1. Introduction
1.1 Motivation and Objectives
Ice storms are among the most hazardous, disruptive, and costly winter weather phenomena. From a socioeconomic perspective, ice storms endanger human life and safety, undermine public infrastructure, and adversely impact local and regional economies. During the 5–9 January 1998 North American ice storm, prolonged freezing rain resulted in widespread power outages, human fatalities and injuries, and catastrophic damage to forests and agricultural operations throughout northern New York, New England, and southeastern Canada (DeGaetano 2000). Power disruption affected nearly 4.5 million people, and the region’s dairy, maple syrup, and timber industries all suffered considerable losses. Insurance claims totaled at least $175 million across New York, Vermont, New Hampshire, and Maine. Despite common public opinion that the 1998 ice storm was an unprecedented event, New York and New England have endured similar magnitude events since the early twentieth century (DeGaetano 2000; Gyakum and Roebber 2001).
More recently (2008), a severe ice storm in south-central China caused $22.3 billion in direct economic losses, 129 human fatalities, and power failure and structural damage that displaced 1.7 million people (Zhou et al. 2011). Electrical power disruption and heavy ice accretion compromised the nation’s infrastructure and jeopardized the distribution of food and other necessities. Extensive forest damage eventually led to soil erosion, landslides, insect infestation, and numerous forest fires. The Chinese ice storm exemplifies the socioeconomic and ecological consequences that can result from unsustainable practices such as concentrated industrial development and short-term growth policies (Zhou et al. 2011). Call (2010) argues that a growing dependence on electricity has also increased the magnitude and duration of ice storm-related power outages, thereby elevating societal vulnerability to costly disasters.
From a geographical perspective, ice storms are historically prevalent and destructive across the United States, particularly in the Northeast. Changnon (2003a) identified 87 ice-storm catastrophes during the 1949–2000 period that caused an estimated $16.3 billion in insured losses throughout the contiguous U.S. As illustrated by Figs. 1.1 and 1.2, the Northeast region experienced the highest frequency of ice storm catastrophes (39) and suffered the greatest insured losses (roughly $4.84 billion total; $124 million per event). More than 20 ice-storm catastrophes occurred in nine states, including North Carolina, Virginia, Maryland, Pennsylvania, New Jersey, New York, Connecticut, Rhode Island, and Massachusetts. Changnon (2003a) notes that locations along and east of the Appalachian Mountains are climatologically favorable for cold-air damming and low-level frontal development, both of which are commonly observed in association with freezing rain.
A separate 9-cool-season (1928–1937) database of ice accretion measurements taken by railroad personnel (Hay 1957) also revealed a maximum in the frequency of damaging ice-storm areas in the northeastern U.S. National Weather Service (NWS) Warning Coordination Meteorologists (WCMs) estimate that ice storm recurrence intervals range between 0.5 and 2 years throughout the NWS Eastern Region, with the shortest recurrence intervals observed across northern and interior sections (Call 2009). The geographical distributions of ice storms from Changnon (2003a) are consistent with studies on the climatological frequency of freezing rain. For instance, Houston and Changnon (2007) determined that the greatest annual average number of freezing rain hours occurs in upstate New York and New England, with an additional distinct maximum east of the southern Appalachians (Fig. 1.3).
From a meteorological perspective, ice storms present a major operational forecast challenge due to the combined influence of synoptic, mesoscale, and microphysical processes on precipitation type. Based on a 2006 survey completed by 15 WCMs in the NWS Eastern Region, most offices would issue watches 24–48 h preceding an ice storm, but only eight (53%) would issue warnings more than 24 h in advance (Call 2009). These responses convey how the difficulty in forecasting precipitation type and intensity undermines the ability to issue precise forecasts beyond 24 h. Moreover, empirical evidence suggests that operational models and forecasters often underestimate the extent and duration of freezing rain in complex winter storms.
In recent years, researchers have developed precipitation type algorithms utilizing parameters such as wet bulb temperature, surface temperature, relative humidity, ice fraction, and cold/warm layer depth (Wandishin et al. 2005). Despite our knowledge of the meteorological conditions and physical mechanisms that influence precipitation type, we cannot accurately observe nor model critical thermodynamicvariables and microphysical processes at sufficiently high spatial and temporal resolutions to precisely forecast precipitation type. Wandishin et al. (2005) nevertheless showed that ensemble forecasts incorporating various precipitation-type algorithms can increase short-range (0–48 h) forecast skill. Utility companies, emergency managers, and the general public would all benefit from improved short-to-medium range prediction of freezing rain (Call 2010). As DeGaetano et al. (2008) note, utility companies encounter many challenging decisions before an impending ice storm, and the lack of forecast tools for predicting ice accretion further complicates these decisions. Thus, adequate preparation and risk management require that accurate forecasts of icing amount and location be issued with ample lead time.
In consideration of socioeconomic and forecast issues highlighted above, we outline three primary objectives for this study. First, we seek to create long-term climatologies of freezing rain and ice storms in the northeastern U.S. These climatologies will examine the temporal and spatial variability of freezing rain and ice storms, as well as characterize ice storms based on spatial properties and relevant meteorological features. Second, we aim to identify the antecedent environments conducive to ice storms and the physical mechanisms that govern their evolution. Although freezing rain typically occurs under preferred synoptic conditions, we need to consider how mesoscale processes modify dynamically and thermodynamically forced large-scale circulations and associated quasi-geostrophic (QG) forcing on regional and local scales. Third, we hope to increase situational awareness and provide scientific insights that will ultimately improve the future prediction of ice storms. The anticipated findings will build upon existing conceptual models and help facilitate critical decisions made by operational forecasters, utility companies, and emergency managers.
1.2 Background
1.2.1 Microphysical Processes and Precipitation Type Issues
Although this study focuses on the synoptic and mesoscale dynamics associated with freezing rain, it is worth noting how microphysical processes and the vertical distribution of thermodynamic variables ultimately determine precipitation type and evolution. Bocchieri (1980) analyzed rawindsonde observations from 48 upper-air stations to objectively identify which physical parameters most influenced precipitation type. Figure 1.4 shows composite temperature and dewpoint profiles for 94 freezing rain and 127 freezing drizzle observations. The classical freezing rain sounding is characterized by a nearly saturated vertical profile, a surface-based subfreezing layer, and a distinct melting layer (T > 0°C) aloft, whereas freezing drizzle typically occurs in colder, drier environments lacking well-defined warm layers. Based on his statistical analysis, Bocchieri proposed using six critical parameters to diagnose precipitation type: 1) the mean layer temperature in the lowest 1000 m, 2) the mean layer temperature between 500 m and 2500 m, 3) the warm layer depth, 4) the warm layer area (bounded by the temperature profile and the 0°C isotherm), 5) thecold layer depth, and 6) the cold layer area (bounded by the wet-bulb temperature profile and the 0°C isotherm). Although this method effectively discriminated between liquid and frozen precipitation, it failed to accurately predict freezing precipitation. Errors likely arose from the fact that Bocchieri neglected the impacts of microphysical processes on precipitation type.
Stewart and King (1987) examined the mesoscale substructure and precipitation evolution in three midlatitude winter storms. Freezing rain occurred in a variety of environments, including warm frontal bands, bands along the cold front, and bands trailing the low center. Model simulations demonstrated that initial snowflake size, precipitation rate, and the depths of the melting and refreezing layers largely determined whether precipitation fell as freezing rain/drizzle or ice pellets. In addition, relative humidity can influence hydrometeor phase changes via diabatic effects (evaporational cooling). Their results suggested that smaller particles, lower precipitation rates, extensive (shallow) melting (refreezing) layers, and saturated conditions increase the likelihood of freezing rain versus ice pellets.
Raga et al. (1991) investigated the relationship between microphysical processes and thermodynamic and kinematic properties of the transition region in a midlatitude winter storm. Aircraft observations indicated an enhanced horizontal temperature gradient at the cold edge of the transition region, likely driven by diabatic warming associated with the freezing of cloud droplets and partially melted particles below the melting layer. The enhanced meridional temperature gradient coincided with an increase in vertical wind shear and the formation of a low-level easterly jet. Observations also revealed strong southerly flow rising above the subfreezing air at lowlevels and transporting warm, moist air into the transition region. These easterly and southerly jets correspond to the cold and warm conveyor belts, respectively,noted by Browning (1986). As illustrated by Fig. 1.5, Raga et al. (1991) developed a conceptual diagram highlighting the predominant microphysical processes, thermodynamic features, and kinematic perturbations associated with transition regions.
Szeto and Stewart (1997) evaluated the effects of melting on surface frontogenesis and related thermodynamic and kinematic fields. Transition regions are often located near low-level frontal boundaries, and microphysical processes can modify these boundaries via latent heating/cooling. Model simulations suggested that cooling by melting accelerates low-level frontogenesis, induces downdrafts and outflows that enhance convergence and ascent on both sides of the transition region, and increases vertical wind shear in the along-front and cross-front directions. These results are consistent with Szeto et al. (1988), who concluded that melting-induced cooling triggers a thermally indirect mesoscale circulation centered near the cold edge of the rain–snow boundary. Furthermore, in the presence of large vertical wind shear, cooling by melting will oppose differential temperature advection and thereby influence the evolution of the vertical temperature profile.
Zerr (1997) examined 34 soundings and described the microphysical and thermodynamic features associated with freezing rain versus ice pellets. In general, shallowmelting layers, deep refreezing layers, colder profiles, and subsaturated conditions appeared to inhibit complete melting and thus prevent freezing rain from occurring. Zerr also defined a melting parameter (βM = TmaxΔZM) and a refreezing parameter (βF = TminΔZF), where Tmax (Tmin) is the maximum (minimum) temperature in the melting (refreezing) layer, and ΔZM (ΔZF) is the depth of the melting (refreezing) layer. Refreezing parameters were large regardless of precipitation type, but melting parameters were notably smaller during ice pellet events than during freezing rain events. Most soundings were characterized by veering winds and warm advection aloft, but several soundings indicated backing winds and low-level coldadvection (likely accompanied by a cold frontal passage).
Rauber et al. (2000) compared the relative importance of warm cloud rain and melting processes during freezing precipitation events. Freezing precipitation can occur via the classical “melting process” (ice particles undergo melting in an above-freezing layer and become supercooled after passing through a surface-based subfreezing layer) or the “warm rain process” (cloud droplets undergo collision and coalescence in subfreezing air that lacks a melting layer). The authors determined that the warm rain process was likely active in about 75% of the 972 analyzed soundings,but typically resulted in freezing drizzle. On the contrary, only 25% of soundings were characterized by conditions favorable for the classical melting process (cloud top temperatures < −10°C and a warm layer aloft). Soundings of the latter type were most frequently observed during freezing rain events in the Midwest U.S. and east of the Appalachians.
Rauber et al. (2001a) assessed the effectiveness of a nondimensional parameter (Czys et al. 1996) that distinguishes between freezing rain and ice pellets. Defined as , this parameter uses values of warm layer depth, ambient temperature, particle fall speed, and critical ice particle radius to estimate the ratio of residence time in the warm layer (tres) to the time required for complete melting (tmelt). Employing a 25-year climatology of soundings during freezing rain, freezing drizzle, and ice pellet events, Rauber et al. (2001a) exposed two key issues with this parameter. First, only 306 of the 1052 soundings conformed to the vertical profile (cold cloud tops, elevated warm layers, and subfreezing surface layers) considered by Czys et al. (1996). Second, this parameter poorly discriminated between freezing rain/drizzle and ice pellets.
Theriault et al. (2010) employed a one-dimensional kinematic cloud model to simulate the behavior of hydrometeors and evaluate precipitation type sensitivity to both temperature and precipitation intensity. Differentiating between freezing rain and ice pellets is especially difficult because supercooled droplets entering the subfreezing layer may remain unfrozen, undergo refreezing, or interact with locally generated ice crystals. Their sensitivity analysis demonstrated that higher precipitation rates and larger initial snowflakes would inhibit completing melting. Slight changes (±0.5°C) in the temperature profile would also have major implications for precipitation type (warmer profiles favor complete melting and, consequently, freezing rain). Since precipitation type at the surface strongly depends on temperature, precipitation intensity, and poorly resolved microphysical processes, precise forecasts of precipitation type are a likely major weakness of operational models.
1.2.2 Climatological Perspective
Numerous studies have examined the climatological aspects of freezing rain and ice storms across North America. Branick (1997) constructed a 13-year (1982–1994) climatology of significant winter weather events impacting the contiguous U.S. He determined that 25% of all events featured icing, but only 12% met the 0.25 in (0.64 cm) minimum ice storm criterion. Approximately, 84% of all events impacted areas less than 250,000 km2, and 70% of precipitation events lasted between 6 and 24 h locally. These spatial and temporal characteristics suggest that although ice storms typically occur in preferred synoptic settings, they are predominantly mesoscale phenomena.
Bernstein (2000) investigated local and regional influences on the type of freezing precipitation (freezing rain, freezing drizzle, and ice pellets) at various sounding sites in the continental U.S. The 30-year (1961–1990) climatology revealed high freezing precipitation frequencies (> 30 h of freezing precipitation per year) along and east of the Appalachian Mountains, where freezing rain represented 30–50% of all freezing precipitation. In the lee of the southern Appalachians, north-northeasterly surface winds and a meridional pressure dipole (higher pressure to the north and lower pressure to the south) are consistent with cold-air damming. Across New England, wind directions varying between northeasterly and northwesterly suggest the importance of both cold-air damming and frontal boundaries enhanced by land–ocean thermal contrasts. Cold cloud-top temperatures (< −10°C), well-defined melting layers, and relatively shallow subfreezing layers in both regions generally support freezing rain formation via the melting process.
Focusing on the 1976–1990 period, Cortinas (2000) evaluated the spatial and temporal distribution of freezing rain throughout the Great Lakes Region, and identified synoptic-scale features characteristic of freezing rain events. Freezing rain occurred more frequently across the eastern Great Lakes than across the western Great Lakes, with regional maxima over interior sections of southeastern Ontario, west-central Pennsylvania, and upstate New York, and regional minima along the western lakeshores. These spatial variations are likely influenced by: 1) the frequency of surface cyclone tracks and associated frontal boundaries, 2) availability of Atlantic moisture, and 3) local and regional topography. Moreover, the notable decrease in frequency near the lakeshores implies that large bodies of water can modify the thermodynamic environment and ultimately reduce the likelihood of freezing rain. Freezing rain reports exhibited significant diurnal variability, with the highest (lowest) frequencies observed during the morning (afternoon). Most events were short-lived and accompanied by a transient surface cyclone tracking east-northeast through the Midwest, anomalously high pressure over eastern North America, and easterly surface winds.
Rauber et al. (2001b) discussed common synoptic-scale features associated with 411 freezing precipitation events east of the Rocky Mountains during the 1970–1994 period. Based on the locations and orientations of surface cyclones, anticyclones, and frontal boundaries, the authors defined seven archetypical patterns conducive to freezing precipitation. Freezing precipitation occurred most frequently on the poleward side of a warm front or within the occluded region of a surface cyclone. Cold-air damming (east of the southern and central Appalachians) and cold-air trapping (along the northern Appalachians) were also present in a substantial number of freezing precipitation events. Surface winds varied between easterly and northwesterly, whereas winds aloft were predominantly southwesterly.
Adopting the same period (1976–1990) as Cortinas (2000), Robbins and Cortinas (2002) analyzed the local and synoptic environments associated with freezing rain events in different regions. Approximately 68% of freezing rain events in the Piedmont region (interior North Carolina and Virginia) were associated with cold-air damming east of the Appalachians. Low-level northeasterly winds and coldadvection established a thermal trough and pressure ridge at the surface, while moist, southerly flow and warm advection produced a warm layer aloft. Most freezing rain events in the Allegheny–Catskill region (west-central Pennsylvania and east-central New York) were associated with a surface cyclone passing to the west, minimal thermal advection at the surface, and pronounced warm advection aloft via southwesterly flow. The composite surface wind field also reveals a second area of cyclonic vorticity near the mid-Atlantic coast, and thereby suggests that some events may occur in conjunction with Miller Type A (cyclogenesis near the Gulf Coast) or Miller Type B (secondary cyclogenesis along the East Coast) cyclone tracks (Miller 1946). Roughly 57% of all freezing rain events occurred along a stationary or quasi-stationary warm front, and 33% occurred within the northern, northwestern, or western sector of surface cyclones.