• Ahn, M. H., , Sohn E. H. , , and Hwang B. J. , 2003: A new algorithm for sea fog/stratus detection using GMS-5 IR data. Adv. Atmos. Sci., 20, 899913.

    • Search Google Scholar
    • Export Citation
  • Anderson, J., 1931: Observations from airplanes of cloud and fog conditions along the southern California coast. Mon. Wea. Rev., 59, 264270.

    • Search Google Scholar
    • Export Citation
  • Ballard, S. P., , Golding B. W. , , and Smith R. N. B. , 1991: Mesoscale model experimental forecasts of the haar of northeast Scotland. Mon. Wea. Rev., 119, 21072123.

    • Search Google Scholar
    • Export Citation
  • Bendix, J., , Thies B. , , Cermak J. , , and Nauß T. , 2005: Ground fog detection from space based on MODIS daytime data—A feasibility study. Wea. Forecasting, 20, 989996.

    • Search Google Scholar
    • Export Citation
  • Byers, H. R., 1930: Summer Sea Fogs of the Central California Coast. University of California Press, 338 pp.

  • Chen, F., , and Dudhia J. , 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model description and implementation. Mon. Wea. Rev., 129, 569585.

    • Search Google Scholar
    • Export Citation
  • Cho, Y. K., , Kim M. O. , , and Kim B. C. , 2000: Sea fog around the Korean Peninsula. J. Appl. Meteor., 39, 24732479.

  • Coakley, J. A., , and Bretherton F. P. , 1982: Cloud cover from high-resolution scanner data: Detection and allowing for partially filled fields of view. J. Geophys. Res., 87 (C7), 49174932.

    • Search Google Scholar
    • Export Citation
  • Deng, J., , Bai J. , , Liu J. W. , , Wang X. W. , , and Shi H. Y. , 2006: Detection of daytime fog using MODIS multispectral data (in Chinese). Mater. Sci. Technol., 34, 188193.

    • Search Google Scholar
    • Export Citation
  • Ellrod, G. P., 1995: Advances in the detection and analysis of fog at night using GOES multispectral infrared imagery. Wea. Forecasting, 10, 606619.

    • Search Google Scholar
    • Export Citation
  • Findlater, J., , Roach W. T. , , and McHugh B. C. , 1989: The haar of north-east Scotland. Quart. J. Roy. Meteor. Soc., 115, 581608.

  • Fitzpatrick, M. F., , Brandt R. E. , , and Warren S. G. , 2004: Transmission of solar radiation by clouds over snow and ice surfaces: A parameterization in terms of optical depth, solar zenith angle, and surface albedo. J. Climate, 17, 266275.

    • Search Google Scholar
    • Export Citation
  • Fu, G., , Zhang M. G. , , Duan Y. H. , , Zhang T. , , and Wang J. Q. , 2004: Characteristics of sea fog over the Yellow Sea and the East China Sea. Kaiyo Mon.,38, 99–108.

  • Fu, G., , Guo J. , , Xie S. P. , , Duan Y. , , and Zhang M. , 2006: Analysis and high-resolution modeling of a dense sea fog event over the Yellow Sea. Atmos. Res., 81, 292303.

    • Search Google Scholar
    • Export Citation
  • Fu, G., , Li P. Y. , , Crompton J. G. , , Guo J. , , Gao S. H. , , and Zhang S. P. , 2010: An observational and modeling study of a sea fog event over the Yellow Sea on 1 August 2003. Meteor. Atmos. Phys., 107, 149159.

    • Search Google Scholar
    • Export Citation
  • Fu, G., , Xu J. , , and Zhang S. Q. , 2011: Comparison of modeling atmospheric visibility with visible satellite imagery (in Chinese). J. Ocean Univ. China,41, 001–010.

  • Fu, G., , Zhang S. P. , , Gao S. H. , , and Li P. Y. , 2012: Understanding of Sea Fog over the China Seas. China Meteorological Press, 220 pp.

  • Gao, S. H., , Lin H. , , Shen B. , , and Fu G. , 2007: A heavy sea fog event over the Yellow Sea in March 2005: Analysis and numerical modeling. Adv. Atmos. Sci., 24, 6581.

    • Search Google Scholar
    • Export Citation
  • Gao, S. H., , Wu W. , , Zhu L. L. , , Fu G. , , and Huang B. , 2009: Detection of nighttime sea fog/stratus over the Huanghai Sea using MTSAT-1R IR data. Acta Oceanol. Sin., 28, 2335.

    • Search Google Scholar
    • Export Citation
  • Gao, S. H., , Qi Y. L. , , Zhang S. B. , , and Fu G. , 2010: Initial conditions improvement of sea fog numerical modeling over the Yellow Sea by using cycling 3DVAR. Part I: WRF numerical experiments (in Chinese). J. Ocean Univ. China,40, 001–009.

  • Gultepe, I., and Coauthors, 2007: Fog research: A review of past achievements and future perspectives. Pure Appl. Geophys., 164, 11211159.

    • Search Google Scholar
    • Export Citation
  • Heidinger, A. K., , and Stephens G. L. , 2000: Molecular line absorption in a scattering atmosphere. Part II: Application to remote sensing in the O2 a band. J. Atmos. Sci., 57, 16151634.

    • Search Google Scholar
    • Export Citation
  • Hong, S. Y., , and Lim J. O. J. , 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129151.

  • Hong, S. Y., , Noh Y. , , and Dudhia J. , 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341.

    • Search Google Scholar
    • Export Citation
  • Hu, J. F., , Guo K. C. , , and Yan L. N. , 1996: Discriminate prediction of marine fog occurrence using a model output statistics scheme (in Chinese). J. Ocean Univ. Qingdao, 26, 439445.

    • Search Google Scholar
    • Export Citation
  • Hu, R. J., , and Zhou F. X. , 1997: A numerical study of the effects of air–sea conditions on the process of sea fog (in Chinese). J. Ocean Univ. Qingdao, 27, 282290.

    • Search Google Scholar
    • Export Citation
  • Hu, R. J., , and Zhou F. X. , 1998: Effects of advection, turbulence and radiation on formation of sea fog (in Chinese). J. Ocean Univ. Qingdao, 20, 2530.

    • Search Google Scholar
    • Export Citation
  • Huang, J., , and Zhou F. X. , 2006: The cooling and moistening effect on the formation of sea fog in the Huanghai Sea. Acta Oceanol. Sin., 25, 4962.

    • Search Google Scholar
    • Export Citation
  • Hunt, G. E., 1973: Radiative properties of terrestrial clouds at visible and infrared thermal window wavelengths. Quart. J. Roy. Meteor. Soc., 99, 346369.

    • Search Google Scholar
    • Export Citation
  • Iacono, M. J., , Delamere J. S. , , Mlawer E. J. , , Shephard M. W. , , Clough S. A. , , and Collins W. D. , 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, doi:10.1029/2008JD009944.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., , and Fritsch J. M. , 1990: A one-dimensional entraining/detraining plume model and its application in convective parameterization. J. Atmos. Sci., 47, 27842802.

    • Search Google Scholar
    • Export Citation
  • Kästner, M., , Kriebel K. T. , , Meerkötter R. , , Renger W. , , Ruppersberg G. H. , , and Wendling P. , 1993: Comparison of cirrus height and optical depth derived from satellite and aircraft measurements. Mon. Wea. Rev., 121, 27082715.

    • Search Google Scholar
    • Export Citation
  • Kim, C. K., , and Yum S. S. , 2010: Local meteorological and synoptic characteristics of fogs formed over Incheon international airport in the west coast of Korea. Adv. Atmos. Sci., 27, 761776.

    • Search Google Scholar
    • Export Citation
  • Kim, C. K., , and Yum S. S. , 2012: A numerical study of sea-fog formation over cold sea surface using a one-dimensional turbulence model coupled with the Weather Research and Forecasting Model. Bound.-Layer Meteor., 143, 481505.

    • Search Google Scholar
    • Export Citation
  • Koračin, D., , Lewis J. , , Thompson W. T. , , Dorman C. E. , , and Businger J. A. , 2001: Transition of stratus into fog along the California coast: Observation and modeling. J. Atmos. Sci., 58, 17141731.

    • Search Google Scholar
    • Export Citation
  • Koračin, D., , Businger J. A. , , Dorman C. E. , , and Lewis J. M. , 2005a: Formation, evolution, and dissipation of coastal sea fog. Bound.-Layer Meteor., 117, 447478.

    • Search Google Scholar
    • Export Citation
  • Koračin, D., , Leipper D. F. , , and Lewis J. M. , 2005b: Modeling sea fog on the U.S. California coast during a hot spell event. Geofizika, 22, 5982.

    • Search Google Scholar
    • Export Citation
  • Kunkel, B., 1984: Parameterization of droplet terminal velocity and extinction coefficient in fog models. J. Climate Appl. Meteor., 23, 3441.

    • Search Google Scholar
    • Export Citation
  • Lee, T. F., , Turk F. J. , , and Richardson K. , 1997: Stratus and fog productions using GOES-8–9 3.9-μm data. Wea. Forecasting, 12, 606619.

    • Search Google Scholar
    • Export Citation
  • Leipper, D., 1948: Fog development at San Diego, California. J. Mar. Res., 7, 337346.

  • Lewis, J. M., , Koračin D. , , Rabin R. , , and Businger J. , 2003: Sea fog off the California coast: Viewed in the context of transient weather systems. J. Geophys. Res., 108, 4457, doi:10.1029/2002JD002833.

    • Search Google Scholar
    • Export Citation
  • Lewis, J. M., , Koračin D. , , and Redmond K. T. , 2004: Sea fog research in the United Kingdom and United States. Bull. Amer. Meteor. Soc., 85, 395408.

    • Search Google Scholar
    • Export Citation
  • Li, R., , Gao S. H. , , and Wang Y. M. , 2012: Numerical study on direct assimilation of satellite radiances for sea fog over the Yellow Sea (in Chinese). J. Ocean Univ. China,42, 10–20.

  • Liang, W. F., , and Hou Z. X. , 2001: The characteristics and forecasts of heavy fog occurred over Qingdao (in Chinese). Shandong Meteor., 84, 1117.

    • Search Google Scholar
    • Export Citation
  • Lim, K. S. S., , and Hong S. Y. , 2010: Development of an effective double-moment cloud microphysics scheme with prognostic cloud condensation nuclei (CCN) for weather and climate models. Mon. Wea. Rev., 138, 15871612.

    • Search Google Scholar
    • Export Citation
  • Lin, Y. L., , Farley R. D. , , and Oriville H. D. , 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl. Meteor., 22, 10651092.

    • Search Google Scholar
    • Export Citation
  • Liu, X., , and Hu X. Q. , 2008: Sea fog automatic detection over the East China Sea using MTSAT data (in Chinese). J. Oceanogr. Taiwan, 27, 112117.

    • Search Google Scholar
    • Export Citation
  • Nakanishi, M., , and Niino H. , 2006: An improved Mellor–Yamada level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397407.

    • Search Google Scholar
    • Export Citation
  • Nicholls, S., 1984: The dynamics of stratocumulus: Aircraft observation and comparison with a mixed layer model. Quart. J. Roy. Meteor. Soc., 110, 783820.

    • Search Google Scholar
    • Export Citation
  • Otkin, J. A., , and Greenwald T. J. , 2008: Comparison of WRF model-simulated and MODIS-derived cloud data. Mon. Wea. Rev., 136, 19571970.

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

    • Search Google Scholar
    • Export Citation
  • Petterssen, S., 1936: On the causes and the forecasting of the California fog. J. Aeronaut. Sci., 3, 305309.

  • Petterssen, S., 1938: On the causes and the forecasting of the California fog. Bull. Amer. Meteor. Soc., 19, 4955.

  • Skamarock, W., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN–475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3_bw.pdf.]

  • Sorli, B., , Pascal-Delannoy F. , , Giani A. , , Foucaran A. , , and Boyer A. , 2002: Fast humidity sensor for high range 80%–95% RH. Sens. Actuators, 100A, 2431.

    • Search Google Scholar
    • Export Citation
  • Stoelinga, M. T., , and Thomas T. W. , 1999: Nonhydrostatic, mesobeta-scale model simulations of cloud ceiling and visibility for an East Coast winter precipitation event. J. Appl. Meteor., 38, 385403.

    • Search Google Scholar
    • Export Citation
  • Taylor, G., 1917: The formation of fog and mist. Quart. J. Roy. Meteor. Soc., 43, 241268.

  • Thompson, G., , Field P. R. , , Rasmussen R. M. , , and Hall W. D. , 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 50955115.

    • Search Google Scholar
    • Export Citation
  • Trémant, M., 1987: La prèvision du brouilliard en mer. Meteorologie Maritime et Activities, Oceanograpiques Connexes Paport, Vol. 20, WMO, 127 pp.

  • Wang, B. H., 1985: Sea Fog. China Ocean Press, 330 pp.

  • Wang, X., , Barker D. , , Snyder C. , , and Hamill T. M. , 2008a: A hybrid ETKF–3DVAR data assimilation scheme for the WRF model. Part I: Observing system simulation experiment. Mon. Wea. Rev., 136, 51165131.

    • Search Google Scholar
    • Export Citation
  • Wang, X., , Barker D. , , Snyder C. , , and Hamill T. M. , 2008b: A hybrid ETKF–3DVAR data assimilation scheme for the WRF model. Part II: Real observation experiments. Mon. Wea. Rev., 136, 51325147.

    • Search Google Scholar
    • Export Citation
  • WMO, 1966: International Meteorological Vocabulary.World Meteorological Organization, 276 pp.

  • Yoo, J. M., , Jeong M. J. , , Hur Y. M. , , and Shin D. B. , 2010: Improved fog detection from satellite in the presence of clouds. Asia-Pac. J. Atmos. Sci., 46, 2940.

    • Search Google Scholar
    • Export Citation
  • Zhang, H. Y., , Zhou F. X. , , and Zhang X. H. , 2005: Interannual change of sea fog over the Yellow Sea in spring (in Chinese). Oceanol. Limnol. Sin., 36, 3642.

    • Search Google Scholar
    • Export Citation
  • Zhang, S. P., , and Bao X. W. , 2008: The main advances in sea fog research in China (in Chinese). J. Ocean Univ. China, 38, 359366.

  • Zhang, S. P., , Xie S. P. , , Liu Q. Y. , , Yang Y. Q. , , Wang X. G. , , and Ren Z. P. , 2009: Seasonal variations of Yellow Sea fog: Observations and mechanisms. J. Climate, 22, 67586772.

    • Search Google Scholar
    • Export Citation
  • Zhao, Y. P., , Chen Y. L. , , and Wang P. G. , 1997: Analysis of atmospheric and oceanic conditions for marine fog formation over the Yellow Sea and East China Seas (in Chinese). Stud. Mar. Sin., 38, 6979.

    • Search Google Scholar
    • Export Citation
  • Zhou, B., , and Du J. , 2010: Fog prediction from a multimodel mesoscale ensemble prediction system. Wea. Forecasting, 25, 303322.

  • Zhou, F. X., , and Liu L. T. , 1986: Comprehensive survey and research report on the water areas adjacent to the Changjing River estuary and Chejudo Island marine fog (in Chinese). J. Shandong Coll. Oceanol., 16, 114131.

    • Search Google Scholar
    • Export Citation
  • Zhou, F. X., , Wang X. , , and Bao X. W. , 2004: Climatic characteristics of sea fog formation of the Huanghai Sea in spring (in Chinese). Acta Oceanol. Sin., 26, 2837.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Geographic map of the Yellow Sea. The two domains with solid lines labeled D1 and D2 are used for the first simulation and the two domains with dashed lines are used for the second simulation. Locations of named and unnamed radiosonde stations (dots) and a buoy (filled triangle) are marked.

  • View in gallery

    Flowchart for the Yellow Sea fog detection algorithm using MTSAT data.

  • View in gallery

    Observed sea fog (brown shading with visibility in meters) evolution detected from MTSAT data. The filled circle, plus sign, and open circle represent the observed visibility with values 0–1, 1–5, and 5–10 km, respectively. The light blue shading indicates high-cloud overcast areas. The observed sea fog from (top) 1200 UTC 5 Mar to 0000 UTC 6 Mar is used for assimilation; and that from 0100 UTC 6 Mar (not shown) to 0000 UTC 8 Mar (not shown) is used for evaluation.

  • View in gallery

    Schematic diagram for analyzing and digitizing (discretizing) sea fog humidity soundings in a west–east vertical cross section, where Hi, Pi, and Ti represent the height, pressure, and temperature, respectively, at the ith level. Above point X is shown a sea fog humidity sounding. Information from such soundings would be used to allocate sea fog humidity data to grid points with the horizontal and vertical grid intervals marked as ΔH and ΔZ.

  • View in gallery

    Flowchart of the cycling 3DVAR method. See text for details.

  • View in gallery

    Illustration of assimilating a sea fog humidity sounding in one 3DVAR cycle. The meanings of most symbols can be determined from Fig. 5 and the text. Additionally, “analysis” is the result of the first run, and MTSAT-RH represents a sea fog humidity sounding.

  • View in gallery

    Example showing the forecast sea fog areas with fog-top heights (shading) from (top) Exp-A and (bottom) Exp-B for (left to right) 0600 UTC 6 Mar to 0600 UTC 7 Mar. A sketch of the transient cyclone (L)–anticyclone (H) couplet system is shown with 1000-hpa geopotential heights. The arrows in (a) indicate the couplet tracks in the next 24 h; and (b) the solid triangle marks the moored buoy site. The lines A–B in (b),(e) show the location for the vertical cross sections in Fig. 8.

  • View in gallery

    Vertical cross section along the line A–B in Fig. 7 at 1800 UTC 6 Mar 2006 for (a) Exp-A and (b) Exp-B. Colors and vectors show temperatures (°C) and winds (m s−1), respectively. Contours—solid, dashed, and dotted—show mixing ratios with values of 0.016, 0.1, and 0.2 g kg−1, respectively.

  • View in gallery

    Time series of air temperature (T, labeled curves ) and relative humidity (RH, unlabeled curves) measured and simulated at the buoy Idedo.

  • View in gallery

    Vertical profiles of the RMSE (solid lines) and biases (dashed lines) between the initial analysis and the radiosonde observations for Exp-A (curves no symbols) and Exp-B (curves with symbols) over the verification domain for (a) mixing ratio, (b) temperature, and (c) u and (d) υ wind components.

  • View in gallery

    As in Fig. 10, but for only the named rawinsdonde stations in Fig. 1 and only for (a) mixing ratio and (b) temperature.

  • View in gallery

    As in Fig. 10, but for 12-h forecast fields.

  • View in gallery

    Comparison of the time series of the average RMSE from the surface to 850 hPa of the forecast fields and the radiosonde observations over the verification domain for Exp-A (solid line) and Exp-B (dashed line) of (a) mixing ratio, (b) temperature, and the (c) u and (d) υ wind components.

  • View in gallery

    Observed sea fog with top heights detected from MTSAT data from (top left) 1900 UTC 28 Apr to (bottom right) 0600 UTC 29 Apr 2007. The symbols are as in Fig. 3.

  • View in gallery

    Forecast sea fog with top height (shading) and 10-m winds (vectors) for (top) Exp-OBS and (bottom) Exp-RH for (left to right) 2100 UTC 28 Apr to 0600 UTC 29 Apr.

  • View in gallery

    Comparison of 3-h forecast vertical profiles of Exp-RH (dashed lines) and Exp-OBS (dotted lines) with radiosonde observations (solid lines) at (top) Qingdao and (bottom) Chengshantou stations. Shown are (a),(c) temperatures and (b),(d) mixing ratios.

  • View in gallery

    Average moisture biases from the surface to 850 hPa between the initial analysis and the observations at the nine named radiosonde stations around the Yellow Sea (see Fig. 1).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 108 108 8
PDF Downloads 91 91 5

Assimilating MTSAT-Derived Humidity in Nowcasting Sea Fog over the Yellow Sea

View More View Less
  • 1 Key Laboratory of Physical Oceanography, Department of Atmospheric Science, Ocean University of China, Qingdao, China
  • | 2 Department of Atmospheric Science, Ocean University of China, Qingdao, China
© Get Permissions
Full access

Abstract

An extended three-dimensional variational data assimilation (3DVAR) method based on the Weather Research and Forecasting Model (WRF) is developed to assimilate satellite-derived humidity from sea fog at its initial stage over the Yellow Sea. The sea fog properties, including its horizontal distribution and thickness, are retrieved empirically from the infrared and visible cloud imageries of the Multifunctional Transport Satellite (MTSAT). Assuming a relative humidity of 100% in fog, the MTSAT-derived humidity is assimilated by the extended 3DVAR assimilation method. Two sea fog cases, one spread widely over the Yellow Sea and the other spread narrowly along the coast, are first studied in detail with a suite of experiments. For the widespread-fog case, the assimilation of MTSAT-derived information significantly improves the forecast of the sea fog area, increasing the probability of detection and equitable threat scores by about 20% and 15%, respectively. The improvement is attributed to a more realistic representation of the marine boundary layer (MBL) and better descriptions of moisture and temperature profiles. For the narrowly spread coastal case, the model completely fails to reproduce the sea fog event without the assimilation of MTSAT-derived humidity. The extended 3DVAR assimilation method is then applied to 10 more sea fog cases to further evaluate its effect on the model simulations. The results reveal that the assimilation of MTSAT-derived humidity not only improves sea fog forecasts but also provides better moisture and temperature structure information in the MBL.

Corresponding author address: Shanhong Gao, Key Laboratory of Physical Oceanography, Dept. of Atmospheric Science, Ocean University of China, 5 Yushan Rd., Qingdao 266003, China. E-mail: gaosh@ouc.edu.cn

Abstract

An extended three-dimensional variational data assimilation (3DVAR) method based on the Weather Research and Forecasting Model (WRF) is developed to assimilate satellite-derived humidity from sea fog at its initial stage over the Yellow Sea. The sea fog properties, including its horizontal distribution and thickness, are retrieved empirically from the infrared and visible cloud imageries of the Multifunctional Transport Satellite (MTSAT). Assuming a relative humidity of 100% in fog, the MTSAT-derived humidity is assimilated by the extended 3DVAR assimilation method. Two sea fog cases, one spread widely over the Yellow Sea and the other spread narrowly along the coast, are first studied in detail with a suite of experiments. For the widespread-fog case, the assimilation of MTSAT-derived information significantly improves the forecast of the sea fog area, increasing the probability of detection and equitable threat scores by about 20% and 15%, respectively. The improvement is attributed to a more realistic representation of the marine boundary layer (MBL) and better descriptions of moisture and temperature profiles. For the narrowly spread coastal case, the model completely fails to reproduce the sea fog event without the assimilation of MTSAT-derived humidity. The extended 3DVAR assimilation method is then applied to 10 more sea fog cases to further evaluate its effect on the model simulations. The results reveal that the assimilation of MTSAT-derived humidity not only improves sea fog forecasts but also provides better moisture and temperature structure information in the MBL.

Corresponding author address: Shanhong Gao, Key Laboratory of Physical Oceanography, Dept. of Atmospheric Science, Ocean University of China, 5 Yushan Rd., Qingdao 266003, China. E-mail: gaosh@ouc.edu.cn

1. Introduction

Fog that occurs over the ocean or a coastal region is usually termed sea fog. It significantly reduces low-level visibilities, which can play a role in severe marine accidents (Trémant 1987; Gultepe et al. 2007). Sea fog is likely to form over regions where cold sea surface temperatures (SSTs) are adjacent to warm currents (Lewis et al. 2004). The Yellow Sea of China (shown in Fig. 1), located north of the warm Kuroshio Current, is exactly such a region and experiences a high frequency of sea fog. The sea fog season in this region usually starts in March and ends in August, with the dominant type being advection fog. The average number of fog days annually along the coastal region of Qingdao, which is located right over northern Yellow Sea, is about 51 but reached 89 in 2006 (Zhang and Bao 2008; Fu et al. 2012).

Fig. 1.
Fig. 1.

Geographic map of the Yellow Sea. The two domains with solid lines labeled D1 and D2 are used for the first simulation and the two domains with dashed lines are used for the second simulation. Locations of named and unnamed radiosonde stations (dots) and a buoy (filled triangle) are marked.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

The pioneering research into the sea fog phenomenon can be traced back to as early as Taylor’s original work (Taylor 1917). By 1950, several important contributions to sea fog formation had been published (e.g., Byers 1930; Anderson 1931; Petterssen 1936, 1938; Leipper 1948). Although the pioneering research tended to be phenomenological, researchers laid the groundwork for the numerical modeling studies that came afterward (Lewis et al. 2004). Results of the pioneering work on sea fog study in China were published in the notable book Sea Fog by Wang (1985). Since then, there have been extensive observational and modeling studies of sea fog over the Yellow Sea. These studies can be classified into three categories: 1) statistical analysis of spatial distributions and temporal variations of sea fog based on observations along the coast, as well as buoy measurements and ship reports and satellite images (Zhou and Liu 1986; Zhao et al. 1997; Cho et al. 2000; Kim and Yum 2010); 2) statistical sea fog forecasting by using multivariate regression equations (Hu et al. 1996; Liang and Hou 2001)—note that this method is usually applied to near-coastal regions and its application in other areas is quite limited; and 3) numerical modeling studies of sea fog formation and the physical–dynamical processes involved (e.g., Hu and Zhou 1997, 1998; Zhou et al. 2004; Fu et al. 2004, 2006, 2010; Zhang et al. 2005, 2009; Huang and Zhou 2006; Gao et al. 2007; Kim and Yum 2012). Among all of these approaches, numerical modeling is especially effective in studying the mechanisms for sea fog formation and evolution and, hence, plays a leading role in forecasting sea fog over the Yellow Sea.

However, sea fog forecasting can be tricky for many reasons. Previous studies on sea fog have indicated that sea fog forecasting is very sensitive to the initial conditions and the formation and evolution of sea fogs are largely controlled by the development of the marine boundary layer (MBL) (Nicholls 1984; Findlater et al. 1989; Ballard et al. 1991; Lewis et al. 2003; Koračin et al. 2001, 2005a,b; Fu et al. 2006; Gao et al. 2007, 2010). Sea fogs over the Yellow Sea are mostly advection fog events, which form when warm moist air passes over cold water (Wang 1985). During the sea fog season from March to August, the dominant weather pattern over the Yellow Sea is either a transient cyclone–anticyclone couplet system or an isolated anticyclone (Wang 1985; Zhou et al. 2004; Gao et al. 2007). Consequently, the prevailing southerly winds carry warm moist air from warm waters in the south toward the cold northern Yellow Sea (shown in Fig. 1), resulting in rapid changes in the MBL structure over the Yellow Sea. The mechanical turbulence cooling over the cold sea surface leads to the building of a shallow and stable moist MBL in which sea fog is likely to form. However, such an MBL is usually not embodied in the initial conditions that are derived mostly from a global or regional analysis, and which includes few routine observations from over the Yellow Sea.

It is well known that accurate initial conditions, especially realistic representations of temperature and humidity profiles, are crucial for successful sea fog simulation. However, since routine surface and upper-air observations (see the dots in Fig. 1) are mostly available over the land, it is difficult to generate optimal initial conditions for modeling studies of fog over the sea. To capture the MBL structure, Gao et al. (2010) designed a cycling three-dimensional variational data assimilation (3DVAR) scheme based on the Weather Research and Forecasting Model (WRF) and its 3DVAR module. When utilizing this scheme to assimilate satellite radiance, Li et al. (2012) found that for most sea fog cases over the Yellow Sea, the inversion thermal structure in the MBL is better represented than that by assimilating routine observations only. However, little is changed in the MBL humidity structure. Based on our experience, failures in sea fog simulations are often due to insufficient MBL moisture in the initial conditions. Thus, this study will focus on the assimilation of the satellite-derived humidity, which is supposed to help generate better initial conditions for sea fog modeling studies.

As an important approach to detecting sea fog, geostationary-orbit satellite imagery with high spatiotemporal resolution (i.e., 1–4-km resolution and 30-min to 1-h interval)—for example the products of the Multifunctional Transport Satellite (MTSAT) of Japan and the Fengyun-2 (FY-2) of China—have already been used to monitor the process of sea fog evolution over the Yellow Sea. Over the last two decades, detection of sea fog using satellite brightness temperature and albedo has been greatly improved (Ellrod 1995; Bendix et al. 2005; Deng et al. 2006; Liu and Hu 2008; Gao et al. 2009; Yoo et al. 2010; Fu et al. 2011). However, whether it is possible to derive realistic humidity information from satellite-detected sea fog still remains an open question. If the humidity can be derived from a sea fog event in its initial stage and then assimilated into a model for the next forecast cycle, particularly in a nowcast system, forecasting accuracy is expected to improve. Based on this idea, we attempted to first detect sea fog events using the MTSAT data that included visible albedo (VIS) and infrared channel brightness temperature (IR), derive the humidity, and then assimilate it into the WRF using the cycling 3DVAR scheme mentioned previously.

This paper is organized as follows. Section 2 describes the procedure developed to derive the humidity from MTSAT data, and introduces the cycling 3DVAR scheme, which is used to assimilate the derived humidity into the WRF. Section 3 discusses two case studies and illustrates the impact of humidity assimilation on sea fog forecasts by conducting a suite of assimilation experiments. In section 4, we apply the cycling 3DVAR scheme to assimilate humidity into the WRF and perform sea fog simulations for 10 cases. Effects of humidity assimilation and its potential applications in sea fog forecasting are evaluated based on results of these experiments. Summary and conclusions are given in section 5.

2. Methodology

It can be assumed that the air is saturated with 100% relative humidity (RH) inside sea fog. Derivation of humidity is equivalent to the determination of the horizontal distribution and vertical thickness of sea fog. The cycling 3DVAR scheme (Gao et al. 2010) is extended to assimilate the MTSAT-derived humidity.

a. Derivation of sea fog humidity

The methods of sea fog detection during daytime and nighttime are different due to the solar radiation effect. The dual infrared channel difference (DCD) method (Hunt 1973; Ellrod 1995; Lee et al. 1997) is used to detect nighttime sea fog based on MTSAT data. The brightness temperature difference between the shortwave channel (IR4 for MTSAT) and longwave channel (IR1 for MTSAT) are calculated (shortwave minus longwave brightness temperature, simply SLTD; i.e., IR4 − IR1). An SLTD value ranging from −5.5° to 2.5°C can be used as a criterion to determine the area of sea fog over the Yellow Sea (Gao et al. 2009). The thickness (or top height) of the sea fog (i.e., the depth from fog top to the sea surface) is calculated following Ellrod (1995):
e1
where H is the thickness in unit of hundreds of meters.

However, this method is not valid during the daytime because solar radiation significantly affects the SLTD. An alternative is to use the fog-top pressure and the mean sea level pressure provided by the International Satellite Cloud Climatology Project (ISCCP) to estimate the fog-top height (e.g., Kim and Yum 2012). However, the ISCCP data are not used for daytime sea fog detection in this study due to its coarse spatial–temporal resolution and difficulty in accessing real-time data.

Liu and Hu (2008) proposed a method to detect daytime sea fog over the East China Sea using the MTSAT data. In their method, first an area where the shortwave brightness temperature and SST difference (hereafter ISTD) are less than 10 K is considered to be a high-altitude cloud or cloud-free area and is discarded. The “noisy” nonfog areas are then removed based on albedo and SLTD that vary with solar zenith angles (hereafter dynamic SLTD). Finally, areas with rough cloud tops are discarded by a texture-filtering method (Coakley and Bretherton 1982; Ahn et al. 2003). The remaining area is regarded as an area of sea fog. Unfortunately, two weaknesses of this approach (i.e., the use of monthly SST and its failure over land near a coastal region) greatly limit its application in sea fog studies.

Based on the study of Bendix et al. (2005), as well as a statistical regression analysis of 46 sea fog cases over the Yellow Sea, Fu et al. (2011) developed a new empirical formula, which is a function of satellite visible albedo to calculate the surface visibility (hereafter Fvis).

In this study, we developed a new method to detect sea fog during either daytime or nighttime. This method not only overcomes the weaknesses of the above-mentioned studies but also takes advantage of their strengths (Fig. 2). The SLTD method with Eq. (1) is employed for the detection of nighttime sea fog, and a detailed description of sea fog detection for daytime is given below.

Fig. 2.
Fig. 2.

Flowchart for the Yellow Sea fog detection algorithm using MTSAT data.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

First, high-altitude cloud and cloud-free areas are discarded when the ISTD is less than 4 K. The daily SST (0.25° × 0.25°) from the North-East Asian Regional–Global Ocean Observing System (NEAR-GOOS) is used here. Second, sea fog patches are determined if Fvis is less than 1.0 km and SLTD is within the range of 3–45 K when the solar zenith angle varies between 10°–80°; or SLTD is within −2–3 K, when the solar angle is larger than 80° or smaller than 10°. For sea fog patch detection over land near the coastal region, the same method is utilized, but the visible albedo used for calculating Fvis is multiplied by 1.8 (http://www.climatedata.info/Forcing/Forcing/albedo.html). Finally, areas with rough textured cloud tops (defined as having the standard deviation of the visible albedo for 3 × 3 pixels larger than 0.08) are discarded. After the above steps, the remaining areas are regarded as sea fog patches. Sea fog thickness is calculated by the following formula (Heidinger and Stephens 2000):
e2
where H is the sea fog thickness (in km) and is the optical thickness, which is determined by the satellite visible albedo and the solar zenith angle (Kästner et al. 1993; Fitzpatrick et al. 2004).

A heavy sea fog event occurred over the Yellow Sea from 6 to 8 March 2006. From 1200 UTC 5 March to 0000 UTC 6 March, sea fog patches were sporadic due to high cloud overshadowing (Figs. 3a–d). There were few high-altitude clouds covering the sea fog area during 0300 UTC 6 March to 0600 UTC 7 March (Figs. 3e–j), meaning that the detected sea fog patches were slightly underestimated. In Figs. 3k,l, a large area of the southern Yellow Sea became overcast by high-altitude clouds from 1200 to 1800 UTC 7 March. However, under these high-altitude clouds there might exist sea fog since the observed visibility was less than 1 km over some areas (e.g., Fig. 3l). Thus, it is possible that the detected sea fog areas in Figs. 3k,l may have been seriously underestimated.

Fig. 3.
Fig. 3.

Observed sea fog (brown shading with visibility in meters) evolution detected from MTSAT data. The filled circle, plus sign, and open circle represent the observed visibility with values 0–1, 1–5, and 5–10 km, respectively. The light blue shading indicates high-cloud overcast areas. The observed sea fog from (top) 1200 UTC 5 Mar to 0000 UTC 6 Mar is used for assimilation; and that from 0100 UTC 6 Mar (not shown) to 0000 UTC 8 Mar (not shown) is used for evaluation.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Sometimes sea fog develops and spreads upward rapidly, leaving the sea surface; and changes into low-level stratus that is usually mingled within sea fog patches. Unfortunately, it is impossible for our current detection methods to distinguish between low-level stratus and sea fog due to their similar top characteristics. Hereafter, both sea fog and low-level stratus are called sea fog. Note that long-lived low-level stratus is rarely found over the Yellow Sea, because the balance between the subsidence and the shallow MBL over cold waters is hard to maintain. In fact, during the sea fog season, the weather systems crossing the Yellow Sea often change quickly; this makes the balance even harder to set up.

b. Assimilation of MTSAT-derived humidity

Figure 4 shows a schematic illustration of how sea fog humidity data is processed before being assimilated. Sea fog is assumed to be adjacent to the sea surface, as is usually the case.

Fig. 4.
Fig. 4.

Schematic diagram for analyzing and digitizing (discretizing) sea fog humidity soundings in a west–east vertical cross section, where Hi, Pi, and Ti represent the height, pressure, and temperature, respectively, at the ith level. Above point X is shown a sea fog humidity sounding. Information from such soundings would be used to allocate sea fog humidity data to grid points with the horizontal and vertical grid intervals marked as ΔH and ΔZ.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Sea fog humidity information is first put into a discrete numerical format in order to be recognized by the WRF-3DVAR. The sea fog data is then allocated to grid points with the horizontal and vertical grid intervals marked as ΔH and ΔZ in Fig. 4, respectively. For a given location in a sea fog area, grid point X in Fig. 4, the vertical humidity profile inside the sea fog can be regarded as a routine sounding profile that only contains humidity and elevation information (hereafter “sea fog humidity sounding”); thus, the whole sea fog space consists of a large number of sea fog humidity soundings but without pressure and temperature information.

Figure 5 shows the cycling 3DVAR flowchart designed by Gao et al. (2010). In a given assimilation window, the 3DVAR process is cycled. For instance, the assimilation time window in Fig. 5 is 0–2Δt. During this window, the WRF preprocessing system (WPS) and the initialization program are run to generate the background (bg) of the initial conditions for the first 3DVAR cycle at the time 0; then, the 3DVAR is conducted with a background error covariance (be) to assimilate observations (obs) around the time period between −0.5Δt and +0.5Δt to generate a realistic analysis, which is used to initialize the WRF. The WRF program wrf.exe starts its integration from this analysis to the next time Δt, and the 3DVAR routine is conducted again. The process is repeated during the entire assimilation period.

Fig. 5.
Fig. 5.

Flowchart of the cycling 3DVAR method. See text for details.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

This cycling 3DVAR has the capability to assimilate various observations, for example, soundings, surface measurements, sea surface winds, and temperature profiles derived from satellite. To distinguish these observations from sea fog humidity soundings, they are collectively called obs (seen in Figs. 5 and 6). Some modifications are made in this study to the cycling 3DVAR approach to expand its capability for assimilating sea fog humidity soundings. The cycle interval of the cycling 3DVAR assimilation process is adjusted to match that of the MTSAT data available with a time interval of 1–3 h. The result (i.e., analysis) generated by a cycle of the cycling 3DVAR method at one specific analysis time is taken as the three-dimensional background for those sea fog humidity soundings of the same time. Temperature and pressure are extracted from the three-dimensional background analysis. To assimilate sea fog humidity, 3DVAR has to run twice in each cycle of the cycling 3DVAR scheme. However, if the forecast field from the previous cycle is directly taken as the background, then the second 3DVAR run can be skipped but the assimilation effect might be less helpful for model simulation. For clarity, Fig. 6 only demonstrates the modification on one cycle, which is the identical cycle illustrated by the dashed line in Fig. 5. The analysis from the first run is taken as the background for the sea fog humidity sounding, which only includes the humidity and elevation (RH and Zi in Fig. 6). After the temperature and pressure (Ti and Pi in Fig. 6) are extracted, sea fog humidity sounding data are incorporated into the obs and assimilated into the WRF via the second cycle of the 3DVAR run.

Fig. 6.
Fig. 6.

Illustration of assimilating a sea fog humidity sounding in one 3DVAR cycle. The meanings of most symbols can be determined from Fig. 5 and the text. Additionally, “analysis” is the result of the first run, and MTSAT-RH represents a sea fog humidity sounding.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

3. Case studies

The extended cycling 3DVAR scheme was applied to assimilation experiments with a focus on a heavy sea fog event covering a large area over the Yellow Sea. Assimilation effects on the sea fog area forecast and the impacts on physical fields were analyzed. In addition, a heavy costal sea fog event that covered a small area was also examined to further evaluate the validity of this assimilation scheme.

a. A heavy sea fog event that covered a large area

The sea fog event occurred during 6–8 March 2006, and the facts of this event are illustrated in Fig. 3 (see brown shading for surface visibility observations). The sea fog already formed over the southwest Yellow Sea at 0300 UTC 6 March (Fig. 3e), and spread northeastward toward the region near the Shandong Peninsula (Figs. 3f–i). Starting from 0600 UTC 7 March, the sea fog occupied a large part of the Yellow Sea (Figs. 3j,k), indicating that the sea fog had reached its mature stage. The fog started to retreat southward during the morning of 8 March (Fig. 3l).

b. Experiment design

The Advanced Research core of the WRF, version 3.3 (ARW; Skamarock et al. 2008), was employed in this study. Sea fog is treated as low-level clouds that form in the MBL in the WRF. The low-level cloud properties are found to be very sensitive to the planet boundary layer (PBL) scheme while the upper-level clouds are sensitive to both the microphysics and PBL schemes (Otkin and Greenwald 2008). To evaluate the performance of the PBL and microphysics schemes in the simulation of sea fog properties, ensemble experiments were conducted for 15 Yellow Sea fog cases using different combinations of PBL and microphysics schemes. The combinations consist of the Yonsei University scheme (YSU; Hong et al. 2006), the Mellor–Yamada–Janjić (MYJ) and Mellor–Yamada–Nakanishi–Niino (MYNN; Nakanishi and Niino 2006) PBL schemes, and the the Purdue Lin (PLIN; Lin et al. 1983), the WRF single-moment six-class (WSM6; Hong and Lim 2006), the WRF double-moment six-class (WDM6; Lim and Hong 2010), and the Thompson (Thompson et al. 2008) microphysics schemes. The result revealed that the best combination of the PBL and microphysics schemes wasYSU–PLIN. Therefore, the YSU and PLIN schemes were chosen in this study. Table 1 summarizes the model configurations, including the domains (see Fig. 1), resolutions, and physics.

Table 1.

WRF configuration.

Table 1.

The initial and lateral boundary conditions were derived from the National Centers for Environmental Prediction (NCEP) Final Analysis (FNL; 1° × 1°, 6 hourly), and the SST data are extracted from daily NEAR-GOOS dataset. The cycling 3DVAR assimilation was performed with a cycle interval of 3 h1 in the window from 1200 UTC 5 March to 0000 UTC 6 March 2006 (12 h in total). The next 48 h from 0000 UTC 6 March to 0000 UTC 8 March 2006 were set as the forecasting period. Because the density of the observation sites was more than 30 km, assimilation was performed for the outer domain only; while the forecast was done over two domains, and the initial conditions for the inner domain weree provided by the outer-domain information. The hourly output of the model results for the inner domain will be used later for analysis.

Since the default WRF 3DVAR NCEP covariance may not have been the optimal static background error covariance for the current data assimilation experiments, the WRF was initialized every 12 h and run 24 h for a period of 15 days. The results were used to generate a new static background error covariance file by using the National Meteorological Center (NMC) method (Parrish and Derber 1992). During the 12-h assimilation window, four time periods of sea fog humidity sounding data were available2 and were numerically discretized based on the detected three-dimensional sea fog shown in Figs. 3a–d. In the discretizing process, the horizontal grid size ΔH was set to 0.1°, and the vertical grid size ΔZ was kept the same as that of the background (i.e., the analysis in Fig. 6), and the RH in sea fog was set to 100%. This is supported by Sorli et al. (2002) who reported that RH values usually ranged from 95%–98% in saturated conditions. In addition, the RH from sea fog humidity soundings will be converted into specific humidity (a prognostic variable in WRF) that is not sensitive to RH varying between 95–100%. To further test this sensitivity, different RHs (95%, 97%, 99%, and 100%) were also selected. The results show that the RH value had very little impact on the assimilation experiments.

Six experiments are performed (Exp-A–Exp-F in Table 2) in this study. Exp-A adopts the flowchart shown in Fig. 5 without assimilating sea fog humidity soundings, and serves as a baseline for evaluating the effects of the assimilation. Exp-B through Exp-E all adopt the improved flowchart shown in Fig. 6 to assimilate both the obs and sea fog humidity sounding data. Among these experiments, Exp-B was specifically designed to see the effects of assimilating sea fog humidity soundings. To account for the errors in the detected sea fog thickness, Exp-C and Exp-D were designed to investigate the assimilation sensitivity to errors in the detected sea fog top; while Exp-E was designed to test the assimilation’s sensitivity to errors at the sea fog bottom. These errors are due to the fact that perhaps the detected sea fog patches included low-level stratus with bottoms above the sea surface. Exp-C through Exp-E were the same as Exp-B except for different fog thicknesses. Assuming the thickness in Exp-B is Hfog, the thicknesses in Exp-C and Exp-D were 1.25 and 0.75 times Hfog, respectively, while Exp-E had the same top as Exp-B except that its fog bottom was lifted by 0.25Hfog. Exp-F was also the same as Exp-B, but the forecast field from the previous cycle (the result from WRF real.exe for the first cycle) was taken as the analysis instead of that from the first 3DVAR run to process MTSAT-derived humidity with elevation only (Zi in Fig. 6); thus, the second 3DVAR run was skipped.

Table 2.

List of experiments for the widespread sea fog case.

Table 2.

c. Assimilation effect on sea fog forecast

1) Sea fog diagnosis

The most important aspect of sea fog forecasting is the fog area. It is difficult to determine the sea fog area solely by a few observations from ships and buoys over the Yellow Sea. Here, the method described in section 2 to detect the horizontal distribution of sea fog is used to obtain observational information from the sea fog area (hereafter termed “observed sea fog”). The horizontal visibility at a certain height above sea surface (usually 10 m above sea level) was taken as the criterion to identify the sea fog area from numerical modeling results (hereafter termed “forecasted sea fog”). The visibility is calculated by the formula (Stoelinga and Thomas 1999)
e3
where is the visibility (km) and β is the extinction coefficient (km−1). In (3), β is related to cloud liquid water, rain, cloud ice, and snow. In this study, only cloud liquid water was included because no precipitation or snow were observed and no model cloud rain, snow, and ice in the low-level atmosphere were found. According to Kunkel (1984), β is related to cloud liquid water content (g m−3):
e4
where C is the mass concentration of cloud liquid water (CLW, g m−3). The region where ≤ 1 km was defined as the sea fog area, which is consistent with the World Meteorology Organization (WMO 1966) definition of fog. However, this forecasted sea fog, detected in terms of the visibility, is not suitable for the comparison with observed sea fog, since the latter is actually defined for cloud (sea fog or low-level stratus) tops instead of sea fog over the sea surface.

To make a fair comparison between forecasted sea fog and observed sea fog, the former was redefined to those areas where the following criteria were met: the CLW at the model’s lowest level is ≥0.016 g kg−1 or the cloud top is ≤400 m. With Eqs. (3) and (4), CLW ≥ 0.016 g kg−1 is equivalent to a visibility ≤ 1 km. The cloud top is calculated based on the threshold of CLW ≥ 0.016 g kg−1; and observations indicate that advection sea fogs are deeper than other types of fog, but rarely exceed 400 m (Zhou and Du 2010).

2) Evaluation method

A simple and direct way to visually compare forecasted sea fog and observed sea fog is the so-called eyeball method, which is qualitative and subjective. A more objective and comprehensive method is to mesh both the observed and forecasted sea fog areas onto the same grids in a given verification area, where point-to-point comparisons are conducted. Sea fog area prediction can be regarded as a binary event (yes or no; 1 or 0). Statistical scores, including the probability of detection (POD), false alarm ratio (FAR), bias, and equitable threat score (ETS) are used for evaluation. They are defined as follow:
e5
e6
e7
e8
where H, F, and O, respectively, refer to the numbers of correctly forecast points (hits), forecast points, and observed points with fog occurrence (i.e., binary value is 1); is a random hit penalty, and N is the total number of grid points in a verification domain.

The inner domain (i.e., D2 in Fig. 1) was taken as the verification domain, and its grid size was set to 0.1°, which is comparable to the 10-km resolution of the inner domain. Note that both the land and sea areas covered by high-altitude clouds were excluded from the verification domain. Verification was carried out for all model outputs at 1-h intervals.

3) Evaluation results

The forecasted sea fog for three typical times (i.e., 0600 UTC 6 March, 1800 UTC 6 March, and 0600 UTC 7 March) from Exp-A and Exp-B are shown in Fig. 7. Visually comparing the forecasted sea fog shown in Fig. 7 with the corresponding observed sea fog shown in Fig. 3 (cf. Fig. 3f with Figs. 7a,d, Fig. 3h with Figs. 7b,e, and Fig. 3j with Figs. 7c,f), it can be seen that the sea fog evolution was well captured by both Exp-A and Exp-B. However, the forecasted sea fog in Exp-A is smaller than the observed sea fog (cf. Figs. 7a–c and Figs. 3f–j, respectively); whereas in Exp-B, the sea fog simulation is significantly improved. The forecasted sea fog shown in Figs. 7d–f is much closer to the observed sea fog than that in Figs. 7a–c.

Fig. 7.
Fig. 7.

Example showing the forecast sea fog areas with fog-top heights (shading) from (top) Exp-A and (bottom) Exp-B for (left to right) 0600 UTC 6 Mar to 0600 UTC 7 Mar. A sketch of the transient cyclone (L)–anticyclone (H) couplet system is shown with 1000-hpa geopotential heights. The arrows in (a) indicate the couplet tracks in the next 24 h; and (b) the solid triangle marks the moored buoy site. The lines A–B in (b),(e) show the location for the vertical cross sections in Fig. 8.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Since the areas covered by high clouds have been excluded from the verification region, forecasted sea fogs located in those areas were correspondingly removed in the verification process. The temporal mean statistical scores (i.e., POD, FAR, bias, and ETS) for all of the experiments are shown in Table 3. Compared to Exp-A, Exp-B through Exp-F all show improvements to some extent for the above scores, except for FAR. However, the bias values are all less than 1.0. This result means that sea fog is not overpredicted. The ETS and POD improvements of Exp-F are only about half of those of Exp-B, indicating that it is important and necessary to run the 3DVAR cycle to get a high quality background, which is later used for generating sea fog humidity soundings. Results of Exp-C through Exp-E demonstrate that increasing sea fog top height and lifting sea fog bottom height both produce better gains compared with Exp-B; implying that there might exist errors in detecting sea fog information. However, these further gains were only 4.5%–8.2% for POD and only 2.1%–4.8% for ETS, respectively, even though the amplitude of increasing or lifting reached 25%.

Table 3.

Statistical scores of the experiments in the widespread sea fog case.

Table 3.

d. Impact of assimilation on physical fields

The model configuration and lateral conditions for all experiments were the same except for their initial conditions. The impact of assimilation on physical fields, including initial conditions and forecast errors, was analyzed in detail with a focus on Exp-A and Exp-B only.

The synoptic weather pattern over the Yellow Sea during the period 6–7 March 2006 is shown in Fig. 7. The sea fog is tightly controlled by a cyclone–anticyclone couplet system. The sea fog over the southern Yellow Sea gradually extended northeastward when the low pressure withdrew eastward and the high pressure approached the Yellow Sea (Fig. 7a; their tracks are indicated by arrows). During the 0800 UTC 6 March–0600 UTC 7 March period, high pressure had dominated the Yellow Sea (Figs. 7b and 7c). The high pressure simulated by Exp-B is stronger and a bit northeastward than that by Exp-A (cf. Figs. 7e,f and Figs. 7b,c, respectively). Thus, in Exp-B the sea fog expanded farther northeastward. This is one major reason why Exp-B had higher POD and ETS values than did Exp-A. By viewing the vertical cross sections along the line A–B that crossed the high pressure in Figs. 7b,e, it can be clearly seen how the high pressure affected the sea fog development. There existes a strong inversion layer and downward movement (Fig. 8) over the sea fog (area with CLW ≥ 0.016 g kg−1). It is apparent that both the inversion and downward movement in Exp-B (Fig. 8b) were stronger than those in Exp-A (Fig. 8a), which indicates that the inward-fog winds along the right edge of the sea fog become weaker after assimilating the sea fog humidity sounding data (see vectors in the dashed frames in Figs. 8a and 8b, respectively).

Fig. 8.
Fig. 8.

Vertical cross section along the line A–B in Fig. 7 at 1800 UTC 6 Mar 2006 for (a) Exp-A and (b) Exp-B. Colors and vectors show temperatures (°C) and winds (m s−1), respectively. Contours—solid, dashed, and dotted—show mixing ratios with values of 0.016, 0.1, and 0.2 g kg−1, respectively.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

In addition to the better-simulated sea fog area, measurements at a moored buoy also provide evidence to show that Exp-B outperformed Exp-A. At the buoy site (marked with solid triangles in both Fig. 7b and Fig. 1), a comparison of the measured and simulated temperature and RH was conducted. Figure 9 shows the results. Here, an RH value of 95% was used as the criterion for fog presence (Kim and Yum, 2010). If RH is less than this criterion, no fog forms. There is a dip with 6-h width and RH of 91% (no fog) in the RH measurement curve. A dip can also be seen in the simulated curves of Exp-A and Exp-B. The dip of Exp-A is too wide (24 h) and too deep (with an RH of 85%, no fog), while the dip of Exp-B has 6-h width and an RH of 93% depth (no fog), which agrees well with the measured but with a 5-h lag. For temperature, the Exp-B result is much closer to the measured than Exp-A, particularly during the first 24-h forecast.

Fig. 9.
Fig. 9.

Time series of air temperature (T, labeled curves ) and relative humidity (RH, unlabeled curves) measured and simulated at the buoy Idedo.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

The improvement of the Exp-B performance was due to its better initial conditions. The agreement between the initial analysis and the radiosonde observations (simply analysis error) over the verification domain were evaluated, focusing on bias and root-mean-square error (RMSE). There were 25 radiosonde stations in the verification domain. Figure 10 shows the average vertical profiles of temperature, wind, and moisture (represented by the water vapor mixing ratio) over all stations. For the temperature profile (Fig. 10b), Exp-B has a smaller RMSE than Exp-A, particularly below ~500 m. The profile of the υ-component wind shows a similar feature (Fig. 10d). However, the RMSE of the u-component wind in Exp-B is about 0.10.8 m s−1 larger than that in Exp-A (Fig. 10c). Furthermore, Exp-B has a larger RMSE and bias in the mixing ratio than does Exp-A below ~500 m, indicating that the MBL is about 0.5 g kg−1 wetter in Exp-B than in Exp-A due to the assimilation of sea fog humidity sounding data. Of particular note, is that Exp-A is about 0.2 g kg−1 drier than the observed below ~200 m.

Fig. 10.
Fig. 10.

Vertical profiles of the RMSE (solid lines) and biases (dashed lines) between the initial analysis and the radiosonde observations for Exp-A (curves no symbols) and Exp-B (curves with symbols) over the verification domain for (a) mixing ratio, (b) temperature, and (c) u and (d) υ wind components.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Since the assimilated humidity soundings are mostly distributed over the Yellow Sea, the MBL over the Yellow Sea should be under the strong influence of these assimilated soundings. Agreement between the initial analysis and the nine radiosonde observations around the Yellow Sea, which are marked with names in Fig. 1, were also evaluated. The results for the mixing ratio and temperature are shown in Fig. 11. Profiles of winds are not shown because they are quite similar to those in Figs. 10c,d. The improvement in temperature shown in Fig. 11b is greater than that in Fig. 10b. For example, the maximum value of the improved RMSE is about 1°C for the former while it is only about 0.6°C for the latter. Similar to the results shown in Fig. 10a, Exp-B still has a greater RMSE and bias than Exp-A (Fig. 11a), but the mixing ratio bias of Exp-A is about −0.2 g kg−1 below ~600 m, whereas it is about 0.3 g kg−1 in Exp-B. This result indicates that the MBL of Exp-A is much drier than observed, whereas Exp-B is wetter than the observations.

Fig. 11.
Fig. 11.

As in Fig. 10, but for only the named rawinsdonde stations in Fig. 1 and only for (a) mixing ratio and (b) temperature.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

The forecast agreement with the observations (simply forecast error) is an important index to assess the impact of assimilation on physical fields. Figure 12 illustrates the comparison of the 12-h forecast errors between Exp-A and Exp-B, including the vertical profiles of RMSE and bias for the mixing ratio, temperature, and wind. It is apparent that Exp-B performs better than Exp-A at almost all vertical levels. Additionally, the forecast errors at other times, for (e.g., 24, 36, and 48 h) were analyzed as well. Figure 13 shows the time series of average RMSE between the surface and 850 hPa. It is found that the RMSE is overall smaller in Exp-B than in Exp-A during the entire forecast period.

Fig. 12.
Fig. 12.

As in Fig. 10, but for 12-h forecast fields.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Fig. 13.
Fig. 13.

Comparison of the time series of the average RMSE from the surface to 850 hPa of the forecast fields and the radiosonde observations over the verification domain for Exp-A (solid line) and Exp-B (dashed line) of (a) mixing ratio, (b) temperature, and the (c) u and (d) υ wind components.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

The above results demonstrate clearly that assimilating sea fog humidity soundings can significantly improve the initial conditions for the MBL moisture, temperature, and wind. Once the moisture field is modified by assimilating the sea fog humidity sounding, the temperature and wind are adjusted accordingly by various interactions during the model integration.

e. A heavy costal sea fog event that covered a small area

As we have discussed previously, the aforementioned sea fog is a typical widespread event. The assimilation result shows that the forecast of the sea fog area is significantly improved, and the forecast errors of humidity and temperature in the MBL are greatly reduced. How does the extended cycling 3DVAR scheme affect a coastal sea fog forecast?

To discuss this issue, a small area of dense sea fog along the southern coast of Shandong Peninsula (see Fig. 1) was selected for another case study. Figure 14 shows the evolution of this sea fog from 1900 UTC 28 April to 0600 UTC 29 April 2007. The sea fog formed over the Qingdao sea area after midnight (0300 LST 29 April; Fig. 14a), then spread and moved northeastward slowly along the coast (Figs. 14b–f). Two numerical experiments were conducted for this sea fog: one with assimilation of obs only, the other with assimilation of both sea fog humidity soundings and obs (hereafter, they are referred to as Exp-OBS and Exp-RH, respectively; see Table 4). The model configuration is the same as in the widespread case (see Table 1), except that the model domains (shown in Fig. 1) are different and have finer horizontal resolutions of 10 and 2.5 km, respectively. Unlike the widespread case, which has a 12-h assimilation window, in this case a 9-h window from 1200 to 2100 UTC 28 April is set up for the cycling 3DVAR process with an hourly cycle interval based on the following considerations. On one hand, we had hoped to examine the assimilation effect using only a few sea fog humidity soundings. On the other hand, we needed radiosonde data to verify the improvement of the temperature and humidity in the MBL simulation. As shown in Fig. 14, the sea fog is in its initial stage before 2100 UTC 28 April. Three hours later (i.e., 0000 UTC 29 April), radiosonde data are available at Qingdao (QD) and Chengshantou (CS) stations (shown in Fig. 1). Within the assimilation window, the sea fog humidity soundings derived from the detected sea fog are only available at 1900 and 2000 UTC 28 April (Figs. 14a,b) and are assimilated in Exp-RH. The forecast is initialized at 2100 UTC 28 April and is integrated for 9 forecast hours.

Fig. 14.
Fig. 14.

Observed sea fog with top heights detected from MTSAT data from (top left) 1900 UTC 28 Apr to (bottom right) 0600 UTC 29 Apr 2007. The symbols are as in Fig. 3.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Table 4.

List of experiments for the narrowly spread coastal sea fog case.

Table 4.

Figure 15 shows the forecasted sea fog in Exp-OBS and Exp-RH. Exp-OBS completely fails to reproduce the sea fog; while Exp-RH successfully simulates the sea fog though its forecasted sea fog area is larger than that observed (Fig. 14). It is noteworthy that even after Doppler radar velocity and reflectivity data3 are assimilated for Exp-OBS, the sea fog cannot be well reproduced. A comparison of the 3-h forecasts with the radiosonde observations at 0000 UTC 29 April at QD and CS is shown in Fig. 16. Although an inversion layer is well represented (Figs. 16a,c), it doesn’t seem helpful for a successful simulation of sea fog by Exp-OBS. The simulated MBL below ~1 km is much drier than observation with a drying bias of ~1.5 g kg−1 at both QD and CS below ~100 m (Figs. 16b,d). The failure of Exp-OBS to adequately simulate sea fog is probably due to this insufficient amount of moisture in the MBL. Exp-RH does not improve temperatures very much in the MBL (Figs. 16a,c), but it remarkably corrects the moisture error below ~170 m (Figs. 16b,d), where the sea fog occurs (Figs. 15 and 16). The assimilation of MTSAT-derived humidity helps in the successful simulation of sea fog by Exp-RH. However, comparing Figs. 14 and 15, it is clear that the forecasted sea fog patch is much larger than observed. Analysis of wind fields in Exp-RH shows that the forecast onshore winds are ~3 m s−1 weaker than observed, which results in much less moistures being transported from ocean to land and hence an expansion of sea fog offshore.

Fig. 15.
Fig. 15.

Forecast sea fog with top height (shading) and 10-m winds (vectors) for (top) Exp-OBS and (bottom) Exp-RH for (left to right) 2100 UTC 28 Apr to 0600 UTC 29 Apr.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

Fig. 16.
Fig. 16.

Comparison of 3-h forecast vertical profiles of Exp-RH (dashed lines) and Exp-OBS (dotted lines) with radiosonde observations (solid lines) at (top) Qingdao and (bottom) Chengshantou stations. Shown are (a),(c) temperatures and (b),(d) mixing ratios.

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

4. Application

To further evaluate the effects and applicability of the proposed assimilation scheme, a suite of experiments were conducted for 10 additional sea fog cases that occurred over the Yellow Sea during 2008–12. All experiments have a 24-h forecast length and a 12-h assimilation window at 3-h intervals, and their initial times are listed in Table 5. To obtain a reliable statistical result, the initial time for each case was set to ensure that the observed sea fog during the forecast period was as clear as possible; that is, the sea fog area detected by the method described in section 2 or seen from visible satellite images were distinguishable. Once the initial time was determined, the previous 12 h was set to be the assimilation window. By looking at the observations (e.g., surface measurements and satellite images) during the assimilation window of each sea fog case, sea fogs were at the developing stage during the assimilation window for cases 1–8, and at the mature stage for cases 9 and 10. Large areas of dense high-altitude cloud were also found in case 10.

Table 5.

Statistical result of the experiments on the 10 sea fog cases (Q, mixing ratio; T, temperature). Letters indicate different sea fog development stages during the assimilation window (12 h before the listed time and date): developmental (D), mature (M), and high-altitude cloud overcast (H). The improvements (%) in group B relative to group A are in parentheses and set in boldface.

Table 5.

The experiments have the same model configuration as Exp-A (Table 1). For comparison purposes, they were divided into two groups. For one group, only obs were assimilated, just like Exp-A (hereafter group A). For the other group, both obs and sea fog humidity soundings were assimilated like Exp-B (hereafter group B). Similar to Exp-A, a new static background error covariance file was generated for each case by using the NMC method (Parrish and Derber 1992). Both forecasted sea fog and observed sea fog were obtained by the method described in section 2c. Given the fine-resolution domain (i.e., D2 in Fig. 1) as the verification region, the scores POD, FAR, bias, and ETS were calculated over the entire forecast period. Additionally, the average agreement between the surface and 850-hPa variables at the initial time and all of the radiosonde observations in D2 were also calculated. The results are outlined in Table 5, in which the cases are sorted by their ETS improvements of group B compared to that of group A.

It is apparent that group B performs much better than group A (Table 5). The average values in Exp-B for all scores and the RMSEs of mixing ratio (Q) and temperature (T) are improved, indicating that assimilation of sea fog humidity sounding produces a positive impact on the sea fog forecast. For all cases except case 10, the ETSs in group B are higher than those in group A. The ETSs of cases 1–8 are improved by at least 30%. In particular, sea fogs in cases 1 and 3 are nearly not predicted in group A with near-zero ETSs; but they are well represented in group B with ETS improvements of 1500%, 918%, and 733%, respectively. Comparing the columns of POD, ETS, Q, and T in Table 5, it can be seen that the ETS improvement can be mainly attributable to the POD improvement, and benefits from the improvements of both the mixing ratio and temperature in the MBL. In some cases FAR improvements are negative, but the maximum bias is no more than 1.56.

The impacts of mixing ratio and temperature assimilations are positive for almost all cases in group B. The assimilation of sea fog humidity soundings improves not only the moisture in the MBL but also the temperature. These improvements result in successful simulations of some sea fog patches, which are not predicted at all in group A due to insufficient MBL moisture. Figure 17 shows the average moisture biases within the surface to 850-hPa layer between the analysis and soundings over nine radiosonde sites (shown in Fig. 1) around the Yellow Sea. The MBL is drier than the observed in group A except for cases 2 and 8. However, these dry biases are all reduced to some extent in group B and even reversed to become wet biases in cases 5 and 8. It should be noted that sea fog in case 8 had already been successfully predicted in group A with the assimilation of obs only. The wet bias in case 2 decreases to near zero, while those of cases 5 and 8 increase to 0.2 and 0.3 g kg−1, respectively. The negative improvement of biases for cases 5 and 8 (see Table 5) is due to overestimated wet biases.

Fig. 17.
Fig. 17.

Average moisture biases from the surface to 850 hPa between the initial analysis and the observations at the nine named radiosonde stations around the Yellow Sea (see Fig. 1).

Citation: Weather and Forecasting 29, 2; 10.1175/WAF-D-12-00123.1

The longer the assimilation time window is, the greater number of sea fog humidity soundings will be assimilated. To investigate the impact of the assimilation window length on the assimilation effect, another two group experiments (i.e., groups C and D) were carried out. Group C is identical to group A and group D is identical to group B expect that the assimilation window is 6 h only. Their results are shown in Table 6. ETSs in group D except for case 10 are smaller than those in group B, which is understandable since the assimilated MTSAT-derived humidity data in group D are much less than those in group B. However, all ETS, Q, and T values are still improved in the group D results compared to those in group C. The improvements of the average values in Table 6 are generally comparable to those in Table 5.

Table 6.

As in Table 5, but the assimilation window is 6 h.

Table 6.

The results above clearly demonstrate that the proposed assimilation method is positively effective for sea fog prediction. Assimilation of sea fog humidity soundings will generally yield an improvement, no matter whether the assimilation window covers the entire initial stage, a partial initial stage, or the mature stage of a sea fog. In addition to the cycling 3DVAR scheme in this study, it is not difficult to apply this method to other data assimilation systems that are based on the 3DVAR method. It is noteworthy that two weaknesses—information of a predetected, existing sea fog event must be available before running the forecast, and the almost doubled computation cost—might limit the application of this method. However, since the entire system can run in a parallel mode and with increasing access to advanced supercomputers, the computation costs that previously prevented its operational application no longer need be a concern. Furthermore, sea fog layers are usually detected and processed into a sea fog humidity soundings right before the forecast is run; thus, the assimilation process is controllable and practical with sea fog information available in advance.

5. Conclusions and discussion

The formation and evolution of sea fogs over the Yellow Sea are strongly controlled by the development of the MBL. Due to the lack of routine observations over this vast sea area, it is hard to accurately depict the initial structure of an MBL, particularly the initial humidity profiles. As a follow-up to the sea fog satellite detection (Gao et al. 2009; Fu et al. 2011) and the cycling 3DVAR scheme (Gao et al. 2010), a method for sea fog satellite detection over the Yellow Sea has been improved in this study. An extended 3DVAR method has been developed to assimilate the MTSAT-derived humidity from the detected sea fog.

A series of experiments with different combinations of PBL and microphysics schemes were carried out. The YSU and PLIN schemes were chosen for the numerical modeling according to the results of these experiments. Two sea fog events over the Yellow Sea, one spread widely and the other spread narrowly along the coastal area, were studied using the detection and assimilation methods developed in this study. Model results were evaluated and the mechanisms for improvements to sea fog forecasting were analyzed in detail. Assimilation of MTSAT-derived sea fog humidity within the MBL is believed to be the major reason for the improvements. The extended 3DVAR method was further applied to 10 additional sea fog cases to evaluate its effectiveness. The main conclusions of this work are:

  1. For the widespread-fog case, the assimilation significantly improves the forecasting of sea fog areas, raising POD and ETS by up to about 20% and 15%, respectively. A realistic depiction of the MBL, particularly its moisture and temperature profiles, is achieved by assimilation of the MTSAT-derived humidity.
  2. Assimilation of MTSAT-derived humidity plays a key role in sea fog forecasting. Without such assimilation, the narrowly spread coastal sea fog event is completely missed.
  3. Results of the 10 sea fog cases studied have clearly demonstrated that the proposed assimilation method is effective. Assimilation of MTSAT-derived humidity not only improves the performance of sea fog forecasting but also makes the moisture and temperature structure in the MBL more realistic compared to that without assimilation. The average moisture and temperature simulations can be improved by about 16% and 10%, respectively. The average POD and ETS values can be improved by at least 60% and 70%, respectively. For some cases, similar to the narrowly spread coastal case, ETS can be improved dramatically by 300%–1500%.
  4. Assimilation of MTSAT-derived humidity will generally gain improvement, no matter whether the assimilation window covers the entire initial stage, a partial initial stage, or the mature stage of a sea fog. Since information of a predetected sea fog must be available before running a forecast, the proposed assimilation method is suitable for nowecasting of sea fog.

Despite the encouraging results in this study, several issues still need to be addressed in the future. The method for detecting sea fog will be improved with a focus on distinguishing low-level stratus decks from sea fog, which will help to get more accurate information about fog depth. The possibility of utilizing information provided by other geostationary satellite data sources (e.g., FY-2) need to be explored, and the capability to incorporate high spatial resolution and multichannel data of polar satellites need to be developed. It is also worth trying to incorporate satellite-derived humidity data into an advanced assimilation system. Within the assimilation windows, spatial spreading of humidity information related to the detected sporadic sea fog patches depends on circulation pattern; hence, it is better to employ flow-dependent background error covariance instead of the static one used in the current study. We will persue the possibility of solving this problem by applying the hybrid 3DVAR method (Wang et al. 2008a,b).

Sea fog forecasting is extremely sensitive to the initial conditions. The simulation results in this study, particularly for narrowly spread coastal fog events, clearly demonstrate the importance of initial conditions. In addition to errors in the initial conditions, insufficient physical–dynamical schemes in WRF may also contribute to simulation bias and sea fog forecasting failure. In real-time forecasting, such kinds of systematic errors or biases are hard to address by using a single modeling system. Thus, ensemble forecasting should also be considered in the future for sea fog events, especially multimodel ensemble forecasts. So far, little has been done to apply multimodel ensemble forecasting to sea fog studies. However, pioneering work using multimodel ensemble forecasts to study fog in general has been published recently (Zhou and Du 2010). Finally, this study clearly illustrates that MBL moisture is an important factor for successful sea fog simulations over the Yellow Sea, which will be helpful for designing ensemble studies. For instance, if a random perturbation method is used to generate an ensemble of the initial conditions, humidity perturbations in the MBL could be more intensely studied, and their perturbation amplitude could be determined based upon the analysis by the proposed assimilation method.

Acknowledgments

This research was financially supported by the National Natural Science Foundation of China (41276009 and 41275049), the National Special Fund for Public Sector Research of China (GYHY201106006). The computation work in this paper was supported by the Ocean University of China Center for High Performance Computing. The data management agencies, Kochi University, NCEP, and NEAR-GOOS, are greatly appreciated. Last but not least, the suggestions from Drs. Yanluan Lin and Yonghua Chen, as well as two anonymous reviewers, gave us opportunities to improve our final version.

REFERENCES

  • Ahn, M. H., , Sohn E. H. , , and Hwang B. J. , 2003: A new algorithm for sea fog/stratus detection using GMS-5 IR data. Adv. Atmos. Sci., 20, 899913.

    • Search Google Scholar
    • Export Citation
  • Anderson, J., 1931: Observations from airplanes of cloud and fog conditions along the southern California coast. Mon. Wea. Rev., 59, 264270.

    • Search Google Scholar
    • Export Citation
  • Ballard, S. P., , Golding B. W. , , and Smith R. N. B. , 1991: Mesoscale model experimental forecasts of the haar of northeast Scotland. Mon. Wea. Rev., 119, 21072123.

    • Search Google Scholar
    • Export Citation
  • Bendix, J., , Thies B. , , Cermak J. , , and Nauß T. , 2005: Ground fog detection from space based on MODIS daytime data—A feasibility study. Wea. Forecasting, 20, 989996.

    • Search Google Scholar
    • Export Citation
  • Byers, H. R., 1930: Summer Sea Fogs of the Central California Coast. University of California Press, 338 pp.

  • Chen, F., , and Dudhia J. , 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model description and implementation. Mon. Wea. Rev., 129, 569585.

    • Search Google Scholar
    • Export Citation
  • Cho, Y. K., , Kim M. O. , , and Kim B. C. , 2000: Sea fog around the Korean Peninsula. J. Appl. Meteor., 39, 24732479.

  • Coakley, J. A., , and Bretherton F. P. , 1982: Cloud cover from high-resolution scanner data: Detection and allowing for partially filled fields of view. J. Geophys. Res., 87 (C7), 49174932.

    • Search Google Scholar
    • Export Citation
  • Deng, J., , Bai J. , , Liu J. W. , , Wang X. W. , , and Shi H. Y. , 2006: Detection of daytime fog using MODIS multispectral data (in Chinese). Mater. Sci. Technol., 34, 188193.

    • Search Google Scholar
    • Export Citation
  • Ellrod, G. P., 1995: Advances in the detection and analysis of fog at night using GOES multispectral infrared imagery. Wea. Forecasting, 10, 606619.

    • Search Google Scholar
    • Export Citation
  • Findlater, J., , Roach W. T. , , and McHugh B. C. , 1989: The haar of north-east Scotland. Quart. J. Roy. Meteor. Soc., 115, 581608.

  • Fitzpatrick, M. F., , Brandt R. E. , , and Warren S. G. , 2004: Transmission of solar radiation by clouds over snow and ice surfaces: A parameterization in terms of optical depth, solar zenith angle, and surface albedo. J. Climate, 17, 266275.

    • Search Google Scholar
    • Export Citation
  • Fu, G., , Zhang M. G. , , Duan Y. H. , , Zhang T. , , and Wang J. Q. , 2004: Characteristics of sea fog over the Yellow Sea and the East China Sea. Kaiyo Mon.,38, 99–108.

  • Fu, G., , Guo J. , , Xie S. P. , , Duan Y. , , and Zhang M. , 2006: Analysis and high-resolution modeling of a dense sea fog event over the Yellow Sea. Atmos. Res., 81, 292303.

    • Search Google Scholar
    • Export Citation
  • Fu, G., , Li P. Y. , , Crompton J. G. , , Guo J. , , Gao S. H. , , and Zhang S. P. , 2010: An observational and modeling study of a sea fog event over the Yellow Sea on 1 August 2003. Meteor. Atmos. Phys., 107, 149159.

    • Search Google Scholar
    • Export Citation
  • Fu, G., , Xu J. , , and Zhang S. Q. , 2011: Comparison of modeling atmospheric visibility with visible satellite imagery (in Chinese). J. Ocean Univ. China,41, 001–010.

  • Fu, G., , Zhang S. P. , , Gao S. H. , , and Li P. Y. , 2012: Understanding of Sea Fog over the China Seas. China Meteorological Press, 220 pp.

  • Gao, S. H., , Lin H. , , Shen B. , , and Fu G. , 2007: A heavy sea fog event over the Yellow Sea in March 2005: Analysis and numerical modeling. Adv. Atmos. Sci., 24, 6581.

    • Search Google Scholar
    • Export Citation
  • Gao, S. H., , Wu W. , , Zhu L. L. , , Fu G. , , and Huang B. , 2009: Detection of nighttime sea fog/stratus over the Huanghai Sea using MTSAT-1R IR data. Acta Oceanol. Sin., 28, 2335.

    • Search Google Scholar
    • Export Citation
  • Gao, S. H., , Qi Y. L. , , Zhang S. B. , , and Fu G. , 2010: Initial conditions improvement of sea fog numerical modeling over the Yellow Sea by using cycling 3DVAR. Part I: WRF numerical experiments (in Chinese). J. Ocean Univ. China,40, 001–009.

  • Gultepe, I., and Coauthors, 2007: Fog research: A review of past achievements and future perspectives. Pure Appl. Geophys., 164, 11211159.

    • Search Google Scholar
    • Export Citation
  • Heidinger, A. K., , and Stephens G. L. , 2000: Molecular line absorption in a scattering atmosphere. Part II: Application to remote sensing in the O2 a band. J. Atmos. Sci., 57, 16151634.

    • Search Google Scholar
    • Export Citation
  • Hong, S. Y., , and Lim J. O. J. , 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129151.

  • Hong, S. Y., , Noh Y. , , and Dudhia J. , 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341.

    • Search Google Scholar
    • Export Citation
  • Hu, J. F., , Guo K. C. , , and Yan L. N. , 1996: Discriminate prediction of marine fog occurrence using a model output statistics scheme (in Chinese). J. Ocean Univ. Qingdao, 26, 439445.

    • Search Google Scholar
    • Export Citation
  • Hu, R. J., , and Zhou F. X. , 1997: A numerical study of the effects of air–sea conditions on the process of sea fog (in Chinese). J. Ocean Univ. Qingdao, 27, 282290.

    • Search Google Scholar
    • Export Citation
  • Hu, R. J., , and Zhou F. X. , 1998: Effects of advection, turbulence and radiation on formation of sea fog (in Chinese). J. Ocean Univ. Qingdao, 20, 2530.

    • Search Google Scholar
    • Export Citation
  • Huang, J., , and Zhou F. X. , 2006: The cooling and moistening effect on the formation of sea fog in the Huanghai Sea. Acta Oceanol. Sin., 25, 4962.

    • Search Google Scholar
    • Export Citation
  • Hunt, G. E., 1973: Radiative properties of terrestrial clouds at visible and infrared thermal window wavelengths. Quart. J. Roy. Meteor. Soc., 99, 346369.

    • Search Google Scholar
    • Export Citation
  • Iacono, M. J., , Delamere J. S. , , Mlawer E. J. , , Shephard M. W. , , Clough S. A. , , and Collins W. D. , 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, doi:10.1029/2008JD009944.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., , and Fritsch J. M. , 1990: A one-dimensional entraining/detraining plume model and its application in convective parameterization. J. Atmos. Sci., 47, 27842802.

    • Search Google Scholar
    • Export Citation
  • Kästner, M., , Kriebel K. T. , , Meerkötter R. , , Renger W. , , Ruppersberg G. H. , , and Wendling P. , 1993: Comparison of cirrus height and optical depth derived from satellite and aircraft measurements. Mon. Wea. Rev., 121, 27082715.

    • Search Google Scholar
    • Export Citation
  • Kim, C. K., , and Yum S. S. , 2010: Local meteorological and synoptic characteristics of fogs formed over Incheon international airport in the west coast of Korea. Adv. Atmos. Sci., 27, 761776.

    • Search Google Scholar
    • Export Citation
  • Kim, C. K., , and Yum S. S. , 2012: A numerical study of sea-fog formation over cold sea surface using a one-dimensional turbulence model coupled with the Weather Research and Forecasting Model. Bound.-Layer Meteor., 143, 481505.

    • Search Google Scholar
    • Export Citation
  • Koračin, D., , Lewis J. , , Thompson W. T. , , Dorman C. E. , , and Businger J. A. , 2001: Transition of stratus into fog along the California coast: Observation and modeling. J. Atmos. Sci., 58, 17141731.

    • Search Google Scholar
    • Export Citation
  • Koračin, D., , Businger J. A. , , Dorman C. E. , , and Lewis J. M. , 2005a: Formation, evolution, and dissipation of coastal sea fog. Bound.-Layer Meteor., 117, 447478.

    • Search Google Scholar
    • Export Citation
  • Koračin, D., , Leipper D. F. , , and Lewis J. M. , 2005b: Modeling sea fog on the U.S. California coast during a hot spell event. Geofizika, 22, 5982.

    • Search Google Scholar
    • Export Citation
  • Kunkel, B., 1984: Parameterization of droplet terminal velocity and extinction coefficient in fog models. J. Climate Appl. Meteor., 23, 3441.

    • Search Google Scholar
    • Export Citation
  • Lee, T. F., , Turk F. J. , , and Richardson K. , 1997: Stratus and fog productions using GOES-8–9 3.9-μm data. Wea. Forecasting, 12, 606619.

    • Search Google Scholar
    • Export Citation
  • Leipper, D., 1948: Fog development at San Diego, California. J. Mar. Res., 7, 337346.

  • Lewis, J. M., , Koračin D. , , Rabin R. , , and Businger J. , 2003: Sea fog off the California coast: Viewed in the context of transient weather systems. J. Geophys. Res., 108, 4457, doi:10.1029/2002JD002833.

    • Search Google Scholar
    • Export Citation
  • Lewis, J. M., , Koračin D. , , and Redmond K. T. , 2004: Sea fog research in the United Kingdom and United States. Bull. Amer. Meteor. Soc., 85, 395408.

    • Search Google Scholar
    • Export Citation
  • Li, R., , Gao S. H. , , and Wang Y. M. , 2012: Numerical study on direct assimilation of satellite radiances for sea fog over the Yellow Sea (in Chinese). J. Ocean Univ. China,42, 10–20.

  • Liang, W. F., , and Hou Z. X. , 2001: The characteristics and forecasts of heavy fog occurred over Qingdao (in Chinese). Shandong Meteor., 84, 1117.

    • Search Google Scholar
    • Export Citation
  • Lim, K. S. S., , and Hong S. Y. , 2010: Development of an effective double-moment cloud microphysics scheme with prognostic cloud condensation nuclei (CCN) for weather and climate models. Mon. Wea. Rev., 138, 15871612.

    • Search Google Scholar
    • Export Citation
  • Lin, Y. L., , Farley R. D. , , and Oriville H. D. , 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl. Meteor., 22, 10651092.

    • Search Google Scholar
    • Export Citation
  • Liu, X., , and Hu X. Q. , 2008: Sea fog automatic detection over the East China Sea using MTSAT data (in Chinese). J. Oceanogr. Taiwan, 27, 112117.

    • Search Google Scholar
    • Export Citation
  • Nakanishi, M., , and Niino H. , 2006: An improved Mellor–Yamada level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397407.

    • Search Google Scholar
    • Export Citation
  • Nicholls, S., 1984: The dynamics of stratocumulus: Aircraft observation and comparison with a mixed layer model. Quart. J. Roy. Meteor. Soc., 110, 783820.

    • Search Google Scholar
    • Export Citation
  • Otkin, J. A., , and Greenwald T. J. , 2008: Comparison of WRF model-simulated and MODIS-derived cloud data. Mon. Wea. Rev., 136, 19571970.

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

    • Search Google Scholar
    • Export Citation
  • Petterssen, S., 1936: On the causes and the forecasting of the California fog. J. Aeronaut. Sci., 3, 305309.

  • Petterssen, S., 1938: On the causes and the forecasting of the California fog. Bull. Amer. Meteor. Soc., 19, 4955.

  • Skamarock, W., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN–475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3_bw.pdf.]

  • Sorli, B., , Pascal-Delannoy F. , , Giani A. , , Foucaran A. , , and Boyer A. , 2002: Fast humidity sensor for high range 80%–95% RH. Sens. Actuators, 100A, 2431.

    • Search Google Scholar
    • Export Citation
  • Stoelinga, M. T., , and Thomas T. W. , 1999: Nonhydrostatic, mesobeta-scale model simulations of cloud ceiling and visibility for an East Coast winter precipitation event. J. Appl. Meteor., 38, 385403.

    • Search Google Scholar
    • Export Citation
  • Taylor, G., 1917: The formation of fog and mist. Quart. J. Roy. Meteor. Soc., 43, 241268.

  • Thompson, G., , Field P. R. , , Rasmussen R. M. , , and Hall W. D. , 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 50955115.

    • Search Google Scholar
    • Export Citation
  • Trémant, M., 1987: La prèvision du brouilliard en mer. Meteorologie Maritime et Activities, Oceanograpiques Connexes Paport, Vol. 20, WMO, 127 pp.

  • Wang, B. H., 1985: Sea Fog. China Ocean Press, 330 pp.

  • Wang, X., , Barker D. , , Snyder C. , , and Hamill T. M. , 2008a: A hybrid ETKF–3DVAR data assimilation scheme for the WRF model. Part I: Observing system simulation experiment. Mon. Wea. Rev., 136, 51165131.

    • Search Google Scholar
    • Export Citation
  • Wang, X., , Barker D. , , Snyder C. , , and Hamill T. M. , 2008b: A hybrid ETKF–3DVAR data assimilation scheme for the WRF model. Part II: Real observation experiments. Mon. Wea. Rev., 136, 51325147.

    • Search Google Scholar
    • Export Citation
  • WMO, 1966: International Meteorological Vocabulary.World Meteorological Organization, 276 pp.

  • Yoo, J. M., , Jeong M. J. , , Hur Y. M. , , and Shin D. B. , 2010: Improved fog detection from satellite in the presence of clouds. Asia-Pac. J. Atmos. Sci., 46, 2940.

    • Search Google Scholar
    • Export Citation
  • Zhang, H. Y., , Zhou F. X. , , and Zhang X. H. , 2005: Interannual change of sea fog over the Yellow Sea in spring (in Chinese). Oceanol. Limnol. Sin., 36, 3642.

    • Search Google Scholar
    • Export Citation
  • Zhang, S. P., , and Bao X. W. , 2008: The main advances in sea fog research in China (in Chinese). J. Ocean Univ. China, 38, 359366.

  • Zhang, S. P., , Xie S. P. , , Liu Q. Y. , , Yang Y. Q. , , Wang X. G. , , and Ren Z. P. , 2009: Seasonal variations of Yellow Sea fog: Observations and mechanisms. J. Climate, 22, 67586772.

    • Search Google Scholar
    • Export Citation
  • Zhao, Y. P., , Chen Y. L. , , and Wang P. G. , 1997: Analysis of atmospheric and oceanic conditions for marine fog formation over the Yellow Sea and East China Seas (in Chinese). Stud. Mar. Sin., 38, 6979.

    • Search Google Scholar
    • Export Citation
  • Zhou, B., , and Du J. , 2010: Fog prediction from a multimodel mesoscale ensemble prediction system. Wea. Forecasting, 25, 303322.

  • Zhou, F. X., , and Liu L. T. , 1986: Comprehensive survey and research report on the water areas adjacent to the Changjing River estuary and Chejudo Island marine fog (in Chinese). J. Shandong Coll. Oceanol., 16, 114131.

    • Search Google Scholar
    • Export Citation
  • Zhou, F. X., , Wang X. , , and Bao X. W. , 2004: Climatic characteristics of sea fog formation of the Huanghai Sea in spring (in Chinese). Acta Oceanol. Sin., 26, 2837.

    • Search Google Scholar
    • Export Citation
1

Most obs data are usually at an interval of 3 h. We also conducted a modeling simulation with an hourly interval, and found that the result was almost the same as that with a 3-hourly interval.

2

The MTSAT data at 1800 UTC 5 Mar 2006 are absent.

3

The Doppler radar data come from the Qingdao Meteorology Observatory.

Save