• Arakawa, A., , and W. Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment, Part 1. J. Atmos. Sci., 31 , 674701.

    • Search Google Scholar
    • Export Citation
  • Cullen, M., 1993: The Unified Forecast/Climate Model. Meteor. Mag., 122 , 81122.

  • Daley, R., , and E. Barker, 2001: NAVDAS: Formulation and diagnostics. Mon. Wea. Rev., 129 , 869883.

  • DeMaria, M., , S. Aberson, , K. Ooyama, , and S. Lord, 1992: A nested spectral model for hurricane track forecasting. Mon. Wea. Rev., 120 , 16281643.

    • Search Google Scholar
    • Export Citation
  • Derber, J., , and W. Wu, 1998: The use of TOVS cloud-cleared radiances in the NCEP SSI analysis system. Mon. Wea. Rev., 126 , 22872289.

  • Emanuel, K., 1991: A scheme for representing cumulus convection in large-scale models. J. Atmos. Sci., 48 , 23132335.

  • Emanuel, K., , and M. Zivkovic-Rothman, 1999: Development and evaluation of a convection scheme for use in climate models. J. Atmos. Sci., 56 , 17661782.

    • Search Google Scholar
    • Export Citation
  • Goerss, J., , and R. Jeffries, 1994: Assimilation of synthetic tropical cyclone observations into the Navy Operational Global Atmospheric Prediction System. Wea. Forecasting, 9 , 557576.

    • Search Google Scholar
    • Export Citation
  • Goerss, J., , and T. Hogan, 2006: Impact of satellite observations and forecast model improvements on tropical cyclone track forecasts. Preprints, 27th Conf. on Hurricanes and Tropical Meteorology, Monterey, CA, Amer. Meteor. Soc., P5.2. [Available online at http://ams.confex.com/ams/pdfpapers/107291.pdf].

  • Goerss, J., , C. Velden, , and J. Hawkins, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part II: NOGAPS forecasts. Mon. Wea. Rev., 126 , 12191227.

    • Search Google Scholar
    • Export Citation
  • Goerss, J., , C. Sampson, , and J. Gross, 2004: A history of western North Pacific tropical cyclone track forecast skill. Wea. Forecasting, 19 , 633638.

    • Search Google Scholar
    • Export Citation
  • Heming, J., 2008: Tropical cyclone initialisation in the Met Office global model: Is it still necessary and can it be improved? Preprints, 28th Conf. on Hurricanes and Tropical Meteorology, Orlando, FL, Amer. Meteor. Soc., 4A.1. [Available online at http://ams.confex.com/ams/pdfpapers/137330.pdf.].

  • Heming, J., , J. Chan, , and A. Radford, 1995: A new scheme for the initialisation of tropical cyclones in the UK Meteorological Office global model. Meteor. Appl., 2 , 171184.

    • Search Google Scholar
    • Export Citation
  • Hogan, T., , and T. Rosmond, 1991: The description of the Navy Operational Global Atmospheric Prediction System’s spectral forecast model. Mon. Wea. Rev., 119 , 17861815.

    • Search Google Scholar
    • Export Citation
  • Lord, S., 1993: Recent developments in tropical cyclone track forecasting with the NMC global analysis and forecast system. Preprints, 20th Conf. on Hurricanes and Tropical Meteorology, San Antonio, TX, Amer. Meteor. Soc., 290–291.

  • Peng, M., , J. Ridout, , and T. Hogan, 2004: Recent modifications of the Emanuel convective scheme in the Navy Operational Global Atmospheric Prediction System. Mon. Wea. Rev., 132 , 12541268.

    • Search Google Scholar
    • Export Citation
  • Sampson, C., , J. Goerss, , and A. Schrader, 2005: A consensus track forecast for southern hemisphere tropical cyclones. Aust. Meteor. Mag., 54 , 115119.

    • Search Google Scholar
    • Export Citation
  • Velden, C., , C. Hayden, , S. Nieman, , W. Menzel, , S. Wanzong, , and J. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc., 78 , 173195.

    • Search Google Scholar
    • Export Citation
  • Zapotocny, T., , J. Jung, , J. Le Marshall, , and R. Treadon, 2007: A two-season impact study of satellite and in situ data in the NCEP Global Data Assimilation System. Wea. Forecasting, 22 , 887909.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Homogeneous comparison of TC track forecast error (n mi) for (a) 4 Jul–31 Oct 2005 and (b) 1 Aug–30 Sep 2006 for the Atlantic basin. The number of forecasts verified is listed below the forecast length (h) on the x axis.

  • View in gallery

    Homogeneous comparison of (a) NOGAPS TC track forecast error (n mi) and (b) percent improvement in NOGAPS TC track forecast error as a result of the assimilation of different sets of satellite observations for 4 Jul–31 Oct 2005 for the Atlantic basin. The number of forecasts verified in (a) is listed below the forecast length (h) on the x axis.

  • View in gallery

    As in Fig. 2, but for the North Pacific and Atlantic basins.

  • View in gallery

    As in Fig. 2, but for 1 Aug–30 Sep 2006.

  • View in gallery

    As in Fig. 4, but for the North Pacific and Atlantic basins.

  • View in gallery

    As in Fig. 2, but for the 2005 and 2006 test periods.

  • View in gallery

    As in Fig. 6, but for the North Pacific and Atlantic basins.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 32 32 4
PDF Downloads 12 12 1

Impact of Satellite Observations on the Tropical Cyclone Track Forecasts of the Navy Operational Global Atmospheric Prediction System

View More View Less
  • 1 Naval Research Laboratory, Monterey, California
© Get Permissions
Full access

Abstract

The tropical cyclone (TC) track forecasts of the Navy Operational Global Atmospheric Prediction System (NOGAPS) were evaluated for a number of data assimilation experiments conducted using observational data from two periods: 4 July–31 October 2005 and 1 August–30 September 2006. The experiments were designed to illustrate the impact of different types of satellite observations on the NOGAPS TC track forecasts. The satellite observations assimilated in these experiments consisted of feature-track winds from geostationary and polar-orbiting satellites, Special Sensor Microwave Imager (SSM/I) total column precipitable water and wind speeds, Advanced Microwave Sounding Unit-A (AMSU-A) radiances, and Quick Scatterometer (QuikSCAT) and European Remote Sensing Satellite-2 (ERS-2) scatterometer winds. There were some differences between the results from basin to basin and from year to year, but the combined results for the 2005 and 2006 test periods for the North Pacific and Atlantic Ocean basins indicated that the assimilation of the feature-track winds from the geostationary satellites had the most impact, ranging from 7% to 24% improvement in NOGAPS TC track forecasts. This impact was statistically significant at all forecast lengths. The impact of the assimilation of SSM/I precipitable water was consistently positive and statistically significant at all forecast lengths. The improvements resulting from the assimilation of AMSU-A radiances were also consistently positive and significant at most forecast lengths. There were no significant improvements/degradations from the assimilation of the other satellite observation types [e.g., Moderate Resolution Imaging Spectroradiometer (MODIS) winds, SSM/I wind speeds, and scatterometer winds]. The assimilation of all satellite observations resulted in a gain in skill of roughly 12 h for the NOGAPS 48- and 72-h TC track forecasts and a gain in skill of roughly 24 h for the 96- and 120-h forecasts. The percent improvement in these forecasts ranged from almost 20% at 24 h to over 40% at 120 h.

Corresponding author address: James S. Goerss, NRL, 7 Grace Hopper Ave. Stop 2, Monterey, CA 93943-5502. Email: jim.goerss@nrlmry.navy.mil

Abstract

The tropical cyclone (TC) track forecasts of the Navy Operational Global Atmospheric Prediction System (NOGAPS) were evaluated for a number of data assimilation experiments conducted using observational data from two periods: 4 July–31 October 2005 and 1 August–30 September 2006. The experiments were designed to illustrate the impact of different types of satellite observations on the NOGAPS TC track forecasts. The satellite observations assimilated in these experiments consisted of feature-track winds from geostationary and polar-orbiting satellites, Special Sensor Microwave Imager (SSM/I) total column precipitable water and wind speeds, Advanced Microwave Sounding Unit-A (AMSU-A) radiances, and Quick Scatterometer (QuikSCAT) and European Remote Sensing Satellite-2 (ERS-2) scatterometer winds. There were some differences between the results from basin to basin and from year to year, but the combined results for the 2005 and 2006 test periods for the North Pacific and Atlantic Ocean basins indicated that the assimilation of the feature-track winds from the geostationary satellites had the most impact, ranging from 7% to 24% improvement in NOGAPS TC track forecasts. This impact was statistically significant at all forecast lengths. The impact of the assimilation of SSM/I precipitable water was consistently positive and statistically significant at all forecast lengths. The improvements resulting from the assimilation of AMSU-A radiances were also consistently positive and significant at most forecast lengths. There were no significant improvements/degradations from the assimilation of the other satellite observation types [e.g., Moderate Resolution Imaging Spectroradiometer (MODIS) winds, SSM/I wind speeds, and scatterometer winds]. The assimilation of all satellite observations resulted in a gain in skill of roughly 12 h for the NOGAPS 48- and 72-h TC track forecasts and a gain in skill of roughly 24 h for the 96- and 120-h forecasts. The percent improvement in these forecasts ranged from almost 20% at 24 h to over 40% at 120 h.

Corresponding author address: James S. Goerss, NRL, 7 Grace Hopper Ave. Stop 2, Monterey, CA 93943-5502. Email: jim.goerss@nrlmry.navy.mil

1. Introduction

Tropical cyclone (TC) track forecasts derived from the forecasts of global numerical weather prediction (NWP) models have become increasingly important in recent years as guidance to TC forecasters at both the National Hurricane Center (NHC) and the Joint Typhoon Warning Center. Forecasters at NHC routinely utilize forecast aids formed using the interpolated TC track forecasts from the Global Forecast System (GFS; Lord 1993) run at the National Centers for Environmental Prediction (NCEP), the Navy Operational Global Atmospheric Prediction System (NOGAPS; Hogan and Rosmond 1991; Goerss and Jeffries 1994) run at the Fleet Numerical Meteorology and Oceanography Center (FNMOC), and the Met Office global model (UKMO; Cullen 1993; Heming et al. 1995). The improvement in the TC track forecasting skill of the global NWP models since 1992 has been illustrated for both the western North Pacific Ocean (Goerss et al. 2004) and the Southern Hemisphere (Sampson et al. 2005). For both basins, the typical global NWP model 72-h forecast error today is comparable to the typical 48-h forecast error 10 years ago.

Over the past decade, one of the major areas of improvement in the global NWP systems has been the increased and more effective assimilation of satellite observations. For example, the direct use of satellite radiances replaced the use of retrievals in the GFS data assimilation system at NCEP in 1995 (Derber and Wu 1998), and the assimilation of high-density multispectral Geostationary Operational Environmental Satellite-8 (GOES-8) winds (Velden et al. 1997) into NOGAPS was initiated at FNMOC in 1996 (Goerss et al. 1998). Goerss and Hogan (2006) found from their NOGAPS assimilation study for August–September 2004 that the assimilation of all types of satellite observations resulted in a 15%–25% improvement in the NOGAPS TC track forecast error. In another study, which focused more on extratropical impacts, Zapotocny et al. (2007) found for a limited sample that the impact of the assimilation of all types of satellite observations upon the GFS TC track forecast error was greater than or equal to the impact of the assimilation of conventional observations.

The purpose of this study is to explore further the impact that the assimilation of satellite observations has upon the TC track forecasting performance of NOGAPS. Data assimilation experiments have been designed to illustrate the impact on the NOGAPS TC track forecasts from the assimilation of different types of satellite observations. In the next section, we describe the design of these experiments. The results of these experiments for 2005 and 2006 are presented in sections 3 and 4, respectively. The combined results for the two test periods are discussed in section 5 along with the results of withholding all satellite observations. In the final section, we present a summary and the conclusions.

2. Experimental design

The TC track forecasts of NOGAPS were evaluated for a number of data assimilation experiments conducted using observational data from two periods, one in 2005 and the other in 2006. The first period, 4 July–31 October 2005, was an extremely active one covering most of the record-breaking Atlantic season. For the Atlantic, there were 12 hurricanes (including Katrina, Rita, and Wilma) and 9 tropical storms. It was also an active period for the other basins. There were 6 hurricanes and 6 tropical storms for the eastern North Pacific and 12 typhoons and 5 tropical storms for the western North Pacific. The second period, 1 August–30 September 2006, was not nearly as active as the first. For the Atlantic, there were only 4 hurricanes and 3 tropical storms. There were 5 hurricanes and 3 tropical storms for the eastern North Pacific and 7 typhoons and 4 tropical storms for the western North Pacific (for this study we have assigned the central North Pacific storm Ioke to the western North Pacific).

The 2007 operational configuration of NOGAPS—consisting of the Naval Research Laboratory Atmospheric Variational Data Assimilation System (Daley and Barker 2001) with the assimilation of all available conventional and satellite observations and a T239L30 global spectral model (239-wave, triangular truncation, 30 vertical levels) with Emanuel convective parameterization (Peng et al. 2004)—was used as the control run. The conventional observations assimilated in these experiments consisted of rawinsonde and pilot balloon, aircraft, surface (land and marine), Australian bogus, and synthetic TC observations. The satellite observations assimilated in these experiments consisted of feature-track winds from geostationary and polar-orbiting satellites, Special Sensor Microwave Imager (SSM/I) total column precipitable water (PW) and wind speeds, Advanced Microwave Sounding Unit-A (AMSU-A) radiances, and Quick Scatterometer (QuikSCAT) and European Remote Sensing Satellite-2 (ERS-2)1 scatterometer winds.

In Fig. 1a, the TC track forecast errors for the 2005 (NGPS) and 2007 (CNTL) configurations of NOGAPS are compared with those for the NHC official forecasts (OFCL) for the 2005 test period. Except for the 24-h forecasts, the forecast errors for the 2007 configuration of NOGAPS are less than those for the configuration run operationally in 2005. Statistical significance was computed using the modified t test, described by DeMaria et al. (1992), for which the effective sample size was determined by requiring at least a 30-h separation between forecasts for the same storm. The improvements for the 72- and 120-h forecasts were significant at the 87% and 88% levels, respectively. The TC track forecast errors for the 2006 (NGPS) and 2007 (CNTL) configurations of NOGAPS are compared with those for the NHC official forecasts (OFCL) for the 2006 test period in Fig. 1b. Except for the 120-h forecasts, the forecast errors for the 2007 configuration of NOGAPS are less than those for the 2006 operational configuration. The improvement shown by the 2007 configuration for the 48-h forecasts is significant at the 89% level, and the degradation shown for the 120-h forecasts is significant at the 86% level. However, note that the number of verifying forecasts for 2006 was roughly one-half of the number for 2005. In any case, it can be seen from Fig. 1 that performance of the NOGAPS control system to be used for these experiments is clearly adequate.

For each experiment, the NOGAPS data assimilation system was run over the test period using a 6-h update cycle, just as is done operationally at FNMOC. Then, 120-h forecasts were run every 12 h (0000 and 1200 UTC) for the entire period. The experiments were designed to illustrate the impact on the NOGAPS TC track forecasts from the assimilation of different types of satellite observations. First, the NOGAPS control system was run over each test period with the assimilation of all available observations (CNTL). Subsequent experiments were then run removing one set of satellite observations. The second experiment (NOSW) was run without the assimilation of feature-track winds derived from the five geostationary satellites (tropical and midlatitude coverage) as well as those derived for the polar regions using the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument aboard the National Aeronautics and Space Administration (NASA) Aqua and Terra satellites. SSM/I total column PW observations were withheld from the third experiment (NOPW), AMSU-A radiances were withheld from the fourth experiment (NOAM), and the MODIS winds were withheld from the fifth experiment (NOMO). The next two experiments were performed without the assimilation of QuikSCAT and ERS-2 scatterometer winds (NOSC) and SSM/I wind speeds (NOIW), respectively. An additional experiment was conducted in which all satellite observations were withheld (NSAT).

The assimilation of synthetic tropical cyclone observations (Goerss and Jeffries 1994) is a critical part of the NOGAPS data assimilation system. Except in the Atlantic basin, where aircraft surveillance is routine, the information used to generate these observations, TC position and intensity, is almost entirely obtained using satellite imagery. As a final experiment, the NOGAPS control system was run without the assimilation of these synthetic observations from 4 July to 30 September 2005.

3. 2005 results

The results of the experiments for the Atlantic basin for 4 July–31 October 2005 are summarized in Fig. 2a, where the NOGAPS TC track forecast errors are displayed. The impact of the different sets of satellite observations is more clearly illustrated in Fig. 2b, where the percent degradation with respect to the control experiment resulting from the removal of the different sets of satellite observations is displayed. Henceforth, we will use this percent degradation to define the percent improvement resulting from the assimilation of the different sets of satellite observations. The observations with the greatest impact on NOGAPS TC track forecasting performance were the feature-track winds derived from the geostationary satellites and the MODIS instrument aboard Aqua and Terra (NOSW). The percent improvement resulting from the assimilation of these observations ranged from about 8% at 48 h to about 19% at 120 h. The improvements for the 24-, 48-, 72-, and 120-h forecasts were significant at the 99%, 95%, 97%, and 92% levels, respectively. The insignificant improvements/degradations due to the assimilation of the polar winds from Aqua and Terra (NOMO) indicate that this improvement is due to the assimilation of the winds from the geostationary satellites. The improvements resulting from the assimilation of SSM/I PW (NOPW) were consistently positive and were significant at the 96% and 97% levels for the 24- and 48-h forecasts, respectively. The only other significant improvement was due to the assimilation of SSM/I wind speeds (NOIW), which was significant at the 92% level for the 72-h forecasts. There were no significant improvements/degradations resulting from the assimilation of the other observation types (e.g., AMSU-A radiances, MODIS winds, and scatterometer winds from QuikSCAT and ERS-2).

The results of these experiments for the eastern and western North Pacific and Atlantic basins for 4 July–31 October 2005 are summarized in Fig. 3a, where the TC track forecast errors are displayed. The impact of the different sets of satellite observations is more clearly illustrated in Fig. 3b, where the percent improvement from the assimilation of the different sets of satellite observations is displayed. As we saw for the Atlantic basin in Fig. 2b, the observations with the greatest impact on NOGAPS TC track forecasting performance were the feature-track winds derived from the geostationary satellites and the MODIS instrument aboard Aqua and Terra. The percent improvement from the assimilation of these observations ranged from just over 10% at 24 h to about 27% at 120 h. These improvements were significant at the 99% level for all forecast lengths except 72 h, where they were significant at the 98% level. As we saw for the Atlantic basin, the impact of the assimilation of the polar winds from Aqua and Terra was not significant, indicating that this improvement is due to the assimilation of the winds from the geostationary satellites. As in the Atlantic, the improvements resulting from the assimilation of SSM/I PW were consistently positive and were significant at the 98%, 99%, and 90% levels for the 24-, 48-, and 120-h forecasts, respectively. The only other significant improvement was due to the assimilation of AMSU-A radiances, which was significant at the 97% level for the 48-h forecasts. There were no significant improvements/degradations resulting from the assimilation of the other observation types (e.g., MODIS winds, scatterometer winds, and SSM/I wind speeds).

The NOGAPS control system was run without the assimilation of synthetic TC observations from 4 July to 30 September 2005. The improvements resulting from the assimilation of the synthetic observations for the eastern and western North Pacific and Atlantic basins were 55%, 38%, 29%, 21%, and 16% for the 24-, 48-, 72-, 96-, and 120-h forecasts, respectively. These improvements were all significant at the 95% level. These results are consistent with those found by Goerss and Hogan (2006). When they conducted a similar experiment for 14 August–30 September 2004, they found that the respective improvements from the assimilation of the synthetic TC observations were 49%, 39%, 23%, 15%, and 7% with all but the 120-h improvement significant at the 95% level. These results are also consistent with those reported by Heming (2008), who found that the TC track forecast improvements resulting from the assimilation of synthetic TC observations into UKMO ranged from 30% at 24 h to 15% at 120 h. When these results are compared with those displayed in Fig. 3b, it is seen that, for all forecast lengths except 120 h, the synthetic TC observations have more impact on NOGAPS TC track forecasts than do any of the satellite observation types. For the 120-h forecasts, they have less impact than the feature-track winds do. It is clear that the assimilation of these observations is of vital importance to the TC track forecast performance of the NOGAPS data assimilation system.

4. 2006 results

The results of the experiments for the Atlantic basin for 1 August–30 September 2006 are summarized in Fig. 4a, where the NOGAPS TC track forecast errors are displayed. When Figs. 2a and 4a are compared, it is seen that the sample sizes for 2006 were about one-half of what they were for 2005. The impact of the different sets of satellite observations is more clearly illustrated in Fig. 4b, where the percent improvement resulting from the assimilation of the different sets of satellite observations is displayed. Except at 120 h, the observations with the greatest impact on NOGAPS TC track forecasting performance were the AMSU-A radiances (NOAM). The percent improvement from the assimilation of these observations ranged from about 9% at 72 h to about 17% at 24 h. These improvements were significant at the 99%, 97%, and 93% levels for the 24-, 48-, and 96-h forecasts, respectively. The only other significant improvements were due to the assimilation of scatterometer winds at 24 h (99%), the assimilation of MODIS winds at 24 h (93%), and the assimilation of SSM/I PW at 96 h (95%). The improvements from the assimilation of satellite feature-track winds were about 14% at 96 h and 24% at 120 h, but because of the small sample size they were only significant at the 85% and 89% levels, respectively.

Note that although the 96-h SSM/I PW improvement was only 10% and was less than that for the satellite feature-track winds, it was significant at a higher level because it was more consistent over the course of the test period.

The results of the experiments for the eastern and western North Pacific and Atlantic basins for 1 August–30 September 2006 are summarized in Fig. 5a, where the TC track forecast errors are displayed. Like was seen for the Atlantic basin, a comparison of Figs. 3a and 5a shows that the sample sizes for 2006 were roughly one-half of what they were for 2005. The impact of the different sets of satellite observations is more clearly illustrated in Fig. 5b, where the percent improvement resulting from the assimilation of the different sets of satellite observations is displayed. At 24 and 48 h, the observations with the greatest impact on NOGAPS TC track forecasting performance were the AMSU-A radiances. The percent improvement from the assimilation of these observations ranged from just over 2% at 96 h to just over 10% at 24 h. These improvements were significant at the 99% and 98% levels for the 24- and 48-h forecasts, respectively. The assimilation of the SSM/I PW observations had the greatest impact at 72 and 96 h and resulted in improvements ranging from 7% at 48 h to about 11% at 96 and 120 h. These improvements were significant at the 98%, 96%, 96%, 99%, and 98% levels for the 24-, 48-, 72-, 96-, and 120-h forecasts, respectively. The assimilation of the satellite feature-track winds had the greatest impact at 120 h and resulted in improvements that ranged from 2% at 48 h to just over 18% at 120 h. The improvement for the 120-h forecasts was significant at the 97% level. The only other significant improvements were at 24 h and were due to the assimilation of scatterometer and MODIS winds. These improvements were significant at the 92% and 97% levels, respectively.

5. Combined results

The results of the 2005 and 2006 experiments for the Atlantic basin are summarized in Fig. 6a, where the NOGAPS TC track forecast errors are displayed. The percent improvement resulting from the assimilation of the different sets of satellite observations is shown in Fig. 6b. At all forecast lengths, the observations with the greatest impact on NOGAPS TC track forecasting performance were the feature-track winds derived from the geostationary satellites and the MODIS instrument aboard Aqua and Terra. The percent improvement from the assimilation of these observations ranged from 7% at 48 h to just over 20% at 120 h. These improvements were significant at the 99%, 93%, 95%, 90%, and 97% levels for the 24-, 48-, 72-, 96-, and 120-h forecasts, respectively. The insignificant improvements/degradations resulting from the assimilation of the MODIS winds indicate that these improvements are due to the assimilation of the geostationary satellite winds. The improvements from the assimilation of SSM/I PW and AMSU-A radiances were consistently positive. The SSM/I improvements ranged from nearly 3% at 72 h to about 8% at 120 h and were significant at the 98% and 97% levels for the 24- and 48-h forecasts, respectively. The AMSU-A improvements ranged from about 2% at 96 h to 7% at 120 h and were significant at the 95% and 96% levels for the 24- and 48-h forecasts, respectively. The only other significant improvements were due to the assimilation of scatterometer winds (92%) and SSM/I wind speeds (92%) at 24 h.

The results of the experiments for 2005 and 2006 for the North Pacific and Atlantic basins are summarized in Fig. 7a, where the NOGAPS TC track forecast errors are displayed. The impact of the different sets of satellite observations is illustrated in Fig. 7b, where the percent improvement resulting from the assimilation of the different sets of satellite observations is displayed. As was seen for the Atlantic basin in Fig. 6b, the observations with the most impact on NOGAPS TC track forecasting performance were the geostationary and MODIS feature-track winds. The percent improvement from the assimilation of these observations ranged from over 7% at 24 h to over 24% at 120 h. These improvements were significant at the 99%, 99%, 98%, 99%, and 99% levels for the 24-, 48-, 72-, 96-, and 120-h forecasts, respectively. As was seen for the Atlantic basin, the impact of the MODIS winds is insignificant, indicating that the positive impact is primarily due to the assimilation of the geostationary satellite winds. Second in impact to the feature-track winds were the SSM/I PW observations. The percent improvement resulting from the assimilation of these observations ranged from just over 3% at 72 h to over 9% at 120 h. These improvements were significant at the 99%, 99%, 92%, 98%, and 97% levels for the 24-, 48-, 72-, 96-, and 120-h forecasts, respectively. The improvements from the assimilation of AMSU-A radiances were also consistently positive, ranging from about 1% at 96 h to 7% at 48 h. These improvements were significant at the 91%, 99%, and 90% levels for the 24-, 48-, and 120-h forecasts, respectively. There were no significant improvements/degradations resulting from the assimilation of the other observation types (e.g., MODIS winds, scatterometer winds, and SSM/I wind speeds).

We also examine the impact of all satellite observations upon the NOGAPS TC forecasts for the 2005 and 2006 test periods. It is seen in Fig. 6a that, at the longer forecast ranges, the assimilation of all satellite observations results in the gain of roughly a day of skill for the NOGAPS TC track forecasts. The 120-h TC track forecast errors for NOGAPS when all satellite observations were assimilated (CNTL) were smaller than the 96-h errors when no satellite observations were assimilated (NSAT). The CNTL 96-h forecast errors were just a little larger than the NSAT 72-h forecast errors. The gains in skill for the 48- and 72-h forecasts resulting from the assimilation of all satellite observations are on the order of 12 h. In Fig. 6b it is seen that the percent improvement from the assimilation of all satellite observations ranges from just over 18% at 24 h to about 47% at 120 h. Except at 24 h, the sum of the percent improvements for each individual satellite observation type is actually less than the percent improvement for all satellite observations. Thus, although the assimilation of some of the types of satellite observations was not statistically significant, their assimilation contributes additively and even multiplicatively at some forecast lengths to reduction in NOGAPS TC track forecast error. A very similar picture is seen when the impact of the assimilation of all satellite observations for 2005 and 2006 for the North Pacific and Atlantic basins is examined. As in the Atlantic, Fig. 7a shows that there is a gain in skill of roughly 12 h for the 48- and 72-h forecasts and of roughly 24 h for the 96- and 120-h forecasts resulting from the assimilation of all satellite observations. The percent improvement from the assimilation of all satellite observations, as shown in Fig. 7b, ranges from almost 17% at 24 h to just over 40% at 120 h. The sum of the percent improvements for each individual satellite observation type is either less than or roughly equal to the percent improvement for all satellite observations at all forecast lengths except 48 h.

6. Summary and conclusions

A number of data assimilation experiments designed to determine the impact of different types of satellite observations upon NOGAPS TC track forecasts were run over two different test periods, one in 2005 and the other in 2006. The first period, 4 July–31 October 2005, was an extremely active one for the Atlantic and Pacific basins. The second period, 1 August–30 September 2006, was not nearly as busy, with roughly one-half as much activity as the first period in the Atlantic and North Pacific basins.

The 2007 operational configuration of NOGAPS was used to conduct these experiments. The TC track forecast performance for this version of NOGAPS was found to be better than that for the versions run operationally in both 2005 and 2006. For each test period, eight data assimilation experiments were conducted by running the NOGAPS data assimilation system using a 6-h update cycle and then making 120-h forecasts every 12 h (0000 and 1200 UTC). A control experiment was run that used all available observations, and then six experiments were run while withholding a type of satellite observations. The observation types withheld were tropical and midlatitude feature-track winds derived from geostationary satellites and polar feature-track winds derived using the MODIS instrument aboard Aqua and Terra, SSM/I total column PW, AMSU-A radiances, MODIS polar feature-track winds, QuikSCAT and ERS-2 scatterometer winds, and SSM/I wind speeds. Experiments were also run while withholding all satellite observations and while withholding only the NOGAPS synthetic TC observations.

The 4 July–31 October 2005 test period was a very active one for the Atlantic and North Pacific basins. The satellite observations found to have the most impact upon NOGAPS TC track forecasts were the feature-track winds derived from geostationary satellites. The impact of these observations was found to be statistically significant at most forecast lengths. The TC track forecast improvements resulting from the assimilation of SSM/I PW observations were consistently positive and were statistically significant for some forecast lengths. With an occasional exception, the forecast improvements/degradations from the assimilation of the other satellite observation types (e.g., AMSU-A radiances, MODIS winds, SSM/I wind speeds, and scatterometer winds) were not statistically significant. In an additional experiment run from 4 July to 30 September 2005, it was found that, for all forecast lengths except 120 h, the synthetic TC observations have more impact on NOGAPS TC track forecasts than do any of the satellite observation types. For the 120-h forecasts, they have less impact than do the feature-track winds.

The 1 August–30 September 2006 test period was not a particularly active one for the Atlantic and North Pacific basins. The number of verifying NOGAPS TC track forecasts for the 2006 test period was roughly one-half of that for the 2005 test period for the Atlantic and North Pacific basins. For the Atlantic, the assimilation of the AMSU-A radiances had the most impact upon the TC track forecasts. This impact was statistically significant at most forecast lengths. The impact of the feature-track winds at 96 and 120 h was greater than or equal to that for the radiances but was not statistically significant because of the small sample size. The impact of the assimilation of SSM/I PW observations was consistently positive but was only significant at 96 h. For the Atlantic and North Pacific basins, the assimilation of the AMSU-A radiances had the most impact upon the NOGAPS TC forecasts at 24 and 48 h, and this impact was statistically significant. The SSM/I PW observations had the most impact at 72 and 96 h and resulted in a consistently positive impact that was significant at all forecast lengths. The assimilation of the feature-track winds from the geostationary satellites had the most impact at 120 h, and this impact was also significant. With just a few exceptions, the forecast improvements/degradations resulting from the assimilation of the other satellite observation types (e.g., MODIS winds, SSM/I wind speeds, and scatterometer winds) were not statistically significant.

For the combined results for the 2005 and 2006 test periods for the North Pacific and Atlantic basins, it was found that the assimilation of the feature-track winds from the geostationary satellites had the most impact, ranging from 7% to 24% improvement in NOGAPS TC track forecasts, and that the impact was statistically significant at all forecast lengths. The impact of the assimilation of SSM/I PW was consistently positive and was statistically significant at all forecast lengths. The improvements resulting from the assimilation of AMSU-A radiances were also consistently positive and were significant at most forecast lengths. There were no significant improvements/degradations from the assimilation of the other satellite observation types (e.g., MODIS winds, SSM/I wind speeds, and scatterometer winds).

The European Centre for Medium-Range Weather Forecasts (ECMWF) global forecast system is the only one used for TC track forecasting that does not utilize synthetic TC observations or vortex relocation in its assimilation process. Whereas the short-range ECMWF TC track forecasts are routinely worse than those from the other global models (e.g., NOGAPS, GFS, and UKMO), the 3–5-day ECMWF TC track forecasts are routinely equal to or better than those from the other models. The impact of the assimilation of synthetic TC observations in both the NOGAPS and UKMO systems is considerably greater for their short-range forecasts than for their 3–5-day forecasts. This would suggest that the assimilation of observations in the vicinity of a TC has much to do with the performance of short-range forecasts and much less to do with the validity of longer-range forecasts. The exceptional long-range forecast skill of the ECMWF model suggests that getting the “big picture” or large-scale environment right is more important than getting the environment right in the vicinity of the TC for the 3–5-day forecasts.

The satellite observations with the most impact on NOGAPS TC track forecasts were the feature-track winds from geostationary satellites, and that impact was greatest for the longer-range forecasts. For each analysis (every 6 h) of the NOGAPS data assimilation cycle, these observations provide complete coverage of the tropics and the midlatitudes. Furthermore, at the same location, we often find wind observations at low level and at mid- to upper levels. This fact would suggest that assimilation of these observations greatly contributes to getting a better picture of the large-scale steering environment.

The impact of the SSM/I PW observations was significant at all forecast lengths. Except for rawinsonde observations, these are the only atmospheric moisture observations available to the NOGAPS data assimilation system. Whereas rawinsonde observations are relatively sparse over the tropical oceans, SSM/I PW observations provide coverage over more than one-half of the globe for each NOGAPS analysis. Goerss and Hogan (2006) found that replacing the Navy version of the relaxed Arakawa–Schubert (1974) convective parameterization scheme with the Emanuel scheme (Emanuel 1991; Emanuel and Zivkovic-Rothman 1999; Peng et al. 2004) resulted in a 10%–15% improvement in NOGAPS TC track forecasts. Given the sensitivity of NOGAPS track forecasts to better depiction of atmospheric moist processes in the tropics and the paucity of conventional moisture observations over the tropical oceans, it is not surprising that the assimilation of SSM/I PW observations has a significant positive impact on NOGAPS TC track forecasts.

The impact of the assimilation of AMSU-A radiances was consistently positive and was significant at most forecast lengths. These observations provide coverage over almost the entire globe for each NOGAPS analysis and provide almost all of the atmospheric thermal information over the tropical ocean areas. It is intuitive that one would not expect that the assimilation of MODIS winds in the polar regions would have much impact on NOGAPS TC track forecasts and, indeed, that was the case. Scatterometer vector-wind observations provide coverage over much less than one-half of the globe for each NOGAPS analysis. Although the coverage provided by SSM/I wind speed observations is better, they provide no information on wind direction. Profile observations like those provided by rawinsondes have much more impact on atmospheric data assimilation systems than do single-level observations, especially single-level surface observations. All of these factors help to explain the lack of impact that scatterometer and SSM/I wind speed observations have on NOGAPS track forecasts.

We examined the impact of the assimilation of all satellite observations upon the NOGAPS TC track forecasts for the 2005–06 test periods. For both the Atlantic basin alone and the combination of the Atlantic and North Pacific basins, similar results were obtained. The assimilation of all satellite observations resulted in a gain in skill of roughly 12 h for the NOGAPS 48- and 72-h TC track forecasts and a gain in skill of roughly 24 h for the 96- and 120-h forecasts. The percent improvement in the TC track forecasts ranged from almost 20% at 24 h to over 40% at 120 h. Furthermore, it was found that although the assimilation of some of the satellite observation types did not result in statistically significant improvements, their assimilation contributed additively and even multiplicatively at some forecast lengths to a reduction in the NOGAPS TC track forecast error.

Acknowledgments

This research was supported by Office of Naval Research 6.2 Base Program project “Assimilation of WindSat, OMPS, and GPS Data into NAVDAS.”

REFERENCES

  • Arakawa, A., , and W. Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment, Part 1. J. Atmos. Sci., 31 , 674701.

    • Search Google Scholar
    • Export Citation
  • Cullen, M., 1993: The Unified Forecast/Climate Model. Meteor. Mag., 122 , 81122.

  • Daley, R., , and E. Barker, 2001: NAVDAS: Formulation and diagnostics. Mon. Wea. Rev., 129 , 869883.

  • DeMaria, M., , S. Aberson, , K. Ooyama, , and S. Lord, 1992: A nested spectral model for hurricane track forecasting. Mon. Wea. Rev., 120 , 16281643.

    • Search Google Scholar
    • Export Citation
  • Derber, J., , and W. Wu, 1998: The use of TOVS cloud-cleared radiances in the NCEP SSI analysis system. Mon. Wea. Rev., 126 , 22872289.

  • Emanuel, K., 1991: A scheme for representing cumulus convection in large-scale models. J. Atmos. Sci., 48 , 23132335.

  • Emanuel, K., , and M. Zivkovic-Rothman, 1999: Development and evaluation of a convection scheme for use in climate models. J. Atmos. Sci., 56 , 17661782.

    • Search Google Scholar
    • Export Citation
  • Goerss, J., , and R. Jeffries, 1994: Assimilation of synthetic tropical cyclone observations into the Navy Operational Global Atmospheric Prediction System. Wea. Forecasting, 9 , 557576.

    • Search Google Scholar
    • Export Citation
  • Goerss, J., , and T. Hogan, 2006: Impact of satellite observations and forecast model improvements on tropical cyclone track forecasts. Preprints, 27th Conf. on Hurricanes and Tropical Meteorology, Monterey, CA, Amer. Meteor. Soc., P5.2. [Available online at http://ams.confex.com/ams/pdfpapers/107291.pdf].

  • Goerss, J., , C. Velden, , and J. Hawkins, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part II: NOGAPS forecasts. Mon. Wea. Rev., 126 , 12191227.

    • Search Google Scholar
    • Export Citation
  • Goerss, J., , C. Sampson, , and J. Gross, 2004: A history of western North Pacific tropical cyclone track forecast skill. Wea. Forecasting, 19 , 633638.

    • Search Google Scholar
    • Export Citation
  • Heming, J., 2008: Tropical cyclone initialisation in the Met Office global model: Is it still necessary and can it be improved? Preprints, 28th Conf. on Hurricanes and Tropical Meteorology, Orlando, FL, Amer. Meteor. Soc., 4A.1. [Available online at http://ams.confex.com/ams/pdfpapers/137330.pdf.].

  • Heming, J., , J. Chan, , and A. Radford, 1995: A new scheme for the initialisation of tropical cyclones in the UK Meteorological Office global model. Meteor. Appl., 2 , 171184.

    • Search Google Scholar
    • Export Citation
  • Hogan, T., , and T. Rosmond, 1991: The description of the Navy Operational Global Atmospheric Prediction System’s spectral forecast model. Mon. Wea. Rev., 119 , 17861815.

    • Search Google Scholar
    • Export Citation
  • Lord, S., 1993: Recent developments in tropical cyclone track forecasting with the NMC global analysis and forecast system. Preprints, 20th Conf. on Hurricanes and Tropical Meteorology, San Antonio, TX, Amer. Meteor. Soc., 290–291.

  • Peng, M., , J. Ridout, , and T. Hogan, 2004: Recent modifications of the Emanuel convective scheme in the Navy Operational Global Atmospheric Prediction System. Mon. Wea. Rev., 132 , 12541268.

    • Search Google Scholar
    • Export Citation
  • Sampson, C., , J. Goerss, , and A. Schrader, 2005: A consensus track forecast for southern hemisphere tropical cyclones. Aust. Meteor. Mag., 54 , 115119.

    • Search Google Scholar
    • Export Citation
  • Velden, C., , C. Hayden, , S. Nieman, , W. Menzel, , S. Wanzong, , and J. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc., 78 , 173195.

    • Search Google Scholar
    • Export Citation
  • Zapotocny, T., , J. Jung, , J. Le Marshall, , and R. Treadon, 2007: A two-season impact study of satellite and in situ data in the NCEP Global Data Assimilation System. Wea. Forecasting, 22 , 887909.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Homogeneous comparison of TC track forecast error (n mi) for (a) 4 Jul–31 Oct 2005 and (b) 1 Aug–30 Sep 2006 for the Atlantic basin. The number of forecasts verified is listed below the forecast length (h) on the x axis.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

Fig. 2.
Fig. 2.

Homogeneous comparison of (a) NOGAPS TC track forecast error (n mi) and (b) percent improvement in NOGAPS TC track forecast error as a result of the assimilation of different sets of satellite observations for 4 Jul–31 Oct 2005 for the Atlantic basin. The number of forecasts verified in (a) is listed below the forecast length (h) on the x axis.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

Fig. 3.
Fig. 3.

As in Fig. 2, but for the North Pacific and Atlantic basins.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

Fig. 4.
Fig. 4.

As in Fig. 2, but for 1 Aug–30 Sep 2006.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

Fig. 5.
Fig. 5.

As in Fig. 4, but for the North Pacific and Atlantic basins.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

Fig. 6.
Fig. 6.

As in Fig. 2, but for the 2005 and 2006 test periods.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

Fig. 7.
Fig. 7.

As in Fig. 6, but for the North Pacific and Atlantic basins.

Citation: Monthly Weather Review 137, 1; 10.1175/2008MWR2601.1

1

Note that the preponderance of scatterometer winds come from QuikSCAT and that ERS-2 winds are located in the high latitudes of the Northern Hemisphere.

Save