1. Introduction
Long-term temperature observations show warming trends of variable strength throughout the Arctic, and ample evidence indicates that the Arctic hydrological cycle, including that for Greenland and the Greenland ice sheet (GrIS), is changing (e.g., Serreze et al. 2000; Hinzman et al. 2005; Hanna et al. 2008; Ettema et al. 2009; Mernild et al. 2010). Since the early 1990s, the increase in GrIS surface runoff has followed atmospheric warming (Hanna et al. 2008), explaining half of the recent mass loss of the GrIS (van den Broeke 2009), a mass loss that by 2100 may contribute up to 54 cm sea level equivalent (SLE) (Pfeffer et al. 2008), since model simulations of future climate scenarios point to higher temperatures in this region (Solomon et al. 2007). In addition to the sea level contribution, terrestrial runoff from Greenland is also important for ocean density, the Atlantic meridional overturning circulation (AMOC) (Rahmstorf et al. 2005), and their impacts on the climate system (Bryden et al. 2005).
Model studies have shown that the AMOC may be sensitive to changes in terrestrial freshwater; freshening the surface waters in the northern North Atlantic inhibits deep convection feeding the deep southward branch of the AMOC (e.g., Bryan 1986; Rahmstorf 1995). The AMOC carries warm upper waters into far-northern latitudes and returns cold deep waters southward across the equator. This transport of heat makes a substantial contribution to, for example, the climate of continental Europe, and any slowdown in the overturning circulation could have implications for climate change (Bryden et al. 2005). One of the potential freshwater sources that might have an effect on the AMOC is enhanced freshwater runoff from Greenland.
In spite of the critical role runoff from Greenland plays in Earth’s climate system, there has been limited quantitative information about the spatial and temporal patterns of drainage basins and associated freshwater flux from Greenland, including the GrIS, to the adjacent seas. In addition, information about changes in Greenland spatial runoff patterns, as well as temporal changes spanning years to decades, is still lacking. Only a few watersheds in Greenland [e.g., Mittivakkat in southeast Greenland (65°N) (Mernild and Hasholt 2006; Liston and Mernild 2012), Zackenberg in northeast Greenland (74°N) (Hasholt et al. 2008; Mernild et al. 2008a; Jensen and Rasch 2010), and Kangerlussuaq in west Greenland (67°N) (Mernild et al. 2011a; Hasholt et al. 2012)] include runoff observations. Unfortunately, even these observational datasets have limitations and, taken together, transfer less than 1% of the Greenland runoff to the surrounding oceans. The Zackenberg discharge (volumetric water flow rate expressed, e.g., as m3 s−1) dataset spans 15 years, the Mittivakkat dataset covers 12 years, and the Kangerlussuaq dataset only covers 4 years. For these three locations, the observations only span parts of the runoff season, ranging between one week and approximately three months. In addition, the Zackenberg, Mittivakkat, and Kangerlussuaq data suffer from difficulties measuring accurate runoff values due to, for example, seasonal and interannual changes in bed profile. Kangerlussuaq is particularly problematic because of the strong turbulence and presence of hydraulic jumps at the measurement point, which makes the location less than ideal for measuring precise discharge quantities. For example, the updated runoff values from Kangerlussuaq appear to include uncertainties of up 125% when compared to independent acoustic Doppler current profiler (ADCP) observations (Hasholt et al. 2012). These uncertainties are due to inaccurate soundings of the observed cross section profile at the catchment outlet because the level of the bed can vary dramatically in response to changes in sand deposition and erosion throughout the runoff season. A study by Rennermalm et al. (2012) confirms that changing bed elevations over time, a phenomenon observed at several Watson River tributaries upstream of Kangerlussuaq, are associated with discharge uncertainties of up to 14%–47%. In addition, the updated Kangerlussuaq runoff time series were only compared against four ADCP point observations (Hasholt et al. 2012) and not against the full range of independent observations required for a statistically rigorous analysis. With these weaknesses in mind, the limited river discharge values available for Greenland should be used with caution, especially if one is trying to understand the role Greenland runoff plays in Earth’s climate system.
A recent study by Rignot and Kanagaratnam (2006) estimated GrIS drainage basins areas (related to GrIS calving fluxes), and Lewis and Smith (2009) estimated a hydrological drainage network for the GrIS. Lewis and Smith (2009) suggested that the GrIS drainage-network routes varying amounts of water to the GrIS edge and that the spatially averaged GrIS annual volumetric runoff production, regionally speaking, is highest in southwest and lowest in northeast Greenland, with greater hydrologic activity in western regions than in eastern regions, for a given latitude.
In the study described herein, detailed analyses of both simulated runoff magnitudes and spatial patterns from individual Greenland catchments, as well as runoff changes through time, were performed and evaluated—all in an effort to understand the climate impact and whether changes in melt rates, melt area, and melt duration were the main reason for the increase in runoff from Greenland to the adjacent seas. In addition, this study introduces modeling tools capable of providing the missing connection between terrestrial water fluxes and ocean circulation features such as the AMOC. This climate-feature connection represents a critical link between Greenland snow and ice melt, ocean processes, and the associated interactions and feedbacks within Earth’s climate system.
Here, we examined the GrIS surface mass balance conditions (surface hydrological conditions), including GrIS and Greenland surface runoff, the spatial distribution of Greenland runoff to the adjacent seas, and their changes from 1960 through 2010. This was accomplished by coupling the HydroFlow runoff routing model (Liston and Mernild 2012) with SnowModel, a spatially distributed, meteorological and snow and ice evolution model (Liston and Elder 2006a, 2006b; Mernild et al. 2006b). The coupled modeling system was run over the GrIS and all surrounding land and peripheral glaciers and ice caps for the period 1960 through 2010 (Fig. 1a). As part of the model simulations, HydroFlow divided all of Greenland, including the GrIS, into individual drainage basins (Fig. 1b) and simulated the associated grid connectivity—its water routing network—within each individual watershed. SnowModel and HydroFlow were then forced with observed meteorological data, and the overall trends and annual variability in air temperature and runoff were related to both variations and trends in the Atlantic multidecadal oscillation (AMO) index (e.g., Folland et al. 1986; Schlesinger and Ramankutty 1994; Kerr 2000; Chylek et al. 2009, 2010) to illustrate the impact from regional weather systems and the impacts from major episodic volcanic eruptions as part of an effort to understand the runoff response to natural forcings. Further, we examined whether the spatial runoff distribution from the warmest decade on record (2001–10) (Hansen et al. 2010) was different than the runoff distribution from both the average of 1960–69 and the average of 1960–2010.
(a) Greenland simulation domain with topography (500-m contour interval), the location of the coastal and GrIS meteorological tower stations (red dots; station specifications are provided in Table 1), and names of the surrounding seas and oceans. Greenland has been divided into three regions (east, west, and north) based on what oceans and seas watershed runoff flow into (see dashed lines and circles). The GrIS is marked with a color scale from gray to white (related to elevation), and the area outside the GrIS with black. (b) Simulated individual Greenland drainage basins (represented by multiple colors). Also, a specific region is illustrated from where examples of catchment runoff and hydrographs are illustrated (see bold square). (c) A closeup example of the individual drainage basins and flow network for the Helheim Glacier region, at the innermost part of the Sermilik Fjord, southeast Greenland, including the location for the runoff values and hydrographs illustrated in Fig. 5.
Citation: Journal of Climate 25, 17; 10.1175/JCLI-D-11-00592.1
2. Methodology and data
a. HydroFlow
Equations (1)–(3) describe a coupled system of equations whose solution yields a discharge hydrograph for each grid cell. These equations can be solved for any grid cell whose up-network inputs are known. Given knowledge of which grid cells flow into down-network grid cells, and first solving the grid cells at the head of a watershed (the grid cells that make up the watershed boundary) where there are no inflows and continuing to solve grid cells that are fed with cells that have already have a solution, the entire solution matrix can be solved at any given time step. As part of the flow network generation, only a single flow outlet into the ocean is allowed for each individual watershed. Also, conservation of mass principles between inflow, storage change, transit times, and outflow from each cell in the routing network must be defined to simulate the catchment runoff and generate discharge hydrographs for the routing grid cells. The residence time/flow velocity of a fluid element passing through the model grid cells depends on, for example, travel distance (e.g., grid cell size); surface slope and roughness (e.g., density of depression storage such as superglacial lakes, crevasses, and moulins); characteristics of the snow and ice matrix that the fluid is flowing through and over (e.g., temperature or cold content and porosity); temporal evolution of the snow and ice matrix; and changes in superglacial, englacial, and subglacial channel dimensions. Since the terrestrial snow distribution and associated characteristics vary in space and time, the transit time of each fluid element also have temporal and spatial evolutions.
Since the bedrock elevation is multiplied by a factor of 0.1 in Eq. (4), the role of bedrock topography on controlling the potentiometric surface and the associated flow direction is secondary. Identification of the flow network and subbasins by Lewis and Smith were therefore dominated by differences in surface topography unless elevation differences in the bedrock topography were sufficiently greater than the differences in surface topography. Therefore, defining the flow network using strictly surface topographic controls (based on 5-km grid cells), as done in the current application, appears to be acceptable owing to the smoothness of the surfaces.
A detailed description and schematic illustrations of HydroFlow, including flow paths and storage changes, can be found in Liston and Mernild (2012). The performance of HydroFlow was verified against runoff observations from a Mittivakkat Glacier test area, an area that included snow-free and snow-covered glacier surfaces, and snow-free and snow-covered land peripheral to the GrIS in southeast Greenland. HydroFlow successfully simulated flow conditions and spatial runoff distribution to the adjacent ocean, with simulated runoff variations, including peaks, reproducing observed runoff (r2 between 0.63 and 0.77) in both timing and volume (Liston and Mernild 2012).
b. SnowModel simulation setup and data analysis
HydroFlow requires temporally evolving, gridded inputs of rainfall and snowmelt and icemelt runoff, over the simulation domain. In this study, these contributions were provided by SnowModel (Liston and Elder 2006a,b; Mernild et al. 2006b), a spatially distributed modeling system that simulates meteorological conditions, surface energy, and moisture exchanges including snow and glacier melt, multilayer heat- and mass-transfer processes within the snow (e.g., snowpack temperature and density evolution), and surface runoff. Required SnowModel inputs include temporally varying fields of precipitation, wind speed and direction, air temperature, and relative humidity obtained from meteorological stations located within the simulation domain and spatially distributed, time-invariant fields of topography and land-cover type. Gridded meteorological forcings required by SnowModel were provided by MicroMet (Liston and Elder 2006b), a quasi-physically based high-resolution (e.g., 10-m to 10-km horizontal grid increment) meteorological distribution model. MicroMet is a data assimilation and interpolation model. The model uses known relationships between meteorological variables and the surrounding landscape (primarily topography) to distribute those variables over any given landscape in physically plausible and computationally efficient ways. At each time step, MicroMet calculates and distributes air temperature, relative humidity, wind speed, wind direction, incoming solar radiation, incoming longwave radiation, surface pressure, and precipitation and makes them accessible to SnowModel.
Previously, SnowModel and its various submodels, MicroMet (Liston and Elder 2006b), EnBal (Liston 1995; Liston et al. 1999), SnowTran-3D (Liston and Sturm 1998, 2002; Liston et al. 2007), and SnowPack-ML (Liston and Hall 1995; Liston and Mernild 2012), have been used successfully to simulate snow and ice accumulation and ablation processes throughout the Arctic, including Greenland (Mernild et al. 2006a; Liston et al. 2007; Mernild and Liston 2010; Liston and Hiemstra 2011). Specifically, for the GrIS, SnowModel has been tested and sufficient explained variances were found when comparing model output against independent in situ observations of meteorological variables (Mernild et al. 2008c, 2010), passive microwave-derived melt extent (Mernild et al. 2008c, 2009, 2011b), Moderate Resolution Imaging Spectroradiometer (MODIS) satellite-derived melt extent (Mernild et al. 2010), and runoff (Mernild et al. 2011a). Therefore, based on these studies, the combination of MicroMet and SnowModel generated gridcell runoffs are assumed to be of sufficient quality to drive the HydroFlow simulations presented herein.
SnowModel and HydroFlow simulations were performed for Greenland using a 5-km grid increment and daily time step over the period September 1959 through December 2010. Digital elevation model (DEM) data were provided by Bamber et al. (2001), and land-cover data were obtained from the U.S. Geological Survey (USGS) North American Land Cover Characteristics Database, version 2.0, updated with Landsat satellite-derived surface characteristics (Mernild et al. 2012a) (resolving glaciers having a glacier size greater than 30 m × 30 m). Glaciers and ice caps were classified as glacier cover in the land-cover file if the individual grid cells were covered by more than 50% of glacier ice. SnowModel was forced with observed atmospheric data from 56 meteorological stations located both in coastal areas and on the GrIS, shown in Fig. 1 and Table 1, (Mernild et al. 2011b). The number of meteorological stations varied from 10 (1960) to 45 (2006) per year between 1960 and 2010 (Table 1). The increase in number of meteorological stations over time likely increased the simulated regional variability. MicroMet (Liston and Elder 2006b) uses a Barnes objective analysis scheme that applies a Gaussian distance-dependent weighting function as part of its horizontal interpolations. In addition, elevation-related lapse rates are applied to the distributed temperature fields. For these reasons, the relatively small number of meteorological stations present during the early part of the simulations does not necessarily imply a degradation in simulated meteorological fields. We therefore assume the station increase does not significantly influence the trends and that the general trends that we produce and describe herein are valid. A detailed description of the model configuration and user-defined constants used in these Greenland simulations are available from Mernild et al. (2009, 2011b).
Meteorological input data for the Greenland runoff simulations. The stations were operated and data provided by Danish Meteorological Institute (DMI), University of Colorado at Boulder (CU), Geological Survey of Denmark and Greenland (GEUS), University of Copenhagen/Institute of Geography and Geology (UC/IGG), and University of Utrecht (UU). The abbreviations indicate Ta: air temperature, RH: relative humidity, Ws: wind speed, Wd: wind direction, and P: precipitation. Observed precipitation at the DMI meteorological station was corrected following Allerup et al. (1998, 2000). See Fig. 1 for station locations.
The spatial distribution of runoff was simulated for Greenland, including the runoff from each of the Greenland individual drainage basins and the routing of that runoff to the surrounding oceans. In addition, Greenland has been divided into eight 45° sectors, or regions, centered on the four cardinal and four ordinal directions (i.e., north, northeast, east, southeast, etc.) to illustrate the regional runoff distribution. The origin of these sectors is located at approximately the center of Greenland at 71.8914°N, 41.7181°W.
The simulated increase in Greenland runoff to the surrounding oceans was compared against increases in melt rates, melt area (melt extent values are illustrated in Fig. 3b of Mernild et al. 2011b), and melt duration or period (based on linear regression) to quantify the relative contributions of these three factors. In addition, calculated changes in Greenland runoff were compared with changes in Greenland runoff duration, also based on linear regression. Also, an additional analysis of the source of the Greenland runoff increase to the surrounding oceans over the period 1960–2010 was conducted. We applied a linear regression to the time series of annual total Greenland runoff, total GrIS runoff, maximum Greenland melt extent, Greenland melt duration, and Greenland average melt rate and calculated the change in those regressions (based on the linear regression slope) over the 50-yr period (based on decadal averages). These results were scaled so that they totaled 100% and are used to provide a relative measure of melt extent, melt duration, and melt rate in governing the changes in Greenland runoff.
3. Results and discussion
a. Individual Greenland drainage basins and flow network
The Greenland simulation domain, individual modeled drainage catchments, and an example of the simulated flow network for the Helheim Glacier region in southeast Greenland are illustrated in Fig. 1. HydroFlow divided the GrIS into ~400 individual drainage basins and all of Greenland (including the GrIS basins) into ~3150 individual basins (Fig. 1b). Each of these basins includes their own flow network that drains runoff to downslope areas and into the adjacent seas (Fig. 1c). For Greenland, the individual simulated drainage basins range in area from 50 to 154 800 km2 (averaging ~700 km2) (Fig. 1b) with 85% of the drainage basins equal to or less than 250 km2; these relatively small basins cover 10% of the total Greenland area and are mainly located in the land area between the GrIS and the oceans. A drainage-network study by Lewis and Smith (2009) identified 293 distinct GrIS hydrological basins ranging in area from 185 to 117 000 km2, values on a similar order of magnitude to those estimated by HydroFlow, even though Lewis and Smith omitted basins less than 100 km2.
The size and the shape of the HydroFlow simulated GrIS drainage basins (Fig. 1b) were compared with drainage basins estimated by Rignot and Kanagaratnam (2006). A comparison of the 20 greatest GrIS drainage basins was carried out only since the catchment division by Rignot and Kanagaratnam did not include midsize or minor catchments. Overall, HydroFlow reproduced the location of the watershed divides and the area of the greatest drainage basins reasonably well. When compared to Rignot and Kanagaratnam (2006), the sizes of the HydroFlow simulated drainage areas were generally within an error of less than 10%, and only one of the areas fell within an error of 30%.
b. Climate and GrIS surface mass balance
In Fig. 2a, the simulated mean annual air temperatures (MAAT) for Greenland are illustrated. Figure 2a also includes the mean summer temperatures for June–August (JJA); these are the temperatures largely associated with summer ablation, including the processes associated with evaporation, sublimation, and surface runoff. Figure 2b displays the AMO index series for 1960–2010 (http://www.esrl.noaa.gov/psd/data/timeseries/AMO/), and Fig. 2c presents the GrIS modeled net precipitation, surface mass balance, and runoff to adjacent seas. From 1960 to 2010 the Greenland mean summer air temperature and MAAT increased an average of 1.9° and 1.2°C (Fig. 2a), respectively. However, before the mid-1980s the trend in mean summer temperature correlates significantly with MAAT and was in antiphase, meaning JJA was on average increasing while MAAT was on average decreasing (r2 = 0.11 and p < 0.1, where r2 is the explained variance and p is level of significance), and hereafter the trends were in phase (significant) (r2 = 0.94, p < 0.01) and increasing (Fig. 2a). Since the mid-1980s, mean summer temperature and MAAT increased an average of 1.5° and 2.2°C, respectively. Hanna et al. (2008) found an increase in coastal Greenland summer temperatures for 1991–2006 of 1.8°C, based on observations. Furthermore, the overall variations in SnowModel simulated mean summer temperature explains the variance significantly (r2 = 0.65, p < 0.01), with the smoothed trends of the AMO index (Fig. 2b) [similar conditions between summer temperature and AMO variations was confirmed in Hanna et al. (2012)]. From 1960 to the mid-1970s, the smoothed AMO index decreased on average and thereafter it increased through 2010, corresponding with the trend in simulated mean summer temperature for Greenland. A study by Chylek et al. (2010) showed that Arctic temperatures were highly correlated with the AMO index, suggesting the Atlantic Ocean as a possible source of Arctic climate variability. This was also the case for the simulated Greenland MAAT for which the explained variance was significant for the periods after the mid-1980s (1986–2010: r2 = 0.95, p < 0.01) and before that time (1960–85: r2 = 0.18, p < 0.01); however, the latter period had a higher r2 value (explained more of the variance).
(a) Simulated mean summer (JJA) and mean annual air temperature (MAAT) Greenland anomaly time series for 1960–2010; (b) unsmoothed and smoothed (10-yr running average) Atlantic multidecadal oscillation (AMO) index; (c) GrIS simulated net precipitation P, surface mass balance (SMB = ΔS), and surface runoff R time series for 1960–2010; and (d) simulated surface GrIS runoff, land strip area (area outside the GrIS) runoff, and Greenland runoff time series for 1960–2010. The Agung (1963; Bali), El Chichón (1982; Mexico), and Mt. Pinatubo (1991; Philippines) volcanic eruptions are marked in (d).
Citation: Journal of Climate 25, 17; 10.1175/JCLI-D-11-00592.1
Figure 2c presents time series (1960–2010) of simulated GrIS surface hydrological conditions: net precipitation (precipitation minus evaporation and sublimation), surface runoff, and surface mass balance (SMB) on an annual basis for the calendar year (1 January–31 December). Mass gain (accumulation) is calculated as positive and mass loss (ablation) is considered negative for the GrIS. The average 1960–2010 simulated GrIS net precipitation was 489 ± 53 km3 yr−1 (here and below, the ± standard deviations are included), varying from 456 ± 46 km3 yr−1 in 1960–69 to 516 ± 38 km3 yr−1 in 2000–10 (Table 2). The simulated average GrIS net precipitation was just below the range of recently reported average net precipitation values: all reporting a similar average trend in precipitation as SnowModel toward higher annual values (Box et al. 2006; Hanna et al. 2005, 2008; Fettweis 2007; Fettweis et al. 2008; Ettema et al. 2009). Averaged for the GrIS, 85% of the SnowModel simulated precipitation fell as snow with the rest falling as rain.
Decadal mean and standard deviation of GrIS simulated net precipitation (Pnet), runoff (R), and surface mass balance (SMB) (change in storage, ΔS), internal GrIS storage, and Greenland runoff from 1960 through 2010. Also provided are the differences between mean Greenland runoff 2000–10 minus 1960–69 and 2000–10 minus 1960–2010.
On a decadal time scale, SnowModel simulated GrIS surface runoff varied an average of 261 ± 45 km3 yr−1 (0.7 ± 0.2 mm SLE yr−1) in 1970–79 to 429 ± 57 km3 yr−1 (1.2 ± 0.2 mm SLE yr−1) in 2000–10 (Table 2). Overall, for the period 1960–2010, GrIS simulated net precipitation and surface runoff rose significantly, averaging 1.5 (r2 = 0.19, p < 0.01) and 3.8 km3 yr−2 (r2 = 0.58, p < 0.01), respectively, leading to an enhanced significant SMB loss of 2.3 km3 yr−2 (r2 = 0.17, p < 0.01) (Table 2), average trends identical to previous studies by, for example, Box et al. (2006), Fettweis (2007), Hanna et al. (2008), and Ettema et al. (2009). These values closely follow air temperature fluctuations (Fig. 2a), indicating that surface mass loss increased as climate warmed with no suggestion of deceleration (Fig. 2c). The described trends for GrIS simulated net precipitation, runoff, and SMB are expected in a warmer climate (Fig. 2c and Table 2) due to enhanced snow accumulation in the relatively cold GrIS interior (where the increase in MAAT was still below freezing); enhanced melt/runoff season and surface melt extent (Fig. 3a); and enhanced ablation at lower elevations, including the GrIS margin areas.
(a) Mean annual simulated GrIS dry-snow line (dotted lines; the maximum average decadal boundary between melt and no melt on the glacier surface) (for definition, see Cuffey and Paterson 2010) on decadal intervals from 1960–69 through 2000–10. The percentages in brackets express the average annual melt extent on decadal scale for GrIS (time series of simulated mean melt extent 1960–2010 can be found in Mernild et al. 2011b). (b) Annual average simulated Greenland spatial surface runoff on decadal intervals for the decade with the lowest (1970–79) and highest (2000–10) runoff and mean (1960–2010). (c) The difference between the 2000–10 annual simulated Greenland runoff and the 1960–69 runoff and between the 2000–10 annual simulated runoff and the 1960–2010 mean.
Citation: Journal of Climate 25, 17; 10.1175/JCLI-D-11-00592.1
The increase in simulated surface runoff led to a cumulative GrIS runoff loss of 16 970 km3 (equal to 47.1 mm SLE) (1960–2010), with an annual average surface runoff of 333 ± 75 km3 yr−1 (0.9 ± 0.2 mm SLE yr−1) (Table 2), just above the range of recently reported average runoff values by, for example, Box et al. (2006), Fettweis (2007), Hanna et al. (2008) (updated), Ettema et al. (2009), and Mernild et al. (2010). The differences between these simulated average surface runoff values were likely primarily due to the different model representations of meltwater retention and refreezing within the snowpack for areas above ~2000 m above mean sea level (MSL). In the simulations presented herein, a multilayer snowpack model (Snowpack-ML) (Liston and Mernild 2012) was coupled to SnowModel and used to simulate the amount of percolation and internal refreezing (storage) from surface snowmelt and rain within the snow and firn layers, making an important contribution to the evolution of snow and ice densities, snow temperatures (cold content: temperatures below freezing), and moisture available for runoff (Liston and Mernild 2012). These simulations produced an internal refreezing value of 25% (129 ± 29 km3 yr−1) (Table 2), which is in the same range produced by the single-layer snowpack model used by Hanna et al. (2002, 2005, 2008) but below the value of 45% estimated by Ettema et al. (2009). If routines for retention and internal refreezing in the snowpack are not included in the simulations, runoff available for internal glacier drainage would be overestimated by approximately 25%–45% (e.g., Hanna et al. 2008; Ettema et al. 2009), depending on the simulation model and/or method.
For the GrIS, subtracting the simulated average surface runoff from the net precipitation yielded a surface mass gain, with an average annual GrIS SMB of 156 ± 82 km3 yr−1 (1960–2010) (Table 2), a mean value just below recent reported values by, for example, Box et al. (2006), Fettweis (2007), Hanna et al. (2008), and Ettema et al. (2009). The GrIS SMB decadal variability ranged from 220 ± 86 km3 yr−1 in 1970–79 to 86 ± 72 km3 yr−1 in 2000–10. The simulations showed the largest (most positive) SMB near the beginning of the simulation period, with a subsequent mass loss as temperatures and runoff increased.
For the GrIS, during 1960–2010 the accumulation zone covered an average of 90% of the total GrIS area, and the ablation zone an average of 10%. In contrast, the simulated area generating surface runoff covered an average of 12% and surface melt an average of 15% of the GrIS (Fig. 3a). A maximum SnowModel simulated ablation zone width of 125 km occurred in the southwest region of the GrIS and was almost as wide for the northeast GrIS. In contrast, the narrowest ablation zone had a maximum width of 10–20 km and occurred in both the northwest and the southeast regions of the GrIS, a distribution predominantly following elevation changes and the spatial variability in precipitation (data simulated in this study but not illustrated). Therefore, the widest ablation zones occurred in relatively low precipitation regions, and the narrowest zones occurred in the high precipitation areas. The spatial variability in simulated GrIS ablation zones were in general agreement with Ettema et al. (2009).
c. Greenland surface runoff
Figure 2d presents the time series of simulated Greenland runoff (1960–2010) and individual runoff contributions from the GrIS and from the land area—including thousands of glaciers and ice caps—located between the ice sheet and the surrounding oceans. The 1960–2010 average, simulated Greenland runoff was 481 ± 85 km3 yr−1 (1.3 ± 0.2 mm SLE yr−1), varying from 413 ± 56 km3 yr−1 (1.1 ± 0.2 mm SLE yr−1) in 1960–69 to 572 ± 53 km3 yr−1 (1.6 ± 0.2 mm SLE yr−1) in 2000–10, following the trends in air temperature and precipitation (Table 2 and Figs. 2a,b). The runoff simulations indicated that 69% of the runoff to the surrounding seas originated from the GrIS and 31% originated from the land area (Table 2; for division between the GrIS and the land area, see Fig. 1a). For the land area, the trend in simulated runoff was constant (Fig. 2d), and the average runoff was 148 ± 41 km3 yr−1 (Table 2). A possible reason for the minimal change in slope of the land-area runoff (0.1 km3 yr−1) in the Fig. 2d curve is because the glaciers and ice caps are already melting all summer, and an enhanced melt season and melt extent were therefore not possible. In contrast, simulated GrIS runoff, on average, has increased 3.9 km3 yr−1 since 1960 (Table 2), and there has been enhanced surface melt extent (Fettweis et al. 2011; Mernild et al. 2011b). Runoff values resolved previously for all of east Greenland indicated a 60% origin from the GrIS and 40% from the land area (Mernild et al. 2008b): these east Greenland runoff values are similar to the Greenland runoff values simulated in the current study.
In Fig. 2d, 1960–2010 simulated runoff time series from both the GrIS and all of Greenland are illustrated. The impact on runoff variability due to major episodic volcanic eruptions, such as Agung (1963), El Chichón (1982), and Mt. Pinatubo (1991) (Fig. 2d), and in the years immediately after do not appear to be systematic. For the year immediately after Agung and Pinatubo, simulated annual runoff values decreased, and they increased after El Chichón. The simulated Greenland runoff variations seems to be due to a combination of annual variations in both temperature and precipitation that are controlled by factors other than volcanic activity. Hanna et al. (2005) stated that global dust veils generated by volcanic activity might cool the polar regions and suppress ice sheet melt, but clearly there are other aspects of the climate system that may offset the volcanic signal. In contrast, the general climate forcing conditions captured by variations in the smooth AMO index time series (Fig. 2b) can be traced in the overall Greenland runoff pattern (Fig. 2d). In general, years with positive AMO index equaled years with relatively high Greenland simulated runoff volume (and relatively high mean summer temperatures), and years with negative AMO index had low runoff volume, with a significant explained variance (r2 = 0.73, p < 0.01) between the AMO index and Greenland runoff.
For Greenland, the spatial distribution of simulated surface runoff is illustrated at the decadal scale for the decade with the lowest (1970–79) and highest (2000–10) runoff and for 1960–2010 in Fig. 3b. Generally, relatively high average surface runoff values were simulated for the southwest and southeast regions of Greenland, and sporadic high values were simulated in the north region with maximum values of 4–6 m water equivalent (w.e.) yr−1. Elsewhere, runoff was less with lowest values in the northeast and northwest regions of less than 0.5 m w.e. yr−1 (Fig. 3b). This spatial simulated surface runoff distribution is largely in agreement with values from Lewis and Smith (2009). This regional pattern in surface runoff can be largely explained by the spatial distribution of precipitation since snowfall (end-of-winter accumulation) and surface runoff are negatively correlated through surface albedo, snow depth, and snow characteristics (e.g., snow cold content) (Hanna et al. 2008; Ettema et al. 2009; Mernild et al. 2009). For dry precipitation regions (west and northeast Greenland), the relatively low end-of-winter snow accumulation melts relatively fast during spring warmup. After the winter snow accumulation (albedo 0.50–0.80) has ablated, the ice surface albedo (0.40) promotes a stronger radiation-driven ablation and surface runoff, owing to the lower ice albedo. For wet precipitation regions (southeast and northwest Greenland) the relatively high end-of-winter snow accumulation, combined with frequent summer snow precipitation events, keeps the albedo high. Therefore, in wet regions it generally takes a longer time to melt the snowpack compared to dry regions before ablating the underlying glacier ice. For glaciers, ice caps, and the GrIS snowpack retention and refreezing processes suggest that regions with relatively high surface runoff are synchronous with relatively low end-of-winter snow accumulation because more meltwater was retained in the thicker snowpack, reducing runoff to the internal glacier drainage system (Hanna et al. 2008; Mernild et al. 2009).
In Fig. 3c, the simulated spatial surface runoff distribution for Greenland is illustrated for both 2000–10 minus 1960–69 and 2000–10 minus 1960–2010. For the GrIS the mean difference in surface runoff between 1960–69 and 2000–10 averaged 150 km3 yr−1 (an increase of 50%), associated with a simulated precipitation increase of 60 km3 yr−1 and a SMB loss of 90 km3 yr−1 (Table 2). Spatially, the difference in simulated GrIS runoff between 1960–69 and 2000–10 was as large as 0.8 m w.e. yr−1 near the ice sheet margin (Fig. 3c). For the land strip area, the differences in surface runoff were more diverse because of the distribution of local glaciers and ice caps. Generally, for the land area in the southwest region, surface runoff had a difference of up to −1.0 m w.e. yr−1 (between 1960–69 and 2000–10), while in the southeast region, for a given latitude, the runoff differences were more variable. Here the difference in surface runoff was less pronounced compared to the southwest areas with differences of only up to −0.6 m w.e. yr−1. However, differences in simulated surface runoff of up to 1.0 m w.e. yr−1 occurred, mainly from local glaciers and ice caps. In north Greenland, simulated surface runoff differed by up to 0.4 m w.e. yr−1 (Fig. 3c). For the land area, the difference in surface runoff during the last five decades was generally quite diverse with both positive and negative differences in runoff (Fig. 3c). These values were highly dependent on the location, distribution, elevation, and size of the local glaciers and ice caps compared to the more homogenous and positive runoff differences simulated for the GrIS.
The last decade (2000–10) has been the warmest decade on record (Hansen et al. 2010) with simulated MAAT and mean summer temperature 1.1° and 1.0°C above average, respectively, for the last five decades (Fig. 2a). For the GrIS, the 2000–10 simulated net precipitation was 27 km3 yr−1 and surface runoff was 97 km3 yr−1 (30%) above the 1960–2010 average, leading to a SMB of 70 km3 yr−1 below average (Table 2). For the GrIS, surface runoff values of up to 0.9 m w.e. yr−1 above average occurred (Fig. 3c). On a regional scale the 2000–10 minus the 1960–2010 runoff distribution generally resulted in a lower than average simulated surface runoff from nonglaciated land areas in the southeast and southwest regions. For glaciated areas, however, surface runoff was generally above this same average, with up to 1.0 m w.e. yr−1 in the southeast and southwest regions. For example, in northeast and north Greenland, surface runoff (2000–10) was as much as 0.7 m w.e. yr−1 greater than the 50-yr average (1960–2010) for the land strip area. A possible explanation for these relatively high, above average, simulated surface runoff values in northeast and north Greenland could be due to changes in the sea ice extent and thickness in the Arctic Ocean and Greenland Sea (Mernild et al. 2011b), and the influence of these changes are captured by the Greenland meteorological station network used to force the model simulations.
d. Spatial runoff distribution from Greenland to adjacent seas
In Fig. 1b, Greenland was divided into individual drainage basins. In addition, the western half (1 133 000 km2; 53% of the total area) and eastern half of Greenland (1 005 000 km2; 47%) were defined based on the main Greenland watershed divide running north to south. Regionally, the average Greenland 1960–2010 simulated runoff to the adjacent seas was greater in the western half of Greenland, 263 km3 yr−1 (equals 55% of the total Greenland runoff), than in the eastern half of Greenland, 218 km3 yr−1 (45%) (indicating an insignificant regional difference).
In Fig. 4, the spatial simulated runoff distribution from Greenland to the adjacent seas is illustrated (where each radial bar in the Greenland runoff figure represents the accumulated runoff contribution from 10 adjacent, individual drainage catchments). The individual drainage catchments route varying amounts of runoff to the surrounding seas. The 1960–2010 average simulated discharge for these drainage catchments varied from <0.01 to 10.1 km3 yr−1 (Fig. 4a). The spatial variability in catchment runoff to the surrounding seas also varied according to catchment size, ice sheet and glacier elevation range, and ice sheet and glacier areal coverage within each individual catchment. For approximately half of the simulated runoff values (colored radial bars) in Fig. 4a, runoff ranged from <0.01 to 1.0 km3 yr−1 (1960–2010) and contributed 15% of the total Greenland runoff. In contrast, 15% of the catchments (Fig. 4a)—catchments having a relatively large ice sheet and/or glacier areal coverage—had a mean annual runoff greater than 2.5 km3 yr−1 and contributed 40% of the Greenland runoff to the adjacent seas.
Spatial distribution of simulated runoff from Greenland’s individual drainage basins [each radial colored bar represents the accumulated runoff of 10 catchments located side by side (in total there are 316 radial bars); this was done to simplify the presentation of spatial runoff trends, since 85% of all catchments are equal or below 250 km2], and from the eight sectors (north, northeast, east, etc.), to the adjacent seas: (a) mean annual Greenland runoff for 1960–2010, where the numbers in brackets indicate the length of the discharge season (in days) for each region; (b) the difference between 2000–10 mean annual Greenland runoff and the 1960–69 runoff, where the numbers in brackets indicate the increase in the length (in days) of the discharge season for each region; and (c) the difference between 2000–10 mean annual Greenland runoff and the 1960–2010 mean. The regional runoff numbers for each sector have been used to scale radial distance of each gray wedge from the coast to the outside of the wedge and not from the center of Greenland to the outside of the wedge. So, for example, the 53 and 57 wedge in Fig. 4a ends are a similar distance from the coast but have very different total wedge sizes [and 49 and 45 are similar (Fig. 4a) because the coast is a similar distance from the center of the projection]. Greenland is slightly distorted from our traditional view in this radial projection.
Citation: Journal of Climate 25, 17; 10.1175/JCLI-D-11-00592.1
In Fig. 4 and in the following discussion, regionally, the average Greenland 1960–2010 simulated runoff to the adjacent seas was greatest from the south sector (88 km3 yr−1) and southwest sector (82 km3 yr−1) and lowest from the east sector (45 km3 yr−1) and southeast sector (49 km3 yr−1) (Fig. 4a). The regional distribution of runoff to the surrounding oceans appears to be in general agreement with the study by Lewis and Smith (2009). Also, Greenland has been divided into three regions based on which oceans and seas watershed runoff flows into (Fig. 1a): the average 1960–2010 simulated runoff to Fram Strait, the Nordic Seas, Denmark Strait, and the Atlantic Ocean (called the eastern oceans and seas of Greenland) was 191 km3 yr−1 (40%); to Smith Sound, Baffin Bay, Davis Strait, and the Labrador Sea (called the western seas and oceans of Greenland) was 237 km3 yr−1 (49%); and to the Wandel Sea and Arctic Ocean (called the northern seas and oceans of Greenland) 53 km3 yr−1 (11%).
In Figs. 4b and 4c, the differences in spatial catchment runoff from Greenland to the surrounding seas for both average 2000–10 minus average 1960–69 and average 2000–10 minus average 1960–2010 are illustrated. Over the last five decades, runoff increased by 142 km3 yr−1 (an increase of 30%) for Greenland (Table 2 and Fig. 4b), with an insignificant increase of 77 km3 yr−1 for the western half of Greenland and 65 km3 yr−1 for the eastern half of Greenland. The greatest change in runoff was in the south (26 km3 yr−1) and southwest sectors (24 km3 yr−1), and the least change was in the east sector (13 km3 yr−1). For the eastern oceans and seas of Greenland the simulated runoff increased by 63 km3 yr−1, for the western seas and oceans by 61 km3 yr−1, and for the northern seas and oceans by 18 km3 yr−1. On an individual catchment scale, runoff increased up to 3.0 km3 yr−1 (Fig. 4b) and was influenced by the catchment size, local climate variability (air temperature and precipitation), and fraction of glacier cover. Further, for the warmest decade on record, 2000–10 (Hansen et al. 2010), Greenland runoff was 91 km3 yr−1 above the 1960–2010 average (Table 2). This time period had the greatest change in the east (17 km3 yr−1) and southwest (16 km3 yr−1) sectors, the least change in the east Greenland sector (8 km3 yr−1), and at the individual catchment scale was up to 1.9 km3 yr−1 greater than average (Fig. 4c). Also, for the eastern oceans and seas of Greenland the runoff was 41 km3 yr−1 above the 1960–2010 average, for the western seas and oceans 39 km3 yr−1 above average, and for the northern seas and oceans 11 km3 yr−1 above.
The length of the simulated discharge season at regional scales (for the eight sectors) was highest in the southern sectors (averaging approximately 4–6 months) and lowest in the northern sectors (averaging approximately 2–3 months) (Fig. 4a). The increase in length of discharge season between 1960–69 and 2000–10 ranged from 11 days in the north sector to 27 days in the south and southwest sectors (Fig. 4b).
The simulated changes (or increase) in Greenland runoff to the surrounding oceans can, in general, be the result of three issues: 1) changes in GrIS melt rates (r2 = 0.03, p > 0.10), 2) changes in GrIS melt area (maximum extent area) (r2 = 0.41, p < 0.01), 3) and changes in GrIS melt duration or period (r2 = 0.21, p < 0.01). In addition, changes in Greenland runoff compared reasonably with changes in Greenland runoff duration (r2 = 0.68; p < 0.01), even though this is not directly related to the production of water like the other three, but it is related to the timing (influenced by, e.g., the snow cold content and snow depth) of how long it takes for the water to reach the ocean. The explained variance between Greenland runoff and runoff duration is relatively high compared to the explained variance between Greenland runoff and the climate-forcing impacts on GrIS melt conditions. This is likely because runoff at the outlet represents an integrated response of the upstream watershed to precipitation and other hydrometeorological processes like snow and glacier melt, to snow cold content, and to glaciohydrological processes like englacial bulk water storage and release, instead of just the climate impact on snow and ice conditions. In addition, this analysis suggests that increases in GrIS melt extent plays a relatively larger role in the simulated runoff increases than do the melt rate and melt duration changes.
As an additional analysis of the source of the Greenland runoff increase to the surrounding oceans over the period 1960–2010, we again assume that the increases are due, in general, to changes in 1) melt rates, 2) melt areas, and/or 3) melt duration. Because each of these is directly proportional to the total runoff (e.g., if the melt rates double, assuming all else is held constant, the runoff doubles), we assume that their contributions are proportional to their individual changes, under the constraint that their total contributions sum to equal the calculated runoff change. Applying linear regressions indicated that 103% of the Greenland runoff increase was due to increases in melt extent, 18% was due to increases in melt duration, and that there was a runoff reduction of 22% due to a decrease in melt rates. Note, that the runoff reduction due to the decreasing melt rates was more than compensated by the increase in melt extent and melt duration (103% + 18% − 22% = 100%). During this period (also based on decadal averages), the runoff from all of Greenland increased by 30% while, in contrast, GrIS runoff increased by 50% (Table 2).
Continuing with this relatively simple attribution analysis procedure, for the GrIS itself, 87% of the GrIS runoff increase was due to increases in melt extent, 18% was due to increases in melt duration, and a reduction of 5% occurred because of an decrease in melt rates (87% + 18% − 5% = 100%). For the land area surrounding the GrIS, the weak increase in runoff (almost horizontal trend in Fig. 2d) to the surrounding oceans over the period 1960–2010 was due to a 0% change in melt extent, with a 108% increase due to an increase in melt duration and a runoff reduction of 8% due to a decrease in melt rates (0% + 108% − 8% = 100%). In summary, the strong increase in GrIS runoff was largely due to increases in melt extent, while the relatively small increase in land area runoff was mainly due to changes in melt duration. This and the air temperature increases noted in Fig. 2a further suggests that the increase in discharge from Greenland to the surrounding oceans is primarily the result of increasing air temperatures that allow melt to occur over more area of the GrIS.
This can also be shown using absolute runoff and runoff extent values. To quantify the absolute contributions between increasing runoff and increasing runoff extent, specific runoff can be used (runoff volume per unit drainage area per time, L s−1 km−2; to convert to mm yr−1, multiply by 31.6). For the GrIS, melt occurring at higher and colder altitudes and latitudes means that the average specific runoff was decreasing, with absolute values of the specific runoff varying an average of 64 L s−1 km−2 in 1970–79 to 41 L s−1 km−2 in 2000–10. This indicates that the increase in runoff area extent has increased faster than the increasing runoff amount.
Here, the observed temperature increases have had a much larger contribution to increasing the available melt area than they do in increasing Greenland melt rates. The energy available to melt snow and ice in Greenland comes primarily from the incoming solar radiation component of the surface energy budget, and the air temperature distributions largely just define where and when the snow and ice melts (Marks and Dozier 1992; Liston and Hiemstra 2011).
e. A case example: Runoff distribution from Helheim glacier drainage basin
In Fig. 5a, the simulated MAAT and mean summer temperatures for June–August for the Helheim glacier drainage basin outlet in southeast Greenland are illustrated for 1960–2010 (see geographical location in Fig. 1c). Figure 5b displays the precipitation distribution time series for 1960–2010, Fig. 5c the modeled annual watershed runoff, Fig. 5d the daily average runoff on decadal intervals, and Fig. 5e the daily simulated runoff time series for 1960–2010. From 1960 to 2010 the Helheim mean summer air temperature and MAAT increased an average of 0.3° and 0.8°C, respectively, indicating less changes in temperature than the average trends for Greenland (Fig. 2a). As for Greenland in general, the trends in mean summer temperature and MAAT for Helheim before the mid-1980s were significant (r2 = 0.23, p < 0.01) and was, in general, in antiphase, meaning JJA was on average decreasing while MAAT was on average increasing, and hereafter the trends were in phase (significant) (r2 = 0.72 and p < 0.01) and increasing (Fig. 2a). Also for Helheim, the precipitation was almost in antiphase with the JJA temperature anomaly (Figs. 5a and 5b); during relatively dry years at Helheim (precipitation below average); for example, before 1970 and after 1995, the temperature was generally above average and between these years the temperature was generally below average (Figs. 5a,b): a similar pattern for Sermilik Fjord from 1900 to 2008 is illustrated in Mernild et al. (2012b). The variability between temperature and precipitation may be explained by deep Icelandic lows associated with stronger westerly winds bringing dry and relatively warm air masses to southeast Greenland during positive NAO scenarios (Bromwich et al. 1999; Hurrell and Deser 2009)—a pattern linked with a positive AMO index (Fig. 2b).
(a) An example of Helheim Glacier catchment outlet mean summer (JJA) and mean annual air temperature (MAAT) anomaly time series for 1960–2010; (b) corrected annual precipitation anomaly 1960–2010; (c) simulated annual runoff anomaly 1960–2010; (d) daily average simulated Helheim Glacier catchment runoff (5-day running mean) on decadal intervals from 1960–69 through 2000–10; and (e) daily simulated Helheim Glacier catchment runoff from 1960 through 2010. The geographical location is illustrated in Fig. 1c.
Citation: Journal of Climate 25, 17; 10.1175/JCLI-D-11-00592.1
These runoff variations are due to a combined effect of climate impacts from temperature (snow and ice melt) and precipitation (rain and snow) (Fig. 5c). Here, simulated runoff is illustrated in Figs. 5c–e. Clearly, the amount of runoff changed over time, together with the date of breakup and the runoff period. The displayed runoff (Figs. 5c,e) indicates definite seasonal, interannual, and decadal cycles. For example, it appears as though, in the illustrated 10-yr running average (Fig. 5c), the variability in runoff before 1973 (r2 = 0.85, p < 0.01) and after 1986 (r2 = 0.92, p < 0.01) were dominated by changes in MAAT followed atmospheric warming (Fig. 5a), and between 1973 and 1986 by the changes in precipitation (r2 = 0.71, p < 0.01) (Fig. 5b). In addition, Fig. 5d shows an approximately 4-week shift in length of discharge season as the decades progress. The daily discharge hydrographs for the 50-yr simulation period (Fig. 5e) highlight the detailed interannual variations in runoff simulated as part of these MicroMet/SnowModel/HydroFlow model simulations.
4. Conclusions
We have investigated the impact of changes in Greenland weather and climate conditions on surface hydrological processes and runoff for the 50-yr period 1960–2010. This included quantifying the spatial distribution and trends of meltwater and rainwater discharge into the adjacent seas from both the GrIS and the land, ice cap, and glacier areas between the GrIS and surrounding oceans. The merging of observed atmospheric forcing datasets with SnowModel—a snow and ice evolution system—and HydroFlow—a runoff routing system—allowed a detailed (5-km, daily) analysis and mapping of spatial variations in Greenland discharge to the adjacent seas and provided insights into the regional distribution of runoff features and quantities. Individual drainage catchments (~3150), each with an individual flow network, were estimated for Greenland before simulating runoff to downflow areas and the surrounding oceans. Given the severe dearth of Greenland discharge observations, runoff simulations are crucial for understanding Greenland spatial and temporal runoff variations; this runoff explains half of the recent mass loss of the GrIS (van den Broeke 2009).
Overall, Greenland has warmed and the runoff has increased during the last 50 years with the greatest runoff increase occurring in southwest Greenland and lower runoff increases occurring in northeast Greenland. The spatial runoff distributions show greater hydrological activity in southwest Greenland and lowest for the northeast Greenland region, supporting the hypothesis that discharges into the adjacent seas are greatest in regions where snowfall (end-of-winter snow accumulation) is generally low and discharge is least in regions where snowfall is high. These processes and relationships are crucial for understanding the spatial distribution of runoff to the surrounding oceans, and the linkages among a changing climate and the associated changes in runoff magnitudes and distributions.
The Greenland simulations showed distinct regional-scale runoff variability throughout the simulation domain. A similar uneven runoff pattern is expected to be present throughout the pan-Arctic because, like Greenland, the pan-Arctic landscape has numerous individual mountain glaciers and ice caps dominating the local to regional freshwater budget, including the spatial distribution of runoff to the oceans and the contribution to present rates of global sea level rise. Gardner et al. (2011) confirmed that mountain glaciers and ice caps are contributing significantly to present rates of sea level rise and will continue to do so during the next century and beyond. In addition, the Canadian Arctic Archipelago could potentially play a key role in future sea level changes. It is located off the northwestern shore of Greenland and contains one-third of the global volume of land-based ice outside the ice sheets (Radić and Hock 2010). Its contribution to sea level rise remains largely unknown.
In addition to this Greenland application, SnowModel and HydroFlow could be used to address water resource issues associated with irrigation and drinking water in regions where snow and glacier melt are key components of the hydrologic cycle, such as the Andes and the Tibetan Plateau and its surrounding mountain ranges. The state of the Tibetan Plateau region’s hydrosphere and cryosphere affects a considerable portion of the world’s population, principally the inhabitants of South and East Asia (Joswiak et al. 2011); the future economic and social health and development of these populations depends critically upon snow- and ice-related hydrologic resources, especially runoff. Other potential applications of these tools include using them to estimate water supply potential for hydropower production in snow-covered and glaciated areas like Greenland and throughout the pan-Arctic and using them to quantify present and future glacier mass balance values from Earth’s estimated 300 000–400 000 glaciers and ice caps (Dyurgerov and Meier 2005).
In the future, under an expected warmer climate regime, the GrIS margin, including its outlet glaciers, will retreat farther inland. Under these conditions, surface runoff from the ice sheet, local glaciers, and ice caps will likely dominate calving as the primary Greenland mass loss mechanism. This makes it crucial to understand future changes in runoff, the spatial distribution of that runoff to the seas surrounding Greenland, and the associated contribution to global sea level rise since glaciers will continue to melt and contribute to sea level rise throughout the next century and beyond (e.g., Gardner et al. 2011).
Understanding the linkages among Greenland local meltwater and runoff, freshwater discharge into the surrounding oceans, ocean circulation features such as the AMOC, and other components of the climate system will provide critical insights into the workings of the Greenland climate system and its interactions with the rest of Earth’s atmosphere, hydrosphere, and cryosphere. Historically, the representation of Greenland freshwater discharge into the oceans has been either nonexistent or unrealistically simplistic. For example, ocean models have traditionally placed the freshwater runoff flux directly into the midocean areas (Weijer et al. 2012) rather than accurately accounting for the spatial and temporal distributions of actual Greenland runoff. With their ability to transport freshwater from the interior and margins of Greenland to the surrounding seas, models such as SnowModel and HydroFlow are now available to fill this gap in climate-system representation. As a consequence, these tools are expected to make valuable contributions toward answering scientific questions related to future Greenland spatial runoff distributions, freshwater discharge into the surrounding oceans, contributions to sea level rise, and other climate-related impacts associated with changes in climate features such as ocean salinity and density stratification, ocean temperature, sea ice distributions, ocean circulation patterns such as the AMOC, and the attendant feedbacks to the atmosphere and the rest of the climate system.
Acknowledgments
We extend a very special thanks to the two anonymous reviewers for their insightful critique of this article. This work was supported by the Climate Change Prediction Program and Scientific Discovery for Advanced Computing (SciDAC) program within the U.S. Department of Energy Office of Science, Los Alamos National Laboratory (LANL) Director’s Fellowship, LANL Institute for Geophysics and Planetary Physics, and NASA Grant NNX08AV21G. LANL is operated under the auspices of the National Nuclear Security Administration of the U.S. Department of Energy under Contract DE-AC52-06NA25396. Thanks are given to the Program for Monitoring of the Greenland Ice Sheet (PROMICE), Geological Survey of Denmark and Greenland; the Danish Meteorological Institute; the University of Utrecht; the Greenland Climate Network, University of Colorado at Boulder; and the Department of Geography and Geology, University of Copenhagen for providing meteorological station observations.
REFERENCES
Allerup, P., H. Madsen, and F. Vejen, 1998: Estimating true precipitation in arctic areas. Nordic Hydrological Programme Rep. 44, 9 pp.
Allerup, P., H. Madsen, and F. Vejen, 2000: Correction of precipitation based on off-site weather information. Atmos. Res., 53, 231–250.
Bamber, J. L., S. Ekholm, and W. B. Krabil, 2001: A new, high-resolution digital elevation model of Greenland fully validated with airborne altimetry data. J. Geophys. Res., 106, 6733–6745.
Box, J. E., and Coauthors, 2006: Greenland ice sheet surface mass balance variability (1988–2004) from calibrated Polar MM5 output. J. Climate, 19, 2783–2800.
Bromwich, D. H., Q.-S. Chen, Y. Li, and R. L. Cullather, 1999: Precipitation over Greenland and its relation to the North Atlantic Oscillation. J. Geophys. Res., 104, 22 103–22 115.
Bryan, F. O., 1986: High-latitude salinity effects and interhemispheric thermohaline circulations. Nature, 323, 301–304.
Bryden, H. L., H. R. Longworth, and S. A. Cunningham, 2005: Slowing of the Atlantic meridional overturning circulation at 25°N. Nature, 438, 655–657, doi:10.1038/nature04385.
Chylek, P., C. K. Folland, G. Lesins, M. K. Dubey, and M. Wang, 2009: Arctic air temperature change amplification and the Atlantic multidecadal oscillation. Geophys. Res. Lett., 36, L14801, doi:10.1029/2009GL038777.
Chylek, P., C. K. Folland, G. Lesins, and M. K. Dubey, 2010: Twentieth century bipolar seesaw of the Arctic and Antarctic surface air temperatures. Geophys. Res. Lett., 37, L08703, doi:10.1029/2010GL042793.
Cuffey, K. M., and W. S. B. Paterson, 2010: The Physics of Glaciers. 4th ed. Elsevier, 707 pp.
Dyurgerov, M. B., and M. F. Meier, 2005: Glaciers and the changing earth system: A 2004 snapshot. University of Colorado at Boulder INSTAAR Occasional Paper 58, 116 pp.
Ettema, J., M. R. van den Broeke, E. van Meijgaard, W. J. van den Berg, J. L. Bamber, J. E. Box, and R. C. Bales, 2009: Higher surface mass balance of the Greenland ice sheet revealed by high-resolution climate modeling. Geophys. Res. Lett., 36, L12501, doi:10.1029/2009GL038110.
Fettweis, X., 2007: Reconstruction of the 1979–2006 Greenland ice sheet surface mass balance using the regional climate model MAR. Cryosphere, 1, 21–40.
Fettweis, X., E. Hanna, H. Gallee, P. Huybrechts, and M. Erpicum, 2008: Estimation of the Greenland ice sheet surface mass balance during 20th and 21st centuries. Cryosphere, 2, 117–129.
Fettweis, X., M. Tedesco, M. R. van den Broeke, and J. Ettema, 2011: Melting trends over the Greenland ice sheet (1958–2009) from spaceborne microwave data and regional climate models. Cryosphere, 5, 359–375.
Folland, C. K., T. Palmer, and D. E. Parker, 1986: Sahel rainfall and worldwide sea temperatures. Nature, 320, 602–607.
Gardner, A. S., and Coauthors, 2011: Sharply increased mass loss from glaciers and ice caps in the Canadian Arctic Archipelago. Nature, 473, 357–360, doi:10.1038/nature10089.
Hanna, E., P. Huybrechts, and T. Mote, 2002: Surface mass balance of the Greenland ice sheet from climate-analysis data and accumulation/runoff models. Ann. Glaciol., 35, 67–72.
Hanna, E., P. Huybrechts, I. Janssens, J. Cappelen, K. Steffen, and A. Stephens, 2005: Runoff and mass balance of the Greenland ice sheet: 1958–2003. J. Geophys. Res., 110, D13108, doi:10.1029/2004JD005641.
Hanna, E., and Coauthors, 2008: Increased runoff from melt from the Greenland ice sheet: A response to global warming. J. Climate, 21, 331–341.
Hanna, E., J. M. Jones, J. Cappelen, S. H. Mernild, L. Wood, K. Steffen, and P. Huybrechts, 2012: The influence of North Atlantic atmospheric and oceanic forcing effects on 1900–2010 Greenland summer climate and ice melt/runoff. Int. J. Climatol., doi:10.1002/joc.3475, in press.
Hansen, J., R. Ruedy, M. Sato, and K. Lo, 2010: Global surface temperature change. Rev. Geophys., 48, RG4004, doi:10.1029/2010RG000345.
Hasholt, B., and Coauthors, 2008: Hydrology and transport of sediment and solutes at Zackenberg. High-Arctic Ecosystem Dynamics in a Changing Climate: Ten Years of Monitoring and Research at Zackenberg Research Station, Northeast Greenland, H. Meltofte et al., Eds., Advances in Ecological Research, Vol. 40, Academic Press, 111–149.
Hasholt, B., A. B. Mikkelsen, M. H. Nielsen, and M. A. D. Larsen, 2012: Observations of runoff and sediment and dissolved loads from the Greenland ice sheet at Kangerlussuaq, west Greenland, 2007 to 2010. Z. Geomorphol., in press.
Hinzman, L. D., and Coauthors, 2005: Evidence and implications of recent climate change in northern Alaska and other Arctic regions. Climatic Change, 72, 251–298.
Hurrell, J. W., and C. Deser, 2009: North Atlantic climate variability: The role of the Northern Atlantic Oscillation. J. Mar. Syst., 78, 28–41.
Jensen, L. M., and M. Rasch, 2010: Zackenberg Ecological Research operations, 15th annual report, 2009. Aarhus University National Environmental Research Institute Rep., 134 pp.
Joswiak, M., T. Yao, and D. Joswiak, 2011: Moving forward on glacier retreat. Eos, Trans. Amer. Geophys. Union, 92, 188, doi:10.1029/2011EO220006.
Kerr, R. A., 2000: A North Atlantic climate pacemaker for the centuries. Science, 288, 1984–1985.
Lewis, S. M., and L. C. Smith, 2009: Hydrological drainage of the Greenland ice sheet. Hydrol. Processes, 23, 2004–2011, doi:10.1002/hyp.7343.
Liston, G. E., 1995: Local advection of momentum, heat, and moisture during the melt of patchy snow covers. J. Appl. Meteor., 34, 1705–1715.
Liston, G. E., and D. K. Hall, 1995: An energy-balance model of lake-ice evolution. J. Glaciol., 41, 373–382.
Liston, G. E., and M. Sturm, 1998: A snow-transport model for complex terrain. J. Glaciol., 44, 498–516.
Liston, G. E., and M. Sturm, 2002: Winter precipitation patterns in arctic Alaska determined from a blowing-snow model and snow-depth observations. J. Hydrometeor., 3, 646–659.
Liston, G. E., and K. Elder, 2006a: A distributed snow-evolution modeling system (SnowModel). J. Hydrometeor., 7, 1259–1276.
Liston, G. E., and K. Elder, 2006b: A meteorological distribution system for high-resolution terrestrial modeling (MicroMet). J. Hydrometeor., 7, 217–234.
Liston, G. E., and C. A. Hiemstra, 2011: The changing cryosphere: Pan-Arctic snow trends (1979–2009). J. Climate, 24, 5691–5712.
Liston, G. E., and S. H. Mernild, 2012: Greenland freshwater runoff. Part I: A runoff routing model for glaciated and nonglaciated landscapes (HydroFlow). J. Climate, 25, 5997–6014.
Liston, G. E., J.-G. Winther, O. Bruland, H. Elvehøy, and K. Sand, 1999: Below surface ice melt on the coastal Antarctic ice sheet. J. Glaciol., 45, 273–285.
Liston, G. E., R. B. Haehnel, M. Sturm, C. A. Hiemstra, S. Berezovskaya, and R. D. Tabler, 2007: Simulating complex snow distributions in windy environments using SnowTran-3D. J. Glaciol., 53, 241–256.
Marks, D., and J. Dozier, 1992: Climate and energy exchange at the snow surface in the alpine region of the Sierra Nevada, 2, Snow cover energy balance. Water Resour. Res., 28, 3043–3054.
Mernild, S. H., and B. Hasholt, 2006: Climatic control on river discharge simulations, Mittivakkat Glacier catchment, Ammassalik Island, SE Greenland. Nord. Hydrol., 37 (4–5), 327–346.
Mernild, S. H., and G. E. Liston, 2010: The influence of air temperature inversions on snowmelt and glacier mass balance simulations, Ammassalik Island, southeast Greenland. J. Appl. Meteor. Climatol., 49, 47–67.
Mernild, S. H., B. Hasholt, and G. E. Liston, 2006a: Water flow through Mittivakkat Glacier, Ammassalik Island, SE Greenland. Danish J. Geogr., 106, 25–43.
Mernild, S. H., G. E. Liston, B. Hasholt, and N. T. Knudsen, 2006b: Snow-distribution and melt modeling for Mittivakkat Glacier, Ammassalik Island, southeast Greenland. J. Hydrometeor., 7, 808–824.
Mernild, S. H., B. Hasholt, and G. E. Liston, 2008a: Climatic control on river discharge simulations, Zackenberg River drainage basin, northeast Greenland. Hydrol. Processes, 22, 1932–1948, doi:10.1002/hyp.6777.
Mernild, S. H., G. E. Liston, and B. Hasholt, 2008b: East Greenland freshwater runoff to the Greenland-Iceland-Norwegian Seas 1999–2004 and 2071–2100. Hydrol. Processes, 22, 4571–4586, doi:10.1002/hyp.7061.
Mernild, S. H., G. E. Liston, C. A. Hiemstra, and K. Steffen, 2008c: Surface melt area and water balance modeling on the Greenland ice sheet 1995–2005. J. Hydrometeor., 9, 1191–1211.
Mernild, S. H., G. E. Liston, C. A. Hiemstra, K. Steffen, E. Hanna, and J. H. Christensen, 2009: Greenland ice sheet surface mass-balance modeling and freshwater flux for 2007, and in a 1995–2007 perspective. Hydrol. Processes, 23, 2470–2484, doi:10.1002/hyp.7354.
Mernild, S. H., G. E. Liston, C. A. Hiemstra, and J. H. Christensen, 2010: Greenland ice sheet surface mass-balance modeling in a 131-yr perspective, 1950–2080. J. Hydrometeor., 11, 3–25.
Mernild, S. H., G. E. Liston, C. A. Hiemstra, J. H. Christensen, M. Stendel, and B. Hasholt, 2011a: Surface mass-balance and runoff modeling using HIRHAM4 RCM at Kangerlussuaq (Søndre Strømfjord), west Greenland, 1950–2080. J. Climate, 24, 609–623.
Mernild, S. H., T. Mote, and G. E. Liston, 2011b: Greenland ice sheet surface melt extent and trends, 1960–2010. J. Glaciol., 57, 621–628.
Mernild, S. H., J. K. Malmros, N. T. Knudsen, and J. C. Yde, 2012a: Multi-decadal marine- and land-terminating glacier retreat in the Ammassalik region, southeast Greenland. Cyrosphere, 6, 625–639.
Mernild, S. H., M.-S. Seidenkrantz, P. Chylek, G. E. Liston, and B. Hasholt, 2012b: Climate-driven fluctuations in freshwater to Sermilik Fjord, east Greenland, during the last 4000 years. Holocene, 22, 155–164, doi:10.1177/0959683611431215.
Pfeffer, W. T., J. T. Harper, and S. O’Neel, 2008: Kinematic constrains on glacier contribution of 21st-century sea-level rise. Science, 321, 1340–1343.
Radić, V., and R. Hock, 2010: Regional and global volumes of glaciers derived from statistical upscaling of glacier inventory data. J. Geophys. Res., 115, F01010, doi:10.1029/2009JF001373.
Rahmstorf, S., 1995: Bifurcations of the Atlantic thermohaline circulation in response to changes in the hydrological cycle. Nature, 378, 145–149.
Rahmstorf, S., and Coauthors, 2005: Thermohaline circulation hysteresis: A model intercomparison. Geophys. Res. Lett., 32, L23605, doi:10.1029/2005GL023655.
Rennermalm, A. K., L. C. Smith, V. W. Chu, R. R. Forster, J. E. Box, and B. Hagedorn, 2012: Proglacial river stage, discharge, and temperature datasets from the Akuliarusiarsuup Kuua River northern tributary, southwest Greenland, 2008–2011. Earth Syst. Sci. Data, 4, 1–12, doi:10.5194/essd-4-1-2012.
Rignot, E., and P. Kanagaratnam, 2006: Changes in the velocity structure of the Greenland ice sheet. Science, 311, 986–990.
Schlesinger, M. E., and N. Ramankutty, 1994: An oscillation in the global climate system of period 65–70 years. Nature, 367, 723.
Serreze, M. C., and Coauthors, 2000: Observational evidence of recent change in the northern high-latitude environment. Climatic Change, 46, 159–207.
Solomon, S., D. Qin, M. Manning, M. Marquis, K. Averyt, M. M. B. Tignor, H. L. Miller Jr., and Z. Chen, Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.
van den Broeke, M. R., and Coauthors, 2009: Partitioning recent Greenland mass loss. Science, 326, 984–986.
Weijer, W., M. E. Maltrud, M. W. Hecht, H. A. Dijkstra, and M. A. Kliphuis, 2012: Response of the Atlantic Ocean circulation to Greenland Ice Sheet melting in a strongly-eddying ocean model. Geophys. Res. Lett., 39, L09606, doi:10.1029/2012GL051611.
