• Aberson, S. D., 2008: Large forecast degradations due to synoptic surveillance during the 2004 and 2005 hurricane seasons. Mon. Wea. Rev., 136, 31383150.

    • Search Google Scholar
    • Export Citation
  • Chen, Y., , and Snyder C. , 2007: Assimilating vortex position with an ensemble Kalman filter. Mon. Wea. Rev., 135, 18281845.

  • Dvorak, V. F., 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103, 420430.

  • Dvorak, V. F., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Tech. Rep. NESDIS 11, 47 pp.

  • Guard, C. P., , Carr L. E. , , Wells F. H. , , Jeffries R. A. , , Gural N. D. , , and Edson D. K. , 1992: Joint Typhoon Warning Center and the challenges of multibasin tropical cyclone forecasting. Wea. Forecasting, 7, 328352.

    • Search Google Scholar
    • Export Citation
  • Hamill, T. M., , Whitaker J. S. , , Fiorino M. , , and Benjamin S. G. , 2011: Global ensemble predictions of 2009’s tropical cyclones initialized with an ensemble Kalman filter. Mon. Wea. Rev., 139, 668688.

    • Search Google Scholar
    • Export Citation
  • Kleist, D. T., , Parrish D. F. , , Derber J. C. , , Treadon R. , , Errico R. M. , , and Yang R. , 2009a: Improving incremental balance in the GSI 3DVAR analysis system. Mon. Wea. Rev., 137, 10461060.

    • Search Google Scholar
    • Export Citation
  • Kleist, D. T., , Parrish D. F. , , Derber J. C. , , Treadon R. , , Wu W.-S. , , and Lord S. J. , 2009b: Introduction of the GSI into the NCEP Global Data Assimilation System. Wea. Forecasting, 24, 16911705.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., , Correa-Torres R. , , Rohaly G. , , and Oosterhof D. , 1997: Physical initialization and hurricane ensemble forecasts. Wea. Forecasting, 12, 503514.

    • Search Google Scholar
    • Export Citation
  • Liu, Q., , Marchok T. , , Pan H.-L. , , Bender M. , , and Lord S. J. , 2000: Improvements in hurricane initialization and forecasting at NCEP with global and regional (GFDL) models. NWS Tech. Procedures Bull. 472, 7 pp. [Available online at http://www.nws.noaa.gov/om/tpb/472.htm.]

    • Search Google Scholar
    • Export Citation
  • Lord, S. J., 1991: A bogusing system for vortex circulations in the National Meteorological Center Global Forecast Model. Preprints, 19th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 328–330.

    • Search Google Scholar
    • Export Citation
  • Pu, Z., , and Braun S. A. , 2001: Evaluation of bogus vortex techniques with four-dimensional variational data assimilation. Mon. Wea. Rev., 129, 20232039.

    • Search Google Scholar
    • Export Citation
  • Rappaport, E. N., and Coauthors, 2009: Advances and challenges at the National Hurricane Center. Wea. Forecasting, 24, 395419.

  • Serrano, E., , and Undén P. , 1994: Evaluation of a tropical cyclone bogusing method in data assimilation and forecasting. Mon. Wea. Rev., 122, 15231547.

    • Search Google Scholar
    • Export Citation
  • Torn, R. D., 2010: Performance of a mesoscale ensemble Kalman filter (EnKF) during the NOAA High-Resolution Hurricane Test. Mon. Wea. Rev., 138, 43754392.

    • Search Google Scholar
    • Export Citation
  • Wu, C.-C., , Lien G.-Y. , , Chen J.-H. , , and Zhang F. , 2010: Assimilation of tropical cyclone track and structure based on the ensemble Kalman filter (EnKF). J. Atmos. Sci., 67, 38063822.

    • Search Google Scholar
    • Export Citation
  • Wu, W.-S., , Parrish D. F. , , and Purser R. J. , 2002: Three-dimensional variational analysis with spatially inhomogeneous covariances. Mon. Wea. Rev., 130, 29052916.

    • Search Google Scholar
    • Export Citation
  • Zou, X., , and Xiao Q. , 2000: Studies on the initialization and simulation of a mature hurricane using a variational bogus data assimilation scheme. J. Atmos. Sci., 57, 836860.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Sea level pressure (contour, hPa) and lowest model-level wind speed (shaded, m s−1) analysis valid at 0000 UTC 4 Sep 2008 for the (top) operational T382 GFS, (middle) T574 GFS control, and (bottom) T574 experiments. The advisory intensities for Hanna (westward storm) and Ike (eastward storm) at this time were 989 hPa (26 m s−1) and 956 hPa (54 m s−1), respectively.

  • View in gallery

    As in Fig. 1, but for the 72-h forecast initialized at 0000 UTC 4 Sep 2008. The triangle represents the National Hurricane Center (NHC) best-track position for Hurricane Ike, which had an advisory intensity of 948 hPa (59 m s−1) at this verification time.

  • View in gallery

    The average T574 GFS tropical cyclone track error (km) for the control (black) and experimental (gray) runs for the period covering 1 Jul–10 Nov 2008. The average track error was computed from a homogeneous sample of cases for storms in the Atlantic, eastern Pacific, and western Pacific basins. The number of cases for each lead time is identified below each forecast hour. Error bars indicate the 5th and 95th percentiles of a resampled block bootstrap distribution.

  • View in gallery

    As in Fig. 3, but for the average GFS bias in the tropical storm near-surface maximum wind (m s−1).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 22 22 4
PDF Downloads 22 22 4

Assimilation of Tropical Cyclone Advisory Minimum Sea Level Pressure in the NCEP Global Data Assimilation System

View More View Less
  • 1 Environmental Modeling Center, National Centers for Environmental Prediction, Camp Springs, Maryland
© Get Permissions
Full access

Abstract

The assimilation of official advisory minimum sea level pressure observations has been developed and tested in the National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) to address forecaster concerns regarding some tropical systems being far too weak in operational Global Forecast System (GFS) analyses. The assimilation of these observations has been operational in the GFS since December 2009. Using the T574 version of the NCEP GFS model, it is demonstrated that the assimilation of these observations results in a substantial reduction in the initial intensity bias for tropical systems, resulting in improved track and intensity guidance for lead times out to 5 days.

Additional affiliation: I. M. Systems Group, Camp Springs, Maryland.

Corresponding author address: Daryl T. Kleist, NOAA Science Center, No. 207, 5200 Auth Rd., Camp Springs, MD 20746-4304. E-mail: daryl.kleist@noaa.gov

Abstract

The assimilation of official advisory minimum sea level pressure observations has been developed and tested in the National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) to address forecaster concerns regarding some tropical systems being far too weak in operational Global Forecast System (GFS) analyses. The assimilation of these observations has been operational in the GFS since December 2009. Using the T574 version of the NCEP GFS model, it is demonstrated that the assimilation of these observations results in a substantial reduction in the initial intensity bias for tropical systems, resulting in improved track and intensity guidance for lead times out to 5 days.

Additional affiliation: I. M. Systems Group, Camp Springs, Maryland.

Corresponding author address: Daryl T. Kleist, NOAA Science Center, No. 207, 5200 Auth Rd., Camp Springs, MD 20746-4304. E-mail: daryl.kleist@noaa.gov

1. Introduction

Deterministic numerical weather prediction (NWP) guidance for tropical cyclone prediction, particularly for track forecasting, has steadily improved for the past several decades. Improvements in NWP guidance have resulted in a substantial increase in the skill of the National Hurricane Center track forecasts (Rappaport et al. 2009). Many factors have contributed to the noted improvement including advanced data assimilation algorithms, new observing systems, improved modeling techniques (including physical parameterizations), and advancements in computing power allowing for higher spatial resolution.

Initializing a representative vortex in the correct position and of appropriate intensity remains a serious challenge. Historically, operational global NWP models have been run at resolutions such that they are unable to represent the scale and intensity of tropical cyclones, as is the case with the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) model. Initializing a realistic vortex that is representative of the scales that are resolvable and consistent with the model dynamics can prove difficult.

In addition to the (scale) representativeness issue, there remains the problem of a lack of observations within the storm environment itself. This is especially true for systems that do not have reconnaissance missions sent out to collect observations. Although there is a wealth of satellite observations taken in or near a tropical storm for any given initialization time, most of these observations are currently not assimilated into operational NWP models because of cloud and rain contamination issues. The assimilation of cloudy (particularly infrared) and rain-affected radiances remains an area of active research. Even those storms for which there are dropsonde observations available, most of the observations within the near-storm environment are not assimilated, due to the insufficient resolution and potential for those observations to have a detrimental impact on the subsequent forecasts (Aberson 2008).

As computing power increases and operational NWP centers are able to run models at higher spatial resolution, there comes a necessity to be able to initialize more realistic vortex structures for tropical cyclone prediction. The current operational global models are not run at sufficiently high resolution to fully resolve tropical cyclones.1 Despite the fact that these models are at relatively coarse resolution for tropical cyclone prediction, they are able to maintain fairly realistic storms in terms of intensity (especially those systems that are of larger scale). The systems being modeled are to the point that they are realistic enough to modify the near-storm environment, and are more than simply being advected by the large-scale, mean steering currents in the model forecast.

Historically, tropical cyclone initialization has been accomplished through a variety of means. One such method is to use a vortex bogusing scheme, where an artificial vortex, prescribed through some combination of empirical or dynamical methods, is placed into a large-scale background field (Lord 1991; Serrano and Undén 1994; Krishnamurti et al. 1997). Another method involves generating and assimilating synthetic observations (pressure, temperature, and wind) based upon some assumed or prescribed vortex (Zou and Xiao 2000; Pu and Braun 2001). The assimilation of synthetic observations in combination with real observations results in an analysis of a more realistic vortex. Finally, to correct for position errors, methods of vortex relocation have been developed (Liu et al. 2000). These methods are designed to extract a misplaced vortex from the model-forecasted large-scale flow and, then, place the vortex back at the observed position prior to data assimilation. So unlike the bogus vortex and synthetic observation methods, the relocation methods have an advantage in the fact that they maintain the vortex in a state that is consistent with the scale and dynamics of the numerical model.

Each of the initialization methods relies on some estimate of the position of the center of the tropical cyclone, as well as perhaps some estimate of its intensity. It is standard operational procedure for the operational agencies responsible for monitoring tropical cyclone activity (i.e., NCEP’s National Hurricane Center and Joint Typhoon Warning Center) to issue regular estimates of position and intensity for tropical cyclones. This information is typically estimated through a combinational of available satellite estimates (Dvorak 1975, 1984), as well as observations retrieved on reconnaissance missions if available (Guard et al. 1992; Rappaport et al. 2009). These estimates provide another source of information (observations) to be assimilated into an NWP model. The assimilation of tropical cyclone position and intensity has been demonstrated to be successful in a variety of previous work (Chen and Snyder 2007; Torn 2010; Wu et al. 2010). Along these lines and motivated by concerns from forecasters at the National Hurricane Center about tropical cyclone analyses and forecasts being too weak in operational GFS forecasts (2008), the assimilation of minimum sea level pressure (MinSLP) from the tropical cyclone official advisories into the NCEP Global Data Assimilation System (GDAS) was pursued.

The rest of the manuscript is organized as follows. Section 2 provides a description of the NCEP GDAS–GFS system and supplemental MinSLP observations. Section 3 describes the impacts of assimilating the MinSLP observations in the GFS–GDAS system, which is followed by a brief summary.

2. Configuration

The NCEP GFS model used in this study is the same version that became operational in July 2010. This version of the GFS is a T574 spectral model (triangular truncation) and has 64 hybrid sigma-pressure vertical levels. The July 2010 GFS implementation contained considerable changes to the model physics. [A description of the GFS model version 9.0.0 is available online from NCEP’s Environmental Modeling Center (EMC; http://www.emc.ncep.noaa.gov/GFS/doc.php).] The same version of the model is used for both cycling within the GDAS as well as making 2-week GFS forecasts.

Data assimilation is performed within the GDAS using the gridpoint statistical interpolation (GSI), a physical space three-dimensional variational data assimilation (3DVAR) analysis scheme (Wu et al. 2002; Kleist et al. 2009b). A variety of observations are assimilated into the analysis including radiosondes, surface pressure, ship/buoy reports, reconnaissance dropsondes, satellite-derived atmospheric motion vectors, wind profilers, ozone retrievals, GPS radio occultation, and a variety of microwave and infrared satellite radiances (Kleist et al. 2009b). Very few observations within a tropical cyclone are assimilated [though all of the National Oceanic and Atmospheric Administration (NOAA) Gulfstream IV (G-IV) based environment dropsondes that are part of synoptic surveillance are assimilated].

Prior to assimilation, a vortex relocation procedure is performed on the short-term GFS forecast that is used as the first guess (Liu et al. 2000). Part of this process involves applying tracking software to identify the positions of storms within the short-term forecast. The tracking software attempts to identify all storms that are of at least tropical depression strength (as identified in the official advisories). Despite the advances in data assimilation and NWP, there are occasionally cases in which the operational short-term forecasts are unable to develop observed (weak) tropical systems. In the event there is no identifiable system within the short-term forecast consistent with the official advisory, synthetic wind observations are generated (assuming an axisymmetric vortex consistent with the observed intensity and size) and assimilated. As the spatial resolution of the GFS continues to increase and as advances continue to be made in the assimilation of observations, the number of instances in which this actually occurs operationally has been decreasing.

To supplement the relocation procedure and synthetic observations, a method of assimilating official advisory MinSLP observations simultaneously with the regular observing system was developed and tested for the GDAS–GFS system. The MinSLP observations are treated like regular surface pressure observations and assimilated for any storm that is identified to be of at least tropical depression strength. Standard quality control is relaxed to ensure the observations are not tossed out by a gross check when there are large deviations between the background and observation. The GSI currently uses a simple model for the multivariate components (wind, temperature, and surface pressure) of the background error covariance (Kleist et al. 2009b), which may not necessarily be appropriate for the background errors associated with a tropical cyclone. The spatial scale of the impacts on the analysis through the assimilation of the MinSLP observations is controlled by the background error specification, set to be consistent with the scales of the GFS model first-guess fields.

Because a single surface pressure observation is assimilated without supplementary wind observations, a large portion of the signal projects onto gravity modes. The tangent linear normal-mode constraint (Kleist et al. 2009a) that is run routinely as part of the global GSI is designed to correct for this type of signal and minimize the potential negative impacts. To counter this effect, the observation error for MinSLP observations was set to be between 0.75 and 2.25 hPa,2 similar to the specification for normal surface pressure observations (which are assigned errors between 1.0 and 1.6 hPa in the GDAS). In addition to the filtering of some of the large pressure signal, the constraint also acts to induce a more balanced wind and temperature increment (though the balance enforced may not be entirely appropriate for tropical cyclones).

3. Analysis and forecast impacts

Although the assimilation of MinSLP observations was originally tested and implemented into an older, lower-resolution version of the GFS model (implemented in December 2009), a clean experiment using the July 2010 version of the GFS was run and evaluated. The control and experiment were started from the same initial files on 20 June 2008, and cycled on their own independently through 10 November 2008. Both the control and experiment runs of the GFS assimilated all operational (at the time) observations, with the additional advisory MinSLP observations added to supplement the observing system in the experimental run. The control and experiment runs both applied the vortex relocation of the first-guess procedure prior to data assimilation.

An example of a storm being much too weak in operational GFS analyses can be seen with Hurricane Ike from 2008, where the operational GFS had an analyzed storm with a MinSLP of 1008 hPa at 0000 UTC 4 September 2008 (Fig. 1, top) while the advisory valid at that same time for Hurricane Ike was 956 hPa. Due to inadequate spatial resolution and a lack of observations within the storm itself, the operational GFS was unable to capture the rapid deepening of this intense cyclone (deepening from a MinSLP of 989 hPa to 956 hPa in the previous 12 h).

Fig. 1.
Fig. 1.

Sea level pressure (contour, hPa) and lowest model-level wind speed (shaded, m s−1) analysis valid at 0000 UTC 4 Sep 2008 for the (top) operational T382 GFS, (middle) T574 GFS control, and (bottom) T574 experiments. The advisory intensities for Hanna (westward storm) and Ike (eastward storm) at this time were 989 hPa (26 m s−1) and 956 hPa (54 m s−1), respectively.

Citation: Weather and Forecasting 26, 6; 10.1175/WAF-D-11-00045.1

The T574 control run of the GFS exhibited the same difficulty in capturing the rapid deepening of this storm and analyzed a storm that was very similar to the (then) operational GFS despite better spatial resolution in the model (Fig. 1, middle). The T574 experimental run that included the assimilation of the MinSLP observations did have a stronger vortex for that same analysis (Fig. 1, bottom), with an analyzed MinSLP 6 hPa deeper than the control. Although the new procedure did not produce a strong enough analysis of the tropical cyclone itself, it was certainly a conservative step in the right direction. The cyclone in the experimental run was slightly stronger than the control and allowed for a deeper storm that the model was able to maintain and more rapidly intensify in the model forecast (Fig. 2). For this very same analysis time, the assimilation of the MinSLP observations had relatively little impact on the nearby Tropical Storm Hanna (Fig. 1), since the control experiment already had relatively small analysis errors (relative to the much stronger Hurricane Ike) due to the longevity, modest intensity, and fairly large spatial scale of that particular storm.

Fig. 2.
Fig. 2.

As in Fig. 1, but for the 72-h forecast initialized at 0000 UTC 4 Sep 2008. The triangle represents the National Hurricane Center (NHC) best-track position for Hurricane Ike, which had an advisory intensity of 948 hPa (59 m s−1) at this verification time.

Citation: Weather and Forecasting 26, 6; 10.1175/WAF-D-11-00045.1

The assimilation of the MinSLP observations did lead to improved track (Fig. 3) and intensity (Fig. 4) forecasts with the T574 GFS model for all lead times out to 5 days. The improvement in track (intensity) was statistically significant from initialization out to lead times of 48 h (72 h), as quantified by applying a paired block bootstrapping algorithm (Hamill et al. 2011). The most notable improvement was in the initial intensity bias (as measured by the maximum near-surface winds), which was maintained through the forecast period. Both the control and experimental forecasts demonstrated an improved intensity bias with increased forecast lead time, indicative of the initial vortices being too weak and the model trying to (slowly) spin up more intense storms (though this effect was reduced in the experimental forecasts). The initial position errors were also found to be slightly better in the MinSLP experiment versus the control, despite the fact that both systems had the vortex relocation procedure on the first guess prior to assimilation. The assimilation of the MinSLP observations, valid at each analysis time, helped to anchor the analyzed position of the tropical cyclones. This can be particularly important within 3DVAR, where all observations are assumed to be valid at analysis time (potentially causing problems with the assimilation of observations not taken at the analysis time near transient features such as tropical cyclones) and as static background errors without flow dependence are typically applied.

Fig. 3.
Fig. 3.

The average T574 GFS tropical cyclone track error (km) for the control (black) and experimental (gray) runs for the period covering 1 Jul–10 Nov 2008. The average track error was computed from a homogeneous sample of cases for storms in the Atlantic, eastern Pacific, and western Pacific basins. The number of cases for each lead time is identified below each forecast hour. Error bars indicate the 5th and 95th percentiles of a resampled block bootstrap distribution.

Citation: Weather and Forecasting 26, 6; 10.1175/WAF-D-11-00045.1

Fig. 4.
Fig. 4.

As in Fig. 3, but for the average GFS bias in the tropical storm near-surface maximum wind (m s−1).

Citation: Weather and Forecasting 26, 6; 10.1175/WAF-D-11-00045.1

4. Summary and discussion

To address forecaster concerns regarding tropical storms being too weak in GFS initial conditions, a method has been developed and implemented to assimilate official advisory MinSLP observations. To get reasonable impacts on analyses and forecasts, it was necessary to specify relatively low observation errors since a large portion of the signal from these observations is filtered by the incremental normal-mode constraint that is applied in the operational global GSI. Using the T574 version of the GFS model for the 2008 hurricane season, the assimilation of the MinSLP observations has been demonstrated to significantly reduce the initial intensity bias at no additional cost to the assimilation system. The improved initial storms in the MinSLP experiment also resulted in improved track and intensity forecasts for all lead times. The assimilation of these observations has been operational in the NCEP GFS since December 2009 and has led to improved tropical storm prediction. The vortex relocation procedure that runs outside of the analysis system remains operational. Hopefully, by relying more heavily on the MinSLP observations and perhaps by assimilating the storm positions directly, the relocation procedure can be phased out in a future GDAS implementation.

Acknowledgments

I would like to thank my many colleagues at EMC and the University of Maryland for their continued collaboration and support. Thanks also to Andrew Collard, Vijay Tallapragada, and three anonymous reviewers for comments that helped improve the manuscript.

REFERENCES

  • Aberson, S. D., 2008: Large forecast degradations due to synoptic surveillance during the 2004 and 2005 hurricane seasons. Mon. Wea. Rev., 136, 31383150.

    • Search Google Scholar
    • Export Citation
  • Chen, Y., , and Snyder C. , 2007: Assimilating vortex position with an ensemble Kalman filter. Mon. Wea. Rev., 135, 18281845.

  • Dvorak, V. F., 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103, 420430.

  • Dvorak, V. F., 1984: Tropical cyclone intensity analysis using satellite data. NOAA Tech. Rep. NESDIS 11, 47 pp.

  • Guard, C. P., , Carr L. E. , , Wells F. H. , , Jeffries R. A. , , Gural N. D. , , and Edson D. K. , 1992: Joint Typhoon Warning Center and the challenges of multibasin tropical cyclone forecasting. Wea. Forecasting, 7, 328352.

    • Search Google Scholar
    • Export Citation
  • Hamill, T. M., , Whitaker J. S. , , Fiorino M. , , and Benjamin S. G. , 2011: Global ensemble predictions of 2009’s tropical cyclones initialized with an ensemble Kalman filter. Mon. Wea. Rev., 139, 668688.

    • Search Google Scholar
    • Export Citation
  • Kleist, D. T., , Parrish D. F. , , Derber J. C. , , Treadon R. , , Errico R. M. , , and Yang R. , 2009a: Improving incremental balance in the GSI 3DVAR analysis system. Mon. Wea. Rev., 137, 10461060.

    • Search Google Scholar
    • Export Citation
  • Kleist, D. T., , Parrish D. F. , , Derber J. C. , , Treadon R. , , Wu W.-S. , , and Lord S. J. , 2009b: Introduction of the GSI into the NCEP Global Data Assimilation System. Wea. Forecasting, 24, 16911705.

    • Search Google Scholar
    • Export Citation
  • Krishnamurti, T. N., , Correa-Torres R. , , Rohaly G. , , and Oosterhof D. , 1997: Physical initialization and hurricane ensemble forecasts. Wea. Forecasting, 12, 503514.

    • Search Google Scholar
    • Export Citation
  • Liu, Q., , Marchok T. , , Pan H.-L. , , Bender M. , , and Lord S. J. , 2000: Improvements in hurricane initialization and forecasting at NCEP with global and regional (GFDL) models. NWS Tech. Procedures Bull. 472, 7 pp. [Available online at http://www.nws.noaa.gov/om/tpb/472.htm.]

    • Search Google Scholar
    • Export Citation
  • Lord, S. J., 1991: A bogusing system for vortex circulations in the National Meteorological Center Global Forecast Model. Preprints, 19th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 328–330.

    • Search Google Scholar
    • Export Citation
  • Pu, Z., , and Braun S. A. , 2001: Evaluation of bogus vortex techniques with four-dimensional variational data assimilation. Mon. Wea. Rev., 129, 20232039.

    • Search Google Scholar
    • Export Citation
  • Rappaport, E. N., and Coauthors, 2009: Advances and challenges at the National Hurricane Center. Wea. Forecasting, 24, 395419.

  • Serrano, E., , and Undén P. , 1994: Evaluation of a tropical cyclone bogusing method in data assimilation and forecasting. Mon. Wea. Rev., 122, 15231547.

    • Search Google Scholar
    • Export Citation
  • Torn, R. D., 2010: Performance of a mesoscale ensemble Kalman filter (EnKF) during the NOAA High-Resolution Hurricane Test. Mon. Wea. Rev., 138, 43754392.

    • Search Google Scholar
    • Export Citation
  • Wu, C.-C., , Lien G.-Y. , , Chen J.-H. , , and Zhang F. , 2010: Assimilation of tropical cyclone track and structure based on the ensemble Kalman filter (EnKF). J. Atmos. Sci., 67, 38063822.

    • Search Google Scholar
    • Export Citation
  • Wu, W.-S., , Parrish D. F. , , and Purser R. J. , 2002: Three-dimensional variational analysis with spatially inhomogeneous covariances. Mon. Wea. Rev., 130, 29052916.

    • Search Google Scholar
    • Export Citation
  • Zou, X., , and Xiao Q. , 2000: Studies on the initialization and simulation of a mature hurricane using a variational bogus data assimilation scheme. J. Atmos. Sci., 57, 836860.

    • Search Google Scholar
    • Export Citation
1

The current NCEP GFS operational horizontal resolution (as of June 2010) is approximately 27 km (details regarding the last resolution change are available online at http://www.weather.gov/os/notification/tin10-15gfs.txt).

2

The observation error is actually specified to be a linear function of the difference between the minimum sea level pressure of the 6-h forecast and the observed storm. The actual errors in the MinSLP observations are larger than the values used for assimilation.

Save