Characterising and reducing uncertainties in all-sky radiative transfer - Vasileios Barlakas Annual Fellow Day 01.03.2021
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Characterising and reducing uncertainties in all-sky radiative transfer Vasileios Barlakas Annual Fellow Day 01.03.2021
Preface Project description ! CREATE – Characterizing and REducing uncertainties in All-sky microwave radiative TransfEr ⚈ Sensors ⚈ Applications o Sensors with frequencies up to ~183 GHz o Stand–alone retrievals o Ice Cloud Imager (ICI): 183.31 – 664 GHz o Data assimilation (DA) ⚈ Open issues o Three-dimensional radiative transfer is ingored o Hydrometeor orientation and polarization are ignored o RTTOV-SCATT accuracy o Two-stream delta-Eddington approximation o Sub-grid variability and cloud overlap scheme
Preface Project description ! CREATE – Characterizing and REducing uncertainties in All-sky microwave radiative TransfEr ⚈ Sensors ⚈ Applications o Sensors with frequencies up to ~183 GHz o Stand–alone retrievals o Ice Cloud Imager (ICI): 183.31 – 664 GHz o Data assimilation (DA) ⚈ Open issues þ Three-dimensional radiative transfer is ingored 1st year1 þ Hydrometeor orientation and polarization are ignored 1st & 2nd year2 o RTTOV-SCATT accuracy 1st & 2nd year2 o Two-stream delta-Eddington approximation o Sub-grid variability and cloud overlap scheme 1Barlakas and Eriksson., Remote Sens., 2020 2Barlakas et al., AMTD, 2020 2Geer, Barlakas et al., to be submitted to GMD, 2021 2Barlakas et al., in preparation, JQSRT, 2021
Overview Part 1 – Hydrometeor orientation !!!!!!!!!!! Introducing hydrometeor orientation into all-sky millimeter/sub-millimeter assimilation Barlakas et al., AMTD, 2020 Geer, Barlakas et al., to be submitted to GMD, 2021
Introduction Motivation !!!!!!!!!!! ⚈ Why ice clouds? Nasa/JPL-Caltech IWP = 2 gm−2 o Model uncertainties (shape and orientation). ⚈ Why MW and sub-mm? o Sensitive to large and small ice hydrometeors o Ice Cloud Imager (ICI) o 183–664 GHz + dual polarization o Improved ice retrievals + extend data assimilation
Introduction Motivation !!!!!!!!!!! ⚈ Why ice clouds? Nasa/JPL-Caltech IWP = 2 gm−2 o Model uncertainties (shape and orientation). ⚈ Why MW and sub-mm? o Sensitive to large and small ice hydrometeors o Ice Cloud Imager (ICI) o 183–664 GHz + dual polarization o Improved ice retrievals + extend data assimilation + Polarization 6 July 2019
Introduction Hydrometeor orientation !!!!!!!!!!! ⚈ Assumptions in DA o Totally randomly oriented (TRO) hydrometeors only o “Scalar” simulations, i.e., V– or H–polarization o Invariant scattering properties between V– and H–simulations Oriented ice hydrometeors induce polarization PD = TBV − TBH
Introduction Hydrometeor orientation !!!!!!!!!!! o Attenuation driven by the extinction matrix ⎛ K 11 K 12 K 13 K 14 ⎞ ⎜ ⎟ ⎜ K 21 K 22 K 23 K 24 ⎟ ✓ incident direction K= ⎜ ⎟ K 31 K 32 K 33 K 34 ⎟ ✓ orientation ⎜ ⎜ ⎝ K 41 K 42 K 43 K 44 ⎟⎠
Introduction Hydrometeor orientation !!!!!!!!!!! o Attenuation driven by the extinction matrix ⎛ K 11 K 12 K 13 K 14 ⎞ ⎜ ⎟ ⎜ K 21 K 22 K 23 K 24 ⎟ ✓ incident direction K= ⎜ ⎟ K 31 K 32 K 33 K 34 ⎟ ✓ orientation ⎜ ⎜ ⎝ K 41 K 42 K 43 K 44 ⎟⎠ o Total random orientation (TRO) ✗ favored direction ✗ orientation ⎛ K 11 0 0 0 ⎞ ⎜ ⎟ ⎜ 0 K 11 0 0 ⎟ K= ⎜ ⎟ ⎜ 0 0 K 11 0 ⎟ ⎜ 0 0 0 K 11 ⎟⎠ ⎝ K11, extinction cross section
Introduction Hydrometeor orientation !!!!!!!!!!! o Attenuation driven by the extinction matrix ⎛ K 11 K 12 K 13 K 14 ⎞ ⎜ ⎟ ⎜ K 21 K 22 K 23 K 24 ⎟ ✓ incident direction K= ⎜ ⎟ K 31 K 32 K 33 K 34 ⎟ ✓ orientation ⎜ ⎜ ⎝ K 41 K 42 K 43 K 44 ⎟⎠ o Total random orientation (TRO) o Azimuthal random orientation (ARO) ✗ favored direction ✓ orientation based on tilt angle β ✗ orientation ✗ orientation in the azimuth ⎛ K 11 0 0 0 ⎞ ⎛ K K 12 0 0 ⎞ ⎜ ⎟ ⎜ 11 ⎟ ⎜ 0 K 11 0 0 ⎟ ⎜ K 12 K 11 0 0 ⎟ K= ⎜ ⎟ K= ⎜ ⎟ ⎜ 0 0 K 11 0 ⎟ ⎜ 0 0 K 11 K 34 ⎟ ⎜ 0 0 0 K 11 ⎟⎠ ⎜ 0 0 −K 34 K 11 ⎟ ⎝ ⎝ ⎠ K11, extinction cross section K12, cross section for linear polarisation K34, cross section for circular polarisation
Polarized scattering correction Extinction at 166.9 GHz !!!!!!!!!!! ⚈ TRO vs ARO Large plate aggregate Brath et al., 2020 Barlakas et al., AMTD, 2020
Polarized scattering correction Extinction at 166.9 GHz !!!!!!!!!!! ⚈ TRO vs ARO Large plate aggregate Brath et al., 2020 Conical scanners view o Extinction cross section, K11 => At θinc ~ 55o, the differences are close to 0. o Linear polarization cross section, K12 => Large differences up to 18 % between V & H Barlakas et al., AMTD, 2020
Polarized scattering correction ARO approximation !!!!!!!!!!! ⚈ Orientation approximation o The parameterization is governed by the polarization ratio ρ: Large plate aggregate Brath et al., 2020 τ H τ TRO (1+ α ) ρ= = , τ V τ TRO (1− α ) α is a correction factor approximating the cross section for linear polarization, i.e., K12 * Control (current framework): ρ (α = 0) = 1 Barlakas et al., AMTD, 2020 Geer, Barlakas et al., to be submitted to GMD, 2021
Data assimilation experiments Setup !!!!!!!!!!! ⚈ Based on the IFS1 of ECMWF2 1) Passive monitoring experiments o Sensors: GMI3 Which is the best ρ? o Duration: Jun. – Jul. 2019 2) Cycled DA experiments o Sensors: AMSR-24, GMI3, SSMIS5 o Duration: Febr. – May 2019 & What is impact of ρ on the forecast? Jun. – Aug. 2019 o Microphysical setup following: Geer, to be submitted to AMT, 2021 1IntegratedForecast System 2European Centre for Medium-Range Weather Forecasts 3Global Precipitation Mission microwave imager 4Advanced Microwave Scanning Radiometer-2 Barlakas et al., AMTD, 2020 5Special Sensor Microwave Imager/Sounder Geer, Barlakas et al., to be submitted to GMD, 2021
Fitting model to observations Choosing the best ρ !!!!!!!!!!! ⚈ Polarization differences (PD) o Observed (o): PDo = TBV o o − TBH PDo − PD b ! o Simulated (b): b PD = T b BV −Tb BH * Careful cloud screening method Barlakas et al., AMTD, 2020
Fitting model to observations Choosing the best ρ !!!!!!!!!!! ⚈ Polarization differences (PD) o Observed (o): PDo = TBV o o − TBH PDo − PD b ! o Simulated (b): b PD = T b BV −T b BH * Careful cloud screening method #simulated h = ∑ log10 / # bins Control bins #observed o Best fit: 166.5 GHz, ρ = 1.4 (α = 0.167) 89.0 GHz, ρ = 1.5 (α = 0.2) Barlakas et al., AMTD, 2020
Fitting model to observations Global performance of ρ = 1.4 !!!!!!!!!!! o Universal arch-like shape of relation o Control run completely fails to reproduce it o A ρ of 1.4 gives a reasonable representation of such relation Barlakas et al., AMTD, 2020
Fitting model to observations Global performance of ρ = 1.4 !!!!!!!!!!! Observations Simulations ρ = 1.4 o Universal arch-like shape of relation o Control run completely fails to reproduce it o A ρ of 1.4 gives a reasonable representation of such relation Barlakas et al., AMTD, 2020
Forecast impact Forecast quality – ρ = 1.4 !!!!!!!!!!! Equal & opposite Unilateral in V α = 0.167 γ = 0.286 τ τ (1+ α ) τ TRO (1+ β ) τ (1) ρ = H = TRO = = TRO = 1.4 τ V τ TRO (1− α ) τ TRO (1) τ TRO (1− γ ) Unilateral in H β = 0.4 1Advanced Technology Microwave Sounder 2Winds at 850 hPa, AMSUA (53.7H GHz), geostationary H2O radiances Barlakas et al., AMTD, 2020
Forecast impact Forecast quality – ρ = 1.4 !!!!!!!!!!! Equal & opposite Unilateral in V α = 0.167 γ = 0.286 τ τ (1+ α ) τ TRO (1+ β ) τ (1) ρ = H = TRO = = TRO = 1.4 τ V τ TRO (1− α ) τ TRO (1) τ TRO (1− γ ) Unilateral in H β = 0.4 ⚈ Validation o Independent: ATMS1 & conventional2 data þ Neutral impact 1Advanced Technology Microwave Sounder 2Winds at 850 hPa, AMSUA (53.7H GHz), geostationary H2O radiances Barlakas et al., AMTD, 2020
Forecast impact Forecast quality – ρ = 1.4 !!!!!!!!!!! Equal & opposite Unilateral in V Instrument(s):Instrument(s): GPM − GMI(ALL−SKY) − TB DMSP−17,18 Area(s): N.Hemis − SSMIS(ALL−SKY) S.Hemis Tropics − TB α = 0.167 γ = 0.286 (a) From 00Z 15−Feb−2019 Area(s): to 12Z 31−Aug−2019 N.Hemis S.Hemis Tropics From 00Z 15−Feb−2019 to 12Z 31−Aug−2019 13 190.3V τ H τ TRO (1+ α ) τ TRO (1+ β ) τ TRO (1) (b) 17 ρ= = = = = 1.4 12 186.9V τ V τ TRO (1− α ) τ TRO (1) τ TRO (1− γ ) 16 11 Frequency (pol.) [GHz] 166.5H 14 Unilateral in H number 10 166.5V number β = 0.4 13 8 89.0V Channel 12 Channel ⚈ Validation 6 11 5 10 36.5V 23.8V o Independent: ATMS1 & conventional2 data 49 18.7V GMI þ Neutral impact 38 18.7H 90 96 92 98 94 10096 98102 100 104102 104 106 o Dependent: AMSR-2, GMI, SSMIS Background Background std. std. dev. dev. [%, [%, normalised] normalised] Pol. Pol. ratio ratio of of 1.4 1.4 applied applied to to V þ Increased consistency between V & H channels (b) Pol. Pol. ratio ratio of of 1.4 1.4 applied applied to V to V V and and H H Pol. ratio of 1.4 applied to Pol. ratio of 1.4 applied to HH þ Increased consistency between instruments 100% 100% = = Control Control + Tuning the overall level of scattering 1Advanced Technology Microwave Sounder 2Winds at 850 hPa, AMSUA (53.7H GHz), geostationary H2O radiances Barlakas et al., AMTD, 2020
Conclusions Summary & Outlook !!!!!!!!!!! ⚈ Final configuration of RTTOV-SCATT v13.0 o Minor software adaption and no calculation burdens o Maximum modeling errors are reduced by 10–15 K o Errors are now ~ symmetrical o Neutral forecast impact, with increasing consistency between instruments o Polarization ratio of 1.4 in RTTOV-SCATT v13.0 + future IFS cycle ⚈ Outlook MHS ICI o Polarization correction scheme o Cross-track sounders o Radar backscattering 3rd year Photo: esa.int Photo: Airbus o Ice retrievals, e.g., in GMI o Frequencies above 183 GHz, i.e., Ice Cloud Imager Barlakas et al., AMTD, 2020 Geer, Barlakas et al., to be submitted to GMD, 2021
Overview Part 2 – RTTOV-SCATT accuracy !!!! Model inter-comparison in all-sky millimeter/sub-millimeter radiative transfer Barlakas et al., in preparation, JQSRT, 2021 Objectives: o Are there any systematic or random errors? o Where does the two-stream delta-Eddington “falls apart”? o Most studies focus on clear-sky or frequencies below ~90 GHz o Prepare RTTOV–SCATT for ICI
Overview Part 2 – RTTOV-SCATT accuracy !!!! Model inter-comparison in all-sky millimeter/sub-millimeter radiative transfer Barlakas et al., in preparation, JQSRT, 2021 Objectives: o Are there any systematic or random errors? o Where does the two-stream delta-Eddington “falls apart”? o Most studies focus on clear-sky or frequencies below ~90 GHz o Prepare RTTOV–SCATT for ICI ARTS vs RRTOV (–SCATT): o Clear–sky o All–sky (i.e. snow, ice, rain) Sensors: ARTS scattering database o Advanced Microwave Sounding Unit A (AMSUA) o 34 freq.: 1-886.4 GHz o Microwave Humidity Sounder (MHS) o 34 particle models (PM) o Ice Cloud Imager (ICI) o 35-45 sizes per PM Eriksson et al., ESSD, 2018
Results MHS: Clear–sky comparison !!!! Settings: o US standard atmosphere o Surface emissivity = 1 o An excellent agreement is found ± 0.1 K Correct spectroscopy! o Exceptions: ICI at 243.2 GHz (~0.2 K) and 448 GHz (~0.4 K) Ongoing by Met Office Barlakas et al., in preparation, JQSRT, 2021
Results ICI: Ice–only comparison !!!! Settings: o Gaussian cloud: 9-11 km o Large plate aggregate habit o Field et al., 2007 PSD ΔΤΒ = ΤΒ,cloudy – TB,clear Barlakas et al., in preparation, JQSRT, 2021
Results ICI: Ice–only comparison !!!! Settings: o Gaussian cloud: 9-11 km o Large plate aggregate habit o Field et al., 2007 PSD ΔΤΒ = ΤΒ,cloudy – TB,clear 325 GHz Barlakas et al., in preparation, JQSRT, 2021
Conclusions Summary & Outlook !!!! ⚈ Performance of RTTOV-SCATT o Clear–sky conditions o An excellent agreement is found (±0.1 K). For ICI, there is still some ongoing work. o All–sky conditions o ICI: ΔΤΒ ~3 K, with a relative error up to ΔΤΒ ~10 % o MHS and AMSUA: ΔΤΒ ~7 K, with a relative error up to ΔΤΒ ~15 % ⚈ Outlook o Where does the two-stream delta-Eddington “falls apart”? 3rd year o Finalize the manuscript o Sub-grid variability and cloud overlap scheme. Barlakas et al., in preparation, JQSRT, 2021
To–do–list Year 3 ! ⚈ Polarized correction scheme o Cross-track sounders MHS o Radar backscattering o Ice cloud retrievals (ongoing) Photo: esa.int ⚈ Performance of RTTOV-SCATT o Finalize the intercomparison (ongoing) o Sub-grid variability and cloud overlap scheme. Thank you for your attention !
You can also read