Quarterly Journal of the Royal Meteorological Society Q. J. R. Meteorol. Soc. 137: 1773–1789, October 2011 A Orographic influence of east Greenland on a polar low over the Denmark Strait Jón Egill Kristjánsson,a * Sigurdur Thorsteinsson,b Erik W. Kolstadc and Anne-Marlene Blechschmidtd a University of Oslo, Norway Meteorological Office, Reykjavik, Iceland c Bjerknes Centre for Climate Research, Bergen, Norway d NCAS-Weather, Lancaster University, UK *Correspondence to: J. E. Kristjánsson, University of Oslo, Department of Geosciences, P.O.Box 1022, Blindern, Oslo, 0315, Norway. E-mail: jegill@geo.uio.no b Icelandic We present a numerical study of a polar low which hit western Iceland in January 2007, with heavy snowfall and mean wind speeds exceeding 20 m s−1 in several locations. The operational models at the time captured the polar low formation rather well, but there was a large spread in their predictions of the subsequent evolution and track of the polar low. The objective of this study is to investigate possible orographic forcing from Greenland as a trigger for the polar low development. In addition to an analysis of surface observations and satellite imagery, sensitivity studies using HIRLAM were carried out with various degradations of Greenland’s orography, as well as with modifications to the sea-surface temperature (SST), surface roughness and the data assimilation scheme. Despite the presence of an upper-level trough and weak static stability in all the simulations, the polar low development was found to be very sensitive to the presence of the high mountains of eastern Greenland. Whereas the control run captured well the main features of the polar low, simulations with parts of east Greenland’s orography removed gave a southward-displaced polar low which moved rapidly eastward, resulting in substantially underestimated nearsurface winds and snowfall amounts. Setting the orographic heights over all of Greenland to zero led to the complete disappearance of the polar low. On the other hand, artificially increasing the SST by 4 K in the Denmark Strait, reducing the orographic roughness or replacing the four-dimensional variational assimilation scheme (4D-Var) by 3D-Var had only a small effect on the polar low. We suggest that hitherto unreported interactions between the high mountains of east Greenland and polar low development over the Denmark Strait may be more important for polar low formation than katabatic flow from valleys in east Greenland that was c 2011 Royal Meteorological Society highlighted in earlier studies. Copyright Key Words: polar low; Denmark Strait; orographic influence Received 27 October 2010; Revised 7 March 2011; Accepted 24 March 2011; Published online in Wiley Online Library 24 May 2011 Citation: Kristjánsson JE, Thorsteinsson S, Kolstad EW, Blechschmidt A-M. 2011. Orographic influence of east Greenland on a polar low over the Denmark Strait. Q. J. R. Meteorol. Soc. 137: 1773–1789. DOI:10.1002/qj.831 1. Introduction Polar lows are mesoscale weather phenomena that evolve at high latitudes during the winter, in connection with marine c 2011 Royal Meteorological Society Copyright cold air outbreaks (MCAOs) over relatively warm seas. They invariably occur on the cold side of the ‘polar front’, i.e. well inside the polar air mass. In the Northern Hemisphere, the most favoured regions for polar low formation are over 1774 J. E. Kristjánsson et al. the northernmost extent of the warm ocean currents (Gulf Stream, Kuroshio) and in regions with frequent MCAOs, such as over the Labrador, Irminger, Norwegian and Barents Seas, as well as near Japan and over the Sea of Okhotsk (Kolstad, 2011). Recently, substantial changes in polar low frequency in a future warmer climate have been suggested (Kolstad and Bracegirdle, 2008; Zahn and von Storch, 2010), drawing attention to the importance of a better understanding of this intriguing phenomenon. Polar lows usually develop in an atmosphere with weak static stability, strong surface-to-air fluxes of sensible and latent heat and considerable low-level baroclinicity. Some polar lows may appear to be mainly driven by latent heat release, resembling tropical cyclones, hence the term ‘Arctic hurricanes’ introduced by Emanuel and Rotunno (1989). In other cases, the life cycle of a polar low can be largely described as a shallow baroclinic wave in a troposphere with weak static stability and a low tropopause (e.g. Reed and Duncan, 1987). In the 1980s and early 1990s there was a debate in the scientific literature concerning this distinction, but there is now more acceptance of the view that, depending on the atmospheric conditions, some polar lows are mainly convective in nature while others are more baroclinic (Rasmussen and Turner, 2003). Recently, Bracegirdle and Gray (2008) found evidence for a gradual transition from a mainly baroclinic phase to a more convective phase during the life cycles of polar lows over the Nordic Seas. Already in the 1980s it was pointed out that in order to spin up a polar low, some ‘trigger mechanism’ was needed, i.e. some factor that helps to organize the convective elements on scales of 1–10 km into a cyclonic system with a horizontal scale of 100–500 km. It is common to express this trigger in terms of a pre-existing upperlevel potential vorticity anomaly (e.g. Montgomery and Farrell, 1992; Grønås and Kvamstø, 1995). Other triggers have also been suggested, such as for instance orographic effects related to the southern tip of Greenland (Rasmussen, 1981) or the Antarctic Peninsula (Gallée, 1995). Such links to orography seem to be rather uncommon though and, according to a statistical analysis of polar lows off northern Norway by Wilhelmsen (1985), only two out of 32 polar lows that were considered had an orographic trigger. Greenland’s enormous ice sheet, located at high elevation, serves as a huge source of cold air, which is frequently drained down valleys and fjords in the form of katabatic winds or ‘piteraqs’ that frequently reach strengths of 20 m s−1 or more (e.g. Heinemann and Klein, 2002). Klein and Heinemann (2002) suggested that convergence of the outflowing air from the katabatic flow in the valleys of east Greenland might be responsible for the formation of mesocyclones over the Denmark Strait, and found support for this view from model simulations. Due to its size, Greenland also has a major impact on the North Atlantic weather and climate through its influence on storm tracks, as shown in model studies by Petersen et al. (2004), Junge et al. (2005) and Tsukernik et al. (2007). The mechanisms for this interaction include various forms of flow distortion, depending on wind direction (Petersen et al., 2005), resulting in e.g. lee vortex formation (e.g. Petersen et al., 2003), cyclone splitting (Kurz, 2004) and phase-locking (Kristjánsson et al., 2009). In this study, we describe a polar low that hit western Iceland on 11–12 January 2007. The low, which rapidly developed over the Denmark Strait in the early hours of 11 January, was reasonably well captured by the major c 2011 Royal Meteorological Society Copyright numerical weather prediction models, but uncertainties concerning the cyclone track and strength nevertheless made accurate short-term (6–12 hour) forecasting for Iceland very difficult (forecaster-on-duty Óli Thór Árnason, personal communication). Considering the fact that the polar low developed only 400 km south of Greenland’s highest mountain, Mt Gunnbjørn (3700 m elevation), we have explored the possibility for orographic forcing from that feature acting as a trigger. A series of model simulations was carried out, in order to address the following questions: • Was the polar low development linked to a propagating upper-level potential vorticity (PV) anomaly? • What was the role of orographic forcing? • Can we distinguish orographic forcing from noise, due to a large sensitivity of the initial state to random perturbations? In the next section, we describe the synoptic weather situation leading up to and during the polar low event. This is followed by a section describing the model tool that was used for the numerical experiments, as well as the experimental set-up. Section 4 deals with the results from the model simulations, followed by a discussion section. Finally, section 6 summarizes the main features of the study, and presents the conclusions. 2. Synoptic description The polar low formation was preceded by a deep (< 960 hPa) synoptic-scale cyclone that approached southern Iceland on 10 January 2007, moving steadily east-northeast, reaching maximum strength of 951 hPa at 0000 UTC 11 January, then gradually filling over the next 48 hours as it continued its northeasterly track past Iceland (Figure 1). In its aftermath, from 0000 UTC 11 January (Figure 1(c)) onwards, Arctic air was advected over the Denmark Strait and surroundings, creating favourable conditions for polar low formation over the relatively warm waters of the Irminger current west of Iceland. The infrared satellite imagery showed the first sign of an incipient polar low at 0505 UTC on 11 January near 65.5◦ N, 30◦ W (Figure 2(a)). At this stage no clear structure was seen, but rather a distinct north–south oriented cloud band from about 64◦ N to 66◦ N along the 30◦ W meridian (Figure 2(a)). Eight hours later, at 1319 UTC (Figure 2(b)), a well-developed polar low was seen at 65.5◦ N, 27◦ W, west of the Vestfirðir peninsula in northwest Iceland. At this stage, the winds had started to pick up from a south-southeasterly direction over western Iceland and it had started to snow in some areas, e.g. Keflavı́k airport in southwest Iceland. The sounding from there at 1200 UTC (Figure 3(a)) shows high relative humidity, closely following the moist adiabat all the way from about 900 hPa to the tropopause, which was very low at about 475 hPa. All these features are indications of deep moist convection in an Arctic air mass, characteristic of polar lows (e.g. Rasmussen and Turner, 2003). At 1502 UTC, the polar low had deepened further (Figure 2(c)) and the associated southwesterly wind field was now causing heavy snow showers over the whole western part of Iceland (not shown). Over the next eight hours or so, the polar low was almost stationary (Figure 2(d)), so that the strongest winds were still at sea, while sustained Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low (a) (b) (c) (d) 1775 (e) Figure 1. HIRLAM analyses of sea-level pressure (hPa, isolines) and temperature at 700 hPa (K, shaded) at (a) 1200 UTC 10 January 2007; (b) 0000 UTC 11 Jan 2007; (c) 1200 UTC 11 Jan 2007; (d) 0000 UTC 12 Jan 2007; (e) 1200 UTC 12 Jan 2007. southwesterly winds of 10–17 m s−1 were found on the west coast of Iceland (not shown). Scatterometer-based wind speed retrievals by QuikSCAT (Quick Scatterometer) (Figure 4(a)) indicated surface wind speeds of as much as 30 m s−1 at this time, and a similar reading was obtained c 2011 Royal Meteorological Society Copyright by the corresponding QuikSCAT image 12 hours later (not shown). The reliability of QuikSCAT winds in the region near Greenland has been assessed by other studies (e.g. Kolstad, 2008; Renfrew et al., 2009; Winterfeldt et al., 2010), and they seem to agree that strong winds may be Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1776 J. E. Kristjánsson et al. (a) (b) (c) (d) (e) (f) Figure 2. NOAA AVHRR infrared (channel 4) images at (a) 0505 UTC 11 January 2007; (b) 1319 UTC 11 Jan 2007; (c) 1502 UTC 11 Jan 2007; (d) 2310 UTC 11 Jan 2007; (e) 0455 UTC 12 Jan 2007; (f) 1256 UTC 12 Jan 2007. The white line (red in the online version) in (b) indicates the position of the cross-sections in Figures 11 and 12. This figure is available in colour online at wileyonlinelibrary.com/journal/qj overestimated. Still, the spatial distribution of QuikSCAT winds seems to be reliable enough, at least for use in case-studies such as this one. The Keflavı́k airport sounding at 0000 UTC 12 January (Figure 3(b)) was distinctly different from the one 12 hours earlier. While the air in the lowest 250 hPa of the atmosphere was still well-mixed and rather humid, the remainder of the troposphere was now dry and much warmer than before. This suggests that at this time strong subsidence was taking place in the lee of Greenland in the vigorous westerly flow that was now found through the whole troposphere (Figure 3(b)) and lower stratosphere (not shown). In the early hours of 12 January, the polar low moved slowly c 2011 Royal Meteorological Society Copyright eastward (Figure 2(e)), hitting the coast of northwest Iceland at 1200 UTC (Figure 2(f)). At this time a surface pressure measurement of 962 hPa was taken just ahead of the polar low at Bjargtangar (location indicated by B in Figure 4(b)). This reading was about 10 hPa lower than in the HIRLAM analysis (Figure 1(e)). As the polar low made landfall on 12 January, it rapidly weakened (Figure 2(f)), but nevertheless heavy snow showers, as well as winds exceeding 20 m s−1 , were observed in several locations in northwest Iceland on that day (Figure 4(b)). Interestingly, at 2246 UTC (not shown), after the polar low had dissipated, satellite images showed several new mesoscale vortices in the same area west of Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low (a) 1777 (b) Figure 3. Skew-T diagrams displaying radiosonde soundings from Keflavı́k, Iceland (64.0◦ N, 22.6◦ W) at (a) 1200 UTC 11 January 2007; (b) 0000 UTC 12 Jan 2007. (Figures obtained from the University of Wyoming). Iceland where the original polar low formed, but none of these developed into a polar low. 3. Model and experimental set-up In order to investigate the role of various factors for the evolution of the polar low on 11–12 January 2007, several simulations were carried out using HIRLAM (HIgh Resolution Limited-Area Model), version 7.2. The simulations used HIRLAM analyses as initial conditions, and 6-hourly forecasts from the European Centre for Medium-range Weather Forecasts (ECMWF) were used at the lateral boundaries. The model grid mesh consisted of 306 × 306 horizontal grid points at 22 km grid spacing and 40 vertical levels, covering an area consisting of northern Europe, the northern North Atlantic and the north-easternmost part of the Canadian Arctic. The HIRLAM analyses were based on three-dimensional variational assimilation (3D-Var: Gustafsson et al., 2001; Lindskog et al., 2001) for one experiment, while fourdimensional variational assimilation (4D-Var: Huang et al., 2002; Gustafsson, 2006) was used for the other five experiments. The HIRLAM 4D-Var applies a multiincremental minimization (Veersé and Thépaut, 1998) and includes the simplified physical parametrization scheme of Janisková et al. (1999). Conventional observational data as well as satellite data from the Advanced Microwave Sounding Unit A (AMSU-A) of the Advanced TIROS Operational Vertical Sounder (ATOVS) were assimilated in a 6 h assimilation cycle. The HIRLAM grid-point forecast model is hydrostatic, and it utilizes a semi-implicit, semiLagrangian two-time-level time integration scheme (Undén et al., 2002). The physical parametrizations used were, for example, the radiation scheme of Savijärvi (1990), the Cuxart–Bougeault–Redelsperger (CBR) turbulence scheme (Cuxart et al., 2000), the Kain–Fritsch convection scheme (Kain, 2004), the Rasch–Kristjánsson (1998) prognostic cloud water scheme, and the Interaction Soil-BiosphereAtmosphere (ISBA: Noilhan and Mahfouf, 1996) surface scheme. Surface friction is treated using a surface roughness parametrization, which has separate formulations over sea c 2011 Royal Meteorological Society Copyright (Charnock’s formula), over vegetation and over orography (Undén et al., 2002). In addition to a CONTROL run in which all the model features were as described above, a series of sensitivity simulations was carried out. Firstly, three simulations were made to investigate the sensitivity to the orography of Greenland; in NOGREEN all orographic heights over Greenland were set to 0 m above sea level; in NOEAST the orographic heights over the easternmost part of Greenland were set to 0 m, while other parts of Greenland were left intact; in NOGUNN only orographic heights around Mt Gunnbjørn in eastern Greenland were set to 0 m. The different orographic height fields are shown in Figure 5(a)–(d). To explore a possible link between the orographic effects and the model’s formulation of orographic roughness, a simulation was carried out (SMOOTH), in which the model’s orographic roughness length was reduced to 1% of the nominal values over all land areas. From Figure 5(a), we see that this would be expected to mainly influence the flow over Greenland, and to a lesser extent over Iceland. We also investigated the sensitivity to the data assimilation scheme. In the simulation called 3DVAR, the fourdimensional data assimilation was replaced by the threedimensional variational data assimilation scheme. Finally, a simulation was carried out in which the sea-surface temperatures (SSTs) in a rectangular area (65◦ –68◦ N, 25◦ –35◦ W) west of northwest Iceland were increased by 4 K. This simulation will be referred to as SST + 4. 4. Results from the HIRLAM analyses and simulations 4.1. Static stability and upper-level conditions As deep convection is one of the main ingredients of mature polar lows, the low-level static stability is a good indicator of polar low potential (e.g. Kolstad, 2006, 2011). Empirical data suggest that the temperature difference between the surface (i.e. SST) and at 500 hPa (T500 ) tends to be well above 40 K upon polar low formation (e.g. Noer and Ovhed, 2003) in Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1778 J. E. Kristjánsson et al. (a) (b) Figure 4. Observed wind speeds (m s−1 ) associated with the polar low: (a) QuikSCAT level 2 winds (coloured arrows; resolution of 12.5 km; obtained from Remote Sensing Systems) around 2030 UTC on 11 January 2007; (b) maximum observed sustained (10-minute average) winds at 10 m height at various locations in Iceland on 12 January 2007. Letters V, K and R refer to geographical locations mentioned in the text: V = Vestfirðir peninsula; B = Bjargtangar; K = Keflavı́k airport; R = Reykjavı́k. the Nordic Seas region. Kolstad et al. (2009) chose to express the criterion in terms of the potential temperature difference between 700 hPa and the surface: the so-called MCAO index. The majority of polar lows investigated by Blechschmidt et al. (2009) reached a temperature difference SST-T 500 of 48 K. Compared to polar lows in other parts of the Nordic Seas, the strongest anomalies of this parameter were found for polar lows that developed near Greenland. In Figure 6 we show SST-T 500 from HIRLAM analyses at different times. Starting with the evening of 10 January, about 12 hours before the formation of the polar low, we see (Figure 6(a)) that the values all around Iceland were in the range 30–40 K, except for a small area near 65◦ N, 35◦ W and another one further southwest, with values between 40 and 45 K. Twelve hours later, just one hour after the first clear signs of a polar low initiation in the satellite imagery (Figure 2(a)), much larger values were found, and in the area of the incipient polar low the values were higher than 48 K (Figure 6(b)), in excellent agreement with Blechschmidt et al..’s suggestion. The temperature difference then decreased somewhat, but nevertheless values well over 40 K persisted for another c 2011 Royal Meteorological Society Copyright 30 hours in the area around the polar low (Figure 6(c) and (d)), thereafter gradually decreasing as the polar low dissipated (not shown). As discussed in the introduction, another crucial ingredient in polar low developments is the existence of an upper-level PV anomaly that can serve as a trigger. In order to investigate whether such a trigger was present, we studied the analysed upper-level PV every 6 hours from 0000 UTC on 8 January, i.e. almost 2.5 days before the polar low formation and until it dissipated at 1800 UTC on 12 January. A subset of these results, along with the height of the 500 hPa pressure level is shown in Figure 7(a)–(f). In the days preceding the polar low event, there was a rather weak (1–2 PVU) west–east oriented upper-level PV anomaly over Greenland near 70◦ N (Figure 7(a) and (b)). The westernmost branch of this anomaly, which was associated with a trough at 500 hPa, was located at 70◦ N, 50◦ W at 1200 UTC on 9 January (Figure 7(a)), then gradually moved southeastwards and increased in strength, so that at 0000 UTC 11 January, just about the time when the polar low started forming, a rather sharp trough was found along the coast of eastern Greenland at 65◦ N, 40◦ W (Figure 7(c)). This is in very good agreement with Fig. 6 (top right) of Blechschmidt et al. (2009). Over the following 24 hours the trough and the associated PV anomaly continued their cyclonic progression and deepened further, possibly due to mutual interaction with the developing polar low (Figure 7(d) and (e)). The upper-level features gradually became more vertically aligned with the polar low at the surface (comparing Figures 1(d) and 7(e)), as expected in a baroclinic development. In summary, it is clear that both the conditions at the surface and near the tropopause were favourable for a polar low development west of Iceland on 11 January 2007. Consequently, one might expect the polar low development to be a foregone conclusion and that it would require drastic changes in the initial or boundary conditions to significantly alter the course of events. 4.2. Sensitivity runs In order to understand the possible role of Greenland’s orography in triggering the polar low, we start by investigating the evolution of sea-level pressure in the six simulations (Figures 8–10). First at +24 h, we see that while the CONTROL run has a distinct polar low in approximately the correct position at 66◦ N, 29◦ W (Figure 8(a)), large deviations from this are found in the three simulations with degraded Greenland orographies (Figure 8(b)–(d)). In the NOGREEN case, the result (Figure 8(b)) is strikingly similar to the results of previous studies by Kristjánsson and McInnes (1999) and Skeie et al. (2006), displaying a strong dipole of orographically enhanced surface pressure over northeast Greenland and a corresponding reduction of as much as −25 hPa in the region around the Denmark Strait. There is no polar low present at either +24 h or at +36 h in the NOGREEN simulation (Figures 8(b), 9(b) and 10(b)). This indicates that the very existence of the polar low is dependent on the presence of Greenland’s orography, in a similar way as the ‘residual low’ in the study by Kristjánsson and McInnes (1999). The removal of all of Greenland’s orography is a very drastic perturbation, so we now investigate to what extent only parts of the orography may play a role. Hence, in Figures 8(c), 9(c) and 10(c) we compare the results of the NOEAST simulation to those of CONTROL. In the NOEAST Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low (a) (b) (c) (d) 1779 Figure 5. Orographic heights above sea level (m) in simulations (a) CONTROL; (b) NOGREEN; (c) NOEAST; (d) NOGUNN. simulation, the sea-level pressure is several hPa higher than in CONTROL in the area where the polar low was located both in reality and in CONTROL (Figures 8(c) and 9(c)). An opposite signal is found north of Iceland, where the high mountains of east Greenland cause the sea-level pressure to be several hPa higher in CONTROL than in NOEAST. A more moderate orographic modification is imposed in the NOGUNN simulation (Figure 5(d)), but even so, the polar low is greatly weakened also in this case (Figures 8(d) and 9(d)). The pressure signals are quite similar to those in NOEAST, but with a slightly smaller amplitude. The results from the SMOOTH simulation (not shown) exhibit very small differences compared to those of CONTROL. This means that the orographic effect indicated by NOGREEN, NOEAST and NOGUNN is not related to the enhanced surface friction of the mountains, but rather to the general flow distortion induced by them. Returning now to the discussion at the end of section 4.1, all the experiments except NOGREEN exhibit large similarities in both the upper-level PV and tropospheric c 2011 Royal Meteorological Society Copyright static stability (not shown) to the corresponding fields in the analysis and the CONTROL run (cf. Figures 6–7). This indicates that the polar low is a shallow disturbance, most pronounced in the lowermost part of the troposphere, and that the features in the upper-level PV and static stability are not strongly affected by the polar low development. Rather, they help set the stage for such a development. In NOGREEN, on the other hand, the PV anomaly moves much more rapidly eastwards than in the other simulations (not shown), which does not allow the polar low enough time to develop. Such an eastward propagation was explored in detail by Kristjánsson et al. (2009) in a case of a synoptic-scale lee cyclone over the Denmark Strait. Before considering orographic influence as the true cause of the results from simulations NOGREEN, NOEAST and NOGUNN, one must also consider the possibility that these results are a pure coincidence. It is conceivable that the atmospheric state near the initial time on 10 January 2007 was so sensitive to perturbations that almost any random perturbation of the initial state or the boundary conditions Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1780 J. E. Kristjánsson et al. (a) (b) (c) (d) Figure 6. The temperature difference between the sea surface and 500 hPa (K) from HIRLAM analyses at different times: (a) 1800 UTC 10 January 2007; (b) 0600 UTC 11 Jan 2007; (c) 1800 UTC 11 Jan 2007; (d) 0600 UTC 12 Jan 2007. would have yielded a strong response. In order to explore this possibility, and thereby avoid a misinterpretation of the results shown so far, results from the experiments 3DVAR and SST + 4 will now be discussed. It turns out that, in both cases, the sensitivity is quite small compared to the sensitivity of modifying east Greenland’s orography. First, in the case of changes to the data assimilation scheme (3DVAR), we find the largest impact at the time of analysis, diminishing with increasing forecast length (Table I; Figures 8(e) and 9(e)). This can be interpreted as being caused by the importance of making use of the observations at the right time in the analysis, which is done better in 4D-Var than in 3D-Var. Further into the simulation the boundaries have an increasing impact on the simulation, with reduced importance of how the observations were used initially. In this simulation alone, and in contrast to what was observed, a secondary polar low southwest of Iceland developed on 11 January, causing the positive anomaly in that area seen in Figure 8(e). Nevertheless, apart from a somewhat delayed deepening (Table I), the main polar low developed similarly to that in the CONTROL run. As polar lows are partly driven by surface fluxes, increasing the SST by 4 K as in simulation SST + 4 has the potential to significantly deepen the polar low, through increased fluxes of both sensible and latent heat. However, despite an enhancement of the sensible and latent heat fluxes of about 50–100 W m−2 each in the area of enhanced c 2011 Royal Meteorological Society Copyright SSTs (not shown), only rather modest changes in the polar low evolution were obtained in the SST + 4 simulation, compared to CONTROL (Figures 8(f) and 9(f)). We then repeated the experiment but with even larger perturbations of the SST, i.e. 8 K and 12 K, respectively. A considerably larger response was then found, with the polar low deepening by 5 hPa, relative to CONTROL at +36 h in the former case and by 8 hPa in the latter case (Table I). However, clearly an 8 K or 12 K enhancement of the SST in such a small area is a much larger perturbation than the uncertainty in the initial state would represent, and even a 4 K enhancement is probably excessive (e.g. Garand, 2003). For more detailed studies of the sensitivity of polar low development to SST, we refer the reader to two recent studies of the influence of SST on polar low development, which show widely different results: Linders et al. (2011), using an axisymmetric model, found a rather weak sensitivity to SST variations of −0.6 hPa maximum deepening per degree warming, while Adakudlu and Barstad (2011) found a much larger sensitivity of −2 hPa/K in their simulations of a Barents Sea polar low during the 2008 International Polar Year–THe Observing system Research and Predictability EXperiment (IPY-THORPEX) Andøya campaign. By comparison, our results correspond to a varying sensitivity of −0.5, −0.6 and −1.0 hPa/K, respectively for the SST + 4, SST + 8 and SST + 12 experiments. Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low (a) (b) (c) (d) (e) (f) 1781 Figure 7. HIRLAM analyses of geopotential height at 500 hPa (isolines, every 50 m) and potential vorticity at 300–500 hPa (shaded) : (a) 1200 UTC 9 January 2007; (b) 1200 UTC 10 Jan 2007; (c) 0000 UTC 11 Jan 2007; (d) 1200 UTC 11 Jan 2007; (e) 0000 UTC 12 Jan 2007; (f) 1200 UTC 12 Jan 2007. In order to explore the results of the sensitivity runs in more detail, Figure 10 shows near-surface winds and accumulated precipitation, in addition to sea-level pressure at 0000 UTC 12 January (+36 h simulation time). Looking first at the sea-level pressure, we see that only the runs CONTROL, 3DVAR and SST + 4 (Figure 10(a), (e) and (f)) have a well-developed polar low in approximately the correct position near Vestfirðir peninsula. All three simulations have 12 h accumulated precipitation between 4 and 16 mm over large areas of western and southern Iceland, west of the Vestfirðir peninsula, as well as in the westerly flow over the warm waters off the south coast. These results are c 2011 Royal Meteorological Society Copyright in quite good agreement with observations (not shown), taking into account the lack of observations over the sea and the well-known underestimation from conventional precipitation measurements in windy conditions with dry snow. In the SST + 4 run the low is about 2 hPa deeper than in CONTROL, and the associated wind and precipitation fields are somewhat stronger than in CONTROL (Figure 10(f)), but otherwise the main features are very similar to those in the CONTROL run (Figure 10(a)). Much larger differences are found in the simulations with degraded orography: Firstly, as noted in connection with Figures 8 and 9, the polar low is completely absent in the NOGREEN run, and Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1782 J. E. Kristjánsson et al. (a) (b) (c) (d) (e) (f) Figure 8. Sea-level pressure at 1200 UTC 11 January 2007 (+24 h): (a) CONTROL; (b) CONTROL minus NOGREEN; (c) CONTROL minus NOEAST; (d) CONTROL minus NOGUNN; (e) CONTROL minus 3DVAR; (f) CONTROL minus SST + 4. The contour interval is 4 hPa in (a) and 2 hPa in the other panels. we note in Figure 10(b) the greatly suppressed precipitation over Iceland. The near-surface wind is influenced to a lesser extent than the precipitation, because in the absence of Greenland’s orography and the polar low, an unrealistically strong northwesterly airflow emanating from Greenland impinges on Iceland, creating jets along the coast and in the lee of Iceland (Figure 10(b)). In the NOEAST run the polar low, in addition to being far too weak, takes a much c 2011 Royal Meteorological Society Copyright too southerly course, hitting Reykjavik in southwest Iceland at 1800 UTC on 11 January (not shown), and moving rapidly eastward. Therefore, in Figure 10(c), the heaviest precipitation and strongest winds are located offshore with unrealistic dry and calm conditions over much of west and southwest Iceland. In the NOGUNN simulation, only a trough forms west of Iceland, while unrealistically a closed low forms north of Iceland, in agreement with the Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low (a) (b) (c) (d) (e) (f) 1783 Figure 9. Sea-level pressure at 0000 UTC 12 January 2007 (+36 h): (a) CONTROL; (b) CONTROL minus NOGREEN; (c) CONTROL minus NOEAST; (d) CONTROL minus NOGUNN; (e) CONTROL minus 3DVAR; (f) CONTROL minus SST + 4. The contour interval is 4 hPa in (a) and 2 hPa in the other panels. positive sea-level pressure anomaly north of Iceland in Figures 8(d) and 9(d). This secondary feature is clearly seen in the precipitation and wind patterns north of Iceland in Figure 10(d), while over northwest Iceland the weather is relatively calm and dry, in stark contrast to what was actually observed at this time. orography of Greenland was set to zero, did not produce any trace of a polar low, while the polar low was greatly weakened in the simulations with a degradation of east Greenland’s orography: NOEAST and NOGUNN. Is this because the MCAO from Greenland onto the warm sea surface fails to materialize in these three simulations or is it because the absence of orographic features leads to a 5. Discussion low-level vorticity deficit in the region where the polar low forms? These questions are relevant not just for polar lows, The results presented above show that the topography of but for synoptic-scale cyclones as well. In Figure 11, we explore the cold air outbreak associated Greenland had a crucial influence on the development of the polar low. The NOGREEN simulation, in which the with the polar low in the CONTROL run. It turns out that c 2011 Royal Meteorological Society Copyright Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1784 J. E. Kristjánsson et al. Table I. Simulated sea-level pressure in the centre of the polar low near northwest Iceland on 11–12 January 2007. hPa +24 h +30 h +36 h +42 h CONTROL NOGREEN NOEAST NOGUNN 3DVAR SMOOTH SST + 4 SST + 8 SST + 12 975 – (975) (977) 978 975 975 973 967 975 – (972) (976) 977 975 973 970 963 974 – (976) (978) 975 973 972 969 966 974 – – – 975 974 973 972 973 The initial time for all the simulations is 1200 UTC 10 January 2007. In NOGREEN no polar low developed, therefore no values are given. In NOEAST and NOGUNN, only a rapidly eastward-moving polar low some 200 km further south was obtained, and the values in parentheses are for this polar low. the cold air below 800 hPa from east Greenland was advected toward the south-southeast more or less along the section, and this transport was amplified between 0000 UTC and 0600 UTC on 11 January (Figure 11(a) and (b)), during the incipient stage of the polar low. At 1200 UTC, when the polar low was rapidly developing, we note that below 800 hPa the cold air outbreak had come to a halt (Figure 11(c)), while there was (not shown) an increased westerly flow of warmer air perpendicular to the cross-section. In the northern part of the section, between about 650 hPa and 850 hPa, there is a tendency for northerly flow and sinking motion. These indications of a lee effect associated with Greenland are much less pronounced when we take away the east Greenland orography (Figure 12(a) and (b)). Furthermore, compared to NOEAST (Figure 12(a)) and NOGUNN (Figure 12(b)), we note that above 850 hPa the potential temperature is higher in CONTROL (Figure 11(c)), especially in the northern part of the section, i.e. in the vicinity of the developing polar low. Twelve hours later (Figure 11(d)), corresponding to the time of the sounding in Figure 3(b), we note that there is now strong rising motion in the northern part of the section (near 66◦ N, 25◦ W), in association with the polar low. Here the troposphere is well-mixed, all the way up to 500 hPa, while further south (to the right in Figure 11(d)) the mixed layer containing snow showers is gradually shallower due to a ridge of high pressure here (Figure 1(e)). A very different situation is found in the NOEAST and NOGUNN simulations (Figure 12(c) and (d)), which below 800 hPa display strong cold advection throughout and a strongly stratified troposphere above 750 hPa (especially in NOEAST), in the absence of the polar low. In NOGUNN, a pronounced sinking motion is found in the southern part of the section, but that is unrelated to Greenland. Having seen the large sensitivity of the polar low evolution to Greenland’s orography, despite the favourable conditions both at the surface and at upper levels, one may ask to what extent these conditions are modified in the simulations with degraded orography. Therefore, we show and compare in Figure 13, from the six simulations, the upper-level height field and the low-level temperature at the time of maximum strength of the polar low at 1800 UTC on 11 January. Firstly, in the CONTROL run (Figure 13(a)), there is at 500 hPa a pronounced trough along Greenland’s east coast northwest of Iceland, while the 850 hPa temperature field shows a well-defined tongue of warm air west of Iceland (where the c 2011 Royal Meteorological Society Copyright polar low develops), as well as another tongue of cold air stretching eastward from the coast of Greenland at about 65◦ N, 35◦ W. In the NOGREEN simulation (Figure 13(b)), the 500 hPa trough is much stronger than in CONTROL, while the cold air advection at 850 hPa is far more advanced than in the CONTROL run. In NOGREEN, the warm tongue found in CONTROL is replaced by a wedge of cold air that is being effectively advected from Greenland. This big difference in the ability of cold air to advance from Greenland toward Iceland was also found in earlier studies (Kristjánsson and McInnes, 1999; Kristjánsson et al., 2009). As a consequence of Greenland’s high elevation the cold air there can only be brought toward Iceland after warming it adiabatically, which makes it warmer than the surroundings, thereby lowering the surface pressure. In this way the cold air over Greenland becomes isolated in a way that would not happen in the absence of the high elevation, but with the same degree of coldness (as in NOGREEN). This also to some extent explains the tendency of Greenland to create cyclones in its vicinity. While previous model studies removing Greenland’s orography have demonstrated a strong influence on synoptic-scale cyclones off southeast Greenland (e.g. Kristjánsson and McInnes, 1999; Skeie et al., 2006), in this study we have obtained a similar result for a polar low. The features that distinguish such a ‘polar lee low’ from the lee lows studied earlier are: (i) that it forms in the Arctic air mass poleward of the main baroclinic zone, rather than due to interaction with the main baroclinic zone; (ii) that the static stability is weak and the tropopause low, resulting in a mesoscale system, rather than a synoptic-scale low (Montgomery and Farrell, 1992). Interestingly, in NOEAST and NOGUNN, which failed completely in simulating a polar low with any resemblance to that observed, the 500 hPa height field is not dramatically different from the corresponding fields in CONTROL. The trough along east Greenland at 500 hPa in CONTROL is replaced by a deeper, more circular low in both NOEAST (Figure 13(c)) and NOGUNN (Figure 13(d)), probably due to stronger cold advection in this area, as the obstacle provided by the orography of eastern Greenland is removed. On the other hand, in both NOEAST and NOGUNN the 850 hPa temperature field is significantly different from CONTROL, with warmer air off northeast Iceland than in CONTROL, probably in connection with the secondary meso-cyclone there, and colder air west of the Vestfirðir peninsula which is where the polar low is present in Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low (a) (b) (c) (d) (e) (f) 1785 Figure 10. Sea-level pressure (black isolines, every 4 hPa), wind speed at 10 m height (blue dashed isolines, every 4 m s−1 ) and accumulated precipitation over 12 h preceding the simulation time (colour shading) at 0000 UTC 12 January 2007 (+36 h): (a) CONTROL; (b) NOGREEN; (c) NOEAST; (d) NOGUNN; (e) 3DVAR; (f) SST + 4. CONTROL, but absent in NOEAST and NOGUNN at this time (viz. Figures 8 and 9). In 3DVAR, on the other hand (Figure 13(e)), which had a surface pressure field similar to CONTROL, the 500 hPa trough over east Greenland is somewhat weaker than in CONTROL, whereas the 850 hPa temperature field c 2011 Royal Meteorological Society Copyright is similar to that in CONTROL. In the SST + 4 run the 500 hPa height field (Figure 13(f)) is almost identical to that of CONTROL, while the 850 hPa temperature field shows a slightly more pronounced warm tongue west of Iceland, as might be expected from the enhanced surface fluxes. Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1786 J. E. Kristjánsson et al. (a) (b) (c) (d) Figure 11. Potential temperature (red isolines, every 2 K) and the velocity component along the section (black arrows) in the cross-section between 68◦ N, 26◦ W and 61◦ N, 22◦ W (cf. Figure 2(b)), from the CONTROL run at: (a) 0000 UTC 11 January 2007 (+12 h); (b) 0600 UTC 11 Jan 2007 (+18 h); (c) 1200 UTC 11 Jan 2007 (+24 h); (d) 0000 UTC 12 Jan 2007 (+36 h). To summarize, the large sensitivity to the orography of east Greenland does not seem to be caused by different pressure fields at upper levels. Rather the differences there are likely to be caused by the different low-level flows that result from the differences in orography. The sensitivity of the polar low development to east Greenland’s orography appears to be a lee effect associated with northerly flow interacting with the steep orography associated with Greenland’s highest mountains near 69◦ N, 30◦ W, northwest of Iceland. 6. Summary and conclusions A polar low that struck the western part of Iceland on 11–12 January 2007 has been investigated using available observations in the area, model analyses and dedicated simulations with the HIRLAM numerical weather prediction model. The polar low developed in an Arctic air mass in the aftermath of a deep synoptic-scale cyclone moving northeast past Iceland. The polar low had the characteristic features of polar lows in this area, previously documented by Blechschmidt et al. (2009), i.e. a temperature difference between the surface and 500 hPa of more than 48 K and an upper-level PV anomaly approaching from the northwest. c 2011 Royal Meteorological Society Copyright The control model run gave a polar low evolution that was quite close to the observed one. Artificially enhancing the sea-surface temperature in the area of polar low development by 4 K had a small effect on the polar low, and this was also the case for a simulation in which the orographic roughness was reduced by a factor of 100, as well as a run in which the initial state was based on 3D-Var, instead of the operational 4D-Var data assimilation scheme. Conversely, in three simulations in which the orography of Greenland was degraded, the polar low was greatly weakened or even absent. Interestingly, the simulations revealed a particular sensitivity to the area of eastern Greenland northwest of Iceland, the site of Greenland’s highest mountain (Mt Gunnbjørn at 3700 m), only about 50 km from Greenland’s east coast. Cross-sections through the air masses between Greenland and Iceland revealed features that are known to characterize lee cyclone formation, such as flow away from Greenland and adiabatic warming of sinking air. In the absence of east Greenland’s orography, these ‘lee cyclone’ features are absent, and there is much stronger cold advection between Iceland and Greenland, while the upper-level flow is quite similar in all the simulations. This leads us to conclude that the orography provides Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low 1787 (b) (a) (c) (d) Figure 12. Potential temperature (red isolines, every 2 K) and velocity wind component along the section (black arrows) in the cross-section between 68◦ N, 26◦ W and 61◦ N, 22◦ W (cf. Figure 2(b)) at 1200 UTC 11 January 2007 (+24 h) from: (a) run NOEAST; (b) run NOGUNN, and at 0000 UTC 12 Jan 2007 (+36 h) from: (c) run NOEAST; (d) run NOGUNN. a trigger that is needed in order to spin up the lowlevel circulation into a vigorous polar low. We further hypothesize that flow distortion associated with this part of Greenland, possibly in the form of subsidence with associated adiabatic warming and vortex stretching, may play a larger role in polar low formation east of Greenland than katabatic flows, previously suggested to be a trigger by Klein and Heinemann (2002). More studies are needed to test this hypothesis, and to explore in more detail the exact mechanism for the orographic triggering. Unfortunately, no Denmark Strait polar lows were captured during the Greenland Flow Distortion experiment (Renfrew et al., 2008), which would otherwise have been a useful test bed for such a hypothesis. While this study has sought to provide new insight into the trigger mechanisms for polar low developments, it has not explicitly dealt with forecasting improvements. A possible follow-up would be to investigate the importance of model resolution for simulations of such ‘polar lee lows’, because it is clear that the ability to resolve Mt Gunnbjørn is resolutiondependent. A recent case study by McInnes et al. (2011) found a significant sensitivity to model resolution in the polar low simulations, but in that study orography was not of c 2011 Royal Meteorological Society Copyright importance. Another issue worth exploring is the sensitivity to the choice of lateral boundary conditions, as the different prediction centres may have different representations of Greenland’s orography. Acknowledgements This study was supported by the Norwegian Research Council’s project ‘THORPEX-IPY: Improved forecasting of adverse weather in the Arctic – present and future‘ (grant no. 175992). The first author would like to thank Hans von Storch for helpful suggestions that led to significant improvements in the experimental set-up. QuikSCAT data were obtained from the Physical Oceanography Distributed Active Archive Center (PO.DAAC) at the NASA Jet Propulsion Laboratory, Pasadena, California (http://podaac.jpl.nasa.gov). We acknowledge support from the Swedish Meteorological and Hydrological Institute (SMHI), concerning computer power, HIRLAM and graphics. We thank Laura Rontu for advice concerning the set-up of the SMOOTH experiment. Two anonymous reviewers are thanked for constructive comments that led to improvements of the manuscript. Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) 1788 J. E. Kristjánsson et al. (a) (b) (c) (d) (e) (f) Figure 13. Simulated 500 hPa heights (isolines, every 60 m) and 850 hPa temperature (colour shading, K) at 1800 UTC 11 January 2007 (+30 h) in: (a) CONTROL run; (b) run NOGREEN; (c) run NOEAST; (d) run NOGUNN; (e) run 3DVAR; (f) run SST + 4. References Adakudlu M, Barstad I. 2011. Impacts of the ice-cover and sea-surface temperature on a polar low over the Nordic seas: a numerical case study. Q. J. R. Meteorol. Soc. 137: 1716–1730, DOI: 10.1002/qj.856. Blechschmidt A-M, Bakan S, Graßl H. 2009. Large-scale atmospheric circulation patterns during polar low events over the Nordic seas. J. Geophys. Res. 114: D06115, DOI: 10.1029/2008JD010865. Bracegirdle TJ, Gray SL. 2008. An objective climatology of the dynamical forcing of polar lows in the Nordic Seas. Int. J. Climatol. 28: 1903–1919. Cuxart J, Bougeault P, Redelsperger J-L. 2000. A turbulence scheme allowing for mesoscale and large-eddy simulations. Q. J. R. Meteorol. Soc. 126: 1–30. c 2011 Royal Meteorological Society Copyright Emanuel KA, Rotunno R. 1989. Polar lows as Arctic hurricanes. Tellus 41A: 1–17. Gallée H. 1995. Simulation of the mesocyclonic activity in the Ross Sea, Antarctica. Mon. Weather Rev. 123: 2051–2069. Garand L. 2003. Toward an integrated land–ocean surface skin temperature analysis from the variational assimilation of infrared radiances. J. Appl. Meteorol. 42: 570–583. Grønås S, Kvamstø NG. 1995. Numerical simulations of the synoptic conditions and development of Arctic outbreak polar lows. Tellus 47A: 797–814. Gustafsson N. 2006. ‘Status and performance of HIRLAM 4D-Var.’ HIRLAM Newsletter 51: 8–17. Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011) Orographic Influence of East Greenland on a Polar Low Gustafsson N, Berre L, Hörnquist S, Huang X-Y, Lindskog M, Navascués B, Mogensen KS, Thorsteinsson S. 2001. Threedimensional variational data assimilation for a limited area model. Part I: General formulation and the background error constraint. Tellus 53A: 425–446. Heinemann G, Klein T. 2002. Modelling and observations of the katabatic flow dynamics over Greenland. Tellus 54A: 542–554. Huang X-Y, Yang X, Gustafsson N, Mogensen KS, Lindskog M. 2002. ‘Four-dimensional variational data assimilation for a limited area model.’ HIRLAM Technical Report 57, 44 pp. Available from HIRLAM-5, c/o Per Undén, SMHI, S-60176 Norrköping, Sweden. Janisková M, Thépaut J-N, Geleyn J-F. 1999. Simplified and regular physical parameterizations for incremental four-dimensional variational assimilation. Mon. Weather Rev. 127: 26–45. Junge MM, Blender R, Fraedrich K, Gayler V, Luksch U, Lunkeit F. 2005. A world without Greenland: Impacts on the Northern Hemisphere winter circulation in low- and high-resolution models. Clim. Dyn. 24: 297–307. Kain JS. 2004. The Kain–Fritsch convective parameterization: An update. J. Appl. Meteorol. 43: 170–181. Klein T, Heinemann G. 2002. Interaction of katabatic winds and mesocyclones near the eastern coast of Greenland. Meteorol. Appl. 9: 407–422. Kolstad EW. 2006. A new climatology of favourable conditions for reversed-shear polar lows. Tellus 58A: 344–354. Kolstad EW. 2008. A QuikSCAT climatology of ocean surface winds in the Nordic seas: Identification of features and comparison with the NCEP/NCAR reanalysis. J. Geophys. Res. 113: D11106, DOI: 10.1029/2007JD008918. Kolstad EW. 2011. A global climatology of favourable conditions for polar lows. Q. J. R. Meteorol. Soc. 137: this issue (in press). Kolstad EW, Bracegirdle TJ. 2008. Marine cold-air outbreaks in the future: An assessment of IPCC AR4 model results for the Northern Hemisphere. Clim. Dyn. 30: 871–885. Kolstad EW, Bracegirdle TJ, Seierstad IA. 2009. Marine coldair outbreaks in the North Atlantic: Temporal distribution and associations with large-scale atmospheric circulation. Clim. Dyn. 33: 187–197. Kristjánsson JE, McInnes H. 1999. The impact of Greenland on cyclone evolution in the North Atlantic. Q. J. R. Meteorol. Soc. 125: 2819–2834. Kristjánsson JE, Thorsteinsson S, Røsting B. 2009. Phase-locking of a rapidly developing extratropical cyclone by Greenland’s orography. Q. J. R. Meteorol. Soc. 135: 1986–1998. Kurz M. 2004. On the dynamics of the splitting process of cyclones near southern Greenland. Meteorol. Z. 13: 143–148. Linders T, Saetra Ø, Bracegirdle TJ. 2011. Limited polar low sensitivity to sea-surface temperature. Q. J. R. Meteorol. Soc. 137: 58–69. Lindskog M, Gustafsson N, Navascués B, Mogensen KS, Huang X-Y, Yang X, Andrae U, Berre L, Thorsteinsson S, Rantakokko J. 2001. Three-dimensional variational data assimilation for a limited area model. Part II: Observation handling and assimilation experiments. Tellus 53A: 447–468. McInnes H, Kristiansen J, Kristjánsson JE, Schyberg H. 2011. The role of horizontal resolution for polar low simulations. Q. J. R. Meteorol. Soc. 137: DOI: 10.1002/qj.849 (in press). Montgomery MT, Farrell BF. 1992. Polar low dynamics. J. Atmos. Sci. 49: 2484–2505. c 2011 Royal Meteorological Society Copyright 1789 Noer G, Ovhed M. 2003. ‘Forecasting of polar lows in the Norwegian and the Barents Sea.’ Proc. 9th meeting of the EGS Polar Lows Working Group. European Geophysical Society: Cambridge, UK. Noilhan J, Mahfouf J-F. 1996. The ISBA land surface parameterisation scheme. Global and Planetary Change 13: 145–159. Petersen GN, Ólafsson H, Kristjánsson JE. 2003. Flow in the lee of idealized mountains and Greenland. J. Atmos. Sci. 60: 2183–2195. Petersen GN, Kristjánsson JE, Ólafsson H. 2004. Numerical simulations of Greenland’s impact on the Northern Hemisphere winter circulation. Tellus 56A: 102–111. Petersen GN, Kristjánsson JE, Ólafsson H. 2005. The effect of upstream wind direction on atmospheric flow in the vicinity of a large mountain. Q. J. R. Meteorol. Soc. 131: 1113–1128. Rasch PJ, Kristjánsson JE. 1998. A comparison of the CCM3 model climate using diagnosed and predicted condensate parameterizations. J. Climate 11: 1587–1614. Rasmussen E. 1981. An investigation of a polar low with a spiral cloud structure. J. Atmos. Sci. 38: 1785–1792. Rasmussen EA, Turner J (eds). 2003. Polar Lows. Cambridge University Press. Reed RJ, Duncan CN. 1987. Baroclinic instability as a mechanism for the serial development of polar lows: A case study. Tellus 39A: 376–384. Renfrew IA, Moore GWK, Kristjánsson JE, Ólafsson H, Gray SL, Petersen GN, Bovis K, Brown PRA, Føre I, Haine T, Hay C, Irvine EA, Lawrence A, Ohigashi T, Outten S, Pickart RS, Shapiro M, Sproson D, Swinbank R, Woolley A, Zhang S. 2008. The Greenland Flow Distortion Experiment. Bull. Am. Meteorol. Soc. 89: 1307–1324. Renfrew IA, Outten SD, Moore GWK. 2009. An easterly tip jet off Cape Farewell, Greenland. I: Aircraft observations. Q. J. R. Meteorol. Soc. 135: 1919–1933. Savijärvi H. 1990. Fast radiation parameterization schemes for mesoscale and short-range forecast models. J. Appl. Meteorol. 29: 437–447. Skeie RB, Kristjánsson JE, Ólafsson H, Røsting B. 2006. Dynamical processes related to cyclone development near Greenland. Meteorol. Z. 15: 147–156. Tsukernik M, Kindig DN, Serreze MC. 2007. Characteristics of winter cyclone activity in the northern North Atlantic: Insights from observations and regional modeling. J. Geophys. Res. 112: D03101, DOI: 10.1029/2006JD007184. Undén P, Rontu L, Järvinen H, Lynch P, Cavlo J, Cats G, Cuxart J, Eerola K, Fortelius C, Gardia-Moya JA, Jones C, Lenderlink G, McDonald A, McGrath R, Navascues B, Nielsen NW, Ødegaard V, Rodriguez E, Rummukainen M, Room R, Sattler K, Sass BH, Savijärvi H, Schreur BW, Sigg R, The H, Tijm A. 2002. ‘HIRLAM-5 scientific documentation.’ Technical Report HIRLAM-5 project, c/o Per Undén, SMHI, SE-60176 Norrköping, Sweden, 144 pp. Veersé F, Thépaut J-N. 1998. Multiple-truncation incremental approach for four-dimensional variational data assimilation. Q. J. R. Meteorol. Soc. 124: 1889–1908. Wilhelmsen K. 1985. Climatological study of gale-producing polar lows near Norway. Tellus 37A: 451–459. Winterfeldt J, Andersson A, Klepp C, Bakan S, Weisse R. 2010. Comparison of HOAPS, QuikSCAT, and buoy wind speed in the eastern North Atlantic and the North Sea. IEEE Trans. Geosci. Remote Sensing 48: 338–348. Zahn M, von Storch H. 2010. Decreased frequency of North Atlantic polar lows associated with future climate warming. Nature 467: 309–312. Q. J. R. Meteorol. Soc. 137: 1773–1789 (2011)