The role of albedo and accumulation in the 2010 melting record in Greenland

Analyses of remote sensing data, surface observations and output from a regional atmosphere model point to new records in 2010 for surface melt and albedo, runoff, the number of days when bare ice is exposed and surface mass balance of the Greenland ice sheet, especially over its west and southwest regions. Early melt onset in spring, triggered by above-normal near-surface air temperatures, contributed to accelerated snowpack metamorphism and premature bare ice exposure, rapidly reducing the surface albedo. Warm conditions persisted through summer, with the positive albedo feedback mechanism being a major contributor to large negative surface mass balance anomalies. Summer snowfall was below average. This helped to maintain low albedo through the 2010 melting season, which also lasted longer than usual.


Introduction and objectives
Large positive near-surface temperature anomalies occurred along the coast of the Greenland ice sheet during the year 2010 (e.g., Box et al 2010). For example, at Aasiaat (68 • 42 35 N 52 • 52 10 W) along Greenland's west coast, 2010 as a whole was the warmest since records began in 1951, with records also set for winter, spring, May and June. Narsarssuaq (61 • 09 39 N 45 • 25 32 W) in southern Greenland, saw its warmest winter and spring, its warmest May, and its warmest annual average since records began in 1951; Nuuk, the capital of Greenland, located along the southwest coast (64 • 10 30 N, 51 • 44 20 W), with a temperature time series extending back to 1873, saw record warmth for winter, spring, summer and the year as a whole (Cappelen 2010).
Surface melting over the Greenland ice sheet, which can be estimated from satellite data, ground observations or models (Abdalati and Steffen 1997, Mote 2007, Nghiem et al 2001, Hall et al 2009, Tedesco 2007, Hanna et al 2008, Fettweis et al 2010b, Ettema et al 2010 was also exceptional in 2010 . Results obtained by applying the algorithm reported in Tedesco (2007) to spaceborne microwave brightness temperatures (e.g., Armstrong et al 1994, Knowles et al 2002 are consistent with those reported in Box et al (2010), showing that large areas of the ablation zone in south Greenland underwent melting up to 50 days longer in 2010 compared to the 1979-2009 average, with melting in 2010 starting exceptionally early at the end of April and ending quite late in mid September (figure 1). These results are confirmed by surface measurements.
Near-surface air temperature is often used as a proxy for surface melting. Previous studies have analyzed exceptional melting events mainly focusing on the relationship between melt and near-surface temperatures (e.g., Mote 2007, Tedesco 2007). However, melting and the surface mass balance (SMB), also depend on accumulation, radiation conditions, refreezing and sublimation, the latter relatively small and constant in time (Box and Steffen 2001, Fettweis 2007, van den Broeke et al 2008. Surface melt and albedo are intimately linked: as melting increases, so does snow grain size, leading to a decrease in surface albedo which then fosters further melt. Also, changes in accumulation can affect the seasonal evolution of surface albedo, which influences the SMB. It is hence not sufficient to analyze near-surface temperature trends to understand the driving mechanisms of extreme mass loss, and studying the role of albedo and accumulation becomes crucial to provide a more robust understanding of the exceptional melting detected by satellite microwave sensors.
In this study, we report results derived from spaceborne sensors, surface glaciological observations and regional atmospheric model outputs regarding the surface albedo, accumulation and bare ice exposure over the Greenland ice sheet during the summer of 2010. Our results indicate that negative surface albedo anomalies were especially prominent over west Greenland, with bare ice exposed earlier than previous years. In addition, increased runoff and reduced accumulation likely contributed to a strongly negative SMB.

Satellite data
We use the moderate-resolution imaging spectroradiometer (MODIS) 16-day gridded albedo product (http://www-modis. bu.edu/brdf/userguide/intro.html) to study anomalies in both directional hemispherical reflectance (black-sky albedo) and bi-hemispherical reflectance (white-sky albedo) in the shortwave, visible and near-infrared bands. Black-sky albedo is the albedo under direct illumination (or direct beam contribution), whereas white-sky albedo is the one under diffuse or indirect light. Black-and white-sky albedos can be combined as a function of the diffuse skylight for a representation of the actual albedo such as measured by field instruments (Lucht et al 2000). The MODIS product, which has been evaluated over Greenland (e.g. Stroeve et al 2005), is provided every 8 days using 16 days acquisitions from both the TERRA and AQUA satellites. Data are provided on a 0.05 • latitude/longitude grid. While we discuss only the shortwave white-sky albedo, the conclusions that follow also apply to black-sky, visible and near-infrared albedos.

Glaciological data
The Institute of Marine and Atmospheric Research Utrecht, (IMAU) installed several automatic weather stations (AWSs) SMB can be estimated at the local spatial scale from stakes and acoustic height rangers. Apart from atmospheric radiation, the stations measure basic meteorological variables like wind speed and direction, temperature, relative humidity and air pressure. SMB measurements are performed along the Ktransect at selected sites. The three AWS sites (S5, S6 and S9) are located, respectively, 6 km (S5), 37 km (S6) and 91 km (S9) from the ice sheet margin at an elevation of 490 m (S5), 1020 m (S6) and 1520 m (S9) a.s.l. Mass balance sites (S4, SHR, S7, S8, S10) are located 3 km (S4), 14 km (SHR), 52 km (S7), 63 km (S8) and 143 km (S10) from the ice sheet margin at an elevation of 340/390 m (S4, 340 m was the initial value), 710 m (SHR), 1110 m (S7), 1260 m (S8) and 1850 m (S10) a.s.l. Mass balance data are available over the period 1991-2010. Weather station data are available for the period 2003-2010. Further information on the stations can be found in van de Wal et al (2005).

The surface and energy balance model
The surface and energy balance model is the regional climate model MAR (Modèle Atmosphérique Régional), coupled to the 1D Surface Vegetation Atmosphere Transfer scheme SISVAT (Soil Ice Snow Vegetation Atmosphere Transfer). The schemes and the set-up employed for the present study are fully described in Fettweis et al (2010a), Fettweis (2007) and Lefebre et al (2005). Differently from most existing models, MAR is coupled to a snow physical model called CROCUS (Brun et al 1992). CROCUS is a one-dimensional multilayered energy balance model consisting of a thermodynamic module, a water balance module taking into account the refreezing of meltwater, a snow metamorphism module, a snow/ice discretization module and an integrated surface albedo module. For this study, MAR outputs are derived at a spatial resolution of 25 km for the period 1958-2010. The ERA-40 reanalysis , the ERA-INTERIM reanalysis (2002( -June 2010 and after that, the operational analysis (July 2010-September 2010) from ECMWF are used to initialize the meteorological fields at the beginning of the simulation in September 1957 and to force every 6 h the lateral boundaries with temperature, specific humidity and wind components during the simulation. The sea surface temperatures (SST) and the sea-ice extent in the SISVAT module are also prescribed by the reanalyses.
No reinitialization or correction are applied to the model outputs. The (re)analysis data are available every 6 h at a resolution of at least 1 • . The MAR spatial domain is represented by an area of 2000 km × 3500 km centered at 70 • N latitude and 40 • W longitude. There are at least ten pixels between the Greenland and the model boundaries. The integration domain is shown in Fettweis et al (2005). An ice sheet mask is then applied to analyze only those pixels belonging to the ice sheet (Fettweis 2008). Outputs from the MAR model have been shown to be consistent with results obtained from passive microwave observations (Fettweis et al 2010b). (2) enhanced melting; and (3) exposure of bare ice for a longer period of time during the melting season. Figure 2(b) provides the standardized anomaly map (the departure from the mean divided by the standard deviation of the distribution) of the number of days when bare ice was exposed in 2010 based on MAR outputs. According to MAR, in 2010 bare ice was exposed for a period up to four times longer than the standard deviation computed over the 1979-2009 period along the western region of the ice sheet, with absolute anomalies peaking up to 50 days. Figure 2(c) shows the summer snowfall anomalies (in mm water equivalent, WE) from MAR, still relative to 1979-2009 means. The model suggests that summer accumulation in 2010 over the southwestern part of the Greenland ice sheet was two standard deviations below the 1979-2009 mean. This, together with large positive nearsurface temperature anomalies, very likely led to the premature exposure of bare ice (figure 2(b)) and the reduced albedo (figure 2(a)).  melting season, with the absence of sporadic peaks that would result from snowfall events (which are present during years when positive anomalies of summer snowfall are measured and modeled). The match between the seasonal trends derived from MODIS and those measured on the ice lends credence to the MAR results, pointing to negative snowfall anomalies for 2010 (figure 2(c)) and premature exposure of bare ice ( figure 2(b)).

Results and discussion
Early exposure of bare ice and reduced snowfall in 2010 are also evident from surface height measurements along the K-transect. Figure 3(c) illustrates surface height changes over the years 2003-2010 at the S9 station. We focus on the station S9 as it is situated at 1500 m elevation, at the approximate  Figure 4 shows monthly standardized anomalies for 2010 (using 1979-2009 as a reference for consistency with the passive microwave data time series) from MAR of near-surface temperature, albedo, snowfall, meltwater, bare ice area and melt area excluding bare ice. Snowfall and nearsurface temperature anomalies for the autumn and winter of 2009 (September-April) are also reported as a reference. The model indicates that surface albedo was persistently below the average (two standard deviations) for the summer. The meltwater produced in May (August) was ∼3 (∼2.5) standard deviations above the mean. Bare ice area in May was ∼3.5 standard deviations above the mean and it remained persistently around two standard deviations above average through summer. By contrast, the melt area excluding bare ice was exceptional in May, August and September.

Conclusions
Our analysis of remote sensing data, surface glaciological observations and model outputs paints a portrait of strongly negative surface mass balance during 2010 promoted by a strong warmth and enhanced by large negative anomalies of albedo and accumulation and large positive anomalies of days when bare ice was exposed. Our results clearly indicate that, beside near-surface temperature, other factors must be considered to properly analyze extreme events such as the one that occurred in 2010. The melt season started in mid April, after a warm and dry winter. Early melt onset, triggered by large positive near-surface temperature anomalies during May 2010 (up to +4 • C above the mean) contributed to accelerated snowpack metamorphism and premature bare ice exposure, with the consequence of rapidly reducing the surface albedo. Reduced accumulation in 2010, and the positive albedo feedback mechanism are likely responsible for the premature exposure of bare ice. While June and July temperature anomalies were not exceptional, being +1.5 • C above the 1979-2009 average, anomalously warm conditions persisted with the positive albedo feedback mechanism contributing to large negative SMB anomalies. Summer snowfall, which helps to increase surface albedo, was below average. Melt during August and September was also exceptional, consistent with low surface albedos and near-surface temperature anomalies of up to +3 • C, yielding a long ablation season.
The surface melt time series from the combined passive microwave record, 1979-present, is characterized by an upward trend, with considerable year-to-year variability primarily related to atmospheric conditions. Viewed in this context, the unusually warm conditions over the Greenland ice sheet in 2010 and reduced snowfall can be related to persistence of a 500 hPa high ridge from late spring through summer. Averaged for May-August, 500 hPa heights were above normal over all of Greenland, with the largest anomalies of up to 80 m over the south-central ice sheet. This anomalous ridge was associated with 700 hPa temperature anomalies over south-central Greenland of up to +3 • C. The ridge and associated 700 hPa temperature anomaly was best expressed in May and August, coinciding with near-surface air temperature records.