6. Discussion and Conclusions

6.1 Climatological Aspects

In Section 3, we examined the climatological aspects of freezing rain and ice storms in the northeastern U.S. Section 6.1 will summarize key results, compare our findings with those of previous studies, and further discuss which factors contributed to the observed temporal and spatial variability of freezing rain and ice storms.
6.1.1 Freezing Rain

From a climatological perspective, freezing rain is quite common throughout the eastern U.S. and southeastern Canada. Sections of the Blue Ridge Mountains and Piedmont, the northern Appalachian Plateau, upstate New York and New England, and the Ottawa and St. Lawrence River Valleys experience the highest frequencies of both freezing rain hours (> 15 h per season) and significant freezing rain events (> 1 per season). These frequencies decrease notably toward the Atlantic coast, the Cumberland Plateau, and the Great Lakes (Figs. 3.1 and 3.6). Whereas the geographical distribution of freezing rain hours and significant freezing rain events appears related to prominent topographic features and land–water boundaries, the distribution of total freezing rain events varies primarily with latitude. On average, areas north of a line extending from the Ohio Valley to southern New England experience at least four freezing rain events per season. South of this line, the frequency of freezing rain events steadily decreases with decreasing latitude (Fig. 3.5). Although freezing rain events occur relatively infrequently across the southeastern U.S., these events have longer mean durations and are thus more likely to produce damaging ice accumulations than freezing rain events elsewhere (Figs. 3.7 and 3.8). The temporal distribution of freezing rain hours exhibits both seasonal and interannual variability (Figs. 3.2 and 3.3.). Throughout much of the domain, a majority of freezing rain hours are observed during the Dec–Jan period, suggesting that freezing rain is climatologically preferred in late fall/early winter.

The geographical distribution of freezing rain hours in Fig. 3.1 is consistent with previous studies, including Bernstein (2000), Cortinas (2000), and Cortinas et al. (2004). Spatial variations in the climatological frequency of freezing rain are strongly influenced by local and regional topography, proximity to large bodies of water, and extratropical cyclone activity. For instance, cold-air damming east of the Appalachians (Bernstein 2000), cold-air trapping in protected valleys (Robbins and Cortinas 2002), and orographic channeling within major river valleys (Carrera et al. 2009), can firmly entrench subfreezing air at the surface and thereby maintain vertical temperature profiles conducive to freezing rain. While the moderating effect of maritime air generally reduces the occurrence of freezing rain near major coastlines and lakeshores (Changnon 2003b), large bodies of water (e.g., the Atlantic Ocean) also play a key role in the transport of warm, moist air aloft, a necessary ingredient for freezing rain (Cortinas 2000). Bernstein (2000) and Rauber et al. (2001b) found that freezing rain often occurs on the cold side of warm or stationary fronts, within the northwestern or northeastern quadrants of surface cyclones, and in regions of cold-air damming. Given the high frequency of cyclogenesis and cold-air damming in the eastern U.S., locations along and east of the Appalachians are climatologically predisposed to synoptic weather patterns associated with freezing rain. Changnon and Karl (2003) also demonstrated that the frequency of freezing rain days varies on interannual to decadal time scales and exhibits strong seasonality, with the Dec–Jan period comprising more than 50% of all freezing rain days throughout much of the eastern U.S. As Bosart (1975) and Bell and Bosart (1988) suggest, enhanced baroclinicity along the East Coast in December and January creates a favorable environment for coastal frontogenesis and cold-air damming, two processes that increase the likelihood of freezing rain over interior sections.
6.1.2 Ice Storms

During the 1993–2010 period, 137 ice storms impacted the 14 CWAs in our domain, yielding an average of approximately eight ice storms per season. The temporal and spatial distributions of ice storms are generally similar to those of freezing rain hours. Ice storm frequency varies on interannual time scales and exhibits strong seasonal dependence, with a significantly greater number of ice storms during the Dec–Jan period than during the Feb–Mar period (Fig. 3.10). Ice storms are quite common across the interior Northeast, particularly over elevated terrain, along prominent mountain ranges, and within protected valleys, but occur rather infrequently near the Atlantic coast and the Great Lakes (Fig. 3.11). Portions of the Appalachian Plateau and Valley-and-Ridge zones, the Lehigh Valley, the Berkshires, the Lake George–Saratoga region, and Down East Maine are especially vulnerable to ice storms (≥ 1 per season). More than 80% of all ice storms were classified as local, regional, or subsynoptic events, whereas less than 20% qualified as synoptic events (Fig 3.12). Furthermore, the number of ice storms is inversely related to the number of CWAs impacted (Fig. 3.13). Ice storms in the northeastern U.S. are predominantly associated with Type G (65 events), Type BC (30 events), and Type EF (25 events) synoptic weather patterns (Fig. 3.14). These three synoptic patterns accounted for roughly 87.6% of all ice storms, while the remaining 12.4% of ice storms were associated with Type A, Type D, or unclassifiable synoptic patterns. Northern portions of the domain were primarily impacted by Type BC and Type G events, while the southernmost areas were nearly exclusively affected by Type EF and Type G events.

Although existing literature on the geographical and temporal distribution of ice storms is very limited, several studies have analyzed the spatial characteristics of significant icing events and investigated the synoptic environments commonly associated with freezing rain. While freezing rain generally occurs under preferred synoptic-scale conditions, the distribution of ice storms by spatial coverage suggests that ice storms are predominantly mesoscale phenomena. According to Changnon (2003a), 16 of the 39 ice storm catastrophes impacting the Northeast climate region during the 1949–2000 period occurred exclusively within this region. Moreover, Branick (1997) determined that severe winter weather events (including ice storms) typically affect areas less than 250,000 km². The findings from these studies suggest that mesoscale topographical features and thermal boundaries largely determine which locations experience ice storm conditions. Therefore, it is crucial that we consider important linkages between antecedent synoptic-scale environments and mesoscale dynamical processes. Consistent with the results of Rauber et al. (2001b), ice storms in the northeastern U.S. most frequently occur in regions of cold-air damming or cold-air trapping along and east of the Appalachians (Type EF and Type G patterns), or on the cold side of warm, stationary, or occluded fronts associated with surface cyclones (Type BC pattern). Changnon (2003a) also determined that southwest-to-northeast oriented ice storm areas (characteristic of Type BC events) were common throughout the Ohio Valley and interior Northeast, while south-southwest-to-north-northeast oriented ice storm areas (indicative of Type EF and Type G events) were common along and east of the Appalachians.
6.2 Composite Analysis

In Section 4, we presented a composite analysis of Type G and Type BC ice storms impacting the ALY CWA. Section 6.2 will discuss the synoptic-scale precursors and synoptic–mesoscale linkages commonly associated with Type G and Type BC ice storms. Conceptual diagrams highlighting the key synoptic ingredients and dynamical processes characteristic of each event type will be introduced in Section 6.4.

As previously suggested by Rauber et al. (2001b), Type G events are typically preceded by a cold air intrusion from interior Canada. The associated 1000–500-hPa thickness trough is initially co-located with a 300-hPa jet maximum and an anomalous 500-hPa trough, whose juxtaposition with an amplifying 500-hPa ridge over the eastern U.S. establishes confluent flow throughout central and eastern North America (Figs. 4.1a and 4.3a). Following the cold air intrusion, a surface anticyclone approaches northern New England from the west, and a surface trough develops downstream of a second 500-hPa trough over the Rocky Mountains (Fig. 4.3b). Northeasterly flow associated with the anticyclone allows a cP air mass to become firmly established within the boundary layer east of the Appalachians. Meanwhile, low- to midlevel southwesterly flow ahead of the surface trough provides warm advection aloft and northeastward moisture transport throughout the Mississippi and Ohio Valleys (Figs. 4.2a and 4.2b). This poleward surge of warm, moist air from the Gulf of Mexico eventually overspreads the northeastern U.S. (Fig. 4.2c). Robbins and Cortinas (2002) note that warm advection aloft is crucial for QG ascent and the formation of an elevated melting layer during freezing rain events in the Allegheny–Catskill region. In the presence of a southwest-to-northeast oriented synoptic-scale pressure gradient, flow blocking by the Appalachians results in a narrow pressure ridge and thermal trough along the eastern slopes (Fig. 4.3b). While this cold-air damming signal is a prominent feature 24 h prior to Type G events, given the transitory nature of the anticyclone, the cold-air damming is relatively short-lived and most closely resembles the “hybrid” type discussed by Bailey et al. (2003).

As the anticyclone drifts offshore, the developing cyclone intensifies and tracks from the Mississippi Valley toward the eastern Great Lakes (Fig. 4.3c). Consistent with our finding that Type G events comprised nearly 70% of ice storms in the ALY CWA, Robbins and Cortinas (2002) determined that freezing rain events in the Allegheny–Catskill region are primarily associated with surface cyclones tracking west of the Appalachians. Meanwhile, the continued amplification and eastward propagation of the 500-hPa ridge enhances the upper-level confluence over southeastern Canada, placing upstate New York near the equatorward jet entrance region and thus in a favorable location for QG ascent (Figs. 4.1b and 4.1c). As expected, an analysis of the Q-vector forcing reveals QG forcing for ascent (Q-vector convergence) over upstate New York (Fig. 4.4). Moreover, the orientation of the Q-vectors with respect to the 700-hPa isotherms (pointing from colder air to warmer air) is indicative of low- to midlevel frontogenesis. A vertical cross section through the jet entrance region depicts low-level frontogenesis south of Albany, NY, and deep ascent associated with a broad, thermally direct ageostrophic transverse circulation (Fig. 4.5). The horizontal branch of this circulation is dominated by northerly ageostrophic winds below 925 hPa. Northerly ageostrophic winds extend equatorward to the warm front, driving near-surface ageostrophic cold advection over Albany, NY, and helping maintain a vertical temperature profile conducive to freezing rain.
6.2.2 Type BC Events

The synoptic-scale precursors and QG forcing associated with Type BC events are quite different from those associated with Type G events. Unlike Type G events, Type BC events are characterized by a relative absence of cold air throughout much of the lower and middle troposphere. During the 48 h preceding Type BC ice storms, 1000–500-hPa thickness values consistently remain near or above 540 dam across the northeastern U.S. Moreover, the 300-hPa jet maximum over the Great Lakes is initially co-located with a thickness ridge rather than a thickness trough (Fig. 4.8a). The large-scale circulation exhibits a split-flow regime consisting of two distinct 500-hPa short-wave troughs over the southern Rockies and south-central Canada, and an anomalous 500-hPa ridge near the East Coast (Fig. 4.6a). A surface anticyclone emerges behind the northern 500-hPa trough, while a surface trough begins to undergo cyclogenesis downstream of the southern 500-hPa trough (Fig. 4.8b). This synoptic pattern closely resembles the Shapiro–Keyser model, whereby cyclogenesis occurs within a region of confluent background flow and results in the formation of a prominent and zonally extensive warm front (Schultz et al. 1998). Ahead of the southern 500-hPa trough, persistent low- to midlevel southwesterly flow enables prolonged transport of warm, moist air aloft across the eastern U.S. (Figs. 4.7a and 4.7b). However, the orientation of the 850–700-hPa wind vectors relative to the 850–700-hPa 0°C isotherm suggests that the synoptic-scale warm advection is notably weaker (albeit longer-lived) prior to Type BC events than prior to Type G events. Southerly return flow associated with the subtropical anticyclone over the western Atlantic likely enhances the poleward surge of warm, moist air along the East Coast (Fig. 4.8b).

As time progresses, the surface cyclone intensifies and tracks northeastward toward the mid-Atlantic coast (Fig. 4.8c). The presence of a well-defined warm front south of Albany, NY, suggests that frontal overrunning provides an important mechanism for freezing rain during Type BC events (Rauber et al. 2001b). Moreover, the juxtaposition of the deepening cyclone with the continental anticyclone implies confluent flow near the warm front, and thus a strengthening of the surface baroclinicity. As the northern 500-hPa trough and the 500-hPa ridge amplify and phase with each other, the meridional geopotential height gradient and upper-level confluence increase dramatically over southern Quebec (Figs. 4.6b and 4.6c). Meanwhile, the southern 500-hPa trough rapidly progresses eastward and becomes negatively tilted, and the low- to midlevel warm advection and poleward moisture transport intensify across eastern New York and New England (Fig. 4.7c). A second 300-hPa jet maximum forms along the East Coast, thereby establishing a coupled jet configuration that fosters surface cyclogenesis and QG ascent (Fig. 4.8c). Similar to Type G events, Type BC events are associated with QG forcing for ascent over upstate New York and low- to midlevel frontogenesis near Albany, NY (Fig. 4.9). However, based on the relative magnitude of the Q-vectors, the implied ascent and frontogenesis are considerably stronger during Type BC events. This inference is confirmed by a vertical cross section centered on Albany, NY, which reveals an intense, meridionally confined thermally direct ageostrophic transverse circulation that tilts poleward with height following a region of sloping frontogenesis (Fig. 4.10). Strong northerly ageostrophic winds below 925 hPa reinforce cold air on the poleward side of the warm front, while sloped ascent above the warm front implies overrunning precipitation at Albany, NY.
6.3 Case Studies

In Section 5, we presented a case study analysis of the 3–4 Jan 1999 (Type G) and 11–12 Dec 2008 (Type BC) ice storms. Section 6.3 will briefly discuss how the evolution of the synoptic-scale circulation, thermal boundaries, and moisture transport established favorable conditions for prolonged freezing rain within the ice storm regions. Additionally, this section will evaluate the role of mesoscale dynamical processes (frontogenesis and terrain–flow interactions) and the local thermodynamic environment (vertical distribution of temperature, moisture, and wind) in determining which areas experienced ice storm conditions.

The 3–4 Jan 1999 ice storm was preceded by cold-air damming east of the Appalachians following a cold-air outbreak throughout eastern North America. As surface cyclogenesis unfolded over the Midwest U.S., low- to midlevel southerly flow ahead of the developing cyclone provided warm advection and poleward moisture transport west of the Appalachians. Meanwhile, cold-air damming preserved the cP air mass over the mid-Atlantic and New England states by delaying the eastward progression of warm, moist air across the Appalachians (Figs. 5.2–5.5). Although the synoptic-scale pressure gradient was not especially conducive to cold-air damming, land–ocean frictional differences and flow blocking by the Appalachians forced the surface winds to turn east-northeasterly, thereby reinforcing subfreezing surface air along the eastern slopes (Figs. 5.10 and 5.11). Eventually, the surface warm front crossed the Appalachians, and cold-air damming rapidly dissipated east of the southern and central Appalachians. Furthermore, enhanced baroclinicity near the mid-Atlantic coast helped fuel secondary cyclogenesis along the existing warm frontal boundary, southeast of the parent low (Fig. 5.6). In response to strengthening low- to midlevel southerly flow east of the Appalachians, poleward moisture transport and warm advection increased dramatically over the mid-Atlantic and New England states (Fig. 5.7). Despite significant warming throughout the lower and middle troposphere, the abundance of very cold air near the region of surface high pressure helped prolong subfreezing surface temperatures over the interior Northeast. Moreover, the southwest-to-northeast pressure gradient between the coastal cyclone and the anticyclone east of Maine reinforced cold-air damming along the eastern slopes of the northern Appalachians and helped sustain the cP air mass across northeastern New England (Figs. 5.12 and 5.13).

Sounding profiles from Pittsburgh, PA, and Albany, NY, were characterized by well-defined surface-based subfreezing layers, pronounced low-level inversions between 925 hPa and 850 hPa, and elevated melting layers (Figs. 5.14 and 5.15). At Pittsburgh, PA, the presence of dry air aloft implies evaporational cooling within the warm layer, but the magnitude and depth of this warm layer also suggests that any hydrometeors reaching the inversion would have melted completely and subsequently remained unfrozen. In comparison to the Pittsburgh, PA, profile, the Albany, NY, profile is colder and has a significantly larger (smaller) cold-layer (warm-layer) depth. However, Robbins and Cortinas (2002) found that the cold-layer properties are often quite pronounced during instances of freezing rain at Albany, NY. Air parcels arriving within the subfreezing layer (500 m AGL) and near the top of the inversion (1500 m AGL) traveled southeastward from interior Canada and turned anticyclonically upon reaching the East Coast (Figs. 5.16 and 5.17). Meanwhile, air parcels arriving above the inversion (3000 m AGL) originated over interior North America and descended into the boundary layer as they approached the Southeast coast. These parcels then remained within the boundary layer for 48–60 h and likely underwent heating and moistening before turning northward and ascending into the ice storm region. The low-level wind profile at both locations exhibited significant speed shear, with very strong southerly winds at 850 hPa denoting the presence of a low-level jet and warm conveyor belt. Additionally, veering winds below 850 hPa were indicative of low-level warm advection over Pittsburgh, PA.
6.3.2 11–12 Dec 2008 Ice Storm

The initial stages of the 11–12 Dec 2008 ice storm featured an antecedent low-level baroclinic zone between the Ohio Valley and the Canadian Maritimes, with 1000–500-hPa thickness values exceeding 540 dam indicating the early presence of warm air aloft throughout the northeastern U.S. While surface cyclogenesis occurred near the Gulf Coast, a broad region of low- to midlevel southwesterly flow extending northeastward from the Gulf of Mexico enabled the poleward transport of very warm, moist air east of the Appalachians. A separate region of southerly flow within the western quadrant of a subtropical anticyclone suggests that the western Atlantic also served as an important moisture source during the 11–12 Dec 2008 ice storm (Figs. 5.20–5.23). As the surface cyclone intensified and tracked northeastward, low- to midlevel poleward moisture transport and warm advection intensified ahead of the cold front and across the warm front (Figs. 5.24 and 5.25). Meanwhile, the strengthening pressure gradient between the deepening cyclone and an anticyclone over southeastern Quebec resulted in pressure-driven channeling down the St. Lawrence, Champlain, and Hudson Valleys, and flow blocking along the eastern slopes of the northern Appalachians. Northerly and northeasterly surface winds throughout the ice storm region reinforced subfreezing surface air on the poleward side of the warm front. Furthermore, confluent flow across the warm front (northeasterly winds to the north; southerly winds to the south) contributed to surface frontogenesis along the mid-Atlantic and southern New England coasts (Figs. 5.28 and 5.29). Roebber and Gyakum (2003) and Ressler et al. (2012) previously showed that orographically enhanced frontogenesis can play a critical role in prolonging freezing rain events within the St. Lawrence and Champlain River Valleys.

Sounding profiles from Albany, NY, and Gray, ME, were characterized by surface-based subfreezing layers, elevated melting layers, and saturated conditions throughout the lower and middle troposphere (Figs. 5.32 and 5.33). The vertical distribution of temperature and moisture at both locations clearly supported freezing rain formation via the classical melting process (Bernstein 2000). In particular, the Gray, ME, profile featured a relatively shallow (deep) cold (warm) layer, and thus closely resembled the idealized freezing rain sounding described by Bocchieri (1980). Air parcels arriving within the subfreezing layer (500 m AGL) originated over interior Canada (cP air mass) and gradually descended into the boundary layer before turning anticyclonically across New England (Figs. 5.34 and 5.35). Meanwhile, air parcels arriving above the inversion (3000 m AGL) had extensive contact with the western Atlantic Ocean (mT air mass) before undergoing rapid ascent as they approached the ice storm region. Gyakum and Roebber (2001) demonstrated that heavy freezing rain during the 5–9 Jan 1998 ice storm was associated with air parcels that experienced substantial diabatic heating within the boundary layer over the subtropical Atlantic. The low-level wind profiles (north-northeasterly winds near the surface and southerly winds at 850 hPa) at both locations closely matched the vertical distribution of wind typically observed during freezing rain events (Zerr 1997; Rauber et al. 2001b; Robbins and Cortinas 2002). Consistent with previous case studies (e.g., Rauber et al. 1994), veering winds between 1000 hPa and 850 hPa also suggest that warm advection aloft was critical in building the elevated melting layer on the poleward side of the surface warm front.
6.4 Conceptual Models

6.4.1 Type G Events

Figure 6.1 illustrates the synoptic environment characteristic of Type G events in the ALY CWA. The synoptic-scale circulation pattern features a 500-hPa trough over the Midwest and a 500-hPa ridge over extreme southeastern Canada. A 300-hPa jet maximum is co-located with the northern periphery of the 500-hPa ridge, placing upstate New York near the equatorward jet entrance region and thus in a favorable location for QG ascent and low-level frontogenetical forcing. Low- to midlevel warm advection driven by southwesterly flow downstream of the 500-hPa trough builds a thermal ridge aloft, with 850–700-hPa layer-averaged temperatures exceeding 0°C across sections of upstate New York and New England. Moreover, the region of deep southwesterly flow along the East Coast enhances the poleward transport of moisture-laden air (PW > 25 mm) from the Gulf of Mexico. In advance of the surface cyclone approaching the eastern Great Lakes, cold-air damming along the eastern slopes of the Appalachians delays the northeastward progression of the surface warm front. The cold-air damming is represented by a highly distorted warm front, with enhanced baroclinicity on either side of the damming region. Meanwhile, the surface anticyclone, which plays a crucial role in establishing cold-air damming prior to Type G events, rapidly exits the Canadian Maritimes.

Figure 6.2 illustrates the synoptic environment characteristic of Type BC events in the ALY CWA. The synoptic-scale circulation pattern features a negatively tilted 500-hPa short-wave trough over the Ohio Valley and a broad 500-hPa ridge over the North Atlantic Ocean. As in the case of Type G events, upstate New York lies within the equatorward entrance region of a 300-hPa jet maximum, but unlike Type G events, Type BC events often feature a second jet maximum along the Southeast coast. This coupled jet configuration enhances surface cyclogenesis over the mid-Atlantic region and dynamical forcing for ascent and low-level frontogenesis near Albany, NY. Similar to Type G events, strong southwesterly winds in the 850–700-hPa layer provide low- to midlevel warm advection and poleward moisture transport throughout the region. However, during Type BC events, the 850–700-hPa thermal ridge is broader and the main PW axis (PW > 25 mm) more closely resembles the comma cloud structure associated with mature extratropical cyclones. High PW values are concentrated along the East Coast (corresponding to the warm conveyor belt) and extend northeastward along the warm front. Additionally, synoptic-scale southerly return flow associated with the subtropical anticyclone implies that the western Atlantic serves as a primary moisture source during Type BC events. North of the approaching surface cyclone, an area of surface high pressure over southeastern Canada allows cold air to bleed southward across the northeastern U.S. Northeasterly flow on the poleward side of the warm front and southerly flow on the equatorward side strengthens the surface confluence and enhances frontogenesis along the mid-Atlantic and southern New England coasts.

Now that a detailed climatology of significant icing events in the northeastern U.S. exists, it would be advantageous to maintain an up-to-date ice storm database for future use. In particular, we recommend periodically expanding upon the database to include ice storms occurring after the 1993–2010 period, as well as ice storms affecting regions beyond the climatology domain. An ongoing effort involves utilizing geographic information systems (GIS) software to create maps highlighting the counties impacted by each ice storm. These maps will allow us to further evaluate the spatial characteristics of ice storm regions, and subsequently compare our findings with Changnon’s (2003a) analysis of damaging ice storm areas during the 1928–1937 period. As previously noted, Storm Data is sensitive to discrepancies in population density among different WFOs. Thus, adopting additional techniques that correct for population biases will likely bolster the reliability of our preliminary climatology. It may also prove beneficial to investigate possible connections between event type and the spatial properties of ice storms. Insights gained from these observations may help operational meteorologists forecast the location, scope, and severity of an icing event based on the prevailing synoptic weather pattern.

Figure 6.2: Conceptual model illustrating the synoptic-scale environment characteristic of Type BC events in the ALY CWA. Plotted features include the 500-hPa geopotential height (solid black contours), 300-hPa wind speed (shaded; J1 denotes the primary jet maximum and J2 denotes the secondary jet maximum), precipitable water (values > 25 mm shaded in green), 850–700-hPa 0°C isotherm (dashed red contour), the surface 0°C isotherm (dashed magenta contour), and 850–700-hPa wind vector (speeds > 20 m s⁻¹). Concentrated regions of QG forcing for ascent and low-level (925-hPa) frontogenesis are indicated by the closed orange and purple contours, respectively. The “H” and “L” symbols represent relative maxima and minima in sea level pressure, respectively, and the warm front symbol denotes the location of the surface baroclinic zone.