Freshwater flux to Sermilik Fjord , SE Greenland

Introduction Conclusions References


Introduction
Global atmospheric temperatures showed a warming trend since the 1970s, and all years during the present century (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) have been among the warmest since the beginning of instrumental records (Allison et al., 2009).Surface air temperature observations reveal the strongest increase occurring over Northern Hemisphere land Figures There is clear evidence of increased melting of the GrIS and marginal glaciers in Greenland since the mid-1990s (e.g., Mote, 2007;Tedesco, 2007;Knudsen and Hasholt, 2008;Steffen et al., 2008), and rapid mass loss has been observed and simulated (e.g., Hanna et al., 2008;Mernild et al., 2008a;Allison et al., 2009;Khan et al., 2010).In a warming climate, we would expect an accelerating freshwater flux: ice discharge (calving) and runoff to the ocean, manifested by, e.g., decreasing ocean salinity, and increasing global eustatic sea level rise (e.g., ACIA, 2005;Box et al., 2006;IPCC, 2007).A few freshwater runoff measurements are available for Greenland from 1990s at the Sermilik Research Station, Sermilik Fjord (65 W Greenland, since 2008 (e.g., Mernild andHasholt, 2006, 2009;Jensen and Rasch, 2009).These data series are important tools for assessing and quantifying the impact of climate change and variability on freshwater runoff from glaciated landscapes such as Greenland.
The first documentation of glaciers in the Sermilik Fjord catchment basin was carried out in 1933, and in 1970 the Sermilik Research Station was established close to the Mittivakkat Glacier to study the control of climate on a low-arctic (Born and B öcher, 2001), partly glaciated landscape.An automated terrestrial monitoring program was initiated for the Mittivakkat Glacier catchment in 1993, which presents today the longest continuous monitoring program in E Greenland.Data on observed climate conditions have been presented by Mernild et al. (2008b) and Jakobsen et al. (2008).Seasonal and annual observations on the Mittivakkat Glacier include: winter, summer, and net mass-balance (Knudsen andHasholt, 2004, 2008), freshwater runoff (e.g., Hasholt, 1980;Hasholt andMernild, 2004, 2008), solute export (Yde et al., 2010), and sediment transport (Hasholt and Walling, 1992;Busskamp and Hasholt, 1996;Hasholt and Mernild, 2008).Modeling studies for this region include seasonal and annual climate processes (Mernild and Liston, 2010), snow cover distribution (Hasholt et al., Introduction Conclusions References Tables Figures

Back Close
Full 2003; Mernild et al., 2006Mernild et al., , 2008a)), glacier surface mass-balance (Mernild et al., 2006(Mernild et al., , 2008a)), and runoff (Mernild and Hasholt, 2006;Mernild et al., 2008a).This collection of extensive observations and model results from the Mittivakkat Glacier catchment was used to simulate the terrestrial surface runoff for the Sermilik Fjord.Not only runoff but also ice discharge from e.g., the Helheim Glacier (one of the most conspicuous calving outlet glaciers from the GrIS) at the innermost part of the fjord, seems to be an important source of freshwater for both, the Sermilik Fjord and the Irminger Sea.We present a 10-year time series (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) of freshwater flux to the Sermilik Fjord in order to assess variability and trend thereof due to changes in air temperature, net precipitation (hereafter referred to as precipitation), and ice dynamics.In particular, we address the simulated temporal and spatial distribution of terrestrial surface freshwater runoff to the fjord and also on a sub-catchment scale.The runoff was simulated in SnowModel (Liston and Elder, 2006a;Mernild et al., 2006), based on in situ meteorological data within the Sermilik Fjord area.Runoff was initially simulated for the Mittivakkat Glacier catchment area of ∼18 km 2 and tested against observed runoff data from the Mittivakkat Glacier catchment outlet which is the only place in the Sermilik Fjord where runoff is observed.The simulated runoff was verified (bias corrected) against runoff observations, before runoff simulations were scaled up to the entire Sermilik Fjord catchment area.The following objectives are addressed: (1) assess the performance of SnowModel by comparing simulated runoff against observed runoff for the Mittivakkat Glacier catchment; (2) simulate the spatial runoff variability and quantify whether the annual freshwater runoff to the Sermilik Fjord has been increasing throughout the simulation period; (3) compare simulated runoff with observed Helheim ice discharge to illustrate the respective distribution from each freshwater source; and (4) merge different sources of freshwater input, e.g., simulated runoff, simulated precipitation over the fjord with satellite-derived ice discharge and geothermal and frictional melting due to basal ice motion to quantify the freshwater flux to the Sermilik Fjord.Introduction

Conclusions References
Tables Figures

Back Close
Full

Physical settings and climate
The Sermilik Fjord catchment (58 045 km 2 ) is located on the east coast of Greenland (65 • N; 37 • W), connected to the Irminger Sea (Fig. 1a).The fjord is 1103 km 2 in area, 85 km in length, and the largest fjord system in SE Greenland.The catchment drains a part of the GrIS, including the Helheim, Fenris, and Midg ård Glaciers (the three major outlet glaciers in Sermilik Fjord catchment), and marginal glaciers, among these the Mittivakkat Glacier on Ammassalik Island (see Figs. 1c and 3a for location), where long-term monitoring of climate, mass-balance, and runoff was observed (Mernild and Hasholt, 2006;Knudsen and Hasholt, 2008).The Sermilik Fjord catchment ranges in elevation from sea level to ∼2900 m a.s.l.The lower parts of the terrain (elevation below 700-1000 m a.s.l.) are dominated by exposed bedrock, sporadic thin soil layers, and sparse vegetation.Landscapes above 700-1000 m a.s.l. are mostly covered by glaciers and the GrIS (Fig. 1c).For the purposes of this study, the Sermilik Fjord catchment has been divided into 7 sub-catchments, each draining into specific parts of the fjord.These areas also represent characteristic variations in glacier ice coverage from approximately 10% (area 2) to 87% (area 4) (Figs.1c and 4f).
Our simulated precipitation values were in the same order of magnitude compared to values from Ettema et al. (2009).Mean annual catchment relative humidity and wind speed were 83% and 4.1 m s −1 , respectively.Sermilik Fjord is dominated by seasonal Introduction

Conclusions References
Tables Figures

Back Close
Full and local-to-regional variations in climate.During summer, the low lying coastal areas, on, e.g., Ammassalik Island (approximately below 300 m a.s.l.), are influenced by air temperature inversions which are common in Arctic coastal landscapes due to the effect of thermal differences between land and ocean (e.g., Kozo, 1982;Weick and Rouse, 1991;Mernild and Liston, 2010).The climate and its seasonal variability are illustrated in Fig. 2b using positive summer air temperature lapse rates in the near coastal areas.Apart from this temperature inversion in the lower lying near coastal areas during summer, meteorological data from all seven stations in the Sermilik Fjord catchment (Fig. 1) showed constantly negative mean monthly lapse rates, very similar to the high-elevation GrIS air temperature lapse rates (Fig. 2a) (e.g., Steffen and Box, 2001;Mernild et al., 2009).
3 Model description and satelite data

SnowModel and model simulations
SnowModel (Liston and Elder, 2006a), is a spatially-distributed snow-evolution, ice melt, and runoff modeling system useful in all landscapes, climates, and conditions where snow and ice play an important role in hydrological cycling (Mernild et al., 2006;Mernild and Liston, 2010).For a detailed description of SnowModel, including its subprograms: MicroMet, EnBal, SnowPack, SnowTran-3D, and SnowAssim see Liston and Elder (2006a,b), Liston et al. (2008), Liston and Hiemstra (2008), and Mernild and Liston (2010).SnowModel is a surface model simulating first-order effects of atmospheric forcing on snow, glacier ice, and runoff, but processes related to glacier dynamics are not included.

Input data, model verification, and uncertainties
Meteorological data of air temperature, relative humidity, wind speed, wind direction, and precipitation were obtained from seven meteorological stations at different elevations within the simulation domain (Fig. 1b).Four stations were located on the GrIS, Introduction

Conclusions References
Tables Figures
The Greenland topographic data at 625-m resolution from Bamber et al. (2001) was used with the image-derived correction published by Scambos and Haran (2002), and interpolated to a 500-m grid increment covering a 400.5 by 300.5 km simulation domain for the Sermilik Fjord catchment (Fig. 1a).The location of the Sermilik Fjord coast line, GrIS terminus, and marginal glaciers were estimated by using Moderate Resolution Imaging Spectroradiometer (MODIS) satellite images (observed on 30 August 2009).User-defined constants for SnowModel are shown in Mernild et al. (2009), and parameter definitions are given in Liston and Sturm (1998).
SnowModel simulated runoff was tested and verified (bias corrected) against observed runoff from the outlet of the Mittivakkat Glacier catchment for the period 1999-2005 (Fig. 2).The cumulative simulated runoff was initially underestimated by 34-43%, averaging 38% according to runoff observations.Therefore, a linear regression (r 2 =0.95;where r 2 is the explained variance) was used for verification as shown in Fig. 3a.The verified cumulative annual Mittivakkat Glacier runoff is illustrated in The assumed accuracy of single outlet discharge measurements is within 5-10%, whereas calculated stage-discharge values might deviate up to 25% from simultaneous manual measurements.However, long-term discharges (monthly and annual) are typically accurate within approximately 5-15% (Hasholt et al., 2006;Mernild and Hasholt, 2009).Statistical analysis from previous SnowModel studies on snow distributions, snow and glacier melt, and runoff from marginal glaciers in Greenland and the GrIS (e.g., Mernild et al., 2006Mernild et al., , 2009)), along with uncertainties in observed discharge used for verification, indicates that simulated influx of runoff to the Sermilik Fjord might be influenced by the same order of uncertainties: We assume an error of 10-25%.This includes uncertainties related to not using routines for air temperature inversion in low lying near coastal areas (Mernild and Liston, 2010) and the associated influence on snow and glacier ice melt and glacier mass-balance simulations; unfortunately no radiosonde data exist for the inner part of the Sermilik Fjord.near the front the glacier to provide a time series of average velocity at the ice front.Due to the high speeds observed, we assumed that speed was constant with ice depth.Averaged glacier width over the region of velocity sampling was obtained from 15-m Landsat imagery.Ice thickness for Helheim Glacier was obtained in 2001 and 2008 by the University of Kansas Coherent Radar Depth Sounder (CoRDS) (Gogineni et al., 2001;Howat et al., 2005).For Fenris and Midg ård glaciers, for which no thickness data are available, ice thickness at the start of the time series was estimated from the height of the calving front assuming a grounded ice front at hydrostatic equilibrium and densities of ice and sea water of 910 and 1027 kg m 3 , respectively.We assume an error of ±50 m in this thickness estimate.Changes in ice thickness through time were then measured from repeat ASTER digital elevation models produced by the LP DAAC and vertically coregistered using tie points over ice-free terrain.These data have a relative error of ±10 m in the vertical (Fujisada et al., 2005).

Geothermal and frictional melting due to basal ice motion
The upper-bounds for the melt water generated through melting at the ice bed due to: (a) geothermal heating; and (b) frictional heating due to basal ice motion were estimated.For (a), it was liberally assumed that the bed was at the melting temperature over all regions with surface elevations below 1200 m a.s.l., and area of 2300 km 2 .A typical geothermal heat flux of 0.05 W m −2 gives a basal melt rate of 5 mm w.eq.y −1 (Cuffey and Paterson, 2010, p. 118) for ice at the melting temperature, totaling approximately 0.01×10 9 m 3 y −1 produced by geothermal heating over this area, which was 2 orders of magnitude less than the contributions from runoff and ice discharge, and can therefore be ignored (Table 2).For (b), the maximum rate of basal melt due to frictional heating caused by ice sliding over the bed is Eq. ( 1):

Conclusions References
Tables Figures

Back Close
Full where, tb is the basal shear stress, ub is sliding speed, ρ is the ice density, and Lf is the latent heat.Again, it was assumed that the bed was at the melting temperature over the drainage area with surface elevations below 1200 m a.s.l.We also assumed that 100% of the ice motion needed to maintain mass balance (i.e., the balance velocity) was accomplished through basal sliding in this region.Balance velocities were obtained from Bamber et al. (2000).Finally, we assumed that the basal drag was equal to the driving stress, which we calculated from the ice thickness and surface elevation maps from Bamber et al. (2000Bamber et al. ( , 2001)).From this we obtain a total melt volume rate of approximately 0.5×10 9 m 3 y −1 , which is approximately 1% of the average total freshwater flux (Table 2).

Terrestrial surface runoff to Sermilik Fjord
Annual (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) cumulative simulated runoff from all seven sub-catchments, and from the entire catchment, to Sermilik Fjord is shown in Fig. 4b.Total runoff to Sermilik Fjord for the modeled decade averaged 4.6×10 9 m 3 y −1 , from a minimum of 2.9×10 9 m 3 y −1 in 1999 to a maximum of 5.9×10 9 m 3 y −1 in 2005; these values were expected to be among the highest since the simulation period included the warmest years since the beginning of instrumental records.For the simulation period, data showed an average insignificant increase in runoff of 1.0×10 9 m 3 (r 2 =0.14, p<0.25;where p level of significance) (Fig. 4b), due to both increasing annual precipitation defined as the melting area times the number of melting days for areas above 2000 m in elevation (Tedesco, 2007) -followed by 2005 as the fourth highest for the simulation period.The largest amount of modeled runoff to the Sermilik Fjord occurred in 2005, and not in 2007 (Fig. 4b).This discrepancy between the GrIS melting conditions and the Sermilik Fjord runoff conditions was due to a record high annual precipitation for 2005 of ∼1800 mm w.eq.y −1 combined with the second highest mean annual summer air temperatures of 2.2 • C (influencing the melting snow and ice conditions) (Fig. 2a).
The record high 2005 precipitation combined with the relatively high percentage of rain (∼65% of the total annual precipitation) was the reason why less precipitation accumulated as snow during winter, and more streamed directly into the fjord as runoff.
The connection between snow melting, melt water retention and refreezing within the snowpack, and runoff is described, e.g., in Hanna et al. (2008), Mernild et al. (2009) related to the variation in annual snow accumulation/precipitation.Weather conditions for the Sermilik Fjord are strongly affected by low pressure systems, especially the associated wind and precipitation which varies significantly due to year-to-year changes in the storm tracks.Most low pressure centers affecting Greenland arrive from directions between south and west, steered by an upper level cyclone, the "polar vortex".During winter these are normally centered over the Canadian Cold Pole and during summers they are less pronounced and centered over the Arctic Ocean (Hansen et al., 2008).This pressure system is of utmost importance for the specific weather conditions at a given position in the southeastern Greenland sector.Therefore, it should be kept in mind, even though maximum melting conditions occurred for the GrIS as in 2007, local variability in precipitation can be the reason for annual runoff peaks, as illustrated for the Sermilik Fjord catchment for 2005.
On a sub-catchment scale, the interannual runoff variability generally followed the variability of the overall runoff to the fjord, showing lowest runoff values in 1999 and highest values in 2005 (Fig. 4b).In Fig. 4b the spatial distribution of runoff to the Sermilik Fjord is illustrated, displaying that sub-area 7 contributed, on average, the lowest annual runoff volume of 0.4×10 9 m 3 y −1 , and sub-area 4 the highest value of Introduction

Conclusions References
Tables Figures

Back Close
Full 1.4×10 9 m 3 y −1 .Besides the general effect of precipitation and summer air temperatures on runoff from all sub-catchments, both the percentage of glacier cover and its hypsometry within each sub-catchment strongly influenced simulated runoff within the region.Generally, sub-catchments showing high fractions of glacier cover and glaciers at low elevations show stronger positive runoff effects during years with high temperatures (Fig. 4c).
The Sermilik Fjord accumulated catchment and sub-catchment runoff (1999-2008) are illustrated in Fig. 4c, showing an overall cumulative freshwater runoff volume of 46.0×10 9 m 3 .The lowest cumulative runoff contribution occurred from sub-area 7, with a total of 3.5×10 9 m 3 , which equalled about 8% of the overall freshwater runoff to the fjord.The highest contribution of 10.4×10 9 m 3 came from sub-area 4 -the Helheim sub-catchment -, which equalled about 25% of the overall runoff.The percentage of cumulative freshwater runoff from the other sub-areas (area 1-3 and 5-6) averaged from 9% to 17% of the overall runoff (Fig. 4d).Obviously sub-catchments with the greatest glacier coverage, combined with the highest percentage of glaciers at low elevations, were the sub-areas where the greatest freshwater runoff contribution to the fjord occurred, and vice versa.In Fig. 4e the differences between sub-catchments 4 and 7 in glacier cover (km 2 ) and in glacier cover (%) within each 100-m elevation interval are shown.Area 4 was the sub-catchment having the greatest glacier area of 910 km 2 from where runoff occurred: ∼30% of the area was located below 500-600 m a.s.l.Area 7 was, however, a sub-catchment having a glacier cover of only 65 km 2 of which only ∼10% was found below 500-600 m a.s.l.Areas 1-3 and 5-6 represent a mixture of the main characteristics found in areas 4 and 7.Even though areas 3 and 6 both have a relatively high glacier cover of 67% and 79%, respectively, the cumulative runoff only accounted for 14% and 9% of the overall runoff to the fjord (Fig. 4d).The reason for these relatively low percentages of runoff values from subarea 3 and 6 were due to the high elevated glacier cover within each sub-catchment.For sub-area 5 the situation was however different: In area 5 the amount of runoff accounted for 17% of the overall runoff to Sermilik Fjord, even though the sub-catchment Introduction

Conclusions References
Tables Figures

Back Close
Full  2008), and consistent with previous GrIS runoff simulations by Mernild et al. (2009).The amount of simulated runoff decreased with increasing altitude, on average by ∼250 mm w.eq. 100 m −1 from the ice margin all the way to the boundary where runoff occurs (Fig. 5); for Jakobshavn drainage area, W Greenland ( 69• N), the value was similar with ∼220 mm w.eq. 100 m −1 (Mernild et al., 2010b).On the GrIS within the Sermilik Fjord catchment (for a latitude range of 65-66 • N) this annual runoff boundary line was located about 25-40 km from the GrIS terminus at an elevation of 1140 m a.s.l. to 1600 m a.s.l., averaging 1150(±140) m a.s.l.

Freshwater flux to Sermilik Fjord
To account for the freshwater flux to the Sermilik Fjord, not only terrestrial surface runoff needs to be addressed, but also: (1) ice discharge influenced by GrIS dynamical processes (as described by Howat et al., 2005Howat et al., , 2008) ) and temperature of near-coastal ocean currents (Holland et al., 2008;Straneo et al., 2010); (2) seasonal changes in internal drainage system due to melting; (3) runoff from subglacial geothermal melting and frictional melting due to basal ice motion; (4) submarine melting at tidewater glacier margins; and (5) precipitation (rain and snow accumulation on sea ice) at the Sermilik Fjord surface area.Unfortunately, seasonal changes in internal drainage system was omitted, due to missing data (values probably insignificant related to the overall 1207 Introduction

Conclusions References
Tables Figures

Back Close
Full terrestrial freshwater flux budget to the Sermilik Fjord), while submarine melting at tidewater glacier margins was integrated in the ice discharge values.Contributions of ice discharge from minor GrIS outlet glaciers, e.g., glaciers located in Johan Petersens Fjord were ignored due to lack of available data, while ice discharge from the three major outlet glaciers: the Helheim Glacier, Fenris Glacier, and Midg ård Glacier at the innermost part of the fjord were included (Figs. 6 and 7).Based on satellite-derived observations, the Helheim average ice discharge for the period 1999-2008, was estimated to be 25.9±2.6×10 9m 3 w.eq.y −1 (Table 2), and for the Fenris and Midg ård Glaciers ice discharge were 2.5±0.5 and 5.5±1.0×10 9 m 3 w.eq.y −1 , respectively (Table 2 and Fig. 6).In Fig. 6 an example of variations in surface ice velocity, ice thickness, and ice discharge for the Helheim Glacier, Fenris Glacier, and Midg ård Glacier are illustrated, showing substantial variations in velocity, ice thickness, and discharge, with a general increase in velocity and discharge after 2002 and peaking in 2005 and 2006.Due to both decreased ice velocity and thickness, ice discharge at Helheim Glacier decreased to earlier levels by 2007 (Howat et al., 2007).However, for the simulation period, ice discharge from the three major outlet glaciers showed an average significant increase of 13.2×10 9 m 3 w.eq.(r 2 =0.49; p<0.01) (Fig. 7).In Table 2 mean annual ice discharge values were combined with annual SnowModel simulated precipitation at the Sermilik Fjord surface area, terrestrial surface runoff, and subglacial geothermal and frictional melting, to deduce the freshwater flux: (1) from the Helheim Glacier Introduction

Conclusions References
Tables Figures
The Helheim Glacier represents one of the major outlet glaciers from the GrIS, which dominated 65% of the freshwater flux into Sermilik Fjord.Runoff only forms a minor part (11%) of the overall freshwater flux to the fjord.Similar conditions were found for the Jakobshavn drainage area (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007), where 7% of the average annual freshwater flux originated from surface runoff (Mernild et al., 2010b).For two of the major GrIS outlet glacier sub-catchments (Helheim and Jakobshavn) it can be concluded that runoff was a minor contributor to the freshwater flux, which was highly dominated by ice discharge.For both Helheim and Jakobshavn, it should be kept in mind that the freshwater flux to the Sermilik Fjord and the Illulissat Fjord, respectively, does not include values for a melting internal drainage system through out runoff season.The freshwater flux should be seen as first estimate, even though, omitted freshwater inputs from changes in englacial and subglacial internal drainage system due to melting only counts for a minor part of the overall flux to the fjords.

Summary and conclusion
The amount of freshwater runoff reaching the ocean from marginal glaciers, the GrIS, and ice free landscapes depends on the precipitation and storage changes in reservoirs of ice, snow, and water on land.In many places around Greenland, glaciers calve directly into the sea and the overall flux of freshwater from specific catchments, e.g., the Sermilik Fjord catchment, will be influenced by or even dominated by a discharge of calving ice.At the Sermilik Fjord catchment, 85% of the average annual freshwater flux of 40.4×10Full  -2.3±0.5 2.5±0.5 2.6±0.5 2.8±0.5 2.9±0.5 2.8±0.5 2.2±0.5 2.5±0.5 2.4±0.5 2.5±0.5 (6%) 33.9 (85%) Satellite estimated ice discharge from the Midg ård Glacier including error, 10 9 m 3 y −1   Mernild et al., 2006), and for different areas around the GrIS are illustrated (Mernild et al., 2008(Mernild et al., , 2009)).Introduction

Conclusions References
Tables Figures

Back Close
Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 3b .
This underestimation of runoff is expected to be due to: (1) englacial and subglacial water flow to neighboring glacier sub-catchments; (2) uncertainties associated with model inputs; (3) unrepresented or poorly-represented processes in SnowModel; and (4) uncertainties related to runoff observations (see below).A dye trace study on the Mittivakkat Glacier by Mernild (2006) confirmed englacial and subglacial water flow through crevasses, moulins, and tunnels to neighboring glacier sub-catchments, and modeling studies further indicated that the location of the northern catchment divide on the Mittivakkat Glacier itself may vary because of crevasse and tunnel development due to, e.g., glacier dynamic activity.Therefore, the glacier may deliver and, in this case, receive freshwater water across this divide from neighboring glacier sub-catchments.Discussion Paper | Discussion Paper | Discussion Paper | Time series of ice flux to the calving front, which we term the ice discharge, for Helheim Glacier, Midg ård Glacier, and Fenris Glacier were calculated from observed average surface velocity, glacier width and estimated ice thickness.Speeds were measured from automated Repeat-Image Feature Tracking (RIFT) using pairs of orthorectified images from: (1) Landsat 7 Enhanced Thematic Mapper Plus (panchromatic band) distributed by the United States Geological Survey; (2) visible to near-infrared bands of the Advanced Spaceborne Thermal Emissivity and reflection Radiometer (ASTER) distributed by the NASA Land Processes Distributed Active Archive (LP DAAC); and (3) SPOT-5 panchromatic images distributed through the SPIRIT program.Landsat and ASTER images have a pixel resolution of 15 m and the 5 m SPOT-5 images were down-sampled to 15 m for RIFT processing.The Multi-Image/Multi-Chip (MIMC) RIFT algorithm (Ahn and Howat, 2010) was used to measure surface displacements every 100 m.Individual displacement vectors were then averaged over a 1-km wide band 1202 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 13) and increasing mean annual summer air temperature (June through August) (r 2 =0.32).An increase in precipitation for the Sermilik Fjord catchment of ∼15% decade −1 , which was above the average increase of ∼1% decade −1 estimated by ACIA (2005).For the simulation period, 2007 showed the largest satellite-derived GrIS cumulative melt extent followed by 2005 (Steffen et al., 2008), but also the largest melt index -Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | areal and the glacier cover area were relative low.The reason for the relatively high runoff volume from sub-area 5 was because of the low elevated glacier cover in the sub-area.In Fig. 5 the spatial distributions of annual cumulative runoff to Sermilik Fjord are illustrated for 1999 through 2008.Those parts of the fjord catchment exhibiting glaciers covering low altitudes, e.g., both marginal glaciers and the Helheim glacier terminus showed the highest simulated runoff values.At the Helheim glacier terminus the areally-averaged annual maximum runoff ranged from ∼1.8 m w.eq. in 2003 to more than ∼3.8 m w.eq. in 2007.Simulated runoff values which seemed to be in line with previously published values, e.g., by Ettama et al. ( Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 1999-2008) originated from ice discharge, 11% from terrestrial surface runoff, 3% from net precipitation at the fjord area, and 1% from subglacial Discussion Paper | Discussion Paper | Discussion Paper | frictional melting.In period of a warmer climate, as for example during the recent decade an increase in runoff (r 2 =0.14) and ice discharge (r 2 =0.49) occurred.The Sermilik Fjord increasing runoff was caused by increasing mean annual summer air temperature and precipitation: even though maximum melting conditions occurred in 2007, local variability in precipitation can be the reason for annual runoff peaks, as illustrated for the Sermilik Fjord for 2005Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 2 .
Fig. 2. (a) Time series of observed mean monthly air temperature from coastal (Station Tasiilaq, Coast, Nunatak, TAS U, TAS L, and KULU) and GrIS stations (NASA-SE).Observed cumulative monthly precipitation from coastal stations (Tasiilaq; light color, and Coast and Nunatak; dark colors) are illustrated; and (b) mean monthly air temperature lapse rates for all the meteorological stations in the simulation domain, for Station Coast and Nunatak (classified as coastal region;Mernild et al., 2006), and for different areas around the GrIS are illustrated(Mernild et al., 2008(Mernild  et al.,  , 2009)).
Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 3 .
Fig. 3. (a) Observed and simulated Mittivakkat Glacier catchment runoff from 1999-2005.The linear regression was used for verification of the Sermilik Fjord simulated runoff.Be aware that the annual observed runoff periods are different.The inset figure indicates the general location of the Mittivakkat Glacier catchment (black polygon) within sub-catchment area 1 (for a general location of the sub-catchments see Fig. 4a) inside the Sermilik Fjord catchment; and (b) observed, simulated, and verified runoff from the Mittivakkat Glacier catchment from 1999-2005.

Fig. 4 .Fig. 5 .
Fig. 4. (a) Area of interest including the seven sub-catchments for the Sermilik Fjord (simulated in River Tools); (b) time series (1999-2008) of annual sub-catchment simulated runoff and annual cumulative runoff, including trend line (linear) for cumulative runoff; (c) cumulative sub-catchment runoff and overall runoff; (d) percentage of sub-catchment runoff of total runoff; (e) glacier cover distribution in percentage and square kilometer within the elevations from where runoff occurred for the sub-catchment with the lowest cumulative runoff (sub-catchment 7) and the highest (sub-catchment 4); and (f) the percentage of sub-catchment glacier cover within the area of interest.

Table 1 .
Meteorological input data for the Sermilik Fjord simulation based on meteorological station data on the GrIS: KULU and NASA-SE (provided by University of Colorado at Boulder) and TAS L and TAS U (by Geological Survey of Denmark and Greenland (GEUS)), and from the outside the GrIS: Station Tasiilaq (by Danish Meteorological Institute (DMI)) and Station Coast and Station Nunatak (by University of Copenhagen, Department of Geography and Geology).The abbreviations indicate: (Ta) air temperature, (Rh) relative humidity, (Ws) wind speed, (Wd) wind direction, and (P) precipitation.For station locations see Fig. 1b.

Table 2 .
Freshwater flux from the Helheim Glacier catchment and to the Sermilik Fjord based on SnowModel simulated freshwater runoff, precipitation at the Sermilik Fjord surface area, subglacial geothermal melting and subglacial frictional melting due to basal ice motion, and satellite-derived ice discharge from the Helheim Glacier, Fenris Glacier, and Midg ård Glacier (the three major outlet glaciers in Sermilik Fjord catchment).Ice discharge is recalculated from