Spectral Resolution Observations

Effect of error correlation Workshop for Soundings from High Spectral Resolution Observations

To Prepare for Advanced IR sounders Aqua Has Been Launched! Workshop for Soundings from High Spectral Resolution Observations n Aqua was launched from Vandenburg AFB, USA at 10. 55 am BST on 4 th may 2002. n It carries the AIRS spectrometer. The Met Office started to receive AIRS data in October 2002 to enable us to assimilate these data in NWP models.

How do we use advanced sounders in NWP? 1. Radiative transfer 2. Cloud detection 3. Bias tuning 4. Real time data monitoring 5. Data assimilation Workshop for Soundings from High Spectral Resolution Observations

Cloud detection n Currently can only simulate accurately clear sky IR radiances as representation of clouds in NWP models and their radiative properties requires improvement. n Therefore must identify those sounder fields of view which have significant cloud within them and screen them out. n Several techniques developed to do this: – Inter-channel tests + SST check – Local spatial variance – Variational O-B checks – PCA Workshop for Soundings from High Spectral Resolution Observations cloud cost

Var cloud cost (English et al. , 1999) Principal Component Analysis (PCA) of the cloud cost The i-th partial cloud cost Var scheme uses simple summation of all partial cloud cost The i-th PCA components of S depends on profile by profile, then. . . Workshop for Soundings from High Spectral Resolution Observations

PCA of simulated O-B difference S is constructed from clear O-B statistics Principal Component Analysis (PCA) of the cloud cost The i-th PCA components of Workshop for Soundings from High Spectral Resolution Observations for each profile

AIRS channel selection for cloud detection - LW-IR, SW-IR, AMSU-A ch. 3, 15 are used for cloud detection 1) SOUND 02 AIRS ch. 261 13. 80 micron, ch. 453 12. 61 micron, ch. 672 11. 48 micron, ch. 787 10. 90 micron, ch. 843 10. 66 micron, ch. 914 10. 35 micron, ch. 1221 8. 96 micron, ch. 1237 8. 90 micron AMSU-A ch. 3 50. 3 GHz, ch. 15 89. 0 GHz 2) MIX 02 SOUND 02 + AIRS ch. 2328 3. 83 micron, ch. 2333 3. 82 micron Workshop for Soundings from High Spectral Resolution Observations

PCA components of O-B difference - Cloudy with much Ice Water IR PCA 02 (-MW-IR) + Cloudy with much Liquid Water - PCA 01 (+MW-IR) M W + Workshop for Soundings from High Spectral Resolution Observations Clear

PCA components of O-B difference PCA 04 (+LWIR-SWIR) Cloudy with much Liquid Water Cloudy with much Ice Water PCA 01 (+MW-IR) Clear Cloud distribution in O-B (and its PCA) space is inhomogeneous and asymmetric!! Workshop for Soundings from High Spectral Resolution Observations

PCA 12 (+ch. 2328 -ch. 2333) PCA components of O-B difference PCA 11 (+ch. 914 -ch. 843) Higher components are no use for cloud detection !! Workshop for Soundings from High Spectral Resolution Observations

Cloud detection: Validation Blue Jc<2 Green Jc>2 Red Jc>20 PCA scheme Workshop for Soundings from High Spectral Resolution Observations VAR scheme Blue Jc<. 94 Green Jc>. 94 Red Jc>20

How do we use advanced sounders in NWP? 1. Radiative transfer 2. Cloud detection 3. Bias tuning 4. Real time data monitoring 5. Data assimilation Workshop for Soundings from High Spectral Resolution Observations

Bias tuning for AIRS To remove biases, predictors from NWP model fields and/or instrument parameters are used. Predictors for AIRS being used are: n Scan angle n Model Tskin n Model Thickness 850 -300 h. Pa n Model Thickness 200 -50 h. Pa n Simulated brightness temperature Workshop for Soundings from High Spectral Resolution Observations

Example of AIRS bias tuning Corrected biases Uncorrected biases AIRS channel 227 Peaks at 700 h. PA Workshop for Soundings from High Spectral Resolution Observations

Example of AIRS bias tuning Corrected biases Uncorrected biases AIRS channel 1574 Upper trop wv Workshop for Soundings from High Spectral Resolution Observations

How do we use advanced sounders in NWP? 1. Radiative transfer 2. Cloud detection 3. Bias tuning 4. Real time data monitoring 5. Data assimilation Workshop for Soundings from High Spectral Resolution Observations

NWP Radiance Monitoring Observed minus Simulated n Continuous global view of data n Good for spotting sudden changes in instruments n Can compare with other satellites and in situ obs But NWP model has errors: (LST, water vapour, ozone, clouds, stratosphere) so bias correction and cloud detection important and care in interpretation Workshop for Soundings from High Spectral Resolution Observations

Monitoring web page Available to the AIRS team in mid-December via password protected page on Met Office site. http: //www. metoffice. com/research/ nwp/satellite/infrared/sounders/airs /index. html Userid: airspage Passwd: &Graces Workshop for Soundings from High Spectral Resolution Observations

Time series of observations Rejects caused by Channel 2357 Workshop for Soundings from High Spectral Resolution Observations

Ozone 20 -70 N O-B st. dev plots Stratosphere 20 N-20 S Water vapour Workshop for Soundings from High Spectral Resolution Observations 20 -70 S

Tartan Plots: O-B clear mean bias Global Observations AMSU 1&2 CO 2 H 2 O O 3 CO 2 Workshop for Soundings from High Spectral Resolution Observations

Day & Night spectra Workshop for Soundings from High Spectral Resolution Observations

O-B difference - Large positive bias in the SW-IR in the day-time due to Non LTE effect in upper sounding chs and sunglint in window 2387 cm-1 (4. 19 micron) SW-IR chans difficult to use for cloud detection Workshop for Soundings from High Spectral Resolution Observations 2392 cm-1 (4. 18 micron) 2618 cm-1 (3. 82 micron)

How do we use advanced sounders in NWP? 1. Radiative transfer 2. Cloud detection 3. Bias tuning 4. Real time data monitoring 5. Data assimilation Workshop for Soundings from High Spectral Resolution Observations

Assimilation of satellite data Observation space Forecast observations B (Forward modelled TBs) H(X) (Radiative transfer model + map to observation space) Workshop for Soundings from High Spectral Resolution Observations O-B Ob increments Quality control Adjoint of observation operator H Observation operator H Background NWP forecast X Observations O or yo Bias tuning Map back to model space Adjusted model fields NWP model space Profile increments Analysis X Forecast

Data assimilation For variational assimilation we want to minimise a cost function J: J(X) = 0. 5(yo - H(X)) (O+F)-1 (yo - H(X))T + 0. 5(X-Xb) B-1 (X-Xb)T To minimise equation above, assuming the observations yo to be. RT linearly related to X then Hois observation operator (e. g. fast model) y arevalue observations (e. g. radiances) b is background the minimum for J(X) is when: X is atmospheric state vector, X (O+F)-1 is observation + forward model error B is background error covariance X = Xb + BHT. (H. B. HT + O+F)-1. (yo - H(Xb)) H is derivative of H wrt X often called jacobian matrix Workshop for Soundings from High Spectral Resolution Observations

Observation errors: AIRS channel covariance AIRS Net Meeting, 24 th February 2003

Forward model error correlation matrix for RTIASI Workshop for Soundings from High Spectral Resolution Observations

Background errors and correlation matrix for T and q Workshop for Soundings from High Spectral Resolution Observations

Options for Data Assimilation (1 - type of observation) n Assimilation of retrievals – T(p), q(p), O 3(p) Lowest cost but inconsistent FG and no control of retrieval process. Must also have retrieval error covariance. n Assimilation of 1 DVar retrievals – T(p), q(p), O 3(p) More optimal but radiances used in isolation n Direct radiance assimilation in 3 DVar or 4 DVar – Radiances for limited number of channels Most expensive but most optimal and is current operational use of ATOVS but only limited use of data n Use combination of channels – pseudo channels or EOFs Possible for ‘day-2’ assimilation, needs more research Workshop for Soundings from High Spectral Resolution Observations

n Initially: Options for Data Assimilation (2 - coverage) – clear sky, tropospheric radiances over sea – stratospheric radiances globally n Medium Term: – cloudy radiances over uniform low cloud – more radiances over land sea-ice n Longer term: – include cloud fully in state vector and provide cloud variables back to model Workshop for Soundings from High Spectral Resolution Observations

Measuring Impact of Satellite Data on Forecasts We can run experiments where satellite data are not used and observe the consequent degradation in the forecast skill relative to a system which has used all data (observing system experiment or OSE). But. Can only do this using today’s satellite data + processing and current data assimilation + forecast model systems. N. B. In 1995 OSE’s suggested satellite data had a negative impact on forecast skill. Workshop for Soundings from High Spectral Resolution Observations

Satellite Vs Conventional: NH Height Forecast skill Sat data largest impact ~10 hr gain at 5. 5 days Workshop for Soundings from High Spectral Resolution Observations

Conventional Vs Satellite: SH Height Forecast skill Sat data largest impact ~48 hr gain at 5 days Workshop for Soundings from High Spectral Resolution Observations

1999 -2002 NOAA-16 +1. 5 1990 -1999 Tuning of 3 D-Var +1. 5 Direct radiance assimilation +2 ATOVS sea ice +2 ATOVS over Siberia +0. 75 ATOVS & 3 D-Var +4 1996 1997 Workshop for Soundings from High Spectral Resolution Observations 2 nd DMSP sat +0. 75 1998 1999 2000 2001

Advanced sounder data volumes Estimates Actual values Workshop for Soundings from High Spectral Resolution Observations

Summary of Current Status(1) n The ATOVS sounder data has led to a significant improvement in forecast skill over the last 4 years n Satellite data now have a larger impact than radiosondes in N. Hemisphere. n New variational data assimilation techniques allowing direct use of satellite radiances has contributed enabling better use to be made of satellite data. n Advanced IR sounder data show promise to improve the temperature, water vapour and ozone fields in the model. Workshop for Soundings from High Spectral Resolution Observations

Summary of Current Status(2) n Cloud and surface parameters should also be updated by using these data. n AIRS is providing an excellent test bed for use of advanced IR sounder data. n Data volumes will remain a challenge for NWP centres (e. g. only 324 channels used from AIRS). n More research on using compressed forms of data and cloud affected data. Workshop for Soundings from High Spectral Resolution Observations

Workshop for Soundings from High Spectral Resolution Observations