These results are synthesized into an operational method which assesses cyclogenesis, deformation, frontogenesis, and conditional stability in a down-scale approach. This method will be illustrated through application to the 6–7 January 2002 snowstorm. This storm featured an intense snowband which was responsible for 30–40 cm of snowfall accumulation in eastern New York. Illustration of the method will draw on available forecast guidance and observations from this case from 24 h preceding band formation through band dissipation.
 

Snow developed across central and northeastern Pennsylvania during the late afternoon on January 6^th, 2002 as a major storm developed along the mid-Atlantic coast. Localized, narrow bands of heavy snow developed within the main snow shield, and produced total snow accumulations of 10 to 20 inches across a region from central Pennsylvania through eastern New York. The storm developed as an intense, compact mid-tropospheric short-wave trough lifted northward along the East Coast downstream from a major long-wave trough located over the Mississippi Valley. This study will compare and contrast that storm to a second storm that produced a single band of heavy snow across northern Pennsylvania and southern New York on January 19^th. Snow accumulations within that band reached up to 12 inches across the southern tier of upstate New York. In contrast to the storm on January 6^th, the storm on the 19^th was associated with a relatively weak area of low-pressure, and a flat, progressive mid-tropospheric flow pattern.
 

The most striking similarity between the storm on January 6^th and the storm on January 19^th was that localized bands of heavy snow developed and played havoc with snowfall forecasts. A comparison of the observed and model forecast data from both storms indicate that in both cases, the heavy snow bands developed in areas where the quasi-geostrophic forcing for upward motion was relatively weak. In the case on the 6^th, a pronounced maximum of quasi-geostrophic forcing for upward motion was forecast, however the maximum was south of the observed heavy snow area. On the 19^th, strong, organized areas of quasi-geostrophic upward motion forcing were completely absent. In both cases, the heavy snow developed in association with frontogenesis associated with frontal boundaries that sloped upward from south to north across Pennsylvania and southern New York.
 

Cross-sectional analysis of each event indicated that the snow banding in both cases was likely enhanced by instability located along and just above the sloping frontal boundary. In the case on January 19^th, a cross-sectional analysis of theta-e indicated that conditional instability was present above the frontal surface. By contrast, in the case on January 6^th a cross-sectional analysis of theta-e indicated a relatively statically stable environment above the frontal surface. In that case, evidence is shown that the environment near the heavy snow bands may have been associated with inertial instability in the geostrophic wind field. The resulting horizontal accelerations, as the real wind field attempted to adjust to this unstable condition, could have resulted in areas of slantwise convection. It is hypothesized that the small scale of the mid-level short-wave trough with this event was associated with very sharp downstream ridging, which resulted in the inertial instability.
 

As was the case with on January 19^th, the data shown for this case indicates the importance of utilizing cross-sections when diagnosing model data associated with a potential snowstorm. Comparing and contrasting the data from January 6^th and January 19^th also illustrates that a variety of environments can produce banded snowstorms.
 

Cutoff cyclones are identified objectively. For our purposes, a cutoff cyclone is defined as a geopotential height minimum surrounded by at least one closed 30 m interval contour. Cutoffs are identified and catalogued, and cyclone tracks are constructed, to delineate favored areas for genesis/lysis and to locate "cutoff freeways." Frequency diagrams showing total number of cutoff cyclones and number of "cutoff 6 h periods" are presented for the Northern and Southern Hemispheres and for eastern North America. Also shown are maps of observed genesis/lysis, the "cutoff grid point of the year," as well as the "cutoff day of the year." Histograms of cutoff activity for the Northern and Southern Hemispheres are presented as well.
 

Cutoff cyclones are associated with many significant forecasting problems throughout the world, but particularly so in the northeastern United States. Given the complex terrain found in the Northeast, the precipitation distribution associated with slow-moving cutoff cyclones is often a challenge to predict and can wreak havoc among local communities.
 

A climatology of warm season (1 May – 30 September) 500 hPa cutoff cyclones is derived from four times daily (0000, 0600, 1200, and 1800 UTC) 500 hPa gridded geopotential height fields. The data used are analyses from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis dataset: 1948–2002.
 

A cutoff cyclone is defined as a geopotential height minimum surrounded by at least one closed contour based on a 30 m interval. Cutoff cyclones meeting this objective criterion are identified and catalogued. Cyclone tracks are determined to delineate favored areas for genesis/lysis and to locate "cutoff cyclone corridors." Frequency diagrams showing total number of cutoff cyclones and number of "cutoff 6 h periods" are presented for the Northern and Southern Hemispheres and for eastern North America.
 

In-progress and future work includes creating objectively derived tracking products to see if there are favored trends for the orientation of the "cutoff cyclone corridors" on an annual basis. Our cutoff climatology will also be used in conjunction with the Unified Precipitation Dataset (UPD) to map precipitation distributions in cutoff cyclones over the northeastern United States. A case study will be presented to document the evolution and environmental impact of a cutoff cyclone.
 

Over the northeastern United States, northwesterly lower-tropospheric flow regimes are ocasionally associated with production of heavy precipitation, especially over the considerable orography of northern New York, Vermont, New Hampshire and western Maine. Two cutoff 500 hPa cyclone events in the autumn of 1999 produced heavy snowfall over the mountains of northern New York State and northern Vermont. These two cases provided an operational impetus for studying the characteristics of upslope snowfall events. The goal of this research is to produce ingredients-based conceptual models and operational methodologies for the purpose of improved prediction of the precipitation patterns produced by these flow regimes in the complex terrain of the northeastern United States.
 

Cases selected for this study were limited to events occurring with prevailing deep-tropospheric northwesterly flow, which excluded any cases involving rapid-genesis coastal cyclones (i.e., Nor’easters). Northwest-flow scenarios generally produce a significant low-level flow component orthogonal to the Green Mountains and Adirondack Range, which is favorable for the generation or enhancement of heavy precipitation by orographic lift. To date, six scenarios have been examined in this study: three events which produced heavy snowfall northern Vermont and northern New York State; two events which had been forecast to produce heavy snowfall, yet significant precipitation failed to occur; and one very weak northwest-flow event.
 

NCEP/NCAR Reanalysis data were used in determining the synoptic-scale characteristics of each of the cases, supplemented with ETA model BUFR sounding analysis data in order to interrogate the mesoscale structure of each event. Additionally, 5-km grid spacing mesoscale ETA model simulations have been performed both to isolate mesoscale signatures, and for comparison with their coarser-resolution operational ETA model (40-km) counterparts. Diagnostic findings from this study suggest several meteorological factors significant to the development of heavy precipitation from this type of flow regime: (a) the low-level moisture profile; (b) the strength and orientation of the low-level wind with respect to the orography; (c) the low-level static stability profile. Low- and high-resolution model comparisons have yielded some measure of forecast success by the 5-km ETA. Operational forecast procedures and techniques currently in development will be presented, as will direction for future investigation.
 

Warm season closed lows provide a variety of severe weather across the Northeast ranging from large hail events to significant flash flooding. On 28 May 2001, Memorial Day, a closed low moving equatorward from Hudson Bay into the Great Lakes Region initiated numerous large hail producing thunderstorms. The severe weather event consisted primarily of hail from penny (1.9 cm) to golf ball size (4.4 cm). The National Weather Service at Albany’s County Warning Area had over a half dozen of its counties impacted. Some counties were hit by large hail multiple times. Late spring crop and orchard damage occurred in portions of Eastern New York. Current CSTAR warm season closed low climatology research has shown that there are several categories or tracks of closed lows.
 

A strong "closed" upper low and its occluded front triggered the severe weather on Memorial Day. The thunderstorms began to develop around noontime with the maximum diurnal heating. The instability generated between the cold pool aloft and the surface heating helped initiate the thunderstorms. Over the next several hours many pulse thunderstorms developed over the Greater Capital Region, the upper Hudson River Valley, and Western New England.
 

A multi-scale analysis will be shown for the 28 May 2001 event to address what caused the hail episode in the Northeast. Various ETA and AVN model grids, surface observations, upper-air data including soundings, cross-sections and Doppler radar data including Vertically Integrated Liquid (VIL) values will be examined. The importance of the cold pool aloft, lapse rates, vorticity advection, and upper-level jet streaks will be stressed in the generation of the hail producing pulse thunderstorms.
 

Approximately fifteen years of hourly meteorological data from several moored buoys along the northeast U.S. coast have been examined, with the goal of developing an operationally useful climatology of winds and waves. The buoys examined were 44009 (Delaware Bay), 44025 (south of Long Island), 44013 (near Boston) and 44007 (near Portland Me). The data were obtained from the National Data Buoy Center’s Internet Web site as text files, which were imported into a spread sheet program for sorting, analysis, and creation of chart displays. Each hourly report contains the wind direction, wind speed, significant wave height, dominant wave period, and various other meteorological information.
 

The primary goal was to determine average wave height as a function of wind speed and direction. Theoretically, the height of wind-driven waves depends mainly on wind speed, fetch length, and time. For near-shore locations, the fetch length varies greatly with wind direction. Thus wind speed and direction account for two of the three factors (excluding wind duration). Besides duration, this analysis does not account for waves generated by distant storms.
 

Therefore, the presentation of results will include primarily wave height means and standard deviations as a function of wind direction (eight compass points) and wind speed (five knot bins) for the four coastal buoys along the mid-Atlantic and northeast U.S. coast. At the Delaware Bay buoy (44009), for example, the highest waves for a given wind speed tend to occur with a northeast wind, and the lowest with a southwest or west wind (see attached chart). Similar results can be shown for the other buoys. This likely results from the greater frequency of sustained northeast winds with coastal storms. However, each buoy has a unique wind-wave profile depending on the nearby coastline. In addition to wave heights, wind frequency as a function of speed and direction will also be shown.
 

It is hoped that the results of this study will be useful as guidance for marine forecasters at offices along the northeast U.S. coast, and perhaps could be incorporated into "Smart Tools" for the marine portion of the Interactive Forecast Preparation System.
 

The challenge of flash flood forecasts continues to perplex forecasters. This system offers a paradigm shift in our thinking about flash flood predictions. To date, the main focus of advances in assessing the possibility of flash floods has been on operational mesoscale models and improved quantitative precipitation forecasts, despite the fact that some studies indicate the serious shortcomings of convective parameterization schemes (Gallus, 1999). Even with rapid advances in processing speeds, the forecast community has reached a plateau in flash flood prediction which will likely not be changed until cumulus scale models are quasi_operational, perhaps within five years. While cumulus scale models may hold high hopes, many areas will still be at risk of flash flooding during the next half decade of development. This system offers a bold initiative to increase advance watches and warnings of flash flooding, specifically for the Middle Atlantic River region, but with applications elsewhere in the nation.
 

The foundation of this system is the research by Grumm and Hart (1999) establishing a real_time operational assessment of climatological anomalies of various model forecast fields based on derived monthly means from the NCEP reanalysis work (1948_2000). Since October 1999, the National Weather Service Office in State College has been producing graphical displays of model forecast anomaly fields for the Eta, Aviation, and the locally run MM5 computer models. Anomaly fields consist of the vertical mass_weighted mean anomaly for height, temperature, wind and moisture. These parameters are compared to the climatological mean to determine the anomaly. This anomaly system, developed by Grumm and Hart, is being used to create anomaly fields for historically significant rainfall events over the MARFC domain dating back to 1948. These significant rainfall events will be chosen based on the observed/estimated rainfall values. Once anomaly fields have been created, a pattern recognition exercise will be conducted to classify these events by time of year and significant types (ie. narrow cold frontal rain bands, mesoscale convective systems, etc.)
 

On 17 June 2001, Father’s Day, the remnants of once Tropical Storm Allison tracked across southern New England, dumping as much as six inches of rain in a three to nine hour period. Widespread flash flooding affected portions of southeastern New York, Connecticut, Rhode Island and eastern Massachusetts. Mainstem rivers also sharply rose. Many mainstem rivers approached bankfull and some exceeded flood stage, most notably the Yantic River at Yantic, Connecticut and Neponset River at Norwood, Massachusetts. The extreme intensity of the rainfall over such a short period of time, coupled with difficulties in obtaining real-time measurements of river stage and precipitation resulted in significant challenges forecasting the magnitude of the flooding along mainstem rivers in Southern New England.
 

Through use of sound hydrometeorological practices, forecasters at the Northeast River Forecast Center (NERFC) utilized modified unit hydrographs in order to issue operational forecasts that were significantly better than model forecasts using the default unit hydrographs contained in the river model.
 

Early on the morning of September 16^th 1999 hurricane Floyd made landfall on the North Carolina coast and moved rapidly northeastward. The winds associated with Floyd weakened rapidly after landfall, but a swath of heavy precipitation developed ahead of the storm from North Carolina to New England. Over the Northeast, the heaviest precipitation (20-40 cm) fell in 12-18 h across northern New Jersey, southeastern New York, and central Connecticut, resulting in over one billion dollars in flood damage and 16 fatalities over the Northeast. Even 24 hours leading up to the event, the operational NCEP models failed to properly forecast the magnitude and location of heavy precipitation. The purpose of this study is to use the Penn State-NCAR Mesoscale Model (MM5) down to 1.3 km grid spacing to objectively evaluate how well this modeling system can simulate the evolution of Floyd and the heavy precipitation over the Northeast. The MM5 was also used to diagnose the physical mechanisms for the heavy rainfall by conducting simulations to separate the effects of latent heating, surface heat fluxes, and local topography on the circulations and strengthening baroclinic zone near the coast.
 

The MM5 was initialized at 0000 UTC 16 September 1999 using the NCEP Eta model interpolated on a 40-km grid. The MM5 was nested using a one-way interface over the Eastern U.S. at 36,12, 4, and 1.3 km grid spacing, with the highest resolution over southern New England. The control run of the MM5 used the Grell convective scheme (at 36- and 12-km grid spacing), MRF PBL scheme, Reisner2 microphysics (which includes graupel and super-cooled water), and the Dudhia radiation scheme. The MM5 precipitation was verified against all available Cooperative observer and NWS rain gauges by interpolating the MM5 precipitation to the observation points.
 

Even though the MM5 was initialized using the NCEP Eta, the 36-km MM5 produced a much better quantitative precipitation forecast than the 32-km Eta over the Northeast U.S. By trying several different convective and PBL parameterizations in the MM5, it was determined that the BMJ convective scheme in the Eta may have contributed to a much weaker storm and under-predicted precipitation amounts. The precipitation distribution was also sensitive to the Eta sea surface temperatures (SSTs), since using the higher resolution Navy OTIS SSTs in the MM5 resulted in a better forecast of the maximum precipitation location. The MM5 at 4- and 1.3 km grid spacing realistically simulated the 20-40 cm of precipitation over the flooded areas, but over-predicted the amounts at many locations. A MM5 simulation without topography still resulted in 20-40 cm of precipitation at high resolution and very little difference in storm track; therefore, the inland terrain played a secondary role in enhancing the heavy precipitation when compared to the deep frontogenetical circulation. A simulation without surface heat fluxes resulted in a 5-10 mb weaker storm shifted 100-200 km farther east, with 30-50% less precipitation. Another simulation without diabatic precipitation effects resulted in a 20-30 mb weaker storm that did not propagate up the coast. Therefore, diabatic effects from precipitation and surface fluxes were critical in maintaining the storm motion and frontogenetical circulations, even as Floyd transitioned to extra-tropical along the coast.
 

This research focuses on the physical processes relating to a recent event of precipitation in which at least 25 mm of liquid equivalent precipitation occurred within a 12h period. The event of 7 January 2002 was that of heavy snow in southern Vermont and the Albany, New York regions that was evidently unrelated to topographic forcing. In the Burlington County Warning Area, the intensity of the precipitation was not forecasted by the operational models 24 hours before the event. Although the case occurred in a large_scale flow that was favorable for cyclogenesis, several crucial details relating to the extreme mesoscale development are identified and studied. Additionally, cyclogenetic mechanisms are studied.
 

The mesoscale environment is studied in the context of both upright and slantwise convection. The processes that were responsible for this destabilization include upper_level cooling associated with the advance of an upper_level short_wave trough, and the development of conditional symmetric instability and coupling with the dynamic tropopause.
 

A synoptic overview of the 7 January 2002 heavy snow event will be presented showing the distribution of the heavy snow, radar reflectivity, and upper air signatures. Mesoscale model output from the NCEP ETA and the McGill MC2 depicting the utility of dynamic tropopause for tracking the intensity of the upper level short_wave trough and the coupling to the lower troposphere will be shown. Finally, suggestions for use by the operational forecaster will be presented.
 

Category F3 or greater intensity tornadoes are rare in eastern New York and western New England with only six since 1950. On 31 May 1998, an F3 tornado struck Mechanicville, New York, injuring 66 people and causing 71 million dollars in damage. The tornado was part of a widespread, severe weather outbreak across the Northeast that killed five, produced 30 tornadoes, 369 reports of wind damage and 151 reports of large hail. This paper will review the synoptic conditions that spawned the severe weather, as well as examine the mesoscale and storm scale environments that produced the Mechanicville tornado.
 

The upper level pattern that produced the 31 May 1998 tornado outbreak began to evolve during the week prior to the event. An upper level closed low moved south into Hudson’s Bay on 26 May and remained there for much of the next week, finally moving through the Canadian Maritimes on 4-5 June. Short waves rotating around it spawned three major northeastern United States severe weather outbreaks (29 May, 31 May and 02 June). The 31 May outbreak was enhanced by a second short wave that ejected out of another closed upper level low that moved south from the Gulf of Alaska (20-24 May) to the coast of California (30 May). At 250 hPa, the coupling of two strong (130 kt) jets helped generate a large area of enhanced upward vertical motion. In response to the strong upper-level dynamics, unseasonably strong surface low pressure (986 hPa) strengthened as it moved from the northern Plains States through the Great Lakes into eastern Canada. A very strong low-level jet (55 kt) provided the mechanism to rapidly transport a very warm, moist airmass over the Ohio Valley into the Northeast and contributed to high shear and helicity in the lower troposphere.
 

During the afternoon of 31 May, the warm front associated with the surface low moved through eastern New York. In the warm sector, CAPE of around 2000 J kg^-1, storm-relative helicity of greater than 450 m² s^-2 and 0-6 km shear greater than 25 m s^-1 indicated the potential for supercells and tornadoes. Terrain may have played a role in further enhancing the tornadic potential of the Mechanicville storm as it moved into the immediate Hudson Valley. The southwest low-level flow over the high terrain to the west was channeled to a more southerly direction in the valley, increasing low-level shear and helicity. The southerly flow also advected warm, moist air poleward more rapidly than in surrounding high terrain areas increasing instability.
 

A line of thunderstorms in western New York at 1600 UTC intensified and moved east producing locally severe weather as it reached central New York at 1824 UTC, about two hours prior to the Mechanicville tornado touchdown. At that time, a small cluster of isolated storms developed about 55 km to the east of the line. One of these storms intensified as it moved east-northeast. By 1942 UTC it had turned severe and was about 30 km ahead of the line. The lead cell moved into the somewhat more unstable and more highly sheared environment of the Hudson Valley about 2000 UTC. Tornado touchdown occurred at 2022 UTC, about the time the outflow from the line of storms to the west caught up with the lead cell.
 

In recent years forecast train wrecks have often been associated with errors involving the representation of mesoscale boundaries and organized deep convection in numerical models as much as synoptic_scale forecast errors. Mesoscale boundaries serve as loci for concentrated thermal, moisture and vorticity gradients. Failure to resolve these boundaries properly in numerical models can adversely affect the timing, location, intensity and duration of forecast precipitation. Organized deep convection is important in enhancing downstream ridge/jet development. Failure to properly resolve the bulk upscale effects of organized deep convection in numerical models can hinder the self development process during cyclogenesis.
 

A significant additional problem is that critical mesoscale boundaries often hide in plain sight, often making themselves known as the forecast is going down in flames. This state of affairs can arise because our enormous database of surface observations is vastly underutilized by models and forecasters alike. Reasons for this problem will be discussed and illustrative examples will be shown.
 

 
Historically, an average of four tornadoes occur each year across the northern New England states of Maine, New Hampshire and Vermont. In recent years, tornado frequency has diminished. It contradicts the expectation of increased tornado detection commonly associated with the deployment of highly sensitive Doppler radars and population increases. Most are classified as "weak" (F0-F1) on the Fujita scale, but "strong" tornadoes (F2-F3) do occur.
 

Our understanding of northern New England tornadogenesis is limited. They tend to be short lived, ill-defined and often do not exhibit radar and environmental characteristics of their Midwest tornado counterparts. They are elusive and accompanied by little or no warning lead time. Storms that produce tornadoes in northern New England sometimes do not even exhibit radar characteristics that typically result in the issuance of severe thunderstorm warnings. Automated detection schemes have not faired well either. Other complications, such as visibility restrictions due to rugged terrain and the deeply forested and rural nature of the region, degrade the ability for visual detection.
 

To help mitigate the impact of detection limitations and improve preparedness in the warning process, this study was conducted to produce a composite tornado day in northern New England. Environmental conditions will be discussed and compared to the Johns and Dorr (1996) study of strong and violent tornado episodes across an expanded New England-wide domain. A frequency distribution created from the Storm Prediction Center tornado data set will be presented. Mesoscale and synoptic features will be combined with radar information to derive a composite tornado environment. Topographical maps will be superimposed with tornado touchdowns to identify correlations of varied terrain with tornadogenesis. Finally, commonly used convective parameters derived from modeled soundings will be examined to demonstrate a tornado case in a high shear, low instability regime.
 

Given the growing volume of forecast and observational data, it is becoming increasingly important for forecasters to extract scientifically relevant information quickly before and during severe weather forecast and warning operations. During this presentation, composite charts of model forecast fields will be supplemented with composite charts of hourly surface analysis to highlight the physical processes supporting severe storm development, convective organization, and storm mode. To capture this information, composite charts are designed to illustrate 1) measures of instability and vertical wind shear, 2) three dimensional moisture availability, content, and distribution and, 3) synoptic and mesoscale forcing mechanisms. Composite charts are effective because they highlight important information, rather than exposing all possible information. By using color, images, and contours effectively, the presented charts allow forecasters to quickly focus on the most important information, resulting in quick recognition of the convective potential.
 

This presentation highlights a severe convective wind event that effected parts of Ohio, Pennsylvania, West Virginia, New York, and Maryland on 9 March 2002. It will be shown that visually enhanced composite charts promote quick recognition of the contributing forecast and real-time severe weather processes. Composite charts presented here focus on evaluating the large scale potential for severe weather with the Eta, and monitoring the evolution of the convective environment with the hourly Rapid Update Cycle (RUC). In this context, information from the hourly RUC provides a critical link to the model forecasts.
 

Successive runs of the Eta indicated 1) significant vertical wind shear associated with an anomalously strong low-level jet, 2) moderate moisture and, 3) considerable low-level forcing from a combination of strong moisture flux convergence and frontogenesis along a vigorous frontal system. Instability was forecast to be the primary limiting factor; the warm sector environment was expected to become only weakly unstable, with CAPE forecast to remain less than 500 jkg^-1, and the Best Lifted Index (BLI) expected to reduce no further than -2. The large scale potential thus suggested the possibility of a narrow frontal squall line with strong convective wind gusts, but weak instability would probably limit the areal extent of severe weather.
 

A self-contained meteorological emergency response vehicle (MERV) is described that provides emergency management weather support during natural disasters and/or homeland defense incidents by integrating this new technology into operational weather support. Fully automated on-site surface and upper air meteorological data is required to feed dispersion models during nuclear, chemical, or biological weapon attacks, as well as during natural disasters where conventional weather observations are unavailable (forest fires, earthquake aftermath, etc.). The system provides continuous automated in-situ measurements of surface pressure, temperature, humidity (PTU), and winds, as well as full upper air profiles from radiosondes and low level PTU and winds from a tethersonde. In addition, it provides remote sensing of boundary layer winds from Doppler lidar and present weather images from a color total sky imager. The sky imager with its horizon-to-horizon view is used to provide users with near real-time images of sky cover and visibility.
 

The system is envisioned to be deployed by unskilled personnel and can be operated with minimal user training, avoiding the need for a trained meteorologist. Internet connectivity provides remote monitoring, control and diagnostics, and real time data is output in standard meteorological formats. The core of the system is an automated radiosonde launcher mounted on a trailer or pickup truck, which supports the other met sensors. The launcher is capable of automatically inflating and releasing sequences of multiple radiosondes at predetermined times or remotely on demand. The launcher uses standard radiosondes from most manufacturers. Alternately, a tethersonde can be deployed providing PTU and winds about every 250 m up to 1.5 km altitude. An eye-safe Doppler lidar is used to obtain high-resolution boundary layer winds, providing u, v, and w wind components in the boundary layer. Measuring the vertical moment of the wind will provide a valuable input to transport models to determine the three dimensional dispersion of a substance such as radioactive fallout.
 
 

During the period November 1, 2001 through May 1, 2002 the National Weather Service (NWS) conducted a Winter Weather Experiment (WWE). Participants included eight Eastern Region Weather Forecast Offices (WFOs), three components of the National Centers for Environmental Prediction (NCEP) - the Hydrometeorological Prediction Center (HPC), Environmental Modeling Center (EMC), and NCEP Central Operations (NCO), with project oversight from the NWS Office of Climate Weather and Water Services (OCWWS). The goal of the experiment was to enhance winter weather services to the public via a suite of enhanced products from HPC for use by the WFOs in support of the their winter weather watch and warning program.
 

Two secondary objectives were also addressed during the experiment. The first was to test NCEP’s newly implemented experimental Short Range Ensemble Forecast (SREF) system and its application to winter weather forecasting. The second was to test HPC’s anticipated role of collaboration with the WFOs as a prototype to the paradigm envisioned to be utilized during the upcoming National Forecast database (NDFD) era.
 

The suite of enhanced products produced by HPC incorporated output from the 10 member 48km SREF system in addition to model output routinely available to HPC. The combination of SREF and operational model output allowed HPC to produce ensemble based winter storm watch/warning graphics for snow and ice accumulations. These graphics depicted the percent the ensemble based prediction of snow and ice accumulation would exceed 24 hour winter storm watch criteria for each county within the WWE area. In addition, an ensemble based low tracks graphic was produced and depicted the expected paths of major surface lows through 72 hours at 12 hour intervals. This graphic also depicted the spatial uncertainty of the location of the surface low. In addition, a text discussion conveying meteorological reasoning for the graphics, the storm potential and uncertainty for forecast day 4 through 7 was also provided. Results from the experiment will be presented and will include subjective feedback from WWE participants (WFO and NCEP HPC) on the utility of the WWE and the performance of the SREF during the WWE. Further, an introduction to this year’s WWE (2002-2003) will be provided as well.

During the afternoon and early evening hours of 19 January 2002, a narrow band of heavy snowfall affected portions of northern Pennsylvania and southern New York. This band was roughly 30 kilometers wide, with the most intense and persistent snowfall rates oriented west to east from near Elmira, New York across the Binghamton, New York area, then further eastward into the Catskill mountain region. Storm totals of eight to 12 inches in about a six hour period were common within this small corridor.
 

This event had some unique characteristics that likely contributed to its "surprise" nature. First, the prevailing flow pattern was flat and progressive as indicated by observed upper air and well initialized model data. This led to weak patterns of vorticity and thermal advection at levels where standard meteorological analysis are typically performed. Second, a strong upper level speed maximum was located from New York state into southern New England, with the more favorable entrance region of this upper level jet core across the Mid-Atlantic region, well south of the heavy snow band. Third, despite uniform liquid equivalent precipitation totals (generally 0.20" to 0.30") observed across New York and Pennsylvania for this event, actual snow accumulations varied widely across the region. Interestingly, the models’ quantitative precipitation forecasts (QPF) were consistent and fairly accurate.
 

So why did heavy snow develop? A cross-sectional inspection of observed and model data during the time of heavy snowfall revealed a frontal boundary that sloped upward from south to north, with the northern extent of the front located across New York and Pennsylvania. Strong frontogenetical forcing in mid-levels of the atmosphere was associated with the frontal zone over southern New York. Additionally, pronounced diffluence developed just above the zone of frontogenesis as the area came under the exit region of a 500 mb jet streak moving through the central Appalachians. This juxtaposition created a shallow, but vigorous circulation promoting strong lift between 600 and 500 mb. Cross-sectional analysis also revealed an area of weak conditional instability in a shallow layer just above the frontal zone, near 500 mb. This instability would have acted to enhance and localize the circulation associated with the frontogenesis, and could have resulted in localized areas of convection, reinforcing the intensity of the banded feature associated with frontogenesis. Finally, another key contributor to the heavy snow appeared to have been a thermal structure supportive of favorable snow growth mechanisms within a deep and saturated layer. This aspect was well portrayed by hourly sounding profiles. It is theorized that each of the above factors played a significant role, with the absence of any one of them possibly resulting in far less snowfall.
 

This case study showed the potential value of viewing model fields using cross-sections. Particularly when mechanisms of interest are shallow in nature, they could fall between mandatory or significant levels or be "smoothed out" if looking at a mean layer in plan view.
 

During the past several years, real-time numerical weather prediction has spread rapidly from operational centers such as the National Centers for Environmental Prediction (NCEP) to universities, government agencies, and private industry. For example, the Stony Brook University (SBU) has been running the Penn State-National Center for Atmospheric Research mesoscale model (MM5) in real-time since the fall of 1999. The SBU MM5 is run for both the 0000 and 1200 UTC cycles at 36-,12-, and 4-km horizontal grid spacing. Data are available in NetCDF format for ingestion and integration with other data sets in meteorological workstations, such as the National Weather Service’s (NWS) Advanced Weather Interactive Processing System (AWIPS). The data can also be visualized via the Internet.
 

In order for these regional modeling efforts to help operational forecasters and model development, it is imperative that objective verification of the model forecasts be completed, with the results effectively passed to the forecaster. SBU has been verifying the MM5 against all available observations over the Eastern U.S. The results have also been compared with the NCEP eta model. SBU has also teamed with the NWS Upton, NY to learn how to use mesoscale model output in an operational mode and to learn the strengths and weaknesses of the MM5 modeling system. A Cooperative Program for Operational Meteorological Education and Training (COMET) Partners Project was completed and some of the results will be presented at the workshop.
 

Recently, SBU, the NWS Forecast Offices at Upton, New York, Taunton, Massachusetts, and Mt. Holly, New Jersey, as well as, the Northeast River Forecast Center and the Eastern Region Headquarters were awarded a new, three-year contract from COMET to examine this problem in more detail. In addition to scientific aspects, the goals of this ongoing effort are both operational forecasting and educationally oriented. Scientific goals include evaluation of model sensitivity to initial conditions, parameterized physics, as well as, quantifying the benefits of event-based verification versus conventional seasonal average verification. Operational forecasting goals include developing operationally oriented visualization tools for mesoscale models, ensembles and verification data sets. Lastly, educational goals include training forecasters to understand and utilize mesoscale models, ensembles and real-time verification more effectively in the forecast process.
 

This presentation will discuss the results from previous work and plans for the three-year COMET collaborative effort that has recently started.
 

Full implementation of a real-time forecasting system at McGill University was completed on 30 May 2002. The forecast office of Environment Canada in St. Laurent, Quebec, and the National Weather Service Forecast Office in Burlington, Vermont, have been providing feedback on this forecast effort. The system includes a once-daily (0000 UTC) 48-h run of the Mesoscale Compressible Community (MC2) model employing a triple-nesting strategy with grid spacings of 36, 12, and 3km. An accounting of the system’s performance timings and statistics will be presented. The 36-km domain extends from the central plains to beyond the Atlantic continental shelf. The 12-km domain encompasses a region bounded by Lake Michigan to the west, and the eastern seaboard. The high-resolution 3-km domain focuses on the Champlain and St. Lawrence Valleys. Data for initial and boundary conditions are supplied by the 0000 UTC forecast run of the operational Canadian Global Environmental Multiscale Model. The system’s domains focus in on the southwestern Quebec and Champlain Valley region with the goal of evaluating high-resolution deterministic forecasts of precipitation amounts for southern Quebec and the northeastern states. Specifically, we wish to examine the influence of channeled flows on precipitation patterns and mesoscale frontogenesis in the St. Lawrence and Champlain Valleys.
 

We present a diagnosis of the output from the real-time forecasting system in the form of a case study validation for the 0000 UTC run of 27 September 2002. A heavy cloud shield and rainband on the northern leading edge of the transitioned ex-hurricane Isidore transited the forecast region from 1200 UTC 27 to 1200 UTC 28 September, and resulted in 24-hour precipitation accumulations of greater than 50 mm at numerous upstate New York and New England stations. We present validation of the three forecast domains, and compare dynamic tropopause analyses from the forecast cycle with those produced by the Rapid Update Cycle (RUC). The later verification will ensure that the large-scale model dynamics resemble the observed evolution of the case study, while the former deals with the interaction of synoptic to mesoscale dynamics and thermodynamics.
 

Additionally, a case of severe squall-line convection occurring at 0000 UTC 18 July 2002 is investigated and verified against enhanced satellite imagery and two special radiosonde launches from the McGill-operated St. Hilaire Research Station. One of the ascents took place in the pre-storm environment, and the other during the convective burst. The latter balloon penetrated the updraft core (vertical velocities in excess of 15 m/s were observed), providing a unique opportunity for us to verify the storm’s intensity on the high-resolution sub-domain with observational data. The MC2 real-time forecasting system is shown to have predicted the storm remarkably well for a 24-h lead time.
 

 
 
Since the introduction of the Interactive Computer Worded Forecast system (ICWF), and later the Interactive Forecast Processing System/Graphical Forecast Editor (IFPS/GFE) system, for public forecast preparation, forecast verification has become a secondary concern to most forecasters at WFO Albany, and probably most other offices. For the past two years, the focus of most forecasters has been on learning how to use these systems to prepare their forecasts, and on getting through the labor intensive day-to-day task of preparing grid field forecasts for the public forecast elements, without much attention being paid to the accuracy of their forecasts. Compared to long-term trends, MAX/MIN verification scores for Albany (ALB) have remained relatively stable in terms of their percent improvement over FWC (NGM MOS) forecasts for the first four periods. However, for the past two years, the percent improvement over the MAV (AVN MOS) forecasts has remained substantially negative. In addition, during the past two years, MAX/MIN forecasts for Glens Falls (GFL) and Poughkeepsie (POU) have also been verified, as well as fifth period forecasts for all three stations. These scores show that, overall, forecasters at Albany generally are have greater errors than the FWC guidance, and even greater errors than the MAV guidance at GFL and POU. Fifth period temperature verification scores for all three stations show that the MAV MOS guidance ranges from 10 to 50 percent better than Albany forecasters.
 

Despite having the superior MAV guidance available, improving model forecasts, and two years to become familiar with verifying MAX/MIN temperatures at GFL and POU, and for the fifth period, Albany forecasters have made little progress in improving their scores during the ICWF/IFPS/GFE transition period.
 

Now that Albany forecasters have become familiar with how to prepare the grid field forecasts, a plan was recently implemented in order to help Albany forecasters increase their focus on verification of their temperature forecasts. This plan will involve a multi-step approach, and will include the creation of locally derived smart tools to create better temperature grid field forecasts.
 

The plan will also involve raising forecaster awareness as to which temperature forecast guidance is best, and how to incorporate it into their grid field forecasts. For example, MAV MOS guidance has routinely done better than the FWC guidance since its inception. However, verification scores show that on a routine basis, the average of the MAV and FWC temperature forecasts provides the best forecast overall.
 

Finally, individual one on one training will be provided to each forecaster in order to insure that they understand how the current verification system works within the IFPS/GFE framework. Key components of this training will be discussed.
 

Geographic Information Systems (GIS) lets users link relational databases to spatial analysis. It is this link which has great applications when it comes to dealing with large quantities of meteorological, climatological and hydrologic data. The primary purpose of this presentation will be to introduce the audience to GIS through a spatial analysis of data sets commonly used in the NWS ranging from verification data to forecast analysis and procedural practices. Detailed spatial analysis will be presented which focuses on the Cooperative Observer Program, Forecast Grid Fields across Eastern Region, as well as Procedural Issues including Winter Weather Thresholds and the Determination of the Growing Season. Applications of GIS in Hydrology will also be examined from looking at basinwide average precipitation fields to highlighting individual creeks which may be susceptible to flooding.
 

Two concepts gleaned from the Summer 2002 Warning Decision Making Workshop will be presented. The presentation will emphasize the need, and techniques for conducting post mortems of severe weather events. The presentation will show lead forecasters or other supervisors how to recognize fatigue symptoms of front line staff during a severe weather event, and how to best utilize the available staff. Various techniques such as rotating duties among available employees and others will be discussed.