Fields from the NCEP/NCAR reanalysis are used in conjunction with simulations from Polar MM5 to examine regional processes contributing to cyclone development in the sub-Arctic/Arctic sector of the North Atlantic. Strong deepening of systems (>6 hPa/6 hours) is especially common south of the tip of Greenland and in the vicinity of the mean Icelandic Low (IL). A secondary maximum is located just south of Svalbard. Composite analyses of strongly deepening systems in these areas point to the influence of pronounced preexisting low-level temperature gradients linked to ocean currents, the sea ice margin and ocean-land contrasts. The area of preferred deepening near Svalbard corresponds to the northernmost penetration of open water in the North Atlantic. Impacts of Greenland's orography are prominent in the synoptic complexity of the region. Typical synoptic situations, examined as case studies, include splitting, or “bifurcation” of cyclones at the southern tip of Greenland; orographic cyclogenesis in the lee of Greenland, at the location of the mean IL; and deepening of preexisting systems near the IL. Orographic influences are clearly captured in Polar MM5 simulations for which control runs of the three cases are compared to runs for which the orography of Greenland is removed.
 The dominant feature of the mean sea level circulation in the northern North Atlantic during winter is the Icelandic Low (IL) and an associated pressure trough extending far into the Arctic (Figure 1). The region encompassing the IL and this trough can be broadly thought of as representing the terminus of the North Atlantic cyclone track. It is one of the most synoptically active and variable areas of the planet, especially during winter.
 The North Atlantic storm track penetrates much further north than its Pacific counterpart. As determined from the number of closed surface lows on the basis of analysis of 6-hourly sea level pressure (SLP) fields (see later discussion), cyclone activity is maximized within the IL with a secondary peak in the Norwegian Sea (Figure 2). Consistent with this poleward penetration of cyclone activity, the region between Greenland and Scandinavia is a primary gateway for the transport of moist static energy (sensible heat, latent heat and geopotential) into the Arctic that is required to balance the high-latitude radiation deficit at the top of the atmosphere [Nakamura and Oort, 1988; Overland and Turet, 1994]. The IL and surrounding region modulates the Arctic's freshwater budget, especially through impacts on net precipitation over the Arctic Ocean [Walsh et al., 1995; Serreze et al., 1995; Rogers et al., 2001] and the export of sea ice and low-salinity waters out of the Arctic and into the North Atlantic via Fram Strait, between northern Greenland and Svalbard [Vinje, 2001]. This flux is the principal mechanism by which freshwater inputs to the Arctic Ocean from river runoff, Bering Strait inflow, and net precipitation over the Arctic Ocean are balanced. The flux is closely linked to the SLP gradient in the vicinity of the strait [Brummer et al., 2003], which varies with the strength and location of the IL and high-latitude pressure trough.
 The strength and location of the IL is, of course, allied with the phase of the well-known North Atlantic Oscillation (NAO), which can be described as a mutual strengthening/weakening of the IL and Azores High. Both are strong (weak) in the positive (negative) phase of the NAO. The NAO has pronounced correlations with moisture transport into the Arctic [Dickson et al., 2000; Rogers et al., 2001], the Fram Strait ice flux [Kwok and Rothrock, 1999] and high-latitude surface air temperature (SAT) anomalies [Hurrell, 1996]. From the early 1970 through the mid 1990s, the winter NAO (and its hemispheric-scale counterpart, the Northern Hemisphere Annular Mode, or NAM) showed a general positive trend, helping to explain patterns of recent Arctic warming [Hurrell, 1996; Thompson and Wallace, 1998], declines in sea ice extent [Rigor et al., 2002; Rigor and Wallace, 2004] and an increased inflow of warm, salty Atlantic-derived waters into the Arctic Ocean [Dickson et al., 2000]. Whether increased greenhouse gas concentrations may favor a more positive NAO/NAM is a matter of continuing debate [e.g., Thompson et al., 2000; Hoerling et al., 2001].
 An important facet of circulation variability in the northern North Atlantic is the strong impact of local and regional processes. The IL and surrounding area is synoptically complex (as first noted by Petterssen [1956, see Figure 13.2.2]) and is characterized by a high frequency of cyclogenesis and rapid cyclone deepening [Serreze et al., 1997]. Factors contributing to this complexity include the orographic influences of Greenland and strong low-level temperature gradients, associated with ocean currents, the sea ice margin and the cold ice sheet. It is reasonable to expect that the intensity of local development processes relates to differences in the general synoptic environment associated with the positive and negative phases of the NAO, which amplify or dampen its winter imprint on the circulation. Two modeling studies [Alexander et al., 2004; Deser et al., 2004] provide evidence that retreat of the sea ice margin in the Atlantic invokes a local change in heating, which involves at least a weak negative response of the NAO.
 The objective of the present study is to highlight some of the features of the IL and surrounding region that contribute to its synoptic complexity. In particular, we assess the role of the Greenland's orography in the synoptic evolution of the North Atlantic cyclones. A description of analysis tools is presented in section 2, it is followed by a description of the general synoptic environment in section 3. In section 4 we conduct an observational analysis of strongly deepening cyclones and investigate links with regional processes, employing a composite approach. Section 5 describes three case studies of cyclone development based on observations and simulations from a regional model. Findings are summarized in section 6.
2. Analysis Tools
 The observational analysis focuses on winter months (November–March) for the 21-year period 1979–1999 and draws on various fields from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis (hereafter NCEP) [Kalnay et al., 1996], including SLP, 500 hPa height, 1000–500 hPa thickness, temperatures, and 500 hPa vertical motion (omega). Comparisons show that these NCEP fields are very similar to those provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA-40).
 Use is also made of output from a cyclone detection and tracking algorithm applied to 6-hourly NCEP SLP fields for the Northern Hemisphere. It represents the most recent incarnation of the algorithm first described by Serreze  and improved for the study of Serreze et al. . This version was subsequently used in studies of Northern Hemisphere cyclone trends [McCabe et al., 2001] and precipitation variability over northern Eurasia [Serreze and Etringer, 2003]. The algorithm provides the location of each cyclone center, cyclone central pressure, and 6-hour deepening rates. Cyclone detection is based on a series of search patterns, testing whether a grid point SLP value is surrounded by grid point values higher than the central point tested. Cyclone tracking employs a nearest neighbor approach that compares system positions at a given 6-hour chart with those for the next 6-hour chart. Deepening is defined as the difference in SLP of a cyclone center between these two charts, with the deepening ascribed to the analysis time of the later chart. Figure 2 is based on the algorithm output.
 The modeling studies employ the polar version of the Mesoscale Model 5 (Polar MM5). Our investigations compare control runs against simulations in which the orography of Greenland is removed. Further details are provided in section 5. MM5, which has found wide acceptance by the community, was developed by NCAR and Pennsylvania State University. Dudhia  and Grell et al.  provide general descriptions. Polar modifications are reviewed by Bromwich et al.  and Cassano et al. . Our simulations employ a 40 km horizontal resolution with 28 vertical sigma levels, the first eight of which represent the bottom 500 m of the boundary layer. We use the MM5 SLAB surface parameterization, which allows for fractional sea ice cover, prescribed using satellite data from the National Snow and Ice Data Center. Simulations employ the Cumulus Parameterization of Grell, the Eta Boundary Layer scheme, the NCAR Community Climate Model 2 radiation scheme and the explicit Reisner 2 mixed phase moisture scheme. The horizontal domain consists of 85 × 82 grid points centered at 75°N 25°W (on the east coast of Greenland). MM5 is driven at the domain boundaries with NCEP data on a 6-hourly basis.
3. General Synoptic Environment
 As discussed, the IL is part of a broad area of low pressure extending well into the Arctic. A number of processes account for the existence of the IL. One of these is the low-level thermal effect of the comparatively warm, ice free underlying ocean adjacent to the sea ice margin and cold land. A feature of the IL (as well as the Aleutian Low in the North Pacific) that supports this idea is that the mean isobars around these lows are parallel to and tightly crowded along the coastlines [Wallace, 1983] (Figure 1). Second, the IL is located downstream of the major eastern North American stationary longwave trough where eddy activity will be favored. Strong regional influences are the third factor in cyclone development.
 The presence of the Greenland ice sheet is the single most important regional influence on the behavior of synoptic systems in the North Atlantic. These influences include splitting, or “bifurcation” of lows moving in from the south and southwest by the high ice sheet barrier (elevations at about 64°N, near the southern tip of Greenland exceed 2800 m, roughly the 700 hPa level) [Kurz, 2004], distortion of wind and temperature field, and leeside cyclogenesis off the southeast coast [e.g., Doyle and Shapiro, 1999; Kristjansson and McInnes, 1999; Petersen et al., 2003].
 For example, Kristjansson and McInnes  performed a modeling case study of a cyclone for January 1995, using different orographic conditions. The control simulation used the observed orography of Greenland. Other runs included “no Greenland” (orography removed, as will be the case in our simulations), “half Greenland” (elevations cut by 50%) and “double Greenland” (elevations doubled). When passing over Greenland, the case study cyclone splits into two systems in both the control and double Greenland simulations. The original (baroclinic) cyclone weakens, whereas the secondary (leeside) cyclone becomes rather strong. In the “no Greenland” and “half Greenland” simulation, the system does not split and the baroclinic system is comparatively stronger. Kristjansson and McInnes  conclude that weakening of the baroclinic low is a result of the orographic deflection of cold air, which weakens cold advection behind the cyclone. This effect is further enhanced by warm air advection ahead of the lee cyclone.
 As a start to documenting regional processes, Figure 3 shows a highly smoothed depiction of the distribution of winter events of cyclone deepening of at least 6 hPa/6 hours, based on the algorithm described earlier. Smoothing relies on averaging deepening counts at each grid cell with counts for neighboring grid cells. The deepening rate is a good indicator of synoptic development. A deepening rate of 6 hPa/6 hours is rather large, and if maintained over a day, would classify the event as a synoptic “bomb” [Sanders and Gyakum, 1980]. It is evident that high-latitude deepening tends to occur in preferred regions along the primary North Atlantic cyclone track. Strong deepening tends to be especially common south of Greenland, at about 54°N. There is a second, more localized center of roughly equal peak magnitude near the locus of the mean IL and a subsidiary maximum just south of Svalbard. Figure 3 has no latitude weighting. Adjusting the counts of strong deepening to 60°N (to account for the fact that storms at different latitudes with identical pressure gradients do not produce the same geostrophic wind) shows the same three relative maxima, but another maximum near the map edge south of Newfoundland. The same relative maxima also appear in the raw (unsmoothed) counts.
 The area of preferred deepening south of Greenland is located in a region of maximum height gradients just east of the axis of the mean eastern North American 500 hPa trough (Figure 1), where there will tend to be positive vorticity advection increasing with height, promoting vertical motion and divergence. This general relationship is supported by the mean patterns of omega at 500 hPa (not shown), which has a local peak in upward motion slightly east of the preferred area of deepening. However, the largest mean upward values are actually found to the north near the IL. Sanders and Gyakum , Roebber , Rogers and Bosart  and others show that explosively deepening systems (“bombs”) tend to occur in the proximity of warm ocean currents in the North Atlantic and North Pacific Oceans. As seen in Figure 4, the mean surface skin and sea surface temperature (SST) field from NCEP, the area of preferred deepening south of Greenland is just to the south of a strong gradient in SST. This gradient is associated with the contrast between the cold Labrador Current and the warmer North Atlantic Drift. Strong low-level temperature gradients also extend far into the Arctic. These are associated with the contrast between relatively warm open ocean water (associated with the North Atlantic drift) and the sea ice margin, the Greenland ice sheet, and the cold southward flowing East Greenland current. That such temperature gradients help support cyclone development in this area (including mesocale polar lows) has been documented in a number of past studies [e.g., Shapiro et al., 1987; Serreze et al., 1993, 1997; Serreze, 1995; Deser et al., 2000].
 Given that winter cyclone counts are highest near the locus of mean IL, it follows naturally that this region appears as a maximum in the frequency of strong deepening and cyclogenesis events (Figures 2 and 3). However, especially strong development in this region finds support in the analysis by Serreze , who used SLP data over the winter period 1973–1992 to determine the location and magnitude of maximum 12-hour deepening rates over the lifecycle of individual cyclones. The winter mean 12-hour maximum deepening rate (adjusted to 60°N latitude) in the IL region of 6.8 hPa (and the 90th percentile value of 16.8 hPa) was found to far exceed corresponding values in other high-latitude regions. Frequent development of systems migrating into the IL region from the south has in turn been noted in a number of earlier studies [e.g., U.K. Meteorological Office, 1964; Whittaker and Horn, 1982].
 The area of preferred deepening just south of Svalbard corresponds to the most northward penetration of open water (Figures 3 and 4), in close proximity to very cold air over the Arctic Ocean. As expected, this area is also associated with a local peak in mean upward vertical motion at 500 hPa.
4. Composite Analyses
 Referring back to Figure 3 and section 3, three areas can be identified with an especially high frequency of strong deepening (>6 hPa/6 hours): to the south of Greenland, off the southeast coast of Greenland (corresponding to the IL), and just south of Svalbard. We term these region 1, region 2 and region 3, respectively. Rapid development in these regions is considered from a composite approach. We identified the grid cell in each region with the highest counts of strong deepening. These are from the raw grid cell counts, not the smoothed counts used to compile Figure 3. Using the analysis time of each deepening event, we then extracted diagnostic fields from the NCEP reanalysis for 6-hour periods up to 72 hours before and 72 hours after each deepening event. These fields were then averaged, allowing us to assess the composite evolution of lows deepening in each region. Time 0 hours in each composite member is taken to be the time of strong deepening (the deepening based on the SLP change during the previous 6 hours). While there will obviously be considerable variability between the different composite members selected for each region, as will become evident below, the approach captures meaningful information regarding the basic features of synoptic development. Results for the three regions are not directly comparable, as the deepening threshold is not latitude weighted. The primary concern was to adopt a threshold that yielded a sufficient number of cases for analysis.
4.1. Region 1, 57°N 53°W, South of Greenland
 The winter data for 1979–1999 reveal 26 deepening events at this location exceeding the selected deepening threshold of 6 hPa/6 hours. The systems comprising the composites originated variously along the Atlantic coast, the open Atlantic, and the lee of the Rocky Mountains. One was traced as a closed low all the way from the Gulf of Alaska. Figure 5 shows composite mean fields based on these 26 events for −12, −6, 0 and +12 hours. Recall that 0 hours is taken to be the time of strong deepening. The plots include SLP, 500 hPa height and the 1000–500 hPa thickness. At −24 hours (not shown) the composite field features three cyclones, one along the Scandinavian coast, one near the mean locus of the IL, and a third, the system of interest, in the Gulf of St. Lawrence. Twelve hours later, corresponding to time −12 hours, the other two systems have weakened, but the system of interest has moved to north of Newfoundland and has strengthened. It is associated with a pronounced region of upward vertical motion at 500 hPa (not shown). Note the prominent cold advection behind the system and the warm advection ahead of it, consistent with deepening of the 500 hPa trough and strengthening of the downstream ridge between −24 and −12 hours. The trough is associated with a 500 hPa closed low to the north. Between −12 and −6 hours the cyclone moves over open water and further deepens, then rapidly deepens (−6 to 0 hours) in association with further development of the temperature advection pattern.
Figure 6 illustrates relationships with low-level thermal influences. SLP is plotted along with temperatures at the lowest model level (sigma 0.995, about 50 m above the surface assuming a 1000 hPa surface pressure) for the same periods as in Figure 5. In being driven by ocean currents, the sea ice margin and the ocean-land contrasts, the low-level temperature pattern remains relatively fixed through the evolution of the cyclone. Note how the system is strongly drawing in cold Arctic air at low levels behind it, while drawing warm maritime air ahead of it during the period of rapid deepening (between −6 and 0 hours). The salient point is that the developing system has strong preexisting temperature gradients to draw from. It is reasonable to surmise that latent heat release is contributing to the development of the composite system (as illustrated in the case study by Skeie et al. ). Twelve hours later, at time +12 hours, the composite system shows evidence of backing into the cold air at 500 hPa and losing its vertical tilt, indicative of occlusion (Figure 5).
 As time progresses, the composite system splits, or bifurcates near the tip of southern Greenland. Initiation of this process is seen in the +12 hour charts in Figures 5 and 6. The center of the distorted low is colocated with a closed pocket of warm air in the Labrador Sea. By +48 hours (not shown), two distinct systems are present, one to the west of Greenland in Davis Strait, and another along the east coast of Greenland. The western cyclone subsequently dissipates. In part, this is simply an artifact of the compositing. Our algorithm shows that, after deepening southwest of Greenland, about half of the systems tracked west into Baffin Bay, with the remainder tracking east into the IL and surrounding region. Bifurcation of the composite low reflects this. However, inspection of numerous NCEP fields reveals that bifurcation of individual systems as they interact with the southern coast of Greenland is quite common. In such bifurcation cases, the tracking algorithm must select one of the lows as a continuation of the original system, while the other is counted as a cyclogenesis event. This is part of the reason why the algorithm shows a high frequency of cyclogenesis both to the east of Greenland in the IL and to the west in Baffin Bay. Characteristics of these bifurcating systems will be examined as a case study in section 5 with references to the study of Kurz .
4.2. Region 2, 63°N 27°W, East of Greenland
 Region 2 is near the locus of the climatological IL. The composite fields in Figure 7, again for −12, −6, 0 and +12 hours are based on the 32 strong deepening events observed in this region over the 21 year period. Most of these systems formed over the Atlantic. One was traced as a closed low back to the Gulf of Mexico, while another one was traced to the Gulf of Alaska. After deepening in region 2, the individual cyclones tracked to the area just north of the IL, into the Arctic Ocean, or into northern Europe.
 The composite plot for −24 hours (not shown) is broadly similar to that for the region 1 composite at the same time in showing a pronounced trough with its axis roughly aligned along Labrador and Newfoundland coasts, with a closed low to the north centered over Baffin Island. Cold advection is evident behind the trough. The trough then migrates east northeast. At −12 hours, the southern part of Greenland and the region extending into Baffin Bay appears as a broad area of relatively low pressure. Between −12 and −6 hours the composite system becomes more compact. Rapid deepening of over 10 hPa then occurs between −6 and 0 hours near the mean location of the IL. The 0 hour plot shows strong cold advection behind the low, and strong warm advection ahead of it. Warm advection increases thickness, thus builds the 500 hPa ridge, while the cold advection decreases thickness and deepens the 500 hPa trough; consequently the composite low intensifies. At time 0 hours, the mean upward motion at 500 hPa is maximized just east of the surface low where the positive temperature advection is strong. By time +12 hours, the composite system has moved slightly north and shows evidence of occlusion as the temperature advection weakens and the system begins to lose its vertical tilt.
 As with the region 1 composite system, the composite SLP and 0.995 sigma level temperature plots (Figure 8) show clear relationships with the largely standing pattern of low-level temperature gradients. One can see how the composite system draws strongly on low-level cold air from Greenland as it moves into the IL region. Unlike the region 1 composite, the region 2 composite continues to develop in the lee of Greenland. At +12 hours after the deepening (Figures 7d and 8d) the composite low is rather broad, suggesting that different cyclones comprising the composite move at different speeds.
4.3. Region 3, 77°N 5°E, Near Svalbard
 Although there were only 11 cases in region 3, this area is interesting in appearing as a high-latitude local maximum in strong deepening. The composite fields for −12, −6, 0 and +12 hours are given in Figures 9 and 10. The cyclones that deepened near Svalbard came from various places. According to our algorithm, some can be traced as closed lows across the Greenland ice sheet. A few came from the Arctic Ocean and Canadian Archipelago, while one originated off the coast of Newfoundland.
 The most notable features at −24 hours (not shown) are the pronounced meridional flow at 500 hPa along the eastern side of Greenland, associated with a broad 500 hPa trough to the west of the island and a pronounced ridge to the east. The ridge axis runs roughly from the British Isles into the Norwegian Sea. Compared to the region 1 and region 2 plots for the same time, there are much stronger geopotential height and thickness gradients along the eastern side of Greenland. The area around the IL and Baffin Bay/Davis Strait appears as a broad area of relatively low pressure.
 By −12 hours (Figure 9) a weak disturbance has developed along the east coast of Greenland, about halfway up the island. The composite disturbance then grows, deepening rapidly between −6 and 0 hours just south of Svalbard. At time 0 hours, there is a pronounced temperature advection pattern, associated with a distinct peak in 500 hPa vertical motion (not shown) just east of the surface feature. The SLP at the center of the composite cyclone at time 0 hours is comparable to that for regions 1 and 2. After the period of rapid deepening, the composite system moves to the east. It can be tracked for about 2 more days until dissipating near the island of Novaya Zemlya.
 As discussed, a feature of region 3 consistent with it appearing as a local maximum in strong deepening is that it represents the northernmost penetration of open ocean water (see Figure 4). As a result, there will be strong preexisting low-level temperature gradients as well as a moisture source for systems to draw from. Note the tight low-level temperature gradients along the sea ice margin (Figure 10). The low-level advection pattern is enhanced by relatively warm air penetrating into high latitudes east of the low associated with a surface anticyclone centered over the British Isles.
5. Case Studies
 We consider three case studies, selected from inspection of many charts as representative of common patterns of synoptic evolution in the area. The first represents a bifurcation event in the IL region, with 0 hours (the time of bifurcation) taken to be at 1200 UT, 17 December 1988. The second documents what we interpret as a “classic” leeside cyclogenesis in the IL region. Time 0 hours, the formation of the leeside low, corresponds to 0600 UT, 11 March 1992. The third represents a situation with no bifurcation, but strong deepening (>6 hPa/6 hours) in the IL region, at 1800 UT, December 21 1990 (time 0 hours) followed by development of a second low to the east. Section 5.2 provides an observational perspective using NCEP fields. Results from Polar MM5 simulations are discussed in section 5.3.
5.2. Observational Perspective
Figure 11 shows fields for the bifurcation event of SLP, 500 hPa height and 1000–500 hPa thickness at −24, −12, 0 and +12 hours. At −24 hours, a deep, mature low is located south of Greenland, very close to region 1 examined previously in the composite analyses. By −12 hours, the cyclone has moved to the northeast and has begun to distort as it nears the tip of Greenland. Time 0 hours sees bifurcation of the system into two closed lows, the second appearing in the lee of Greenland (as it is quite weak, it does not readily appear as a closed low in Figure 11). At the same time, the system on the west side of Greenland is weakening. Our tracking algorithm considered the deeper low on the west side of Greenland as a continuation of the original system, with the weaker low to the east as a cyclogenesis event. While the first appearance of a closed low (a 1 hPa isobar in our algorithm) is a reasonable definition of cyclogenesis, it seems that the eastern low might be more properly termed a bifurcation. By +12 hours, the eastern low has deepened, while the system to the west has dissipated. The eastern low then migrates east out of the IL region.
 It might be argued that distortion/bifurcation of the system as depicted in the NCEP data is an artifact of reducing pressures to sea level over the ice sheet. However, such events are in fact real. Kurz  evaluated the bifurcation phenomenon through a case study for February 2001. In summary, as a cyclone approaches Greenland from the southwest, the warm air flow ahead of the low is blocked by the island and two pools of warm air, one to the west of the island and the other to the east, are formed. Simultaneously, the “original” vorticity maximum tracks northward (to the west of Greenland), and a separate vorticity maximum forms at the southern tip of Greenland and migrates to the northeast. This orographically induced splitting is crucial for the formation of a second cyclone to the east of the island.
 Fields for the leeside development event are provided in Figure 12 for −12, −6, 0 and +12 hours. At −24 hours (not shown) the system of interest is located well west of Greenland, near the Labrador coast. It moves quickly to the east, and at −12 hours, is found along the southwest coast of Greenland, where it deepens. The low moves very little between −12 and −6 hours. By 0 hours the low on the west side of Greenland has largely dissipated, while a new system has developed along the southeast Greenland coast in the region of strong thickness gradients. Note the strong zonal flow at 500 hPa over the southern part of Greenland. Twelve hours later, the system has moved well to the east and has become more baroclinic, as seen from the well developed temperature advection pattern.
 Fields for the final case study are provided in Figure 13. This is a complex case and as such we show fields from −36 to +36 hours. The fields for −36 hours depict a developing low approaching southern Greenland from the southwest. The system tracks more to the east as compared to the first case study, so that instead of bifurcating at the tip of Greenland, it moves smoothly into the IL region. At −24 hours, the system has strengthened and is centered just to the southeast of Greenland. By −12 hours, it has expanded and shows a well-developed baroclinic pattern. The low further deepens and settles at the lee of Greenland. By 0 hours (the time of maximum deepening) the system has become very broad and a closed low appears at 500 hPa. The low then stalls as it becomes isolated from the advecting flow to the south. A secondary low then begins to develop along the warm front because of preexisting baroclinic conditions and strong vertical motion associated with the larger system. Further development of the secondary low is evident in the fields for +6, +12 and +24 hours. The second system is strongly baroclinic, while the original, still deep system remains stalled in the lee of Greenland and becomes barotropic. Deep low pressure in the IL region is still present at +36 hours, by which time the secondary low has also lost its temperature advection fields.
5.3. Modeling Perspective
 Building on work by Kristjansson and McInnes  and others, we compared the evolution of each case study cyclone in a control run (CONTROL) to a run in which the orography of Greenland is removed (NOGREEN) with Polar MM5. The surface of Greenland is covered with ice in the NOGREEN case. The basic model setup for these Polar MM5 simulations was described in section 2. Polar MM5 was initialized 72 hours prior to time 0 hours for each case. Model fields were output every 6 hours. Fields of sea level pressure, 500 hPa height and 1000–500 hPa thickness for each case at selected times (with the same designation as in the observational analysis) follow in Figures 14–16. Because of difficulties in reducing pressures over the ice sheet to sea level, sea level pressures over the ice sheet in the CONTROL case are masked. In all three case studies, the CONTROL simulations follow quite closely the synoptic evolution seen in the NCEP fields. Recall that fractional sea ice cover is prescribed on the basis of satellite data. Sea ice cover is the same in the CONTROL and NOGREEN runs.
 The first case study (case 1) represents the bifurcation event at the southern tip of Greenland. Maps are presented for 0, +6 and +12 hours (Figure 14). The map for 0 hours illustrates distortion (bifurcation) of the original cyclone at the southern tip of Greenland in the CONTROL case and a classical midlatitude system with well developed temperature advection patterns in the NOGREEN case. The NOGREEN system is located further to the east and is also considerably deeper. Therefore one can say that the orography retards the development and the motion of this system. By time +6 hours the CONTROL system has split into separate lows, while the much deeper NOGREEN low has moved somewhat to the east. By time +12 hours, the system to the west of Greenland in the CONTROL case has dissipated, leaving only the low to the east. The original NOGREEN system remains much stronger and lies further to the north.
Figure 15 compares the CONTROL and NOGREEN runs for the lee cyclogenesis event (case 2). Fields are given for −6, 0 and +6 hours. At −6 hours, the system of interest in the CONTROL run is located along the southwest coast of Greenland. It is quite weak and by time 0 hours it has dissipated, with the new low appearing off the coast of southern Greenland. Note the zonal flow at 500 hPa across the southeastern part of Greenland, suggestive of vortex stretching in the lee. The leeside low then deepens, and, driven by the strong zonal flow, moves fairly rapidly to the east. The evolution of this system in the NOGREEN simulation is fundamentally different. At −6 hours, the system is somewhat further east of its original position in the CONTROL simulation and is also much deeper; in the absence of a barrier, there would be no spin-down of the system in response to reduction in the effective fluid depth on the windward side of the Greenland ice sheet. The NOGREEN system then migrates smoothly over the island, remaining deeper than its CONTROL counterpart.
Figure 16 compares the simulations for the more complex case 3, the strong deepening event in the vicinity of the IL, subsequent stalling of the cyclone, and development of a secondary low. Fields are shown for 0, +18 and +30 hours. Initially, there is little difference between the tracks of the cyclone in the CONTROL and NOGREEN runs. Differences emerge as the system moves into the IL region. By time 0 hours, the cyclone of interest in the CONTROL simulation is located in the vicinity of the IL, while the NOGREEN low is tracking to the north over the ice sheet. By +18 hours, the CONTROL low is still in the IL region and has lost its vertical tilt, while the NOGREEN system is located well to the north, over the center of Greenland, and another separate low is found southeast of Greenland. As in the NCEP fields, the CONTROL fields at +18 hours show development of a secondary cyclone to the northeast of Iceland in the Norwegian Sea. Such a development is also seen in the NOGREEN run, but it is less obvious in Figure 16. By +30 h, the CONTROL simulation shows the initial cyclone still stalled in the vicinity of the IL and vertically stacked, while the second baroclinic system is well to the north, near the coast of Greenland. In NOGREEN, the original system travels over Greenland, while the secondary system moves further north. The NOGREEN simulation also shows yet another system developing near Iceland, at the same location where the original cyclone in the CONTROL run is stalled.
 Tracks of the central pressure for all the three case studies (Figure 17) at every six hours for the duration of each simulation help to summarize the above finding. The case 1 cyclone in the CONTROL simulation bifurcates at the southern tip of Greenland. When Greenland's orography is removed, the cyclone does not bifurcate, but instead tracks smoothly over southern Greenland and then to the northeast along the Greenland coast. The track in the case 2 CONTROL simulation points to leeside cyclogenesis. The “parent low” stalls along the west side of Greenland and dissipates while a new system forms to the southeast in the lee of the ice sheet. By contrast, in the NOGREEN simulation, the system instead tracks smoothly over the ice sheet and then rapidly to the east with the strong advecting flow. The case 3 cyclone in the CONTROL simulation stalls in the vicinity of the IL. Without orography, the system instead tracks over the ice sheet. In both cases, a second low form to the east, but the track of the secondary low differs between the CONTROL and NOGREEN simulations.
Table 1 summarizes the central sea level pressures of the principal cyclones at times −12 to +12 hours. For cases 1 and 2, the NOGREEN system is generally deeper; the orography of Greenland not only strongly influenced the tracks, but weakened the systems. By contrast, the case 3 system tended to be somewhat deeper in the CONTROL simulation than in NOGREEN simulation. This particular storm seems to be somewhat unusual in that it continues to deepen in the IL region even after stalling. However, as seen in Table 2, there is no difference in central pressure between the primary low in CONTROL and NOGREEN when both reach maturity (+18 hours). The secondary system that forms in case 3 follows the same strengthening pattern as in the work by Kristjansson and McInnes  in that the cyclone in the NOGREEN case is (in general) slightly stronger than in the CONTROL simulation (Table 2). However, unlike the Kristjansson and McInnes  case study, the formation of the secondary cyclone is not associated with the orography of Greenland.
Table 1. Central Pressure of Principal Cyclones From −12 to +12 Hours
Table 2. Central Pressure of Case 3 Cyclones (Primary and Secondary) From +6 to +42 Hours
6. Summary and Conclusions
 This paper has documented some of the regional processes that contribute to the synoptic complexity in the sub-Arctic/Arctic North Atlantic. We find that strong deepening of cyclones (>6 hPa/6 hr) is especially common in three regions: (1) south of the tip of Greenland, (2) in the vicinity of the mean IL region, and (3) just south of Svalbard. Composite analyses of strongly deepening systems in these three areas point to the influences of pronounced preexisting low-level temperature gradients that can be linked to ocean currents, the sea ice margin and the ocean-land contrasts. The area of preferred deepening near Svalbard is interesting given its very high latitude (about 77°N). Correspondence with the northernmost penetration of open water in the Atlantic is not coincidental; this location results in strong low-level temperature gradients and a moisture source conducive to system development.
 Impacts of Greenland's orography are complex and we have only touched on some of the major issues. We can identify at least three fairly common synoptic situations in which orographic influences are prominent. One is splitting, or “bifurcation” of cyclones at the southern tip of Greenland. From inspection of numerous NCEP fields, whether or not bifurcation occurs seems to be sensitive to only minor differences in system tracks as they approach Greenland. The second is cyclogenesis in the lee of southern Greenland, typically at the mean IL location. This generally occurs under conditions of strong midtropospheric zonal flow across the southern part of the island, where ice sheet elevations reach 2800 m. The third synoptic situation regards systems moving in from the south that, in tracking east of the southern tip of Greenland, do not bifurcate but instead commonly deepen to the north and east of the IL. Orographic impacts are abundantly evident in simulations using Polar MM5 where control runs are compared to runs for which the orography is removed.
 The region between Greenland and Scandinavia is a key gateway for the transport of moist static energy into the Arctic. It seems likely that without the high topography of the Greenland ice sheet, the nature of these transports would be quite different. A subsequent effort will examine aspects of poleward energy transports into the Arctic with and without Greenland's orography under contrasting phases of the NAO from monthlong MM5 integrations.
 This study was supported by NSF grants OPP-0240948 and OPP-0138018. This includes graduate student support for M. Tsukernik and D. Kindig. John Cassano is thanked for help in setting up the MM5 simulations and comments on the manuscript. Authors also want to thank the anonymous reviewers for providing thoughtful comments and suggestions that improved the manuscript.