March 2-3, 2014 Storm Summary

March 2-3, 2014 Light Snow

radarA slow moving frontal boundary shifted south through the region as an anomalously cold air mass entered the Northeast US, with a weak wave of low pressure developing along this frontal boundary and producing widespread precipitation from the central US into the Mid Atlantic. The system was originally forecast to produce moderate accumulations in the NYC area, but the southward push of cold, dry air was underestimated, with snow accumulations ending up less than an inch across the area.

Jump to: Storm Synopsis | Forecasting the Storm | Storm Timeline | Storm Impacts



March 2-3, 2014 Storm Archive

NWS Regional Radar
WPC Surface Analysis
GFS Model Archive


Storm Synopsis

GFS initialized 500mb geopotential heights and vorticity, showing 12z Sunday, 3/2 on the left and 12z Monday, 3/3 on the right, depicting the evolution of the upper level setup (image credit: NCEP Model Analyses and Guidance)


No organized surface low pressure developed from this system, although the precipitation event was fueled by multiple factors, including a southern stream shortwave which earlier brought much-needed rain to California and provided lifting over the central US, a strong piece of the polar vortex over southeast Canada aiding in suppressing the system, and a strong baroclinic zone separating an anomalously cold and dry air mass to the north from a warm and moist air mass to the south. With the addition of the placement of the right entrance quadrant of a jet streak over the central US, strong upward vertical motion occurred near the frontal boundary supporting the development of heavy precipitation near Arkanas and Tennessee, which then continued ENE along with the mid-upper level flow into the Mid Atlantic region.

Despite the time of the year, the snow from this system was significantly suppressed as far south as Virginia with an anomalously cold air mass to the north. The polar vortex was positioned over the eastern Hudson Bay into northern Quebec, with a strong vorticity lobe extending to its north over the central Hudson Bay. This vort lobe then quickly dived southeast, suppressing geopotential heights over the northeastern US ahead of the southern shortwave trough, allowing for a southward push of strong cold, dry air into the region as the system arrived, which later resulted in another round of record cold temperatures, with some locations in the Mid Atlantic setting records for the coldest March temperatures in recorded history. This lowering of heights prevented additional amplification of the storm system, and aided to further shear out the southern shortwave as the frontal boundary continued to be pushed southwards, continuously suppressing the precipitation shield south of the NYC area.


Forecasting The Storm

February 27 – Near Record Cold Tonight
February 28 – Cold, Snowy Start to March

March 1 – Snow Returns Monday
March 2 – Light Snow Tomorrow
March 2 Storm Updates


Post-Storm Analysis – Model Guidance & Forecast Analysis

This section is split into two parts; the first analyzes the model guidance leading up to the event, specifically the ECMWF, GFS and CMC model guidance, and the second reviews the forecast verification leading up to the event and where the forecast went wrong.


2/24 12z ECM at hour 168, depicting a low pressure over Ohio with snow to rain for the NYC area and heavy snow to the north. Note the modeled positioning of the polar vortex east of Canada near the Davis Strait (image credit: PSU e-Wall).

f168Long Range: The long range model guidance has consistently hinted at a storm potential during this time period, as far back as the 2/22 12z run of the ECMWF, nearly 10 days away, and even the GFS as far back as 2/19 in its extreme long range, when it tends to have much lower accuracy, has consistently centered a storm system around the March 3-4 time frame. As the time frame entered the somewhat more reliable 192 hour range on February 24, both the GFS and ECMWF depicted a low pressure tracking through the region, primarily driven by the southern stream shortwave with little influence from the northern stream, producing either a mix or rain in the area, and heavy snow to the north.

During this time period, both models incorrectly handled the positioning of the polar vortex, depicting it east of Canada, near the Davis Strait, thus exerting little influence over the region with no suppression to keep the storm significantly further south. Considering this was in the longer range, many small errors likely resulted in this significant difference later into the runs, although one source of error appears to originate from poor handling of the upper level flow in Canada, as well as slight mishandling of the shortwave troughs over the NE US on 2/24 and 2/26, the former which rapidly deepened offshore later and further east than modeled. This shortwave was thus modeled too slowly as it approached Greenland, allowing for more interaction between the two shortwaves earlier noted and lowering heights too much over the Davis Strait, which along with an underestimation of the coverage area of ridging over Greenland also resulted in the depiction of the polar vortex departing too quickly.


Medium Range: The 2/25 runs of the model guidance continued to depict a significant storm system affecting the region; the ECMWF depicted a fairly fast moving wave of low pressure producing rain to snow on the 0z run and heavy snow on the 12z run, with the event ending on Monday night (3/3), while the GFS and CMC were too slow with the system, depicting a long duration significant snow event from Sunday (3/2) through Tuesday (3/4). The GFS especially incorrectly handled the vort lobe extending into western Canada and was also a little too slow with the timing of the southern shortwave, extending the event into Tuesday, while the CMC was the only major model to depict more of a frontal passage type event, with a SW-NE oriented front steadily progressing east rather than a stalled frontal boundary producing a west to east prolonged snow event.

By February 26, the southern stream was handled relatively well, despite relatively minor timing, track and strength issues of the shortwave trough approaching California. The more significant errors in the model guidance, however, originated in Canada with the model guidance continuing to show difficulty in correctly handling the polar vortex and the elongated vorticity lobe extending into southwestern Canada. The model guidance particularly struggled to correctly handle the upper level setup beginning around the 3/1 frame, which as of 2/26 was still 60-72 hours away, with the errors primarily focused on the strength of the northern end of the polar vortex near the Hudson Bay, determined by how low geopotential heights would be under the northern end of the upper level low, and the strength of a vort lobe in this region. This would later help to determine the orientation of the polar vortex and the southward extent of the frontal boundary, along with the strength of the main storm system once it arrives in the northeast US depending on the interaction between the north and south streams.

Below is one of the time frames where the models diverged with the upper level setup, focusing on 3/1 12z (7am); the initialized 500mb heights/vorticity from the GFS is to the bottom left, and the forecast from the 2/26 18z run is to the right. The strength of the polar vortex and vort lobe over the Hudson Bay, denoted by the darker orange colors and the “X” in the bottom left map over the northern Hudson Bay, was underestimated, which when run forward through 3/3 allowed for less suppression and a more significant storm affecting the region (image credit: NCEP MAG).



2/26 18z run of the GFS at hour 120, on 18z Monday (1pm), depicting the polar vortex positioned favorably to interact with the southern shortwave rather than to suppress it, with a relatively consolidated wave of low pressure producing moderate to heavy snow over the area (image credit: PSU e-Wall).

f120As of the 2/26 runs, the GFS and ECMWF consistently showed a major snowstorm affecting the region, with over 10-12 inches of snow, with the GFS correcting by the 2/26 18z run to depict a shorter event through Monday evening (3/3) rather than a long duration event into Tuesday (3/4). This continued through 2/27 as well. As previously noted, the divergence point for the model guidance was on 3/1, when the polar vortex entered the Hudson Bay; these model runs underestimated the strength of the polar vortex, particularly of the northern vort lobe, which was modeled to take longer to swing south and do so further west than actually observed. This set up for more of an elongated orientation of the vortex from SW to NE, which aligned favorably with the timing of the southern shortwave to prevent suppression of heights in the northeast US, and with some runs actually allowed for some slight amplification downstream of the shortwave trough. As a result, this allowed for more interaction between the two streams, preventing the main storm system from becoming suppressed or significantly sheared and placing the frontal boundary in a favorable position for a 18-24 hour period of moderate to heavy snow training over the region without the precipitation shield shifting south throughout the event.

While both models continued to slightly correct the positioning and strength of the polar vortex and the northern vort lobe on the 2/27 runs, these changes were not substantial enough to result in a significant change in the surface reflection yet, continuing to depict a major snowstorm over Pennsylvania into New Jersey but with the northern end of the precipitation shield slowly trending south over the central New England region. Notably, despite mishandling the exact setup and timing of the polar vortex and the southern shortwave, the CMC model was generally the most suppressed model and was more consistent in depicting a SW-NE oriented precipitation shield, although it was notably too slow with the system with precipitation extending well into Tuesday, 3/4, and with some of its runs depicted heavy snow accumulations over PA/NJ as well, but not to the extent of the GFS and ECM. The CMC’s 2/27 12z run was the first to show a suppressed outcome closer to reality, despite being too slow, with somewhat better handling of the northern stream, depicting the precipitation shield gradually shifting southeast with the heavier snow focused over West Virginia and Virginia.


2/28 12z run of the GFS at hour 78, on 18z Monday (1pm), depicting a more suppressed and elongated system as it gradually began to gain a better handle on the polar vortex strength and positioning, but still underestimated the magnitude of the southern trend (image credit: PSU e-Wall).

f78Short Range: The slight corrections towards a stronger northern vort lobe continued with the 2/28 runs, eventually reaching a point where these slight differences began to have a more significant impact downstream on the region and begin the south trend. The GFS continued to depict a significant snowstorm with over 10 inches for the area through the 2/28 6z run, only 66-78 hours prior to the onset of precipitation, but by the 12z run the correction towards a slightly stronger northern vort lobe than the 6z run, which while still not as strong as was actually observed, resulted in this stronger vort lobe swinging south faster and thus further east than the earlier runs. Instead of favorably aligning the northern stream with the southern shortwave allowing for more interaction between the two, supporting a relatively strong system with no suppression of the precipitation shield throughout the duration of the event, this trend ended up allowing the northern stream to outrun the southern shortwave, reaching the northeast US ahead of the main event and suppressing heights aloft, preventing additional amplification of the storm and instead resulting in some shearing and weakening of the system as it spread east. The remainder of the 2/27 runs of the GFS did not depict the vort lobe as strong as it was in reality despite continuing to trend slightly stronger with this feature, which still resulted in a moderate snow event producing at least 4-8 inches of snow across the area.

2/28 0z run of the CMC at hour 90, on 18z Monday (1pm), depicting the northern shortwave passing through the NE US ahead of the southern shortwave, allowing for constant suppression of the system with the heaviest snow remaining north of the area. Despite its timing and consistency issues, this CMC run handled the suppression aspect better than the GFS and ECM (image credit: PSU e-Wall).

f90The ECMWF, which was significantly north with its 2/27 12z run, going as far as depicting the heavy snow remaining north of NYC, shifted south with its 2/28 runs but not to the magnitude of the GFS, with the 0z run still depicting at least 10 inches of snow in NYC and the 12z run with over 7-8 inches. Meanwhile, the CMC gained a better handle on the timing and the strength of the vort lobe, especially with its 2/28 0z run, which resulted in a somewhat more accurate depiction of the heavy snow sliding south of the NYC area, although consistency and timing issues with the CMC’s handling of the event, along with its verification record in the medium-long range, precluded more confidence in the CMC at that time.

As noted earlier, the model guidance diverged more significantly from actual conditions at the 3/1 time frame, the critical point after which the impacts of the model errors downstream were more significant. With the 3/1 0z runs, the model guidance initialized with more accurate conditions closer to reality, allowing for a more sustained southward trend throughout most of the day and into 3/2. By this point, the model guidance more correctly depicted the stronger vort lobe than originally modeled, which then tracked south and later SE faster and further south/east, allowing for more significant lowering of heights over the northeast US. Along with a slightly slower southern shortwave than modeled earlier, this supported the northern trough outrunning the southern stream, preventing further amplification of the system while resulting in more shearing of the low while continuously shifting the precipitation shield south as the frontal boundary was able to more easily progress southward with a stronger push of anomalously cold air. By this point, the main errors in the model guidance mostly resulted from minor errors in the exact position and strength of the polar vortex and the southern shortwave.

3/1 18z run of the GFS at hour 48, on 18z Monday (1pm), depicting the continued trend towards a more suppressed storm with most of the snow remaining south of the area, despite still showing too much snow over the area (image credit: PSU e-Wall).

f48The 0z ECMWF began with a notable south shift, depicting only 4-5 inches of snow in NYC with its 0z run, while the GFS showed similar totals, with only the front end of the main wave of precipitation on Monday (3/3) producing moderate snow in the area as the precipitation shield was continuously suppressed to the south throughout the day. The 6z run of the GFS temporarily shifted back north, depicting more significant snowfall again with over 7-8 inches in NYC, but continued the southward trend with its following runs, with the 3/1 18z run only brushing the area in the northern end of the moderate snow zone, with at least 1-3 inches in the northern half of the area and 3-5 inches elsewhere. Even this was too far north; the south trend continued throughout 3/2, with both the GFS and ECMWF depicting at least 2-5 inches with the morning runs and 1-3 inches in the afternoon and evening runs. Both models never fully caught on to the magnitude of the southern trend until the event was already ongoing; the entire NYC area ended up with less than 0.10 inch of liquid-equivalent precipitation, with only a maximum of 0.8 inch of snow near Islip, NY and 0.1 inch in Central Park, NY.


Forecast Analysis: What Went Wrong?

As noted in the section above, only relatively minor differences with the handling of the polar vortex near the Hudson Bay resulted in more significant impacts downstream, differentiating between a major snowstorm with over 10-12 inches to a weaker and more suppressed system with only light to moderate accumulations. These differences presented larger than usual uncertainties leading up to the storm, and while confidence was nearly certain that precipitation would fall, the snow forecasts were especially of high uncertainty and low confidence, and were continuously adjusted downwards right until the start of the event.

The first mention of the storm was made in the 2/24 forecast discussion, a little over a week away, when the usual uncertainties regarding the handling of the southern shortwave and the upper level flow over Canada precluded a higher confidence and more detailed outlook, but with strong model and ensemble support for a widespread precipitation event. This uncertainty was further expanded upon with the 2/25 forecast discussion, which specifically noted that with the key players over the Pacific Ocean and Canada, both regions of poor upper level data sampling, additional changes were expected with confidence too low to forecast specific details except for a high probability of precipitation. The model guidance on 2/27, including the earlier CMC run, were especially consistent with a major snowstorm affecting the region; despite these differences, the 2/27 forecast discussion continued to note the uncertainties and did not issue any high confidence snow outlook, but failed to account for the possibility of a significantly more suppressed outcome.

Given the model guidance struggling to correctly handle the setup, the forecast was especially difficult. While it is always important to take the model guidance with caution, especially in the medium to long range when accuracy tends to decrease with time, this storm especially reinforced the notion that model consistency does not necessarily reflect accuracy, regardless of how many runs depicted a similar outcome. The GFS showed a stretch of nearly 10 consecutive runs of over 10 inches of snow near NYC from 2/26 through 2/28, with the ECM also fairly consistent with heavy snow near or north of NYC, but especially considering that the key features were still over the Pacific Ocean and Canada, regions known for a lack of upper level data sampling, as well as the time range of the storm still beyond 84-96 hours when the model guidance occasionally tends to have more significant errors, the modeled scenario of a 10+ inch snowstorm was not a guarantee for the area given the greater than usual uncertainty.

In this case, the main uncertainty originated from the upper level flow in Canada, which was not properly resolved until 2/28 and 3/1 when the significant south trend became evident. Until this occurred, the possibility did exist for a somewhat further south and more suppressed storm, but with not enough confidence to expect such a scenario, especially given the consistency of the ECM and GFS models which are typically considered reliable, along with frequent trends throughout the winter for even marginal upper level setups to trend more favorable for significant snowstorms in the short range, as was best displayed with the 1/21, 2/3 and 2/14 storms. Even though the shorter range forecast discussions from 2/28 onwards did highlight the main bust risk as a suppressed and weaker storm system which would lead to less snow than forecast, more emphasis should have been placed on this possibility, and as a result of this uncertainty, snow totals were lowered too slowly, with 1-4 inches still forecast on the morning of 3/2. While the model guidance is generally considered fairly reliable within the 5-day range, once in a while there are still cases where significant uncertainty exists with only minor changes in certain features leading to more significant differences downstream for the area which presents a higher than normal bust risk, as with this event where this low-end bust risk ended up verifying to its full potential.



Storm Timeline in the Northeast

Regional radar image from 7:38 AM EDT 3/3 from the National Weather Service, showing light snow brushing the NYC area with more widespread moderate snow south of central New Jersey.

radarPrecipitation affected the region in two waves; the first wave produced light precipitation over the NYC area between 3:30 and 7:00 pm on Sunday, 3/2. Despite the frontal boundary remaining south of the area, temperatures were initially slow to fall, remaining in the mid 30s to low 40s, supporting a mix of light rain, snow and sleet, with little to no measurable precipitation. Further south, heavier precipitation fell towards Virginia, Maryland and southern NJ, but was in the form of rain with temperatures in the 50s and 60s ahead of the cold front. As this wave moved out, dry conditions continued through the remainder of the evening hours.

The second and more organized wave of precipitation affected the region overnight into Monday, 3/3. Precipitation became more widespread over Maryland towards 10-11pm, spreading ENE with the northern fringe of the precipitation shield reaching the NYC area around 1-2am. Light snow showers fell as temperatures steadily dropped into the low-mid 20s, ranging from as little as isolated flurries towards SE NY, parts of northern NJ and southern CT, to as much as 1/2 inch or locally higher near NYC and parts of Long Island. Snow slowly tapered off from north to south, ending at 5am in northern locations and 8am near coastal southern locations. As the arctic air mass continued to surge further south, the frontal boundary continued to slowly slide south with the northern end of the precipitation shield reaching Philadelphia by 10am as the changeover to snow reached central-southern Virginia.



Storm Impact and Precipitation Totals

3.2.14_snowAfter a period of frequent major snowstorms earlier in the winter, this storm continued the trend for less significant snowstorms than modeled beginning with a nor’easter on 2/15 which produced less snow than forecast over the area and into parts of New England. As noted earlier, the storm was initially depicted by the model guidance as producing over 12 inches of snow, primarily over Pennsylvania into New Jersey and southern New England, before gradually trending further south and drier during the two days leading up to the onset of precipitation. In the end, snow accumulations were less significant and further south than forecast, generally peaking in the 4-8 inch range across West Virginia into northern Virginia, Maryland, Delaware and southern New Jersey, with localized totals over 8 inches in northern Virginia and eastern West Virginia.

In the NYC area, despite the forecasts leading up to the storm, still forecasting at least 1-3 inches on the day of the storm, less than 1 inch of snow ended up falling across the area. The highest observed accumulation was 0.8 inch in Islip, NY, decreasing further north and west with only 0.1 inch of snow in Central Park. At least a trace of snow was observed across most of the area.

Below is a list of selected snow reports across the area from the National Weather Service:

Northern NJ:
0.2″ – Newark, NJ

New York City:
0.6″ – JFK Airport, NY
0.4″ – LaGuardia Airport, NY
0.1″ – Central Park, NY

Long Island and South CT:
0.8″ – Islip, NY
0.6″ – Upton, NY


Leave a Reply