ITSC-12 Cloud processing in IASI context Lydie Lavanant Météo-France, Centre de Météorologie Spatiale, BP 147, 22300 Lannion Cedex France Purpose: Retrieval of cloud information and atmospheric profiles in cloudy conditions Steps: » Test case description » CO2-slicing method » Avhrr cloud description in IASI fov » IASI channel selection in cloudy conditions » Preliminary results of profile retrieval in cloudy conditions 01/03/02 Test Case Global IASI orbit simulation. Feb. 1996 13000 situations with: • simulated IASI cloudy and noisy spectra 0.25 cm-1 (R. Rizzi model) • Colocated atmospheric profiles: NWP analyses in T, Q, O on RTIASI levels Cloud description (cover, CLWV, CIWV) on 31 NWP levels Dataset provided by Eumetsat (ISSWG) CO2 slicing method Ref: Menzel and Stewart 1983, Smith and Frey 1990 [(Rclr – Rmeas)k / (Rclr – Rmeas)ref] – [Nek (Rclr - Rcld)k / Neref(Rclr - Rcld)ref] = fpc Rmeas: measured radiance Rclr: clear radiance computed from the colocated forecast Rcld: black-body radiance at the cloud level n k= channel from 690 cm-1 to 810 cm-1 Ref= reference channel = 899.75 cm-1 For each channel k: cloud pressure = pressure which minimises equation P_co2 = S (p_co2(k) w2(k)) / Sw2 W = dfpc / dlnp Ne = (Rclr – Rmeas)ref / (Rclr - Rcld)ref Assumption: one thin cloud layer Rejections: (Rclr – Rmeas) < sqrt(2)*radiometric noise Ne < 0 Preliminary results of CO2 method using CDS cloudy spectra 100-200 23 158.4 102.4 200-300 358 47.1 103.3 300-400 396 -9.7 132.7 400-500 658 -31.8 131.7 500-600 802 -51.3 120.2 600-700 913 -97.2 112.3 700-800 1057 -126.6 125.9 800-900 793 -147.6 135.8 900-1000 1534 -129.9 164.0 1000-1050 72 -51.4 71.4 Method: • • • • Adapt RTIASI for implementing RTTOV7 cloudy routines developed by F. Chevallier and al. (2001) Simulate cloudy noisy IASI spectra Rmeas for all CDS situations using: » NPW profiles (T,H2O,.., CC, CLWV, CIWV) » radiometric noise Compute clear noisy radiances Rclr for the same fov using: » RTIASI clear » noisy NWP profiles (apply forecast errors) Apply CO2-slicing method Examples of IASI cloudy spectra Variation with the number of channels 1 cloud layer Ne > 0.3 Variation with emissivity 1 cloud layer RTIASI cloudy + noise Profile= analysis 24 channels. resolution:5cm-1 Variation with emissivity Several cloud layers RTIASI cloudy + noise Profile= analysis Variation with emissivity 1 cloud layer RTIASI cloudy Profile=forecast P_Co2, e_Co2 s = 0.2 - 0.25 1 cloud layer RTIASI cloudy + noise Profile=forecast CDS dataset cloud pressure Cloud top pressure CO2 retrieved cloud pressure AVHRR Cloud mask in IASI fov Operational routine for HIRS fov (inside AAPP) Based on a threshold technique applied . every AVHRR pixel in sounder fov . to various combinations of channels » » Combinations of channels depend on: . geographical location of the pixel . solar illumination and viewing geometry » Thresholds computed in-line with: . constant values from experience . tabulated functions defined off-line through RTTOV simulations on climatological data-set . TWVC retrieved from colocated AMSU-A Current products: » percentage clear AVHRR in FOV » surface temperature from AVHRR split-window » black body cloud coverage in FOV » cloud top temperature for the black body layer » clear/cloudy flag for each AVHRR pixel Next version: »Ts, Tcld, Cloud type for each Avhrr pixel »-> number of clouds AVHRR Cloud mask in IASI fov AVHRR Cloud mask in IASI fov Validation over Europe correctly detected Cloudy targets Cloud free targets sea ; day 1774 (99.8%) 584 (87.8%) sea ; glint 269 (98.8%) 72 (89%) sea ; twilight 59 ( 98.3%) 12 (90%) land ; day 995 (99.5%) 638 (80.9%) land ; twilight 27 (100%) 11 (67.4%) Comparison of satellite obs. and Hirs 8 RTTOV6 Tbs using: March 2001 Clear land Clear sea Cloudy Cloudy sea land Cv>50 Cv>50 * NWP profile, N 937 103 136 55 * AVHRR clear cover +Ts, Bias (K) 0.02 0.64 0.49 0.15 Std (K) 0.99 0.39 0.98 0.65 7007 targets of 5x5 AVHRR pixels Noaa12, 14, 15 for 3 years 38 cloud types Mask comparison with visual analysis of satellite imagery by CMS nephanalysts * AVHRR black-body cloud cover +Tn Channels selection and retrieval in clear conditions on CDS the 300 most informative channels Clear situations nbsit= 187 (1/10) • Rodgers DFS selection • Guess error matrix = forecast • Use a mean profile for mid-latitude conditions Channels selection above the cloud Select channels from the 300 most informative channels in clear conditions Ex: for p_cloud=850 hPa . uncontaminated channels above the cloud top level: about 65% channels selected . cloud contaminated channels with (Tbobs – Tbgucld) < 0.3K : about 85% channels selected Profile retrieval in cloudy conditions. CDS dataset un-contaminated channels above the cloud 600 < p_cloud < 700 Nbsit= 132 (1/3) 700 < p_cloud < 800 Nbsit= 146 (1/4) 800 < p_cloud < 900 Nbsit= 138 (1/7) 900 < p_cloud < 1000 Nbsit= 166 (1/5) Profile retrieval in cloudy conditions with cloud information as control variable Cloud control variables: ln(p_cld), e_cld Cloud guess: CO2 p_cld and e_cld selected channels: Tbobs – Tbgucld < 0.3K -> more than 80% channels selected all P_cld > 800hPa 1807 situations (15% of situations) before 1d_var P_cld after 1d_var e_cld forecast as background with uncontaminated channels above cloud: 1DVar in clear conditions all selected channels: 1DVar cloudy Summary: • • • • Create IASI cloudy spectra using NWP analyses (T,H2o,.., CC, CLWV, CIWV) Use CO2 method to determine the cloud top pressure and emissivity Retrieve temperature profile in cloudy condition with CO2 cloud parameters as guess validate on CDS dataset Future: • • • • Consolidate the results on recent NWP data (with cloud profile information on 60 levels) -> package » add the water vapor profile » Combine IASI, AVHRR and AMSU information Validate on AIRS observations Adapt the method to the IASI stand-alone package Test a cloud-clearing method (J. Joiner)