DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Assimilation of Geostationary Infrared Satellite Data to Improve Forecasting of Mid-level Clouds Curtis J. Seaman

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Acknowledgements Advisor: u Thomas H. “Tom” Vonder Haar Committee: u William R. “Bill” Cotton u Wayne H. Schubert u Mahmood R. “Mo” Azimi (Electrical and Computer Engineering) u Tomislava “Tomi” Vukicevic (University of Colorado-Boulder) u Adam Kankiewicz (WindLogics, Inc.) u Manajit Sengupta, Andy Jones, Steve Fletcher, Scott Longmore, Laura Fowler, Cindy Combs, Louie Grasso, John Forsythe u Karll Renken, Michael Hiatt, Adam Carheden u Department of Defense

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Outline u Introduction u RAMDAS and GOES u Assimilation of GOES Imager (ch. 3 & 4) u Assimilation of GOES Sounder (ch. 7 & 11) u Variation of decorrelation lengths and variance (background error covariance) u Conclusions and Future Work

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Introduction Mid-level, layered clouds are poorly forecast u CLEX and CloudNet results u Vertically thin and most models have poor vertical resolution in mid-troposphere u Complex microphysics u Weak vertical velocity tricky to model u Models have poor moisture information (radiosonde dry bias: Soden et al. 2004) Mid-level, layered clouds (altocumulus and altostratus) are important u Cover 20-25% of the globe u Important for military, civilian aviation t Source of icing and turbulence t Obscure military targets, pilot visibility t Interfere with infrared (IR) and laser communication u Importance for climate? u Cause of blown forecasts Image: Illingworth et al. (2007) Image: Illingworth et al. (2007)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Satellite Data Assimilation u Operational forecast centers typically assimilate only cloud- free radiances t Ease, computational cost t McNally and Vespirini (1995); Garand and Hallé (1997); Ruggiero et al. (1999); McNally et al. (2000); Raymond et al. (2004); Fan and Tilley (2005) u Clouds cover 51-67% of the globe t ISCCP, CLAVR, Warren et al. (1986,1988) u Area of recent research t Marecal and Mahfouf (2002), Bauer et al. (2006), Weng et al. (2007) t Vukicevic et al. (2006) Image: NASA

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Regional Atmospheric Modeling Data Assimilation System (RAMDAS) u Mesoscale 4-DVAR assimilation system designed to assimilate clear and cloudy scene VIS, IR satellite data u Forward model: RAMS u Observational operator: VISIROO t SHDOM, OPTRAN, Deeter and Evans (1998) t Developed for GOES (Imager & Sounder) u Full adjoints for VISIROO and RAMS t Includes RAMS microphysics, but not convective parameterizations

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Regional Atmospheric Modeling Data Assimilation System (RAMDAS) Forward NWP model Satellite radiance over assimilation time window Forward Radiative Transfer to compute radiance from model output Reverse or adjoint of radiative transfer to compute adjoints Reverse or adjoint of NWP model to compute adjoints Update initial condition using propagated gradient of cost function. Compute gradient of cost function and decide whether it is small enough yet Generate forecast NoYes start end Greenwald et al. (2002) Zupanski et al. (2005)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, The 2 Nov 2001 Altocumulus Image:

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Imager Ch. 1 VIS 0.63  m Ch. 4 window 10.7  m Ch. 3 vapor 6.7  m Sounder Ch. 19 VIS 0.70  m Ch. 7 window 12.0  m Ch. 11 vapor 7.0  m

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment Set-Up u 75 x 75 (6 km horiz.) x 84 (stretched-z vert.) grid centered on North Platte, NE u Lateral boundaries masked to 50 x 50 x 84 to ignore boundary condition errors u RAMS initialized with 00 UTC FNL (GDAS) reanalysis data u 11 UTC output used to initialize RAMDAS u 1145 UTC GOES observations assimilated Control Variables: p  il u,v,w r ice r snow r total pressure (pert. Exner function) ice-liquid potential temperature winds ice water mixing ratio snow water mixing ratio total water mixing ratio

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Initial Forward Model Run Initial model is poor, but we’ll use it (and get valuable results) Dashed = Observed Sounding; Solid = Model Sounding Dashed = Observed Sounding; Solid = Model Sounding

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m) Before Assimilation After Assimilation Observed T b

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m) Before assimilation After assimilation

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m) Surface wind (before) Surface wind (after)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #1: Imager only GOES Imager channels 3 (6.7  m) & 4 (10.7  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m) Before Assimilation After Assimilation Observed T b

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m) Before assimilation After assimilation

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m) Surface wind (before) Surface wind (after)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #2: Sounder only GOES Sounder channels 7 (12.02  m) & 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Control Variable r l x,y [km] r l z [km] pressure temperature u-wind v-wind w-wind501.0 total water mixing ratio501.0 rain water mixing ratio500.5 snow water mixing ratio500.5 Decorrelation Lengths u Background error covariance matrix, B, in RAMDAS based on decorrelation length and variance u Assumed values of decorrelation length for each control variable shown u Decorrelation lengths and variances doubled and halved

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, u Increasing (decreasing) decorrelation length increases (decreases) impact of observations u Variance (not shown) had little effect Decorrelation Lengths Note: dashed lines correspond to soundings based on default values of decorrelation length

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Imager Experiment Sounder Experiment Decorrelation Lengths Doubled Mid-level Cloud

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Summary u GOES Imager experiment t Cooled the surface, increased upper-tropospheric humidity t Increased fog t No closer to producing mid-level cloud u GOES Sounder experiment t Produced subsidence inversion t Cooled, humidified atmosphere near 2 km AGL t Some surface cooling t No mid-level cloud, but closer to producing one u Decorrelation lengths are important t Increasing (decreasing) decorrelation length increases (decreases) effect of observations u Variance not important u Significant innovations achieved with only one observation time t Biggest changes to temperature, dew point and winds

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Conclusions u RAMDAS is designed to minimize the difference between the modeled and observed brightness temperatures t The assimilation can only modify the model state where the observations are sensitive to model variables t The mathematics doesn’t have a sense of the physics u When no cloud is present in the model, the adjoint calculates sensitivities based on having no cloud t In the GOES Imager case, ch. 4 is most sensitive to surface temperature in the absence of cloud, ch. 3 is most sensitive to upper-tropospheric humidity t GOES Sounder ch. 7 & 11 are more sensitive to low- to mid- troposphere temperature and humidity u Additional constraints are needed to ensure an appropriate physical solution

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Back to the Future u Add constraints t Surface temperature t Additional cloud information u Ideal decorrelation lengths u Log-normal distributions (Fletcher and Zupanski 2007) u More channels u More case studies t CLEX-9, CLEX-10

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Questions?

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Backup Slides

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Geostationary Operational Environmental Satellites (GOES) u Imager t Higher spatial, temporal resolution u Sounder t More channels u Experiments use window and water vapor channels t Imager ch. 3 & 4 t Sounder ch. 7 & 11 t Water vapor-only (3 & 11) Theoretical Weighting Functions

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Nov 2001 Altocumulus from CLEX-9 u RAMS run with 100 m vertical resolution initialized with Eta 40 km reanalysis t Peak RH of 85% too low to form cloud u Radiosonde dry bias? u Will assimilation of IR water vapor radiances help?

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Initial (No Assimilation) Forward Model Run

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m) Before Assimilation After Assimilation Observed T b

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m) Before assimilation After assimilation

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m) Surface wind (before) Surface wind (after)

DoD Center for Geosciences/Atmospheric Research at Colorado State University February 2, Experiment #3: water vapor only GOES Imager ch 3 (6.7  m) & Sounder ch 11 (7.02  m)