In order to calculate the vertical motion over some high mountain regions, such as Greenland, an ω-equation without the quasigeostrophic approximation in σ-coordinates has been developed. A dynamic method for retrieving precipitation over Greenland is based on this ω-equation. The retrieved annual mean precipitation distribution for 1987 and 1988 is in very good agreement with the observed annual accumulation pattern over the Greenland Ice Sheet.
The major weather system producing precipitation over Greenland is the frontal cyclone. Based on the precipitation characteristics, Greenland can be divided into five subregions. Precipitation over the north coastal and central interior regions primarily occurs in summer. For the three other subregions, if the composite monthly mean sea level pressure charts for high and low monthly precipitation amounts are constructed, a clear relationship between precipitation and cyclonic activity emerges. If a mean cyclone exists in the Labrador Sea, heavy precipitation will fall over Greenland during that month. By contrast, if a mean cyclone exists near Iceland, precipitation over Greenland will be reduced. This is an important relationship between Greenland precipitation and cyclonic activity.
The cyclonic tracks near Greenland are established. A synoptic example is used to show the relation between precipitation and a cyclone moving up the west side of Greenland (track B) combined with movement across the southern tip of the island (track C). In this example, lee cyclogenesis is caused by the southern part of the Greenland Ice Sheet. The lee cyclone develops on the east coast along track C. During lee cyclogenesis, heavy precipitation falls over the southern region. The “parent” cyclone moves along track B, and precipitation falls along the west coast of Greenland.
A possible feedback between cyclonic activity and the mass balance of the Greenland Ice Sheet is proposed. On the one hand, cyclonic activity has a significant influence on snow accumulation over the ice sheet. The development of Icelandic cyclones is not favorable for precipitation over Greenland. On the other hand, the Greenland Ice Sheet has an important dynamic effect in producing lee cyclogenesis and affecting the frequency of Icelandic cyclones. This possible feedback may be important for understanding how the mass balance of the Greenland Ice Sheet and the Icelandic low are maintained in the present climate state.
Recently, dramatic and rapid climate change during and near the end of the last ice age has been found by analyzing layers of ice from deep cores drilled from the Greenland Ice Sheet (e.g., Johnsen et al. 1992; Alley et al. 1993; Taylor et al. 1993). In order to help understand such events, it is necessary to investigate the controls on the present precipitationregime over Greenland.
The observations of Greenland precipitation are limited and generally inaccurate (Bromwich and Robasky 1993). In particular, gauge measurements of the mostly solid precipitation are strongly contaminated by wind effects and are primarily confined to the complex coastal environment. Snow accumulation (the net result of precipitation, sublimation/evaporation, and drifting) determinations from the ice sheet are limited in space and time. Thus, precipitation must be determined indirectly. However, the analyzed wind, geopotential height, and moisture fields are available for recent years. The precipitation should be retrievable from these fields by a dynamic approach.
The annual precipitation over the Greenland Ice Sheet has been studied by Bromwich et al. (1993); they used a statistical diagnostic model to calculate precipitation based upon the advection of relative geostrophic vorticity at the 500-hPa level. The computed mean annual precipitation over Greenland and its adjacent region is shown in Fig. 1a. The observed mean annual accumulation distribution over Greenland Ice Sheet in centimeters of water equivalent from Bender (1984) is shown in Fig. 1b. In comparison to the spatial distribution of mean annual accumulation (Bender 1984; Ohmura and Reeh 1991), the modeled annual mean result of Bromwich et al. (1993) captures the main features. The major shortcoming of the simulated distribution is the lack of a band of higher values that runs along the western coast of Greenland, from Jakobshavn to Thule, where the precipitation is substantially underestimated. This shortcoming is primarily caused by orographic effects. In addition, their method is not accurate enough to study precipitation on a daily basis. In order to obtain a more realistic depiction of precipitation over Greenland, the precipitation retrieval method needs to be further enhanced.
A natural method for retrieving precipitation over Greenland from the observed wind, geopotential height, and moisture fields is using a four-dimensional data assimilation (FDDA) system. The National Centers for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR) (Kalnay et al. 1996) are cooperating in a project (denoted “reanalysis”) to produce a 40-yr (1957–96) record of global analyses of atmospheric fields, which includes the global recovery of precipitation. The model used for data assimilation has a resolution of T62 (triangular truncation at 62 waves), which is equivalent to about 210 km, with 28 vertical levels. Except for the horizontal resolution, the model is identical to the NCEP global operational model implemented on 10 January 1995, which has a resolution of T126 (105 km). The 40 yr of reanalysis will be completed in 1997.
Recently the mean results of the reanalysis for 1982–94 have been published (Kalnay et al. 1996). The mean annual precipitation over the Greenland region for 1982–94 from the reanalysis is shown in Fig. 1c. Comparing Fig. 1c with Fig. 1b, it can be seen that there are at least two important shortcomings in the distribution of the mean precipitation over Greenland produced by the reanalysis. One is that the mean annual precipitation is too large, with the highest amount in the southeast coast being 388 cm yr−1. The precipitation amounts from the reanalysis are about twice those observed along the southeast and southwest coasts where high precipitation is usually observed. The other shortcoming is that there is a high precipitation area in the centralregion of Greenland with a maximum of 110 cm yr−1. However, the observed accumulation in the central interior of Greenland is very low (Fig. 1b). The mean annual precipitation in this region is about 10 cm yr−1. This amount is also verified by Bolzan and Strobel (1994) in a study over a 180 km × 180 km region centered at Summit, Greenland, for the mean distribution and interannual variation of accumulation during 1964–86. The precipitation over central Greenland near Summit from the reanalysis is about 10 times that of the observed.
From Fig. 1c, it can be seen that a wave train pattern of the mean annual precipitation emits toward the southeast from the southeast coast of Greenland. This wave train pattern does not exist in the real atmosphere. Because the topography of the Greenland Ice Sheet is very steep, especially along its southeastern coast, the model used in the assimilation must be able to deal properly with these complicated mountains and the horizontal resolution of the model needs to be higher than 210 km. Recently, Higgins et al. (1996) also found that NCEP/NCAR reanalysis overestimates daily mean (May 1985–89) precipitation rates by a factor of almost 2 over the southeastern United States. However, in general, the errors of precipitation over plains in the midlatitudes from the reanalysis are not as large as those over Greenland.
From the above, it seems that a high resolution limited-area FDDA system that can deal properly with the complicated mountains is needed to retrieve precipitation over Greenland. However, FDDA systems for limited regions are still under development. On the other hand, this method is computationally very demanding. Because we only want to determine some basic features of the precipitation over Greenland, a relatively simple approach based on a vertical motion calculation is adopted to improve upon the method of Bromwich et al. (1993). This is the first goal of this paper.
The ω-equation based on the quasigeostrophic approximation is a useful tool for computing the vertical velocity. Because the quasigeostrophic approximation is not accurate enough, a number of generalized ω-equations with fewer assumptions have been developed (e.g., Krishnamurti 1968a,b; Tarbell et al. 1981; DiMego and Bosart 1982; Iversen and Nordeng 1984; Hirschberg and Fritsch 1991; Pauley and Nieman 1992; Raisanen 1995). However, the quasigeostrophic ω-equation and its generalizations mentioned above always use p-coordinates in which the boundary condition at the earth’s surface is usually given at the 1000-hPa level. The topography over Greenland is very complex and the 850-hPa level is actually below the earth’s surface. Orographic effects are difficult to handle with the ordinary p-coordinate system in the Greenland region.
The σ-coordinate is popular and has been used successfully to represent orographic effects in numerical models. In order to improve the capability of σ-coordinates to deal with complicated mountains, an isobaric geopotential height has been proposed by Chen and Bromwich (1996). This concept can be used to extend the generalized ω-equation to σ-coordinates. For this purpose, an ω-equation based on the isobaric geopotential height in σ-coordinates without the quasigeostrophic approximation has been developed (Chen and Bromwich 1996). The dynamicmethod for retrieving precipitation is based on this study. A brief description of the dynamic method is given in section 2. The retrieved precipitation for 1987 and 1988 is discussed in section 3.
Bromwich et al. (1993) used the region of high synoptic-scale (2–6 day) variability of the 500-hPa height field to represent the “storm track” and found that the interannual variation of precipitation over Greenland was related to the strength and position of the storm track. Kapsner et al. (1995) compared accumulation rates and proxy temperatures derived from ice in the deep core obtained by Greenland Ice Sheet Project 2 (GISP2). They found that atmospheric circulation, not temperature, seems to have been the primary control on snow accumulation in Greenland over the past 18 kyr. During both warm (Holocene) and cold (Younger Dryas, Last Glacial Maximum) climate regimes, the sensitivity of accumulation to temperature changes is less than expected if accumulation is controlled thermodynamically by the ability of warmer air to deliver more moisture. During the transitions between warm and cold climate states, by contrast, accumulation varies more than can be explained by temperature changes. Thus, circulation changes may be more important than direct temperature effects in determining snow accumulation in Greenland.
It can be seen from the above that the relation of precipitation over Greenland to the large-scale circulation and cyclonic activity is important. The second purpose of this paper is to use the daily retrieved precipitation and synoptic analysis to investigate this relationship. Questions to be considered include the following. What is the major weather system producing precipitation over Greenland? What are the basic features of cyclonic activity near Greenland? How is this cyclonic activity related to the precipitation over the ice sheet? Is there any interaction between the precipitation produced by cyclonic activity and the Greenland Ice Sheet? The relationship between precipitation and cyclonic activity is discussed in section 4. Cyclone tracks near Greenland are shown in section 5, and their relationship to precipitation over the ice sheet is illustrated by a synoptic example in the appendix. Based on the analyzed results, a possible feedback between cyclonic activity and the mass balance of the Greenland Ice Sheet is proposed in section 6. This feedback can be used to understand how the accumulation over the Greenland Ice Sheet and the climate around the North Atlantic Ocean are maintained in the present climate state.
The thermal and dynamic effects of the Greenland Ice Sheet on the synoptic systems and climate over the North Atlantic Ocean as well as the atmospheric controls of the mass balance of the Greenland Ice Sheet have received little evaluation. Meteorologically, Greenland is one of the least studied areas in the Northern Hemisphere. This paper is intended to draw attention to these problems.
2. Brief description of a generalized ω-equation in σ-coordinates and a simplified method for computing precipitation
In order to calculate the vertical motion over some high mountain regions, such as Greenland, the Tibetan Plateau, and Antarctica, an isobaric geopotential height in σ-coordinates has been proposed (Chen and Bromwich 1996). By using this concept, a generalized ω-equation without the quasigeostrophic approximation in σ-coordinates was developed. This method can improve ω calculation over high mountain regions and is used to retrieve precipitation over Greenland in the present paper. A brief description of this generalized ω-equation method is givenbelow.
a. An isobaric geopotential height in σ-coordinates
The vertical coordinate is defined by σ = p/p∗, where p∗(x, y, t) is the surface pressure. The vertical distribution of variables is shown in Fig 2a. In the vertical, 16 σ-levels at σ = 0.025, 0.075, 0.125, 0.170, 0.240, 0.325, 0.430, 0.550, 0.610, 0.670, 0.775, 0.855, 0.915, 0.955, 0.980, and 0.995 are used. Let U, V, Ω, and D be
where m is the map scale factor, δ the horizontal divergence, and ζ the vertical component of the relative vorticity. A variable X ↓ is a column vector denoted by
where (. . .)T is for transpose. Using the continuity equation and vertical finite differencing, the pressure vertical velocity in σ coordinates can be expressed by
where C is a lower-triangular matrix (Chen et al. 1997).
Equation (3) can be used to compute ω directly from the observed wind, and it can be referred to as a kinematic method. However, the horizontal divergence derived from the kinematic method is sensitive to small errors in the observed wind. The divergence derived from the following vorticity, divergence, and thermodynamic equations is determined from terms of similar magnitude and is less sensitive to observational errors than the kinematic method; it is referred to as an ω-equation method. Because the ω-equation method involves computations of higher-order derivatives than the kinematic method, this diminishes the advantages if the computation of derivatives is not accurate enough. Thus, in order to reduce computational errors, a harmonic-Fourier spectral method (Chen and Kuo 1992a,b) is used.
where the inner parts satisfy the Poisson equations
with zero Dirichlet boundary values. The harmonic parts, ψh and χh, are the harmonic functions satisfying the Laplace equations
with the coupled boundary condition
whereuE | Σ and vE | Σ are the external wind at the boundary.
In order to extend the ω-equation to σ-coordinates, it is important to know how the horizontal pressure gradient force is expressed in σ-coordinates. The transformation of the horizontal pressure gradient from p- into σ-coordinates is expressed by
where ϕ(x, y, p, t) is the geopotential height in p-coordinates, and ϕ(x, y, σ, t) is the geopotential height in σ-coordinates. It is seen from (8) that the vector on the left-hand side is irrotational, but the total vector on the right-hand side of (8) is not irrotational because T is a function of x and y. Partitioning a wind field into the irrotational and nondivergent parts can easily be done by Chen and Kuo’s (1992a,b) method. If the total vector on the right-hand side of (8) is partitioned by this method, its irrotational part is expressed by ∇ϕ′i(x, y, σ, t), where ϕ′i is referred to as the inner part of an isobaric geopotential height in σ-coordinates. Based on this method, the inner part ϕ′i can be derived from the solution of the following Poission equation:
with zero Dirichlet boundary value. The isobaric geopotential in σ-coordinates can also be utilized to improve synoptic analysis and model predictions for high mountain regions. Over the Tibetan Plateau, analyses on isobaric surfaces below the 500-hPa level hardly identify weather systems because data are extrapolated from higher levels. However, these systems do exist and they can affect the weather in the downstream region. It is found (Chen and Bromwich 1996) that the contours of the isobaric geopotential at the σ = 0.995 level over the Tibetan Plateau are very similar to those of the streamfunction for synoptic and mesoscale systems, which is the same as the geostrophic wind approximation used on isobaric surfaces over plains.
b. A generalized ω-equation in σ-coordinates
The Coriolis parameter and map scale factor are separated into
where f0 and (m2)0 are the average values in the integration region, and f′ and (m2)′ are their deviations. The inner part of the isobaric geopotential height can be separated into geostrophic and ageostrophic parts. The geostrophic part is expressed by ϕ′ig = f0ψi, while the ageostrophic part is denoted by
If the atmospheric motion is quasigeostrophic, the ageostrophic isobaric geopotential always vanishes.
The vorticity and divergence equations can be transformed into the equations of the inner parts of the streamfunction and velocity potential, respectively, and they are expressed by
where Ei ↓ = [m2(U2 + V 2)i/2] ↓, and the terms ψadv,i ↓ and χadv,i ↓ are the variation rates of the inner parts of the streamfunction and velocity potential caused by advection, respectively. The equation of the ageostrophic isobaric geopotential can be written as
and Φ′had,ia ↓ is referred to as the variation rate of the ageostrophic isobaric geopotential caused by advection and diabatic heating, and A is a matrix (Chen et al. 1997).
If the tendencies of the velocity potential and ageostrophic geopotential in (13) and (14) are neglected, this approximation is referred to as a balanced ageostrophic approximation (Chen et al. 1996). Thus, (14) becomes
Equation (16) is a velocity potential form of the generalized ω-equation for the balanced ageostrophic approximation in σ-coordinates. In this equation, the diabatic and advection terms computed by the ageostrophic wind are the same as those in the generalized ω-equation in p-coordinates (Pauley and Nieman 1992), but the effect of orography on the vertical motion is better described. If (16) is transformed into p-coordinates, and the term Φ′had,ia ↓ is computed by the geostrophic wind and expressed by Φ′had,ia,g ↓, (16) becomes a velocity potential form of the quasigeostrophic ω-equation.
c. A simplified method for computation ofprecipitation
The thermodynamic equation is written as
and the continuity equation for specific humidity is
where C and E are the rates of condensation and evaporation per unit mass and L is the condensation heat per unit mass. The vertical advection of a variable X is approximated by
The advection and adiabatic variations of the temperature and specific humidity are computed by
where the horizontal advection is computed by a spectral method (Chen et al. 1997) and the vertical advection is computed by (20). The vertical velocity σ̇ is obtained from (17) and ω is derived from (16) and (3). Time differencing for (21) and (22) is computed by the Euler-backward or Matsuno scheme.
Large-scale condensation occurs when qi,j,k(t + Δt) is greater than rcq*i,j,k, where qi,j,k(t + Δt) is the specific humidity predicted by (22); q*i,j,k is the saturation specific humidity at temperature Ti,j,k(t + Δt) and pressure pi,j,k. Here, Ti,j,k(t + Δt) is computed from (21) and rc is a critical relative humidity. Excess water in a layer can be obtained by removing supersaturation from the specific humidity field and modifying the temperature through latent heat release. The excess water removed from an atmospheric layer precipitates into the layer immediately below. The falling precipitation either evaporates completely in that layer or brings the layer to saturation and then the excess water passes to the next layer below again. The process is repeated. When the lowermost layer is saturated, the condensed water precipitates onto the ground as snow or rain. The computation procedure is similar to the large-scale condensation and precipitation discussed by Arakawa and Lamb (1977). In this method, both the precipitation and latent heat release are obtained.
At the beginning of the computation for the pressure and sigma vertical velocity, the diabatic heating (L/Cp)(C − E) ↓ vanishes. After the latent heat release is derived, the diabatic heating, (L/Cp)(C − E) ↓, is set to be equal to the computed latent heat release, and then the vertical velocity and precipitation process is calculated again. The whole process is repeated once. Because the effect of latent heat release on the vertical motion is important (Smith et al. 1984; Pauley and Nieman 1992), the above method is used to get a corrected precipitation amount that accounts for this effect.
3. The annual precipitation over Greenland
a. The distribution of the annual precipitation for 1987 and 1988
Data from the EuropeanCentre for Medium-Range Weather Forecasts (ECMWF) at 2.5° × 2.5° resolution (TOGA Archive II from NCAR) for 1987 and 1988 are used. The precipitation is calculated twice per day based on the analyzed data at 0000 and 1200 UTC. The topography of Greenland is presented in Fig. 2b, which shows that the steepest slopes are near the southeast coast. The mesh size of the limited region is 111 × 81 and the grid spacing is 50 km. The time step Δt used here is 30 min. The precipitation rate is only computed for one time step and then it is applied to a 12-h period. According to tests of the precipitation over Greenland, it is better to assume that the critical relative humidity is 0.85.
The retrieved precipitation for 1987 and 1988 is shown in Figs. 3a and 3b, respectively. Figure 3c is the mean annual precipitation for these two years. In order to check if the computed annual precipitation distribution is reasonable, we compare it with the observed mean annual accumulation over the Greenland Ice Sheet shown in Fig. 1b. Although two years are too short for a climatic depiction, the comparison gives a preliminary evaluation of the method.
From Figs. 3a–c, it can be seen that the maximum values of more than 100 cm yr−1 are located along the southeastern coast and southwestern edge of Greenland. A secondary band of relatively high precipitation is present along the western coast of Greenland from Jakobshavn to Thule, which is underestimated in the simulation by Bromwich et al. (1993). In comparison with the observed distribution in Fig. 1b and the results of Bromwich et al. (1993) shown in Fig.1a, it is found that the major shortcoming in the simulation by Bromwich et al. (1993) is corrected by the new method. A large area of very low precipitation about 10 cm yr−1, which dominates the central interior region of Greenland to the north of 70°N, is also shown in Figs. 3a–c.
The annual precipitation over the Greenland region simulated by the NCEP/NCAR reanalysis for 1987 and 1988 is shown in Figs. 3d and 3e, respectively. It can be seen that the annual precipitation pattern over Greenland for an individual year shown in Figs. 3d or 3e is very similar to that of the mean annual precipitation for 1982–94 in Fig. 1c. Comparing Figs. 3d and 3e with Figs. 1b, 3a, and 3b, it can be seen that the two important shortcomings in the mean precipitation over Greenland shown by Fig. 1c are also present in Figs. 3d and 3e. The precipitation amounts from the reanalysis are about twice those observed in Fig. 1b and those retrieved in Figs. 3a–c. There are high precipitation areas in the central region of Greenland with maxima of 111 and 101 cm yr−1 in Figs. 3d and 3e, respectively. They are about 10 times larger than the observed amount and the retrieved values in Figs. 3a–c. A wave train pattern extending southeastward from the southeast coast of Greenland is also found in the annual precipitation in Figs. 3d and 3e for 1987 and 1988.
It is seen fromthe above that the precipitation distributions over Greenland retrieved for 1987 and 1988 by the simple dynamic method presented in this paper are superior to those obtained by NCEP/NCAR reanalysis and are accurate enough to be used for study of the basic features of precipitation over Greenland.
b. The annual mean circulation at the 700-hPa level and vertical motion for 1987 and 1988
The height of the Greenland Ice Sheet in the central interior region is approximately at the 700-hPa level. The annual mean wind, geopotential height, and temperature at the 700-hPa level are shown in Fig. 3f. There is a pronounced minimum in the temperature field over the central interior region. Figure 3g is the annual mean vertical velocity at the 700-hPa level. The vertical velocity is calculated twice per day at 0000 and 1200 UTC, and then these computed values are averaged over two years of 1987 and 1988 to obtain the annual mean vertical motion. Figure 3g shows that a relatively large area with mean descending motion occurs in the central interior region.
The large area of very low precipitation that dominates the central interior region of Greenland to the north of 70°N (Figs. 3a–c) may be closely related to the mean descending motion in the central region shown in Fig. 3g.
Snow and ice surfaces typically have an albedo larger than 80%, implying that more than three-quarters of the incident shortwave radiation from the sun is reflected. Little of this reflected solar radiation is absorbed by the overlying atmosphere. In addition, the ice surface acts nearly like a blackbody for longwave radiation. Thus, the ice surface efficiently radiates thermal energy to outer space. The cold source at the ice and snow surface will reduce the air temperature in the boundary layer by downward sensible heat transfer. On the other hand, an adiabatic warming must be produced by the descending motion in the central region shown in Fig. 3g, and it may primarily be compensated by the temperature decrease due to the diabatic cooling. Furthermore, the descending motion would minimize cloud cover and promote radiative cooling. Thus, the cold center shown in Fig. 3f is probably also related to the intense cooling at the ice surface, which markedly contrasts with the relatively warm air at some horizontal distance away in the free atmosphere.
Scorer (1988) published a paper entitled “Sunny Greenland.” In comparison to other places along the same parallel of latitude, the precipitation over the central interior region of Greenland is very small, thus the weather in the summer half year in this region may be very sunny. However, it is also seen from Figs. 3a–c that most of the southern part of Greenland, south of about 68°N, has heavy precipitation of more than 100 cm yr−1. Thus, “Sunny Greenland” does not apply to the entire island. Scorer (1988) also mentioned that two major parties of explorers were trapped by deep snowfalls for a few months in the 1920s. These falls had not been expected because of the belief that the southern part had the same sunny weather.
c. The annual precipitation over the entire Greenland Ice Sheet
As a check on the magnitudes of the simulated annual precipitation, the annual precipitation over the entire Greenland Ice Sheet was computed. This was done by finding all grid points that lie over Greenland, multiplying their annual precipitation values by the area of the grid square centered on that point, and then summing all these products. Our computed resultsand those obtained by other investigators including those based on glaciological observations are compared in Table 1. Examination of Table 1 shows that the model’s values of annual accumulation are close to those obtained previously.
The interannual variations of precipitation over Greenland were studied by Bromwich et al. (1993) and also estimated from the atmospheric moisture budget by Robasky and Bromwich (1994). These two computationally independent estimates matched closely for the period 1980–89. The 2-yr mean values for 1987 and 1988 from Bromwich et al. (1993) and Robasky and Bromwich (1994) are 651 and 664 km3, respectively, while in our computation, this 2-yr mean value is 638 km3. The values for 1987 from Bromwich et al. (1993) and Robasky and Bromwich (1994) are 19 and 66 km3 larger than those for 1988, respectively, while in our computation, the value for 1987 is also 26 km3 larger than that for 1988. Thus, the mean values for these two years computed by the above three methods are very close, and the tendency of the variations between these two years has the same sign.
d. Precipitation over Greenland assimilated by the NCEP/NCAR reanalysis and T126 model
To determine if the precipitation over Greenland predicted by the global model of NCEP is affected by the horizontal resolution of the model, the precipitation rates derived from the operational global T126 (105 km) model for NCEP (Kanamitsu et al. 1991) are also examined. The precipitation rates with a 1° × 1° resolution predicted by the T126 model are obtained from the NCEP GRIB (Gridded Binary) dataset. In this paper we do not want to study the difference of the precipitation over Greenland predicted by the T62 and T126 models, and we only have the NCEP GRIB data based on the T126 model since 1994. Thus, only some examples of the precipitation retrieved by the simple dynamic method are compared with the computed precipitation rates derived from the operational global T126 model.
The NCEP GRIB dataset only has daily precipitation rates. The precipitation rates at 1200 UTC 24 and 25 July 1994 retrieved by the dynamic method are shown in Figs. 4a and 4b, while the corresponding precipitation rates from the NCEP GRIB data based on the T126 model are shown respectively in Figs. 4c and 4d. Comparing Figs. 4a and 4b with 4c and 4d, respectively, it can be seen that the distribution of the major precipitation centers obtained from these two methods is similar. These dates for comparison were chosen at random. The monthly precipitation for July 1994 retrieved by the dynamic method is shown in Fig. 4e. The monthly precipitation computed from the precipitation rates of the NCEP GRIB dataset is based on a method that is the same as that used in the dynamic method, and its value for July 1994 is shown in Fig. 4f. The monthly precipitation for July 1994 from the NCEP/NCAR reanalysis based on the T62 model is shown in Fig. 4g. Comparing Figs. 4e and 4f, it is seen that the distribution of monthly precipitation amounts obtained from the dynamic method and from the NCEP GRIB data based on the T126 model are similar, but the precipitation over the southeast coast of Greenland retrieved by the dynamic method is a little larger. As shown in Fig. 4g, the two important shortcomings and thefalse wave train pattern shown in Fig. 1c are also present in the monthly precipitation for July 1994 based on the NCEP/NCAR reanalysis. However, these shortcomings do not occur in the predicted total precipitation for July 1994 based on the T126 model shown by Fig. 4f. Therefore, in order to obtain a better assimilation of precipitation over Greenland in the NCEP/NCAR reanalysis, it seems necessary to use a higher resolution model, such as T126, instead of T62.
Recently, Genthon and Braun (1995) used the predicted precipitation of the ECMWF T106 model (about 1.125°) from 6 yr of the ECMWF archive (May 1985 to April 1991) to study the annual mean accumulation (precipitation minus evaporation or sublimation) over Greenland. Comparing Fig. 3c with the annual mean accumulation shown in Fig. 6 of Genthon and Braun (1995), the distributions of these two figures are quite similar. The amounts in the high precipitation region shown in Fig. 3c are about 10 cm yr−1 larger than those shown in Fig. 6 of Genthon and Braun (1995) because the precipitation in Fig. 3c does not take the evaporation/sublimation loss into account.
4. Intraannual variations of precipitation over different regions of Greenland and their relation to cyclonic activity
a. The major system producing precipitation over Greenland
Based on evaluation of the daily isobaric analyses and retrieved precipitation for 1987 and 1988, it is found that the major weather system producing precipitation over Greenland is the frontal cyclone. Cyclonic activity, in general, includes the distribution of cyclone frequency, the locations of cyclogenesis and the tracks followed by cyclones. Cyclonic activity is very important for the climatological features of precipitation over Greenland.
b. Three major regions of Greenland
According to the synoptic survey of precipitation and possible effects of orography on moving cyclones, the Greenland Ice Sheet may be divided into three major regions separated by the solid lines shown in Fig. 5. The first is referred to as the sourthern region and mostly lies to the south of about 68°N. When a cyclone passes close to the southwest coast of Greenland or Cape Farewell on the southern coast, heavy precipitation often occurs over the southern region. This situation sometimes causes lee cyclogenesis near the east coast.
The second region is located to the north of the southern region, and it is referred to as the central region. Its width is about 1200 km, and the highest elevation is more than 3000 m. Based on the survey of 1987 and 1988, it is found that the central region of Greenland has an important blocking effect on moving cyclones. No cyclone moving from west to east across this region was found during these two years. According to the precipitation characteristics, the central region can also be divided into three subregions, which are the central west coastal region, central interior region, and central east coastal region, as shown in Fig. 5. It has been pointed out in the previous section that the central interior region has annual average descending motion and very low precipitation.
The third region is the north coastal region located to the north of 80°N and is a narrow area that slopes steeply down to sea level. A few cyclones influence this region during summer.
c. Seasonal variations of precipitation over the north coastal and central interiorregions, and composite charts for high precipitation months in summer
In order to show some intraannual variations of precipitation over Greenland, the monthly mean precipitation for 1987 and 1988 averaged for the five regions and for all of Greenland are given in Figs. 6a–f. This mean precipitation is computed by finding all N grid points that lie in the region, summing their monthly precipitation values, and then dividing by N.
Figures 6a and 6b show the mean monthly precipitation averaged over the north coastal and central interior regions. It is seen from these two figures that mean monthly precipitation has an obvious seasonal variation, especially in the north coastal region. The precipitation in the north coastal region is much larger in summer (June, July, and August) than in winter (December, January, and February). In the central interior region, the amplitude of the variation is very small, but the values in summer are also larger.
Figures 7a and 7b are the composite monthly mean sea level pressure (SLP) and monthly mean wind, temperature, and geopotential height at the 700-hPa level, respectively, for the high precipitation months in the north coastal and central interior regions. Because the sea level pressure over Greenland can only be extrapolated and computed from the temperature at upper levels, it is too high over the Greenland Ice Sheet.
The monthly total precipitation is produced by the sum of the behavior of individual synoptic systems within the month and cannot be completely deduced from the monthly mean synoptic maps. However, a relationship between the monthly total precipitation and monthly mean circulation approximately exists, similar to what has been found for the coastal regions of Antarctica (Bromwich et al. 1995). It is seen from Figs. 7a and 7b that there is a mean cyclone located in the Arctic Ocean close to the northeast coast of Greenland and Spitzbergen during the high precipitation months in summer. In this case, the north coast is affected by northwesterly winds associated with the mean cyclone. This flow is forced to ascend over the slopes of the north coastal region and causes precipitation.
In order to show possible cyclone track information for high precipitation months over the north coastal region, a Lanczos filtering scheme (Duchon 1979) for 2.5–6-day variations of SLP was applied to the twice daily ECMWF data for the summer months (June–August) of 1987–88. The average synoptic SLP variability for summer is shown in Fig. 7c. The axis of maximum variability corresponds to the cyclone track. It is seen from Fig. 7c that a relatively large center is located over Baffin Bay and trends northeastward through Nares Strait. These features reveal that the cyclones often develop over Baffin Bay in summer and then move northeastward through Nares Strait. There are other centers of variability in Fig. 7c located to the southeast of Iceland and over the Arctic Ocean and Scandinavia. These maximum value regions are related to the tracks of the Icelandic and other cyclones, whose intensity is stronger than that of the cyclones over Baffin Bay. It should be noted that axes of variance maxima do not always coincide with traditional storm tracks because alterations in the position and intensiy of anticyclones can also contribute significantly to variance maxima.
d. Composite charts for high and low precipitation months over the southern region: Labrador Sea and Icelandic cyclones
Except for the monthly precipitation over the north coastal and central interior regions discussed above, it is very difficult from Figs. 6c–f to find obvious seasonal precipitation variations. However, if composite charts for high and low monthly precipitation over a region are constructed without making any allowance for seasonal variation, some interesting results are found.
Figure 6c is the mean monthly precipitation averaged over the southern region of Greenland. The majority of the precipitation over all of Greenland occurs in the southern region; thus, the monthly mean precipitation over this region can be used to represent the monthly mean precipitation over the entire ice sheet. Figure 6f is the mean monthly precipitation averaged over all of Greenland. Comparing Figs. 6c and 6f, it can be seen that the values are smaller in Fig. 6f due to the averaging over a larger area, but the months with maximum and minimum values are nearly the same.
A month for which the monthly mean precipitation averaged over the southern region of Greenland is larger than 80 cm yr−1 or smaller than 50 cm yr−1 is referred to as a high or low precipitation month, respectively. Figures 7d and 7eare the composite monthly mean SLP and wind, geopotential height, and temperature fields at the 700-hPa level, respectively, for the high precipitation months (4 months in the sample) over the southern region. It is seen from Fig. 7d that there is a mean cyclone located in the Atlantic Ocean and Labrador Sea close to the southwest coast of Greenland. At the 700-hPa level shown in Fig. 7e, there is a trough in the Labrador Sea close to the east coast of Labrador. The southern region of Greenland has warm air advection and is located in advance of this trough, where the cyclonic vorticity advection is often larger at upper levels than lower in the troposphere. Such conditions are favorable for dynamically supported ascending motion on individual synoptic maps (Chen 1981, 1987; Sanders and Hoskins 1990) and may approximately be applied to monthly mean conditions. In addition, both the southwest and southeast coasts of this region are directly impacted by the onshore flow from the Atlantic Ocean. This flow would have relatively high humidity and produces precipitation over the coastal slopes by orographic lifting.
Figures 7f and g are the composite monthly mean SLP and 700-hPa level charts, respectively, for the low precipitation months (5 months in the sample) over the southern region. Figure 7f shows that a monthly mean cyclone is located in the Atlantic Ocean close to Iceland. Comparing Figs. 7e and 7g, the northern part of the trough at the 700-hPa level shown in Fig. 7e has moved eastward and is located over Davis Strait, but its southern part is not clear and seems to have moved to the east of Greenland. With this circulation, the southern part of Greenland is located to the rear of the southern part of the 700-hPa trough and to the northwest of the surface cyclone, and the temperature advection is not clear and very small. Both the baroclinic and orographic effects favorable for precipitation in Figs. 7d and 7e are not present in Figs. 7f and 7g.
The average synoptic SLP variability for high and low precipitation months over the southern region are shown in Figs. 7h and 7i, respectively. The variability for high precipitation months over the Labrador Sea is larger than 6 hPa, and the axis of maximumvalues (cyclone track) has two routes. One goes northward through Davis Strait along the west coast of Greenland. The other turns eastward and is associated with lee cyclogenesis. This lee cyclogenesis is similar to the cyclones that form in the lee of the Rocky Mountains, known as “Colorado” and “Alberta” cyclones (Palmen and Newton 1969). Lee cyclogenesis is usually associated with a “parent” cyclone upstream along the west coast of the southern region. When the parent cyclone moves eastward across the southern region, a discontinuous movement may occur in the sea level pressure field. The parent cyclone often decays along the west coast or sometimes moves northward through Davis Strait, while a lee cyclone develops along the east coast. This lee cyclogenesis process is revealed by Fig. 7h, and an example will be shown in the appendix.
Figure 7i shows that the axis of maximum variability is over the Atlantic Ocean to the southeast of Greenland. This is the cyclone track for low precipitation over the southern part of Greenland, and it passes through the region near the southeast coast of Iceland from southwest to northeast.
The mean cyclone located to the southwest of Greenland over the Labrador Sea as shown in Fig. 7d is referred to as a Labrador Sea cyclone. For the same reason, the mean cyclone located to the southeast of Greenland over the Atlantic Ocean near the south coast of Iceland as shown in Fig. 7f is referred to as an Icelandic cyclone. The Icelandic cyclone in Fig. 7f is similar to the Icelandic low present on the long-term averaged mean SLP charts.
e. Composite charts for high precipitation months over the central west and central east coastal regions
Figures 6d and 6e are the mean monthly precipitations averaged over the central west coastal and central east coastal regions of Greenland, respectively. A month for which the monthly mean precipitation averaged over one of these coastal regions is larger than 40 cm yr−1 is referred to as a high precipitation month for that coastal region. Figure 7j is the composite monthly mean SLP for the high precipitation months (6 months in the sample) over the central west coastal region. Figure 7j shows that there is a mean trough along Davis Strait and Baffin Bay and a small mean cyclone center located in Baffin Bay. This cyclone produces the precipitation over the central west coastal region.
Figure 7k is the composite monthly mean SLP for the high precipitation months (3 months in the sample) over the central east coastal region. It shows that there is a mean cyclone located near the southeast coast, but a mean trough extends northward along Denmark Strait. The easterly flow associated with this trough onto the east coast causes precipitation over this region.
f. A relationship between precipitation over Greenland and cyclonic activity
An SLP monthly mean cyclone located in a region means that, during this month, cyclones often pass through or they persist or develop with high frequency. Because most of the precipitation over Greenland falls in the southern region, from Figs. 7d and 7f, it can be seen if the Labrador Sea cyclone occurs, enhanced precipitation will fall over Greenland during that month. In contrast, if the Icelandic cyclone occurs, precipitation over Greenland will be reduced. This is an important relationship between precipitation over Greenland and cyclonicactivity.
5. Cyclone tracks near Greenland
Based on the position of the centers of the cyclones plotted on the twice daily maps and the synoptic SLP variability in Figs. 7c, 7h, and 7i, the cyclone tracks around the Greenland region are summarized by the five classes shown schematically in Fig. 8. The solid lines represent primary cyclone tracks, while the dashed denote secondary tracks.
Cyclone tracks for North America and the Northern Hemisphere were studied by Whittaker and Horn (1981, 1984) using 20 yr of data for the period 1958–77. They primarily concentrated on the middle latitudes and the results are shown for January, April, July, and October. Tracks A and B shown in Fig. 8 are also present in Whittaker and Horn’s (1984) analyses for all 4 months.
Track A is related to the sea level circulation dominated by the Icelandic low, for which most cyclones come from the North American mainland and coastal area. Track B is a major storm track into Baffin Bay from the south as well as from Hudson Bay. This track sometimes extends into the Arctic Ocean.
Track C represents a cyclone moving across the southern part of the Greenland Ice Sheet and tracking through Denmark Strait. A closed low that has approached Greenland disappears at the west coast and redevelops along the east coast and over Denmark Strait. The dotted line expresses the discontinuity of the cyclone movement on the SLP chart. Sometimes a cyclone “splits” into two centers, one center traveling northward into Baffin Bay and the other northeastward across the southern part of Greenland into Denmark Strait. Tracks B and C are related to the maximum variability of SLP shown in Fig. 7h, which is for high precipitation over the southern region of Greenland. Track A is related to the maximum variability of SLP shown in Fig. 7i for reduced precipitation over southern Greenland.
During summer, the polar vortex in the upper troposphere contracts rapidly and the mean positions of the polar and arctic front jets move northward. Some cyclones approach Greenland from the west. Track D represents these cyclones but very few of them cross the Greenland Ice Sheet. Once in Baffin Bay, these lows tend to persist and decay gradually.
Some cyclones are formed within Baffin Bay itself. Track E is closely related to the maximum SLP variability shown in Fig. 7c, and it sometimes causes summer precipitation over the north coastal region. The cyclones along this track are formed in Baffin Bay and move northeastward through Nares Strait.
A synoptic example that combines tracks B and C is shown in the appendix.
6. A possible feedback between the Greenland Ice Sheet and cyclonic activity
The thermal and dynamic effects of the topography of the Greenland Ice Sheet have a strong influence on the weather and climate, in the same manner as the Rocky Mountains and the Tibetan Plateau. However, the Greenland Ice Sheet is quite different from other mountains because its topography is generated by the climate system itself. The change of the mass balance of the ice sheet is produced by the difference between two aspects: mass addition from precipitation; and mass loss from runoff and iceberg calving from the margins. In the mean state for the present climate, these two components are roughly in equilibrium. Thus, precipitation over Greenland is an important factor affecting ice sheet mass balance. If precipitation over Greenland were markedly increased or decreased, the topography of the Greenland Ice Sheet would also change. The amount of ice contained by theGreenland Ice Sheet is also directly related to global sea level.
As shown in section 5, precipitation over Greenland is primarily determined by the cyclonic activity (see also Robasky and Bromwich 1994). Thus, the mass balance of the Greenland Ice Sheet is closely related to the cyclonic activity near Greenland.
If the relationship between temperature change and snow accumulation over Greenland needs to be studied, it is necessary first to investigate how the cyclonic activity is influenced by the temperature change; then the effect of temperature change on snow accumulation may easily be understood. Based on the comparison of past snow accumulation amounts with inferred temperature changes from a deep Greenland ice core, Kapsner et al. (1995) also found that circulation changes may be more important than direct temperature effects in determining snow accumulation in Greenland.
On the one hand, cyclonic activity has a significant influence on the snow accumulation over Greenland. As shown in section 5, if Labrador Sea cyclones develop more frequently, precipitation will increase over Greenland. By contrast, if Icelandic cyclones develop more frequently, precipitation over Greenland will be reduced. On the other hand, the Greenland Ice Sheet also has an important dynamic effect on cyclonic activity. For example, many of the Icelandic lows are lee cyclones produced by the Greenland Ice Sheet, and they are clearly and repeatedly positioned to the east of Greenland near Iceland. Thus, there is an important feedback between the Greenland Ice Sheet and cyclonic activity. If the Greenland Ice Sheet were to be lower than at present, the frequency of Icelandic cyclones would be reduced. This may cause more precipitation over Greenland and lead to ice sheet growth. Then, Icelandic cyclones would develop more frequently, limiting the size of the ice sheet. The feedback between the Greenland Ice Sheet and the frequency of Icelandic cyclones must be an important factor in the present climate regime for maintaining the mass balance of the Greenland Ice Sheet and for positioning the Icelandic low.
In the present climate, many climatic features near Greenland are related to the Icelandic low. It produces the warm moist air flow and the warm North Atlantic current moving toward western Europe and Scandinavia. These processes continue nearly all the year round and are the main causes of the mild climate of western Europe and Scandinavia (Scorer 1988). The Icelandic low also produces the ocean current that moves sea ice southward along Greenland’s east coast. Therefore, maintenance of the present climate around the North Atlantic region is closely related to the feedback between the Greenland Ice Sheet and cyclonic activity.
Of course, the present climate around the North Atlantic region is maintained by a series of interactions and feedbacks among the atmospheric circulation, oceanic circulation, and Greenland Ice Sheet. However, they can only be studied step by step. The possible feedback between the Greenland Ice Sheet and cyclonic activity may play an important role in the present climate as well as climate change around the North Atlantic region, and it needs to be studied further.
Based on the studies shown in the above sections, the following conclusions can be reached.
1) In order to calculate the vertical motion over some high mountain regions, such as Greenland, a generalized ω-equation without the quasigeostrophic approximation in σ-coordinates was developed (Chen and Bromwich 1996). In this method, an isobaric geopotential height in σ-coordinates is utilized. The dynamic method for retrieving precipitation over Greenland is based on this ω-equation. In comparison with the precipitationderived from the NCEP/NCAR reanalysis based on a T62 model, it is found that the precipitation distributions over Greenland retrieved for 1987 and 1988 by the dynamic method presented in this paper are superior. Thus, the dynamic method is very useful for studying the basic features of precipitation over Greenland.
The precipitation distribution over Greenland for July 1994 retrieved by the dynamic method is also compared with that produced by the operational global T126 (105 km) model of NCEP. The precipitation distributions obtained by these two methods are similar. In order to obtain a better assimilation of precipitation over Greenland in the reanalysis, it seems necessary to use a model with higher resolution than T62, such as T126.
2) The retrieved distribution of annual mean precipitation over Greenland for 1987 and 1988 is checked against the observed annual accumulation distribution over the Greenland Ice Sheet, and good agreement is obtained.
From the computed results, maximum values of more than 1000 mm yr−1 are located along the southeastern coast and southwestern edge of Greenland. A secondary band of relatively high precipitation is present along the western coast of Greenland. The major shortcoming in the simulation by Bromwich et al. (1993) is thus corrected by the new method. A large area of very low precipitation dominates the central interior region of Greenland to the north of 70°N. Based on ω calculations, mean descending motion is also found in this region.
3) The major weather system producing precipitation over Greenland is the frontal cyclone. Based on the precipitation characteristics, Greenland can be divided into five subregions shown by Fig. 5. Precipitation over the north coastal region and central interior region has a noticeable seasonal variation with larger values during summer than the other seasons. For the other three regions, if the composite monthly mean SLP charts for high and low monthly precipitation over these regions are constructed, the relationship between precipitation and cyclonic activity appears very clearly. The monthly mean composite charts show that if Labrador Sea cyclones develop more frequently, more precipitation will fall over Greenland during that month. By contrast, if Icelandic cyclones develop more frequently, precipitation over Greenland will be reduced. This is an important relationship between precipitation over Greenland and cyclonic activity.
4) The cyclonic tracks have been studied and shown in Fig. 8. A synoptic example in the appendix is used to show the relation of combined tracks B and C to the precipitation over Greenland. In this example, lee cyclogenesis is caused by the southern part of the Greenland Ice Sheet. The lee cyclone develops on the east coast of Greenland along track C. During lee cyclogenesis, heavy precipitation falls over the southern region. The “parent” cyclone moves along track B, and precipitation falls along the west coast of Greenland.
5) Based on the analyzed results, a possible feedback between cyclonic activity and the mass balance of the Greenland Ice Sheet has been proposed. On the one hand, cyclonic activity has a significant influence on snow accumulation over Greenland. The development of Icelandic cyclones is not favorable for precipitation over Greenland. On the other hand, the Greenland Ice Sheet has an important dynamic effect by causing lee cyclogenesis that influences the frequency of Icelandic cyclones. This is a feedback between the mass balance of the Greenland Ice Sheet and the cyclonic activity. It may be important for maintaining the present climate around the North Atlantic region.
This study is based on only two years of data and is preliminary. The results need to be checked by analysis of a much longer period.
We would like tothank Dr. E. M. Rasmusson and two anonymous reviewers for their valuable comments. Special thanks go to R. I. Cullather for his helpful discussions and for performing some of the computations and producing some of the figures. This research was sponsored by NOAA/Office of Global Programs via Grant NA36GP0314-01 and by NASA under Grant NAGW-3677, Supplement 3.
A Synoptic Example Combining Tracks B and C
The purpose of this example is to show how the precipitation over Greenland is influenced by a Labrador Sea cyclone and cyclonic tracks B and C.
Figures 9a–i show an example of the development and movement of a cyclone near Greenland from 0000 UTC 15 January to 0000 UTC 17 January 1988. It is a combined example of tracks B and C. In Fig. 9a, a cyclone moves to the region near the southwest coast of Greenland, and it is equivalent to the Labrador Sea cyclone shown in the composite chart Fig. 7d. Figure 9e is the wind, geopotential height, and temperature fields at the 500-hPa level for 0000 UTC 15 January. It is seen from this figure that a trough was located over Labrador with a diffluence of the height contours over the southern part of Greenland. A relatively strong arctic frontal jet with a thermal ridge and positive temperature advection is present in the advance of this trough. The surface Labrador Sea cyclone continues to develop in the advance of the upper trough in the next 24 h as shown by Fig. 9b. After another 12 h, a low pressure center develops on the lee side of the southern part of the Greenland Ice Sheet shown in Fig. 9c. At the same time, the cyclone at the west coast (“parent” cyclone) weakens and moves northward. About 12 h later at 0000 UTC 17 January, shown in Fig. 9d, the lee cyclone continues to develop but the parent cyclone moves further northward through Davis Strait and Baffin Bay.
From Fig. 9e, it can be seen that there is a southerly current over Davis Strait, Baffin Bay, and central and northern parts of Greenland. The parent cyclone moving northward is caused not only by the effect of the orography but also by the southerly flow in the upper troposphere. In this example, the trajectory of the lee cyclone is the same as track C, but the center of the parent cyclone moves along track B.
The precipitation is calculated every 12 h by the method shown in section 2. The 12-h precipitation amounts for the periods ending 0000 UTC 15, 0000 UTC 16, 1200 UTC 16, and 0000 UTC 17 January 1988, which correspond to the time of Figs. 9a–d, are shown in Figs. 9f–i, respectively. From Figs. 9g and 9h, it is seen that heavy precipitation falls on both the west and east coasts of the southern region of Greenland when a deepening cyclone passes along track C. The maximum 12-h precipitation amounts exceed 30 mm. Figures 9h and 9i also show that when a cyclone moves along track B, the precipitation falls along the west coast of Greenland from Jakobshavn to Thule. This example represents a typical distribution of the precipitation over Greenland associated with tracks B and C. It is also an example to show how the precipitation over Greenland is affected by the development of a Labrador Sea cyclone. Twenty-four hours after Fig. 9d, the lee cyclone moved away from the east coast and the precipitation over the southern region ceased (figures omitted).
Corresponding author address: Qiu-shi Chen, Byrd Polar Research Center, Ohio State University, 1090 Carmark Road, Columbus, OH 43210.
* Byrd Polar Research Center Contribution Number 1003.