Atmospheric and oceanic climate forcing of the exceptional Greenland ice sheet surface melt in summer 2012

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. (2013) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: /joc.3743 Atmospheric and oceanic climate forcing of the exceptional Greenland ice sheet surface melt in summer 2012 Edward Hanna, a * Xavier Fettweis, b Sebastian H. Mernild, c,d John Cappelen, e Mads H. Ribergaard, f Christopher A. Shuman, g,h Konrad Steffen, i,j,k Len Wood l, and Thomas L. Mote m a Department of Geography, University of Sheffield, Sheffield, UK b Laboratory of Climatology, Department of Geography, University of Liège, Liège, Belgium c Climate, Ocean and Sea Ice Modelling Group, Computational Physics and Methods, Los Alamos National Laboratory, Los Alamos, NM, USA d Glaciology and Climate Change Laboratory, Center for Scientific Studies/Centre de Estudios Cientificos (CECs), Valdivia, Chile e Danish Meteorological Institute, Data and Climate, Copenhagen, Denmark f Centre for Ocean and Ice, Danish Meteorological Institute, Copenhagen, Denmark g Joint Center for Earth Systems Technology, University of Maryland, Baltimore County, Baltimore, MD, USA h Cryospheric Sciences Laboratory, NASA Goddard Space Flight Center, Greenbelt, MD, USA i Swiss Federal Research Institute WSL, Birmensdorf, Switzerland j Institute for Atmosphere and Climate, Swiss Federal Institute of Technology, Zürich, Switzerland k Architecture, Civil and Environmental Engineering, École Polytechnique Fédéral de Lausanne, Lausanne, Switzerland l School of Marine Science and Engineering, University of Plymouth, Plymouth, UK m Department of Geography, University of Georgia, Athens, GA, USA ABSTRACT: The NASA announcement of record surface melting of the Greenland ice sheet in July 2012 led us to examine the atmospheric and oceanic climatic anomalies that are likely to have contributed to these exceptional conditions and also to ask the question of how unusual these anomalies were compared to available records. Our analysis allows us to assess the relative contributions of these two key influences to both the extreme melt event and ongoing climate change. In 2012, as in recent warm summers since 2007, a blocking high pressure feature, associated with negative NAO conditions, was present in the mid-troposphere over Greenland for much of the summer. This circulation pattern advected relatively warm southerly winds over the western flank of the ice sheet, forming a heat dome over Greenland that led to the widespread surface melting. Both sea-surface temperature and sea-ice cover anomalies seem to have played a minimal role in this record melt, relative to atmospheric circulation. Two representative coastal climatological station averages and several individual stations in south, west and north-west Greenland set new surface air temperature records for May, June, July and the whole (JJA) summer. The unusually warm summer 2012 conditions extended to the top of the ice sheet at Summit, where our reanalysed ( ) DMI Summit weather station summer (JJA) temperature series set new record high mean and extreme temperatures in 2012; 3-hourly instantaneous 2-m temperatures reached an exceptional value of 2.2 C at Summit on 11 July These conditions translated into the record observed ice-sheet wide melt during summer However, 2012 seems not to be climatically representative of future average summers projected this century. Additional Supporting information may be found in the online version of this article. KEY WORDS climate change; global warming; Greenland; surface melt extent; temperature Received 11 November 2012; Revised 29 April 2013; Accepted 4 May Introduction The Greenland Ice Sheet (GrIS) is a highly sensitive indicator of regional and global climate change, and has been undergoing rapid warming and mass loss during the last 5 20 years (Hanna et al., 2008, 2012a, 2012b; Rignot et al., 2011). Much attention has been given to the NASA announcement of record surface melting of the GrIS in mid-july, This event, where 98.6% of the * Correspondence to: E. Hanna, Department of Geography, University of Sheffield, Sheffield, UK. ehanna@sheffield.ac.uk Now retired. entire ice sheet surface including at the highest-altitude, Summit region was found from multiple independent satellite sensors, field verification and model results to have melted at times primarily on 11 and 12 July 2012, was clearly unprecedented in the satellite record of observations dating back to the 1970s and probably unlikely to have occurred previously for well over a century (Nghiem et al., 2012). A further melt event where 79.2% of the surface melted on 29 July 2012 was also unusual, although Summit experienced a similarly brief melt event very much like this in mid-july 1995 when 60% of the ice-sheet surface melted according to co-author TM s long-running satellite-based melt dataset 2013 Royal Meteorological Society

2 E. HANNA et al. used in Nghiem et al. (2012). Even though over half the ice sheet surface up to 2500 m had already been melting in recent previous warm summers (e.g. Mernild et al., 2011; Mernild and Liston 2012a), and therefore this process only had to reach up 500 m vertically to reach the summit of the ice sheet, the 2012 melt event raises the questions of what caused this surface melt of nearly the entire ice sheet and how often such an event is likely to occur in the future. Two clear and possibly interlinked candidates for broad GrIS surface melting are considered: meteorological forcing from a warmer atmosphere (i.e. warmer air masses being advected over the ice sheet) and/or the influence of a warmer ocean. There has been considerable discussion of both these factors in the recent literature in the context of Greenland climate change and links with GrIS mass balance (e.g. Hanna et al., 2012b). Here, first we ask how unusual were (1) the summer 2012 GrIS melt and (2) surface air temperatures over Greenland, and second we examine each of the above potential forcing factors (atmosphere and ocean) in turn, before asking what the record 2012 melt event and recent climate variability tells us about likely future climate change in Greenland. 2. How unusual was the summer 2012 GrIS melt? Previous research utilizing passive microwave observations available nearly continuously since 1972 (Mote 2007; Abdalati 2008) has clearly shown that GrIS melt extent has been increasing over the past few decades, with several new extent records being set in the satellite record in the past eight years (e.g. Hanna et al., 2008; Mernild et al., 2011). To understand the conditions that led to such widespread melt, we first looked at the glaciological pre-conditioning by snow accumulation simulated by the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis-driven regional climate model (RCM) MAR (Fettweis, 2007; Fettweis et al., 2013b) during winter (September 2011 April 2012). We found that accumulation was close to the long-term mean with a small positive anomaly in north-east Greenland (Figure S1) but the anomalies are within two standard deviations of mean conditions and are therefore not significant. However, as might be anticipated from the NASA satellite results discussed above, from June to August 2012 the surface melt extent derived from passive microwave satellite data (Tedesco et al., 2013) and simulated by MAR and SnowModel was exceptional (Snow- Model is a physically energy-balance based, spatially distributed, meteorological and snow and ice evolution modelling system, see Liston and Elder 2006; Mernild et al., 2010, 2011; Mernild and Liston, 2012b for additional material). Relative to , the MAR meltwater (runoff) production in JJA 2012 was 120% (130%) above average and indeed was the highest (i.e. 3.5 standard deviations above the mean ) modelled meltwater and runoff since the beginning (1958) of the ECMWF-forced MAR simulations. According to the melt retrieval algorithm from Mote (2007) based on passive microwave satellite data, the observed maximum 2012 melt extent occurred on 11 July [Day of year (DOY) 193] and covered 90% of the GrIS surface area (Figure 1(a)). On the following two days, 12 and 13 July (DOY 194/195), the satellite-derived melt extent covered 88 and 81% respectively. The total area that experienced any melt during the week of 8 15 July was 98%, which agrees well with the NASA melt map (Nghiem et al., 2012), which is a union, rather than an intersection, of the several individual satellite products, including passive microwave satellite-based data from co-author TM. About two weeks later on 29 July (DOY 211), an additional melt event was observed to reach a total area of 67%. The equivalent simulated daily GrIS melt extents from SnowModel both in mid- and late-july were well in excess of two standard deviations above the mean ( ) (Figure 1(a)). Relative to the period (Figure 1(b), updated from Mernild et al., 2011), 2012 saw by far the maximum ice sheet melt extent, and by any measurement technique was significantly above the previous melt extent area record of 52% in In Figure 1(c), the daily SnowModel simulated melt extent is illustrated for This figure illustrates how the annual melt duration (i.e. with >10% melt extent) culminated in an extended melting period of 70 d (Mernild et al., 2011; updated), where the entire 2012 melt duration (>10% melt extent) for the Greenland ice sheet was 130 d. The record surface melt in summer 2012 is likely to have been influenced by dominant anti cyclonic conditions, with such circulation centred over southern Greenland (Figure 2), Such conditions are generally indicated by a highly negative North Atlantic Oscillation (NAO) index: in this case of 1.60 for JJA 2012 according to the Hurrell et al. (2012) NAO dataset. Such atmospheric conditions favour warmer and drier than normal summers by enhancing warm southerly air advection along the west coast and shifting precipitation towards the north of the ice sheet (Box et al., 2012; Hanna et al., 2012b; Fettweis et al., 2013a). This anomaly in the general circulation induced: (i) Exceptionally high JJA temperatures at 600 hpa (T600 at 4 km altitude, i.e. above the ice sheet), inducing record surface melt more extreme than that previously observed for the previous five summers (Fettweis et al., 2013a); (ii) A significant lack of precipitation during summer 2012 in south-east Greenland: both observed (e.g. 0 mm at Tasiilaq, 0.7 mm at Narsarsuaq and 21.6 mm at Qaqortoq in June 2012, and only 1.4 mm at Tasiilaq in July 2012) and as simulated by MAR (Figure S2). Combined with sunnier conditions than normal in southern Greenland (Figure S3), this caused a reduction of the surface albedo [as derived from Moderate-Resolution Imaging Spectroradiometer

3 CLIMATE FORCING OF 2012 GREENLAND ICE MELT Figure 1. GrIS surface melt extent: (a) Time series of daily simulated melt extent from 2012, simulated average with standard deviations (Mernild et al., 2011; Mernild and Liston, 2012a), and satellite-observed daily melt extent (Mote 2007, updated); (b) time series of simulated maximum annual melt extent ( ) with fitted trend line; and (c) daily simulated melt extent from January through December for (MODIS) satellite data and simulated by MAR] during the whole 2012 summer, which enhanced the melt induced by the warmer-than-normal atmosphere (Tedesco et al., 2011, 2013). Furthermore, summer 2012 succeeded a series of warm summers (2007, 2008, 2010 and 2011), and therefore, once the winter accumulation substantially melted in June 2012, lower albedo areas (including bare ice zones) resulting from previous summers reappeared, which accentuated the above-normal temperatures and consequent melt. A MAR simulation forced by 6-hourly means of the ECMWF ERA-Interim reanalysis-based (Dee et al., 2011) sea-surface temperature (SST) and sea-ice concentration (SIC) averaged over (when the SST was significantly lower than in 2012 and SIC much higher) does not significantly affect the near-surface temperature simulated by MAR over the GrIS in Summer 2012 (Figure 3), where for that scenario runoff declines by only 1%. This finding supports a similar result previously shown by Hanna et al. (2009). Our MAR sensitivity experiment suggests that the general circulation, and in particular the JJA temperatures in the

4 E. HANNA et al. Figure 2. Anomaly of JJA 2012 temperature at 600 hpa (T600) simulated by the NCEP/NCAR Reanalysis with respect to Also shown are the JJA mean geopotential height at 600 hpa (blue lines) and the anomalies of the JJA 2012 wind at 600 hpa (in mauve). Figure 3. Sensitivity of the JJA 2012 near-surface temperature simulated by MAR using the climatological mean sea-surface temperature (SST) and sea-ice concentration (SIC) instead of JJA 2012 SST and SIC values. The JJA 2012 mean near-surface wind is shown in orange. free atmosphere (e.g. at hpa), is the dominant factor impacting the melt over the GrIS as indicated by Fettweis et al. (2013a) and that the SST and SIC anomalies (also affected by the general circulation) do not significantly enhance the net melt over Greenland. It should be noted that the low dependency of MARsimulated melt on SST and SIC might be an artefact of the RCM, since atmospheric conditions at the MAR lateral boundaries are not affected by the changes made to SST and SIC in the MAR km integration domain shown as Figure 1 in Fettweis et al. (2005). However, bearing in mind the general circulation which prevails above GrIS (with winds typically blowing from south-west to north-east), air masses come in on average from Canada and are therefore typically little affected by oceanic conditions around Greenland. In addition, air masses in the boundary layer (i.e. those most influenced by SSTs) do not generally reach the ice sheet because of katabatic winds which often flow down the ice sheet and limit the penetration of these near-surface marine air masses to the tundra area, as shown in Figure 3. However, according to Overland et al. (2012), it is possible that the oceanic conditions in the Arctic and North Atlantic may alter the NAO and general atmospheric circulation over Greenland and so may influence the melt record. 3. Greenland ice sheet meteorology observational analysis summer Coastal Greenland temperatures We used monthly values of mean daily, mean daily maximum and daily minimum, and extreme maximum surface air temperature (SAT) data from Greenland coastal (or near-coastal) synoptic weather stations operated by the Danish Meteorological Institute (DMI) (Cappelen et al., 2001; Cappelen, 2011, 2012). DMI station locations and time periods are given in Figure 4 and Table I. Updated data to include summer 2012 (i.e. until August 2012 inclusive) were used in the analysis. All these temperature data have been homogenized using the standard normal homogeneity test (Alexanderson, 1986), including comparisons with neighbouring stations records where available (Cappelen et al., 2001; Cappelen, 2011, 2012). Five long-running sites, Upernavik, Ilulissat, Nuuk, Ivituut/04270 Narsarsuaq (often referred to by DMI as a single merged temperature series) and Tasiilaq, have SAT records extending back before 1900 giving a reasonable distribution around the ice-sheet margin whereas many other sites began recording around 1960 (Table I). Following Hanna et al. (2012a, 2012b) we use Composite Greenland Temperature 2 and 3 (CGT2 & CGT3) seasonal SAT time series, where CGT2 is based on means of data from nine stations asterisked in Table I and begins in 1961, and CGT3 is based on means of SAT data from a subset of these stations, i.e. the five above-mentioned long-running stations: Upernavik, Ilulissat, Nuuk, Ivituut/04270 Narsarsuaq, and Tasiilaq. CGT2 is more comprehensive but CGT3 has the advantage of summarising coastal temperature changes since 1895; both CGT records are very strongly correlated for their common overlap period (Figure 5). Figure 5 summarizes summer (JJA) and July mean SAT for the DMI Greenland stations and CGT averages,

5 CLIMATE FORCING OF 2012 GREENLAND ICE MELT Table I. All DMI stations except for Summit ( 3200 m above sea-level) are within 100 m of sea-level. The GC-Net AWS at Summit (Swiss Camp) is at 3200 (1176) m elevation. Station name Data source World Meteorological Organization (WMO) code Latitude ( N) Longitude ( W) Available data period *#Upernavik DMI 34210/04210/ Sep Aug *Aasiaat DMI Jan Aug *#Ilulissat DMI 34216/04216/ Jul Aug *Sisimiut DMI 04230/ Jan Aug Kangerlussuaq DMI May 1973 Aug *#Nuuk DMI 34250/ Jan Aug *Paamiut DMI Jan Aug Ivittuut DMI Jan Dec *#Narsarsuaq DMI Jan Aug *Qaqortoq DMI Jan Aug Danmarkshavn DMI Jan Aug Ittoqqortoormiit DMI 34339/ Jan Aug Aputiteeq DMI Jan Aug *#Tasiilaq DMI 34360/ Oct Aug Ikermiuarsuk DMI Jan Aug Ikerasassuaq DMI Jan Aug Summit DMI Nov Aug Summit GC-Net May 1996 Jul Swiss Camp GC-Net May 1991 Jul Sources: Cappelen (2011), Cappelen et al. (2001), Steffen and Box (2001). All DMI stations except for Summit (3202 m) are within 100 m of sea-level. The GC-Net AWS at Summit (Swiss Camp) is at 3199 (1176) m elevation. Stations marked * are used in our CGT2 series, and stations also marked # are used in our CGT3 series, after Hanna et al. (2012a, 2012b). Neighbouring DMI stations Ivittuut and Narsarsuaq are combined to provide a long south Greenland temperature time series. Table II. Five highest mean daily surface air temperature values and 2012 rank (1st = warmest) for DMI Greenland coastal stations and Greenland Temperature Composites for summer (JJA). Year/temperature ( C) 2012 rank CGT2 2012/8.2, 2010/8.1, 2003/7.8, 2007/2.6, 2005&2008/7.6 1st CGT3 2012/8.4, 2010/8.3, 2003/8.0, 1931,2007 &2008/7.9 1st Upernavik 2012/7.4, 2011/7.1, 2008/6.9, 1960&2009/6.7 1st Aasiaat 2012/7.9, 2010/7.0, 1960&2011/6.9, 2009/6.7 1st Ilulissat 1960/8.8, 1931/8.5, 2005&2010/8.1, 1987&2012/8.0 5th= Sisimiut 2012/9.0, 2010/8.4, 2003/8.0, 2007/7.8, 2009&2011/7.7 1st Kangerlussuaq 2010/11.1, 1974/11.0, 2003,2011&2012/10.9 3rd= Nuuk 2012/9.0, 2010/8.1, 2007/7.8, 1948/7.7 1st Paamiut 2010/7.3, 2006/7.1, 2005/7.0, 2003/6.9, 2012/6.6 5th Ivittuut/Narsarsuaq 2012/11.4, 2003&2010/11.3, 1917&1948/11.1 1st Qaqortoq 2003/8.9, 2010/8.7, 1991/8.6, 2007/8.4, 2005&2012/8.3 5th= Danmarkshavn 2008/4.6, 2012/4.5, 2003/4.4, 1966&2002/3.5 2nd Ittoqqortoormiit 1949/7.3, 1932/7.2, 1936/7.0, 1948/6.8, 1947/6.7 16th Aputiteeq 1980/3.6, 2009/3.3, 2008/3.0, 2002,2010&2012/2.6 4th= Tasiilaq 2003/7.9, 2010/7.8, 1932/7.6, 1947/7.4, 1936,1939&1950/7.3 58th= Ikerasassuaq 2010/8.1, 2008/7.7, 2005/7.6, 2007&2012/7.5 4th= and Tables II VII (Tables SI SVI) summarize summer, June and July SAT anomalies for, respectively, mean daily temperature, and extreme daily maximum temperature (mean daily maximum temperature and mean daily minimum temperature). By a number of different measures, summer (JJA) 2012 was clearly the warmest on record. Mean daily SAT for summer (JJA) was greatest in the respective records for CGT2, CGT3, Upernavik, Aasiaat, Sisimiut, Nuuk and Narsarsuaq (Table II), where the five individual stations are mostly along Greenland s west coast or, for Narsarsuaq, in the extreme south. Danmarkshavn in the north east recorded its second warmest summer on record (Table II). On the other hand, Tasiilaq in south-east Greenland recorded only its 58th warmest summer (Table II); this is thought to be due to persistent northerly winds emanating almost directly from the relatively cold Summit (Figures 2 and 3). CGT2 and CGT3 were also highest or equal highest in their respective monthly mean SAT records for June and July 2012, when many west coast stations had either their warmest or second warmest summer on record (Table III and IV). However, note that Tasiilaq (SE Greenland) had its coldest June on record (in 115 years of data), although it recovered somewhat to fifteenth warmest in July 2012.

6 E. HANNA et al C at Ivittuut/Narsarsuaq on 29 May 2012 (Table V) was also a new record for the highest official SAT ever recorded in Greenland in May (DMI, 2012). Figure 4. Location map of Greenland climate stations and oceanographic profiles used in this study, superimposed on summer (JJA) 2012 near-surface temperature anomaly simulated by MAR relative to the period where MAR is forced by ERA-40 over and by ERA-Interim over Areas where temperature anomalies were at least twice the standard deviation are hatched. The inset panel shows the Greenland Blocking Index (GBI) area (60 80 N, W). Mean temperatures for CGT2 & 3 were respectively only sixth and twentieth highest (sixth and tenth highest) for May (August) 2012, and so these late spring and late summer months are not analysed in detail here. A breakdown of the temperature records between mean daily maximum and mean daily minimum reveals high/record 2012 values for both these sets of data, although note that CGT2 & 3 are highest in summer (JJA) 2012 for mean daily maximum temperature but only second highest for mean daily minimum temperature (Tables SI SVI). Also note from these tables that Tasiilaq had its lowest JJA mean daily minimum temperature, and ninth lowest JJA mean daily maximum temperature, in Ilulissat near Jakobshavn Isbrae on Greenland s west coast did not have exceptionally high 2012 SAT values as did the other west coast stations, as it rarely exceeded fifth highest for any of the measured temperature parameters or time periods (Tables II VII, SI SVI). The DMI station data and Figure 4 indicate that the unusual 2012 summer warmth may have been most extreme in the near-coastal areas of west and south Greenland, and also inland in the south, central west and north of the country. An analysis of extreme daily maximum temperatures shows new station records in 2012 for the respective months at Ivittuut/Narsarsuaq (24.8 C) in May, and Upernavik (20.3 C) and Ittoqqortoormiit (18.6 C) in July (Table V VII). The daily maximum temperature of 3.2. Interior Greenland temperatures Most of the longer-running (>5 10 years), easily accessible records from the Greenland ice sheet are from the Greenland Climate Network (GC-Net) of automatic weather stations (AWS) (Steffen and Box, 2001). The longest running GC-Net station is at Swiss Camp ( 1176-m elevation on the western central flank of the ice sheet; Figure 1 and Table I), for which we recently analysed a combined Swiss Camp research tower and AWS surface air temperature record, which dates almost continuously from May 1991 (Hanna et al., 2012b). Here we present an update of the Swiss Camp record, focusing on summer (JJA) and July averages. There are several other GC-Net surface air temperature records dating from the mid-late 1990s ( although most have some breaks in their records. The one of principal interest here is Summit ( 3200-m elevation), near the top/centre of the ice sheet (Figure 1 and Table I), which, as we have seen, experienced a significant melt event during 8 12 July 2012 (Nghiem et al., 2012), indicating surface air temperatures temporarily unusually for this high-elevation, inland, ice sheet site above freezing. Actually there are several different stations and sensors at Summit, run by both United States [University of Colorado/GC-Net and US National Oceanographic & Atmospheric Administration (NOAA)] and Danish (DMI) programs. In addition, the University of Wisconsin has data from May 1987 to July 1997 from this area but using an early passive temperature sensor shield design, essentially a short vertical tube with no side vents, now known to be prone to solar heating during low wind speed periods (Shuman et al., 2001; Genthon et al., 2011). The GC-Net Summit temperature record has data since May 1996 but with significant gaps in 1998 (DOY and ), 1999 (DOY 1 129), 2004 (DOY ), 2011 (DOY ) and 2012 (DOY 1 47 and 53 63). GC Net Summit temperatures are recorded by Type E thermocouples, with an instrument accuracy of 0.1 C (Box and Steffen, 2000), inserted in a Gill 7-plate radiation shield. Apart from the GC-Net site, the other long-running Summit station is operated by the DMI and has monthly data from 1997 present, also with some notable gaps, especially before 1998 (Vaarby Laursen, 2010). DMI use an Aanderaa 3444 platinum resistance thermometer PT100 in an Aanderaa radiation shield. As with many weather stations worldwide, neither the GC-Net nor DMI temperature sensors are actively ventilated, and may therefore sometimes be subject to artificial heating from solar radiation during periods with high solar input, especially when wind speeds are low as that reduces passive ventilation (Shuman et al., 2001; Genthon et al., 2011). Although in general this is a

7 CLIMATE FORCING OF 2012 GREENLAND ICE MELT Figure 5. Greenland (a) summer (JJA) and (b) July mean daily surface air temperatures for DMI coastal meteorological stations and their composites (CGTx). Table III. Five highest mean daily surface air temperature values and 2012 rank for DMI Greenland coastal stations and Greenland Temperature Composites for June. Year/temperature ( C) 2012 rank CGT2 2012/7.0, 2003&2010/6.8, 1997&2005/6.4, 2007&2008/6.0 1st CGT3 2003&2012/7.2, 2008&2010/7.1, 1948/6.8 1st= Upernavik 2008/6.7, 2012/6.5, 2003/5.7, 1931/5.4, 1990/5.1 2nd Aasiaat 2012/7.3, 2003/5.8, 1997/5.5, 2005/5.2, 2010/5.1 1st Ilulissat 1997/8.3, 1990/8.2, 2003 & 2005/7.8, 1931,1948&2012/7.5 5th Sisimiut 2012/8.1, 2003/6.6, 2010/6.4, 2004/6.3, 2002/6.1 1st Kangerlussuaq 1997/11.7, 2000&2012/11.6, 2003/11.5, 1998/11.1 2nd= Nuuk 2012/8.1, 1947/6.8, 1948&2010/6.7, 1924/6.6 1st Paamiut 1987/6.8, 2010/6.5, 2001&2005/6.1, 1991,1997&2004/5.7 9th= Ivittuut/Narsarsuaq 1929/11.4, 2012/11.3, 1991/11.2, 1924,1948,1987&2010/11.0 2nd Qaqortoq 1991/8.4, 2010/8.1, 1987&2005/7.6, 2001&2003/7.4 16th Danmarkshavn 2008/4.1, 2002/3.6, 2012/3.5, 1968/3.0, 1953/2.8 3rd Ittoqqortoormiit 1932/6.7, 1941/5.7, 1934&1938/5.6, 1948/5.2 40th Tasiilaq 1932/7.3, 1916&1947/7.1,1895/6.6, 2002&2010/ th = COLDEST Ikermiuarsuk 1961/5.9, 1958/4.5, 1960&2003/4.0, 1999&2008/3.9 48th (3rd coldest) Ikerasassuaq 2008/6.8, 1961/6.6, 2010/6.3, 2007/6.2, 2006/6.0 19th long-recognized meteorological problem, the severity of which may be greater than previously recognized in polar regions, due to ground-reflected solar radiation and inadequate radiation shields (Genthon et al., 2011), it is considered to be a short-term effect which occasionally biases temperature maxima during conditions of low wind speed and leaves monthly mean temperatures relatively unaffected (e.g. Hanna et al., 2012a). However, due to this factor, NOAA has been running aspirated (actively ventilated) temperature sensors at Summit since August 2005 ( NOAA has utilized two types of actively ventilated sensor shields (Met One and Cambridge) with both the Vaisala and Logan sensors running in Cambridge shields since 2008.

8 E. HANNA et al. Table IV. Five highest mean daily surface air temperature values and 2012 rank for DMI Greenland coastal stations and Greenland Temperature Composites for July. Year/temperature ( C) 2012 rank CGT2 2012/9.7, 2008&2011/8.9, 2007/8.8, 2009/8.6, 2005/8.5 1st CGT3 2012/9.9, 2008&2011/9.3, 1936,2007&2009/9.2 1st Upernavik 2011/9.6, 1960/8.8, 2007/8.7, 2012/8.5, 2009/8.4 4th Aasiaat 2012/9.1, 1960&2011/ /8.2, 2008/8.1 1st Ilulissat 1960/11.0, 2011/10.2, 1908/10.1, 1931/9.9, 2007/9.6 8th= Sisimiut 2012/10.4, 2008/9.9, 2011/9.1, 2009/9.0, 2010/8.7 1st Kangerlussuaq 1974/12.7, 2011/12.5, 2012/12.3, 1999/12.2, 1990/12.0 3rd Nuuk 2012/10.4, 2008/9.5, 1936/9.4, 1948/9.0, 1934/8.8 1st Paamiut 1958/8.0, 2005&2012/7.7, 2003/7.6, 2006/7.5 2nd= Ivittuut/Narsarsuaq 2012/13.0, 1917&1991/12.7, 2007&2009/12.3 1st Qaqortoq 2012/10.4, 2003/9.7, 1991/9.6, 2007/9.4, 2005/9.2 1st Danmarkshavn 1958/6.1, 1976/5.9, 2003/5.6, 2008/5.5, 2012/5.4 5th Ittoqqortoormiit 1936/9.4, 1949/8.9, 1947/8.2, 1948/7.9, 1932&1933/7.7 17th Aputiteeq 2009/4.6, 2005/4.1, 2008/3.7, 1980/3.4, 1991&2012/3.3 5th Tasiilaq 1929/9.2, 1950/9.0, 1939&2003/8.9, 2005/8.7, 1928&1936/8.6 15th Ikerasassuaq 2005/9.9, 1964/8.7, 2010&2012/8.2, 2008/8.1, 2007/8.0 3rd Table V. Five highest extreme daily maximum surface air temperature values and 2012 rank for DMI Greenland coastal stations for May. Year/temperature ( C) 2012 rank Upernavik 1946/ /15.0, 2012/14.5, 1956,1976&1998/14.0 3rd Ilulissat 2004/23.8, , 1954/17.1, 1946/17.0, 1995/ th Nuuk 1954/18.5, 2010/18.3, 2012/18.2, 1935/17.8, 1933/16.8 3rd Ivittuut/Narsarsuaq 2012/24.8, 1998/21.6, 2010/20.1, 1909&1991/20.0 1st Danmarkshavn 1972/11.8, 1987/9.3, 1966/8.4, 1965/8.3, 1992&2004/7.6 10th Ittoqqortoormiit 1947/11.9, 1941/11.8, 2012/9.9, 1969/9.7, 1965&1996/9.5 3rd Tasiilaq 1941/17.9, 1935/16.3, 1915/16.2, 1945/15.9, 1999/ rd Figure 6. Surface air temperature summer (JJA) and July means at GC- Net Swiss Camp, Greenland. Note cooling immediately following the Mt. Pinatubo volcanic eruption of Although these records are not sufficiently long-running by themselves to form a useful climatology, they are useful for calibrating and cross-checking the other temperature sensors at Summit. The above uncertainty is likely to impact mainly on the extreme maxima and short-term sub-daily values recorded during calm, sunny conditions in summer, so we mainly concentrate here on the monthly mean temperature data values and trends, and leave a more detailed analysis of the Summit temperature record for future work. This approach is justified by our analysis following Genthon et al. (2011) of all available GC-Net Summit summer (JJA) hourly 2-m wind speed and temperature data, where taking summer means of 2-m temperatures for various wind-speed thresholds between 0 and 5 ms 1 in 1 ms 1 increments, results in similar trends that agree to within 0.3 C for 3 C trends (Figure S4). Examining first the Swiss Camp GC-Net summer record, we found a mean temperature of 2.0 C for July 2012, which is marginally below the highest July mean temperature of 2.2 C in 2011, and marks a significant (at the 2σ level) 1.9 C underlying temperature trend for the 22-year record (Figure 6). The corresponding temperature trend for summer (JJA) in the same record is 2.6 C, a similarly (2σ ) significant rise. The Summit record analysis was slightly less straightforward due to gaps in both the GC-Net and DMI Summit station temperature records. Following advice from coauthor CAS, who has previously published key studies into the Summit temperature record (e.g. Shuman et al., 2001), and standard statistical methods, we chose to use the DMI Summit record as a baseline and fill gaps in the summer (JJA) series using splicing (regression fitting) with both the GC-Net Summit record (for 1996, 1997, 1999 and 2003) as well as, for just three years

9 CLIMATE FORCING OF 2012 GREENLAND ICE MELT Table VI. Five highest extreme daily maximum surface air temperature values and 2012 rank (1st = warmest) for DMI Greenland coastal stations for June. Year/temperature ( C) 2012 rank Upernavik 1957/18.2, 1959&2003/18.0, 2008/17.4, 1913/17.3, 1956/16.7 7th Ilulissat 2002/21.1, 1949/20.8, 1959/20.7, 1941/20.0, 1960/ th Nuuk 1950/23.0, 1947/22.0, 1892/21.8, 1939/21.4, 1941/ th Ivittuut/Narsarsuaq 1947/23.1, 1897/23.0, 2002/22.6, 2008&2012/22.5 4th= Danmarkshavn 2002/17.1, 1949/16.9, 1995/14.3, 1953/13.8, 2005/ th Ittoqqortoormiit 1996/17.7, 2007/17.2, 1995/16.3, 1963/16.0, 1934&2002/15.8 7th Tasiilaq 1916&1942/25.3, 1958/21.5, 1944/21.0, 1991/ th = 4th coldest Table VII. Five highest extreme daily maximum surface air temperature values and 2012 rank (1st = warmest) for DMI Greenland coastal stations and Greenland Temperature Composites for July. Year/temperature ( C) 2012 rank Upernavik 2012/20.3, 1908/19.8, 1939/19.0, 1924&1940/18.5 1st Ilulissat 1908/21.9, 2012/21.1, 1977&2011/20.6, 1948/20.4 2nd Nuuk 1908/24.2, 1900&1917/21.2, 2008/21.0, 1930/20.4 6th Ivittuut/Narsarsuaq 1976/24.0, 2012/23.6, 1918/23.4, 2008/22.9, 1950/22.4 2nd Danmarkshavn 2002/19.7, 2005/17.6, 2012/17.5, 1990/16.4, 1988/15.5 3rd Ittoqqortoormiit 2012/18.6, 2006/18.5, 1931&2005/17.9, 1995/17.5 1st Tasiilaq 2005/25.3, 1995/25.2, 1902/24.8, 1963/23.5, 1915/ nd (1994, 1995 and 1998), our previously published Summit record monthly means derived in part using older University of Wisconsin data (Shuman et al., 2001; Hanna et al., 2008). This seemed a reasonable approach given the high correlations (r ) of all three Summit summer temperature series [i.e. the Hanna et al. (2008) Summit series and our updated, independent DMI and GC-Net series] for the common overlap periods. We used monthly means provided by DMI, having crosschecked these against the 3-hourly, instantaneous value DMI Summit data, but excluded several suspect high 3-hourly temperatures above 0 C on the basis that they were 3 C above the highest GC-Net hourly mean temperatures recorded on the same day, thus we made small ( C) adjustments to five DMI Summit monthly mean temperature values (Table VIII). Similarly, a few spurious GC-Net Summit hourly temperatures above 5 C were very likely to be erroneous, so were excluded when calculating monthly means. Having performed this simple post-processing, we obtained a complete reanalysed Summit summer (JJA) surface air temperature series spanning (Figure 7). This exhibits an underlying linear trend of 2.2 (1.8) C for JJA ( ). The corresponding Summit temperature trend for the July monthly series is 3.1 C. Our reanalysed average Summit temperature for the 2012 summer (JJA) is 11.4 C, which is a new high record in that series (slightly surpassing the previous record of 11.5 Cset in 2010) and 1.8σ above the mean summer temperature of 13.5 C for Similarly, the July 2012 mean of 9.2 C is a record in the July Summit temperature series (previous record of 9.5 C in 2005) andis1.5σ above the mean July temperature of 11.5 C for These analysis results suggest that 2012 was an unusually warm July and summer at Summit in Figure 7. Surface air temperature summer (JJA) means at Summit, Greenland: DMI, GC-Net, and our new reanalysed Summit record. this (albeit still relatively short) composite instrumental record. Turning briefly to the relatively high Summit temperatures associated with the major Greenland melt in summer 2012, a few values between 5 13 C recorded by GC-Net Summit were discarded in the monthly mean calculations because they are clearly erroneous but several values between 2 5 C remain. These latter cannot definitely be discarded at this time but have minimal influence at the monthly averaging timescale due to the large number of values recorded. We made an intercomparison of extreme high temperatures recorded at the DMI and GC-Net Summit stations. The highest reasonably reliable temperature in the WMO DMI Summit 2-m 3-hourly record (there are no available standard maximum and minimum values) is 2.2 C at 15 UTC on 11 July 2012, i.e. setting a new temperature record during the main 2012 melt event

10 E. HANNA et al. Table VIII. Extreme high 3-h instantaneous surface air temperature summer (JJA) data at DMI Summit and highest GC-Net Summit hourly mean temperatures recorded on the same days, with summary of impact on DMI Summit monthly mean temperatures for relevant months. Date/hour DMI Summit temperature GC-Net Summit TC highest temperature on same day Flag DMI value as suspect? Effect on DMI Summit mean temperature for month 3 Jun. 2004/15z Yes Changed from 14.0 to 14.1 C 9 Jun. 2003/15z Yes Changed from 13.5 to 13.7 C 7 Jun. 2003/12z Yes Ditto 4 Jun. 2012/12z Yes Changed from 10.3 to 10.4 C 11 Jul. 2012/15z No None 13 Jul. 2012/15z No None 16 Jul. 2012/18z No None 11 Jul. 2012/18z No None 23 Jun. 2008, 9z Yes Changed from 13.3 to 13.6 C 14 Jul. 2012/12z No None 14 Jul. 2012/18z No None 26 Jun. 2007, 21z No None 6 Aug. 2011, 15z Yes Changed from 14.8 to 14.7 C 14 Jul. 2012, 21z No None 4 Jul. 1998, 21z 0.8 N/A No None 24 Jul. 1998, 0z 0.8 N/A No None 11 Jul. 2012, 12z No None 2 Aug. 2011, 18z No None 12 Jul. 2012, 12z No None 13 Jul. 2012, 18z No None 13 Jul. 2012, 21z No None 27 Jun. 2005, 21z Yes No effect on monthly mean 14 Jul. 2012, 15z No None 30 Jun. 2003,12z Yes Changed from 13.5 to 13.7 C 29 Jul. 2012, 15z No None 9 Jul. 2004, 16z No None 10 Jul. 2004, 20z No None All temperatures are in degrees Celsius and were recorded at a nominal 2 m height above ground. (Table VIII). It is unusual but not unprecedented, even in the relatively short instrumental record to have surface air temperatures rising above freezing at Summit, putting the above record into context, this happened on only 27 occasions out of readings in the DMI Summit record but as we have seen a number of these values are well out of synch with the nearby GC-Net thermometers, so can be discounted as spurious (Table VIII). There are only five other reliable temperature readings >1.0 C at DMI Summit (as judged through comparison with independent GC-Net Summit readings), and these all occurred between July Seven of the other twelve above-freezing readings ( C) at DMI Summit were also recorded in July 2012, including a value of 0.2 C at 15z on 29 July 2012 (Table VIII), around the time of the other large-scale Greenland melt event noted by passive microwave data as mentioned in the preceding text. The peak 29 July 2012 DMI Summit temperature of 0.2 C agrees well with the NOAA Logan sensor which shows temperatures briefly reaching 0 C on the same date at its nominal 2 m height above the snowpack. Similarly, the NOAA sensor did not record above 1 C during the main melt event on 11 July. This brief comparison suggests that some of the extreme high DMI (and GC-Net) Summit temperatures may be biased C high due to passive solar heating of nonventilated thermometer screens. Of course both the recent 2012 and pre-instrumental [i.e. as measured by ice cores (Alley and Anandakrishnan, 1995), although note the 1995 event observed by passive microwave satellite] extreme Greenland surface melt events may result from warm air masses briefly transiting over the ice sheet. That is due to warmth for a few days, rather than a sustained warm period of at least several weeks. In such cases, spring and/or summer temperatures could be relatively normal for those isolated melt event years. Here we have presented clear evidence that with the 2012 extreme melt event the central Greenland warming was unusual at both the short (daily) and monthly/seasonal timescales; however, further reanalysis of past instrumental and satellite data is needed to put the 2012 extreme warm/melt event more firmly in a longerterm historical context. 4. Atmospheric forcing factors including high geopotential heights and blocking high pressure The monthly 500 hpa geopotential height (Z500) and anomaly hemispheric charts from the US Climate Prediction Center (Figure 8) show an interesting evolution of

11 CLIMATE FORCING OF 2012 GREENLAND ICE MELT Figure 8. Monthly 500 hpa height (black lines annotated in dm) and anomaly (blue-red colours with scalebar) hemispheric charts for March August 2012 (a f), from the Climate Prediction Center. Anomalies are with respect to the atmospheric flow over Greenland from March through to August, A four-wave polar jet stream pattern is evident for March, which slowly develops to five-wave pattern by June, when the positive geopotential height anomaly over Greenland was at its strongest. According to Rossby wave theory (see discussion in Section 3 of Hanna et al., 2012b), the five-wave pattern is more prone to be stationary. The positive anomaly over Greenland in June 2012 was remarkable due to the very strong flow around the upwind trough. The amplitude of the trough increases in July and maintains a southerly component to the wind over Greenland (Figure 2). The associated high geopotential height (high pressure) anomaly over Greenland during summer (JJA) 2012 is clearly greater than similar anomalies during the previous warm/relatively high (but far from near-total) surface melt summers of (Figure S5, Supplementary Materials), and in fact is the highest such anomaly in the NCEP/NCAR

12 E. HANNA et al. frequency of anti cyclonic conditions in southern Greenland, which pushed cloudiness and precipitation further north over Greenland (Figure S3). Figure hpa geopotential height summer (JJA) values , based on NCEP/NCAR Reanalysis data for the Greenland Blocking Index (GBI) region (Hanna et al., 2012a) of N, W (see inset in Figure 4 map). NAO index (NAOI) JJA values (Hurrell et al., 2012) are also plotted on a reversed scale for comparison. Reanalysis record dating back to 1948 (Figure 9). In addition, the JJA 2012 Z500 anomaly over Greenland is the highest across the Northern Hemisphere (Figure 2), indicating that it was local in nature and may well originate from regional NAO forcing, although this Greenland blocking might also be causing the highly negative NAOtype pattern (Woollings et al., 2010). Geopotential height anomalies over Greenland are numerically classified by their intensity and coverage over the area defined in Figure 4, using a measure known as the Greenland Blocking Index (GBI) (Fang 2004; Hanna et al., 2012b). The more local geographic nature of the GBI means that it correlates more directly than the NAO index with GrIS runoff changes (Hanna et al., 2012b). Until now the GBI has been relatively little used, even by Greenland climate researchers, but we anticipate that it may gain wider popularity given the recent enhanced interest in GrIS melting, mass balance changes, and feedbacks of both the ice-sheet changes and Greenland regional atmospheric circulation patterns with the wider climate system. NAO JJA values are also plotted in Figure 9 for comparison, showing the expected strong anti-correlation with GBI and a strong run of negative NAO summers since 2007; 2012 has the fourth lowest value in the Hurrell et al. (2012) principal-component summer (JJA) NAO record since 1948 (the start of the NCEP/NCAR Reanalysis record, for which we analyse GBI), and five of the six lowest NAOI summers of the last 65 years occurred since Another contributory factor to the exceptionally high Greenland melt of July 2012 may be lack of cloud cover and/or resulting changes in cloud/radiation interaction (e.g. Bennartz et al., 2013), which appears to have augmented the warm winds in melting surface ice around Summit; however, reliable cloud cover data are significantly lacking for the ice sheet, although this aspect certainly merits further investigation. Validated MAR RCM results clearly show the signature of the abnormal 5. Oceanic forcing factors and sea-ice extent Temporal changes in water temperatures are considerably slower than air temperature changes due to the much higher heat capacity and slower velocities of sea water masses. Here we focus on the ocean pre-conditions to the abnormally high Greenland air temperatures observed during summer Temperature and salinity measurements off West Greenland have been carried out for the last decades by the Danish Meteorological Institute on behalf of Greenland Institute of Natural Resources. Six sections in total are maintained from Cape Farewell to Sisimiut, with five stations on each, and they are measured in June/July each year, weather and ice conditions permitting. These measurements represent an annual snapshot from up to 30 stations distributed off Southwest Greenland (Figure 4 and Ribergaard, 2012) In general the coldest water masses with the lowest salinities are found on the Greenland continental shelf, and comprise Polar Water mixed with runoff waters from land and partly with warmer Irminger Water from below. Below, at intermediate depths ( m, shallowest in the south) and west of the shelf, is found warmer and saline Irminger Water, which is transported to the area via the Irminger Current a side branch of the North Atlantic Current. This water mass is responsible for the major oceanic heat transport towards southeast and west Greenland waters. Thus changes in the oceanographic conditions off southwest Greenland are likely to be representative of major changes in large part of the Greenland waters. To obtain a robust time series of the general sea surface temperature in Southwest Greenland waters, we calculated a normalized index for salinity and temperature for the upper 40 m of water. These indices were formed by subtracting the long term mean and dividing by the standard deviation for each of the 30 stations. Thereby we were able to combine all the stations in two single indices one for salinity and one for temperature. By using all the stations together, we reduce the influence of individual eddies and frontal movement over the continental shelf, which can alter the water properties quite significantly for individual stations. The temperature and salinity indices were both slightly negative in 2012, indicating slightly lower sea surface temperatures and salinities in the upper 40 m than normal for the past 20 years (Figure 10). This result suggests no major changes in oceanographic conditions in 2012 compared with previous years. Satellite-based SST anomaly plots show values 0 2 C above average around Greenland, with the warmest anomalies offshore to the south and especially south-west of southern Greenland (Figure S6). However,

13 CLIMATE FORCING OF 2012 GREENLAND ICE MELT Figure 10. Temperature (top) and Salinity (bottom) index for the upper 40 m water column for Sisimiut, Maniitsoq, Fylla Bank, Paamiut, Cape Desolation and Cape Farewell sections during the period Bold black lines show the mean of all five individual stations on each of the six sections, and the thin grey lines represent the 30 individual stations. there was a rapidly evolving situation where SSTs around Greenland were generally lower than, or around, normal in May and June, becoming progressively warmer in July and August 2012 (Figure S7). This most likely represents a response of the near-surface ocean layer a thin surface skin in the satellite data to the much warmer than normal atmospheric conditions in June and July. This conclusion is supported by the about normal 0 40 m temperature anomalies observed in June for SW Greenland (see above), and by June 2012 vertical temperature profiles in the top 100 m of ocean (not shown here), which show a clear 2 C temperature gradient in the upper 50 m, with the highest temperatures at the surface. This gradient shows that the atmosphere was heating the sea surface. Arctic sea-ice extent reached a new record low in late summer/early autumn, 2012 (NSIDC, 2012). During much of 2012, Northern Hemisphere sea ice extent was lower than the median but similar to the previous seven years until June, from when the major acceleration of the retreat took place. The only exception was within the Barents Sea north of Norway, where the sea-ice extent was lower than normal since winter, which may indicate a continuation of the increased influence of the relative warm Atlantic water masses during the past fifteen years (e.g. Holliday et al., 2008). On the other hand, the sea ice extent in Davis Strait was further south than normal during March and close to normal conditions for spring as a whole. Combined with slightly lower than average SSTs (0 40 m) off West Greenland in June 2012, this suggests that the main cause of the major warming event over Greenland in summer 2012 was a sudden and abnormal change in atmospheric circulation, as confirmed by the MAR sensitivity experiment (see Section 2). 6. Discussion: what does the record 2012 melt event and recent climate variability tell us about future climate change in Greenland? The JJA 2012 T600 anomaly in the NCEP-NCAR reanalysis was the highest since 1948 (not shown here) and was higher than the most pessimistic Coupled Model Intercomparison Project Phase 5 (CMIP5)-based future projections (10-year averages) over (Figure 11). According to the CMIP5 ensemble mean (including 30 GCMs), the reanalysis-based T600 anomaly in 2012 should be common by the end of this century for the RCP 4.5 scenario and in the middle of this century for the RCP 8.5 scenario. Mean probabilities for future decades of having JJA T600/runoff anomalies as observed in 2012 are shown in Table IX; such runoff anomalies as simulated in 2012 by MAR are not projected to occur before 2040 in most of the CMIP5 models. The CMIP5- based future projections fail to reproduce the observed GrIS melt increase since 2000 because this mainly resulted from NAO and GBI anomalies in the general circulation (Hanna et al., 2012b; Fettweis et al., 2013a), which are not projected over this century by any CMIP5 model (Belleflamme et al., 2012; Fettweis et al., 2013a). Moreover, recent summers ( ) have been drier than normal in Greenland as a result of more frequent anti cyclonic conditions than normal (Tedesco et al., 2011; Box et al., 2012), which has enhanced the surface iceand snow-melt arising from a warmer atmosphere. On the other hand, GCM-based future simulations project a small increase of snowfall for temperature increases lower than 3 C, which (should this occur) would tend to dampen the increase of melt thanks to the albedo-snowfall feedback (Box et al., 2012; Franco et al., 2013). Therefore, this suggests that summer 2012 is not really representative of the future climates projected by the CMIP5 GCMs

14 E. HANNA et al. Table IX. Mean probabilities from 30 CMIP5 GCMs of having the summer (JJA) T600 mid-tropospheric air temperature and runoff of We refer to Moss et al. (2010) for details of the Representative Concentration Pathways (RCP) scenarios. Period Probability of having the JJA T600 anomaly of 2012 Probability of having the JJA runoff anomaly of 2012 RCP4.5 RCP8.5 RCP4.5 RCP % 1% 0% 0% % 11% 1% 1% % 51% 7% 21% % 91% 15% 64% % 99% 21% 92% Figure 11. (a) Anomaly of JJA temperature at 600 hpa (T600) simulated by the NCEP-NCAR reanalyses and by the CMIP5 GCMs with respect to over Greenland for the RCP 4.5 and RCP 8.5 scenarios. We refer to Moss et al. (2010) for more details about the Representative Concentration Pathways (RCP) scenarios. The T600 anomaly is taken over an area covering Greenland (60 85 N and W). The ensemble means as well as the standard deviations of the CMIP5 GCMs are plotted in colour. A 10-year running mean was used to smooth the curves. (b) Same as (a) but for the GrIS meltwater runoff anomaly as simulated by MAR forced by the ERA-INTERIM and derived from the CMIP5 GCMs following the Fettweis et al. (2013b) estimation. We refer to Fettweis et al. (2013b) for more details about these model estimates and the 30 CMIP5 GCM they used. and may more likely result from natural variability of the NAO/GBI. However, such atmospheric circulation changes may be anthropogenically forced (Francis and Vavrus 2012; Overland et al., 2012). The Z500/T600 anomalies over GrIS during summer 2012 were the highest in the Northern Hemisphere, which suggests they are local (linked with negative NAO anomalies) and not global in origin (see Figure 2). 7. Conclusions There were some exceptionally high/record mean and extreme daily maximum surface air temperatures along the western side, and in the extreme south of, Greenland, linked with unusually high geopotential heights and atmospheric pressure anomalies over the ice-sheet surface, during May July, The unusual atmospheric circulation pattern was gauged by negative NAO and exceptionally high GBI index values. Sea-surfacetemperature anomalies close to the Greenland coast, although above average, were not exceptionally high in May/June 2012, and temperature and salinity profiles for the top 40 m off south-west Greenland even showed slightly negative anomalies, suggesting the sea-surface warming was mainly due to atmospheric heating rather than oceanic heat advection. An RCM-based sensitivity study shows little effect of changing sea-surfacetemperature and sea-ice-cover anomalies on GrIS nearsurface temperature and meltwater runoff, in line with previous findings. Taken together, our present results strongly suggest that the main forcing of the extreme GrIS surface melt in July 2012 was atmospheric, linked with changes in the summer NAO, GBI and polar jet stream which favoured southerly warm air advection along the western coast. The next 5 10 years will reveal whether or not 2012 was a one off/rare event resulting from the natural variability of the NAO or part of an emerging pattern of new extreme high melt years. Because such atmospheric, and resulting GrIS surface climate, changes are not well projected by the current generation of GCMs, it is currently very hard to predict future changes in Greenland climate. Yet it is crucial to understand such changes much better if we are to have any hope of reliably predicting future changes in GrIS mass balance, which is likely to be a (if not the) dominant contributor to global sea-level change over the next years (Goelzer et al., 2012). Acknowledgements NASA MEaSUREs program supported the passive microwave surface melt product produced at the University of Georgia. The SnowModel 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, and LANL Institute for Geophysics and Planetary Physics. Thanks to the Program for Monitoring