• Aksoy, A., , D. Dowell, , and C. Snyder, 2009: A multicase comparative assessment of the ensemble Kalman filter for assimilation of radar observations. Part I: Storm-scale analyses. Mon. Wea. Rev., 137, 18051824.

    • Search Google Scholar
    • Export Citation
  • Anderson, J. L., , T. Hoar, , K. Raeder, , H. Liu, , N. Collins, , R. Torn, , and A. Avellano, 2009: The Data Assimilation Research Testbed: A community data assimilation facility. Bull. Amer. Meteor. Soc., 90, 12831296.

    • Search Google Scholar
    • Export Citation
  • Benedetti, A., , and M. Janiskova, 2008: Assimilation of MODIS cloud optical depths in the ECMWF model. Mon. Wea. Rev., 136, 17271746.

  • Brennan, J. I., , Y. J. Kaufman, , I. Koren, , and R. R. Li, 2005: Aerosol-cloud interaction-misclassification of MODIS clouds in heavy aerosol. IEEE Trans. Geosci. Remote Sens., 43, 911915.

    • Search Google Scholar
    • Export Citation
  • Caya, A., , J. Sun, , and C. Snyder, 2005: A comparison between the 4DVAR and ensemble Kalman filter techniques for radar data assimilation. Mon. Wea. Rev., 133, 30813094.

    • Search Google Scholar
    • Export Citation
  • Chang, F.-L., , P. Minnis, , B. Lin, , M. Khaiyer, , R. Palikonda, , and D. Spangenberg, 2010a: A modified method for inferring cloud top height using GOES 12 imager 10.7 and 13.3-μm data. J. Geophys. Res., 115, D06208, doi:10.1029/2009JD012304.

    • Search Google Scholar
    • Export Citation
  • Chang, F.-L., , P. Minnis, , S. Sun-Mack, , L. Nyugen, and Y. Chen, 2010b: On the satellite determination of multi-layered multi-phase cloud properties. Preprints, 13th Conf. on Cloud Physics/13th Conf. on Atmospheric Radiation, Portland, OR, Amer. Meteor. Soc., JP1.10. [Available online at https://ams.confex.com/ams/pdfpapers/171180.pdf.]

  • Chen, Y., , F. Weng, , Y. Han, , and Q. Liu, 2008: Validation of the Community Radiative Transfer Model by using CloudSat data. J. Geophys. Res., 113, D00A03, doi:10.1029/2007JD009561.

    • Search Google Scholar
    • Export Citation
  • Derber, J. C., , and W.-S. Wu, 1998: The use of TOVS cloud-cleared radiances in the NCEP SSI analysis system. Mon. Wea. Rev., 126, 22872299.

    • Search Google Scholar
    • Export Citation
  • Dong, X., , P. Minnis, , G. G. Mace, , W. L. Smith Jr., , M. Poellot, , R. T. Marchand, , and A. D. Rapp, 2002: Comparison of stratus cloud properties deduced from surface, GOES, and aircraft data during the March 2000 ARM Cloud IOP. J. Atmos. Sci., 59, 32563284.

    • Search Google Scholar
    • Export Citation
  • Dong, X., , P. Minnis, , B. Xi, , S. Sun-Mack, , and Y. Chen, 2008: Comparison of CERES-MODIS stratus cloud properties with ground-based measurements at the DOE ARM Southern Great Plains site. J. Geophys. Res., 113, D03204, doi:10.1029/2007JD008438.

    • Search Google Scholar
    • Export Citation
  • Dowell, D. C., , G. S. Romine, , and C. Snider, 2010: Ensemble storm-scale data assimilation and prediction for severe convective storms. Preprints, 25th Conf. on Severe Local Storms, Denver, CO, Amer. Meteor. Soc., 9.5. [Available online at https://ams.confex.com/ams/pdfpapers/176121.pdf.]

  • Errico, R. M., 2000: The dynamical balance of leading singular vectors in a primitive-equation model. Quart. J. Roy. Meteor. Soc., 126, 16011618.

    • Search Google Scholar
    • Export Citation
  • Errico, R. M., , G. Ohring, , F. Weng, , P. Bauer, , B. Ferrier, , J.-F. Mahfouf, , and J. Turk, 2007: Assimilation of satellite cloud and precipitation observations in numerical weather prediction models: Introduction to the JAS special collection. J. Atmos. Sci., 64, 37373741.

    • Search Google Scholar
    • Export Citation
  • Gaspari, G., , and S. E. Cohn, 1999: Construction of correlation functions in two and three dimensions. Quart. J. Roy. Meteor. Soc., 125, 723757.

    • Search Google Scholar
    • Export Citation
  • Heemink, A. W., , M. Verlaan, , and A. J. Segers, 2001: Variance reduced ensemble Kalman filtering. Mon. Wea. Rev., 129, 17181728.

  • Hong, S.-Y., , and H.-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124, 23222339.

    • Search Google Scholar
    • Export Citation
  • Jones, T. A., , and S. A. Christopher, 2010: Satellite and radar remote sensing of southern Plains grass fires: A case study. J. Appl. Meteor. Climatol., 49, 21332146.

    • Search Google Scholar
    • Export Citation
  • Jones, T. A., , and D. J. Stensrud, 2012: Assimilating AIRS temperature and mixing ratio profiles using an ensemble Kalman filter approach for convective-scale forecasts. Wea. Forecasting, 27, 541564.

    • Search Google Scholar
    • Export Citation
  • Kalman, R. E., 1960: A new approach to linear filtering and prediction problems. J. Basic Eng., 82, 3545.

  • Kato, S., , L. M. Hinkelman, , and A. Cheng, 2006: Estimate of satellite-derived cloud optical thickness and effective radius errors and their effect on computed domain-averaged irradiances. J. Geophys. Res., 111, D17201, doi:10.1029/2005JD006668.

    • Search Google Scholar
    • Export Citation
  • Lauwaet, D., , K. D. Ridder, , and P. Pandey, 2011: Assimilating remotely sensed cloud optical thickness into a mesoscale model. Atmos. Chem. Phys. Discuss., 11, 13 35513 380.

    • Search Google Scholar
    • Export Citation
  • Lin, Y., , G. DiMego, , B. Ferrier, , J. Jung, , D. Keyser, , and E. Rodgers, 2003: Assimilation of GOES cloud top pressure data into the Eta Model. JSC/CAS Working Group on Numerical Experimentation (WGNE) Blue Book 2003, 2 pp. [Available online at http://www.wcrp-climate.org/WGNE/BlueBook/2003/individual-articles/01-lin.pdf.]

  • Mace, G. G., , Y. Zhang, , S. Platnick, , M. D. King, , P. Minnis, , and P. Yang, 2005: Evaluation of cirrus cloud properties from MODIS radiances using cloud properties derived from ground-based data collected at the ARM SGP site. J. Appl. Meteor., 44, 221240.

    • Search Google Scholar
    • Export Citation
  • Mansell, E. R., , C. L. Ziegler, , and E. C. Bruning, 2010: Simulated electrification of a small thunderstorm with two-moment bulk microphysics. J. Atmos. Sci., 67, 171194.

    • Search Google Scholar
    • Export Citation
  • McNally, A. P., , J. C. Derber, , W.-S. Wu, , and B. B. Katz, 2000: The use of TOVS level-1B radiances in the NCEP SSI analysis system. Quart. J. Roy. Meteor. Soc., 126, 689724.

    • Search Google Scholar
    • Export Citation
  • McNally, A. P., , P. D. Watts, , J. A. Smith, , R. Engelen, , G. A. Kelly, , J.-N. Thépaut, , and M. Matricard, 2006: The assimilation of AIRS radiance data at ECMWF. Quart. J. Roy. Meteor. Soc., 132, 935957.

    • Search Google Scholar
    • Export Citation
  • McPherson, R. A., and Coauthors, 2007: Statewide monitoring of the mesoscale environment: A technical update on the Oklahoma Mesonet. J. Atmos. Oceanic Technol., 24, 301321.

    • Search Google Scholar
    • Export Citation
  • Mecikalski, J. R., , P. Minnis, , and R. Palikonda, 2013: Use of satellite derived cloud properties to quantify growing cumulus beneath cirrus clouds. Atmos. Res., 120–121, 192201.

    • Search Google Scholar
    • Export Citation
  • Menzel, W. P., , and J. F. Purdom, 1994: Introducing GOES-I: The first of a new generation of geostationary operational environmental satellites. Bull. Amer. Meteor. Soc., 75, 757781.

    • Search Google Scholar
    • Export Citation
  • Migliorini, S., 2012: On the equivalence between radiance and retrieval assimilation. Mon. Wea. Rev., 140, 258265.

  • Milbrandt, J. A., , and M. K. Yau, 2005: A multimoment bulk microphysics parameterization. Part I: Analysis of the role of the spectral shape parameter. J. Atmos. Sci., 62, 30513064.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , J. Huang, , B. Lin, , Y. Yi, , R. F. Arduini, , T.-F. Fan, , J. K. Ayers, , and G. G. Mace, 2007: Ice cloud properties in ice-over-water cloud systems using Tropical Rainfall Measuring Mission (TRMM) visible and infrared scanner and TRMM Microwave Imager data. J. Geophys. Res., 112, D06206, doi:10.1029/2006JD007626.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2008a: Near-real time cloud retrievals from operational and research meteorological satellites. Proc. SPIE Europe Remote Sens., 7107-2, Cardiff, Wales, SPIE, 1–8.

  • Minnis, P., and Coauthors, 2008b: Cloud detection in non-polar regions for CERES using TRMM VIRS and Terra and Aqua MODIS data. IEEE Trans. Geosci. Remote Sens., 46, 38573884.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2010: CERES Edition 3 cloud retrievals. Preprints, 13th Conf. on Atmospheric Radiation, Portland, OR, Amer. Meteor. Soc., 5.4. [Available online at https://ams.confex.com/ams/pdfpapers/171366.pdf.]

  • Minnis, P., and Coauthors, 2011: CERES Edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data. Part I: Algorithms. IEEE Trans. Geosci. Remote Sens., 49, 43744400.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2012: Simulations of infrared radiances over a deep convective cloud system observed during TC4: Potential for enhancing nocturnal ice cloud retrievals. Remote Sens., 4, 30223054.

    • Search Google Scholar
    • Export Citation
  • Otkin, J. A., 2010: Clear and cloudy sky infrared brightness temperature assimilation using an ensemble Kalman filter. J. Geophys. Res., 115, D19207, doi:10.1029/2009JD013759.

    • Search Google Scholar
    • Export Citation
  • Painemal, D., , P. Minnis, , J. K. Ayers, , and L. O'Neill, 2012: GOES-10 microphysical retrievals in marine warm clouds: Multi-instrument validation and daytime cycle over the southeast Pacific. J. Geophys. Res., 117, D19212, doi:10.1029/2012JD017822.

    • Search Google Scholar
    • Export Citation
  • Parrish, D. F., , and J. C. Derber, 1992: The National Meteorological Center's spectral statistical interpolation analysis system. Mon. Wea. Rev., 120, 17471763.

    • Search Google Scholar
    • Export Citation
  • Pavelin, E., , S. English, , and J. Eyre, 2008: The assimilation of cloud-affected infrared satellite radiances for numerical weather prediction. Quart. J. Roy. Meteor. Soc., 134, 737749.

    • Search Google Scholar
    • Export Citation
  • Pincus, R., , R. J. P. Hofmann, , J. L. Anderson, , K. Raeder, , N. Collins, , and J. S. Whitaker, 2011: Can fully accounting for clouds in data assimilation improve short-term forecast by global models? Mon. Wea. Rev., 139, 946956.

    • Search Google Scholar
    • Export Citation
  • Polkinghorne, R., , and T. Vukicevic, 2011: Data assimilation of cloud affected radiances in a cloud resolving model. Mon. Wea. Rev., 139, 755773.

    • Search Google Scholar
    • Export Citation
  • Polkinghorne, R., , T. Vukicevic, , and F. Evans, 2010: Validation of cloud resolving model background data for data assimilation. Mon. Wea. Rev., 138, 781795.

    • Search Google Scholar
    • Export Citation
  • Schmit, T. J., , E. M. Prins, , A. J. Schreiner, , and J. J. Gurka, 2001: Introducing the GOES-M imager. Natl. Wea. Dig., 25, 2837.

  • Schmit, T. J., , M. M. Gunshor, , W. P. Menzel, , J. J. Gurka, , J. Li, , and A. S. Bachmeier, 2005: Introducing the next-generation Advanced Baseline Imager on GOES-R. Bull. Amer. Meteor. Soc., 86, 10791096.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO 80307.]

  • Smith, W. L., Jr., , P. Minnis, , L. Nguyen, , D. Spangenberg, , R. Austin, , and G. Mace, 2007: Preliminary comparison of CloudSat liquid water path to passive satellite estimates. A-Train Symp., Lille, France, NASA. [Available online at http://www-angler.larc.nasa.gov/site/doc-library/52-Smith.Lille.07.pdf.]

  • Smith, W. L., Jr., , P. Minnis, , H. Finney, , R. Palikonda, , and M. M. Khaiyer, 2008: An evaluation of operational GOES-derived single-layer cloud top heights with ARSCL data over the ARM Southern Great Plains site. Geophys. Res. Lett., 35, L13820, doi:10.1029/2008GL034275.

    • Search Google Scholar
    • Export Citation
  • Smith, W. L., Jr., , P. Minnis, , S. G. Benjamin, , and S. S. Weygandt, 2010: 4-D water content fields derived from operational satellite data. Proc. IEEE 2010 Int. Geoscience and Remote Sensing Symp., Honolulu, HI, IEEE, 308–311.

  • Snyder, C., , and F. Zhang, 2003: Assimilation of simulated Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 131, 16631677.

    • Search Google Scholar
    • Export Citation
  • Stein, T. H. M., , J. Delanöe, , and R. J. Hogan, 2011: A comparison among four different methods for ice-cloud properties using data from CloudSat, CALIPSO, and MODIS. J. Appl. Meteor. Climatol., 50, 19521969.

    • Search Google Scholar
    • Export Citation
  • Thompson, G., , R. M. Rasmussen, , and K. Manning, 2004: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon. Wea. Rev., 132, 519542.

    • Search Google Scholar
    • Export Citation
  • Tong, M., , and M. Xue, 2005: Ensemble Kalman filter assimilation of Doppler radar data with compressible nonhydrostatic model: OSS experiments. Mon. Wea. Rev., 133, 17891807.

    • Search Google Scholar
    • Export Citation
  • Torn, R. D., , G. J. Hakim, , and C. Snyder, 2006: Boundary conditions for limited area ensemble Kalman filters. Mon. Wea. Rev., 134, 24902502.

    • Search Google Scholar
    • Export Citation
  • Vukicevic, T., , T. Greenwald, , M. Zupanski, , D. Zupanski, , T. Vonder Haar, , and A. Jones, 2004: Mesoscale cloud state estimation from visible and infrared satellite radiances. Mon. Wea. Rev., 132, 30663077.

    • Search Google Scholar
    • Export Citation
  • Vukicevic, T., , M. Sengupta, , A. Jones, , and T. Vonder Haar, 2006: Cloud resolving data assimilation: Information content of IR window observations and uncertainties in estimation. J. Atmos. Sci., 63, 901919.

    • Search Google Scholar
    • Export Citation
  • Waliser, D., and Coauthors, 2009: Cloud ice: A climate model challenge with signs and expectations of progress. J. Geophys. Res., 114, D00A21, doi:10.1029/2008JD010015.

    • Search Google Scholar
    • Export Citation
  • Weisz, E., , J. Li, , J. Li, , D. K. Zhou, , H.-L. Huang, , M. D. Goldberg, , and P. Yang, 2007: Cloudy sounding and cloud top height retrieval from AIRS alone single field-of-view radiance measurements. Geophys. Res. Lett., 34, L12802, doi:10.1029/2007GL030219.

    • Search Google Scholar
    • Export Citation
  • Xue, M., , M. Tong, , and K. K. Droegemeier, 2006: An OSSE framework based on the ensemble square root Kalman filter for evaluating the impact of data from radar networks on thunderstorm analysis and forecasting. J. Atmos. Oceanic Technol., 23, 4666.

    • Search Google Scholar
    • Export Citation
  • Yucel, I., , W. J. Shuttleworth, , R. T. Pinker, , L. Lu, , and S. Sorooshian, 2002: Impact of ingesting satellite-derived cloud cover into the Regional Atmospheric Modeling System. Mon. Wea. Rev., 130, 610628.

    • Search Google Scholar
    • Export Citation
  • Yucel, I., , W. J. Shuttleworth, , X. Gao, , and S. Sorooshian, 2003: Short-term performance of MM5 with cloud-cover assimilation from satellite observations. Mon. Wea. Rev., 131, 17971810.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    GOES-13 1-km resolution visible satellite imagery at 2045 UTC 10 May 2010 with Weather Surveillance Radar-1988 Doppler (WSR-88D) reflectivity at 1 km AGL from the KTLX radar overlaid. Also shown are the severe weather reports (red = tornado, green = hail, blue = wind) that occur during this event between 2000 UTC 10 May and 0100 UTC 11 May.

  • View in gallery

    (a) CWP and (b) CTP retrieved from GOES-13 data at 2045 UTC. Black dots denote pixels removed as being potentially dust contaminated.

  • View in gallery

    Mesoscale (domain 1) and nested grid convective scale (domain 2) WRF domains used for both experiments.

  • View in gallery

    Location of observations assimilated into WRF-DART. (a) Conventional surface, aircraft, and radiosonde temperature observations between 1800 and 2100 UTC. Characteristics of humidity, wind, and pressure observations are similar. (b) GOES CWP observations at 2045 UTC with clear-sky (CWP = 0) and cloudy regions (CWP > 0) denoted. White areas indicate regions where no valid observations exist or are rejected by DART.

  • View in gallery

    Observation diagnostics for CWP showing model bias (circles), RMSD (squares), and TSPRD (triangles) between the prior and posterior analysis fields for each 15-min cycle starting at 1800 UTC and continuing until 2100 UTC. The number of observations assimilated during each cycle is also shown (diamonds).

  • View in gallery

    Analysis increments for (a) 700-hPa temperature (T700) and (b) water vapor mixing ratio (Q700) at 1815 UTC, which represents the first assimilation cycle satellite data are assimilated. (c),(d) As in (a),(b), but at 2045 UTC representing the final time satellite data are assimilated.

  • View in gallery

    (a) GOES-13 CWP retrievals at 1945 UTC with corresponding posterior ensemble mean CWP generated from (b) CONV and (c) PATH model analyses at the same time. (d),(e),(f) As in (a),(b),(c), but for 2045 UTC. Black contours in (b),(c),(e),(f) represent 0.5 kg m−2 contour of GOES CWP shown in (a) and (c), respectively.

  • View in gallery

    Difference (PATH − CONV) in CWP at (a) 1945 and (b) 2045 UTC analysis times after satellite data have been assimilated. Blue regions indicate where PATH generates lower values of CWP whereas red regions indicate that PATH generates higher CWP values.

  • View in gallery

    (a) Contour plot of Oklahoma Mesonet SWF at 2100 UTC with corresponding Mesonet locations and wind barbs overlaid (white) with short barbs = 5 m s−1 and long barbs = 10 m s−1. Posterior ensemble mean SWF generated from (b) CONV and (c) PATH at 2100 UTC with Mesonet values overlaid along with model wind barbs (black). The color contrast between Mesonet and model SWF indicates where differences between the two are the greatest.

  • View in gallery

    Time series of bias and RMSD for (a) SWF and (b) 2-m temperature between 2030 and 2230 UTC calculated between Mesonet observations and each experiment's analysis at each time step.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 39 39 13
PDF Downloads 25 25 7

Evaluation of a Forward Operator to Assimilate Cloud Water Path into WRF-DART

View More View Less
  • 1 Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma
  • 2 NOAA/OAR/National Severe Storms Laboratory, Norman, Oklahoma
  • 3 NASA Langley Research Center, Hampton, Virginia
  • 4 Science Systems and Applications, Inc., Hampton, Virginia
© Get Permissions
Full access

Abstract

Assimilating satellite-retrieved cloud properties into storm-scale models has received limited attention despite its potential to provide a wide array of information to a model analysis. Available retrievals include cloud water path (CWP), which represents the amount of cloud water and cloud ice present in an integrated column, and cloud-top and cloud-base pressures, which represent the top and bottom pressure levels of the cloud layers, respectively. These interrelated data are assimilated into an Advanced Research Weather Research and Forecasting Model (ARW-WRF) 40-member ensemble with 3-km grid spacing using the Data Assimilation Research Testbed (DART) ensemble Kalman filter. A new CWP forward operator combines the satellite-derived cloud information with similar variables generated by WRF. This approach is tested using a severe weather event on 10 May 2010. One experiment only assimilates conventional (CONV) observations, while the second assimilates the identical conventional observations and the satellite-derived CWP (PATH).

Comparison of the CWP observations at 2045 UTC to CONV and PATH analyses shows that PATH has an improved representation of both the magnitude and spatial orientation of CWP compared to CONV. Assimilating CWP acts both to suppress convection in the model where none is present in satellite data and to encourage convection where it is observed. Oklahoma Mesonet observations of downward shortwave flux at 2100 UTC indicate that PATH reduces the root-mean-square difference errors in downward shortwave flux by 75 W m−2 compared to CONV. Reduction in model error is generally maximized during the initial 30-min forecast period with the impact of CWP observations decreasing for longer forecast times.

Corresponding author address: Dr. Thomas A. Jones, Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, and NOAA/OAR/National Severe Storms Laboratory, 120 David L. Boren Blvd., Norman, OK 73072. E-mail: thomas.jones@noaa.gov

Abstract

Assimilating satellite-retrieved cloud properties into storm-scale models has received limited attention despite its potential to provide a wide array of information to a model analysis. Available retrievals include cloud water path (CWP), which represents the amount of cloud water and cloud ice present in an integrated column, and cloud-top and cloud-base pressures, which represent the top and bottom pressure levels of the cloud layers, respectively. These interrelated data are assimilated into an Advanced Research Weather Research and Forecasting Model (ARW-WRF) 40-member ensemble with 3-km grid spacing using the Data Assimilation Research Testbed (DART) ensemble Kalman filter. A new CWP forward operator combines the satellite-derived cloud information with similar variables generated by WRF. This approach is tested using a severe weather event on 10 May 2010. One experiment only assimilates conventional (CONV) observations, while the second assimilates the identical conventional observations and the satellite-derived CWP (PATH).

Comparison of the CWP observations at 2045 UTC to CONV and PATH analyses shows that PATH has an improved representation of both the magnitude and spatial orientation of CWP compared to CONV. Assimilating CWP acts both to suppress convection in the model where none is present in satellite data and to encourage convection where it is observed. Oklahoma Mesonet observations of downward shortwave flux at 2100 UTC indicate that PATH reduces the root-mean-square difference errors in downward shortwave flux by 75 W m−2 compared to CONV. Reduction in model error is generally maximized during the initial 30-min forecast period with the impact of CWP observations decreasing for longer forecast times.

Corresponding author address: Dr. Thomas A. Jones, Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, and NOAA/OAR/National Severe Storms Laboratory, 120 David L. Boren Blvd., Norman, OK 73072. E-mail: thomas.jones@noaa.gov

1. Introduction

The characterization of cloud properties in numerical weather prediction (NWP) models plays an important role in forecasting convective systems. To reduce the uncertainty in analyzed cloud properties, the assimilation of nontraditional data sources such as observations from radars (e.g., Snyder and Zhang 2003) and satellites (e.g., Pincus et al. 2011) is needed. In particular, radar reflectivity and radial velocity have been used to estimate the location of precipitation (and the corresponding clouds) and sample the near-storm environment (e.g., Dowell et al. 2010). The assimilation of radar data has proven effective at increasing the accuracy of the model analyses and downstream forecasts.

One disadvantage of radar data assimilation is that precipitation radars are not very sensitive to nonprecipitating clouds. Yet it is important that nonprecipitating clouds be properly analyzed, since they can affect energy balances within the model and also indicate the locations of possible convective initiation (e.g., Mecikalski et al. 2013). Remote sensing observations from satellites provide information on cloud properties on similar horizontal and temporal scales as radar data, but with greater sensitivity to the non- or preconvective clouds that may be present. The question then becomes how best to assimilate these observations, since two different approaches to satellite data assimilation are in use. The first is the direct assimilation of satellite infrared and microwave radiances. Model state variables are generally transformed into simulated satellite radiances via a radiative transfer model (RTM) built into the forward operator (Vukicevic et al. 2004, 2006; McNally et al. 2006; Chen et al. 2008; Otkin 2010; Polkinghorne et al. 2010; Polkinghorne and Vukicevic 2011). Assimilating radiances directly often has been the favored approach in operations and research studies since it avoids uncertainties and discrepancies in the various retrieval algorithms that differ from satellite to satellite (Derber and Wu 1998; Errico 2000). This technique generally performs best in clear-sky regions, but it can be very resource intensive depending on the number and resolution of channels being assimilated (Migliorini 2012). Recent research has begun to focus on assimilating cloudy radiances, but this remains an even more challenging and resource intensive task (Vukicevic et al. 2004, 2006; Errico et al. 2007; Weisz et al. 2007; Pavelin et al. 2008; Polkinghorne et al. 2010; Polkinghorne and Vukicevic 2011).

Despite the inherent challenges, which include uncertainties in the background error and covariance estimate for high-resolution cloudy radiances and uncertainties in the relationship between radiance observations and the three-dimensional (3D) microphysical structures of clouds, positive results have been found when assimilating these data. Vukicevic et al. (2004, 2006), Polkinghorne et al. (2010), and Polkinghorne and Vukicevic (2011) found that assimilating these data using a four-dimensional (4D) variational technique coupled with an RTM improved the characterization of cloud properties within high-resolution NWP analysis and forecasts. In particular, the 3D distribution of cloud ice mass was improved significantly while liquid water cloud mass below ice layers was less affected. While very promising, applying these techniques within an ensemble modeling framework proved too resource intensive for this research requiring a different approach to be used.

A more resource-friendly and easier-to-interpret approach to satellite data assimilation is to use retrieved products (Migliorini 2012). For clouds, these products can include cloud-top pressure (CTP), cloud-top temperature (CTT), cloud water path (CWP), cloud ice path (CIP), cloud optical thickness (COT), and cloud coverage (e.g., Minnis et al. 2011). CWP represents the column-integrated cloud liquid and ice water from cloud base to cloud top, which is often defined in pressure levels [cloud-base pressure (CBP) and cloud-top pressure (CTP; Minnis et al. 2011)]. COT represents the optical extinction resulting from a cloud and is directly related to cloud thickness (Kato et al. 2006). These products have the advantage of being directly related to the cloud properties being analyzed within NWP models and do not require the use of complex RTMs. One challenge that remains is the lack of vertical sensitivity of satellite sensors to cloud properties. For geostationary sensors in particular, only cloud-top and sometimes cloud-base heights can be resolved. The vertical distribution of cloud microphysical properties, such as cloud liquid water, cannot currently be retrieved directly. However, satellite retrievals still provide a wide range of information on clouds not generally available from other sources.

Satellite data assimilation on regional and smaller scales has received much less attention than corresponding radar data assimilation in part due to the difficulties inherent in observing and assimilating cloud properties. There are many nonunique relationships between observed variables and cloud properties and the relationships that do exist are often highly nonlinear, violating many of the underlying error distribution assumptions used in several common forms of data assimilation (Pincus et al. 2011). No single solution exists to these difficulties, but recent research has taken steps to begin addressing this problem. Lin et al. (2003) use satellite cloud fraction and cloud-top pressure to adjust the humidity fields within the model where clouds are detected. For example, atmospheric water vapor is adjusted to a saturated value at that level (and below) where a cloud exists. Above this layer, the water vapor is reduced if it is saturated within the model. Similarly, if a satellite failed to detect a cloud and the model analysis shows a saturated environment, the model water vapor content is reduced. Following this theme, Yucel et al. (2002, 2003) proposed a more complex technique to adjust model cloud (and humidity) variables toward the satellite observed cloud properties using a four-step process. Step 1 determines where both satellite and model data indicate clear-sky conditions. In this case, no changes to the model analysis are made. Step 2 determines where both model and satellite indicate clouds. Here, the total column cloud liquid water and ice from satellite measurements are used to adjust the corresponding values in the model, with the model-analyzed vertical distribution of cloud water/ice used to guide the fit to the total column cloud water reported by the satellite. This assumes the model vertical distribution of cloud properties is correct even if the total column concentrations are off. Step 3 represents the case where no clouds are detected by satellite yet the analysis contains a cloud. In this case, all model-analyzed cloud properties are set to zero. Finally, step 4 is the case where the satellite detects a cloud not present in the model analysis. This is the most challenging case, as it requires the total column satellite-derived cloud properties are assimilated on to the three-dimensional model grid.

Benedetti and Janiskova (2008) and Lauwaet et al. (2011) use satellite COT within a forward operator designed to determine the best model profiles of column-integrated cloud water and cloud ice that fit their respective observations through either best-fit or iterative algorithms. This represents a more direct assimilation approach than described by Yucel et al. (2002, 2003), but relies heavily on the assumptions made within the forward operator to derive the vertical profiles of cloud properties from their column-integrated observations. Results from these studies clearly indicate that assimilating cloud properties in some form improves the cloud analysis fields as well as the energy balance and precipitation forecasts. Results also indicate that the more frequent the assimilation step, the greater the potential improvement.

Many uncertainties and questions remain as to the best cloud properties and assimilation techniques to use when attempting to correctly characterize clouds within high-resolution NWP models. While it would be impossible to provide closure to all these issues in a single study, we hope to provide some answers by assimilating Geostationary Operational Environmental Satellite (GOES) imager cloud property retrievals into a convective-scale NWP model using an ensemble Kalman filter (EnKF) approach. The primary advantage of this approach is that it provides a flow-dependent and dynamically evolving estimate of the background error covariance for assimilated observations (Kalman 1960). This is especially important for satellite-derived cloud properties as they are a function of the observing conditions and the actual characteristics of the cloud, both of which can vary significantly with space and time (e.g., Polkinghorne et al. 2010). The EnKF approach also reduces the reliance on uniform error variance characteristics required by some other assimilation techniques (Heemink et al. 2001).

Finally, cloud ice or liquid phase, determined for the top of the cloud, is also available. For this study, CWP is the primary variable to be assimilated, while CTP and CBP are used to define the vertical extent of the cloud layer during the assimilation process. The cloud properties used in this research are derived from the imager on board the GOES-East (GOES-13) satellite using retrieval algorithms developed for the Clouds and the Earth's Radiant Energy System project (Minnis et al. 2008a,b, 2011). These algorithms are currently capable of retrieving CTP, CBP, CWP, COT, and cloud phase among other properties.

The NWP model and assimilation approach chosen are the Advanced Research Weather Research and Forecasting Model (ARW-WRF; Skamarock et al. 2008) in conjunction with the Data Assimilation Research Testbed (DART) software (Anderson et al. 2009). To the authors' knowledge, actually assimilating the values of CWP, CTP, and CBP for real cases on a storm-scale grid has not been previously attempted. As a result, this research requires the development and testing of a forward operator to transform WRF and DART hydrometeor variables into a parameter that can be compared against GOES satellite retrievals (Pincus et al. 2011).

To assess the viability of assimilating CWP, a southern plains severe weather event occurring on 10 May 2010 is selected for study and two data assimilation experiments are conducted for comparison. One assimilates only conventional observations and is used as a control. The second assimilates the identical conventional observations plus CWP every 15 min over a 3-h window using the newly developed forward operator described in section 4. Results from these experiments are compared during both the assimilation cycle and a 90-min forecast period to determine the potential for satellite CWP data to increase skill in storm-scale forecasting. Oklahoma Mesonet data represent the primary verification observations for this study (McPherson et al. 2007).

The characteristics of the 10 May 2010 case study event are described in section 2. Section 3 provides information on the GOES-based cloud retrievals. The WRF-DART characteristics and the assimilation techniques are outlined in section 4. Section 5 describes the observation diagnostics of the assimilation, with sections 6 and 7 providing an overview and verification of the analysis fields. Summary and conclusions are given in section 8.

2. Event summary: 10 May 2010

A surface low pressure center deepens as it moves eastward through southern Kansas, trailing a dryline southward into central Oklahoma at 2100 UTC 10 May 2010. Convective available potential energy (CAPE) exceeds 2500 J kg−1 and storm relative helicity (SREH) exceeds 500 m2 s−2 throughout much of central and southeastern Oklahoma at this time, indicating an environment favorable for tornadic supercells. Thunderstorms develop ahead of the eastward progressing dryline in southern Kansas around 1800 UTC and in central and southern Oklahoma by 2100 and 2200 UTC, respectively. Visible satellite imagery from GOES-13 at 2045 UTC shows a line of cumulus clouds in west-central Oklahoma ahead of dryline that are in the process of developing into thunderstorms. Farther north in Kansas, a fully developed supercell is suggested from the large cirrus outflow present. Visible imagery also shows a large area of low-level stratus clouds in eastern Oklahoma and the adjacent regions of Kansas and north Texas (Fig. 1). Dozens of damaging wind, severe hail, and tornado reports occur between 2000 UTC 10 May and 0100 UTC 11 May (Fig. 1). In contrast to the cloudy region east of the dryline, clear sky and dry air characterize the environment behind the dryline in western Oklahoma, with some dust even being blown in from west Texas.

Fig. 1.
Fig. 1.

GOES-13 1-km resolution visible satellite imagery at 2045 UTC 10 May 2010 with Weather Surveillance Radar-1988 Doppler (WSR-88D) reflectivity at 1 km AGL from the KTLX radar overlaid. Also shown are the severe weather reports (red = tornado, green = hail, blue = wind) that occur during this event between 2000 UTC 10 May and 0100 UTC 11 May.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

3. GOES imager products

The GOES imager takes multispectral images over the Western Hemisphere every half hour with images over the continental United States (CONUS) occurring every 5–15 min depending on the need (Menzel and Purdom 1994). This allows for high temporal sampling of a particular location, which represents a significant advantage compared to polar-orbiting satellites that can only observe the same location twice per day. GOES-13 views the subject domain with a sensor zenith angle of ~48° ± 4°. The GOES imager measures atmospheric radiances in five spectral bands: a 1-km visible channel at 0.65 μm, a 4-km shortwave infrared channel at 3.9 μm, a 4-km water vapor channel at 6.5 μm, a 4-km atmospheric infrared window channel at 10.8 μm, and an 8-km CO2 absorption channel at 13.3 μm (Menzel and Purdom 1994; Schmit et al. 2001).

a. Retrieval algorithm

Cloud properties are retrieved from 4-km GOES imager data for pixels classified as cloudy. Pixels are determined to be clear or cloudy using the multispectral threshold retrieval algorithm developed by Minnis et al. (2008b) and adapted to geostationary data (Minnis et al. 2008a). A variety of cloud parameters are retrieved using modified versions of the visible infrared shortwave-infrared split-window technique (VISST), and the shortwave-infrared infrared split-window technique (SIST; Minnis et al. 2008a,b, 2011). The first method is used during the day (solar zenith angles ≤82°), when visible spectrum data are available. Otherwise the latter is used. The SIST was modified to accommodate the replacement of the 12-μm split-window channel on the older GOES series (GOES-8 through GOES-12) with the 13-μm channel the newer GOES series (GOES-13 and thereafter). For the altered SIST, the modified CO2-absorption technique (MCAT) of Chang et al. (2010a,b) uses 10.8- and 13.3-μm channels to retrieve CTP, CTT, and cloud emissivity for clouds having COT less than ~4. The 3.9-μm channel is used to estimate the particle effective size. During the day, the MCAT is applied independently of the VISST and its results replace those of the VISST for ice clouds having VISST COT < 4 and VISST CTP > MCAT CTP + 200 hPa. Otherwise, the algorithms are mostly the same as the original versions. Cloud-base pressure is the pressure corresponding to the altitude equal to the difference between cloud-top height and cloud thickness. Cloud thickness is parameterized in terms of CWP, CTT, ln(COT), and, for liquid clouds, droplet effective radius (Minnis et al. 2010). Hereafter, the combined retrieval algorithm is simply referred to as VISST. A full description of the unmodified VISST algorithm is found in Minnis et al. (2008b, 2011). We have chosen to assimilate the retrieved values of CWP, CTP, and CBP at 15-min intervals at the nominal 4-km resolution.

The retrieved values of CWP and CTP for 2045 UTC 10 May (Fig. 2) indicate that a large area of low-top clouds (CTP > 800 hPa) exist over the eastern half of Oklahoma corresponding to the stratus field observed in Fig. 1. The accompanying CWP values are generally small being <0.3 kg m−2. In the northern portion of the domain, the retrieved CTP < 300 hPa corresponds with the cirrus outflow from the early convection with corresponding CWP values being much larger, in excess of 1.0 kg m−2. The line of cumulus ahead of the dryline is also evident from GOES CWP with small areas of elevated CWP values present associated with developing convection.

Fig. 2.
Fig. 2.

(a) CWP and (b) CTP retrieved from GOES-13 data at 2045 UTC. Black dots denote pixels removed as being potentially dust contaminated.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

b. Retrieval uncertainties

Comparison of GOES cloud-top and cloud-base heights with surface-based cloud radar and lidar observations at the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site near Lamont, Oklahoma, found that GOES cloud heights retrieved during daytime hours are generally accurate to within ±1.0 km with a slight low bias (Smith et al. 2008). Uncertainties are greatest for thin cirrus clouds in the upper troposphere and smallest for single layer low-level stratus. Cloud-base heights for deep convection are also sometimes too high, since lower-level information cannot be ascertained because of the very opaque nature of these clouds. Converting ±1 km to pressure coordinates yields an uncertainty of approximately ±50 hPa for cloud heights in the midtroposphere. Height uncertainties increase in the presence of multilayer clouds with retrieved CTP and CBP being the most representative of the top-most cloud layer in the atmosphere.

Since cloud phase is a binary classification, the distinction between CWP and CIP is rather fuzzy for deep, multiphase clouds and when multiple cloud layers are present. This represents a challenge for data assimilation purposes, since WRF resolves this distinction in fine detail. One way to address this issue is to ignore the phase information from the satellite retrieval and assume that CWP be equal to the sum of all cloud water regardless of the cloud phase. Hereafter, CWP from GOES refers to either CWP or CIP, and the phase flag in the retrieval is ignored.

CWP from VISST has been compared with a variety of other types of measurements to help understand the uncertainties. Dong et al. (2002) found a 6% underestimate of the liquid water path from GOES-8 data relative to microwave radiometer retrievals for overcast stratus clouds over the SGP site. For the same conditions during a different period, Dong et al. (2008) found that the CWP from VISST applied to Moderate Resolution Imaging Spectroradiometer (MODIS) data differed from the SGP microwave radiometer retrieval by 7%, on average. Similar differences were found for marine stratus scenes using CWP from the VISST applied to GOES-10 data and compared with three different satellite-borne microwave imager retrievals (Painemal et al. 2012). For thin ice clouds, Mace et al. (2005) determined that the CWP retrieved from VISST applied to MODIS data is 9% greater than that from surface-based cloud radar data at the SGP. For thicker clouds over the CONUS, Smith et al. (2007) determined that the GOES-estimated CWP for thick ice clouds is unbiased compared to CloudSat cloud radar retrievals for areas having snow-free surfaces. The instantaneous uncertainty is on order of a factor of 2, which is comparable to the uncertainty in any given radar retrieval algorithm (e.g., Stein et al. 2011). Over snow, the CWP tends to be overestimated using the limited number of channels available on the current GOES imagers. Waliser et al. (2009) showed that, over the globe, the mean VISST-derived CWPs for ice clouds from MODIS agree well spatially and in magnitude with those from CloudSat.

The differences noted above can be considered as first-order estimates of the uncertainties in the VISST-derived CWP values. The uncertainties are likely smaller because the standard deviations of the differences include errors due to time and space sampling differences, errors in the microwave and radar retrieval methods, and calibration uncertainties. Removal of those error sources would reduce the instantaneous uncertainties. Assessing CWP uncertainties in optically thick ice clouds is difficult because those clouds are often associated with storms that are unfriendly to aircraft and have large internal variabilities in hydrometer size and concentration. Thus, it is not surprising that differences for thick ice clouds are around the 100% level. All of the retrievals mentioned above are for single-layer, vertically contiguous clouds. Greater errors are likely to be realized for multilayered, multiphase clouds (e.g., Minnis et al. 2007).

Finally, there is a potential for the retrieval algorithm to misclassify certain high concentrations of aerosols, generally dust, as clouds since the radiance signatures of both can be similar (Brennan et al. 2005). Evidence of this misclassification exists in western Oklahoma where CTP is retrieved in a region where no clouds are apparent on the visible imagery (Figs. 1 and 2b). A close examination of the visible data, along with meteorological knowledge of a dry atmosphere coupled with strong winds, indicates the presence of atmospheric dust. As a result, the “cloud properties” retrieved in this region are likely in error. For this research, dust-contaminated pixels are defined as those whose (10.8–13.3 μm) brightness temperature difference is less than 24 K and whose visible reflectance is below 0.18. These thresholds are based on GOES observations of a similar dust plume on 9 April 2009, which is described in detail by Jones and Christopher (2010). Pixels classified as dust are denoted in black in Fig. 2b for each event. Future cloud and aerosol retrieval algorithms based on GOES-R data will offer more robust methods to remove spurious data than the simple technique applied here.

4. WRF-DART characteristics

a. Run-time settings

The NWP model selected for this study is the ARW-WRF model version 3.3.1 using the double-moment Milbrandt–Yau cloud microphysics scheme (Skamarock et al. 2008; Milbrandt and Yau 2005). The ensemble consists of 40 members that use the same physical process parameterization schemes, yet have different initial and boundary conditions. A mesoscale analysis is generated using a 15-km domain covering most of the CONUS with 51 vertical levels stretching from the surface up to 50 hPa (Fig. 3). A parallel filter within the DART software assimilates atmospheric observations into the WRF using an ensemble adjustment Kalman filter (Anderson et al. 2009). Initial and boundary conditions are obtained from the 1200 UTC 10 May 2010 NAM model analysis and subsequent forecast fields at 3-h intervals. The background error covariances are estimated by the National Meteorological Center (NMC) method (Parrish and Derber 1992) using the WRF data assimilation software. These samples are then added to each ensemble member to account for uncertainties in the initial and boundary conditions (Torn et al. 2006). The mesoscale data assimilation cycle starts at 1200 UTC 10 May 2010 and continues for 9 h until 2100 UTC using conventional surface, marine, aircraft, and radiosonde observations with hourly forecasts generated thereafter. Additional description of the model characteristics used for the mesoscale assimilation can be found in Jones and Stensrud (2012).

Fig. 3.
Fig. 3.

Mesoscale (domain 1) and nested grid convective scale (domain 2) WRF domains used for both experiments.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

Using the hourly mesoscale analyses and forecasts as boundary conditions, a 3-km one-way nested domain is then generated over the area where the most intense convection occurs (Fig. 3). WRF model characteristics remain the same as for the mesoscale run with the exception that no cumulus parameterization is used. Conventional observations and GOES CWP retrievals are assimilated at 15-min intervals beginning at 1800 UTC and continuing until 2100 UTC. The error covariance for all CWP observations is assumed to be constant with a value of 0.05 kg m−2. The horizontal localization half-radius used by the covariance localization function is set to 100 km for conventional observations and 20 km for satellite-derived observations (Gaspari and Cohn 1999). A similar horizontal localization for satellite observations in cloudy regions was also applied by Vukicevic et al. (2004, 2006) with positive results. For conventional observations, a vertical localization half-radius of 4 km is applied. Since CWP is integrated over the entire atmospheric column, a vertical localization is not applicable to this variable. Adaptive covariance inflation is used in both data assimilation experiments. The predicted variables updated by the data assimilation scheme include the three wind components, perturbation temperature, perturbation geopotential, perturbation surface pressure of dry air, and potential temperature tendency due to microphysics as well as water vapor and hydrometeors. Also updated are the 10-m wind fields, 2-m temperature and water vapor fields, total surface pressure variables, and finally soil moisture, which are diagnosed by the surface and boundary layer schemes using state variables on the model grid.

A 3-h assimilation window is used to assimilate a reasonable time series of observations into the model, which results in the potential to assimilate up to 13 satellite scans. For this event, 1800 and 2100 UTC retrievals are not available; thus, the final number of satellite scans assimilated is 11. Defining the minimum number of scans to assimilate is based on results from several radar data assimilation experiments. These experiments indicate that approximately 10 radar scans are needed to generate an accurate representation of convection for a storm-scale analysis (Snyder and Zhang 2003; Xue et al. 2006; Caya et al. 2005). One difference between radar and satellite data is that the time interval between radar observations is generally 5 min or less, resulting in a total assimilation window less than an hour. Using a 3-h assimilation window at convection-allowing scales may further increase model error over this longer period.

Since variables directly related to the cloud microphysical properties are being assimilated, the selection of the microphysics scheme used in this study is important. The cloud properties generated by several schemes including Thompson, Morrison, Milbrandt–Yau, and the National Severe Storms Laboratory (NSSL) two-moment scheme are compared with satellite observations to determine which scheme best matched the general observed characteristics of thunderstorms (Thompson et al. 2004; Hong and Pan 1996; Milbrandt and Yau 2005; Mansell et al. 2010). Results indicate that the Milbrandt–Yau scheme generates the best representation of the observed ice properties from cirrus outflow compared to the other schemes tested (not shown), leading to the Milbrant–Yau scheme being selected.

b. CWP forward operator

Assimilating satellite retrieved cloud properties such as CTP and CWP directly into DART requires the development of a new forward operator, with the main challenge being to translate from a total column CWP value into a vertical distribution. A forward operator is created that integrates the model mixing ratios of cloud liquid water (QCLOUD), cloud ice (QICE), graupel (QGRUAP), rain (QRAIN), and snow (QSNOW) following the definition used by Otkin (2010). Both liquid and frozen variables are summed into a single CWP variable to better approximate the characteristics of the satellite retrieval. Recall that the GOES-retrieval algorithm only classifies a pixel as liquid or ice cloud based on the cloud height assumption. For mixed-phase clouds, both ice and water are present in the retrieved value of CWP.

The vertical extent of the cloud layer at each observation location is defined by CTP and CBP and is passed to the forward operator and used to constrain the CWP calculated from the model. For example, if CTP = 400 hPa and CBP = 800 hPa, then the mixing ratios of each cloud variable are only summed between these two levels. Cloud water and cloud ice outside this layer are ignored. Thus, the adjustment to model cloud microphysical properties relating to CWP where clouds exist will only occur within the vertical bounds of the cloud as defined by the satellite data. If this vertical constraint is not used, then the observed CWP would be compared with a CWP value calculated from all model levels. The end result of the assimilation would likely be to create a dry bias in the model analysis, since the forward operator implicitly assumes the satellite CWP would be spread across all model layers when in fact it should only be spread over a limited vertical extent. Assimilating clear-sky (CWP = 0) retrievals works in much the same way except that the vertical extent of the observation remains undefined. Thus, assimilating clear-sky regions should dry the analysis by reducing cloud water and ice mixing ratios if the model generates a cloud. Currently, the GOES-retrieval algorithm is not able to resolve multiple cloud layers, representing a possible source of error within the forward operator. Future algorithm enhancements (e.g., Chang et al. 2010a,b) should better resolve cloud layers and reduce the impact of this uncertainty. Once the summation of model variables is complete, the model CWP is compared with the satellite retrieved value and the EnKF makes the appropriate adjustments to the model fields to better correspond with observations. The CWP forward operator described here operates in much the same way as a total precipitable water (TPW) forward operator recently added to the DART software.

c. Assimilation experiments

Two experiments are conducted to test the viability of assimilating GOES CWP and the corresponding cloud layer information during the assimilation cycle of the 40-member 3-km ensemble (Table 1). Both experiments use the same 15-km mesoscale background as initial and boundary conditions for the 3-km ensemble. The first experiment only assimilates conventional observations between 1800 and 2100 UTC and represents the control experiment (CONV). The second assimilates CWP constrained by CTP and CBP in addition to the conventional observations (PATH) over the same 3-h time interval. After the final assimilation cycle at 2100 UTC, forecast fields are output at 5-min intervals out to 2230 UTC for each member using the hourly mesoscale forecasts as boundary conditions. The impact of assimilating the satellite data is then ascertained by comparing the ensemble mean analysis and forecast fields produced from these two experiments.

Table 1.

Observation types assimilated into each experiment.

Table 1.

5. Data assimilation

a. Assimilated observations

Between 1800 and 2100 UTC 10 May 2010, 43 367 observations are assimilated into WRF-DART for the PATH experiment with GOES CWP retrievals accounting for all but 2149 (5%) of that total (Table 2). The remainder originates from conventional observations made from the Automated Surface Observing System (ASOS), Aircraft Communication, Addressing, and Reporting System (ACARS), and radiosonde instruments. These are shown in Fig. 4a for the 3-h assimilation time window. Note the small number of observations and observation sites compared to those available from GOES CWP retrievals, which are shown for 2045 UTC in Fig. 4b. Comparing Fig. 4b with the visible imagery in Fig. 1, it is evident that CWP is being assimilated in both the clear-sky and cloudy regions providing good coverage within this domain. At this time, some form of cloud cover (CWP > 0) exists over large portions of the domain with 2670 cloudy observations being assimilated, whereas clear-sky (CWP = 0) observations only number 835. Cloudy observations primarily occur in eastern Oklahoma associated with the large low-level cloud field with a second area farther west ahead of the dryline where developing cumulus and ongoing convection is present. In between and behind the dryline, no clouds are detected resulting in CWP = 0 observations being assimilated. Fewer observations are assimilated in southwestern Oklahoma, as they are potentially dust contaminated and thus removed from the observation dataset (Figs. 2b and 4b).

Table 2.

Number of available NAv and assimilated NAssim convectional and CWP observations between 1800 and 2100 UTC 10 May 2010 for the PATH experiment. METAR: aviation routine weather report.

Table 2.
Fig. 4.
Fig. 4.

Location of observations assimilated into WRF-DART. (a) Conventional surface, aircraft, and radiosonde temperature observations between 1800 and 2100 UTC. Characteristics of humidity, wind, and pressure observations are similar. (b) GOES CWP observations at 2045 UTC with clear-sky (CWP = 0) and cloudy regions (CWP > 0) denoted. White areas indicate regions where no valid observations exist or are rejected by DART.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

b. Observation diagnostics

The large number of assimilated CWP observations suggests they should have a large impact on the model analysis fields compared to only assimilating conventional observations. Since the direct assimilation of cloud microphysical variables derived from satellite observations has not been previously undertaken within a WRF-DART framework, it is important to verify that these variables are being successfully assimilated and interpreted correctly. For each 15-min assimilation cycle between 1800 and 2100 UTC, the number, bias, root-mean-square difference (RMSD), and total spread (TSPRD) are computed between the CWP observations and the prior and posterior analysis fields. For each cycle, up to 4000 observations are being assimilated, decreasing slightly as a function of time as more dust contaminated pixels are withheld at later time periods. Bias between observations and both prior and posterior analysis fields is slightly negative (−0.02 kg m−2) initially, decreasing to near zero after a few assimilation cycles (Fig. 5). RMSD ranges between 0.1 and 0.7 kg m−2 between 1815 and 2030 UTC with the classic “sawtooth” pattern evident at each time interval (Fig. 5). RMSD gradually increases out to 2000 UTC before decreasing again for the following two cycles. A jump in both prior and posterior RMSD occurs at 2045 UTC with the prior value exceeding 0.9 kg m−2. The posterior RMSD is less than the prior RMSD at each cycle with reductions ranging from <0.1 kg m−2 early in the assimilation period to greater than 0.4 kg m−2 in the case of the final assimilation cycle that includes CWP data at 2045 UTC. The increase in RMSD at 2045 UTC can be partially attributed to a significant increase in convection within the domain after 2030 UTC. The number of high (>1.0 kg m−2) CWP values being assimilated increases nearly 40% from 113 to 152 between 2030 and 2045 UTC. As a result, any errors are likely to be magnified compared to previous assimilation cycle where the number of CWP > 1.0 kg m−2 observations is much lower. Total spread at each cycle closely follows that for RMSD, with values being confined between 0.05 and 0.3 kg m−2 and having the same sawtooth pattern (Fig. 5). The somewhat lower total spread compared to RMSD may indicate that the assumed observation error is slightly too small; however, no filter divergence is observed during the assimilation. This was determined by examining the error curves of several model variables, none of which showed rapid increases in error during the assimilation cycle.

Fig. 5.
Fig. 5.

Observation diagnostics for CWP showing model bias (circles), RMSD (squares), and TSPRD (triangles) between the prior and posterior analysis fields for each 15-min cycle starting at 1800 UTC and continuing until 2100 UTC. The number of observations assimilated during each cycle is also shown (diamonds).

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

The relationship between CWP calculated from model analyzed cloud microphysical variables and observed CWP is expected to be high as shown above. However, the effect on other variables not directly related to cloud microphysics is less clear. Thus, analysis increments of 700-hPa temperature (T700) and water vapor mixing ratio (Q700) are generated at the initial and final times satellite observations are assimilated (1815 and 2045 UTC). At 1815 UTC, differences in T700 are confined to the eastern portions of the domain corresponding to the areas of thicker stratus clouds (Fig. 6a). A small area of warming occurs in north-central Oklahoma coupled with some cooling farther east in Arkansas and Missouri. For the latter area, Q700 is generally greater indicating that assimilating the positive CWP values over this area result in a cooler, but more moist air mass (Fig. 6b). No significant differences are apparent in the western portion of the domain as deep convection has yet to fire in this region.

Fig. 6.
Fig. 6.

Analysis increments for (a) 700-hPa temperature (T700) and (b) water vapor mixing ratio (Q700) at 1815 UTC, which represents the first assimilation cycle satellite data are assimilated. (c),(d) As in (a),(b), but at 2045 UTC representing the final time satellite data are assimilated.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

The location of greatest temperature and water vapor mixing ratio differences shifts toward central Oklahoma associated with the area of developing convection (Figs. 6c,d). The largest differences in T700 (±2.5 K) occur in southern Kansas near the location of the ongoing convection where positive CWP values are being assimilated. Over much of this area, the posterior analysis is cooler than the preassimilation analysis. Smaller differences are apparent southwestward along the line of developing convection (Fig. 6c). The largest differences in Q700 correspond with the location of the temperature differences, though moistening or drying as a result of the assimilation does not always match the cooling or warming in the temperature fields (Fig. 6d). Other differences remain present in the northeastern corner of the domain in southwestern Missouri in the more stratiform region. The statistical evidence strongly indicates that assimilating the CWP observations correctly adjusts the analysis fields related to CWP to better fit the observations during each cycle. The following section describes how the adjusted fields compare with identical fields generated from only assimilating conventional observations.

6. Satellite versus WRF CWP

a. 1945 UTC

To visualize the effects of assimilating GOES CWP retrievals into WRF-DART, the ensemble mean CWP produced by the CONV and PATH experiments is compared with each other and the satellite retrievals. At 1945 UTC, which represents the eighth assimilation cycle, GOES CWP retrievals show evidence of developing convection along the border of Oklahoma and Kansas near 36.8°N, 98.5°W where CWP > 1.0 kg m−2 (Fig. 7a). Additional deep clouds are present farther north and west. In the remainder of central and southern Oklahoma and into north Texas, CWP is <0.05 kg m−2 indicating mostly clear conditions. Modest CWP values (0.3–0.5 kg m−2) exist over southeastern Kansas, northeastern Oklahoma, and northwestern Louisiana associated with the low-level stratus. A stark contrast exists between the observed CWP and that generated from the CONV experiment at the same time (Fig. 7b). The posterior ensemble mean CWP from CONV is often too aggressive at developing convection ahead of the dryline as evidenced by an area of CWP > 1 kg m−2 extending from the Oklahoma–Kansas border south into central and southern Oklahoma. Farther eastward in Arkansas, the analysis CWP is also greater in the stratiform cloud regions as well. The effect of assimilating CWP observations is apparent as CWP from PATH is significantly reduced in central and southern Oklahoma (Fig. 7c). Over this area, CWP = 0 observations are being assimilated, which causes WRF-DART to adjust its cloud water and cloud ice mixing ratios downward to generate a matching CWP. The magnitude of the decrease is evident in Fig. 8a, which displays the difference in CWP between the two experiments (PATH − CONV) at this time. A large area of negative values with magnitudes exceeding 0.25 kg m−2 is present in western Oklahoma where PATH suppresses convective development compared to CONV.

Fig. 7.
Fig. 7.

(a) GOES-13 CWP retrievals at 1945 UTC with corresponding posterior ensemble mean CWP generated from (b) CONV and (c) PATH model analyses at the same time. (d),(e),(f) As in (a),(b),(c), but for 2045 UTC. Black contours in (b),(c),(e),(f) represent 0.5 kg m−2 contour of GOES CWP shown in (a) and (c), respectively.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

Fig. 8.
Fig. 8.

Difference (PATH − CONV) in CWP at (a) 1945 and (b) 2045 UTC analysis times after satellite data have been assimilated. Blue regions indicate where PATH generates lower values of CWP whereas red regions indicate that PATH generates higher CWP values.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

Farther north, the coverage and spatial orientation of the convection near the Kansas–Oklahoma border (37°N, 99°W) is well captured by PATH but not by CONV. PATH is even able to generate convection where none exists in the CONV analysis northwest of the primary cell, matching well with the observations. Here, CWP values exceeding 1.0 kg m−2 are being assimilated where the CONV model analysis generates no significant cloud cover. The corresponding adjustment in WRF-DART is to increase the cloud water and cloud ice concentrations to produce a CWP that is consistent with the satellite observations. The magnitude of this increase exceeds 0.25 kg m−2 north of 37°N where CONV does not generate convection or displaces it too far east (Fig. 8a).

Farther eastward in the stratiform cloud regions, CWP values generated by PATH are generally smaller than for CONV in the northeastern portion of the domain and somewhat larger in the southeastern. However, the magnitude of these differences is generally smaller than present farther west in the convective region of the domain. Assimilating CWP in the PATH experiment clearly produces a more physically realistic representation of CWP than present in CONV midway through the assimilation cycle, which indicates further improvements should be evident by the final assimilation cycle.

b. 2045 UTC

The final cycle where GOES CWP observations are being assimilated occurs at 2045 UTC since no satellite data are available for 2100 UTC. By this time, the convection along and north of the Oklahoma–Kansas border has increased significantly in coverage and intensity compared to an hour earlier at 1945 UTC (Figs. 7a,d). The largest cell is centered near 37.0°N, 97.5°W, with additional clouds and precipitation (CWP > 1.0 kg m−2) farther north. Initial development ahead of the dryline into Oklahoma is also beginning to occur by this time. Similar to results from 1945 UTC, CONV is too aggressive in developing and sustaining convection along the dryline compared to observations (Fig. 7e). In Kansas, where convection is ongoing, its placement in the CONV model analysis is wrong, being too far west.

In the PATH analysis at 2045 UTC, the magnitude and coverage of CWP is much lower along the line of developing convection, which is in much better agreement with the satellite observations (Fig. 7f). The maxima in CWP in central and southern Oklahoma are displaced slightly westward of their location in CONV, indicating that PATH might be somewhat slower in its progression of its convective features. Figure 8b shows this decrease as a large area in central Oklahoma where CWP from the PATH experiment is less than CWP from the CONV experiment by more than 0.25 kg m−2. Overall, PATH is much less aggressive in developing convection during the analysis period as it is suppressed at each assimilation cycle. The decrease in analyzed CWP compared to CONV is a direct result of assimilating zero values of CWP in this experiment. Additional experiments assimilating either zero or positive values of CWP (but not both) support this conclusion. When assimilating only zero values, the modeled convection is properly suppressed, but little change is apparent where convection is ongoing. Similarly, assimilating only positive values of CWP results in analyzed CWP values similar to the CONV experiment in central and southern Oklahoma, while only providing modest improvement where convection was ongoing (not shown). Similar results were found by Vukicevic et al. (2004, 2006), Polkinghorne et al. (2010), and Polkinghorne and Vukicevic (2011) when assimilating cloudy radiances on similar horizontal scales. Assimilating CWP in clear-sky regions is also comparable to assimilating clear-air radar reflectivity (Tong and Xue 2005; Aksoy et al. 2009). Both provide the model information that no clouds are present in certain areas and act to suppress spurious precipitation that may be generated by the model during the assimilation window.

In Kansas, assimilating CWP observations produces a more accurate depiction of the ongoing convection near 37.0°N, 97.5°W from the PATH experiment compared to the CONV (Fig. 7f). Where the satellite observations indicate convection, CWP generated from PATH increases by more than 0.25 kg m−2 compared to CONV. In addition, CWP from PATH is lower farther west where CONV is too aggressive with its convection in Kansas compared to the observations at this time. Thus, assimilating the satellite observations provides both a more accurate depiction of the ongoing convection while at the same time suppressing spurious convection. Additional differences exist in the eastern portion of the domain with PATH generally having somewhat higher CWP values than CONV. However, these differences are generally smaller than in the more convective regions of the domain.

7. Verification against Mesonet data

a. Downward shortwave flux at 2100 UTC

The availability of high spatial and temporal resolution downwelling shortwave flux (SWF) measurements from the Oklahoma Mesonet allows for an independent and quantitative verification of the differences in cloud properties between each model experiment. SWF can be interpreted as a measure of cloud coverage and thickness over a particular location. In clear-sky environments, the maximum possible SWF reaches the surface, whereas increasing cloud cover decreases this amount. If the WRF ensemble is characterizing clouds correctly, then SWF from Mesonet observations and the model analysis should be in good agreement. Where there are large differences, it signals that either the model generates a cloud that does not exist, or the model fails to generate a cloud that does exist. Mesonet SWF observations indicate extensive cloud cover in northern Oklahoma with SWF values generally <400 W m−2 (Fig. 8). Some sites in northern Oklahoma appear almost completely cloud covered with SWF < 200 W m−2 (Table 3). In the southeastern portion of the domain, SWF varies from site to site, which is representative of the wavelike cloud features present over this region (Fig. 1). In the central and far western portions of the domain, SWF values generally exceed 700 W m−2 indicating mostly clear skies (Fig. 9a). The exception occurs in a north–south line near the western edge of the domain boundary corresponding to the cloud field where convection is developing and SWF ranges from 590 to 134 W m−2 (Table 3).

Table 3.

Oklahoma Mesonet SWF (W m−2) at selected sites along with corresponding CONV and PATH ensemble mean SWF at 2100 UTC.

Table 3.
Fig. 9.
Fig. 9.

(a) Contour plot of Oklahoma Mesonet SWF at 2100 UTC with corresponding Mesonet locations and wind barbs overlaid (white) with short barbs = 5 m s−1 and long barbs = 10 m s−1. Posterior ensemble mean SWF generated from (b) CONV and (c) PATH at 2100 UTC with Mesonet values overlaid along with model wind barbs (black). The color contrast between Mesonet and model SWF indicates where differences between the two are the greatest.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

Comparing the SWF observations with values from the CONV ensemble mean reveals several interesting differences (Fig. 9b). The most obvious is the low SWF values from southwest into south-central Oklahoma (SWF < 400 W m−2) associated with spurious model-generated convection (Fig. 7). CONV is also somewhat too far west with the convection in far northern Oklahoma near 98°W where Mesonet SWF is >750 W m−2 corresponding to near-clear-sky values compared to CONV values below 250 W m−2 (Table 3). Smaller differences exist over the eastern portion of the domain where CONV appears to again underestimate SWF in east-central regions, but overestimate SWF farther in the northeast. The magnitude of these differences is generally on the order of 100 W m−2 (Table 3). Differences also exist in southeastern Oklahoma, though no consistent pattern of under or over analyzing of SWF is apparent. Calculating the difference between Mesonet and ensemble mean CONV SWF results in a bias (CONV − Mesonet) of −48.7 W m−2 for the 75 sites in Fig. 9 with a corresponding RMSD of 240.7 W m−2. The negative bias indicates that CONV is underpredicting SWF compared to ground truth observations. The low bias in SWF is a result of CONV developing convection in western Oklahoma too quickly, generating more and thicker clouds, which is consistent with the CWP plot from CONV shown in Fig. 7b.

Assimilating CWP results in large changes to SWF with PATH generating higher SWF values over several regions as compared to CONV (Fig. 9c). In the southwestern portion of the domain, PATH generates a large area of SWF > 750 W m−2 in a region where SWF from CONV is several hundred watts per meters squared lower. Mesonet stations in this region all indicate clear or mostly clear conditions and report high SWF values. The PATH analysis is in much better agreement with these Mesonet observations compared to CONV, with increases in SWF by up to 500 W m−2 at many locations (Table 3). Farther north, CONV has already developed strong convection resulting in a large area of SWF < 200 W m−2. In PATH, the spatial extent of the convection and resulting cloud cover decreases, but some remains. Mesonet observations of SWF are highly variable in this region, which is a sign of the nonuniform cloud field present. PATH appears to capture this variability more realistically though the skill at a few sites may be reduced owing in part to placement errors. Overall, however, PATH produces much better agreement with the Mesonet SWF observations with an RMSD = 165.5 W m−2 and a smaller, but now positive, bias of 39.0 W m−2.

b. Time series

The improvement in the ensemble mean analyses from assimilating CWP is not limited to a single time at 2100 UTC nor is it only evident in the SWF fields. Comparisons during both the data assimilation and forecast periods for SWF and 2-m temperature show that the assimilation of CWP leads to better agreement with observations (Fig. 10). During the final three assimilation cycles between 2030 and 2100 UTC, RMSD from CONV ranges between 50 and 80 W m−2 higher than the corresponding RMSD from PATH (Fig. 10a). Slightly larger differences are apparent in the bias where CONV is more than 50 W m−2 to low with PATH ranging between 0 and 40 W m−2 too high. Once the assimilation of CWP ends and the free forecast begins, RMSD for both CONV and PATH increases rapidly in the first 15 min. This is a result of the rapid increase in deep convection (high CWP values) in both the model forecasts and observations. The difference in RMSD between CONV and PATH decreases as a function of time becoming negligible 30 min into the forecast at 2130 UTC. This indicates that the increase in accuracy of the SWF fields imparted by assimilating satellite CWP retrievals decreases once the assimilation stops. Similar results have been observed during radar reflectivity assimilation where its effect on the model-simulated reflectivity quickly decreases as a function of time (Snyder and Zhang 2003). Following 2130 UTC, the differences between CONV and PATH are small and continue to decrease as the cirrus shield from the ongoing convection increases in size, uniformly reducing SWF over the eastern half of the domain. The bias is roughly constant between 2100 and 2130 UTC with PATH having the lower absolute value by approximately 10 W m−2. After 2130 UTC, biases for CONV and PATH slowly converge to zero at 2230 UTC. For most of this period the absolute value of the bias is lower for PATH as compared to CONV indicating that the former continues to generate less and/or thinner cloud cover over much of the domain.

Fig. 10.
Fig. 10.

Time series of bias and RMSD for (a) SWF and (b) 2-m temperature between 2030 and 2230 UTC calculated between Mesonet observations and each experiment's analysis at each time step.

Citation: Monthly Weather Review 141, 7; 10.1175/MWR-D-12-00238.1

While the improvement in SWF RMSD appears to become small 30 min into the forecast, its impact on temperature remains for a much longer period of time (Fig. 10b). Changes in temperature fields have a direct impact on convective instability parameters; thus, they have important implications to downstream convective development and evolution. The RMSD between each experiment and Mesonet 2-m temperature shows that PATH consistently has a lower RMSD compared to CONV. Starting at 2030 UTC and continuing into the free forecast period, the difference is approximately 0.5 K and slowly decreases to 0.2 K at 2230 UTC. Bias differs by a similar magnitude with PATH being approximately 0.5 K warmer for the entire period. Thus, it can be said that PATH is consistently warmer at the surface compared to CONV, which is in better agreement with the observations. The primary reason for the warmer temperatures is the greater amount of solar radiation reaching the surface early in the assimilation cycle. Recall that CONV is too aggressive at developing convection in Oklahoma, leading to a reduction in SWF reaching the surface, thereby reducing surface temperature. The response in the temperature fields from changes in SWF is not instantaneous, but rather accumulates over time. Thus, temperature differences are likely to remain apparent for a longer period of time compared to a more rapidly changing parameter such as SWF. The reduction in temperature bias and RMSD over the entire 90-min forecast period strongly indicates assimilating CWP has a lasting impact on the ensemble forecasts.

8. Conclusions

With the upcoming launch of GOES-R and its Advanced Baseline Imager (ABI), retrieved cloud properties are scheduled to become available after 2015 (http://www.goes-r.gov/resources/faqs.html) at spatial and temporal resolutions comparable to that available from operational radar data (Schmit et al. 2005). Furthermore, the increased spectral resolution of the GOES-R ABI allows for more accurate retrievals that will act to reduce potential uncertainties resulting from some of the assumptions made when creating the CWP forward operator for the current generation GOES retrievals. Yet despite these uncertainties, the results from this study strongly indicate that the assimilation of CWP is having a positive impact on the ensemble mean analyses and forecasts. During the assimilation period, the PATH experiment consistently produces a better representation of the CWP field compared to the conventional data-only experiment. These results are consistent with previous research by Vukicevic et al. (2004, 2006), Polkinghorne et al. (2010), and Polkinghorne and Vukicevic (2011) who also showed that assimilating high-resolution GOES satellite observations improved the representation of cloud properties within a model analysis.

The improved characterization of clouds and convection found in this research produces a more accurate SWF analysis by the end of the assimilation at 2100 UTC when verified against Mesonet observations. This improvement decreases during the forecast period as the effects of assimilating specific cloud characteristics decreases as forecast time increases. However, the impact of assimilating CWP on downstream surface temperature remains evident 90 min into the forecast as the lower amount of convection generated in the PATH experiment early on leads to more solar radiation reaching the surface and increasing the 2-m temperature. These warmer temperatures remain even after both experiments have generated large areas of convection and associated cirrus outflow.

Given these positive results, further investigation of assimilating GOES satellite retrievals is underway. GOES VISST retrievals are available in near–real time (Minnis et al. 2008a) and are continually being upgraded as new techniques are added to the analyses. For example, multilayered cloud retrievals are becoming operational for optically thin cirrus clouds over low-level water clouds (Chang et al. 2010a,b). Methods for increasing the range of retrieved CWP at night are also being explored (Minnis et al. 2012) that will allow for more reliable assimilation over the entire diurnal cycle. Techniques for estimating the cloud water content profiles based on the cloud temperature and CWP are also being evaluated (Smith et al. 2010). Yet, several important questions remain to be answered before these data can be assimilated in real time. These questions include analyzing the sensitivity of cloud microphysical schemes on assimilating CWP, determining the best localization and error variance characteristics to use, and improving the representation of multiple cloud layers in the forward operator (e.g., Polkinghorne et al. 2010). Finally, the relative advantages and disadvantages of assimilating satellite retrievals along with other high-resolution datasets such as radar reflectivity and velocity must also be ascertained. The eventual goal is to maximize the use of both satellite and radar observations to generate an accurate convective-scale model analysis and improved very short-range forecasts of convection.

Acknowledgments

We appreciate the two anonymous reviewers whose comments improved the quality of this work. Weather Surveillance Radar-1988 Doppler (WSR-88D) level 2 radar reflectivity were retrieved from National Climatic Data Center archives. Mesonet data were kindly provided by the Oklahoma Climatological Survey. Nancy Collins at the University Corporation for Atmospheric Research provided invaluable assistance in helping debug certain parts of the new CWP forward operator. This research was supported by the NOAA/National Environmental Satellite, Data, and Information Service. Partial funding for this research was also provided by NOAA/Office of Oceanic and Atmospheric Research under NOAA–University of Oklahoma Cooperative Agreement NA17RJ1227, under the U.S. Department of Commerce. P. Minnis and R. Palikonda are supported by the NASA Modeling, Analysis, and Prediction (MAP) Program, the GOES-R Program, and by the Department of Energy Atmospheric Science Research Program under Interagency Agreement DE-SC0000991/003. (The near-real-time satellite analyses can be accessed for a variety of domains at http://angler.larc.nasa.gov/.)

REFERENCES

  • Aksoy, A., , D. Dowell, , and C. Snyder, 2009: A multicase comparative assessment of the ensemble Kalman filter for assimilation of radar observations. Part I: Storm-scale analyses. Mon. Wea. Rev., 137, 18051824.

    • Search Google Scholar
    • Export Citation
  • Anderson, J. L., , T. Hoar, , K. Raeder, , H. Liu, , N. Collins, , R. Torn, , and A. Avellano, 2009: The Data Assimilation Research Testbed: A community data assimilation facility. Bull. Amer. Meteor. Soc., 90, 12831296.

    • Search Google Scholar
    • Export Citation
  • Benedetti, A., , and M. Janiskova, 2008: Assimilation of MODIS cloud optical depths in the ECMWF model. Mon. Wea. Rev., 136, 17271746.

  • Brennan, J. I., , Y. J. Kaufman, , I. Koren, , and R. R. Li, 2005: Aerosol-cloud interaction-misclassification of MODIS clouds in heavy aerosol. IEEE Trans. Geosci. Remote Sens., 43, 911915.

    • Search Google Scholar
    • Export Citation
  • Caya, A., , J. Sun, , and C. Snyder, 2005: A comparison between the 4DVAR and ensemble Kalman filter techniques for radar data assimilation. Mon. Wea. Rev., 133, 30813094.

    • Search Google Scholar
    • Export Citation
  • Chang, F.-L., , P. Minnis, , B. Lin, , M. Khaiyer, , R. Palikonda, , and D. Spangenberg, 2010a: A modified method for inferring cloud top height using GOES 12 imager 10.7 and 13.3-μm data. J. Geophys. Res., 115, D06208, doi:10.1029/2009JD012304.

    • Search Google Scholar
    • Export Citation
  • Chang, F.-L., , P. Minnis, , S. Sun-Mack, , L. Nyugen, and Y. Chen, 2010b: On the satellite determination of multi-layered multi-phase cloud properties. Preprints, 13th Conf. on Cloud Physics/13th Conf. on Atmospheric Radiation, Portland, OR, Amer. Meteor. Soc., JP1.10. [Available online at https://ams.confex.com/ams/pdfpapers/171180.pdf.]

  • Chen, Y., , F. Weng, , Y. Han, , and Q. Liu, 2008: Validation of the Community Radiative Transfer Model by using CloudSat data. J. Geophys. Res., 113, D00A03, doi:10.1029/2007JD009561.

    • Search Google Scholar
    • Export Citation
  • Derber, J. C., , and W.-S. Wu, 1998: The use of TOVS cloud-cleared radiances in the NCEP SSI analysis system. Mon. Wea. Rev., 126, 22872299.

    • Search Google Scholar
    • Export Citation
  • Dong, X., , P. Minnis, , G. G. Mace, , W. L. Smith Jr., , M. Poellot, , R. T. Marchand, , and A. D. Rapp, 2002: Comparison of stratus cloud properties deduced from surface, GOES, and aircraft data during the March 2000 ARM Cloud IOP. J. Atmos. Sci., 59, 32563284.

    • Search Google Scholar
    • Export Citation
  • Dong, X., , P. Minnis, , B. Xi, , S. Sun-Mack, , and Y. Chen, 2008: Comparison of CERES-MODIS stratus cloud properties with ground-based measurements at the DOE ARM Southern Great Plains site. J. Geophys. Res., 113, D03204, doi:10.1029/2007JD008438.

    • Search Google Scholar
    • Export Citation
  • Dowell, D. C., , G. S. Romine, , and C. Snider, 2010: Ensemble storm-scale data assimilation and prediction for severe convective storms. Preprints, 25th Conf. on Severe Local Storms, Denver, CO, Amer. Meteor. Soc., 9.5. [Available online at https://ams.confex.com/ams/pdfpapers/176121.pdf.]

  • Errico, R. M., 2000: The dynamical balance of leading singular vectors in a primitive-equation model. Quart. J. Roy. Meteor. Soc., 126, 16011618.

    • Search Google Scholar
    • Export Citation
  • Errico, R. M., , G. Ohring, , F. Weng, , P. Bauer, , B. Ferrier, , J.-F. Mahfouf, , and J. Turk, 2007: Assimilation of satellite cloud and precipitation observations in numerical weather prediction models: Introduction to the JAS special collection. J. Atmos. Sci., 64, 37373741.

    • Search Google Scholar
    • Export Citation
  • Gaspari, G., , and S. E. Cohn, 1999: Construction of correlation functions in two and three dimensions. Quart. J. Roy. Meteor. Soc., 125, 723757.

    • Search Google Scholar
    • Export Citation
  • Heemink, A. W., , M. Verlaan, , and A. J. Segers, 2001: Variance reduced ensemble Kalman filtering. Mon. Wea. Rev., 129, 17181728.

  • Hong, S.-Y., , and H.-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124, 23222339.

    • Search Google Scholar
    • Export Citation
  • Jones, T. A., , and S. A. Christopher, 2010: Satellite and radar remote sensing of southern Plains grass fires: A case study. J. Appl. Meteor. Climatol., 49, 21332146.

    • Search Google Scholar
    • Export Citation
  • Jones, T. A., , and D. J. Stensrud, 2012: Assimilating AIRS temperature and mixing ratio profiles using an ensemble Kalman filter approach for convective-scale forecasts. Wea. Forecasting, 27, 541564.

    • Search Google Scholar
    • Export Citation
  • Kalman, R. E., 1960: A new approach to linear filtering and prediction problems. J. Basic Eng., 82, 3545.

  • Kato, S., , L. M. Hinkelman, , and A. Cheng, 2006: Estimate of satellite-derived cloud optical thickness and effective radius errors and their effect on computed domain-averaged irradiances. J. Geophys. Res., 111, D17201, doi:10.1029/2005JD006668.

    • Search Google Scholar
    • Export Citation
  • Lauwaet, D., , K. D. Ridder, , and P. Pandey, 2011: Assimilating remotely sensed cloud optical thickness into a mesoscale model. Atmos. Chem. Phys. Discuss., 11, 13 35513 380.

    • Search Google Scholar
    • Export Citation
  • Lin, Y., , G. DiMego, , B. Ferrier, , J. Jung, , D. Keyser, , and E. Rodgers, 2003: Assimilation of GOES cloud top pressure data into the Eta Model. JSC/CAS Working Group on Numerical Experimentation (WGNE) Blue Book 2003, 2 pp. [Available online at http://www.wcrp-climate.org/WGNE/BlueBook/2003/individual-articles/01-lin.pdf.]

  • Mace, G. G., , Y. Zhang, , S. Platnick, , M. D. King, , P. Minnis, , and P. Yang, 2005: Evaluation of cirrus cloud properties from MODIS radiances using cloud properties derived from ground-based data collected at the ARM SGP site. J. Appl. Meteor., 44, 221240.

    • Search Google Scholar
    • Export Citation
  • Mansell, E. R., , C. L. Ziegler, , and E. C. Bruning, 2010: Simulated electrification of a small thunderstorm with two-moment bulk microphysics. J. Atmos. Sci., 67, 171194.

    • Search Google Scholar
    • Export Citation
  • McNally, A. P., , J. C. Derber, , W.-S. Wu, , and B. B. Katz, 2000: The use of TOVS level-1B radiances in the NCEP SSI analysis system. Quart. J. Roy. Meteor. Soc., 126, 689724.

    • Search Google Scholar
    • Export Citation
  • McNally, A. P., , P. D. Watts, , J. A. Smith, , R. Engelen, , G. A. Kelly, , J.-N. Thépaut, , and M. Matricard, 2006: The assimilation of AIRS radiance data at ECMWF. Quart. J. Roy. Meteor. Soc., 132, 935957.

    • Search Google Scholar
    • Export Citation
  • McPherson, R. A., and Coauthors, 2007: Statewide monitoring of the mesoscale environment: A technical update on the Oklahoma Mesonet. J. Atmos. Oceanic Technol., 24, 301321.

    • Search Google Scholar
    • Export Citation
  • Mecikalski, J. R., , P. Minnis, , and R. Palikonda, 2013: Use of satellite derived cloud properties to quantify growing cumulus beneath cirrus clouds. Atmos. Res., 120–121, 192201.

    • Search Google Scholar
    • Export Citation
  • Menzel, W. P., , and J. F. Purdom, 1994: Introducing GOES-I: The first of a new generation of geostationary operational environmental satellites. Bull. Amer. Meteor. Soc., 75, 757781.

    • Search Google Scholar
    • Export Citation
  • Migliorini, S., 2012: On the equivalence between radiance and retrieval assimilation. Mon. Wea. Rev., 140, 258265.

  • Milbrandt, J. A., , and M. K. Yau, 2005: A multimoment bulk microphysics parameterization. Part I: Analysis of the role of the spectral shape parameter. J. Atmos. Sci., 62, 30513064.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , J. Huang, , B. Lin, , Y. Yi, , R. F. Arduini, , T.-F. Fan, , J. K. Ayers, , and G. G. Mace, 2007: Ice cloud properties in ice-over-water cloud systems using Tropical Rainfall Measuring Mission (TRMM) visible and infrared scanner and TRMM Microwave Imager data. J. Geophys. Res., 112, D06206, doi:10.1029/2006JD007626.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2008a: Near-real time cloud retrievals from operational and research meteorological satellites. Proc. SPIE Europe Remote Sens., 7107-2, Cardiff, Wales, SPIE, 1–8.

  • Minnis, P., and Coauthors, 2008b: Cloud detection in non-polar regions for CERES using TRMM VIRS and Terra and Aqua MODIS data. IEEE Trans. Geosci. Remote Sens., 46, 38573884.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2010: CERES Edition 3 cloud retrievals. Preprints, 13th Conf. on Atmospheric Radiation, Portland, OR, Amer. Meteor. Soc., 5.4. [Available online at https://ams.confex.com/ams/pdfpapers/171366.pdf.]

  • Minnis, P., and Coauthors, 2011: CERES Edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data. Part I: Algorithms. IEEE Trans. Geosci. Remote Sens., 49, 43744400.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2012: Simulations of infrared radiances over a deep convective cloud system observed during TC4: Potential for enhancing nocturnal ice cloud retrievals. Remote Sens., 4, 30223054.

    • Search Google Scholar
    • Export Citation
  • Otkin, J. A., 2010: Clear and cloudy sky infrared brightness temperature assimilation using an ensemble Kalman filter. J. Geophys. Res., 115, D19207, doi:10.1029/2009JD013759.

    • Search Google Scholar
    • Export Citation
  • Painemal, D., , P. Minnis, , J. K. Ayers, , and L. O'Neill, 2012: GOES-10 microphysical retrievals in marine warm clouds: Multi-instrument validation and daytime cycle over the southeast Pacific. J. Geophys. Res., 117, D19212, doi:10.1029/2012JD017822.

    • Search Google Scholar
    • Export Citation
  • Parrish, D. F., , and J. C. Derber, 1992: The National Meteorological Center's spectral statistical interpolation analysis system. Mon. Wea. Rev., 120, 17471763.

    • Search Google Scholar
    • Export Citation
  • Pavelin, E., , S. English, , and J. Eyre, 2008: The assimilation of cloud-affected infrared satellite radiances for numerical weather prediction. Quart. J. Roy. Meteor. Soc., 134, 737749.

    • Search Google Scholar
    • Export Citation
  • Pincus, R., , R. J. P. Hofmann, , J. L. Anderson, , K. Raeder, , N. Collins, , and J. S. Whitaker, 2011: Can fully accounting for clouds in data assimilation improve short-term forecast by global models? Mon. Wea. Rev., 139, 946956.

    • Search Google Scholar
    • Export Citation
  • Polkinghorne, R., , and T. Vukicevic, 2011: Data assimilation of cloud affected radiances in a cloud resolving model. Mon. Wea. Rev., 139, 755773.

    • Search Google Scholar
    • Export Citation
  • Polkinghorne, R., , T. Vukicevic, , and F. Evans, 2010: Validation of cloud resolving model background data for data assimilation. Mon. Wea. Rev., 138, 781795.

    • Search Google Scholar
    • Export Citation
  • Schmit, T. J., , E. M. Prins, , A. J. Schreiner, , and J. J. Gurka, 2001: Introducing the GOES-M imager. Natl. Wea. Dig., 25, 2837.

  • Schmit, T. J., , M. M. Gunshor, , W. P. Menzel, , J. J. Gurka, , J. Li, , and A. S. Bachmeier, 2005: Introducing the next-generation Advanced Baseline Imager on GOES-R. Bull. Amer. Meteor. Soc., 86, 10791096.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO 80307.]

  • Smith, W. L., Jr., , P. Minnis, , L. Nguyen, , D. Spangenberg, , R. Austin, , and G. Mace, 2007: Preliminary comparison of CloudSat liquid water path to passive satellite estimates. A-Train Symp., Lille, France, NASA. [Available online at http://www-angler.larc.nasa.gov/site/doc-library/52-Smith.Lille.07.pdf.]

  • Smith, W. L., Jr., , P. Minnis, , H. Finney, , R. Palikonda, , and M. M. Khaiyer, 2008: An evaluation of operational GOES-derived single-layer cloud top heights with ARSCL data over the ARM Southern Great Plains site. Geophys. Res. Lett., 35, L13820, doi:10.1029/2008GL034275.

    • Search Google Scholar
    • Export Citation
  • Smith, W. L., Jr., , P. Minnis, , S. G. Benjamin, , and S. S. Weygandt, 2010: 4-D water content fields derived from operational satellite data. Proc. IEEE 2010 Int. Geoscience and Remote Sensing Symp., Honolulu, HI, IEEE, 308–311.

  • Snyder, C., , and F. Zhang, 2003: Assimilation of simulated Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 131, 16631677.

    • Search Google Scholar
    • Export Citation
  • Stein, T. H. M., , J. Delanöe, , and R. J. Hogan, 2011: A comparison among four different methods for ice-cloud properties using data from CloudSat, CALIPSO, and MODIS. J. Appl. Meteor. Climatol., 50, 19521969.

    • Search Google Scholar
    • Export Citation
  • Thompson, G., , R. M. Rasmussen, , and K. Manning, 2004: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon. Wea. Rev., 132, 519542.

    • Search Google Scholar
    • Export Citation
  • Tong, M., , and M. Xue, 2005: Ensemble Kalman filter assimilation of Doppler radar data with compressible nonhydrostatic model: OSS experiments. Mon. Wea. Rev., 133, 17891807.

    • Search Google Scholar
    • Export Citation
  • Torn, R. D., , G. J. Hakim, , and C. Snyder, 2006: Boundary conditions for limited area ensemble Kalman filters. Mon. Wea. Rev., 134, 24902502.

    • Search Google Scholar
    • Export Citation
  • Vukicevic, T., , T. Greenwald, , M. Zupanski, , D. Zupanski, , T. Vonder Haar, , and A. Jones, 2004: Mesoscale cloud state estimation from visible and infrared satellite radiances. Mon. Wea. Rev., 132, 30663077.

    • Search Google Scholar
    • Export Citation
  • Vukicevic, T., , M. Sengupta, , A. Jones, , and T. Vonder Haar, 2006: Cloud resolving data assimilation: Information content of IR window observations and uncertainties in estimation. J. Atmos. Sci., 63, 901919.

    • Search Google Scholar
    • Export Citation
  • Waliser, D., and Coauthors, 2009: Cloud ice: A climate model challenge with signs and expectations of progress. J. Geophys. Res., 114, D00A21, doi:10.1029/2008JD010015.

    • Search Google Scholar
    • Export Citation
  • Weisz, E., , J. Li, , J. Li, , D. K. Zhou, , H.-L. Huang, , M. D. Goldberg, , and P. Yang, 2007: Cloudy sounding and cloud top height retrieval from AIRS alone single field-of-view radiance measurements. Geophys. Res. Lett., 34, L12802, doi:10.1029/2007GL030219.

    • Search Google Scholar
    • Export Citation
  • Xue, M., , M. Tong, , and K. K. Droegemeier, 2006: An OSSE framework based on the ensemble square root Kalman filter for evaluating the impact of data from radar networks on thunderstorm analysis and forecasting. J. Atmos. Oceanic Technol., 23, 4666.

    • Search Google Scholar
    • Export Citation
  • Yucel, I., , W. J. Shuttleworth, , R. T. Pinker, , L. Lu, , and S. Sorooshian, 2002: Impact of ingesting satellite-derived cloud cover into the Regional Atmospheric Modeling System. Mon. Wea. Rev., 130, 610628.

    • Search Google Scholar
    • Export Citation
  • Yucel, I., , W. J. Shuttleworth, , X. Gao, , and S. Sorooshian, 2003: Short-term performance of MM5 with cloud-cover assimilation from satellite observations. Mon. Wea. Rev., 131, 17971810.

    • Search Google Scholar
    • Export Citation
Save