1. Introduction 1.1 General Purpose The purpose of this thesis is to document the spatial structure and temporal evolution of heavy rainstorms that sometimes develop well poleward of landfalling or near-coastal-tracking Atlantic basin tropical cyclones (TCs) (refer to the appendix for a list of acronyms used in this thesis). A statistical and composite climatology of these predecessor rain events (PREs) was established using a database of 47 cases occurring downstream of 21 TCs between 1998 and 2006. The PREs from three TCs—two from within the climatological period and one from before—were chosen for detailed synoptic-scale and mesoscale study based on their capacity to illustrate significant findings from the climatology as well as physical mechanisms crucial to their formation. A null case also was selected so the environment downstream of the TC could be contrasted with those cases that spawned PREs. The aim of this type of approach is to continue the mission of the warm-season project of the Collaborative Science, Technology, and Applied Research (CSTAR) program under which the research was conducted—namely, to improve the prediction of heavy precipitation events over the northeastern U.S. through integration of academic research into useful operational frameworks. In order to provide an appropriate background for this particular project, the remainder of this chapter will document previous research on the climatological and physical aspects of warm-season extreme rain events (EREs), extratropical transition (ET), and the precipitation distribution occurring both directly with, and well downstream of, landfalling and transitioning TCs. 1 1.2 Motivation and Overview It has been documented recently that inland freshwater flooding is the main weather-related cause of death when a TC makes landfall in the U.S. (Rappaport 2000). Complicating this issue is the observation that numerous recent landfalling and near-coastal-tracking TCs have been preceded by coherent areas of heavy rainfall, some of which can produce their own flash floods (M. Jurewicz and D. Vallee 2006, personal communication). Area forecast discussions (AFDs) from the National Weather Service (NWS) weather forecast office (WFO) in Taunton, MA, prior to the onset of PREs ahead of Katrina (2005) and Ophelia (2005) suggest that these events pose a critical forecast challenge. Timing and placement of such unexpected heavy rain can have disastrous effects, as occurred when the PRE ahead of Frances (2004) deluged the New York City subway system during rush hour on the morning of 8 September (Luo et al. 2004). It is hypothesized that this lack of anticipation—and thus of preparation—arises from a number of factors, including 1) the inability of current numerical guidance to accurately depict these organized areas of heavy rain, 2) the possibility that subtle features not easily detected may help focus heavy rainfall in the moist tropical air mass ahead of a TC, and 3) the lack of attention an operational forecaster may give toward predicting antecedent precipitation when the rainfall evolution associated with the TC itself is perceived to be the main concern. Consequently, a primary motivating factor of this research is to increase awareness in the meteorological community of the potential for distinct areas of heavy rainfall to occur far from the centers of TCs. 2 Significant progress has been made in our understanding of the ET process, but the operational forecaster is still faced with a number of challenges when trying to predict the sensible weather that may result when a TC approaches the midlatitudes. Jones et al. (2003) explains that these difficulties arise mainly because one set of numerical prediction models is currently used in the midlatitudes, while another set is used in the tropics. Neither of these sets successfully captures the interaction of the two regimes with any consistency. Recent CSTAR-related research, however, has successfully improved the forecast of precipitation accompanying landfalling and transitioning TCs (D. Vallee 2006, personal communication) through the creation of conceptual models (DeLuca 2004; Srock 2005). These works, among others (Bosart and Carr 1978, hereafter BC78; Jones et al. 2003), have noted that heavy rain can form when the large-scale environmental flow allows tropical moisture to stream far ahead of the TC itself. Only BC78, in detailing the excessive rain centered on Wellsville, NY, far ahead of Agnes (1972), have provided a thorough investigation of this type of event in the refereed literature to date (see section 1.4.3). After consulting with a number of current operational meteorologists, the author believes that many are aware of a small number of PREs, particularly if they experienced them directly. However, no known comprehensive study of the overall properties of PREs, nor of the mechanisms that work to bring them about, has been conducted. This gap in understanding can have serious implications both for areas that are directly affected by a TC and those that are not. The current study will attempt to bridge this gap by providing forecasters with new 3 climatological perspectives of PREs and detailed analyses of several cases. With a reported increase in heavy precipitation events in the last century over the U.S. (Karl and Knight 1998), and the suggestion that Atlantic TC activity will remain high for the foreseeable future due to a favorable combination of warm sea surface temperatures and reduced vertical wind shear (Goldenberg et al. 2001), an extensive study of PREs needs to be undertaken now so forecasters can better anticipate their occurrence. 1.3 Significant Warm-Season EREs 1.3.1 Overview Since PREs are separate from the main rain shields associated with TCs, a relevant background literature review should begin with a general discussion of significant warm-season EREs. These events are associated commonly with flash floods, which produce the most fatalities of any element of convective storms (Doswell et al. 1996, hereafter D96; NOAA 2006) and cause more property damage than all other weather-related phenomena in an average year (Fritsch et al. 1998). A significant portion of these statistics likely can be attributed to the lag in forecasting and warning of flash floods in comparison to tornadoes (D96). Relatively poor predictability of warm-season in comparison to cool-season precipitation (Funk 1991; Fritsch et al. 1998) arises from the predominance of forcing for ascent by convective processes (Olson et al. 1995), which often takes place near mesoscale convergence zones (Heideman and Fritsch 1988; Koch and Ray 1996), as opposed to synoptic-scale forcing, which 4 is better portrayed by current numerical models. Flash floods are particularly difficult to forecast due to the complex interaction of meteorological and hydrological factors (D96). The occurrence and inadequate prediction and warning of such events in turn has marked societal and economic impacts on industries that rely heavily on accurate forecasts of significant precipitation, such as government, business, agriculture, entertainment, and transportation (Fritsch et al. 1998). 1.3.2 Climatologies Heavy precipitation and flash flood climatologies have been constructed for decades on both the national (La Rue and Younkin 1963; Maddox et al. 1979, hereafter M79; Karl and Knight 1998; Brooks and Stensrud 2000; Schumacher and Johnson 2005; Schumacher and Johnson 2006) and regional (Winkler 1988; Houze et al. 1990; Giordano and Fritsch 1991; Bradley and Smith 1994; LaPenta et al. 1995; Konrad 1997; Junker et al. 1999; Moore et al. 2003) levels. Recently, Schumacher and Johnson (2006) classified an event as extreme if the total rainfall received by a particular location exceeded its 50-yr recurrence amount, thus limiting the study only to those events considered extreme for a given region. Using this scheme, Schumacher and Johnson (2006) found that EREs were most common within a southwest–northeast band from the southern Plains through the Ohio Valley and Great Lakes states during the period 1999–2003 (see their Fig. 4). ERE occurrence in the southern part of this band peaked in late spring with the predominance of mesoscale convective systems (MCSs), 5 although there was a secondary maximum in early fall when significant synopticscale forcing was more likely (M79; Houze et al. 1990; Bradley and Smith 1994; Schumacher and Johnson 2006). The vast majority of EREs in the north, however, were due to MCSs developing in weakly forced environments (Bradley and Smith 1994) during midsummer, which corresponds well with the overall national peak in ERE occurrence (M79; Giordano and Fritsch 1991; Brooks and Stensrud 2000). The East Coast saw significantly fewer EREs during the period, partially because the 50-yr recurrence amount for the East Coast is significantly higher than that of areas farther inland (Fig. 1.1), indicating that heavy rainfall is more common there overall. In particular, Fig. 1.1 shows that localized maxima in recurrence amount are located along the Southeast U.S. coastlines bordering the Atlantic Ocean and Gulf of Mexico, as well as the eastern slopes of the southern and central Appalachians. EREs most commonly occur in this portion of the U.S. in September due to the direct impacts of TCs (La Rue and Younkin 1963; LaPenta et al. 1995; Brooks and Stensrud 2000) or their interactions with synoptic-scale disturbances (LaPenta et al. 1995). Despite comprising less than ten percent of all EREs east of the Rockies, these tropical-related events produce the largest rainfall totals overall (Schumacher and Johnson 2006). 1.3.3 Meteorological Ingredients 1.3.3.1 General Considerations 6 Numerous studies have discussed the utility of ingredients-based methodologies to augment the output from numerical models when trying to predict summertime EREs (e.g., Funk 1991; D96). D96 state that the largest precipitation totals occur where the rainfall rate is highest for the longest amount of time, thereby reducing this forecast problem to two main variables. Difficulties arise, however, when trying to forecast the different factors that affect rainfall rate and duration. For example, D96 state that high rainfall rates are generally produced by a combination of rapid upward vertical motion, significant moisture through a deep layer, and high precipitation efficiency. Long duration at any one location is favored by slow system movement and upstream cell development The remainder of this section will document previous work on the meteorological processes that cause the aforementioned basic ingredients to combine in midlatitudes, thus creating an environment favorable for EREs in the warm season. The same factors combine to produce heavy precipitation in tropical settings, but the manner in which they are brought together may differ substantially from typical midlatitude situations (D96). 1.3.3.2 Favorable Synoptic-Scale Patterns The high rainfall rates and slow movement contributing to flash floods east of the Rockies described by D96 are usually associated with the elements of at least one of three characteristic large-scale patterns defined by M79. The first is called a synoptic flash flood pattern, in which convection breaks out in the warm, moist air mass ahead of a quasi-stationary surface front often oriented from 7 south-southwest to north-northeast (Fig. 1.2a). The slowing down or stalling of a front with this orientation happens with some regularity along the Appalachians (O’Handley and Bosart 1996), especially when the winds aloft have a significant component parallel to the front (M79) and the mountain range (Schumacher et al. 1996), allowing convection to organize into a linear structure along the front (Dial and Racy 2004). Fig. 1.2b shows that the area at risk for the heaviest rainfall with the synoptic setup is at the nose of a low-level jet streak as described by M79 and Junker et al. (1999), but other studies (e.g., Elsner et al. 1989; Moore et al. 2003) have found the heaviest rainfall near the left-exit region of the low-level jet streak. Significant precipitation often results when either of these sectors of a low-level jet streak becomes positioned ahead of a slow-moving midlevel trough (Fig. 1.2c). Sometimes the exit region of a low-level jet streak can couple with the entrance region of an upper-level jet streak (Uccellini and Johnson 1979). For example, Junker et al. (1999) noted the importance of the low-level jet streak in transporting moisture toward a region of MCS development in the Midwest, but also found that midwestern MCSs are collocated with the equatorward-entrance region of an upper-level jet streak about 60% of the time. The presence of the equatorward entrance region of an upper-level jet streak may allow heavy rain to cover a more extensive area than it would if favorable jet dynamics were not present (Kane et al. 1987). Frontal flash flood events (M79) occur in the cool, moist air mass poleward of a zonally oriented quasi-stationary surface boundary, sometimes near a weak 8 mesolow that can act to enhance low-level convergence (Fig. 1.3a). The orientation of the low-level jet streak perpendicular to the surface front creates an overrunning situation, where warm, moist air is transported above the cooler air at the surface (Fig. 1.3b). The greatest chance for heavy rainfall occurs downstream of a weak 500 hPa short-wave trough and along or just upstream of a ridge axis extending northwest of a midlevel anticyclone (Fig. 1.3c). D96 hypothesize that the margins of such a ridge favor excessive precipitation because of the tendency for sharp temperature and moisture boundaries to collect there. M79 term their third type of event a mesohigh flash flood, in which convection prior to the ERE produces a nearly stationary outflow boundary. New convection develops on the cool side of the outflow boundary southwest of the mesohigh, in an area of significant low-level moisture that is often located ahead of a slow-moving surface front (Fig. 1.4a). Heavy rain is favored further if the exit region of the low-level jet streak becomes juxtaposed with the outflow boundary (Fig. 1.4b) and a weak midlevel short-wave trough located upstream (Fig. 1.4c). The orientation of the greatest flash flood threat near the large-scale ridge axis in Fig. 1.4c is similar to the midlevel frontal pattern depicted in Fig. 1.3c. With the risk of heaviest rain on the warm side of the synoptic-scale front, weak subsidence on the periphery of the ridge may inhibit deep convection long enough for moisture to accumulate there and for low-level diabatic heating to increase the midlevel lapse rate and to decrease the vertical stability (D96). 9 Heavy rainfall may occur over one area in distinct episodes with the mesohigh type of setup in M79, increasing the potential for flash flooding. The initial rain serves to saturate the underlying soil, allowing for significant runoff of the subsequent heavy rainfall. Multiple episodes of heavy rain over one area can hinder the utility of NWS flash-flood guidance and make for an especially challenging meteorological and hydrological forecast, especially if severe weather is also present (Schwartz et al. 1990). Winkler (1988) identified an additional large-scale pattern conducive to flash flooding termed Type IV (Fig. 1.5) when she examined events over Minnesota. In this case, a narrow moisture ridge at the surface intersects a midlevel ridge axis with weak flow aloft, leading to a spatially confined, slowmoving ERE of short duration. With the upper-level trough axis still well to the west, the focus for precipitation comes from a weak surface boundary in the moist air mass. 1.3.3.3 Synoptic-Scale and Mesoscale Forcing for Ascent Synoptic-scale forcing for ascent is usually diagnosed by some form of the quasigeostrophic (QG) omega equation. Using the traditional form (e.g., Bluestein 1992, section 5.6; Holton 2004, section 6.4), previous studies have noted that summertime EREs often are associated with low-level warm-air advection (e.g., Junker et al. 1999), indicated by veering of the wind with height, and/or weak differential cyclonic vorticity advection (e.g., BC78; Junker et al. 1999). Nevertheless, ascent brought about by synoptic-scale processes is 10 usually too slow to raise a parcel of air to its level of free convection (LFC) or produce observed ERE rainfall rates (Doswell 1987). Many EREs therefore occur when mesoscale forcing for ascent dominates over or augments synoptic-scale forcing for ascent (Doswell 1987; Fritsch et al. 1998; Junker et al. 1999). D96 suggest that the background lift provided by synoptic-scale QG forcing mostly aids mesocale processes by moistening and destabilizing the atmosphere near where EREs occur. Low-level convergence can provide the lift needed to bring air parcels to their LFC and result in rapid ascent, especially in the warm season (Heideman and Fritsch 1988). Summertime convergence zones along which heavy precipitation can form span many different scales and types. Only half of all summertime rainfall in the U.S. occurs in proximity to the deep synoptic-scale fronts present in the M79 synoptic and frontal flash flood composites (Heideman and Fritsch 1988). Excluding rainfall from TCs, other, more subtle, triggering mechanisms can include thunderstorm outflow boundaries (M79), sea-breeze fronts (Heideman and Fritsch 1988; Koch and Ray 1996; Fovell 2005), and coastal fronts (Bosart et al. 1972; Appel et al. 2005), as well as prefrontal troughs, airmass boundaries, and heat troughs (Koch and Ray 1996). It is often challenging to detect such triggering mechanisms using traditional surface mesoanalyses; radar and satellite imagery have proven to be more effective (Koch and Ray 1996). If low-level confluence associated with one of the types of boundaries mentioned above acts on a preexisting thermal gradient, a region of 11 frontogenesis will be produced, and the thermal gradient may intensify. The frontogenesis produces a direct thermal circulation, which acts to enhance vertical motion near the front (Moore et al. 2003). Frontogenesis also can arise due to evaporational cooling when precipitation falls and/or to the juxtaposition of cloudy and clear air, leading to differential radiative heating (Langmaid and Riordan 1998). The diabatic cooling caused by falling precipitation and reduced radiative heating also can work to strengthen cold-air damming along the Appalachian Mountains (Fritsch et al. 1992). 1.3.3.4 Forcing for Ascent by Physiographic Processes Orographic processes can be the only source of lift necessary in areas of large terrain gradients (Barros and Kuligowski 1998), such as the EREs that led to the Big Thompson and Rapid City (Maddox et al. 1978) and Fort Collins (Petersen et al. 1999) flash floods. These EREs occurred in upslope flow regions in the Rocky Mountains, in which weak differential cyclonic vorticity advection caused by subtle short waves rotating around a mid- to upper-level trough upstream supplied a background environment of weak synoptic-scale forcing. A “bentback ridge” (Maddox et al. 1978) just to the east contributed to deep southeasterly flow against the mountains. This setup is consistent with the M79 frontal and mesohigh composites, but orographic lift was the main mechanism responsible for raising parcels in a moist, conditionally unstable air mass below a marked temperature inversion to their LFC. A further result of the deep southeasterlies in these cases was to anchor precipitation cells along the 12 foothills, which caused long periods of heavy rain and exacerbated the severity of the flooding. Heavy rain events with significant orographic upward motion also can take place in proximity to the Appalachians, as occurred in the central part of the mountain chain in January 1996 (Barros and Kuligowski 1998) and in western Virginia in June 1995 (Pontrelli et al. 1999). However, if the orographic uplift is not sufficient to bring parcels to their LFC, Barros and Kuligowski (1998) noted that the elevated terrain can still produce or enhance precipitation via the transport of warm, moist air over a cold pool that collects against the topographic barrier (indicative of cold-air damming), or through the seeder–feeder mechanism (see also, e.g., Bell and Bosart 1988). Another feature dependent on regional physiography that can act to produce or enhance precipitation is the coastal front (Bosart et al. 1972). Srock (2005, section 1.5.2) thoroughly reviews previous research on this topic, so only a brief summary will be included here. Coastal fronts arise due to land–sea temperature contrasts, orography, and the shape of the coastline, and are observed most frequently in winter along the coasts of New England, North Carolina, and Texas (Bosart et al. 1972; Bosart 1975). They also can arise due to frictional convergence near the coastline produced by gradients in surface roughness between the land and the sea (Roeloffzen 1986). Coastal fronts tend to act like warm fronts, such that overrunning enhances precipitation on their cold sides (Bosart et al. 1972; Marks and Austin 1979; Srock 2005). 13 More recent studies have examined the occurrence of warm-season coastal fronts (e.g., DeLuca 2004; Appel et al. 2005; Srock 2005). The results indicate that the mechanisms by which they form are similar to the cool season, but the temperature contrasts along warm-season coastal fronts are weaker than their wintertime counterparts. Nevertheless, coastal fronts still can act to enhance precipitation significantly, especially if they are induced ahead of polewardmoving TCs (DeLuca 2004; Srock 2005). In researching summertime coastal fronts in North Carolina, Appel et al. (2005) found that most are diurnal in nature, remain just offshore, and are infrequently associated with cold-air damming. They can develop in response to a mesoscale atmospheric circulation on the western edge of the Gulf Stream, and are usually observed when an anticyclone resides nearby or just east of the North Carolina coast. 1.3.3.5 Moisture, Instability, and Shear Significant moisture through a deep tropospheric layer is necessary to sustain moist convection capable of producing EREs and flash floods (e.g., M79). To illustrate this point, M79 calculated that the mean precipitable water for each of their composites was greater than 37 mm. More recently, Konrad (1997) concluded that a ridge axis in the 700 hPa mixing ratio pattern was the most common ingredient associated with heavy rainfall over the southeastern U.S., even though the magnitude of the atmospheric moisture content showed no direct relationship with observed rainfall amounts. Furthermore, Funk (2003) explained that the source of the moisture must be considered in addition to its 14 magnitude. If the moisture is transported from a maritime tropical air mass characterized by cumulus clouds containing many small drops and a high density and size distribution, higher rainfall totals than expected may occur. D96 noted several reasons why deep moisture was so important to heavy rain development, including that it 1) results in rainfall rates directly proportional to the precipitable water of the column through increased condensation, 2) enhances precipitation efficiency through the reduction of dry-air entrainment, which promotes evaporation, and 3) creates a relatively low LFC, which allows buoyant processes to increase updraft strength if the parcel can be lifted to that level. Convection can be either surface-based or elevated in nature. Surfacebased convection typically occurs in the warm sector ahead of a frontal boundary, as in the M79 synoptic and mesohigh flash flood composites. Positive convective available potential energy (CAPE) and a negative lifted index on an atmospheric sounding indicate the potential for surface-based convection. On a plan-view map, warm-sector convection often develops near a low-level θe ridge axis (Funk 1991), which can be used as a proxy for CAPE (Schwartz et al. 1990). It also has been shown that the vertical distribution of positive area in a long, narrow manner reduces parcel vertical velocity and acceleration (Blanchard 1998). The resultant increase in parcel residence time within the warm cloud layer promotes greater precipitation efficiency due to the collision–coalescence process (Beard and Ochs 1993). 15 Moore et al. (2003) conducted an extensive study of the features that bring about elevated convection, which are summarized in Fig. 1.6. This type of activity occurs on the cool side of a front as warm, moist air is transported over the surface boundary by a low-level jet streak (Colman 1990a,b; Moore et al. 2003; Schumacher and Johnson 2005), making the environment favorable for elevated convection strikingly similar to the M79 frontal flash flood composite. Deep moisture and the equatorward entrance region of an upper-level jet streak also are common features, as previously discussed in this subsection and in subsection 1.3.3.2, respectively. Potential instability [indicated by equivalent potential temperature (θe) decreasing with height] and CAPE can exist in elevated layers above the frontal inversion (Moore et al. 2003), with low-level frontogenesis (Colman 1990b) and/or warm-air advection (Moore et al. 2003) typically providing the lifting mechanism necessary to release the elevated instability. It can be inferred from Henry (1988) that the Showalter index and K index are better-suited for evaluating elevated instability that may promote flooding rains than surface-based CAPE and the lifted index. The Showalter index and K index for the composites in M79 were less than −2 and greater than 35, respectively, indicating at least an 80% chance for thunderstorm development utilizing the forecasting rule of thumb proposed by Henry (1988). Diagnosis of the potential for heavy convective precipitation often begins with an evaluation of moisture content and instability (D96), but surface moisture flux convergence (MFC) also can be an important aid in predicting heavy rainfall location and extent up to three hours in advance (Banacos and Schultz 2005). 16 MFC has a long history of use in ERE case studies (e.g., BC78; Elsner et al. 1989) because it can be helpful for identifying mesoscale boundaries, such as those described in section 1.3.3.3, when they can be resolved by gridded surface analyses. However, use of MFC assumes that the dominant lifting mechanisms are not located significantly above the boundary layer and that CAPE and convective inhibition are favorable for deep, moist convection (Banacos and Schultz 2005). Junker et al. (1999) discussed that MFC also is evident in the exit region of an 850 hPa jet streak that transports moisture toward an area of heavy rainfall, but it is significant spatial extent and duration of MFC that distinguishes large EREs from smaller ones with lesser rainfall rates. Another ingredient that is important to assess when predicting warmseason EREs is the vertical wind shear, which can diagnose the anticipated organization of the convection and the system motion vector. EREs often are associated with weak-to-moderate speed shear, which tends to organize the convection into linear structures (Houze et al. 1990; D96; Schumacher and Johnson 2005) and allow individual convective elements to move slowly. If, in addition to slow system movement, the most favorable ingredients for cell redevelopment discussed previously are located upstream from the active convection, training echoes will be observable on radar, and the duration of heavy rainfall at any one location will be prolonged (D96). Flash flooding can occur in precipitating systems that form in strongly sheared environments, but either the rainfall rates must be exceptionally high or significant upstream 17 redevelopment must take place for an extended period of time to counterbalance the faster movement of individual cells. Significant veering of the wind with height has been positively correlated with rainfall amounts occurring with EREs in the southeastern U.S. (Konrad 1997). The deep (but not necessarily strong) warm-air advection this veering implies is most common in frontal flash flood events because of the orientation of the low-level jet streak with respect to the midlevel ridge axis (see Figs. 1.3b,c). Synoptic flash flood events are associated more commonly with veering winds only in the lowest layers of the troposphere and unidirectional speed shear aloft (see Figs. 1.2b,c). 1.4 Precipitation Distribution Associated with Landfalling and Transitioning TCs 1.4.1 Background on ET TCs have profound effects on many regions of the world, as they bring high wind, heavy rain, and storm surge (Jones et al. 2003), as well as sporadic tornadoes (Edwards and Pietrycha 2006) with them, particularly to coastal and maritime locations. When TCs make landfall, they weaken as the land surface cuts off the surface heat and moisture fluxes that drive them over the warm ocean. They may undergo ET, regardless of whether or not they make landfall, as they encounter the midlatitude westerlies, but the degree of transition and redevelopment as an extratropical cyclone varies on a case-by-case basis. Summaries on the current understanding of ET are presented in Jones et al. (2003), DeLuca (2004), and Srock (2005), so only a brief discussion will be 18 given here. The two-step process of ET illustrated by Fig. 1.7 (Jones et al. 2003) first outlines the environmental changes encountered by a TC when it approaches the midlatitudes. These may include, but are not necessarily limited to, stronger temperature and moisture gradients, increased vertical shear, lower SSTs, and a larger Coriolis parameter. The TC responds through structural changes, most notably the development of significant asymmetries in the temperature, wind, and moisture fields. The overall intensity of the transitioning TC decreases, but its areal extent expands. The TC then may interact with an upper-level trough or low-level baroclinic zone if either is in sufficient proximity, or simply decay in the Tropics without completing ET if neither of these midlatitude features is present. When the TC becomes an extratropical cyclone, the ET process takes an average of 36 h (Evans and Hart 2003). Potential reintensification as an extratropical cyclone is the subject of a third stage in the ET model proposed by Klein et al. (2000), although reintensification is not necessary to have extreme impacts on land- and oceanbased interests (Jones et al. 2003). Redevelopment is often observed when a midlatitude cyclonic circulation is located northwest of the TC (Harr and Elsberry 2000; Harr et al. 2000) at an average distance of 1700 km (Foley and Hanstrum 1994), but can occur more gradually with less intensification if the midlatitude system is northeast of the TC (Harr and Elsberry 2000; Harr et al. 2000). The strength of the subsequent extratropical cyclone generally depends on the degree of interaction between the TC and upper-level dynamical and low-level thermodynamic processes (Klein et al. 2002). 19 1.4.2 Precipitation Directly Associated with a TC 1.4.2.1 Large-Scale Considerations Forecasting rainfall amount, location, and duration is one of the most complicated tasks associated with a landfalling or transitioning TC (Jones et al. 2003). A successful quantitative precipitation forecast requires accurate prediction of TC track, intensity, structural changes, and interactions with midlatitude circulations, which often are represented inaccurately by numerical models. Unreliable predictions can result in excessive rainfall amounts occurring in unexpected locations, causing significant loss of life and property damage. Heavy rainfall is usually symmetric about the TC track in mature storms (Frank 1977; Marks 1985; Rodgers et al. 1994), although asymmetries may occur depending on the environmental shear in the vicinity of the storm, as well as the TC motion (Chen et al. 2006). A left-of-track shift in the rainfall distribution commonly occurs at higher latitudes (Atallah et al. 2007) as the TC undergoes ET and interacts with a midlatitude circulation (Jones et al. 2003). The relevant features involved in this process are illustrated schematically in Fig. 1.8a by Atallah et al. (2007) using the potential vorticity framework (Hoskins et al. 1985). The approach of a strong, positively tilted midlatitude trough enhances the lowlevel thermal gradient north and west of the TC, gradually altering the mechanisms forcing ascent from symmetric diabatic heating to the differential vorticity and temperature advections featured in QG theory. Additionally, downstream ridging induced by diabatic heating can enhance the upper-level divergence occurring within the equatorward entrance region of a southwesterly 20 upper-level jet streak, leading to more favorable jet dynamics and an increased likelihood of intense precipitation, sometimes well downstream of the TC center (Atallah and Bosart 2003; Jones et al. 2003; Atallah et al. 2007). Atallah et al. (2007) explained that the effects of diabatic heating can be inferred qualitatively by the nonconservation of potential temperature on the dynamic tropopause. A right-of-track precipitation distribution occurs more readily at lower latitudes than at higher latitudes as the TC interacts with a well-defined east– west oriented ridge that has likely been enhanced by diabatic heating and upperlevel outflow from the TC (Fig. 1.8b; Atallah et al. 2007). In this case, the upstream trough is much broader and weaker, and the forcing for vertical motion through vorticity advection by the thermal wind shifts to the northeast of the TC. Most storms that exhibit this shift make landfall along the Gulf Coast and then track along the Appalachians, which suggests qualitatively that upslope flow along the eastern slopes will promote a right-of-track rainfall pattern. This type of rainfall pattern tends to occur with TCs of relatively small spatial scale, though the precipitation distribution often shifts from right-of-track to left-of-track as the TCs continue moving poleward (Atallah et al. 2007). 1.4.2.2 Small-Scale Considerations The mesoscale and physiographic features discussed as potential foci for heavy warm-season rainfall in sections 1.3.3.3 and 1.3.3.4, respectively, also can act to enhance precipitation directly within or out ahead of a TC (DeLuca 2004; Srock 2005). Specifically, Srock (2005) discusses how coastal fronts and 21 topographic barriers can modulate the precipitation distribution, especially in the southeastern U.S. Mesoscale coastal fronts played key roles in augmenting rainfall rates with the approach of Agnes (1972) (Bosart and Dean 1991), Marco (1990) (Srock 2005), and Bob (1991) (DeLuca 2004). Floyd (1999) is an example of how an intense coastal front associated with a troposphere-deep baroclinic zone can act in concert with the large-scale factors discussed in section 1.4.2.1 to produce an extensive shield of rain spanning a considerable distance from the TC center (Atallah and Bosart 2003). Orographic contributions to rainfall have been documented in connection with numerous Atlantic basin storms impinging upon the Appalachians, such as Jerry (1995) (Srock 2005), and smaller-scale mountain ranges of the northeastern U.S. (DeLuca 2004). Sinclair (1993) documented Cyclone Bola (1988), which deposited copious amounts of rainfall in upslope regions of New Zealand. In general, any time a TC circulation causes a moist airstream to intersect a topographic barrier, regardless of the distance from the TC center, there is the potential for enhanced rainfall. 1.4.3 Heavy Precipitation Well Downstream of a TC Jones et al. (2003) included a short section in their review paper about the potential for areas of heavy rain separate from the main TC circulation to develop well ahead of the center of the TC, even if it never makes landfall. Noting that this rain is often not anticipated because of its distance from the TC center, they described the increased risk of flooding that can occur if the heavy rainfall directly 22 associated with the TC falls over the same region as the antecedent precipitation. Awareness that heavy rainfall can break out well ahead of a TC is not new, although there has been relatively little work reported in the refereed scientific literature on this type of event. Peripheral awareness of this possibility dates back to at least 1955, when Mook (1955) described a long line of convergence that extended well northeast of a weakening Diane (1955). This line, which DeLuca (2004) suggested may have been a weak coastal front, spawned distinct areas of heavy rain in Pennsylvania and Connecticut that were partially enhanced by orographic lift, and another near Boston that was associated with a weak cyclonic circulation that developed along the convergence line. The most in-depth published analysis of a PRE was on the episode centered near Wellsville, NY, far ahead of Agnes (1972) (BC78). The rainfall distribution in Fig. 1.9a shows that the region in New York occurred as a separate entity from the heavy precipitation directly associated with Agnes. BC78 found that the inland flooding occurred 1) within a midlevel confluence zone between the tropical and midlatitude airstreams (Fig. 1.9b), 2) in an environment characterized by weak differential cyclonic vorticity advection, and 3) as a narrow tongue of deep tropical moisture was transported well poleward of the TC through a low-level ridge line. The heavy rain in New York eventually became embedded within a surface pressure trough as a small-scale cyclonic circulation began to develop just north of the PRE. 23 Subsequent studies of Agnes (DiMego and Bosart 1982a,b; Bosart and Dean 1991) continued to document heavy rainfall in advance of the TC in addition to the Wellsville, NY, heavy rains. They found that the highest precipitation rates were concentrated on the windward slopes of the Appalachians in the presence of orographic lift and moisture, as well as along and just west of a weak coastal front extending well to the north of Agnes. The space–time cross section from Bosart and Dean (1991) shown in Fig. 1.10 demonstrates that the passage of the coastal front was marked by a significant wind shift from easterly to north-northwesterly, an increase in wind speed, and a temperature drop of approximately 3°C, coincident with the onset of the most intense rainfall rates. Vigorous ascent likely resulted from a weak, frontogenetical direct thermal circulation in the presence of a moist tropical air mass characterized by a moist–neutral stability profile. An upper-level jet streak combined with low-level convergence and overrunning along a TC-induced θe gradient to produce significant rainfall in advance of Marco (1990) (Srock 2005). A coastal front in eastern North Carolina ahead of Floyd (1999) helped focus heavy rainfall there as the storm approached (Atallah and Bosart 2003; Colle 2003). The tropical disturbance that eventually would become Leslie (2000) interacted with a stalled frontal boundary in southern Florida to produce extremely heavy rains (Franklin et al. 2001). In studying the accuracy of the Z–R relationships used by the NWS with the Weather Surveillance Radar-1988 Doppler (WSR-88D), Ulbrich and Lee (2002) found that two periods of heavy precipitation preceded the arrival of rainfall directly 24 associated with Helene (2000) in upstate South Carolina: one associated with the initial passage of a cold front (Fig. 1.11a), and another poleward of the front as it became quasi-stationary and oriented east–west (Fig. 1.11b). Finally, Lin et al. (2001) have found that heavy precipitation with a significant orographic component preceded the arrival of a tropical depression (TD) in 1959 and Tropical Storm Rachel (1999) in Taiwan, as well as another TD in Japan in 1974. They concluded that the role of the TCs was to enhance the moist low-level jet streak impinging upon the steep mountain slopes, which aided in the release of conditional instability. The proximity of a slow-moving synopticscale system, such as a quasi-stationary front, helped slow the movement of the convective systems and prolong the rain. 1.5 Goals and Organization of the Thesis This thesis aspires to marry the previous research presented here on EREs and the precipitation distribution associated with TCs to show how the approach of a TC at great distances can help bring about the ingredients necessary for the formation of PREs. An extensive climatology and several detailed case studies will be used to provide forecasters with a practical framework for better predicting the occurrence of these PREs, with an eye toward continuing the recent trend of successful incorporation of CSTAR research into NWS operations. The thesis will be organized as follows. Chapter 2 will provide an overview of the data sources, demonstrate the PRE identification process, and explain the 25 various methods used in constructing the climatology, the three case studies, and the null case study. Chapter 3 will present the first known statistical and composite climatology of PREs, with an emphasis on their overall properties and the main synoptic-scale features associated with them. Chapter 4 will document the governing synoptic-scale and mesoscale dynamics and thermodynamics associated with the three selected case studies and one null case. Chapter 5 will relate the findings of the case studies to the literature review and suggest a forecast strategy for PREs by comparing the different case studies among themselves and with the null case. Technology transfer of this research into NWS operations also will be discussed in this chapter. Chapter 6 will conclude with a synthesis of this research and suggestions for future work. 26