The Greenland Ice Sheet (GrIS) is one of the most dominant orographic obstacles for atmospheric flow in the Northern Hemisphere. Within an idealized framework, we investigate the potential impact of a reduced Greenland topography upon precipitation, snow accumulation, and atmospheric circulation over the GrIS. Using the global atmospheric model ECHAM5-HAM at about 1° spatial resolution (T106) and with present-day climatological mean conditions, we perform four 16-year sensitivity experiments that are identical except for the height of the GrIS topography: one control simulation with the present-day Greenland topography, and three simulations with topographies reduced to 75%, 50%, and 25% of the present-day height. This simple reduction of the GrIS topography (as compared to realistic melt dynamics) leads to an overall increase in annual total precipitation and snow accumulation, composed of significant increases in eastern, northern, and central Greenland and decreases on the western slopes. In principle, this gain in snow accumulation raises the possibility of a negative feedback that would stabilize the height of the GrIS. However, this feedback is likely overcompensated by enhanced ablation (positive feedback), as surface air temperatures strongly increase with reduced topographic height. The analysis of changes in circulation patterns indicates that flatter topographies allow the atmospheric flow to penetrate farther inland, enabling precipitation in regions that are presently desert-like. Prominent circulation features change, in particular the all-season Greenland Anticyclone and wintertime Icelandic Low become weaker with lower topography.
 Greenland is the world's largest island stretching over 2600 km meridionally, between 59° and 84°N, and 950 km zonally, between 11° and 74°W. More than 80% of the island's area is covered by the Greenland Ice Sheet (GrIS) with a total volume of 2.65 × 106 km3 [Ohmura et al., 1996]. The GrIS largely determines the topography and peaks at approximately 3200 meters above sea level. The coastal slopes are very steep, most prominently in the southeast, the wettest region in Greenland, which is directly hit by onshore flow from the northern flank of the Icelandic Low [Ohmura and Reeh, 1991]. The evolution of the GrIS has received considerable attention in climate research, as its mass balance is particularly sensitive to climate change, and a full or partial meltdown of the GrIS would have major repercussions on sea level rise.
 The massive size of the present-day GrIS is known to impinge substantially on synoptic and planetary scale atmospheric flow [e.g., Broccoli and Manabe, 1992; Kristjánsson and Mcinnes, 1999; Petersen et al., 2004]. Moreover, a number of studies point out the relevance of atmospheric circulation for Greenland precipitation, which is mostly orographically induced over the coastal slopes by passing cyclones [Chen et al., 1997; Schuenemann et al., 2009]. The orographic uplift leads to immediate precipitation over the outer margins of the ice sheet, not affecting regions farther inland due to a lack of precipitable water [Ohmura and Reeh, 1991]. Most of Greenland's interior is in the precipitation shadow and represents the so-called “dry slot” in annual mean precipitation amounts [Roe, 2005].
 Although the flow past Greenland is very complex, with a continuously changing upstream flow, some indication of the atmospheric response may be obtained from idealized studies of uniform flow past isolated topography. Sufficiently high orography, as embodied by the present-day GrIS, may induce upstream blocking and splitting of atmospheric flow, which in some cases is associated with vortex shedding in the lee of the obstacle [Smith, 1989; Schär and Durran, 1997]. Depending on the dimensionless obstacle height and the Rossby number, two main flow regimes may be distinguished [Trüb and Davies, 1995; Ólafsson and Bougeault, 1997; Schär, 2002; Petersen et al., 2003]: either (1) the flow passes over the obstacle, remaining unblocked and quasi-stationary for low dimensionless heights, or (2) the flow is blocked and forced to go around with a potential for vortex shedding to the lee of the obstacle. In this framework, present-day Greenland is mainly placed in the “flow around” regime, but also partially in the “flow over” regime [Schär, 2002]. In the latter case, flow that is directed over Greenland may induce the formation of an anticyclone sitting at the ice sheet top, similar to quasi-geostrophic flow solutions. Indeed, the so-called Greenland (glacial) anticyclone [Hobbs, 1945] is a well-known all-season circulation feature over the GrIS, induced by descending air over the central cold region. Following the above theory, the reduction of the GrIS height would place Greenland more often in the “flow over” regime, which in turn would lead to larger vertical and weaker horizontal deflections of the incident flow. As the flow field considerably determines the precipitation pattern over the ice sheet, a feedback upon snow accumulation and surface mass balance is to be expected as well.
 The fourth IPCC assessment report 2007 [Lemke et al., 2007] summarizes observational estimates of the present-day GrIS mass balance from numerous studies, and finds that most results indicate accelerating mass loss from Greenland during the 1990s up to 2005. Various studies using GCM calculations support these findings and indicate that the mass loss of the ice sheet will become even more substantial for the coming century [Meehl et al., 2007]. Ridley et al.  (HadCM3) find that under a 4 × CO2 climate the GrIS undergoes continuous reduction. According to Ridley et al. [2005, Figure 4], the GrIS might shrink to 75% of its volume after approximately 300 years, to 50% after 750 years, and down to 25% after 1500 years.
 Ablation is highly sensitive to temperature changes near the ice sheet margins, where runoff of meltwater is seen to increase in all models. Most studies indicate that the increase in ablation exceeds the increase in accumulation, leading to enhanced GrIS mass loss and a net positive contribution to global sea level in the 21st century.
 The present study investigates in an idealized framework how a reduced Greenland topographic height, as may occur in a warmer climate, would interact with the local atmospheric flow and affect the amount and spatial distribution of precipitation and snow accumulation over the GrIS. The investigation of such atmosphere-topography interactions will add to our understanding of the dynamics of an eventual meltdown of the GrIS in a warmer climate. However, in assessing the potential of such feedbacks, we will assume an otherwise unperturbed climate and investigate the role of Greenland's height under current climatic conditions, i.e., we prescribe global sea-surface temperatures (SST) and greenhouse gas concentrations at current levels.
 Greenland precipitation is expected to change not only due to the topographic effects as discussed in this paper, but will also increase in response to the anticipated global mean warming. There is a scientific consensus that the global specific humidity will rise in a warmer climate by about 6–7%K−1, consistent with the Clausius-Clapeyron relation [e.g., Boer, 1993; Allen and Ingram, 2002; Held and Soden, 2006; Meehl et al., 2007]. The intensity of the hydrological cycle is, however, also constrained by the availability of energy and not by moisture alone [Pall et al., 2007; Wild et al., 2008]. Thus, the increase in global mean precipitation is smaller and estimated to be near 3.4%K−1 (based on model simulations under equilibrium 2 × CO2 [Allen and Ingram, 2002]) or 2%K−1 (based on climate change experiments (A1B) of the fourth IPCC assessment report [Held and Soden, 2006]). This additional effect is not accounted for in the present study. It is also evident that in the short term (i.e. the next 100 years) changes in Greenland precipitation are likely driven by global mean warming, while the potential precipitation changes addressed in the current paper would act on a slower timescale but become relevant as soon as there is a significant net mass loss over the GrIS.
 In Section 2 we give a brief description of the GCM and experimental setup used in this study. A validation of total precipitation (TP) and snow accumulation for the current GrIS is presented in Section 3. Section 4 describes the topographically induced changes in amount and spatial distribution of TP, snow accumulation, and ablation over the flatter Greenland topographies. Changes in circulation patterns, the main reason for the redistribution of Greenland precipitation, are the topic of Section 5. Sections 6 and 7 discuss our results and present our conclusions.
2. Model and Experiments
 We use the fifth generation general circulation model ECHAM5, which was developed at the Max Planck Institute for Meteorology, Hamburg (MPI) [Roeckner et al., 2003] and evolved from the weather forecasting model of the European Centre for Medium Range Weather Forecasts (ECMWF) [Simmons et al., 1989]. In our version, ECHAM5 incorporates the Hamburg Aerosol Model (HAM) [Stier et al., 2005] and sophisticated cloud microphysics [Lohmann et al., 2007], providing a detailed description of tropospheric aerosols and their interaction with clouds. We run the model with spectral resolution of T106 (about 1.1° grid spacing on a corresponding Gaussian grid) with 31 vertical levels from the surface up to 10 hPa (ca. 30 km). It is clear that a resolution of 1.1° is poor compared to the resolution used in RCMs [Box et al., 2006; Fettweis, 2007; Ettema et al., 2009] but only a GCM can evaluate the impact of a lower GrIS topography on the general circulation, which impacts in its turn the precipitation pattern.
 We chose to prescribe climatological conditions in order to reduce the natural variability of the system. Aerosol emissions and greenhouse gas (GHG) concentrations are kept constant at their 1870 values. For the sea surface temperatures (SST) and sea ice coverage (SIC) we use a climatological data set based on 1961–1990 Hadley Centre data [Rayner et al., 2003].
 Four simulations were carried out, each of them covering 16 years, plus 3 months spin up time. The simulations are identical except for the Greenland topographies. We conducted a control simulation (in the following referred to as normGR) with a present-day Greenland topography, and three simulations with Greenland topographies reduced to 75%, 50%, and 25% of normGR (in the following referred to as 75GR, 50GR, and 25GR). The reduction was applied evenly over the entire Greenland topography by multiplying the input orography (bedrock plus ice sheet) at each spatial grid point in the GrIS with the corresponding factors. This simple reduction contrasts with more detailed modeling studies by, e.g., Ridley et al. , who find that the GrIS melt in a 4 × CO2 climate starts in the low-lying coastal regions (i.e., the north and southwest) and is strongly constrained by the eastern mountain ranges. Our approach has been selected as it represents a well controlled and more easily interpretable idealized framework for sensitivity studies about topography-precipitation feedbacks.
 To define the spatial extent of the GrIS in the model data, we combined the model output variables “land sea mask” and “fraction of land covered by glaciers” to keep only those land grid points that are permanently covered by ice. Based on this approach, the horizontal extent of the GrIS is 2 * 106 km2, covered by 493 grid points in our model resolution (T106). We further subdivided the GrIS into five regions: south (S), west (W), central (C), east (E), and north (N) Greenland. The subdivision follows earlier studies [e.g., Chen et al., 1997; Schuenemann and Cassano, 2010] and is shown in Figure 1, along with the Greenland topography as represented by the model resolution (T106).
 Previous studies already demonstrated the suitability of ECHAM5 for studying Greenland's climate [Schuenemann et al., 2009; Mernild et al., 2010]. Comparing 15 GCMs, Walsh et al.  found ECHAM5 to give realistic results with respect to output variables such as temperature, precipitation, and sea level pressure for Greenland, Alaska, pan-Arctic, and the Northern Hemisphere, ranking it among the top two GCMs (MPI ECHAM5 and GFDL CM2.1) for such studies.
 In line with these studies, Greenland precipitation from our normGR simulation is well within the range of published observational estimates, as listed in Table 1, and also compares well with a selection of previous modeling studies. The GrIS-averaged annual mean total precipitation (TP) for normGR is 353 mm yr−1. The published observational estimates (Table 1) range from 340 to 360 mm yr−1 with identical mean and median values of 350 mm yr−1. The modeling study results, as listed in Table 1, range from 336 to 494 mm yr−1, with a mean value of 395 mm yr−1 and median of 388 mm yr−1.
Table 1. Annual Total Precipitation (16-Year Mean) and Mass Balance Parameters Over the NormGR Ice Sheet and Published Estimates From Literaturea
Columns denote model or observational reference, total precipitation (TP), snowfall, snow accumulation (Acc.), ablation (Abl.), and surface mass balance (SMB) in mm yr−1 water equivalent (w.e.). For normGR, ablation and SMB are determined in three ways: (1) based on summer mean temperature [Ohmura, 2001], (2) based on energy balance [Braithwaite and Olesen, 1990], and (3) by combining methods 1 and 2. See text for details.
 The spatial distribution of annual TP over normGR, shown in Figure 2 (left), is found to be in general agreement with the precipitation map by Ohmura and Reeh  (Figure 2, right) based on observations made at 35 meteorological stations in the coastal regions of Greenland, and accumulation measurements of 251 pits and cores obtained from the upper accumulation zone. The resulting pattern also compares well with the distribution map obtained from calibrated regional climate model (RCM) output by Box et al.  (not shown). In accordance with their findings, the normGR precipitation maxima are located in the southeastern coastal regions, exceeding 1900 mm yr−1, and on the western slopes with up to 600 mm yr−1. The smallest amount of precipitation reaches north and central Greenland, with a minimum around 70 mm yr−1.
 The normGR seasonal cycles of TP over the GrIS and for the individual regions (see Section 2) compare quite well with corresponding ERA-Interim climatologies (1979–2011), except for the southern part, where the reanalysis shows a distinct seasonal cycle with a minimum during summer and a maximum during winter (not shown). We attribute the less pronounced seasonal cycle in normGR to the comparatively coarse resolution of the coastal margins [cf. Wild et al., 2003]. ERA-Interim has been found to reproduce the spatial pattern and temporal variability of Greenland TP reasonably well [Chen et al., 2011].
 The simulated annual snow accumulation (snow fall minus evaporation) over the normGR ice sheet is with 290 mm yr−1 in good agreement, although somewhat on the lower side, compared to the observation based estimates listed in Table 1, which have a median of 314 mm yr−1. The spatial pattern closely follows the TP pattern, with lowest rates found in central Greenland around the summit area, local peaks along the west coast (around 500 mm yr−1), and maxima in the southeast exceeding 1400 mm yr−1. The spatial distribution of snow accumulation is in general agreement with the observation-based accumulation map by Ohmura et al.  and the aforementioned RCM mean accumulation pattern for the period 1958–2007 by Burgess et al.  (not shown).
4. Impact on Precipitation and Surface Mass Balance
4.1. Changes in Precipitation
 The GrIS-averaged annual mean total precipitation (16-year mean) increases over a reduced Greenland topography. Reducing the GrIS height from 100% down to 25% of its present-day height, raises the annual mean TP by 36%. The increase is approximately linear with on average +11% per 25% topographic height reduction (Table 2). According to a Kolmogorov-Smirnov-Lilliefors test, the GrIS-averaged annual means are normally distributed at the 5% significance level, as is to be expected for natural variability. To examine whether these changes are statistically significant, we can thus apply the parametric two-sample t test on the GrIS-averaged annual means. We find the differences between normGR and the reduced-GrIS simulations 50GR and 25GR to be significant at the 95% significance level, but insignificant for 75GR. The result is confirmed by the non-parametric Wilcoxon-Mann-Whitney test. All mean values considered in the following passed a Kolmogorov-Smirnov-Lilliefors test and we always obtained very similar results for both the t test and the Wilcoxon-Mann-Whitney test. Only the results of the t test are given in the following.
Table 2. Annual (16-Year Mean) Precipitation, Temperature, and Mass Balance Parameters Over the Entire Ice Sheet From our NormGR and Reduced-Greenland Simulationsa
(1), (2), (3)
(1), (2), (3)
Units are mm yr−1 w.e. unless otherwise indicated. Columns denote simulation, total precipitation (TP), snowfall (absolute and % of TP), 2-meter air temperature (Temp2) (°C), snow accumulation (Acc.), ablation (Abl.) and surface mass balance (SMB), estimated in three different ways (see Table 1). Year-to-year standard deviations of TP are in the range 25–42 mm yr−1.
75, 379, 212
215, −89, 78
114, 547, 293
190, −243, 11
206, 856, 534
113, −537, −215
476, 1091, 1091
−148, −763, −763
 The spatial pattern of annual TP changes significantly with reduced GrIS height (Figure 3, hatched areas indicate significance at the 95% significance level as determined for each individual grid box). Precipitation on the western slopes weakens in all three reduced-Greenland simulations, exceeding −50% around Thule in the northwest of 25GR. East of it, i.e., in most of north, (south-)eastern, and central Greenland, precipitation is enhanced. In terms of percentage, most of the TP increase occurs in central Greenland, where the absolute topography decrease is largest and absolute TP values are smallest (not shown). Regions that are presently very dry receive enhanced precipitation, whereas rather wet regions, especially the western and partially the southeastern slopes (25GR), become drier. The resulting TP pattern of 25GR shows a more distinct north-south gradient instead of a desert in the central summit area and less extreme TP rates along the coastal margins than for normGR (Figure 3).
 The multiyear seasonal cycles of monthly mean TP averaged over the entire GrIS (Figure 4, top row, left) show that TP increases almost uniformly throughout the year, seasonality does not change substantially when reducing the topography. In all simulations, most TP is generated during summer and autumn.
 A look at individual regions (see Section 2) instead of the entire GrIS, reveals substantial differences in normGR TP seasonality (Figure 4, black line). Most precipitation reaches the west during summer, whereas in east, north, and central Greenland the seasonal cycles are overall flatter with weak maxima during summer and autumn. In south Greenland, the normGR TP shows no distinct seasonality. Reducing the GrIS height increases TP in the south throughout the year (75GR, 50GR) with a slight tendency to decrease during winter in 25GR. Precipitation in west Greenland declines during summer but barely changes during the colder months. In east and central Greenland, TP increases significantly (95% significance level of two-sample t test applied to monthly and regional means) throughout the year (25GR-normGR), and the summer and autumn peaks become more distinct. In north Greenland, TP increases abruptly during most of the seasons (except summer) for a topographic reduction down to 25GR.
4.2. Changes in Snow Accumulation
 The snow accumulation over the ice sheet is determined by the amount and spatial distribution of solid precipitation or snowfall, respectively, and to a lesser extent by evaporation. Over normGR, the total precipitation (TP) is mostly solid, exceeding 86% in the annual mean (see Table 2).
 As the solid precipitation fraction decreases over a flatter and thus warmer Greenland, the increase in accumulation cannot be as high as the increase in TP. Reducing Greenland's height increases the GrIS-averaged annual snow accumulation approximately linearly by on average +4% per 25% topographic reduction (Table 2). As with total precipitation, these changes are significant at the 95% level for 50GR and 25GR, but not for 75GR. The overall increase in snow accumulation with decreasing height of topography is a composite of different tendencies in space and season. Snow accumulation decreases most prominently on the western slopes during the warmer months, but increases most distinctly in east and central Greenland during the colder months (Figures 5 and 6).
 The spatial patterns of annual mean snow accumulation and TP are closely related to each other (see Section 3.1). Nonetheless, the seasonality differs because of the influence of surface air temperature on the solid fraction of precipitation. Unlike for TP, accumulation has strong minima during the summer months and is dominated by the colder months. This pattern gets more pronounced as the GrIS is reduced, especially in eastern and central Greenland (Figure 6). The warmer the summers, the lower the associated accumulation rates. In July, the GrIS-averaged multiyear monthly mean surface air temperature reaches close to 0°C in 25GR (not shown). The implied higher frequency in positive degree days not only results in a shift from solid to liquid precipitation, but also strongly increases ablation (see next Section).
4.3. Ablation Estimates
 Ablation is strongly affected by surface air temperature. In the context of global warming it has been shown that beyond a critical temperature increase over Greenland (i.e., +4.5°C [Gregory and Huybrechts, 2006]) ablation accelerates, leading to a decline in surface mass balance (SMB) [e.g., Meehl et al., 2007]. Recent ECHAM5 T213 calculations by Bengtsson et al.  project 200% more ablation for a 5.2°C warming over the GrIS (IPCC scenario A1B, +3°C globally), turning the initially positive SMB negative. In our simulations, the surface warming is due to the reduction of the topographic height. With a topographically induced warming of around +2°C per 25% topographic reduction, up to 6°C over 25GR (Table 2), we expect a comparable increase in ablation, as the melt zone will expand spatially into Greenland's central regions, and temporarily into presently too cold seasons.
 Even though the model resolution of T106 is higher than typically used in climate model simulations, it is still fairly coarse for an adequate determination of ablation [Glover, 1999; Wild et al., 2003]. To derive a rough estimate, we determined ablation in three ways:
 (1) Following Wild et al. , we estimated ablation from the mean summer temperature, based on an empirical formulation after Ohmura , which we applied to each grid point over the GrIS. The computation is based on the observed relationship between the annual ablation and the collocated mean air temperature for the summer months (JJA) [Ohmura et al., 1996]. For normGR, this methodology yields an annual ablation of 75 mm yr−1, which is significantly below published estimates (Table 1), and may be considered as a conservative estimate. For the reduced Greenland topographies, this approach is not fully justified, as the empirical relationship stems from the current-day GrIS. Applying it nevertheless, suggests accelerated mass loss with reduced topography, the enhanced accumulation being overcompensated by the warming of up to +6°C (25GR) and the associated ablation increase (Table 2).
 (2) Based on a simple energy balance model [Braithwaite and Olesen, 1990], we calculated the residual energy available for melting over the GrIS as the sum of the net radiation, sensible, and latent heat fluxes. This approach does not capture the refreezing of meltwater in the firn [Pfeffer et al., 1991; Ohmura et al., 1996] and overestimates ablation strongly, resulting in a negative SMB for normGR (Table 1).
 (3) To overcome the shortcoming of method (2), we combined it with method (1) in calculating the melt based on the energy balance only in grid boxes, in which the temperature criterion of method (1) indicates that ablation takes place and, thus, refreezing does not occur. The GrIS-averaged annual ablation for normGR lies then with 212 mm yr−1 well within the range of observational and model estimates (Table 1).
 All three methods, regardless whether our normGR ablation agrees with published estimates for present-day conditions or not (Table 1), indicate that the ablation increase over a flatter GrIS sooner or later dominates over the gain in snow accumulation (Table 2). The SMB declines (all methods) and eventually turns negative in 50GR (method 3) or 25GR (method 1), respectively.
5. Circulation Changes and TP Patterns
 Various earlier studies find a clear relationship between the circulation and precipitation patterns over Greenland, arguing that TP is mostly due to orographic effects acting upon the mean flow and passing extratropical cyclones [e.g., Ohmura and Reeh, 1991; Chen et al., 1997; Schuenemann et al., 2009; Schuenemann and Cassano, 2010]. Here, we use two methodologies to assess this relationship and the associated changes as the height of Greenland is reduced. First, one common approach is to inspect the mean sea level pressure (SLP) patterns and wind fields in summer (JJA) and winter (DJF). This is well motivated as cyclonic activity highly depends on the season, and as snow accumulation and also TP show a seasonal cycle in at least some of the regions of Greenland (especially the west, Figures 4 and 6). Second, we will identify high precipitation months for each of the five regions introduced in Section 2, and determine the corresponding circulation patterns by composite analysis, following Chen et al. . This approach is particularly appropriate for regions where precipitation shows no distinct seasonality.
5.1. Winter and Summer Flow Patterns
 The analysis is conducted using maps of precipitation, sea level pressure (SLP), 500 hPa geopotential height (gpm500), and low-level wind (Figures 7 and 8). In these plots we exclude SLP in the area of Greenland, as pressure reduction leads to substantial errors.
 The spatial distributions of winter and summer precipitation are determined by characteristic flow patterns and differ substantially: In winter, TP maximizes on the southeastern slopes, while the other regions experience their seasonal minima (Figure 7, first row, left). Reducing Greenland's height significantly increases TP over the higher plateau in north, east, and central Greenland (Figure 7, first row). The TP maxima in the southeast initially amplify (75GR and 50GR), but then decrease again (25GR). In summer, TP is more evenly distributed over Greenland than in winter, maximizing in the very south and on the slopes of western and southeastern normGR (Figure 8, top row, left). A flatter GrIS leads to a decrease in TP on the western slopes and an increase in all other regions, predominantly the summit area (Figure 8, top row).
5.1.1. Winter Circulation
 In normGR, the winter circulation in terms of SLP and wind field (at the sixth terrain-following model level, approximately 860 hPa over sea) depicts a long-stretched, banana-shaped Icelandic Low with its center just east of the southern tip that produces strong onshore winds striking the southeastern coast (Figure 7, second and third rows, left). Once these southeasterlies reach the higher plateau, they are deflected north and transition into anticyclonic motion around the summit area. Commonly referred to as the Greenland Anticyclone, this is an expected flow feature over Greenland's glacial surface [Hobbs, 1945], and in agreement with the quasi-geostrophic “flow over” theory [Schär, 2002].
 The reduction of the GrIS height leads to a weaker Icelandic Low, SLP increases over the Greenland and Norwegian Seas, and decreases to the southwest (50GR and 25GR) (Figure 7, second row, left). The weakening of the Icelandic Low is accompanied by the diversion of the North Atlantic storm track. The storm track difference pattern in terms of 2.5–8 days band-pass filtered variability of the 500 hPa geopotential height (gpm500), indicates enhanced storm activity above and to the west of a flatter Greenland, and weaker activity over the eastern North Atlantic (not shown). Thus, in the case of a flatter Greenland topography, storms move more frequently north along the west coast instead of the east coast, which is their natural track under current conditions, and traverse the island without being forced to go around it. These findings are in line with Junge et al. , who suggest that more cyclones pass over a flat-bottom Greenland topography.
 Previous studies have shown that westerly flow crossing the southern tip of Greenland induces lee-cyclogenesis that significantly strengthens the Icelandic low [Chen et al., 1997; Kristjánsson and Mcinnes, 1999; Doyle and Shapiro, 1999], and that the “lee cyclone” declines or even vanishes in simulations without a Greenland topography [Kristjánsson and Mcinnes, 1999; Petersen et al., 2003; Junge et al., 2005; Tsukernik et al., 2007]. Indeed, the normGR wind field indicates the presence of a winter mean “lee cyclone” or “tip vortex” to the east of the southern tip, which becomes weaker as the height of the GrIS is reduced and essentially disappears in 25GR (Figure 7, second and third rows). In 75GR and 50GR the “lee cyclone” still exists and the TP maxima in the southeast amplify, as onshore winds are less deflected and hit the coast more perpendicular (Figure 7, first and second rows). In 50GR, the TP maxima are shifted inland, which is in accordance with theory, implying that orographic precipitation moves closer to the mountain crest when the topographic height is reduced [Roe, 2005]. In 25GR, the missing “lee cyclone” and weaker Icelandic Low result in weaker easterly onshore winds and less precipitation at the southeastern margin.
 The wind fields are furthermore consistent with a weakening of the Greenland Anticyclone, most prominently in 25GR (Figure 7, third row). The anticyclone weakens over a flatter (quasi-geostrophic “flow over” theory) and warmer (less cooling and descent of low-level air) Greenland. Its absence allows easterly onshore flow to penetrate farther inland (Figure 7, second row), which appears to be the main reason for the increase in TP over eastern and northern Greenland. The effect on west Greenland is reversed. The southerly flow to the west of Greenland (which brings warm moist air to the northern portion of the west coast) is diminished, in response to the weakening of the Greenland Anticyclone. As a result, the flow is more directed over (rather than around) the Greenland topography, yielding increased upstream and reduced downstream precipitation. At the same time, as the orography is reduced, it becomes easier for the southwesterly and -easterly flows to reach the central plateau with the potential for higher TP rates. Looking at the gpm500 contours, the geopotential height increases over and northwest of Greenland, indicating the weakening and spatial expansion of the winter polar vortex for a flatter GrIS (Figure 7, third row).
5.1.2. Summer Circulation
 In summer, the normGR circulation is dominated by low pressure over Northern Canada and Baffin Bay, producing westerly onshore flow. The westerlies sharply deviate north over the western slopes and accelerate around the summit area, again inducing the formation of a Greenland Anticyclone (Figure 8, second and third rows, left). A reduced Greenland topography shifts the mean cyclone southeast into Baffin Bay and over west Greenland in 25GR (Figure 8, second row). The intensification of this Baffin Bay Low [Ohmura and Reeh, 1991] changes the flow pattern quite dramatically, as it produces offshore flow over northwestern Greenland. As in winter, the storm activity increases strongly over Greenland's interior (not shown). The Greenland Anticyclone ultimately vanishes in 25GR (Figure 8, third row).
5.2. Flow Patterns for High TP Months
 For all Greenland regions, composite maps are constructed that relate to rainy (or snowy) periods in the respective regions. To this end, for each simulation and Greenland region, we select the 25% of all months with the highest TP during the 16-year simulation period (on average 3 months per year). We then compute their composite mean SLP patterns, 500 hPa geopotential heights and 860 hPa wind fields. The seasonal distribution of the high TP months per region in normGR is as follows: central, north, and west Greenland show distinct summer maxima (with more than 50% of high TP months) and minima in winter (down to 4% in the west). In east and south Greenland, high TP months are more evenly distributed between summer, autumn, and winter (25–31%), the minimum with 17% lies in spring. The reduction in topographic height does not substantially change the seasonal distributions of high TP months. In general, the 25GR simulation exhibits a broadened seasonal peak of the high TP maxima in west, central, and east Greenland, spanning summer (on average 44%) and autumn (42%).
 The circulation patterns responsible for high TP in south, west, and east Greenland differ substantially (Figure 9), whereas north and central Greenlands' patterns are combinations of the eastern and western high TP flow regimes. Thus, we concentrate on the former three regions.
 Over southern Greenland for normGR, high TP is related to a mean cyclone that engulfs the southern tip with the low pressure center just east of it (Figure 9, first row, left). This cyclone produces onshore flow approaching the southeastern slopes (Figure 9, second row, left). Chen et al.  found that if a cyclone exists in the mean monthly sea level pressure in the Labrador sea (Labrador Low), heavy precipitation occurs in southern Greenland. The pressure pattern we found for high TP in the south appears to be a combination of such a Labrador Low and an additional “lee cyclone”, induced by westerly flow crossing the southern tip and causing lee cyclogenesis [Chen et al., 1997; Kristjánsson and Mcinnes, 1999; Doyle and Shapiro, 1999]. Reducing Greenland's topography changes the pattern noticeably. The mean cyclone's center weakens and relocates westward around the southern tip. In 25GR, the cyclone looks just like the Labrador Low proposed by Chen et al.  (Figure 9, first row, right). The absence of lee cyclogenesis may play a role and the formerly intense onshore winds (Figure 9, second row) and precipitation (Figure 3) decline.
 Over western Greenland, high TP is associated with a summer-like low pressure structure centered over Baffin Island (Figure 9, third row, left). The westerly onshore winds accelerate around the summit area (Figure 9, fourth row, left). Conducting the analysis with reduced Greenland topography yields a pattern with an intensified low pressure system over the Baffin Bay (Figure 9, third row). A more uniform southwesterly wind field dominates over 25GR (Figure 9, fourth row), reduces TP on the western slopes and provides enhanced TP to the central region. These changes do not necessarily imply that the Baffin Bay low becomes stronger, but that such lows become more important in yielding precipitation over western Greenland.
 Over eastern Greenland, high TP is produced in late summer by an Icelandic Low that, compared to the winter mean, stretches broader into the Norwegian sea with its center farther off the southeastern coast of Greenland (Figure 9, fifth row, left). The easterly onshore winds sharply deviate south, as they are blocked by steep orography and forthcoming westerly downslope winds, and run parallel to the shore (Figure 9, sixth row, left). The Icelandic Low in these circulation patterns becomes more compact in 50GR and 25GR (Figure 9, fifth and sixth rows), although the wintertime Icelandic Low itself becomes generally weaker with the reduction of the Greenland topography (see Figures 7 and 8). The result thus suggests that TP in eastern Greenland would become more sensitive to the location of the Icelandic Low in a scenario with reduced topographic height.
5.3. Flow and TP Pattern: Synthesis
 Both analyses in terms of regional high TP and summer/winter circulation patterns yield the following major flow changes as the GrIS height is reduced: (1) The decay of the Greenland Anticyclone reduces the lateral diversion of the incoming flow and thereby increases the lifting and precipitation with the easterly onshore flow (TP: east ↑, north ↑, west ↓), (2) Reduced orographic flow blocking is associated with the shift of orographic TP maxima closer to the mountain crest and increased cross-orographic circulation (slopes ↓, interior ↑), (3) The diversion of the NA storm track contributes to a less intense Icelandic Low (southeast ↓) and to a more intense Baffin Bay Low that wells over a flatter Greenland, inducing offshore flow over the northwest and enhanced inflow into the central regions (interior ↑, west ↓), and (4) Lee-cyclogenesis occurs less frequently (south ↓).
 In this study, we use simplified topographic reductions as compared to realistic melt dynamics of the GrIS. Simulations using dynamic ice sheet models [e.g., Ridley et al., 2005, 2010] show that the GrIS meltdown evolves non-linearly and exposes bedrock asymmetrically, starting in the north and southwest of the GrIS. Along the low-lying ice sheet margins, the initial warming is strongest and the retreat of ice is accompanied by enhanced surface air temperatures due to elevation- and albedo-temperature feedbacks [e.g., Toniazzo et al., 2004; Ridley et al., 2005]. The eastern mountains act as a barrier, restricting the ice flow to lower elevations, and a dome of ice remains in central-east Greenland long after most of the ice sheet has retreated. Thus, the maximum elevation height of the GrIS is barely affected for many hundred years, even for a climate warming up to +10°C [Greve, 2000]. Bedrock elevation maps by Letréguilly et al.  have mean altitudes of 440 meters before and 812 meters after isostatic rebound and are characterized by two main mountain ranges in eastern Greenland and on the southern tip, reaching up to 1000 and 600 meters before, and up to 1400 and 800 meters after isostatic rebound [Letréguilly et al., 1991; Greve, 2000; Bamber et al., 2001; Toniazzo et al., 2004; Ridley et al., 2005]. The mean height of all our reduced-Greenland topographies (461 meters for 25GR) lies above the present-day bedrock mean altitude. However, our Greenland orographies are generally too smooth, i.e., too flat around the summit area, and too high in low-lying coastal areas (maximum height 830 meters for 25GR). Overall, we expect our results to remain basically valid, once the bedrock is fully exposed. Differences may be expected during the different phases of the complex melt [Ridley et al., 2005] or because of other reasons, e.g., globally changing atmospheric conditions in a 4 × CO2 world.
 Considering the impact of changes in Greenland topography upon the atmospheric circulation, both the height and width of the GrIS determine the flow regime [Trüb and Davies, 1995; Ólafsson and Bougeault, 1997; Schär, 2002; Petersen et al., 2003]. Changes in both these parameters associated with the melt of the GrIS can potentially affect the proportion of flows in the “flow over” and “flow around” regimes. If the melt of Greenland proceeds such as to primarily reduce the height of the GrIS (rather than the shape), then the flow will become more likely to flow over the GrIS and thereby affect precipitation as simulated in the presented numerical experiments. If on the other hand the melt proceeds such as to maintain the peak height of Greenland but to reduce its horizontal scale (in particular in the north-south direction), then the flow will change such as to be more often in the “flow around” regime, thereby potentially reducing precipitation. This effect occurs as a reduction of horizontal extent will reduce the impact of the Earth's rotation on the selection of the flow regime (i.e. an increase in effective Rossby number).
 We further introduced uncertainties through the computation of ablation, as we neglect several feedbacks that might either enhance or reduce ablation. On the one hand, we observe a shift from solid to liquid precipitation (see Table 2 and Figure 6) that might enhance ablation during summer through the direct contribution of rain to the melt, which is estimated to be extremely small for present-day conditions [Box et al., 2004], and through the decrease in surface albedo due to less frequent snowfall [Ettema et al., 2009; Screen and Simmonds, 2011]. On the other hand, we have not taken into account the exposure of bare ice during summer that might be delayed due to increased snowfall during the colder months [Tedesco et al., 2011]. Indeed, the GrIS-averaged snow accumulation (Figure 6, top row, left) is enhanced from October to April over a reduced GrIS topography, but there are regional differences. In east, central, and north Greenland a thicker snowpack might build up, whereas in west and partially south Greenland, snow accumulation even decreases (see Figures 5 and 6). In the latter two regions, earlier bare ice appearance might enhance the mass loss of the ice sheet.
 Within this study, we investigate how a reduced height of the Greenland topography would affect the flow and precipitation patterns over the Greenland Ice Sheet (GrIS), and whether there is a significant increase in snow accumulation tied to this, which might act as a negative feedback and slow down the ongoing GrIS mass loss. We use an idealized framework to quantify the feedback at question, rather than a scenario approach that would need to consider the transient melt of the GrIS and specifications of climate forcings. Our approach thus makes several simplifications as we use (1) simplified topographies, (2) prescribe present-day climatological mean conditions, and (3) estimate ablation based on rather vague methods. Nevertheless, the idealized framework has proven to be suitable in isolating the topography-precipitation feedback.
 We conducted four climatological 16-year ECHAM5-HAM T106 simulations that are identical except for the height of the GrIS: 100%, 75%, 50%, and 25% of the present-day height. In the control simulation normGR, total precipitation (TP) and snow accumulation are in reasonable agreement with observations, reanalyses, and previous modeling studies. High precipitation over Greenland occurs mainly in the coastal regions and maximizes in the southeast, predominantly during autumn and winter, and less pronounced on the western slopes, distinctly peaking during summer.
 Reducing the GrIS height increases the total precipitation (+11% per 25% topographic reduction) and snow accumulation (+4% per 25% topographic reduction). The change in total amounts is accompanied by a regional redistribution (wetter central, east, north, and drier west). The seasonal distribution of TP remains largely unchanged, but snow accumulation becomes more concentrated in winter. The annual mean surface air temperature increases by +2°C per 25% topographic reduction. From the current analysis, there is strong evidence that the associated increase in ablation dominates over the small gain in snow accumulation. A flatter GrIS topography, as might occur in a warmer climate, thus may accelerate ablation, which represents a positive feedback on the GrIS mass loss.
 We conclude that there must be a relationship between the changes in precipitation pattern – regional increases and decreases – and the partial transition from the “flow around” to the “flow over” regime as the Greenland topography is reduced. The topographically induced changes in circulation are thus essential for the increase and redistribution in TP.
 The present study demonstrates that feedback processes between Greenland's topography, circulation, and precipitation/snow accumulation may play a relevant role for the evolution of the surface mass balance of the GrIS, potentially both in the past and in the future.
 Based on the presented results and previous publications [e.g., Broccoli and Manabe, 1992; Kristjánsson and Mcinnes, 1999; Dethloff et al., 2004; Petersen et al., 2004; Junge et al., 2005; Tsukernik et al., 2007], it is to be expected that other prominent flow features in Greenland's vicinity (e.g., Greenland tip jet [Doyle and Shapiro, 1999]) and in the far field are affected by a shrinking Greenland topography. Thus, it may be interesting to extend the spatial scope of this study into the Arctic region and the entire Northern Hemisphere. Indeed, preliminary analysis of our data at this spatial scale suggests a weakening and spatial expansion of the polar vortex (strongest in 25GR), accompanied by enhanced surface temperatures over the polar region and negative temperature anomalies over mid-latitudinal regions in North America and Eurasia. These findings resemble the features of the negative Arctic oscillation (AO) index [Wallace and Gutzler, 1981]. A flatter Greenland topography seems to shift the AO permanently toward its negative phase. Further research is needed to scrutinize the interaction between the melting GrIS in a changing climate and the Northern Hemispheric large-scale circulation.
 Computations were performed at the Swiss National Supercomputing Center (CSCS), and the authors are most grateful for the continuous help and assistance of CSCS staff. We would also like to thank Atsumu Ohmura and Heinz Blatter for support and extensive discussions, and Ulrike Lohmann and Sylvaine Ferrachat (ETH) as well as MPI Hamburg and the Center for Climate System Modeling (C2SM) for their support in maintaining this model. ECMWF ERA-Interim data used in this study have been obtained from the ECMWF data server. The work was financially supported by the National Center for Competence in Climate Research (NCCR Climate).