• Avila, L. A., , Pasch R. J. , , and Jiing J.-G. , 2000: Atlantic tropical systems of 1996 and 1997: Years of contrasts. Mon. Wea. Rev., 128, 36953706, doi:10.1175/1520-0493(2000)128<3695:ATSOAY>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Berrisford, P., and et al. , 2011: The ERA-Interim archive version 2.0. ERA Report Series 1, ECMWF, 23 pp. [Available online at http://www.ecmwf.int/sites/default/files/elibrary/2011/8174-era-interim-archive-version-20.pdf.]

  • Berry, G., , and Thorncroft C. , 2005: Case study of an intense African easterly wave. Mon. Wea. Rev., 133, 752766, doi:10.1175/MWR2884.1.

    • Search Google Scholar
    • Export Citation
  • Burpee, R. W., 1974: Characteristics of North African easterly waves during the summer of 1968 and 1969. J. Atmos. Sci., 31, 15561570, doi:10.1175/1520-0469(1974)031<1556:CONAEW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Carlson, T. N., 1969a: Some remarks on African disturbances and their progress over the tropical Atlantic. Mon. Wea. Rev., 97, 716726, doi:10.1175/1520-0493(1969)097<0716:SROADA>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Carlson, T. N., 1969b: Synoptic histories of three African wave disturbances that developed into Atlantic hurricanes. Mon. Wea. Rev., 97, 921925, doi:10.1175/1520-0493(1969)097<0256:SHOTAD>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., , and Stern M. E. , 1962: On the stability of internal baroclinic jets in a rotating atmosphere. J. Atmos. Sci., 19, 159172, doi:10.1175/1520-0469(1962)019<0159:OTSOIB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chen, T.-C., 2006: Characteristics of African easterly waves depicted by ECMWF reanalyses for 1991–2000. Mon. Wea. Rev., 134, 35393566, doi:10.1175/MWR3259.1.

    • Search Google Scholar
    • Export Citation
  • Chen, T.-C., , Wang S.-Y. , , and Clark A. J. , 2008: North Atlantic hurricanes contributed by African easterly waves north and south of the African easterly jet. J. Climate, 21, 67676776, doi:10.1175/2008JCLI2523.1.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Dunkerton, T. J., 1993: Observation of 3–6-day meridional wind oscillations over the tropical Pacific, 1973–1992: Vertical structure and interannual variability. J. Atmos. Sci., 50, 32923307, doi:10.1175/1520-0469(1993)050<3292:OODMWO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Fine, A. S., , Nicholls D. P. , , and Mogul D. J. , 2010: Assessing instantaneous synchrony of nonlinear nonstationary oscillators in the brain. J. Neurosci. Methods, 186, 4251,doi:10.1016/j.jneumeth.2009.10.023.

    • Search Google Scholar
    • Export Citation
  • Fink, A. H., , Vincent D. G. , , Reiner P. M. , , and Speth P. , 2004: Mean state and wave disturbances during Phase I, II, and III of GATE based on ERA-40. Mon. Wea. Rev., 132, 16611683, doi:10.1175/1520-0493(2004)132<1661:MSAWDD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Frank, N. L., 1975: Atlantic tropical systems of 1974. Mon. Wea. Rev., 103, 294300, doi:10.1175/1520-0493(1975)103<0294:ATSO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Frank, W. M., , and Roundy P. E. , 2006: The role of tropical waves in tropical cyclogenesis. Mon. Wea. Rev., 134, 23972417, doi:10.1175/MWR3204.1.

    • Search Google Scholar
    • Export Citation
  • Goldenberg, S. B., , and Shapiro L. J. , 1996: Physical mechanisms for the association of El Nino and West African rainfall with Atlantic major hurricanes. J. Climate, 9, 11691187, doi:10.1175/1520-0442(1996)009<1169:PMFTAO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Goldenberg, S. B., , Landsea C. W. , , Mestas-Nuñez A. M. , , and Gray W. M. , 2001: The recent increase in Atlantic hurricane activity: Causes and implications. Science, 293, 474479, doi:10.1126/science.1060040.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., , and Wheeler M. C. , 2008: Some space–time spectral analyses of tropical convection and planetary-scale waves. J. Atmos. Sci., 65, 29362948, doi:10.1175/2008JAS2675.1.

    • Search Google Scholar
    • Export Citation
  • Hopsch, S. B., , Thorncroft C. D. , , Hodges K. , , and Aiyyer A. , 2007: West African storm tracks and their relationship to Atlantic tropical cyclones. J. Climate, 20, 24682483, doi:10.1175/JCLI4139.1.

    • Search Google Scholar
    • Export Citation
  • Hopsch, S. B., , Thorncroft C. D. , , and Tyle K. R. , 2010: Analysis of African easterly wave structures and their role in influencing tropical cyclogenesis. Mon. Wea. Rev., 138, 13991419, doi:10.1175/2009MWR2760.1.

    • Search Google Scholar
    • Export Citation
  • Huang, N. E., , Steven R. L. , , and Shen Z. , 1996: The mechanism for frequency downshift in nonlinear wave evolution. Adv. Appl. Mech., 32, 59117, doi:10.1016/S0065-2156(08)70076-0.

    • Search Google Scholar
    • Export Citation
  • Huang, N. E., and et al. , 1998: The empirical mode decomposition and Hilbert spectrum for nonlinear and non-stationary time series analysis. Proc. Roy. Soc. London, 454A, 903995, doi:10.1098/rspa.1998.0193.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and et al. , 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267, doi:10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Landsea, C. W., 1993: A climatology of intense (or major) Atlantic hurricanes. Mon. Wea. Rev., 121, 17031713, doi:10.1175/1520-0493(1993)121<1703:ACOIMA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lau, K.-H., , and Lau N.-C. , 1990: Observed structure and propagation characteristics of tropical summertime synoptic-scale disturbances. Mon. Wea. Rev., 118, 18881913, doi:10.1175/1520-0493(1990)118<1888:OSAPCO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lorenz, E., 1963: Deterministic nonperiodic flow. J. Atmos. Sci., 20, 130141, doi:10.1175/1520-0469(1963)020<0130:DNF>2.0.CO;2.

  • Merrill, R. T., 1984: A comparison of large and small tropical cyclones. Mon. Wea. Rev., 112, 14081418, doi:10.1175/1520-0493(1984)112<1408:ACOLAS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Nielsen-Gammon, J. W., , and Gold D. A. , 2006: Dynamical diagnosis: A comparison of quasigeostrophy and Ertel potential vorticity. Synoptic–Dynamic Meteorology and Weather Analysis and Forecasting, Meteor. Monogr., No. 55, Amer. Meteor. Soc., 183–202, doi:10.1007/978-0-933876-68-2_9.

  • North, G. R., 1984: Empirical orthogonal functions and normal modes. J. Atmos. Sci., 41, 879887, doi:10.1175/1520-0469(1984)041<0879:EOFANM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Parker, D. J., and et al. , 2005: The diurnal cycle of the West African monsoon circulation. Quart. J. Roy. Meteor. Soc., 131, 28392860, doi:10.1256/qj.04.52.

    • Search Google Scholar
    • Export Citation
  • Pigorini, A., , Casali A. G. , , Casarotto S. , , Ferrarelli F. , , Baselli G. , , Mariotti M. , , Massimini M. , , and Rosanova M. C. E. , 2011: Time–frequency spectral analysis of TMS-evoked EEG oscillations by means of Hilbert–Huang transform. J. Neurosci. Methods, 198, 236245, doi:10.1016/j.jneumeth.2011.04.013.

    • Search Google Scholar
    • Export Citation
  • Pytharoulis, I., , and Thorncroft C. , 1999: The low-level structure of African easterly waves in 1995. Mon. Wea. Rev., 127, 22662280, doi:10.1175/1520-0493(1999)127<2266:TLLSOA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Reed, R. J., , Norquist D. C. , , and Recker E. E. , 1977: The structure and properties of African Wave disturbances as observed during phase III of GATE. Mon. Wea. Rev., 105, 317333, doi:10.1175/1520-0493(1977)105<0317:TSAPOA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Reed, R. J., , Hollingsworth A. , , Heckley W. A. , , and Delsol F. , 1988: An evaluation of the ECMWF operational system in analyzing and forecasting easterly wave disturbances over Africa and the tropical Atlantic. Mon. Wea. Rev., 116, 824865, doi:10.1175/1520-0493(1988)116<0824:AEOTPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Salisbury, J. I., , and Wimbush M. , 2002: Using modern time series analysis techniques to predict ENSO events from the SOI time series. Nonlinear Processes Geophys., 9, 341345, doi:10.5194/npg-9-341-2002.

    • Search Google Scholar
    • Export Citation
  • Schlurmann, T., 2002: Spectral frequency analysis of nonlinear water waves derived from the Hilbert-Huang transformation. J. Offshore Mech. Arct. Eng., 124, 2228, doi:10.1115/1.1423911.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., 2014: Nonlinear feedback in a five-dimensional Lorenz model. J. Atmos. Sci., 71, 17011723, doi:10.1175/JAS-D-13-0223.1.

  • Shen, B.-W., 2015a: Nonlinear feedback in a six-dimensional Lorenz model: Impact of an additional heating term. Nonlinear Processes Geophys., 22, 749764, doi:10.5194/npg-22-749-2015.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., 2015b: Parameterization of negative nonlinear feedback using a five-dimensional Lorenz model. Fractal Geom. Nonlinear Anal. Med. Biol., 1, doi:10.15761/FGNAMB.1000109.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Atlas R. , , Reale O. , , Lin S.-J. , , Chern J.-D. , , Chang J. , , Henze C. , , and Li J.-L. , 2006: Hurricane forecasts with a global mesoscale-resolving model: Preliminary results with Hurricane Katrina (2005). Geophys. Res. Lett., 33, L13813, doi:10.1029/2006GL026143.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Tao W.-K. , , and Wu M.-L. , 2010: African easterly waves in 30-day high resolution global simulations: A case study during the 2006 NAMMA period. Geophys. Res. Lett., 37, L18803, doi:10.1029/2010GL044355.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Tao W.-K. , , Lin Y.-L. , , and Laing A. , 2012: Genesis of twin tropical cyclones as revealed by a global mesoscale model: The role of mixed Rossby gravity waves. J. Geophys. Res., 117, D13114, doi:10.1029/2012JD017450.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , DeMaria M. , , Li J.-L. F. , , and Cheung S. , 2013a: Genesis of Hurricane Sandy (2012) simulated with a global mesoscale model. Geophys. Res. Lett., 40, 49444950, doi:10.1002/grl.50934.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Nelson B. , , Cheung S. , , and Tao W.-K. , 2013b: Improving NASA’s multiscale modeling framework for tropical cyclone climate study. IEEE Comput. Sci. Eng., 15, 5667, doi:10.1109/MCSE.2012.90.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Cheung S. , , Li J.-L. F. , , and Wu Y.-L. , , and Shen S. S. , 2016: Multiscale processes of Hurricane Sandy (2012) as revealed by the parallel ensemble empirical mode decomposition and advanced visualization technology. Adv. Data Sci. Adapt. Anal., in press.

  • Thorncroft, C. D., , and Hoskins B. J. , 1994: An idealized study of African easterly waves. II: A nonlinear view. Quart. J. Roy. Meteor. Soc., 120, 9831015, doi:10.1002/qj.49712051810.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., , and Rowell D. P. , 1998: Interannual variability of African wave activity in a general circulation model. Int. J. Climatol., 18, 13051323, doi:10.1002/(SICI)1097-0088(1998100)18:12<1305::AID-JOC281>3.0.CO;2-N.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., , and Hodges K. , 2001: African easterly wave variability and its relationship to Atlantic tropical cyclone activity. J. Climate, 14, 11661179, doi:10.1175/1520-0442(2001)014<1166:AEWVAI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Torrence, C., , and Compo G. P. , 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc., 79, 6178, doi:10.1175/1520-0477(1998)079<0061:APGTWA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wang, G., , Chen X.-Y. , , Qiao F.-L. , , Wu Z. , , and Huang N. E. , 2010: On intrinsic mode function. Adv. Adapt. Data Anal., 2, 277293, doi:10.1142/S1793536910000549.

    • Search Google Scholar
    • Export Citation
  • Wang, T., , Zhang M. , , Yu Q. , , and Zhang H. , 2012: Comparing the applications of EMD and EEMD on time–frequency analysis of seismic signal. J. Appl. Geophys., 83, 2934, doi:10.1016/j.jappgeo.2012.05.002.

    • Search Google Scholar
    • Export Citation
  • Wu, Z. H., , and Huang N. E. , 2009: Ensemble empirical mode decomposition: A noise-assisted data analysis method. Adv. Adapt. Data Anal., 1, 141, doi:10.1142/S1793536909000047.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    An idealized tropical wave: (a) zonal wind perturbation (U′), (b) WWB, (c) total wind (U′ plus WWB), (d) IMF6 from the EEMD of (c),(e) IMF9 from the decomposition of (c),(f) IMF6 plus IMF9, (g) correlation of the IMG6 and U′, and (h) correlation of the IMF9 and WWB. Data obtained from Shen et al. (2012).

  • View in gallery

    (a) Time–longitude diagrams for meridional winds at 700 hPa averaged over the latitudes of 7°–20°N from the ERA-Interim 0.75° reanalysis from 2004 to 2008. Green lines indicate tracked AEWs that displayed traceable signals of AEWs and that persisted for more than 24 h without interruption. Dark green lines indicate NHC tracked storms that were associated with AEWs. The dotted portion signifies that storms are moving north of 20°N. (b) As in (a), but for 2009–13.

  • View in gallery

    The maximum of ζ2 (10−5 s−1) over ocean vs over land: (a) all of the AEW cases, (b) the developing cases, and (c) the nondeveloping cases. Plotted values for the IMF3 and the trend mode were multiplied by two. Black dots indicate total wind, blue dots indicate the IMF3, and red dots indicate the trend mode.

  • View in gallery

    As in Fig. 3, but for the maximum of ζ1 (10−5 s−1). The ζ1 for the trend mode had near-zero values for all cases and rendered the correlation nonmeaningful.

  • View in gallery

    A histogram of the longitudinal location for the maximum of ζ2: (a) all AEW cases, (b) the developing cases, and (c) the nondeveloping cases. Gray, blue, and red bars represent total wind, the IMF3, and trend mode, respectively. Note that the West African coast is located at ~17°W.

  • View in gallery

    As in Fig. 5, but for the maximum of ζ1. Note that ζ1 for the trend mode had near-zero values for all cases.

  • View in gallery

    The tendency for horizontal wind shear for the total winds (black line), IMF3 (blue line), and trend mode (red line) and SLP (black dotted line) along the storm track of Helene (2006). The ζ2 is plotted as solid lines, the ζ1 as dashed–dotted lines, and SLP as a black dotted line. The plotted shear of the IMF3 and the trend mode were multiplied by three. The vertical thin dotted line indicates the time the storm was first classified by the NHC as a TD. The two vertical gray lines indicate the beginning and the end of intensification for the downscaling feature. The enhanced black dotted line for SLP indicates storm intensification with an SLP drop greater than 2.5 hPa/6 h. The red dashed–dotted line is plotted as a value of zero.

  • View in gallery

    As in Fig. 7, but for the 12 developing cases that display a downscaling feature in the horizontal shear during storm intensification for the study period.

  • View in gallery

    AEWs associated with named TCs in the Atlantic region from July to September between 2004 and 2013. Boxes divide the tropical Atlantic Ocean into eastern, middle, and western regions, following Hopsch et al. (2010). Green dots indicate cyclones that were first classified as TDs and blue dots indicate TSs, according to the NHC’s best-track data. Colored numbers indicate the total number of storms formed in each box during the 10-yr study period. TC genesis, as defined by the classification of TD, occurs most frequently within the eastern Atlantic Ocean. A higher frequency of stronger storms (TSs and hurricanes) formed farther within downstream region storm tracks (mid-Atlantic) than in the region immediately off the African coast.

  • View in gallery

    Tracks for the 13 storms and the 42 developing AEW cases showing the downscaling transfer of shear from the trend mode IMF to the short wave IMF3. The blue dotted line designates the tracks of the AEWs, the green lines indicate the tracks of TDs/TSs, and the red lines indicate a hurricane. The bolder portion of the tracks indicates storm intensification with a continuous drop in SLP. All of these storms developed into hurricanes.

  • View in gallery

    (left) U wind and (right) vorticity fields (from ERA-Interim global reanalysis) over the tropical Atlantic at 1200 UTC 12 Sep 2006. (top) The original data that represent the sum of all the IMF modes and the trend mode, (middle) the IMF3, and (bottom) the trend mode. The green cross marks the location of Hurricane Helene from the NHC best-track data, and the number in green in the top-left panel shows the SLP at the center of the hurricane. (b) As in (a), but at 1200 UTC 15 Sep 2006.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 107 107 9
PDF Downloads 92 92 6

An Evaluation of the Parallel Ensemble Empirical Mode Decomposition Method in Revealing the Role of Downscaling Processes Associated with African Easterly Waves in Tropical Cyclone Genesis

View More View Less
  • 1 Earth System Science Center, University of Alabama in Huntsville, Huntsville, Alabama
  • | 2 Department of Mathematics and Statistics, Center for Climate and Sustainability Studies, and Computational Science Research Center, San Diego State University, San Diego, California
© Get Permissions
Full access

Abstract

In this study the parallel ensemble empirical mode decomposition (PEEMD) is applied for an analysis of 10-yr (2004–13) ERA-Interim global reanalysis data in order to explore the role of downscaling processes associated with African easterly waves (AEWs) in tropical cyclone (TC) genesis. The focus of the study was aimed at understanding the downscaling process in multiscale flows during storm intensification. To represent the various length scales of atmospheric systems, intrinsic mode functions (IMFs) were extracted from the reanalysis data using the PEEMD. It was found that the nonoscillatory trend mode can be used to represent large-scale environmental flow and that the third oscillatory mode (IMF3) can be used to represent AEW/TC scale systems. The results 1) identified 42 developing cases from 272 AEWs, where 25 of them eventually developed into hurricanes; 2) indicated that the maximum for horizontal shear largely occurs over the ocean for the IMF3 and over land near the coast for the trend mode for developing cases, suggesting shear transfer between the trend mode and the IMF3; 3) displayed opposite wind shear tendencies for the trend mode and the IMF3 during storm intensification, signifying that the downscaling process was active in 13 hurricane cases along their tracks; and 4) showed that among the 42 developing cases, only 13 of the 25 hurricanes were found to have significant downscaling transfer features, so other processes such as upscaling processes may play an important role in the other developing cases, especially for the remaining 12 hurricane cases. In a future study, the authors intend to investigate the upscaling process between the convection scale and AEWs/TCs, which requires data at a finer grid resolution.

Corresponding author address: Dr. Bo-Wen Shen, Department of Mathematics and Statistics, San Diego State University, 5500 Campanile Drive, San Diego, CA 92182-7720. E-mail: bshen@mail.sdsu.edu; wuy@nsstc.uah.edu

Abstract

In this study the parallel ensemble empirical mode decomposition (PEEMD) is applied for an analysis of 10-yr (2004–13) ERA-Interim global reanalysis data in order to explore the role of downscaling processes associated with African easterly waves (AEWs) in tropical cyclone (TC) genesis. The focus of the study was aimed at understanding the downscaling process in multiscale flows during storm intensification. To represent the various length scales of atmospheric systems, intrinsic mode functions (IMFs) were extracted from the reanalysis data using the PEEMD. It was found that the nonoscillatory trend mode can be used to represent large-scale environmental flow and that the third oscillatory mode (IMF3) can be used to represent AEW/TC scale systems. The results 1) identified 42 developing cases from 272 AEWs, where 25 of them eventually developed into hurricanes; 2) indicated that the maximum for horizontal shear largely occurs over the ocean for the IMF3 and over land near the coast for the trend mode for developing cases, suggesting shear transfer between the trend mode and the IMF3; 3) displayed opposite wind shear tendencies for the trend mode and the IMF3 during storm intensification, signifying that the downscaling process was active in 13 hurricane cases along their tracks; and 4) showed that among the 42 developing cases, only 13 of the 25 hurricanes were found to have significant downscaling transfer features, so other processes such as upscaling processes may play an important role in the other developing cases, especially for the remaining 12 hurricane cases. In a future study, the authors intend to investigate the upscaling process between the convection scale and AEWs/TCs, which requires data at a finer grid resolution.

Corresponding author address: Dr. Bo-Wen Shen, Department of Mathematics and Statistics, San Diego State University, 5500 Campanile Drive, San Diego, CA 92182-7720. E-mail: bshen@mail.sdsu.edu; wuy@nsstc.uah.edu

1. Introduction

Recent intense tropical storms have caused substantial financial damage to human societies and have threatened human life (e.g., Shen et al. 2006, 2013a,b). Therefore, improving our understanding of tropical cyclone (TC) genesis and prediction is an active topic in atmospheric research. The association of TCs with tropical waves has now been studied for several decades (e.g., Landsea 1993; Frank and Roundy 2006). Landsea (1993) indicated that over 85% of intense Atlantic Ocean hurricanes from the 1940s through the 1960s were found to spawn from African easterly waves (AEWs). Mechanisms and factors that affect the genesis processes of tropical storms in association with tropical waves have been extensively studied (e.g., Reed et al. 1988; Lau and Lau 1990; Thorncroft and Hodges 2001; Berry and Thorncroft 2005; Hopsch et al. 2010). For example, Thorncroft and Hodges (2001) indicated that AEWs leaving West Africa, signified by low-level amplitudes, may impact Atlantic Ocean tropical cyclone activity. Hopsch et al. (2010) pointed out that the dry signal in mid- to upper-level air just ahead of the AEW trough can impede the development of an AEW into a tropical storm.

An AEW may be generated by the release of barotropic–baroclinic instability from the African easterly jet (AEJ) (Charney and Stern 1962) that appears as a thermal wind balance in response to the temperature gradient between the Sahara and the Gulf of Guinea. Midlevel AEJ wind speeds peak around 600–700 hPa north of the West African monsoon trough that is located north of the Guinea coast (Parker et al. 2005). The integrative system of the AEW–AEJ has been documented to have an impact on TC activities. An AEW is a westward-propagating system that typically has a wavelength ranging from 2000 to 4000 km and periods of 3–5 days (Pytharoulis and Thorncroft 1999) with a maximum amplitude near 600–700 hPa. While AEWs are synoptic systems, they also possess subsynoptic-scale structures that involve nonlinear multiscale processes for development (e.g., Thorncroft and Hoskins 1994). AEW tracks over the West African continent arise in two forms, each occurring along either side of the AEJ (Carlson 1969a,b; Burpee 1974; Reed et al. 1988). The northern track (north of 15°N) is associated with the semipermanent Saharan thermal low, while the southern track (south of 15°N) is related to the rainy zone (Pytharoulis and Thorncroft 1999). AEWs propagating along the northern track have been reported to be far more frequent than those along the southern track (Reed et al. 1988), while most AEWs that reach the Atlantic hurricane main development region (MDR) (Goldenberg and Shapiro 1996) come from the southern track (e.g., Thorncroft and Hodges 2001; Hopsch et al. 2007). The Atlantic hurricane MDR (Goldenberg and Shapiro 1996; Goldenberg et al. 2001), where Atlantic Ocean hurricanes are found to be the most active, is a band (~10°–20°N) that converges along the West African coast and in Central America.

In this study, we used a global reanalysis dataset to investigate the downscaling process of the multiscale interaction for AEW to tropical storm evolution. Previous studies have used multidecade reanalysis datasets such as the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) and the National Centers for Environmental Prediction (NCEP) Global Reanalysis 1 at low levels (e.g., 850–600 hPa) to identify the characteristics of AEJ/AEWs (Kistler et al. 2001; Thorncroft and Hodges 2001; Fink et al. 2004; Chen 2006; Hopsch et al. 2007, 2010). In our study, we used the latest ECMWF global reanalysis data (i.e., the ERA-Interim) that have a horizontal grid spacing of 0.75°, with a focus on revealing the downscaling process from large-scale environmental flows to AEW-/TC-scale systems. Examining the feedback of upscaling processes associated with convective systems to the AEW or to a TC requires much finer-resolution model simulations or reanalysis, and thus the feedback was not included in this study. The study presented here includes an area from West Africa to the Atlantic Ocean and a 10-yr period from 2004 to 2013. Note that the study regions are different from the MDRs used in most previous studies. In hurricane studies, in general, the MDR reaches westward and includes the Caribbean region. The domain targeted in our study included the region from the African coast to Central America (before entering the Caribbean) and the West African continent, where AEWs originate. As outlined by Hopsch et al. (2007), these regions contain the highest AEW activity and the highest density of Atlantic TC genesis. Details regarding the selection of the study domain are provided in section 2b.

While most major Atlantic hurricanes are spawned from AEWs, only a fraction of AEWs have actually developed into named tropical storms. The fraction indicates the annual variation regardless of the consistency in the total number of AEWs from year to year (Burpee 1974; Frank 1975; Avila et al. 2000; Goldenberg et al. 2001). Here, to facilitate discussions and to compare our results with previous studies, we categorized, as defined in section 2, AEWs as either developing or nondeveloping. The statistics and characteristics of hurricanes associated with developing AEWs versus nondeveloping AEWs were documented based on our analysis and are compared with results obtained from previous studies.

Various multiscale analysis methods have been used in climate and weather studies. Familiar methods include 1) principal component analysis (PCA, or empirical orthogonal functions), 2) a spectral analysis with a Fourier transform, and 3) a wavelet analysis (e.g., North 1984; Dunkerton 1993; Hendon and Wheeler 2008; Torrence and Compo 1998). Often, the application of these methods is limited to linear or stationary systems with a requisite of a priori knowledge. However, atmospheric dynamics are often nonlinear and nonstationary, and a priori knowledge is not known in advance. To overcome these issues, we adopted the empirical mode decomposition (EMD), a part of the Hilbert–Huang transform (Huang et al. 1998) for decomposing a nonlinear and nonstationary dataset in order to sift out a set of so-called intrinsic mode functions (IMFs). Previous studies have mainly applied the Hilbert–Huang transform for analyzing 1D time series data in order to examine the time–frequency distribution (e.g., Wang et al. 2012). Limited studies have focused on decomposition using the 2D EMD (e.g., Wu and Huang 2009). In this study, an improved version of the ensemble EMD (EEMD) technique was applied to the two-dimensional spatial field for every 6-h time frame over the entire study period. We then examined the time evolution of the trend mode (basic state) and the TC representing the IMF mode, and thus scale interactions.

The ERA-Interim dataset and the methodology employed in this study are described in section 2. Our results and a discussion are presented in section 3, and a summary and conclusions are provided in section 4.

2. Dataset and methodology

In this study, the EEMD was employed to decompose the ERA-Interim global reanalysis wind data for the analysis of multiscale interaction processes during tropical cyclogenesis development over a 10-yr period. The study domain for our analysis covered latitudes between 7° and 20°N, and longitudes between 15°E over West Africa and 60°W over the tropical Atlantic, the same as for the MDR study by Hopsch et al. (2007) that performed a 22-yr analysis of AEWs and TCs. AEW tracking was first conducted in order to collect all cases of AEWs. The time evolution of the horizontal shear from the original ERA data and the extracted IMFs was then analyzed for AEW cases. In sections 2a and 2b, we briefly describe the ERA-Interim dataset and the tracking techniques, respectively. The shear calculation is described in section 2c, and the method for the EMD is outlined in section 2d.

a. ERA-Interim dataset

The ERA-Interim dataset is the latest global reanalysis product created by the ECMWF and spans a period from January 1979 to the present day. Gridded products include 3-h surface fields and daily vertical integrals, and 6-h upper-air atmospheric fields that have a horizontal resolution of 0.75° (~80 km) and vertical coverage from the lower troposphere to the stratosphere (up to 0.1 hPa) on 60 model layers. Six-hour atmospheric fields are also provided for pressure, potential temperature, and potential vorticity levels. In this study, we used upper-air wind, relative vorticity, and geopotential height on the 700-hPa pressure level from July to September over the 10-yr period from 2004 to 2013. A more detailed description of ERA-Interim data can be found in Dee et al. (2011) and Berrisford et al. (2011).

b. Tracking AEWs and tropical cyclones

AEW movement can be determined by identifying the features or characteristics of AEWs. For example, one of the most significant features of an AEW is its trough, where the meridional wind (the υ component) changes signs in easterly shear flow below the level of the AEJ (e.g., 700 or 850 hPa). Maximum relative vorticity centers at these pressure levels are another feature of pronounced AEWs and have widely been used to track their movement (e.g., Reed et al. 1977). Other approaches for AEW tracking include using streamline charts (e.g., Reed et al. 1988) or locating the streamfunction minimum (e.g., Hopsch et al. 2010) at the levels of the AEJ. Supplementary use of satellite imagery to track clouds associated with AEWs is also often employed. Streamline charts can indicate developed AEW circulation at 700–850 hPa, while the streamfunction minimum reflects a trough in geopotential height as in quasi-geostrophic flow the streamfunction is proportional to geopotential height (Nielsen-Gammon and Gold 2006). Combinations of these approaches are often applied to provide better tracking results. For example, Fink et al. (2004) identified AEW episodes using the trough line located where the meridional wind (υ) changes in sign on Hovmöller diagrams at 700 and 850 hPa, overlapping with the local vorticity maxima and streamline charts. Reed et al. (1988) tracked AEWs using trough lines at 700 hPa superimposed with the local vorticity maximum on streamline charts and also observed clouds from satellites. For our tracking procedure, we identified AEW characteristics using the following tools: 1) the average meridional wind shown in a time–longitude diagram, 2) the instantaneous meridional wind (υ), and 3) the relative vorticity and geopotential height (h) supplemented by wind flowcharts at 700 hPa.

To obtain a time evolution for the westward movement of troughs, our AEW tracking efforts in a particular year began with a time–longitude diagram of the meridional wind component averaged over 7°–20°N for a continuous 3-month period (July–September). The exact AEW trough line was first identified by finding the location of the sign change in the υ-component wind. The wave center was then determined through the superimposition of the local center of negative deviation of geopotential height (Hgp) from the mean over the trough line. We also used the following conditions to ensure the location of the AEW when the local Hgp center is not well defined: 1) the vortex center and/or 2) the vorticity maxima when no clear closed circulations existed. Our procedures for tracking nondeveloping and developing AEWs are discussed below.

For the present study, AEW cases were categorized as either developing or nondeveloping. An AEW that has developed into a tropical depression (TD) or a stronger class storm tracked and named by the National Hurricane Center (NHC), and is thus associated with an NHC tracked TC is referred to as a developing AEW. Otherwise, it is referred to as a nondeveloping AEW. For nondeveloping AEW cases, our tracking strategy searched for wave disturbance around the West African coast near the Guinea Highlands, where AEW signals are the most pronounced (Chen 2006; Hopsch et al. 2010) and can be easily identified. Once identified, waves were backtracked to the African continent and forward tracked by following signals out into the ocean. An AEW was selected for the study only if AEW signals were identifiable and persisted for more than 24 h. For developing AEW cases, we used best tracks from the NHC and then backtracked wave disturbances to the African continent, as for the nondeveloping cases. In some cases a slight but nontrivial displacement occurred between the circulation center resolved using the ERA-Interim wind field and that determined from the NHC’s best-track data. Therefore, an adjustment for determining a storm center in the ERA-Interim data was made in order to properly represent the shear strength of the corresponding AEW and/or the TC.

We calculated statistics for AEWs within selected MDRs and West Africa. We also continued to track AEWs that developed into hurricanes outside of the selected regions and that went beyond 60°W and 20°N. Table 1 lists AEWs and tropical storms within the studied tropical Atlantic Ocean domain (backtracked to the African continent when possible) during the study period.

Table 1.

Breakdown by year of AEWs (second column) and the NHC tracked storms (third column) that developed as a result of AEWs from July to September from 2004 to 2013 for the MDR provided in Fig. 3. Numbers between the parentheses in the third column include TDs of a non-AEW origin. Also shown in the fourth column are hurricanes that developed from these AEW associated storms. As discussed in section 3, the number between the square brackets in this column indicates hurricanes with downscaling features.

Table 1.

c. Calculation of shear magnitude

Relative vertical vorticity (ζrel), defined as ζrel = (), is a characteristic of TCs and AEWs, and its time evolution is an indicator of the growth or decay of TCs and AEWs. For this study, we analyzed the horizontal shear of u-component and υ-component winds by defining ζ1 = and ζ2 = , respectively. To understand the association between the intensification of a TC (or an AEW) and the reduction of shear in environmental flow, one of the focuses of our study was the evolution of wind shear.

Storm strength was examined via an area average of wind shear within a moving domain centered at the storm’s center along storm tracks. The shape and/or size of the domain [e.g., a rectangle vs a square domain (e.g., 9° × 3°, 6° × 6°) and a small vs large size (e.g., 6° × 6°, 12° × 12°), respectively] were tested to ensure that the area size and shape that we choose did not sway the trend of the averaged shear. In general, the shear magnitude averaged over a smaller area tended to have a noisy variation with a larger average magnitude. The value of the average shear magnitude, when averaged over a larger area, tended to be smaller with a smoothed-out variation tendency. Nevertheless, a similar trend (but with differing magnitudes) for average shear was found regardless of the size of the area used to calculate the average. Given the fact that the average hurricane extent is ~3°–6° in radius of the outermost closed isobar (ROCI) (Merrill 1984) in active databases of the ROCI maintained by the NHC and that tropical easterly flow lies elongated in the x direction, in this study an area of 9° × 4.5° (twice and equivalent to the average storm extent in the zonal and meridional directions, respectively) was selected for the areal average of the shear calculation. Thus, the shear magnitude analysis reported in section 3 (the results and discussion section) is based on the averaged shear magnitude over an area of 9° × 4.5° from a storm’s center.

d. Ensemble empirical mode decomposition (EEMD)

The EMD was first introduced as part of the Hilbert–Huang transform (HHT; Huang et al. 1996). In contrast to scale analysis methods that assume stationarity and linearity in the data (e.g., Fourier transform), the EMD method utilizes local minima and maxima to extract IMFs from the data. Thus, the method is adaptive to locality and does not require a priori knowledge regarding the data. The EMD decomposes a given dataset into N oscillatory IMFs and the residual is a nonoscillatory mode that is defined as the trend mode. Each of the extracted (N) IMFs represent a mode at comparable scales, as opposed to a simple harmonic function within the Fourier transform The mode with the highest frequency (shortest wavelength) is extracted first, followed by the extraction of IMF modes with lower frequency in succession. The trend mode is determined by calculating the difference between the original data and the summation of the N oscillatory IMFs. A detailed description of the EMD formulation and the procedure for IMF extraction is provided in appendix A. The EMD method has been applied widely in engineering and in science fields such as neuroscience, biomedical and ocean engineering, seismic studies, atmospheric science, etc. (e.g., Fine et al. 2010; Pigorini et al. 2011; Huang et al. 1998; Salisbury and Wimbush 2002; Schlurmann 2002; Wang et al. 2012).

Although encouraging performance for the EMD in data analysis has been documented, issues such as mode mixing and aliasing have been reported. Mode mixing is the presence of signals of disparate scales in an IMF mode, or a signal of a similar scale in different modes, as a result of signal intermittency (Huang et al. 1998) that could cause aliasing in the time–frequency distribution and lead to indistinct physical meaning of the individual IMFs. To alleviate the mode mixing problem, the EEMD (Wu and Huang 2009) was introduced. By dealing with an ensemble of datasets, each of which contains the original observational data and finite amplitude white noise, the EEMD produces one set of IMFs for each ensemble member using the EMD and performs an ensemble average in order to obtain the final IMFs. Using the EEMD (e.g., Wu and Huang 2009), the chance of scale mixing is reduced and the added noises cancel out one another for the mean IMFs. A brief summary of the methodology of both EMD and EEMD can be found in appendix A. In a recent application, Shen et al. (2016) demonstrated the capability of the EEMD in extracting mixed Rossby–gravity (MRG) waves and the westerly wind belt (WWB) from an idealized 2D wind field (Fig. 1), which is a linear combination of the analytical solutions for an MRG and WWB (e.g., Shen et al. 2012 and references therein). In studying Sandy, the 2012 Atlantic hurricane, Shen et al. (2016) were able to extract oscillatory modes that resembled an MRG and an equatorial Rossby wave using the newly developed parallel EEMD (PEEMD). For this study, the PEEMD, achieving a parallel speedup of 720 times using 200 eight-core processors, was used to decompose the u-component and υ-component wind fields for the multiscale analysis.

Fig. 1.
Fig. 1.

An idealized tropical wave: (a) zonal wind perturbation (U′), (b) WWB, (c) total wind (U′ plus WWB), (d) IMF6 from the EEMD of (c),(e) IMF9 from the decomposition of (c),(f) IMF6 plus IMF9, (g) correlation of the IMG6 and U′, and (h) correlation of the IMF9 and WWB. Data obtained from Shen et al. (2012).

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

In the EMD, as well as the EEMD or PEEMD, the trend mode is obtained by subtracting the first (N) IMFs from the original data. As a result of the iterative nature of the sifting process, representation of the trend mode may depend on the choice of the actual number of N. To examine the dependence on the choice of N, we performed experiments using different numbers of N, including N = 6, 7, and 8, within the EEMD analysis of the 19 680 time frames of the 10-yr ERA-Interim dataset. Here, the corresponding cases are referred to as cases N6, N7, and N8, respectively. Our results indicate that the trend mode for cases N7 and N8 highly resembled one another, with a correlation coefficient of 0.98 or higher for the entire study period. Such a high correlation did not appear in trend mode cases N6 and N8, where the correlation coefficients varied from 0.898 to 0.99 for the various time frames. Apparently, some nonnegligible differences between case N6 and case N8 (and case N7) exist. Therefore, we suggest that case N8 is better suited than case N6 for the determination of trend modes within the 10-yr ERA-Interim dataset. In addition, given better computational performance and comparable trend modes for case N7 as compared to case N8, the obtained results using the PEEMD with N = 7 are discussed below.

In this study, the nonoscillatory trend mode was used to relate the basic-state sheared flow. On the other hand, by parsing through the winds and shear/vorticity fields from all of the AEW cases (see an example in appendix B), particularly the developing ones, we found that the third oscillatory mode (IMF3) poses a significant portion of the total vorticity, as compared to other IMFs, and that the sequence-to-sequence alignment between a TC system and the corresponding IMF3 is the most consistent. Here, the sequence-to-sequence alignment indicates the correspondence in time and space between the two datasets, that is, the observed TC and a specific IMF. These analysis results suggest that the IMF3 is the best representative of a TC system.

3. Results and discussion

The AEW cases we collected for this study are provided in Fig. 2 and Table 1. Figure 2 provides the tracks of AEWs in the time–longitude diagram of meridional winds during the 10-yr study period. Some AEWs propagated across the tropical Atlantic Ocean to the western region of the ocean and experienced enhancement and/or weakening during their journey, while others were early dissipaters. On the other hand, a few AEWs recurved northward and became/or merged with extratropical disturbances. Generally speaking, higher populations of AEWs occur in July and August as compared to September, when AEWs are less active and less organized (Fig. 2). The annual number of AEW occurrences was relatively steady, with the number of total AEWs shifting between 26 and 30 for 9 of the 10 study years during the July–September period (also shown in Table 1). The exception was 2013, which had only 24 AEWs. The near-constant annual number of AEWs seems to suggest “stable” large-scale forcing (e.g., the seasons remained nearly unchanged interannually).

Fig. 2.
Fig. 2.

(a) Time–longitude diagrams for meridional winds at 700 hPa averaged over the latitudes of 7°–20°N from the ERA-Interim 0.75° reanalysis from 2004 to 2008. Green lines indicate tracked AEWs that displayed traceable signals of AEWs and that persisted for more than 24 h without interruption. Dark green lines indicate NHC tracked storms that were associated with AEWs. The dotted portion signifies that storms are moving north of 20°N. (b) As in (a), but for 2009–13.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

For the 10-yr study period, we determined that 42 storms were associated with developing AEWs within the study domain. A breakdown of these storms by year is provided in Table 1, and the statistics of their activities for the ocean portion of the study domain are provided in appendix B. The developing rate was one in 6.5 (42/272 = 1/6.5) AEWs, on average. The ratio is comparable to that of 91 TCs to 603 total AEWs as determined by Hopsch et al. (2010) for a 22-yr study period. Further analysis suggested that 25 storms of the 42 developing cases eventually developed into hurricanes, including 10 hurricanes that developed west of 60°W longitude and/or north of 20°N latitude (three at >25°N).

While the number of detected annual AEWs was nearly constant, the number of developing AEWs varied greatly from year to year (from four to eight TCs annually) during the study period. Most of these storms were identified along the so-called southern track, consistent with previous research (Chen et al. 2008; Hopsch et al. 2007). The varying number of TCs from year to year indicates that the forcing of TC genesis may come from source(s) other than large-scale flows (i.e., AEWs). However, here, to simplify the discussions, we focus on the association of TC genesis with AEWs.

In the following, we examine the location of the maximum horizontal wind shear in the path of westward-moving AEWs in order to illustrate the differences over ocean and over land. Both the total wind and the decomposed IMFs of the zonal (u) and meridional (υ) winds were analyzed (Figs. 3 and 4; Tables 2 and 3). Figures 3 and 4 provide the maxima of horizontal shears associated with zonal and meridional winds, denoted as maxζ2 and maxζ1, respectively, over land versus over ocean. Common wide scattering and low correlations (0.46 for developing and 0.31 for nondeveloping) are presented in Fig. 3. For the total wind and IMF3 (representing TC-scale systems), the most remarkable difference between the developing and nondeveloping cases was that developing cases showed a substantially higher maxζ2 over the ocean as compared to over land (Table 2). For the developing cases, the maxζ2 of the total wind and the IMF3 were determined to be 1.39 and 0.22 × 10−5 s−1 over land and 1.89 and 0.38 × 10−5 s−1 over ocean. For the nondeveloping cases, the variation in the overall average maxζ2 between over land and over ocean was not significant (Fig. 3c), with values changing from 1.16, 0.20, and 0.42 (×10−5 s−1) to 1.28, 0.23, and 0.42 (×10−5s−1) for the total wind, the IMF3, and the trend mode, respectively. For the overall average of the trend mode (environmental flow), while maxζ2 had the same magnitudes over land as for over ocean for nondeveloping cases, they increased only slightly from over land to over ocean for the developing cases. The spatial distribution of maxζ2 based on the trend mode is analyzed in greater detail below using Fig. 5. The maxζ1 analysis displayed similar patterns (Fig. 4) to that of maxζ2. Wide scattering and low correlations (0.34–0.5) can be seen for maxζ1 over land versus over ocean for both the developing and nondeveloping cases. The overall maxζ1 for the total wind and the IMF3 were determined to be 0.45 (0.56) and 0.27 (0.33) × 10−5 s−1 for over land (over ocean), respectively, for the nondeveloping cases; and 0.51 (0.99) and 0.30 (0.51) × 10−5 s−1, respectively, for over land (over ocean) for the developing cases. Note that ζ1 was near zero for the trend mode, as was maxζ1. The increase in the shear magnitude of the total wind and of the IMF3s (both ζ1 and ζ2) from over land to downstream over ocean suggests that intensification in the TC scale mainly occurs after storms move out over the ocean.

Fig. 3.
Fig. 3.

The maximum of ζ2 (10−5 s−1) over ocean vs over land: (a) all of the AEW cases, (b) the developing cases, and (c) the nondeveloping cases. Plotted values for the IMF3 and the trend mode were multiplied by two. Black dots indicate total wind, blue dots indicate the IMF3, and red dots indicate the trend mode.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

Fig. 4.
Fig. 4.

As in Fig. 3, but for the maximum of ζ1 (10−5 s−1). The ζ1 for the trend mode had near-zero values for all cases and rendered the correlation nonmeaningful.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

Table 2.

The sample mean of the maxζ2 (10−5 s−1) for the developing (D) and nondeveloping (ND) cases. TotWind denotes total wind and TrendM denotes trend mode. The last column provides the correlation of maxζ2 between over land and over ocean.

Table 2.
Table 3.

The sample mean of the maxζ1 (10−5 s−1) for the D and ND cases. The last column provides the correlation of maxζ1 between over land and over ocean. The trend mode (TrendM) of the υ wind was near zero, yielding a magnitude of ζ1 with an average of zero.

Table 3.
Fig. 5.
Fig. 5.

A histogram of the longitudinal location for the maximum of ζ2: (a) all AEW cases, (b) the developing cases, and (c) the nondeveloping cases. Gray, blue, and red bars represent total wind, the IMF3, and trend mode, respectively. Note that the West African coast is located at ~17°W.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

To illustrate the spatial variation of maxζ2 along the storm pathway from the African continent across the studied regions to western Atlantic Ocean regions, a histogram of the longitudinal location of maxζ2 occurrence is provided in Fig. 5. For all cases (most of them nondeveloping cases), the maxζ2 of the total wind largely appeared between 10° and 30°W and was most significant around 20°W, just off the North African coast (the shoal line was ~17°W). The result may reflect an AEW intensification scenario over the Guinea Highlands and in the immediate downstream area off the northern West African coast, consistent with results from previous studies (e.g., Hopsch et al. 2007). The longitudinal location of maxζ2 for the trend mode and the IMF3 displayed a similar pattern, with a high appearance rate between 10° and 30°W.

By comparing the developing and nondeveloping cases, we found significant differences in the longitudinal location of maxζ2 (Fig. 5). First, the longitudinal location of maxζ2 for the total wind and the IMF3 (TC scale) occurred overwhelmingly over the ocean for the developing cases, with that of the total wind occurring immediately off the coast (20°W and westward) and with IMF3 spreading across the ocean. While the longitudinal location of maxζ2 of the trend mode also occurred closer to the coast, its inland distribution was higher over the Guinea Highlands for the developing cases. These phenomena suggest that developing cases (AEWs) may obtain strength throughout their interaction with the Guinea Highlands and that associated storms may intensify over the ocean along the course of AEW development. One possible contributor to the strengthening of storms is the downscaling process (i.e., the transfer of shear from the environmental trend mode to the TC-scale IMF3). The longitudinal location of maxζ1 (Fig. 6) displayed similar distribution patterns to that of maxζ2 but with minor differences, namely, the lack of a drastic peak near 20°N as for the maxζ2 for nondeveloping cases. Instead, twin peaks of maxζ1 were determined around 30° and 10°–15°W.

Fig. 6.
Fig. 6.

As in Fig. 5, but for the maximum of ζ1. Note that ζ1 for the trend mode had near-zero values for all cases.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

For the developing cases, the time evolution of the wind shear along their tracks was examined to reveal downscaling processes during intensification. In several cases, the shear of the trend mode began to decrease during storm intensification, departing from the shear tendency of total wind shear. At the same time, the shear of the IMF3 gained strength. As shown in Fig. 7, such an example can be found in the ninth NHC storm (Helene) in 2006 (e.g., Shen et al. 2010), where storm intensification is depicted by the dropping of sea level pressure at the storm’s center. The decrease of the trend mode shear and the increase of shear in the IMF3 during storm intensification suggest downscaling transfer from the trend mode to the IMF.

Fig. 7.
Fig. 7.

The tendency for horizontal wind shear for the total winds (black line), IMF3 (blue line), and trend mode (red line) and SLP (black dotted line) along the storm track of Helene (2006). The ζ2 is plotted as solid lines, the ζ1 as dashed–dotted lines, and SLP as a black dotted line. The plotted shear of the IMF3 and the trend mode were multiplied by three. The vertical thin dotted line indicates the time the storm was first classified by the NHC as a TD. The two vertical gray lines indicate the beginning and the end of intensification for the downscaling feature. The enhanced black dotted line for SLP indicates storm intensification with an SLP drop greater than 2.5 hPa/6 h. The red dashed–dotted line is plotted as a value of zero.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

Among the 42 developing AEWs, 13 cases exhibited different degrees of decrease in the shear of the trend mode and an increase in the shear of the IMF3 during storm intensification. All of them developed into hurricanes. In addition to Helene (2006) in Fig. 7, the rest of the 13 cases are presented in Fig. 8 and Table 4. The storm intensification period is simply defined as the period with a continuous decrease in sea level pressure (SLP) at the storm’s center. Note that two of the cases (storm 5 of 2005 and storm 4 of 2007) experienced a brief disruption in the decrease of SLP, but their pressure tendency displayed a strong decreasing transition. The overall tendency in SLP and shear for the 13 developing cases are discussed below, while their life spans and storm tracks are described in appendix B. The intensification period lasted from 60 to 168 h with the fall of the storm center SLP ranging from 21 for the least intense case (storm 12, 2011) to 75 hPa for the most intense case (storm 4, 2007). The decline in the horizontal shear of the zonal wind (ζ2) for the trend mode ranged from 0.17 to 0.86 × 10−5 s−1, and strengthening of ζ2 for the IMF3 varied from 0.1 to 0.66 × 10−5 s−1. The analysis of the horizontal shear of the meridional wind (ζ1) displayed a similar behavior as for ζ2 during storm intensification. The horizontal shear of the meridional wind for the IMF3 was enhanced by a magnitude ranging from 0.04 to 0.56 × 10−5 s−1. On average, as shown in Fig. 8, SLP decreased by 53 hPa and the horizontal shear of the zonal wind (ζ2) for the trend mode declined by 0.37 × 10−5 s−1 during the storm intensification period. The strength of both ζ1 and ζ2 for the IMF3 were enhanced by 0.28 × 10−5 s−1. Since the upscaling process was not accounted for during our analysis, the reader should note that the sum of the enhancements for wind shear in the IMF3s (ζ1 + ζ2) may be larger than the decline in the trend mode.

Fig. 8.
Fig. 8.

As in Fig. 7, but for the 12 developing cases that display a downscaling feature in the horizontal shear during storm intensification for the study period.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

Table 4.

SLP and wind shear magnitude at the beginning and end of storm intensification, as well as their difference for the 13 storms displaying downscaling features. The ζ2b and ζ1b indicate shear values (×10−5 s−1) at the beginning for the zonal and meridional winds, respectively. The ζ2e and ζ1e indicate shear values at the end for the zonal and meridional winds, respectively. The difference is defined as the ending value minus the beginning value. Hurricanes were categorized from NHC reports based on the Saffir–Simpson hurricane wind scale.

Table 4.

Among the 42 developing AEWs, 25 turned into hurricanes, with maximum shear appearing over the ocean (Fig. 5). Thirteen hurricanes displayed significant downscaling transfer features. Thus, other processes such as upscaling processes may contribute to the intensification of the other developing cases, namely, the remaining 12 hurricanes. However, since the 0.75° resolution of ERA-Interim data may not be appropriate for resolving mesoscale processes, we did not attempt to explore the upscaling processes associated with mesoscale and/or convective-scale processes. These processes will be explored in a future study by applying the PEEMD to high-resolution global simulations and/or global reanalysis.

4. Summary

In exploring multiscale processes of tropical cyclone genesis associated with AEWs, the shear variation of different IMFs from the PEEMD analysis was examined for AEWs over a 10-yr period (July–September from 2004 to 2013) in order to investigate the downscaling process. AEWs were tracked and categorized into two groups: 1) developing cases, in which an AEW eventually developed into an NHC named TC; and 2) nondeveloping cases, in which an AEW did not develop into a named storm. The PEEMD technique was used to decompose raw ERA-Interim wind data into a set of N oscillatory IMFs at different scales and a nonoscillatory residual, the latter of which is used to define the basic-state trend mode. Based on our 10-yr data analysis, we suggest that the trend mode of zonal wind represents environmental basic sheared flow and that the third IMF (IMF3) represents a mode shared by the TC and AEW signals. Since our focus in this study was on scale interactions between basic flow and “perturbations,” a detailed separation of TC and AEW signals will be examined in a future study. The 10-yr multiscale analyses between the IMF3 and the trend mode were documented in order to establish statistics for shear magnitude changes along AEW tracks within the selected MDR domain and to explore the downscaling process in the development of TCs associated with AEWs.

Our analyses indicated that the number of AEWs remained nearly constant (from the ERA-Interim dataset) and that the number of TCs changed greatly from year to year (from the NHC’s best-track dataset). The former seems to suggest “stable” large-scale forcing (e.g., seasonal forcing remains nearly unchanged interannually), while the latter suggests various mechanisms for the initiation and further intensification of an AEW that may or may not develop into a TC. During the 10-yr period analyzed, 42 AEWs developed into NHC classified storms within the selected study domain and 25 further developed into hurricanes. We examined the scale interactions for the developing AEWs and focused on downscaling processes.

Multiscale analyses of the trend mode and IMF3 suggest that 13 of the 42 developing AEWs exhibited a downscaling feature for shear transfer from the trend mode to the IMF3 (Figs. 7 and 8). All of the 13 cases developed into hurricanes. For these 13 cases, the average SLP drop was 53 hPa during the intensification phase. The average decrease of the horizontal shear of the zonal wind for the trend mode, representing a decrease in the basic-state wind shear, was 0.37 × 10−5 s−1, while the average enhancement of the horizontal shear of zonal wind (as well as the horizontal shear of meridonal wind) for the IMF3 was 0.28 × 10−5 s−1 during storm intensification. The total increase in wind shear in the IMF3s (i.e., ζ1 + ζ2), larger than the decrease of wind shear in the trend mode, may suggest that additional processes (e.g., upscaling processes) also contribute to the intensification of the IMF3. In a future study, we will examine these processes using finer-resolution datasets.

The IMF3 for the developing cases displayed a substantially higher maximum shear magnitude over ocean as compared to over land. In comparison, the trend mode for most of the developing cases displayed an increase in shear over the Guinea Highlands, as indicated by the result that the longitudinal distribution of maximum shear magnitude peaked between the Guinea Highlands and 20°W. Thus, a potential for the subsequent downscale transfer from the trend mode to the IMF3 may exist. In other words, after tropical waves pass by the Guinea Highlands and move westward, AEWs may intensify and provide a downscale transfer to storms for their development over the ocean.

Since only 13 of the 25 hurricanes were found to contain significant downscaling transfer features, upscaling processes may play an important role for the other developing cases, especially the other 12 hurricane cases. A future study that employs the EEMD with global mesoscale model simulations at a resolution of 0.25° or finer is being planned so that we can examine the impact of upscaling processes associated with convective systems on tropical cyclogenesis. The performance of the PEEMD in revealing the feedback of small-scale processes will be first examined by decomposing the solutions from the nonlinear Lorenz model (Lorenz 1963), high-order Lorenz models (Shen 2014, 2015a), and a revised Lorenz model with parameterization (Shen 2015b), and then comparing the corresponding IMFs. The newly derived Lorenz modes are chosen because of their capability in simulating additional small-scale processes associated with high-wavenumber modes and/or parameterizations.

Acknowledgments

We are grateful to the following organizations for their support: the College of Sciences at San Diego State University, the Earth System Science Center at the University of Alabama in Huntsville, the NASA Earth Science Technology Office, and the NASA Advanced Information Systems Technology Program (AIST11-NNX14AP05G supported by Goddard Space Flight Center). Resources supporting this work were provided by the NASA High-End Computing Program and the NASA Advanced Supercomputing Division at Ames Research Center.

APPENDIX A

The EMD and EEMD Methods

In this appendix, we discuss the method and procedures for EMD/EEMD for calculation of the IMFs.

We used the following notation. The index i is used to indicate the ith IMFs. We used 1) t as the independent variable that can be either time, longitude, or latitude; 2) X(t) as the original data; 3) R(t) as the residual of the data; 4) S(t) as the data to be shifted out; 5) Ci(t) as the ith IMF; and 6) U(t) [L(t)] as the upper (lower) envelope by connecting maxima (minima) with a cubic spline. Therefore, i = 0, R(t) = X(t), and S(t) = R(t) initially. The S(t) and R(t) are changed after the beginning of the so-called shifting processes that are summarized as follows:

  1. Let i = 0 and , where X(t) is the original data;
  2. Let ;
  3. Identify the extrema (both maxima and minima) of the data S(t);
  4. Generate the upper (lower) envelope, ;
  5. Determine the local mean, , by averaging the envelopes;
  6. Let , that is, subtract out the mean from the data;
  7. Go to step 3 unless stoppage criteria for an IMF are met;
  8. i = i + 1, = S(t), , and go to step 2.

For each IMF, we performed 10 iterations to obtain the responding IMF, namely, steps 2–7 were repeated 10 times.

Based on the above-mentioned sifting processes, IMFs possess the following features. The IMFs are symmetric with respect to the local zero mean and have the same number of zero crossings and extrema. The IMFs are time (spatial) domain functions that represent the local variability of the original signal at a particular range of frequencies (wavelengths). Mathematically, a strict definition of the IMF is as follows: U(t) + L(t) = 0 (Wang et al 2010), namely, its local means are zero.

Using the above-mentioned processes, the IMFs and residual functions are defined at the same grid points as those for the raw data. The residual Rm represents the differences between the raw data and a sum of the first mth IMFs. In other words, the time (or spatial) series of the raw data X(j) can be represented by the first mth IMFs and Rm as follows:
eq1
where X(j) represents the value of X at t = jΔt or x = jΔx: here Δt and Δx are the temporal and spatial increments, respectively. Similarly, X(j) can also be represented by the first (m + 1)th IMFs and Rm+1, as follows:
eq2
The above-mentioned two equations lead to
eq3
With no loss of generality, we have
eq4
Since (j) is purely oscillatory with respect to (j), we can consider (j) as the local mean of (j). Thus, EMD is a “Reynolds type” decomposition for sifting out (or extracting) periodic components from the data by separating the local mean from the fluctuations using spline fits (Huang et al. 1998).

For the ensemble EMD, the follow steps were performed:

  1. Add a noise series to the targeted data;
  2. Decompose the data with added noise into IMFs using the products of the EMD;
  3. Repeat steps 1 and 2 again and again but with a different noise series each time;
  4. Obtain the (ensemble) means of the corresponding IMFs of the decompositions as the final result.
With the EEMD, the chance of scale mixing is reduced, while the dyadic property is preserved. As indicated by Wu and Huang (2009), for the mean IMFs, added noises cancel out one another. We also implemented a three-level parallelism into the EEMD, which is referred to here as the PEEMD.

APPENDIX B

Statistics of the Developing Cases and Characteristic Life Span and Storm Tracks of the Downscaling Cases

Statistics for TC activities were obtained for the developing cases by tallying activities over the ocean portion of the study domain. We found that TC genesis, defined as the time when a storm was first qualified and classified as a TD or a tropical storm (TS), occurred more frequently within the eastern Atlantic Ocean region (i.e., the region immediately off the West African coast; Fig. B1). However, the frequency of TC genesis remained fairly high across the rest of the selected regions. The determined number of TD occurrences was 20, 14, and 17 for the eastern, middle, and western Atlantic Ocean regions, respectively. In contrast, stronger storms, such as TSs and/or hurricane classes, occurred farther downstream within the middle and western Atlantic Ocean regions. The determined number of TSs (hurricanes) in the eastern, middle and western Atlantic Ocean was 12 (4), 17 (7), and 18 (6), respectively. For the 10-yr period, we determined 41 TDs and 37 TSs within the selected study regions. One TS (sixth NHC storm, 2009) was classified directly from a tropical wave over a 6-h period. In total, 42 storms were associated with developing AEWs, and 25 of them developed into hurricanes.

Fig. B1.
Fig. B1.

AEWs associated with named TCs in the Atlantic region from July to September between 2004 and 2013. Boxes divide the tropical Atlantic Ocean into eastern, middle, and western regions, following Hopsch et al. (2010). Green dots indicate cyclones that were first classified as TDs and blue dots indicate TSs, according to the NHC’s best-track data. Colored numbers indicate the total number of storms formed in each box during the 10-yr study period. TC genesis, as defined by the classification of TD, occurs most frequently within the eastern Atlantic Ocean. A higher frequency of stronger storms (TSs and hurricanes) formed farther within downstream region storm tracks (mid-Atlantic) than in the region immediately off the African coast.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

Among the 42 developing cases over the 10-yr study period, 13 exhibited the downscaling feature of shear transfer from the trend mode to the IMF3. The tracks of these storms with downscaling features are provided in Figure B2. All of the 13 cases developed into hurricane-class storms at some point during their life spans. The duration of these storms, from TD/TS formation to dissipation or until becoming an extratropical storm ranged from 8 to 15 days, with an average time span of 11.00 days. In a similar manner, the time it took for these storms to develop from the TD/TS class to the hurricane class varied widely from 1 day 6 h to 7 days 12 h, with an average of 3.34 days. Once they become hurricanes, these storms can take a long journey out of the tropics. The longest journey lasted 10 days (sixth NHC storm, 2004) and the shortest lasted only 1 day 6 h (ninth NHC storm, 2012). The average time these storms remained within the hurricane classification was 6.29 days. The wide range of evolution time frames suggests varying times for the release of shear instability from large-scale flow during the downscaling process; as reflected in the growth/decay of the IMF3/trend mode shear magnitude for these cases shown in Figs. 7 and 8 and Table 4. The wide range may also be associated with upscaling processes involving mesoscale systems that possess a greater degree of variability than larger-scale systems.

Fig. B2.
Fig. B2.

Tracks for the 13 storms and the 42 developing AEW cases showing the downscaling transfer of shear from the trend mode IMF to the short wave IMF3. The blue dotted line designates the tracks of the AEWs, the green lines indicate the tracks of TDs/TSs, and the red lines indicate a hurricane. The bolder portion of the tracks indicates storm intensification with a continuous drop in SLP. All of these storms developed into hurricanes.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

The progression of these developing cases is reflected in space in the spread of locations where they form TDs/TSs and/or hurricane-class storms. Of the 13 cases, 8 first formed TDs or TSs within the eastern Atlantic Ocean (east of 30°W), 3 formed within the mid-Atlantic Ocean (30°–45°W), and 2 formed within the western Atlantic Ocean (45°–60°W). Among those that first formed TDs/TSs within the eastern region, five became hurricane within the mid-Atlantic Ocean (30°–45°W), while the other three became hurricanes west of 45°W. For those forming TDs/TSs within the middle and western Atlantic Ocean, all became hurricane-class storms west of 45°W. Overall, most of the storms developed into hurricane-class storms west of 45°W, while the most rapid intensification (a decrease in SLP) occurred between 30° and 50°W.

To identify a specific IMF signal that can represent a TC system, the time sequence of winds and shear/vorticity for the TC systems (the developing cases) and various IMFs were examined for every case. The matching between the TCs and IMF3s was found to be most consistent. Such an agreement is demonstrated by the two time frames for Hurricane Helene (2006) given in Figure B3 as an example. Further analysis of spatial averaged vorticity showed that IMF3 is the top contributor in most AEW cases, especially the developing cases. In this study, the IMF3 is chosen to represent a TC system for its consistent sequence-to-sequence alignment with the TC system in the wind and vorticity fields and its significant contribution to the vorticity field.

Fig. B3.
Fig. B3.

(left) U wind and (right) vorticity fields (from ERA-Interim global reanalysis) over the tropical Atlantic at 1200 UTC 12 Sep 2006. (top) The original data that represent the sum of all the IMF modes and the trend mode, (middle) the IMF3, and (bottom) the trend mode. The green cross marks the location of Hurricane Helene from the NHC best-track data, and the number in green in the top-left panel shows the SLP at the center of the hurricane. (b) As in (a), but at 1200 UTC 15 Sep 2006.

Citation: Journal of Atmospheric and Oceanic Technology 33, 8; 10.1175/JTECH-D-15-0257.1

REFERENCES

  • Avila, L. A., , Pasch R. J. , , and Jiing J.-G. , 2000: Atlantic tropical systems of 1996 and 1997: Years of contrasts. Mon. Wea. Rev., 128, 36953706, doi:10.1175/1520-0493(2000)128<3695:ATSOAY>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Berrisford, P., and et al. , 2011: The ERA-Interim archive version 2.0. ERA Report Series 1, ECMWF, 23 pp. [Available online at http://www.ecmwf.int/sites/default/files/elibrary/2011/8174-era-interim-archive-version-20.pdf.]

  • Berry, G., , and Thorncroft C. , 2005: Case study of an intense African easterly wave. Mon. Wea. Rev., 133, 752766, doi:10.1175/MWR2884.1.

    • Search Google Scholar
    • Export Citation
  • Burpee, R. W., 1974: Characteristics of North African easterly waves during the summer of 1968 and 1969. J. Atmos. Sci., 31, 15561570, doi:10.1175/1520-0469(1974)031<1556:CONAEW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Carlson, T. N., 1969a: Some remarks on African disturbances and their progress over the tropical Atlantic. Mon. Wea. Rev., 97, 716726, doi:10.1175/1520-0493(1969)097<0716:SROADA>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Carlson, T. N., 1969b: Synoptic histories of three African wave disturbances that developed into Atlantic hurricanes. Mon. Wea. Rev., 97, 921925, doi:10.1175/1520-0493(1969)097<0256:SHOTAD>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Charney, J. G., , and Stern M. E. , 1962: On the stability of internal baroclinic jets in a rotating atmosphere. J. Atmos. Sci., 19, 159172, doi:10.1175/1520-0469(1962)019<0159:OTSOIB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chen, T.-C., 2006: Characteristics of African easterly waves depicted by ECMWF reanalyses for 1991–2000. Mon. Wea. Rev., 134, 35393566, doi:10.1175/MWR3259.1.

    • Search Google Scholar
    • Export Citation
  • Chen, T.-C., , Wang S.-Y. , , and Clark A. J. , 2008: North Atlantic hurricanes contributed by African easterly waves north and south of the African easterly jet. J. Climate, 21, 67676776, doi:10.1175/2008JCLI2523.1.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Dunkerton, T. J., 1993: Observation of 3–6-day meridional wind oscillations over the tropical Pacific, 1973–1992: Vertical structure and interannual variability. J. Atmos. Sci., 50, 32923307, doi:10.1175/1520-0469(1993)050<3292:OODMWO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Fine, A. S., , Nicholls D. P. , , and Mogul D. J. , 2010: Assessing instantaneous synchrony of nonlinear nonstationary oscillators in the brain. J. Neurosci. Methods, 186, 4251,doi:10.1016/j.jneumeth.2009.10.023.

    • Search Google Scholar
    • Export Citation
  • Fink, A. H., , Vincent D. G. , , Reiner P. M. , , and Speth P. , 2004: Mean state and wave disturbances during Phase I, II, and III of GATE based on ERA-40. Mon. Wea. Rev., 132, 16611683, doi:10.1175/1520-0493(2004)132<1661:MSAWDD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Frank, N. L., 1975: Atlantic tropical systems of 1974. Mon. Wea. Rev., 103, 294300, doi:10.1175/1520-0493(1975)103<0294:ATSO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Frank, W. M., , and Roundy P. E. , 2006: The role of tropical waves in tropical cyclogenesis. Mon. Wea. Rev., 134, 23972417, doi:10.1175/MWR3204.1.

    • Search Google Scholar
    • Export Citation
  • Goldenberg, S. B., , and Shapiro L. J. , 1996: Physical mechanisms for the association of El Nino and West African rainfall with Atlantic major hurricanes. J. Climate, 9, 11691187, doi:10.1175/1520-0442(1996)009<1169:PMFTAO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Goldenberg, S. B., , Landsea C. W. , , Mestas-Nuñez A. M. , , and Gray W. M. , 2001: The recent increase in Atlantic hurricane activity: Causes and implications. Science, 293, 474479, doi:10.1126/science.1060040.

    • Search Google Scholar
    • Export Citation
  • Hendon, H. H., , and Wheeler M. C. , 2008: Some space–time spectral analyses of tropical convection and planetary-scale waves. J. Atmos. Sci., 65, 29362948, doi:10.1175/2008JAS2675.1.

    • Search Google Scholar
    • Export Citation
  • Hopsch, S. B., , Thorncroft C. D. , , Hodges K. , , and Aiyyer A. , 2007: West African storm tracks and their relationship to Atlantic tropical cyclones. J. Climate, 20, 24682483, doi:10.1175/JCLI4139.1.

    • Search Google Scholar
    • Export Citation
  • Hopsch, S. B., , Thorncroft C. D. , , and Tyle K. R. , 2010: Analysis of African easterly wave structures and their role in influencing tropical cyclogenesis. Mon. Wea. Rev., 138, 13991419, doi:10.1175/2009MWR2760.1.

    • Search Google Scholar
    • Export Citation
  • Huang, N. E., , Steven R. L. , , and Shen Z. , 1996: The mechanism for frequency downshift in nonlinear wave evolution. Adv. Appl. Mech., 32, 59117, doi:10.1016/S0065-2156(08)70076-0.

    • Search Google Scholar
    • Export Citation
  • Huang, N. E., and et al. , 1998: The empirical mode decomposition and Hilbert spectrum for nonlinear and non-stationary time series analysis. Proc. Roy. Soc. London, 454A, 903995, doi:10.1098/rspa.1998.0193.

    • Search Google Scholar
    • Export Citation
  • Kistler, R., and et al. , 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247267, doi:10.1175/1520-0477(2001)082<0247:TNNYRM>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Landsea, C. W., 1993: A climatology of intense (or major) Atlantic hurricanes. Mon. Wea. Rev., 121, 17031713, doi:10.1175/1520-0493(1993)121<1703:ACOIMA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lau, K.-H., , and Lau N.-C. , 1990: Observed structure and propagation characteristics of tropical summertime synoptic-scale disturbances. Mon. Wea. Rev., 118, 18881913, doi:10.1175/1520-0493(1990)118<1888:OSAPCO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lorenz, E., 1963: Deterministic nonperiodic flow. J. Atmos. Sci., 20, 130141, doi:10.1175/1520-0469(1963)020<0130:DNF>2.0.CO;2.

  • Merrill, R. T., 1984: A comparison of large and small tropical cyclones. Mon. Wea. Rev., 112, 14081418, doi:10.1175/1520-0493(1984)112<1408:ACOLAS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Nielsen-Gammon, J. W., , and Gold D. A. , 2006: Dynamical diagnosis: A comparison of quasigeostrophy and Ertel potential vorticity. Synoptic–Dynamic Meteorology and Weather Analysis and Forecasting, Meteor. Monogr., No. 55, Amer. Meteor. Soc., 183–202, doi:10.1007/978-0-933876-68-2_9.

  • North, G. R., 1984: Empirical orthogonal functions and normal modes. J. Atmos. Sci., 41, 879887, doi:10.1175/1520-0469(1984)041<0879:EOFANM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Parker, D. J., and et al. , 2005: The diurnal cycle of the West African monsoon circulation. Quart. J. Roy. Meteor. Soc., 131, 28392860, doi:10.1256/qj.04.52.

    • Search Google Scholar
    • Export Citation
  • Pigorini, A., , Casali A. G. , , Casarotto S. , , Ferrarelli F. , , Baselli G. , , Mariotti M. , , Massimini M. , , and Rosanova M. C. E. , 2011: Time–frequency spectral analysis of TMS-evoked EEG oscillations by means of Hilbert–Huang transform. J. Neurosci. Methods, 198, 236245, doi:10.1016/j.jneumeth.2011.04.013.

    • Search Google Scholar
    • Export Citation
  • Pytharoulis, I., , and Thorncroft C. , 1999: The low-level structure of African easterly waves in 1995. Mon. Wea. Rev., 127, 22662280, doi:10.1175/1520-0493(1999)127<2266:TLLSOA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Reed, R. J., , Norquist D. C. , , and Recker E. E. , 1977: The structure and properties of African Wave disturbances as observed during phase III of GATE. Mon. Wea. Rev., 105, 317333, doi:10.1175/1520-0493(1977)105<0317:TSAPOA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Reed, R. J., , Hollingsworth A. , , Heckley W. A. , , and Delsol F. , 1988: An evaluation of the ECMWF operational system in analyzing and forecasting easterly wave disturbances over Africa and the tropical Atlantic. Mon. Wea. Rev., 116, 824865, doi:10.1175/1520-0493(1988)116<0824:AEOTPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Salisbury, J. I., , and Wimbush M. , 2002: Using modern time series analysis techniques to predict ENSO events from the SOI time series. Nonlinear Processes Geophys., 9, 341345, doi:10.5194/npg-9-341-2002.

    • Search Google Scholar
    • Export Citation
  • Schlurmann, T., 2002: Spectral frequency analysis of nonlinear water waves derived from the Hilbert-Huang transformation. J. Offshore Mech. Arct. Eng., 124, 2228, doi:10.1115/1.1423911.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., 2014: Nonlinear feedback in a five-dimensional Lorenz model. J. Atmos. Sci., 71, 17011723, doi:10.1175/JAS-D-13-0223.1.

  • Shen, B.-W., 2015a: Nonlinear feedback in a six-dimensional Lorenz model: Impact of an additional heating term. Nonlinear Processes Geophys., 22, 749764, doi:10.5194/npg-22-749-2015.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., 2015b: Parameterization of negative nonlinear feedback using a five-dimensional Lorenz model. Fractal Geom. Nonlinear Anal. Med. Biol., 1, doi:10.15761/FGNAMB.1000109.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Atlas R. , , Reale O. , , Lin S.-J. , , Chern J.-D. , , Chang J. , , Henze C. , , and Li J.-L. , 2006: Hurricane forecasts with a global mesoscale-resolving model: Preliminary results with Hurricane Katrina (2005). Geophys. Res. Lett., 33, L13813, doi:10.1029/2006GL026143.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Tao W.-K. , , and Wu M.-L. , 2010: African easterly waves in 30-day high resolution global simulations: A case study during the 2006 NAMMA period. Geophys. Res. Lett., 37, L18803, doi:10.1029/2010GL044355.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Tao W.-K. , , Lin Y.-L. , , and Laing A. , 2012: Genesis of twin tropical cyclones as revealed by a global mesoscale model: The role of mixed Rossby gravity waves. J. Geophys. Res., 117, D13114, doi:10.1029/2012JD017450.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , DeMaria M. , , Li J.-L. F. , , and Cheung S. , 2013a: Genesis of Hurricane Sandy (2012) simulated with a global mesoscale model. Geophys. Res. Lett., 40, 49444950, doi:10.1002/grl.50934.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Nelson B. , , Cheung S. , , and Tao W.-K. , 2013b: Improving NASA’s multiscale modeling framework for tropical cyclone climate study. IEEE Comput. Sci. Eng., 15, 5667, doi:10.1109/MCSE.2012.90.

    • Search Google Scholar
    • Export Citation
  • Shen, B.-W., , Cheung S. , , Li J.-L. F. , , and Wu Y.-L. , , and Shen S. S. , 2016: Multiscale processes of Hurricane Sandy (2012) as revealed by the parallel ensemble empirical mode decomposition and advanced visualization technology. Adv. Data Sci. Adapt. Anal., in press.

  • Thorncroft, C. D., , and Hoskins B. J. , 1994: An idealized study of African easterly waves. II: A nonlinear view. Quart. J. Roy. Meteor. Soc., 120, 9831015, doi:10.1002/qj.49712051810.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., , and Rowell D. P. , 1998: Interannual variability of African wave activity in a general circulation model. Int. J. Climatol., 18, 13051323, doi:10.1002/(SICI)1097-0088(1998100)18:12<1305::AID-JOC281>3.0.CO;2-N.

    • Search Google Scholar
    • Export Citation
  • Thorncroft, C. D., , and Hodges K. , 2001: African easterly wave variability and its relationship to Atlantic tropical cyclone activity. J. Climate, 14, 11661179, doi:10.1175/1520-0442(2001)014<1166:AEWVAI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Torrence, C., , and Compo G. P. , 1998: A practical guide to wavelet analysis. Bull. Amer. Meteor. Soc., 79, 6178, doi:10.1175/1520-0477(1998)079<0061:APGTWA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wang, G., , Chen X.-Y. , , Qiao F.-L. , , Wu Z. , , and Huang N. E. , 2010: On intrinsic mode function. Adv. Adapt. Data Anal., 2, 277293, doi:10.1142/S1793536910000549.

    • Search Google Scholar
    • Export Citation
  • Wang, T., , Zhang M. , , Yu Q. , , and Zhang H. , 2012: Comparing the applications of EMD and EEMD on time–frequency analysis of seismic signal. J. Appl. Geophys., 83, 2934, doi:10.1016/j.jappgeo.2012.05.002.

    • Search Google Scholar
    • Export Citation
  • Wu, Z. H., , and Huang N. E. , 2009: Ensemble empirical mode decomposition: A noise-assisted data analysis method. Adv. Adapt. Data Anal., 1, 141, doi:10.1142/S1793536909000047.

    • Search Google Scholar
    • Export Citation
Save