Climatology and Predictability of Cool-Season High Wind Events in the New York City Metropolitan and Surrounding Area Michael Layer School of Marine and Atmospheric Sciences Stony Brook University Stony Brook, NY
Outline Motivation Data & Methods Climatology Results Model Verification Results Conclusions
Motivation CountsStatistics WarningsEventsScores TotalVerif NOT Verif TotalWarned NOT Warned PODFAR OKX Poor verification statistics for High Wind Warning events Source: NWS Performance Management [available online at
Motivational Questions What synoptic regimes and meteorological conditions are common to non-convective cool-season high wind events? – Height and MSLP pattern evolution – Low level winds and stability What atmospheric mechanisms might be responsible for the generation of high surface winds? How well does an ensemble forecast system improve the prediction of NCWEs as compared with deterministic forecasts?
Data & Methods Observational data RUC 13 km analyses – 1-hr temporal resolution – Available NARR 32 km analyses – 3-hr temporal resolution – Available ACARS profiles – Variable temporal resolution – Available ASOS METARs – 5-minute temporal resolution – Available for 6 primary climate sites
Data & Methods Climatology study period: through “cool seasons” (15 September – 15 May) Model verification study period: through seasons Use 5-Minute ASOS observations to find observed High Wind Warning criteria (Sustained wind 35 kts or higher and/or wind gust 50 kts or higher) – Observations >=9 hours apart are separate events – Exception: Regime change (frontal passage) Exclude convective events Bin events into 3 common types – Pre-cold frontal – Post-cold frontal – Nor’easter/coastal storm
Event Type Classification Pre-cold frontal (PRF) – 14 observed events – 1 false alarm Post-cold frontal (POF) – 32 observed events – 4 false alarms Nor’easter/Coastal storm (NEC) – 14 observed events – 7 false alarms
Observed PRF events
Observed POF events
Observed post-cold frontal events
Observed NEC events
Large-Scale Synoptic Summary Pre-cold frontal (14 observed events) – Trough deepening and going negatively tilted over the Great Lakes, ridging over the W. Atlantic/SE Canada – Simultaneous strengthening of negative and positive height & MSLP anomalies, oriented in a WSW-ENE direction Post-cold frontal (32 observed events) – Trough deepening and going negatively tilted over the NE U.S./SE Canada – Rapidly strengthening negative height & MSLP anomalies Nor’easter/Coastal (14 observed events) – Trough deepening and going negatively tilted over the Great Lakes, ridging over the W. Atlantic – Simultaneous strengthening of negative and positive height & MSLP anomalies, oriented in a SW-NE direction
Diurnal Climatology of High Winds
Influence of Height Gradient
PRF Vertical Profiles
NEC Vertical Profiles
POF Vertical Profiles
Larger-Scale vs. Smaller-Scale Factors Diurnal factor insignificant except for POF events Correlation between maximum height gradient & maximum wind/gust is positive but not significantly positive LLJ strength/height cannot solely differentiate between an observed event vs. a false alarm PRF profiles – Stable low-levels – Strong LLJ signature POF profiles – Weak low-level stability – Weak LLJ signature NEC profiles – Moderate low-level stability – Strong LLJ signature
SREF Ensemble Verification Verification of the previous version of SREF (October 2009 – August 2012) 32 km horizontal resolution, 21 members, 4 different model cores Two daily model runs (09Z and 21Z) Verified against RUC analyses and ACARS profiles Four forecast lead time periods – 1 st period: 1-6 hour – 2 nd period: 9-24 hrs – 3 rd period: hrs – 4 th period: hrs
2 nd Period Verification
Verification Summary The SREF as a whole has a high bias in low- level ( hPa) wind forecasts The WRF-ARW core has the largest error in over-predicting the LLJ SREF ensemble provides a significant improvement in skill over the deterministic (control) members
Conclusions Highest FAR with NEC events Lowest POD with POF events Large-scale synoptic evolutions can be used as pattern recognition/analog tools Height/pressure gradient and LLJ height & strength cannot solely determine an event vs. a non-event Analyze wind profile & low-level stability plus other factors (if applicable) SREF over-predicts LLJ, but still provides improved forecast skill over any of the 5 control members