• Ancell, B., , and Hakim G. J. , 2007: Comparing adjoint and ensemble sensitivity analysis with applications to observation targeting. Mon. Wea. Rev., 135, 41174134.

    • Search Google Scholar
    • Export Citation
  • Baxter, M. A., 2011: Impacts of very small initial condition errors on mesoscale aspects of two cyclones. Electron. J. Oper. Meteor., 12, 2011EJ01.

    • Search Google Scholar
    • Export Citation
  • Bei, N., , and Zhang F. , 2007: Impacts of initial error scale and amplitude on the mesoscale predictability of heavy precipitation along the mei-yu front of China. Quart. J. Roy. Meteor. Soc., 133, 8399.

    • Search Google Scholar
    • Export Citation
  • Bishop, C. H., , Etherton B. J. , , and Majumdar S. J. , 2001: Adaptive sampling with the ensemble transform Kalman filter. Part I: Theoretical aspects. Mon. Wea. Rev., 129, 420436.

    • Search Google Scholar
    • Export Citation
  • Black, T. L., 1994: The new NMC mesoscale Eta Model: Description and forecast examples. Wea. Forecasting, 9, 265278.

  • Brennan, M. J., , and Lackmann G. M. , 2005: The influence of incipient latent heat release on the precipitation distribution of the 24–25 January 2000 U.S. East Coast cyclone. Mon. Wea. Rev., 133, 19131937.

    • Search Google Scholar
    • Export Citation
  • Bright, D. R., , and Nutter P. A. , 2004: On the challenges of identifying the “best” ensemble member in operational forecasting. Preprints, 16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., J11.3. [Available online at http://ams.confex.com/ams/pdfpapers/69092.pdf.]

  • Caplan, P., , Derber J. , , Gemmill W. , , Hong S.-Y. , , Pan H.-L. , , and Parrish D. , 1997: Changes to the 1995 NCEP operational medium-range forecast model. Wea. Forecasting, 12, 581594.

    • Search Google Scholar
    • Export Citation
  • Chang, E. K. M., 2005: The impact of wave packets propagating across Asia on Pacific cyclone development. Mon. Wea. Rev., 133, 19982015.

    • Search Google Scholar
    • Export Citation
  • Charles, M. E., , and Colle B. A. , 2009: Verification of extratropical cyclones within the NCEP operational models. Part II: The short-range ensemble forecast system. Wea. Forecasting, 24, 11911214.

    • Search Google Scholar
    • Export Citation
  • Clark, A. J., , Gallus W. A. Jr., , Xue M. , , and Kong F. , 2009: A comparison of precipitation forecast skill between small convection-allowing and large convection-parameterizing ensembles. Wea. Forecasting, 24, 11211140.

    • Search Google Scholar
    • Export Citation
  • Clark, A. J., , Gallus W. A. Jr., , and Weisman M. L. , 2010a: Neighborhood-based verification of precipitation forecasts from convection-allowing NCAR WRF model simulations and the operational NAM. Wea. Forecasting, 25, 14951509.

    • Search Google Scholar
    • Export Citation
  • Clark, A. J., , Gallus W. A. Jr., , Xue M. , , and Kong F. , 2010b: Growth of spread in convection-allowing and convection-parameterizing ensembles. Wea. Forecasting, 25, 594612.

    • Search Google Scholar
    • Export Citation
  • Colle, B. A., , and Charles M. E. , 2011: Spatial distribution and evolution of extratropical cyclone errors over North American and adjacent oceans in the NCEP Global Forecast System model. Wea. Forecasting, 26, 129149.

    • Search Google Scholar
    • Export Citation
  • Davis, C. A., , Stoelinga M. T. , , and Kuo Y.-H. , 1993: The integrated effect of condensation in numerical simulations of extratropical cyclogenesis. Mon. Wea. Rev., 121, 23092330.

    • Search Google Scholar
    • Export Citation
  • Davis, C. A., , Brown B. , , and Bullock R. , 2006: Object-based verification of precipitation forecasts: Part I: Methodology and application to mesoscale rain areas. Mon. Wea. Rev., 134, 17721784.

    • Search Google Scholar
    • Export Citation
  • Dickinson, M. J., , Bosart L. F. , , Bracken W. E. , , Hakim G. J. , , Schultz D. M. , , Bedrick M. A. , , and Tyle K. R. , 1997: The March 1993 Superstorm cyclogenesis: Incipient phase synoptic- and convective-scale flow interaction and model performance. Mon. Wea. Rev., 125, 30413072.

    • Search Google Scholar
    • Export Citation
  • Du, J., , DiMego G. , , Tracton M. S. , , and Zhou B. , 2003: NCEP short-range ensemble forecasting (SREF) system: Multi-IC, multi-model and multi-physics approach. Research Activities in Atmospheric and Oceanic Modeling, J. Cote, Ed., CAS/JSC Working Group Numerical Experimentation Rep. 23, WMO/TD 1161, 5.09–5.10.

  • Du, J., and Coauthors, 2004: The NOAA/NWS/NCEP short-range ensemble forecast (SREF) system: Evaluation of an initial condition vs multiple model physics ensemble approach. Extended Abstracts, 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., 21.3. [Available online at http://ams.confex.com/ams/84Annual/techprogram/paper_71107.htm.]

  • Du, J., , McQueen J. , , DiMego G. , , Toth Z. , , Jovic D. , , Zhou B. , , and Chuang H.-Y. , 2006: New dimension of NCEP short-range ensemble forecasting (SREF) system: Inclusion of WRF members. Preprints, WMO Expert Team Meeting on Ensemble Prediction System, Exeter, United Kingdom, World Meteorological Organization, 5 pp. [Available online at http://www.emc.ncep.noaa.gov/mmb/SREF/WMO06_full.pdf.]

  • Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 30773107.

    • Search Google Scholar
    • Export Citation
  • Dudhia, J., 1993: A nonhydrostatic version of the Penn State–NCAR Mesoscale Model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121, 14931513.

    • Search Google Scholar
    • Export Citation
  • Eckel, F. A., , and Mass C. F. , 2005: Aspects of effective mesoscale, short-range ensemble forecasting. Wea. Forecasting, 20, 328350.

  • Evans, M., , and Jurewicz M. L. , 2009: Correlations between analyses and forecasts of banded heavy snow ingredients and observed snowfall. Wea. Forecasting, 24, 337350.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121, 764787.

  • Grimit, E. P., , and Mass C. F. , 2002: Initial results of a mesoscale short-range ensemble forecasting system over the Pacific Northwest. Wea. Forecasting, 17, 192205.

    • Search Google Scholar
    • Export Citation
  • Hakim, G. J., 2005: Vertical structure of midlatitude analysis and forecast errors. Mon. Wea. Rev., 133, 567578.

  • Hakim, G. J., , and Torn R. D. , 2007: Ensemble synoptic analysis. Synoptic–Dynamic Meteorology and Weather Analysis and Forecasting: A Tribute to Fred Sanders, Meteor. Monogr., No. 55, Amer. Meteor. Soc., 147–161.

  • Hakim, G. J., , Keyser D. , , and Bosart L. F. , 1996: The Ohio Valley wave-merger cyclogenesis event of 25–26 January 1978. Part II: Diagnosis using quasigeostrophic potential vorticity inversion. Mon. Wea. Rev., 124, 21762205.

    • Search Google Scholar
    • Export Citation
  • Hawblitzel, D. P., , Zhang F. , , Meng Z. , , and Davis C. A. , 2007: Probabilistic evaluation of the dynamics and predictability of the mesoscale convective vortex of 10–13 June 2003. Mon. Wea. Rev., 135, 15441563.

    • Search Google Scholar
    • Export Citation
  • Hohenegger, C., , and Schär C. , 2007: Predictability and error growth dynamics in cloud-resolving models. J. Atmos. Sci., 64, 44674478.

    • Search Google Scholar
    • Export Citation
  • Hohenegger, C., , Luthi D. , , and Schär C. , 2006: Predictability mysteries in cloud-resolving models. Mon. Wea. Rev., 134, 20952107.

  • Hong, S.-Y., , Dudhia J. , , and Chen S.-H. , 2004: A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation. Mon. Wea. Rev., 132, 103120.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 2003: A nonhydrostatic model based on a new approach. Meteor. Atmos. Phys., 82, 271285.

  • Jones, M. S., , Colle B. A. , , and Tongue J. S. , 2007: Evaluation of a mesoscale short-range ensemble forecast system over the northeast United States. Wea. Forecasting, 22, 3655.

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

    • Search Google Scholar
    • Export Citation
  • Lorenz, E. N., 1963: Deterministic nonperiodic flow. J. Atmos. Sci., 20, 130141.

  • Lorenz, E. N., 1969: The predictability of a flow which possesses many scales of motion. Tellus, 21, 289307.

  • Lorenz, E. N., 1996: Predictability—A problem partly solved. Seminar on Predictability, Vol. 1, ECMWF, 1–19.

  • Mahoney, K. M., , and Lackmann G. M. , 2007: The effect of upstream convection on downstream precipitation. Wea. Forecasting, 22, 255277.

    • Search Google Scholar
    • Export Citation
  • Morgan, M. C., , and Nielsen-Gammon J. W. , 1998: Using tropopause maps to diagnose midlatitude weather systems. Mon. Wea. Rev., 126, 25552579.

    • Search Google Scholar
    • Export Citation
  • Nicosia, D. J., , and Grumm R. H. , 1999: Mesoscale band formation in three major northeastern United States snowstorms. Wea. Forecasting, 14, 346368.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Bosart L. F. , , Keyser D. , , and Waldstreicher J. S. , 2004: An observational study of cold season–banded precipitation in northeast U.S. cyclones. Wea. Forecasting, 19, 9931010.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Waldstreicher J. S. , , Keyser D. , , and Bosart L. F. , 2006: A forecast strategy for anticipating cold season mesoscale band formation within eastern U.S. cyclones. Wea. Forecasting, 21, 323.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Bright D. R. , , and Brennan M. J. , 2008a: Operational forecaster uncertainty needs and future roles. Wea. Forecasting, 23, 10691084.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Colle B. A. , , and Yuter S. E. , 2008b: High-resolution observations and model simulations of the life cycle of an intense mesoscale snowband over the northeastern United States. Mon. Wea. Rev., 136, 14331456.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Colle B. A. , , and McTaggart-Cowan R. , 2009: The role of moist processes in the formation and evolution of mesoscale snowbands within the comma head of northeast United States cyclones. Mon. Wea. Rev., 137, 26622686.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Colle B. A. , , and Aiyyer A. , 2010: Evolution of mesoscale precipitation band environments within the comma head of northeast U.S. cyclones. Mon. Wea. Rev., 138, 23542374.

    • Search Google Scholar
    • Export Citation
  • Ralph, M. F., and Coauthors, 2005: Improving short-term (0–48 h) cool-season quantitative precipitation forecasting: Recommendations from a USWRP workshop. Bull. Amer. Meteor. Soc., 86, 16191632.

    • Search Google Scholar
    • Export Citation
  • Reisner, J., , Rasmussen R. M. , , and Bruintjes R. T. , 1998: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124, 10711107.

    • Search Google Scholar
    • Export Citation
  • Schlemmer, L., , Martius O. , , and Sprenger M. , 2010: Disentangling the forcing mechanisms of a heavy precipitation event along the Alpine south side using potential vorticity inversion. Mon. Wea. Rev., 138, 23362353.

    • Search Google Scholar
    • Export Citation
  • Schumacher, R. S., 2011: Ensemble-based analysis of factors leading to the development of a multi day warm-season heavy rain event. Mon. Wea. Rev., 139, 30163035.

    • Search Google Scholar
    • Export Citation
  • Schumacher, R. S., , Schultz D. M. , , and Knox J. A. , 2010: Convective snowbands downstream of the Rocky Mountains in an environment with conditional, dry symmetric, and inertial instabilities. Mon. Wea. Rev., 138, 44164438.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., , Klemp J. B. , , Dudhia J. , , Gill D. O. , , Barker D. M. , , Wang W. , and Powers J. G. , 2005: A description of the Advanced Research WRF version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp.

  • Stensrud, D. J., , Bao J. , , and Warner T. T. , 2000: Using initial condition and model physics perturbations in short-range ensemble simulations of mesoscale convective systems. Mon. Wea. Rev., 128, 20772107.

    • Search Google Scholar
    • Export Citation
  • Suarez, A., , Reeves H. D. , , Wheatley D. , , and Coniglio M. , 2012: Comparison of ensemble Kalman filter–based forecasts to traditional ensemble and deterministic forecasts for a case study of banded snow. Wea. Forecasting, 27, 85105.

    • Search Google Scholar
    • Export Citation
  • Szunyogh, I., , Toth Z. , , Zimin A. V. , , Majumdar S. J. , , and Persson A. , 2002: Propagation of the effect of targeted observations: The 2000 Winter Storm Reconnaissance Program. Mon. Wea. Rev., 130, 11441165.

    • Search Google Scholar
    • Export Citation
  • Tan, Z.-M., , Zhang F. , , Rotunno R. , , and Snyder C. , 2004: Mesoscale predictability of moist baroclinic waves: Experiments with parameterized convection. J. Atmos. Sci., 61, 17941804.

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

    • Search Google Scholar
    • Export Citation
  • Toth, Z., , and Kalnay E. , 1997: Ensemble forecasting at NCEP and the breeding method. Mon. Wea. Rev., 125, 32973319.

  • Zhang, F., 2005: Dynamics and structure of mesoscale error covariance of a winter cyclone estimated through short-range ensemble forecasts. Mon. Wea. Rev., 133, 28762893.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , Synder C. , , and Rotunno R. , 2003: Effects of moist convection on mesoscale predictability. J. Atmos. Sci., 60, 11731185.

  • Zhang, F., , Bei N. , , Rotunno R. , , Snyder C. , , and Epifanio C. C. , 2007: Mesoscale predictability of moist baroclinic waves: Convection-permitting experiments and multistage error growth dynamics. J. Atmos. Sci., 64, 35793594.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , Weng Y. , , Kuo Y.-H. , , Whitaker J. S. , , and Xie B. , 2010: Predicting Typhoon Morakot’s catastrophic rainfall with a convection-permitting mesoscale ensemble system. Wea. Forecasting, 25, 18161825.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) Time series of the observed MSLP (black), 12–30-h forecast ensemble mean MSLP (white), and individual ensemble member forecast (see legend) MSLP for the 25 Dec 2002 case. MM5 members are solid and WRF members are dashed. (b) As in (a), but for 12–36 h forecasts and observations for the 12 Feb 2006 case. (c) As in (a), but for 12–30-h forecasts and observations for the 14 Feb 2007 case. (d)–(f) Corresponding ensemble member (colored by forecast projection) and observed (black) surface low positions for each case.

  • View in gallery

    (a)–(c) Manually analyzed cooperative observer storm total liquid equivalent precipitation (mm, shaded every 5 mm according to scale) for the (a) 25 Dec 2002, (b) 12 Feb 2006, and (c) 14 Feb 2007 cases. (d)–(f) Corresponding ensemble mean forecast storm total precipitation. (g)–(i) Corresponding ensemble probability (%) of exceeding 50, 25, and 50 mm for the 25 Dec 2002, 12 Feb 2006, and 14 Feb 2007 cases, respectively.

  • View in gallery

    Summary of ensemble band predictions for the (left) 25 Dec 2002, (middle) 12 Feb 2006, and (right) 14 Feb 2007 cases. Ensemble predictions are shown at the time of (a)–(c) and (d)–(f) observed band formation, (g)–(i) band maturity, and (j)–(l) band dissipation.

  • View in gallery

    (a) Time series of the percentage of members forecasting a band during the 25 Dec 2002 case. The timing and duration of the observed banded event is shown by a black bar at the top of the graph. (b) Ensemble distribution of forecasted band duration (h). The observed duration is shown as an asterisk. (c),(d) As in (a),(b), but for the 12 Feb 2006 case. (e),(f) As in (a),(b), but for the 14 Feb 2007 case.

  • View in gallery

    Evolution of the 475–250-hPa PV mean (contoured every 1 PVU) and spread (shaded, PVU) of the six-member MM5 initial condition subensemble over the 36-km domain for the (a)–(c) 25 Dec 2002, (d)–(f) 12 Feb 2006, and (g)–(i) 14 Feb 2007 cases. Select upper-level PV anomalies are labeled.

  • View in gallery

    Evolution of the 800–600-hPa PV mean (contoured every 0.5 PVU, starting at 1 PVU) and spread (shaded, PVU) of the six-member MM5 initial condition subensemble over the 36-km domain for the (a)–(c) 25 Dec 2002, (d)–(f) 12 Feb 2006, and (g)–(i) 14 Feb 2007 cases. Select upper-level PV anomalies are labeled.

  • View in gallery

    Observed (black solid) and forecast band locations from representative northwest (gray dashed) and southeast (gray solid) outlier members at select times during the (a) 25 Dec 2002, (b) 12 Feb 2006, and (c) 14 Feb 2007 cases. (d)–(f) Corresponding comparison of observed (black solid) and predicted band duration from northwest (gray dashed) and southeast (gray solid) members.

  • View in gallery

    Comparison of the MSLP and simulated reflectivity from the (a)–(c) SREF-N1-MM5 and (d)–(f) GFS-MM5 to (g)–(i) the observed radar mosaic and NCEP surface analysis low position for the 25 Dec 2002 case.

  • View in gallery

    Comparison of the 475–250-hPa layered-average PV (shaded according to scale, PVU) evolution over the 36-km domain for the (a)–(c) northwest (SREF-N1-MM5) and (d)–(f) southeast (GFS-MM5) members at 0000 UTC 25 Dec 2002 (initialization), 1200 UTC 25 Dec 2002 (12-h forecast), and 1800 UTC 25 Dec 2002 (18-h forecast). (g)–(i) Corresponding 475–250-hPa layered-average PV difference (SREF-N1-MM5 minus GFS-MM5).

  • View in gallery

    Comparison of the MSLP and simulated reflectivity from the (a)–(c) GFS-MM5 and (d)–(f) SREF-N1-MM5 to (g)–(i) the observed radar mosaic and NCEP surface analysis low position for the 12 Feb 2006 case.

  • View in gallery

    Comparison of the 475–250-hPa layered-average PV (shaded according to scale, PVU) evolution over the 36-km domain for the (a)–(c) northwest (GFS-MM5) and (d)–(f) southeast (SREF-N1-MM5) members at 1200 UTC 11 Feb 2006 (initialization), 0000 UTC 12 Feb 2006 (12-h forecast), and 1200 UTC 12 Feb 2006 (24-h forecast). (g)–(i) Corresponding 475–250-hPa layered-average PV difference (GFS-MM5 minus SREF-N1-MM5).

  • View in gallery

    Comparison of the MSLP and simulated reflectivity from the (a)–(c) GFS-MM5 and (d)–(f) NAM-MM5 to (g)–(i) the observed radar mosaic and NCEP surface analysis low position for the 14 Feb 2007 case.

  • View in gallery

    Comparison of the 475–250-hPa layered-average PV (shaded according to scale, PVU) evolution over the 36-km domain for the (a)–(c) northwest (GFS-MM5) and (d)–(f) southeast (NAM-MM5) members at 0000 UTC 14 Feb 2007 (initialization), 1200 UTC 14 Feb 2007 (12-h forecast), and 1800 UTC 14 Feb 2007 (18-h forecast). (g)–(i) Corresponding 475–250-hPa layered-average PV difference (GFS-MM5 minus NAM-MM5).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 31 31 6
PDF Downloads 25 25 5

Diagnosing Snowband Predictability Using a Multimodel Ensemble System

View More View Less
  • 1 NOAA/NWS/NCEP/Hydrometeorological Prediction Center, Camp Springs, Maryland
  • | 2 School of Marine and Atmospheric Sciences, Stony Brook University, Stony Brook, New York
© Get Permissions
Full access

Abstract

The forecast uncertainty of mesoscale snowband formation and evolution is compared using predictions from a 16-member multimodel ensemble at 12-km grid spacing for the 25 December 2002, 12 February 2006, and 14 February 2007 northeast U.S. snowstorms. Using these predictions, the case-to-case variability in the predictability of band formation and evolution is demonstrated. Feature-based uncertainty information is also presented as an example of what may be operationally feasible from postprocessing information from future short-range ensemble forecast systems. Additionally, the initial condition sensitivity of band location in each case is explored by contrasting the forecast evolutions of initial condition members with large differences in snowband positions. Considerable uncertainty in the occurrence, and especially timing and location, of band formation and subsequent evolution was found, even at forecast projections <24 h. The ensemble provided quantitative mesoscale band uncertainty information, and differentiated between high-predictability (14 February 2007) and low-predictability (12 February 2006) cases. Among the three cases, large (small) initial differences in the upper-level PV distribution and surface mean sea level pressure of the incipient cyclone were associated with large (small) differences in forecast snowband locations, suggesting that case-to-case differences in predictability may be related to the quality of the initial conditions. The complexity of the initial flow may also be a discriminator. Error growth was evident in each case, consistent with previous mesoscale predictability research, but predictability differences were not correlated to the degree of convection. Discussion of these results and future extensions of the work are presented.

Corresponding author address: David R. Novak, NOAA/NWS/NCEP/Hydrometeorological Prediction Center, Rm. 410, 5200 Auth Rd., Camp Springs, MD 20746. E-mail: david.novak@noaa.gov

Abstract

The forecast uncertainty of mesoscale snowband formation and evolution is compared using predictions from a 16-member multimodel ensemble at 12-km grid spacing for the 25 December 2002, 12 February 2006, and 14 February 2007 northeast U.S. snowstorms. Using these predictions, the case-to-case variability in the predictability of band formation and evolution is demonstrated. Feature-based uncertainty information is also presented as an example of what may be operationally feasible from postprocessing information from future short-range ensemble forecast systems. Additionally, the initial condition sensitivity of band location in each case is explored by contrasting the forecast evolutions of initial condition members with large differences in snowband positions. Considerable uncertainty in the occurrence, and especially timing and location, of band formation and subsequent evolution was found, even at forecast projections <24 h. The ensemble provided quantitative mesoscale band uncertainty information, and differentiated between high-predictability (14 February 2007) and low-predictability (12 February 2006) cases. Among the three cases, large (small) initial differences in the upper-level PV distribution and surface mean sea level pressure of the incipient cyclone were associated with large (small) differences in forecast snowband locations, suggesting that case-to-case differences in predictability may be related to the quality of the initial conditions. The complexity of the initial flow may also be a discriminator. Error growth was evident in each case, consistent with previous mesoscale predictability research, but predictability differences were not correlated to the degree of convection. Discussion of these results and future extensions of the work are presented.

Corresponding author address: David R. Novak, NOAA/NWS/NCEP/Hydrometeorological Prediction Center, Rm. 410, 5200 Auth Rd., Camp Springs, MD 20746. E-mail: david.novak@noaa.gov

1. Introduction

A major challenge of cool-season quantitative precipitation forecasting (QPF) is to determine the spatial and temporal variability of precipitation within extratropical cyclones (Ralph et al. 2005). Variability in the location and intensity of cool-season precipitation is often determined by the development and evolution of mesoscale precipitation bands (Ralph et al. 2005). Thus, improving mesoscale band forecasts will help improve cool-season QPF.

Mesoscale precipitation bands are frequently observed in the comma-head portion of extratropical cyclones in the northeast United States (Nicosia and Grumm 1999; Novak et al. 2004; Novak et al. 2010). Prediction techniques have focused on assessing the ingredients of frontogenesis, weak moist symmetric stability, and sufficient moisture. However, Evans and Jurewicz (2009) found that correlations between heavy snowfall and a single model’s forecast values of frontogenetical forcing, weak moist symmetric stability, saturation, and favorable thermodynamic environments (ingredients associated with mesoscale snowbands) decrease substantially as forecast lead time increases beyond 12 h. They hypothesized that model forecast positioning and timing errors are responsible for the lower correlations associated with longer-lead forecasts. In an analysis of three snowband cases, Novak et al. (2009) found that the frontogenetical forcing for the band was highly sensitive to upstream potential vorticity (PV) modification from precipitation/convection. The frontogenetical evolution associated with the band was also sensitive to the precipitation evolution and resulting PV modification both locally and well east of the banding region. This sensitivity is consistent with error growth associated with moist convective processes, as established by previous mesoscale predictability research (e.g., Zhang et al. 2003; Tan et al. 2004; Brennan and Lackmann 2005; Zhang et al. 2007; Hawblitzel et al. 2007; Bei and Zhang 2007; Hohenegger et al. 2006).

In recognition of the predictability limits inherent to mesoscale bands, Novak et al. (2006) proposed a mesoscale band forecast strategy that assesses the band ingredients in a time- and scale-dependent approach utilizing ensembles, deterministic models, and observations. However, they acknowledge that a more complete band forecast strategy would emphasize a probabilistic approach that quantifies the case-to-case variability in band predictability.

Short-range ensemble forecast (SREF) systems can provide objective information regarding the predictability of features by accounting for initial conditions and model uncertainty (e.g., Du et al. 2003, 2004, 2006; Grimit and Mass 2002; Eckel and Mass 2005; Jones et al. 2007; Suarez et al. 2012). For example, recent work has examined extratropical cyclone track performance (e.g., Charles and Colle 2009). However, current operational SREFs have horizontal grid spacing on the order of 20–40 km (Du et al. 2006), which is too coarse to adequately resolve mesoscale snowbands. Thus, explicit mesoscale band uncertainty information is not available from operational SREFs, limiting a probabilistic approach to the band prediction problem.

Given mesoscale models such as the fifth-generation Pennsylvania State University–National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5; Dudhia 1993) and Weather Research and Forecasting model (WRF) (Skamarock et al. 2005) are able to predict the primary aspects of mesoscale precipitation bands at ~10-km grid spacing (Novak et al. 2009; Schumacher et al. 2010), Novak et al. (2006) and Novak et al. (2008b) recommended exploring using the explicit band predictions from members of an ensemble to assess the predictability of band formation and evolution. Such information may help advance band prediction toward a probabilistic format. Following this recommendation, the primary goals of this work are to

  • demonstrate the ability of a 12-km grid-spacing multimodel ensemble using different initial conditions and physics to quantify the case-to-case variability in band predictability; such an ensemble system on a continental domain will be operationally available from the National Centers for Environmental Prediction (NCEP) soon, and will be available prior to more computationally expensive convection-allowing ensembles;
  • illustrate the type of feature-based (explicit mesoscale snowband) uncertainty information possible from such an ensemble system; and
  • explore the sensitivity of band location to initial conditions.
To achieve these goals, ensemble predictions for three banded snowstorms over the northeast United States are conducted and analyzed.

Section 2 describes the cases, ensemble system, and analysis method. The ensemble predictions of band evolution for each storm are summarized in section 3, and forecast location uncertainties are investigated in section 4. The implications and future directions of the work are discussed in section 5.

2. Data and methods

a. Snowband cases

Heavy snowfall occurred in the 25 December 2002, 12 February 2006, and 14 February 2007 cases, with maximum totals of ~100 cm (40 in) during the 25 December 2002 and 14 February 2007 cases in eastern New York, and a record setting 68.3 cm (26.9 in) in New York City during the 12 February 2006 case. Much of the snowfall in each case was associated with the formation and evolution of an intense mesoscale snowband, with sustained snowfall rates of 5–10 cm h−1 reported within each respective band.

Common cyclone features important to the development of snowbands northwest of the surface cyclone (e.g., Novak et al., 2010) were present in each case, including the presence of small stability and the development of intense lower-tropospheric frontogenesis along an inverted trough at the 700-hPa level. Detailed observations and deterministic 4-km model results from these cases are presented in Novak et al. (2008b) and Novak et al. (2009). The three cases were also included in the climatology and potential vorticity (PV) classification of Novak et al. (2010). The 25 December 2002 and 12 February 2006 cases were classified by Novak et al. (2010) as “PV hook,” which was the most common type of banded cyclone events in the northeastern United States. In PV hook events the band forms on the northwest side of a hook-shaped upper-level PV anomaly in a region of cyclonic PV advection. The 14 February 2007 case was classified as “lagging upper-trough.” In this type of system, band formation occurs in the northwest quadrant of the surface cyclone, but more than 400 km to the east of the 400-hPa 2-PVU contour (1 PVU = 10−6 K m2 kg−1 s−1), and within 300 km of a saturated 700-hPa PV maximum (i.e., a diabatic PV anomaly), consistent with persistent convection and diabatic heating (Novak et al. 2010). The upper-level PV perspective of the cases is discussed in section 4.

b. Ensemble design and evaluation

A 16-member multimodel ensemble using different initial conditions and physics was developed to explore the predictability of the three snowband events (Table 1). Member names appearing in the first column of Table 1 will be used throughout the paper. The ensemble system is composed of two models: MM5v3.7 and WRFv2.2. Initial and boundary conditions were used from the operational NCEP North American Mesoscale (NAM) model (Janjić 2003), the Global Forecast Systems (GFS) model (Caplan et al. 1997), and select NCEP SREF system perturbation members (N1, N2, P1, and P2; Du et al. 2003, 2004). Each initial breeding pair was perturbed positively (P) and negatively (N) using the NCEP breeding technique described by Toth and Kalnay (1997). These particular perturbed SREF members were from the Eta Model (Black 1994) subset and were chosen since they were available for all three cases.

Table 1.

Ensemble system configuration.

Table 1.

Although most members of the 16-member multimodel ensemble used here employed simple ice microphysics (Dudhia 1989; Hong et al. 2004) and Grell convective parameterization (Grell 1993), the microphysics of GFS-initialized members were varied between simple ice and Reisner2 (Reisner et al. 1998) in the MM5 and simple ice and Thompson (Thompson et al. 2004) in the WRF. The convective parameterization of SREF-P2-initialized members was varied between the Grell and Kain–Fritsch convective schemes (Kain and Fritsch 1990) (Table 1). The ensemble was integrated over an outer 36-km domain covering the eastern two-thirds of the United States and adjacent coastal waters, and a 12-km (one way) nested grid over the northeast United States (cf. Fig. 1a of Jones et al. 2007). The 12-km ensemble output is used in section 3 to analyze band characteristics, and the 36-km output is used in section 4 to explore the sensitivity of band location to initial conditions. The ensemble was initialized 19, 21, and 15 h prior to observed band formation for the 25 December 2002, 12 February 2006, and 14 February 2007 cases, respectively.

To assess band predictability from the ensemble, each ensemble member’s hourly output was manually inspected. Following Novak et al. (2004), both the observed and simulated bands in the 12-km domain were defined as a reflectivity feature >250 km in length, 20–100 km in width, and with intensities of >30 dBZ maintained for at least 2 h. Band formation was noted at the time when these conditions were first met (after determining that the feature persisted for at least 2 h). The intensity threshold was reduced to 26 dBZ for the 12 February 2006 banded event to document predictions of the occurrence and evolution of weaker bands. The timing and location of the simulated and observed bands were compared for each case. The subjectivity of the band identification is a recognized limitation of the work, and is discussed further in section 5.

The probabilistic skill of the ensemble system cannot be evaluated, since only three cases are analyzed. For the demonstration purposes of this paper, large ensemble spread is assumed to represent the low predictability of the quantity of interest, while small spread is assumed to represent high predictability. “Predictability” in this work refers to practical predictability (Lorenz 1996)—the ability to predict based on current procedures—which incorporates limitations of current ensemble and observational systems. This is opposed to the ideal intrinsic predictability (Lorenz 1969), which is the extent to which prediction is possible if an optimum procedure is used.

3. Ensemble predictions

a. Surface low

The three cyclones exhibited rapid cyclogenesis along the East Coast (Fig. 1), with the 25 December 2002 event exhibiting the greatest deepening [28 hPa (24 h)−1], followed by the 14 February 2006 case [26 hPa (24 h)−1] and the 12 February 2006 case [23 hPa (24 h)−1]. The MSLP and low track of the 14 February 2007 storm were well forecast by the ensemble mean, with the mean MSLP within ±4 hPa of the observed during the 12–30-h forecast evolution (Fig. 1c), and the 24-h forecast low positions within 300 km of the observed (Fig. 1f). In contrast, the MSLP for the 12 February 2006 storm was poorly forecast by most ensemble members. The ensemble mean exhibited positive errors of as much as 10 hPa, and only one member (GFS-Thom-WRF) exhibited an MSLP evolution as deep as the observed evolution during the 0000–1500 UTC period of devel-opment (Fig. 1b). Several members exhibited surface lows that were near the observed coastal track; however, three members were over 500 km farther off the coast (Fig. 1e). The ensemble mean MSLP for the 25 December 2002 storm exhibited more modest positive errors of 4–6 hPa during the forecast (Fig. 1a), and a low track that was comparable in skill to the 14 February 2007 case (Fig. 1d).

Fig. 1.
Fig. 1.

(a) Time series of the observed MSLP (black), 12–30-h forecast ensemble mean MSLP (white), and individual ensemble member forecast (see legend) MSLP for the 25 Dec 2002 case. MM5 members are solid and WRF members are dashed. (b) As in (a), but for 12–36 h forecasts and observations for the 12 Feb 2006 case. (c) As in (a), but for 12–30-h forecasts and observations for the 14 Feb 2007 case. (d)–(f) Corresponding ensemble member (colored by forecast projection) and observed (black) surface low positions for each case.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

b. Liquid equivalent precipitation

Observed liquid equivalent precipitation exceeded 40, 70, and 75 mm in the 12 February 2006, 25 December 2002, and 14 February 2007 cases, respectively (Figs. 2a–c). The simulated precipitation maximum in the 25 December 2002 and 14 February 2006 cases was elongated (Figs. 2d,f), suggesting several members predicted bands in similar locations. However, the 12 February 2006 case did not exhibit a well-defined elongated precipitation maximum. Maximum precipitation amounts from the ensemble mean were only ~50% of the observed in the 25 December 2002 and 12 February 2006 cases. In contrast, the maximum precipitation amounts from the ensemble mean were ~85% of the observed in the 14 February 2007 case, which is impressive given the expected dampening of the precipitation maxima by averaging.

Fig. 2.
Fig. 2.

(a)–(c) Manually analyzed cooperative observer storm total liquid equivalent precipitation (mm, shaded every 5 mm according to scale) for the (a) 25 Dec 2002, (b) 12 Feb 2006, and (c) 14 Feb 2007 cases. (d)–(f) Corresponding ensemble mean forecast storm total precipitation. (g)–(i) Corresponding ensemble probability (%) of exceeding 50, 25, and 50 mm for the 25 Dec 2002, 12 Feb 2006, and 14 Feb 2007 cases, respectively.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

The probability of exceeding 50 mm in the 25 December and 14 February 2007 cases, and 25 mm in the 12 February 2006 case was calculated. These thresholds were chosen at roughly two-thirds of the observed precipitation maximum. At locations along the observed band, the ensemble predicted a 10%–20% probability of exceeding 50 mm in the 25 December 2002 case (Fig. 2g), a 10%–20% probability of exceeding 25 mm in the 12 February 2006 case (Fig. 2h), and over a 90% probability of exceeding 50 mm in the 14 February 2007 case (Fig. 2i). These results highlight the varied uncertainty in the precipitation amounts for these cases, with the 14 February 2007 case exhibiting the highest probability for heavy precipitation.

c. Mesoscale snowband characteristics

Analysis of the simulated reflectivity from the individual ensemble members showed that all members exhibited band formation during the 14 February 2007 case, while >80% of the members predicted bands at some time during the 25 December 2002 and 12 February 2006 cases (Table 2). The locations of band formation irrespective of timing (band formation envelope) are shown in Figs. 3a–c. In each case a favored SW–NE-oriented envelope of band locations was predicted by most of the ensemble members. The narrow envelope (~100 km wide) of band predictions in the 14 February 2007 case, which were clustered near the observed position (Fig. 3c), illustrates the relatively high predictability of band location in this case. In contrast, the particularly broad envelope (~600 km wide) of band predictions in the 12 February 2006 case (Fig. 3b) illustrates the relatively low predictability of band location in this case. Consistent with the band location spread, the surface low positions at the time of band formation also show a relatively large spread in the 12 February 2006 case, while the 25 December 2002 and 14 February 2007 cases exhibit relatively small spread (Figs. 3a–c). Note that there was small band location spread in the 14 February 2007 case, despite having a large MSLP spread (Fig. 1c), while there was large band location spread in the 12 February 2006 case, despite having a relatively small MSLP spread. This result suggests that the predictability of cyclone depth is not a good proxy for the predictability of band location.

Table 2.

Ensemble member band predictions. Members that forecasted a band are denoted by an X for each case.

Table 2.
Fig. 3.
Fig. 3.

Summary of ensemble band predictions for the (left) 25 Dec 2002, (middle) 12 Feb 2006, and (right) 14 Feb 2007 cases. Ensemble predictions are shown at the time of (a)–(c) and (d)–(f) observed band formation, (g)–(i) band maturity, and (j)–(l) band dissipation.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

The differences in predictability among the cases are also evident in the timing of the stages of band evolution. For example, several members predicted a band at the time of observed band formation in the 25 December 2002 (Fig. 3d) and 14 February 2007 cases (Fig. 3f), while none of the members predicted a band at the time of actual band formation in the 12 February 2006 case (Fig. 3e). The observed band at the time of band maturity fell within an envelope of 10 members in the 25 December 2002 case (Fig. 3g) and 9 members in the 14 February 2007 (Fig. 3i) case. However, the observed band was ~100 km southeast of the edge of the only four members predicting bands at the time of observed band maturity in the 12 February 2006 case (Fig. 3h). The small number of members predicting bands at this time was a symptom of a large timing error (shown below) and, once again, illustrates the relatively low predictability of the 12 February 2006 case.

The varied uncertainty in the timing and duration of the predicted bands can be quantified. In the 25 December 2002 case, 75% of the members predicted a band at the time of observed band formation, while none of the members predicted a band after the observed band had dissipated (Fig. 4a). The ensemble correctly suggested the duration of the band, with the mode of the duration distribution (6 h) coinciding with the observed duration (6 h) (Fig. 4b). In contrast, band formation was favored several hours later than observed during the 12 February 2006 case, with none of the members forecasting a band at the time of actual band formation (Fig. 4c), and 75% of the members forecasting a band after the observed band had dissipated. The mode of the duration distribution (4 h) was less than the observed duration (9 h), but was within the ensemble duration envelope (Fig. 4d). As discussed above, the 14 February 2007 case exhibited high certainty in band occurrence and small spread in the location of banding. The mode of the duration distribution (8 h) also coincided with the observed band duration (8 h) (Fig. 4f). However, only 45% of the members predicted a band at the time of observed band formation, while over 70% of the members predicted a band after the observed band had dissipated (Fig. 4e). These results highlight the challenge of predicting not only band occurrence, but band location, timing, and duration.

Fig. 4.
Fig. 4.

(a) Time series of the percentage of members forecasting a band during the 25 Dec 2002 case. The timing and duration of the observed banded event is shown by a black bar at the top of the graph. (b) Ensemble distribution of forecasted band duration (h). The observed duration is shown as an asterisk. (c),(d) As in (a),(b), but for the 12 Feb 2006 case. (e),(f) As in (a),(b), but for the 14 Feb 2007 case.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

4. Sensitivity of band location to initial conditions

a. Method

A conclusive analysis of the sources of the forecast uncertainty in each case is beyond the scope of this work. Given that most members produced bands, our focus is placed on the band location uncertainty. In recognition that initial condition uncertainty fundamentally limits predictability (e.g., Lorenz 1963), the ensemble evaluation is focused on the sensitivity of band location to the initial conditions (ICs). The focus on IC uncertainty necessitated using the six-member MM5 subensemble of IC perturbed members.

A meteorological context for how the ICs affected the band location in each case is explored by

  1. examining the forecast evolution of the IC ensemble means and spreads and
  2. contrasting the forecast evolutions of members with opposing outlier solutions. This approach allows an overall view of the synoptic situation and associated uncertainty, supported by detailed comparison of outlier member evolutions within the context of the IC subensemble mean and spread.

1) Ensemble mean and spread

Upper- (475–250 hPa) and lower- (800–600 hPa) level layer-average PV mean and spread were calculated. The upper-level PV can be used to diagnose upper-level troughs, ridges, and jets (Morgan and Nielsen-Gammon 1998), and is used to compare the evolution of the cases. Such comparison is consistent with Novak et al. (2010), who used the upper-level PV pattern to classify banded cyclones. The lower PV includes diabatically generated PV anomalies, which play an important role in the formation and evolution of mesoscale snowbands (e.g., Novak et al. 2009). Surface potential temperature anomalies and weaker anticyclonic PV anomalies are not examined. The PV analysis is conducted on the 36-km domain to provide a synoptic-scale perspective upstream of the band formation region.

2) Outlier members

The forecast evolution of a member predicting the band farthest northwest of the observed location is contrasted with the forecast evolution of a member predicting the band farthest southeast of the observed location among the IC subensemble. Differences in the upper-level PV evolutions between outlier members are shown. In addition, the MSLP and simulated reflectivity from each outlier member are also shown for reference.

The conclusive source of band location uncertainty cannot be obtained by comparing opposing outlier solutions. The only way to objectively determine how an IC error impacts a forecast metric is through formal sensitivity analysis (e.g., Hakim and Torn 2007; Ancell and Hakim 2007). We considered comparing means of clusters of members, as in Zhang et al. (2010). However, the relatively small number of members in the current work (16 members) precludes a meaningful cluster comparison approach. Others working with smaller ensemble datasets have compared “best” and “worst” members (e.g., Hawblitzel et al. 2007; Schumacher 2011), but the best and worst solutions are sensitive to element and forecast projection (Bright and Nutter 2004). Thus, what follows is a description of the evolution of two opposing outlier members that have bands in different locations using the previously described PV paradigm in Novak et al. (2009, 2010). Such an analysis can highlight the case-to-case variability in the meteorological uncertainties, and provide a meteorological context for how IC uncertainty contributes to the band location uncertainty in each case.

b. Cases

1) 25 December 2002

At the initialization time (19 h prior to observed band formation) the largest spread (~0.6 PVU) in the 475–250-hPa layer-averaged PV among the IC subensemble was found along the downstream edge of PV anomalies A and B in the southern plains (Fig. 5a). These anomalies corresponded to 500-hPa short-wave troughs (not shown). As the anomaly complex moved east, the spread on the downstream edge increased (Figs. 5b,c), suggesting east–west location differences in the anomalies. At lower levels (800–600 hPa), the initial PV spread is small (<0.4 PVU) (Fig. 6a). However, the spread grows along the coast during the 1200–1800 UTC 25 December 2002 period (Figs. 6b,c), as cyclogenesis occurs (e.g., Fig. 1a) and moist processes become more active. The enhanced spread in northern New Jersey at 1800 UTC may be particularly important, as Novak et al. (2009) showed through PV inversion that the lower-level PV anomaly in northern New Jersey was responsible for inducing a majority of the 700-hPa frontogenesis in the band formation region. However, the existence and location of the anomaly were largely dictated by the upper-level PV evolution (Novak et al. 2009, their Fig. 9), justifying the focus on the upper-level PV differences analyzed below.

Fig. 5.
Fig. 5.

Evolution of the 475–250-hPa PV mean (contoured every 1 PVU) and spread (shaded, PVU) of the six-member MM5 initial condition subensemble over the 36-km domain for the (a)–(c) 25 Dec 2002, (d)–(f) 12 Feb 2006, and (g)–(i) 14 Feb 2007 cases. Select upper-level PV anomalies are labeled.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

Fig. 6.
Fig. 6.

Evolution of the 800–600-hPa PV mean (contoured every 0.5 PVU, starting at 1 PVU) and spread (shaded, PVU) of the six-member MM5 initial condition subensemble over the 36-km domain for the (a)–(c) 25 Dec 2002, (d)–(f) 12 Feb 2006, and (g)–(i) 14 Feb 2007 cases. Select upper-level PV anomalies are labeled.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

The SREF-N1-MM5 was the farthest northwest member within the subensemble, whereas the GFS-MM5 was the farthest southeast member. Band predictions from these members were separated by ~250 km at the time of observed band maturity (2200 UTC 25 December; Fig. 7a). Band formation was 3 h earlier than observed in the SREF-N1-MM5 solution, while it was just 1 h early in the GFS-MM5 solution (Fig. 7d).

Fig. 7.
Fig. 7.

Observed (black solid) and forecast band locations from representative northwest (gray dashed) and southeast (gray solid) outlier members at select times during the (a) 25 Dec 2002, (b) 12 Feb 2006, and (c) 14 Feb 2007 cases. (d)–(f) Corresponding comparison of observed (black solid) and predicted band duration from northwest (gray dashed) and southeast (gray solid) members.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

Comparison of the MSLP initialization at 0000 UTC 25 December shows that the primary surface low over Kentucky in the SREF-N1-MM5 was initialized 2 hPa deeper than observed while the GFS-MM5 was initialized 1 hPa weaker than observed (Figs. 8a,d). Both members initialized the low ~75 km too far north (cf. Figs. 8a,d,g). During the subsequent 18-h forecast the coastal low developed and became dominant. Both members deepened the coastal low similarly, such that by 1800 UTC 25 December the SREF-N1-MM5 and GFS-MM5 members were within 1 hPa of each other. However, the low position is ~100 km farther west in the SREF-N1-MM5 than the GFS-MM5 (Figs. 8c,f), which is consistent with the farther westward band location in the SREF-N1-MM5. The observed low was 5–6 hPa deeper and approximately equidistant between the two solutions (Fig. 8i).

Fig. 8.
Fig. 8.

Comparison of the MSLP and simulated reflectivity from the (a)–(c) SREF-N1-MM5 and (d)–(f) GFS-MM5 to (g)–(i) the observed radar mosaic and NCEP surface analysis low position for the 25 Dec 2002 case.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

At 0000 UTC 25 December 2002, the initialized upper-level PV anomaly A over northern Texas is ~0.5 PVU stronger in the SREF-N1-MM5 analysis than the GFS-MM5 analysis (cf. Figs. 9a,d). Additionally, the SREF-N1-MM5 anomaly is ~50 km farther south than the GFS-MM5 analysis. Note that differences (SREF-N1-MM5 minus GFS-MM5) associated with anomaly A are the largest within the entire 36-km domain at this time (Fig. 9g). Farther north, over southeast Missouri, anomaly B in the SREF-N1-MM5 is ~0.25 PVU stronger than the GFS-MM5, possibly corresponding to a deeper surface low in the SREF-N1-MM5 (i.e., Figs. 8a,d). Both anomalies A and B are found within the area of maximum spread (Fig. 5a), suggesting these anomalies played a key role in the band location uncertainty.

Fig. 9.
Fig. 9.

Comparison of the 475–250-hPa layered-average PV (shaded according to scale, PVU) evolution over the 36-km domain for the (a)–(c) northwest (SREF-N1-MM5) and (d)–(f) southeast (GFS-MM5) members at 0000 UTC 25 Dec 2002 (initialization), 1200 UTC 25 Dec 2002 (12-h forecast), and 1800 UTC 25 Dec 2002 (18-h forecast). (g)–(i) Corresponding 475–250-hPa layered-average PV difference (SREF-N1-MM5 minus GFS-MM5).

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

During the next 12 h the positive PV difference associated with anomaly A over Texas translated across the Gulf coast as anomaly A moved east (Fig. 9h). However, by 1200 UTC 25 December 2002, anomaly A (now centered over western North Carolina) in the SREF-N1-MM5 solution became weaker and farther west than the GFS-MM5 anomaly (Figs. 9b,e).

By 1800 UTC 25 December 2002 (18-h forecast), anomaly A in the SREF-N1-MM5 solution was nearly 1 PVU weaker than the GFS-MM5 (Figs. 9c,f). Although both solutions exhibit a PV hook, the eastern flank of the hook in the SREF-N1-MM5 solution was ~100 km farther west than the GFS-MM5 anomaly (Figs. 9c,f), which is consistent with the farther west band solution. The difference in location of the PV hook resulted in a 2-PVU difference across New Jersey (Fig. 9i), and occurred near the area of maximum spread in the IC subensemble (Fig. 5c).

This case illustrates the importance of error growth on time scales as short as 12–18 h, as the initial conditions with the stronger and farther south PV anomaly (SREF-N1) become the forecast with the weaker and farther westward anomaly and associated farther westward snowband. Also, the differences in PV for the two members grow in size and magnitude with time, illustrating growth of the error (Figs. 9g–i).

2) 12 February 2006

Comparison of the initialized upper-level PV at 1200 UTC 11 February 2006 (21 h prior to the observed band formation) showed a notably more complex distribution than the initialization of the 25 December 2002 case. A broad, positively tilted PV tail extended from the northeast United States into the Front Range of the Rockies, and contained three embedded PV anomalies: one in Colorado (A), another in northwest Missouri (B), and an elongation across the lower Great Lakes (C) (Fig. 5d). In general, the spread was larger and covered a larger area than the 25 December 2002 case (Figs. 5a,d). As the anomaly complex moved east over the next 24 h, anomalies B and C merged. Spread increased on both the downstream and upstream edges of anomaly B, and also upstream of anomaly A (Fig. 5e), again suggesting east–west location differences. In contrast to upper levels, the initial spread of the lower-level PV was relatively small (<0.4 PVU), and comparable to the 25 December 2002 case (Figs. 6a,d). The spread in PV grew modestly during the period, and the mean lower-level PV was weak (Figs. 6d–f). Examination of the individual members showed that the lower-level PV anomalies were weak (~1.0 PVU), and did not amplify in earnest until after 1200 UTC 12 February, consistent with the ensemble timing error (e.g., Fig. 4c).

The GFS-MM5 was the farthest northwest member within the IC subensemble, while the SREF-N1-MM5 was the farthest southeast member. Band predictions from these members were separated by ~600 km at the time of observed band dissipation (1800 UTC 12 February) (Fig. 7b), when both members predicted bands. The GFS-MM5 predicted band formation only 1 h late, but it extended the band duration 5 h longer than observed (Fig. 7e). The SREF-N1-MM5 had particularly poor timing, with band formation occurring 9 h later than observed, and a band duration 3 h shorter than observed (Fig. 7e).

Comparison of the MSLP initialization at 1200 UTC 11 February shows that the SREF-N1-MM5 surface low was ~200 km farther northeast and 2 hPa weaker than the GFS-MM5 results (1011 versus 1009 hPa) (Figs. 10a,d). Compared to the observations, the GFS-MM5 had a better initialization of the surface cyclone, with a nearly identical surface low location and pressure (cf. Figs. 10a,d,g). The surface low in the SREF-N1-MM5 solution developed modestly while moving off the coast (Figs. 10d–f). This solution quickly evolved away from the observations (cf. Figs. 10d–f and 10g–i). In stark contrast, the surface low in the GFS-MM5 solution deepened rapidly and occluded near the coast by 1200 UTC 12 February (Figs. 10a–c). The GFS-MM5 cyclone evolution was similar to the observed cyclone evolution (cf. Figs. 10a–c and 10g–i)

Fig. 10.
Fig. 10.

Comparison of the MSLP and simulated reflectivity from the (a)–(c) GFS-MM5 and (d)–(f) SREF-N1-MM5 to (g)–(i) the observed radar mosaic and NCEP surface analysis low position for the 12 Feb 2006 case.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

.

Comparison of the corresponding initialized upper-level PV at 1200 UTC 11 February 2006 shows large differences in the locations of PV anomalies A, B, and C. The difference field (GFS-MM5 minus SREF-N1-MM5) shows that each of the anomalies was farther southeast in the SREF-N1-MM5 solution (Fig. 11g). During the next 12-h, anomaly B strengthened and broadened, while anomaly C elongated to the northeast (Figs. 11b,e). Notably, anomaly B is stronger and extends farther southeast in the GFS-MM5 solution, resulting in a positive difference centered over Alabama (Fig. 11h). This positive difference grew in size and amplitude over the next 12 h as it moved toward the coast, and was associated with a stronger and farther northeastward anomaly B in the GFS-MM5 solution (Figs. 11c,f,i). The stronger and farther northeast position of anomaly B corresponded to a stronger and farther north cyclone and associated banding. Comparison of the difference fields reveals the growth of the differences (Figs. 11g–i). In general, the forecast differences between solutions in this case were large, with local difference maxima exceeding 3 PVU.

Fig. 11.
Fig. 11.

Comparison of the 475–250-hPa layered-average PV (shaded according to scale, PVU) evolution over the 36-km domain for the (a)–(c) northwest (GFS-MM5) and (d)–(f) southeast (SREF-N1-MM5) members at 1200 UTC 11 Feb 2006 (initialization), 0000 UTC 12 Feb 2006 (12-h forecast), and 1200 UTC 12 Feb 2006 (24-h forecast). (g)–(i) Corresponding 475–250-hPa layered-average PV difference (GFS-MM5 minus SREF-N1-MM5).

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

3) 14 February 2007

The IC subensemble upper-level PV mean at initialization (15 h prior to band formation) exhibits a broad, negatively tilted PV tail extending southeastward from the northern plains into the southeast United States (Fig. 5g). Within the PV tail, two embedded PV anomalies were centered near Georgia (A) and northeast Missouri (B). The initial upper-level PV spread was the smallest among the three cases, and was generally <0.6 PVU. Over the next 18 h, anomaly A moved northeast along the coast, well ahead of the primary upper trough (anomaly B) (Figs. 5g–i). Spread increased downstream of each anomaly, but was generally <1 PVU by 1800 UTC 14 February 2007—the smallest spread in the vicinity of the banding among the cases. At lower levels the initial spread was comparable to the other two cases (Fig. 6g); however, by forecast hour 12 a distinct lower-PV anomaly is present off the Virginia coast. This anomaly moved north along the coast over the next 6 h (Figs. 6g–i). Although the spread associated with this feature was relatively large (>1 PVU), the spread was centered on the anomaly, suggesting differences in amplitude rather than location. The dominance of the lower-level PV and the lagging nature of upper-level anomaly B in this case is consistent with the lagging upper-trough classification (Novak et al. 2010).

The GFS-MM5 was the farthest northwest member within the IC subensemble, while the NAM-MM5 was the farthest southeast member. Band predictions from these members were only separated by ~100 km at the time of observed band maturity (1800 UTC 14 February 2007) (Fig. 11c). The GFS-MM5 band duration and timing were within 1 h of the observed, while the NAM-MM5 predicted band formation 5 h late, and band duration was half the observed (4 h) (Fig. 7f).

The GFS-MM5- and NAM-MM5-analyzed surface lows were nearly identical, with positions within 20 km of each other and identical pressures (1003 hPa) (Figs. 12a,d). The observed low was 1001 hPa, and was located in nearly the same position (Fig. 12g). The forecast evolution of the cyclone developed similarly in each member, such that by 1800 UTC 14 February 2007 (18-h forecast) the surface lows are both located just off the New Jersey coast with pressures within 2 hPa of each other (Figs. 12c,f) and within 1 hPa of the observed (Fig. 12i). During the cyclone evolution, widespread convection (evidenced by dBZ > 40 dBZ) was present in the simulations and observations near the low center and along the trailing cold front (Fig. 12).

Fig. 12.
Fig. 12.

Comparison of the MSLP and simulated reflectivity from the (a)–(c) GFS-MM5 and (d)–(f) NAM-MM5 to (g)–(i) the observed radar mosaic and NCEP surface analysis low position for the 14 Feb 2007 case.

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

Differences in the initialized upper-level PV at 0000 UTC 14 February 2007 for the two members were small (~0.25 PV) (Fig. 13g), consistent with the relatively small MSLP differences. Differences 18 h into the forecast were generally <0.75 PVU and were associated with a smaller separation distance between PV anomalies A and B in the GFS-MM5 solution compared to the NAM-MM5 results (cf. Figs. 11c,f). Overall, differences in MSLP and upper-level PV are hard to discern during the evolution of the 14 February 2007 case, consistent with the small band location differences (e.g., Fig. 7c). However, the small differences over the Midwest and southeast United States did grow as anomalies A and B interacted during the 0000–1800 UTC period.

Fig. 13.
Fig. 13.

Comparison of the 475–250-hPa layered-average PV (shaded according to scale, PVU) evolution over the 36-km domain for the (a)–(c) northwest (GFS-MM5) and (d)–(f) southeast (NAM-MM5) members at 0000 UTC 14 Feb 2007 (initialization), 1200 UTC 14 Feb 2007 (12-h forecast), and 1800 UTC 14 Feb 2007 (18-h forecast). (g)–(i) Corresponding 475–250-hPa layered-average PV difference (GFS-MM5 minus NAM-MM5).

Citation: Weather and Forecasting 27, 3; 10.1175/WAF-D-11-00047.1

c. Representativeness

To explore the representativeness of the opposing outlier member evolutions, we compared the second-most extreme northwest and southeast location solutions among the IC subensemble. This analysis confirmed that the key features identified by the previous analysis were also present, but that the particular evolution of the shape and magnitude of the PV anomalies could not be consistently related to band location differences (not shown). In other words, although the key PV anomaly from members with southeast (northwest) band solutions may evolve to be farther east (west) in general, each member may have different strengths and shapes of the key PV anomaly.

5. Discussion and future work

a. Ensemble predictions

The results of this study show that even at forecast projections <24 h there can be considerable uncertainty in the occurrence and, especially, the timing and location of band formation and subsequent band evolution. For example, band occurrence was favored in the ensemble for each case, but the specific timing and location of the bands among the members varied (especially in the 25 December 2002 and 12 February 2006 cases). This result suggests that answering whether a band will occur during a given case may be easier than answering when and where it will occur, even for a relatively short forecast. Even with this uncertainty, a 12-km grid-spacing ensemble identified favored time periods and corridors of band formation threat (e.g., Figs. 3a–c). The current (as of 2011) operational NCEP SREF grid spacing (~30 km) precludes obtaining such information.

Although limited to three cases, the results also demonstrate that a 16-member multimodel ensemble at 12-km grid spacing using different initial conditions and physics may be able to provide useful information on the magnitude of this uncertainty and, thus, help differentiate between a case with high predictability (14 February 2007) and much lower predictability (12 February 2006). Such ensemble systems may help forecasters communicate corresponding unusually high or low confidence of band evolution scenarios. Additional cases are necessary to establish whether this kind of mesoscale ensemble prediction system can reliably discriminate the case-to-case variability in band predictability.

Although the ensemble provided useful uncertainty information, the observed timing of the band formation fell outside the ensemble envelope in the 12 February 2006 case. These errors may be symptomatic of an underdispersive ensemble system (e.g., Stensrud et al. 2000; Grimit and Mass 2002; Jones et al. 2007). Correcting model deficiencies and applying improved perturbation approaches are likely necessary steps to achieve an ensemble system that produces reliable and sharp distributions of band characteristics. Increasing the ensemble system resolution to convection-allowing scales (i.e., <4 km) may also improve snowband prediction, as demonstrated for warm season convection over the central United States (Clark et al. 2009, 2010a) and tropical cyclone rainfall (e.g., Zhang et al. 2010). The development and testing of such convection-allowing ensemble systems and their utility to provide feature-based uncertainty information for cool-season phenomena in a real-time forecast environment is encouraged.

Additional cases are required to assure that the results presented here are representative of a larger sample of banded cyclones, and that the assumed ensemble skill–spread relationship is valid. For example, the case with the shortest lead time (15 h in the 14 February 2007 case) had the smallest spread, whereas the case with the longest lead time (21 h in the 12 February 2006 case) had the largest spread. Although it seems unlikely that the 6-h difference in initialization could be the primary reason for the difference in spread, a larger dataset of cases could help establish the expected skill for particular lead times. Furthermore, the observed cases examined in this study were known to exhibit band formation. Application of the ensemble system to observed nonbanded (null) cases [e.g., as defined and analyzed in Novak et al. (2010)] may reveal false alarms, where the ensemble exhibits a high probability of band formation, yet formation does not occur. Additionally, predictions at longer forecast lead times are necessary to establish current practical predictability limits of snowbands in northeast U.S. cyclones.

The location and timing diagnostics (i.e., Figs. 3 and 4) illustrate the type of band uncertainty information possible from a 12-km grid-spacing ensemble. This information is unavailable from the current generation of regional operational ensemble systems. However, as the operational NCEP SREF evolves toward 12-km grid spacing (J. Du 2010, personal communication), feature-based mesoscale uncertainty information, such as that demonstrated here, may become available operationally. Such feature-based ensemble information was noted as a forecaster need in a recent National Weather Service–sponsored survey, as reported by Novak et al. (2008a). The subjectivity of band identification limits the approach here to a demonstration. However, the use of object-oriented approaches (e.g., Davis et al. 2006) to identify and track band predictions in an objective manner may lead to automating the band identification process. Automated band identification may accelerate the real-time application of SREFs for band predictions, and foster a more probabilistic approach to the band prediction problem.

b. Sensitivity of band location to initial conditions

Comparison of the patterns of forecast evolution of members with opposing outlier solutions of band location within the IC subensemble shows significant case-to-case variability in the meteorological uncertainties among the three cases examined. A critical question is why one case may be more predictable than another.

The quality of the initial conditions is likely a key factor affecting predictability. Among the three cases, large (small) initial differences in the upper-level PV distribution were correlated to large (small) differences in forecast band locations. The PV differences were closely associated with PV anomalies, suggesting the initialized details of the anomalies (i.e., short waves) are important for the subsequent band forecast. This result is consistent with those of Zhang (2005), who showed in his case that the maximum error growth is found in the vicinity of upper-level and surface zones, with the strongest PV gradient over the area of active moist convection. This result is also similar in general to Schlemmer et al. (2010), who found that the representation of the details of PV streamers affected the stability, moisture flux, and orographic forcing for precipitation events along the south side of the Alps. Thus, targeted improvements in the initialization of key PV anomalies may improve band location predictions.

The analysis in this paper focused on short-term forecasts and the associated regional errors. However, remote upstream errors (e.g., over the Pacific) may also contribute to cyclone predictability via downstream development (Szunyogh et al. 2002; Hakim 2005; Chang 2005; Colle and Charles 2011). The relative importance of larger-scale upstream initial condition errors versus local initial condition errors in cyclone (and associated precipitation) predictability is an open question.

The limited case examples also suggest that the complexity of the synoptic pattern (number of interacting PV anomalies) may be another factor associated with the case-to-case differences in predictability. For example, the 25 December 2002 case exhibited at most two PV anomalies and associated difference maxima in the domain, while the highly uncertain 12 February 2006 case exhibited at least three PV anomalies and associated difference maxima. Hakim et al. (1996) suggest that predicting PV interactions is an inherently difficult challenge.

The growth of differences between solutions was evident, consistent with previous mesoscale predictability research (e.g., Zhang et al. 2003; Zhang et al. 2007; Hohenegger and Schär 2007). Diabatic heating associated with deep convection has been established to be important to extratropical cyclone development and associated precipitation (e.g., Davis et al. 1993; Dickinson et al. 1997; Mahoney and Lackmann 2007), and has been shown to contribute to nonlinear error growth and reduced predictability (Zhang et al. 2007; Clark et al. 2010b; Hohenegger et al. 2006; Hohenegger and Schär 2007; Baxter 2011). Thus, could the large uncertainty in the 12 February 2006 event be attributed to the presence of deep convection along the Gulf and southeast U.S. coasts? Radar images show that although convection was present, it was not widespread (Fig. 10). In contrast, the 14 February 2007 case exhibited active convection throughout its life cycle (Fig. 12), yet exhibited the least band location spread. Thus, the degree of convection associated with each system does not discriminate between the band uncertainty for these three cases.

Overall, the results suggest that predictability differences in band location among the cases can be related to the quality of the initial conditions and the complexity of the initial flow, whereas predictability differences were not correlated to the degree of convection. Although initial condition sensitivity was emphasized in this work, sensitivities to model core, physics, and lateral boundary conditions likely also contribute to band location uncertainty. This contribution likely varies on a case-by-case basis. Future work should explore these sensitivities.

Despite the short-time ranges (<24 h), the sign of the initial location and magnitude errors of the key upper-level short waves do not necessarily translate into similarly signed errors in forecast band location. For example, the farther south and stronger PV anomaly initialized at 0000 UTC 25 December 2002 in the SREF-N1-MM5 later evolved to be weaker and farther east. The intrinsic nonlinear nature of the atmosphere and model error may contribute to this result. Thus, forecasters assuming location and magnitude errors in the initial fields will translate into similarly signed errors in short-range forecasts of band location may be mistaken. However, the SREF-N1-MM5 solution in the 12 February 2006 case illustrates that egregious errors in the initial fields (surface low 200 km southwest of the observed) can be used to help forecasters hedge away from the subsequent forecast. On the other hand, this approach does not account for the nonlinear effects of model and physics errors, which may allow a poorly initialized forecast to “self-correct” later in the evolution (e.g., Bright and Nutter 2004).

Only northwest–southeast band location differences were qualitatively explored. Sensitivities in band timing and occurrence also need to be examined. For example, Novak et al. (2010) found that statistically significant differences between the midlevel frontogenesis maximum of the banded and null events only appeared ~2 h prior to band formation. The simplified approach and limited ensemble size employed in this work precluded a more comprehensive examination of these aspects of band predictability. However, future work using objective ensemble sensitivity approaches (e.g., Bishop et al. 2001; Hakim and Torn 2007; Ancell and Hakim 2007) may help to conclusively identify the sources of band formation and evolution uncertainty in a systematic and comprehensive way. Ensemble sensitivity estimates how any arbitrary perturbation at a single point, spread spatially by the covariance relationships of the ensemble, would alter the response function (precipitation in this case). This property allows ensemble sensitivity to identify areas requiring additional observations, and show forecasters meteorological features like fronts or upper-level troughs (i.e., PV anomalies), which may most directly affect the band forecast. This knowledge has the potential to increase a forecaster’s situational awareness of band uncertainties and ultimately improve band predictions.

Acknowledgments

The second author was supported in part by UCAR-COMET (Grant S07-66814) and NOAA-CSTAR (NA10NWS4680003). Martin Baxter provided insightful comments that improved the manuscript. Jun Du provided the archived NCEP SREF perturbations necessary to run the ensemble. Three anonymous reviewers provided constructive comments leading to improvements in the presentation of this work.

REFERENCES

  • Ancell, B., , and Hakim G. J. , 2007: Comparing adjoint and ensemble sensitivity analysis with applications to observation targeting. Mon. Wea. Rev., 135, 41174134.

    • Search Google Scholar
    • Export Citation
  • Baxter, M. A., 2011: Impacts of very small initial condition errors on mesoscale aspects of two cyclones. Electron. J. Oper. Meteor., 12, 2011EJ01.

    • Search Google Scholar
    • Export Citation
  • Bei, N., , and Zhang F. , 2007: Impacts of initial error scale and amplitude on the mesoscale predictability of heavy precipitation along the mei-yu front of China. Quart. J. Roy. Meteor. Soc., 133, 8399.

    • Search Google Scholar
    • Export Citation
  • Bishop, C. H., , Etherton B. J. , , and Majumdar S. J. , 2001: Adaptive sampling with the ensemble transform Kalman filter. Part I: Theoretical aspects. Mon. Wea. Rev., 129, 420436.

    • Search Google Scholar
    • Export Citation
  • Black, T. L., 1994: The new NMC mesoscale Eta Model: Description and forecast examples. Wea. Forecasting, 9, 265278.

  • Brennan, M. J., , and Lackmann G. M. , 2005: The influence of incipient latent heat release on the precipitation distribution of the 24–25 January 2000 U.S. East Coast cyclone. Mon. Wea. Rev., 133, 19131937.

    • Search Google Scholar
    • Export Citation
  • Bright, D. R., , and Nutter P. A. , 2004: On the challenges of identifying the “best” ensemble member in operational forecasting. Preprints, 16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., J11.3. [Available online at http://ams.confex.com/ams/pdfpapers/69092.pdf.]

  • Caplan, P., , Derber J. , , Gemmill W. , , Hong S.-Y. , , Pan H.-L. , , and Parrish D. , 1997: Changes to the 1995 NCEP operational medium-range forecast model. Wea. Forecasting, 12, 581594.

    • Search Google Scholar
    • Export Citation
  • Chang, E. K. M., 2005: The impact of wave packets propagating across Asia on Pacific cyclone development. Mon. Wea. Rev., 133, 19982015.

    • Search Google Scholar
    • Export Citation
  • Charles, M. E., , and Colle B. A. , 2009: Verification of extratropical cyclones within the NCEP operational models. Part II: The short-range ensemble forecast system. Wea. Forecasting, 24, 11911214.

    • Search Google Scholar
    • Export Citation
  • Clark, A. J., , Gallus W. A. Jr., , Xue M. , , and Kong F. , 2009: A comparison of precipitation forecast skill between small convection-allowing and large convection-parameterizing ensembles. Wea. Forecasting, 24, 11211140.

    • Search Google Scholar
    • Export Citation
  • Clark, A. J., , Gallus W. A. Jr., , and Weisman M. L. , 2010a: Neighborhood-based verification of precipitation forecasts from convection-allowing NCAR WRF model simulations and the operational NAM. Wea. Forecasting, 25, 14951509.

    • Search Google Scholar
    • Export Citation
  • Clark, A. J., , Gallus W. A. Jr., , Xue M. , , and Kong F. , 2010b: Growth of spread in convection-allowing and convection-parameterizing ensembles. Wea. Forecasting, 25, 594612.

    • Search Google Scholar
    • Export Citation
  • Colle, B. A., , and Charles M. E. , 2011: Spatial distribution and evolution of extratropical cyclone errors over North American and adjacent oceans in the NCEP Global Forecast System model. Wea. Forecasting, 26, 129149.

    • Search Google Scholar
    • Export Citation
  • Davis, C. A., , Stoelinga M. T. , , and Kuo Y.-H. , 1993: The integrated effect of condensation in numerical simulations of extratropical cyclogenesis. Mon. Wea. Rev., 121, 23092330.

    • Search Google Scholar
    • Export Citation
  • Davis, C. A., , Brown B. , , and Bullock R. , 2006: Object-based verification of precipitation forecasts: Part I: Methodology and application to mesoscale rain areas. Mon. Wea. Rev., 134, 17721784.

    • Search Google Scholar
    • Export Citation
  • Dickinson, M. J., , Bosart L. F. , , Bracken W. E. , , Hakim G. J. , , Schultz D. M. , , Bedrick M. A. , , and Tyle K. R. , 1997: The March 1993 Superstorm cyclogenesis: Incipient phase synoptic- and convective-scale flow interaction and model performance. Mon. Wea. Rev., 125, 30413072.

    • Search Google Scholar
    • Export Citation
  • Du, J., , DiMego G. , , Tracton M. S. , , and Zhou B. , 2003: NCEP short-range ensemble forecasting (SREF) system: Multi-IC, multi-model and multi-physics approach. Research Activities in Atmospheric and Oceanic Modeling, J. Cote, Ed., CAS/JSC Working Group Numerical Experimentation Rep. 23, WMO/TD 1161, 5.09–5.10.

  • Du, J., and Coauthors, 2004: The NOAA/NWS/NCEP short-range ensemble forecast (SREF) system: Evaluation of an initial condition vs multiple model physics ensemble approach. Extended Abstracts, 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., 21.3. [Available online at http://ams.confex.com/ams/84Annual/techprogram/paper_71107.htm.]

  • Du, J., , McQueen J. , , DiMego G. , , Toth Z. , , Jovic D. , , Zhou B. , , and Chuang H.-Y. , 2006: New dimension of NCEP short-range ensemble forecasting (SREF) system: Inclusion of WRF members. Preprints, WMO Expert Team Meeting on Ensemble Prediction System, Exeter, United Kingdom, World Meteorological Organization, 5 pp. [Available online at http://www.emc.ncep.noaa.gov/mmb/SREF/WMO06_full.pdf.]

  • Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46, 30773107.

    • Search Google Scholar
    • Export Citation
  • Dudhia, J., 1993: A nonhydrostatic version of the Penn State–NCAR Mesoscale Model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121, 14931513.

    • Search Google Scholar
    • Export Citation
  • Eckel, F. A., , and Mass C. F. , 2005: Aspects of effective mesoscale, short-range ensemble forecasting. Wea. Forecasting, 20, 328350.

  • Evans, M., , and Jurewicz M. L. , 2009: Correlations between analyses and forecasts of banded heavy snow ingredients and observed snowfall. Wea. Forecasting, 24, 337350.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterizations. Mon. Wea. Rev., 121, 764787.

  • Grimit, E. P., , and Mass C. F. , 2002: Initial results of a mesoscale short-range ensemble forecasting system over the Pacific Northwest. Wea. Forecasting, 17, 192205.

    • Search Google Scholar
    • Export Citation
  • Hakim, G. J., 2005: Vertical structure of midlatitude analysis and forecast errors. Mon. Wea. Rev., 133, 567578.

  • Hakim, G. J., , and Torn R. D. , 2007: Ensemble synoptic analysis. Synoptic–Dynamic Meteorology and Weather Analysis and Forecasting: A Tribute to Fred Sanders, Meteor. Monogr., No. 55, Amer. Meteor. Soc., 147–161.

  • Hakim, G. J., , Keyser D. , , and Bosart L. F. , 1996: The Ohio Valley wave-merger cyclogenesis event of 25–26 January 1978. Part II: Diagnosis using quasigeostrophic potential vorticity inversion. Mon. Wea. Rev., 124, 21762205.

    • Search Google Scholar
    • Export Citation
  • Hawblitzel, D. P., , Zhang F. , , Meng Z. , , and Davis C. A. , 2007: Probabilistic evaluation of the dynamics and predictability of the mesoscale convective vortex of 10–13 June 2003. Mon. Wea. Rev., 135, 15441563.

    • Search Google Scholar
    • Export Citation
  • Hohenegger, C., , and Schär C. , 2007: Predictability and error growth dynamics in cloud-resolving models. J. Atmos. Sci., 64, 44674478.

    • Search Google Scholar
    • Export Citation
  • Hohenegger, C., , Luthi D. , , and Schär C. , 2006: Predictability mysteries in cloud-resolving models. Mon. Wea. Rev., 134, 20952107.

  • Hong, S.-Y., , Dudhia J. , , and Chen S.-H. , 2004: A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation. Mon. Wea. Rev., 132, 103120.

    • Search Google Scholar
    • Export Citation
  • Janjić, Z. I., 2003: A nonhydrostatic model based on a new approach. Meteor. Atmos. Phys., 82, 271285.

  • Jones, M. S., , Colle B. A. , , and Tongue J. S. , 2007: Evaluation of a mesoscale short-range ensemble forecast system over the northeast United States. Wea. Forecasting, 22, 3655.

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

    • Search Google Scholar
    • Export Citation
  • Lorenz, E. N., 1963: Deterministic nonperiodic flow. J. Atmos. Sci., 20, 130141.

  • Lorenz, E. N., 1969: The predictability of a flow which possesses many scales of motion. Tellus, 21, 289307.

  • Lorenz, E. N., 1996: Predictability—A problem partly solved. Seminar on Predictability, Vol. 1, ECMWF, 1–19.

  • Mahoney, K. M., , and Lackmann G. M. , 2007: The effect of upstream convection on downstream precipitation. Wea. Forecasting, 22, 255277.

    • Search Google Scholar
    • Export Citation
  • Morgan, M. C., , and Nielsen-Gammon J. W. , 1998: Using tropopause maps to diagnose midlatitude weather systems. Mon. Wea. Rev., 126, 25552579.

    • Search Google Scholar
    • Export Citation
  • Nicosia, D. J., , and Grumm R. H. , 1999: Mesoscale band formation in three major northeastern United States snowstorms. Wea. Forecasting, 14, 346368.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Bosart L. F. , , Keyser D. , , and Waldstreicher J. S. , 2004: An observational study of cold season–banded precipitation in northeast U.S. cyclones. Wea. Forecasting, 19, 9931010.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Waldstreicher J. S. , , Keyser D. , , and Bosart L. F. , 2006: A forecast strategy for anticipating cold season mesoscale band formation within eastern U.S. cyclones. Wea. Forecasting, 21, 323.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Bright D. R. , , and Brennan M. J. , 2008a: Operational forecaster uncertainty needs and future roles. Wea. Forecasting, 23, 10691084.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Colle B. A. , , and Yuter S. E. , 2008b: High-resolution observations and model simulations of the life cycle of an intense mesoscale snowband over the northeastern United States. Mon. Wea. Rev., 136, 14331456.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Colle B. A. , , and McTaggart-Cowan R. , 2009: The role of moist processes in the formation and evolution of mesoscale snowbands within the comma head of northeast United States cyclones. Mon. Wea. Rev., 137, 26622686.

    • Search Google Scholar
    • Export Citation
  • Novak, D. R., , Colle B. A. , , and Aiyyer A. , 2010: Evolution of mesoscale precipitation band environments within the comma head of northeast U.S. cyclones. Mon. Wea. Rev., 138, 23542374.

    • Search Google Scholar
    • Export Citation
  • Ralph, M. F., and Coauthors, 2005: Improving short-term (0–48 h) cool-season quantitative precipitation forecasting: Recommendations from a USWRP workshop. Bull. Amer. Meteor. Soc., 86, 16191632.

    • Search Google Scholar
    • Export Citation
  • Reisner, J., , Rasmussen R. M. , , and Bruintjes R. T. , 1998: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124, 10711107.

    • Search Google Scholar
    • Export Citation
  • Schlemmer, L., , Martius O. , , and Sprenger M. , 2010: Disentangling the forcing mechanisms of a heavy precipitation event along the Alpine south side using potential vorticity inversion. Mon. Wea. Rev., 138, 23362353.

    • Search Google Scholar
    • Export Citation
  • Schumacher, R. S., 2011: Ensemble-based analysis of factors leading to the development of a multi day warm-season heavy rain event. Mon. Wea. Rev., 139, 30163035.

    • Search Google Scholar
    • Export Citation
  • Schumacher, R. S., , Schultz D. M. , , and Knox J. A. , 2010: Convective snowbands downstream of the Rocky Mountains in an environment with conditional, dry symmetric, and inertial instabilities. Mon. Wea. Rev., 138, 44164438.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., , Klemp J. B. , , Dudhia J. , , Gill D. O. , , Barker D. M. , , Wang W. , and Powers J. G. , 2005: A description of the Advanced Research WRF version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp.

  • Stensrud, D. J., , Bao J. , , and Warner T. T. , 2000: Using initial condition and model physics perturbations in short-range ensemble simulations of mesoscale convective systems. Mon. Wea. Rev., 128, 20772107.

    • Search Google Scholar
    • Export Citation
  • Suarez, A., , Reeves H. D. , , Wheatley D. , , and Coniglio M. , 2012: Comparison of ensemble Kalman filter–based forecasts to traditional ensemble and deterministic forecasts for a case study of banded snow. Wea. Forecasting, 27, 85105.

    • Search Google Scholar
    • Export Citation
  • Szunyogh, I., , Toth Z. , , Zimin A. V. , , Majumdar S. J. , , and Persson A. , 2002: Propagation of the effect of targeted observations: The 2000 Winter Storm Reconnaissance Program. Mon. Wea. Rev., 130, 11441165.

    • Search Google Scholar
    • Export Citation
  • Tan, Z.-M., , Zhang F. , , Rotunno R. , , and Snyder C. , 2004: Mesoscale predictability of moist baroclinic waves: Experiments with parameterized convection. J. Atmos. Sci., 61, 17941804.

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

    • Search Google Scholar
    • Export Citation
  • Toth, Z., , and Kalnay E. , 1997: Ensemble forecasting at NCEP and the breeding method. Mon. Wea. Rev., 125, 32973319.

  • Zhang, F., 2005: Dynamics and structure of mesoscale error covariance of a winter cyclone estimated through short-range ensemble forecasts. Mon. Wea. Rev., 133, 28762893.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , Synder C. , , and Rotunno R. , 2003: Effects of moist convection on mesoscale predictability. J. Atmos. Sci., 60, 11731185.

  • Zhang, F., , Bei N. , , Rotunno R. , , Snyder C. , , and Epifanio C. C. , 2007: Mesoscale predictability of moist baroclinic waves: Convection-permitting experiments and multistage error growth dynamics. J. Atmos. Sci., 64, 35793594.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , Weng Y. , , Kuo Y.-H. , , Whitaker J. S. , , and Xie B. , 2010: Predicting Typhoon Morakot’s catastrophic rainfall with a convection-permitting mesoscale ensemble system. Wea. Forecasting, 25, 18161825.

    • Search Google Scholar
    • Export Citation
Save