Greenland Ice Sheet surface melt: A review

Kang Yang , ManChun Li

1. Department of Geographic Information Science, Nanjing University, Nanjing, Jiangsu 210093, China

2. Department of Geography, University of California, Los Angeles, California 90095, USA

1 Introduction

The Greenland ice sheet (GrIS) is the largest freshwater reservoir in the Northern Hemisphere. The GrIS can make the global sea level rise by 7 m if it melts completely(Gregory and Huybrechts, 2006). The GrIS is losing mass due to a combination of surface meltwater runoff and iceberg calving (Rignot and Thomas, 2002; Dowdeswell, 2006;Vaughan and Arthern, 2007; Nicholls and Cazenave, 2010).Estimates of the former are large and have been calculated to contribute more than 40% of total GrIS mass loss from 2000–2008 (Van den Broekeet al., 2009). However, iceberg calving and its impact on the GrIS mass loss have also been researched in the past few decades (Howatet al., 2005;Joughinet al., 2008b; Thomaset al., 2009). At present,growing attention has been paid to the study of runoff-induced mass loss (Alleyet al., 2008). For instance, the impacts of GrIS surface melt and runoff have been integrated into GrIS mass balance modeling (Mernildet al., 2009), and different global warming scenarios have been employed to simulate the impacts induced by surface melt (Parizek and Alley, 2004). An improved understanding of the GrIS surface melt is critically important for assessing its impact on current and future ice sheet dynamics and global sea level rise(Rennermalmet al., 2013).

Typically, the GrIS surface melt occurs in the following sequence: solar radiation warms the ice sheet and meltwater forms if the temperature is above 0 °C. The meltwater has much lower reflectance than snow or ice, and thus accelerates the melting of its surrounding snow and ice, leading to the formation of a positive feedback loop. Then meltwater is transported on the ice sheet surface mainly under the control of the surface topography, forming a complex supraglacial hydrological system that includes supraglacial lakes, supraglacial streams, crevasses, and moulins. This system facilitates the ice sheet mass loss in two ways (Figure 1): First, some meltwater is transported directly to the ice sheet margin and then to the ocean (Pfefferet al., 1991; Janssens and Huybrechts, 2000); and second, some surface meltwater is transported into the ice sheet through crevasses or moulins,increasing basal sliding and thereby accelerating ice flow(Zwallyet al., 2002; Van de Walet al., 2008).

Figure 1 Typical pattern of the Greenland ice sheet surface melt

However, these processes show a typical pattern of the GrIS surface melt, and the actual situations may work differently. Nevertheless, all the potential processes are strongly connected to the GrIS supraglacial hydrology (Irvine-Fynnet al., 2011; Rennermalmet al., 2013). This has motivated studies of the production, transport, and release processes of surface melt water (Irvine-Fynnet al., 2011). These hydrological processes are the essential concerns of the GrIS surface melt research. It is particularly urgent to address three scientific questions: How much meltwater is produced on the ice sheet?What are the characteristics of the supraglacial hydrology system? How does surface meltwater impact ice sheet motion?

This paper summarizes the current understanding of GrIS surface melt research from the perspective of these three essential supraglacial hydrological processes, emphasizing the most recent findings and highlighting the remaining gaps in knowledge.

2 How much meltwater is produced on the ice sheet?

Large amounts of meltwater are produced on the GrIS each melt season (Ettemaet al., 2010; Liston and Mernild,2012; Mernild and Liston, 2012). To better understand the impacts of meltwater on the ice sheet, the first step is to quantify the surface meltwater volume. Surface melt models are the most commonly used approaches to quantify the meltwater volume (Quincey and Luckman, 2009). Because the meltwater volume shows good positive correlations with certain supraglacial hydrological features, it is possible to derive the meltwater volume information through monitoring of these supraglacial features, such as supraglacial lakes(McMillanet al., 2007).

2.1 Ice sheet surface melt modeling

Surface melt models seek the relationships between surface melt and hydrometeorological conditions to acquire the melt rate and, consequently, the melt volume (Quincey and Luckman, 2009). The two main types of melt models are degree-day models and energy-balance models (Hock, 2005).

Degree-day models, or temperature-index models,briefly describe the ice sheet surface melt, mainly by focusing on building empirical relationships between the near-ground temperature and the surface melt status (Ohmura, 2001; Cuiet al., 2010; Qiaoet al., 2010; Wuet al., 2010). These models are widely used due to their reasonable theoretical assumptions and simple model inputs (e.g., measured temperature)(Reeh, 1994; Hock, 2005). However, these models simplify the ice sheet surface melt to a temperature-controlled process,and only acquire the modeled melt results near the meteorological monitoring stations (Hock, 2005). Therefore, many modified models have been proposed. For instance, in addition to temperature, the overall ablation and accumulation situations of the ice sheet surface are considered as melt controlling-factors (Tedescoet al., 2011). Furthermore, the refreezing factor is also employed to improve these degree-day models (Janssens and Huybrechts, 2000).

Energy-balance models aim to simulate the physical process of ice sheet surface melt. A surface melt equation is the key component of these models, which is acquired by determining the energy input and output parameters (Sunet al., 2011). These energy-balance models are commonly more complex than degree-day models and thus can simulate melt status more accurately (Hock, 2005). For instance, degree-day models can only model the daily melt situation whereas energy-balance models can model the melt variations throughout a day. Furthermore, distributed energy-balance models can demonstrate the broad spatial dynamics of surface melt, whereas degree-day models can only represent the melt status near the meteorological monitoring stations (Sunet al., 2011). A main disadvantage of energy-balance models lies in their strict requirements for the input parameters, some of which are difficult to acquire.Therefore, simplified energy-balance equations are usually employed instead. For instance, McGrathet al. (2011)proposed a simplified energy balance model for a moulin-controlled watershed to simulate the surface runoff in the western part of the GrIS.

Energy-balance models can simulate surface melt in large spatial and temporal ranges (Hock, 2005). However,these models only focus on the possible effects of energy transfer on surface melt; they lack an understanding of refreezing and englacial hydrology processes. There is growing consideration of these factors in energy balance modeling. Moreover, both energy-balance models and degree-day models consider solar radiation as the only control factor of the GrIS surface melt, but other factors can also impact the GrIS surface melt. For instance, the relationships between the extent of Arctic sea ice and the melt area on the ice sheet have been analyzed, showing that the shrinking of sea ice extent may accelerate the GrIS surface melt(Rennermalmet al., 2009).

2.2 Supraglacial lake extraction and analysis

Supraglacial lakes are important hydrological features on the GrIS (McMillanet al., 2007). These lakes can temporarily store large amounts of meltwater during melt seasons (Sneed and Hamilton, 2007; Leesonet al., 2012). To investigate the meltwater storage capability of these lakes, the first step is lake surface area extraction through remotely sensed imagery.

The two main approaches to extracting supraglacial lakes are manual digitization and automatic extraction (McMillanet al., 2007). The latter includes image band ratio and classification methods, both employing the spectral differences between water and ice/snow. A thresholding method has been developed using the band ratio between the red and blue bands of MODIS imagery (Box and Ski, 2007). An improved thresholding method was then developed based on adaptive searching windows (Selmeset al., 2011). Fuzzy logic membership functions are employed to classify and extract supraglacial lakes from MODIS red band imagery(Sundalet al., 2009). MODIS has a short revisiting time and thus can illustrate the dynamics of supraglacial lakes effectively (Lianget al., 2012). However, the supraglacial lakes extracted from MODIS images are in coarse spatial resolution,requiring accuracy validation. ASTER images are commonly used as a validation data source (Sundalet al., 2009). Supraglacial lakes show highly dynamic variations (e.g., a lake drained completely in 2 h) (Daset al., 2008), and therefore it is important to delineate and extract drained lake information.

The number, area, and volume variations of supraglacial lakes can indicate the ice sheet melt status (McMillanet al.,2007; Sundalet al., 2009). The shape characteristics of supraglacial lakes have been studied, representing that the diameters of the supraglacial lakes in southeastern Greenland vary from several hundred meters to 2 km (Box and Ski,2007). Selmeset al. (2011) extracted the supraglacial lakes from all regions of the GrIS by using 3,704 MODIS images for the period 2005–2009, indicating that there are 2,038 lakes formed each year, 55% of which are distributed in the southeastern part of the GrIS, an area rich in supraglacial hydrology features. McMillanet al. (2007) analyzed the supraglacial lake variations in southeastern Greenland using 12 Landsat ETM+ and ASTER images, indicating a positive relationship between the lake numbers and the degree-day indicators. Moreover, Sundalet al. (2009) found a significantly positive relationship between supraglacial lake areas and the derived modeled surface runoff by using 260 MODIS images between 2003 and 2007.

2.3 Supraglacial lake depth and volume estimation

Depth is an important indicator to describe a supraglacial lake shape. Lake volume can be calculated from a combination of lake area and depth (Box and Ski, 2007; McMillanet al., 2007; Sneed and Hamilton, 2007). There are three main approaches to derive lake depth, namely, remote sensing estimation, field measurement, and DEM modeling.

1) Remote sensing estimation. This method uses the Bouguet-Lambert-Beer law or empirical formulae to calculate depth straightforwardly (McMillanet al., 2007). The Bouguet-Lambert-Beer law relates the absorption of light to the properties of the material through which the light is travelling. This estimation method relies on reasonable assumptions about the albedo of the bottom surface of lakes and the optical attenuation characteristics of the meltwater, and it shows relatively high accuracy (Sneed and Hamilton, 2007).However, this method has strict requirements for the input data: the optical images need to be atmospherically corrected and the reflectance for optically deep water should be provided (Box and Ski, 2007). This method is first employed to derive supraglacial lake depths from ASTER images (Sneed and Hamilton, 2007). The root-mean-square departure of the derived lake depths is approximately 0.3 m during cloud-free conditions for ASTER images (Georgiouet al.,2009). The capabilities of MODIS and Landsat ETM+ images to derive supraglacial lake depths have also been vali-dated. Aside from the depth estimations based on the Bouguet-Lambert-Beer law, an empirical approach has also been proposed by using the empirical relationship between the field-measured depth and the MODIS red band DN values (Box and Ski, 2007). WorldView-1 and WorldView-2 or the other high-spatial-resolution images provide great potential to estimate supraglacial lake depth (Rennermalmet al.,2013). However, small spatial coverage and the omission of deep ocean pixels limit its applicability.

2) Field measurement. This method is the most accurate but also the most expensive depth estimation approach. For instance, a depth detector was used to measure two typical supraglacial lake depths in the southeastern GrIS, achieving an accuracy of 0.1 m (Box and Ski, 2007). A remotely controlled boat with a GPS, a sonar, and a spectrometer was employed to measure the depth of Olivia Lake in southeastern Greenland, quantifying an approximately 15% depth difference between the in-situ measurements and the satellite-estimated lake depth values (Tedesco and Steiner, 2011).

3) DEM modeling. This method uses the volume of a topographical depression and a lake area mask to estimate the depth values. For instance, several LiDAR datasets and a degree-day model were used to acquire the empty base topography (McMillanet al., 2007). This approach requires high-spatial-resolution DEMs, and LiDAR-derived DEM is by far the only appropriate choice (Adler, 2010). However,the significant developments of WorldView-1 and WorldView-2 and other high-spatial-resolution sensors with stereo images have great potential in DEM modeling, and thus lake depth derivation.

Among the above three depth-derivation methods, remote sensing estimation is the only approach that can acquire depth measurements in wide areas. Therefore, it is crucial to validate or verify remote-sensing-derived results with field measurements or DEM modeling results (Rennermalmet al.,2013). Supraglacial lake volume is a better surface melt indicator than lake area. A positive correlation between supraglacial lake volume and surface runoff was obtained with LiDAR and GDEM data to calculate supraglacial lake volume, and dH/dtwas used to simulate the surface runoff(Adler, 2010).

3 What are the characteristics of a supraglacial hydrological system?

The GrIS supraglacial hydrological system is extraordinarily complex, consisting of a largely ephemeral patchwork of supraglacial lakes, streams, crevasses, and moulins.Growing attention has been paid to the potential links between supraglacial hydrology systems and the GrIS mass balance (Irvine-Fynnet al., 2011). Supraglacial hydrological features are widely distributed in the western GrIS, especially in the Kangerlussuaq area of southwestern Greenland(Selmeset al., 2011). Consequently, this region has been the subject of intensive recent research where surface melting and ice sheet motion have been well measured (Hannaet al.,2002; Mernildet al., 2010; Mernildet al., 2011). Supraglacial hydrological systems transport and release large amounts of meltwater and thus have great impacts on the GrIS mass balance, although comprehensive study of their formation and spatial and temporal dynamics is still lacking(Rennermalmet al., 2013).

3.1 Supraglacial hydrology system formation

During melt seasons, supraglacial streams form if the incising rate exceeds the melt rate (Nolin and Payne, 2007).Then the meltwater is transported mainly under the control of topography and either forms supraglacial lakes in topographical depressions (Box and Ski, 2007; McMillanet al.,2007; Sundalet al., 2009; Selmeset al., 2011; Tedesco and Steiner, 2011) or flows into the ice sheet interior through moulins (Alleyet al., 2005; Catania and Neumann, 2010;Phillipset al., 2011). Also, some meltwater will directly flow into the ocean. Meltwater transport forms hundreds of watersheds on the GrIS (Hardyet al., 2000; Lewis and Smith,2009). Two types of watersheds are found on the ice sheet,which are mainly controlled by supraglacial lakes or moulins,respectively: the supraglacial lake watersheds can host meltwater, while the moulin watersheds transfer meltwater into the ice sheet directly (McGrathet al., 2011). However, the fast drainage of supraglacial lakes indicates that moulins can form at the bottom surface of these lakes, which can also transfer meltwater into the ice sheet (Krawczynskiet al., 2009; Selmeset al., 2011). Furthermore, two meltwater transport pathways exist on the GrIS, which are controlled by supraglacial streams or crevasses, respectively (Colganet al., 2011).

In summary, supraglacial hydrological features are interconnected, impacting both the meltwater transport and release (Rennermalmet al., 2013). Therefore, the key aim of supraglacial hydrological system analysis is to illustrate the characteristics and interactions of supraglacial hydrological features.

3.2 Surface meltwater transport

Supraglacial streams start with small rills at the beginning of the melt season (Knighton, 1981). Rills combine into channels and thus form a network of supraglacial streams(Hambrey, 1977; Marston, 1983; Knighton, 1985). Most of the supraglacial streams are undistinguishable in coarse- or medium-resolution remotely sensed imagery, except for a few wide streams that link supraglacial lakes (Yang and Smith, 2013). Moreover, the network of supraglacial streams is complex, making it difficult to conduct field measurements.Therefore, the exploration of supraglacial streams is just beginning (Rennermalmet al., 2013), and mapping of supraglacial streams using high-spatial-resolution remote sensing images is the first step to study these features. McGrathet al. (2011) manually delineated the supraglacial streams in a small watershed by using a WorldView-1 image, indicating that the supraglacial streams tended to reform in the same locations. However, an automatic method using both spectral and shape information needs to be developed, due to the large number of images required to study hydrologic processes with high spatial resolution across large areas of the ablation zone (Yang and Smith, 2013).

Crevasses are fractures formed from tension, and their patterns are controlled by the directions of the principal stresses and openings in the direction of maximum tension(Colganet al., 2011; Lampkin, 2011). Crevasses are the terminations of surface meltwater transport, transferring some meltwater into the ice sheet. However, crevasse-type drainage is less efficient than moulin-type drainage in transferring meltwater (Colganet al., 2011). In a study of the Sermeq Avannarleq ablation zone in western Greenland (Phillipset al., 2011), the crevasse distribution area was found to increase 9%–17% during a 25-year time period (1985–2009),showing the process for meltwater flowing into the ice sheet through crevasses. In western Greenland, the spatial distributions of the crevasse- and the moulin-type drainage basins are believed to be independent.

3.3 Surface meltwater release

Supraglacial lakes can host significant amounts of meltwater and are temporary meltwater storage sites during the melt seasons. However, the widespread existence of fast-drained lakes indicates that some of the lakes can release meltwater directly into the ice sheet (Selmeset al., 2011).Therefore, the spatial distribution and dynamics of supraglacial lakes are very important to reveal the process of meltwater release on the ice sheet (Hoffmanet al., 2011; Lampkin and VanderBerg, 2011). The spatial distribution of supraglacial lakes is mainly controlled by topography, latitude, and altitude. McMillanet al. (2007) studied lake areas at different altitudes and found that lakes distributed at higher altitudes represented larger area variations. The dynamics of lake areas over time indicate that low-elevation lakes are the first to form and the first to disappear (Sundalet al., 2009). The potential spatial distributions of supraglacial lakes were modeled using a surface melt model, demonstrating that the distributions of supraglacial lakes are independent of the surface melt rate in summer and are mainly controlled by surface topography(Luthjeet al., 2006). The potential impacts of basal topography on the spatial distribution of supraglacial lakes have also been studied (Lampkin and VanderBerg, 2011).

Moulins are the vertical conduits that transfer meltwater to the subsurface and are thus the meltwater transport termination on the ice sheet surface (Dewart, 1966). As the terminations of surface meltwater transport, moulins can gather and release huge amounts of meltwater; thus, it is crucial to determine the spatial distribution of moulins, leading to a better understanding of supraglacial/englacial meltwater transfer (McGrathet al., 2011). In contrast to supraglacial lakes (polygon) or supraglacial streams (polyline), moulins are point features and thus are very difficult to distinguish using coarse- or medium-spatial resolution images(Holmlund, 1988). Therefore, field validation is the main approach to locating moulins (Dewart, 1966). Recently,with the development of high-spatial-resolution remotely sensed imagery, manual interpretation from images shows great potential to delineate and locate moulins more accurately (Phillipset al., 2011). However, automatic moulin extraction methods are still lacking. It should be noted that the relatively stable locations of moulins may facilitate their delineation. This characteristic can lead to a model approach. For instance, Phillipset al. (2011) modeled the spatial distributions of moulins by using a fuzzy set overlay model with elevation, slope, and aspect as indicators, and they showed 88% success after field data validation. In addition to linking supraglacial and englacial hydrological systems, moulins also work as important englacial meltwater transport paths (Daset al., 2008).

Meltwater can also flow directly into the ocean by proglacial rivers, forming a third and least-known meltwater release approach. It is difficult to directly measure the discharge of these proglacial rivers due to their braided surfaces and significantly varying velocities (Hodgkins, 2001). Thus,certain discharge indicators are employed instead, among which sediment plumes are used most frequently (Mernildet al., 2008; Chuet al., 2009; McGrathet al., 2010). Remote sensing has shown great successes in measuring discharges of braided rivers in terrestrial environments, and its applications in proglacial environments should be exploited in the future (Smith, 1997).

3.4 Integrated research of the GrIS supraglacial hydrological system

The current studies of the GrIS supraglacial hydrological system mainly focus on a specific hydrological feature (e.g.,supraglacial lakes); integrated research focusing on the overall characteristics of the system is still lacking. To solve this problem, McGrathet al. (2011) conducted a preliminary study on the mass balance dynamics in a moulin-controlled watershed on the GrIS.

Hydrological analysis based on DEM data provides a new opportunity for integrated supraglacial hydrology research (Hardyet al., 2000). The theoretical basis for this approach is that topography is the main factor that controls meltwater transport. If the "actual" supraglacial hydrological system (including supraglacial streams, supraglacial lakes,and moulins, as previously mentioned) is well organized, it will match well with the corresponding system modeled by DEM. Otherwise, the actual system will significantly differ from the modeled one, showing as an "unorganized" system.Therefore, comparing an actual and modeled supraglacial hydrological system will reveal the forms and spatial patterns of meltwater forms, transports, and release processes. Lewis and Smith (2009) delineated the GrIS watersheds using coarse surface and bedrock DEMs, dividing the GrIS into 293 watersheds and demonstrating the melt characteristics of all the watersheds. Given the complexity of the GrIS suprag-lacial hydrological system, an integrated approach that combines both observations and DEM modeling is of great importance in future.

4 How does meltwater impact the ice sheet motion?

4.1 Impact patterns

Some of the surface meltwater flows into the ice sheet and impacts the ice motion (Chandleret al., 2013). It is crucial to determine the impact patterns of the surface melt on the ice sheet motion (Liu Q and Liu S, 2012). However, it is difficult to acquire the direct impact pattern information, and hence a simple correlation can be employed instead (Zwallyet al.,2002). More specifically, indicators about the surface melt and the ice sheet motion are derived separately and the correlation between them is determined to reveal the potential impacts of the surface melt on the ice sheet motion (Sundalet al., 2009).

The main approaches to modeling the GrIS surface melt have been represented in Section 2.1. The key method that represents the ice sheet motion is to monitor outlet glaciers,due to their high motion speed (Zwallyet al., 2002). For instance, one of the most representative GrIS outlet glaciers, the Jakobshavn Isbrae glacier, is believed to move at a very high speed (1.3×104m/a in 2003) (Joughinet al., 2004). In the summer of 2008, an outlet glacier transect (35 km) in the western GrIS was equipped with four GPS receivers to continuously monitor the glacier motion speed, and they indicated that the motion speed during the melt season was twice as fast as the baseline values in winter (Bartholomewet al., 2010).Rignot and Kanagaratnam (2006) analyzed the motion speeds of the main GrIS outlet glaciers using two Radarsat-1 InSAR images, indicating that the motion acceleration extent expanded from 66°N in 1996–2000 to 70°N in 2005.

4.2 Contradictory viewpoints

In the typical pattern of GrIS meltwater transport and release processes, meltwater flows into the ice sheet and sometimes even reaches the bottom surface of the ice sheet.This meltwater can lubricate the bottom surface and consequently accelerates the ice sheet motion. This viewpoint was first proposed with the discovery of a positive relationship between surface melt rates and the synchronous ice sheet motion speed (Zwallyet al., 2002). Since then, much evidence has been found to support this viewpoint. For instance, a 100% increase in the ice sheet motion speed was found 2 h after the surface melt achieved a peak value(Shepherdet al., 2009). The motion speed of 10 watersheds in the western GrIS derived from InSAR images yielded positive relationships with corresponding modeled runoff(Palmeret al., 2011). Moreover, Palmeret al. (2011) investigated the potential relationships between motion acceleration and the spatial distributions of supraglacial hydrological features, and Bartholomewet al. (2011) explored the driving forces of hydrological features pertaining to ice sheet motion acceleration.

Although many studies support this "lubrication-acceleration" theory, it is still challenged. Some studies question the contribution capability of surface meltwater to the motion acceleration, while others indicate that the meltwater actually slows the ice sheet motion. For instance, a study of the Jakobshavn Isbrae Glacier by Joughinet al.(2008a) elucidated the lubrication process and indicated that the glacier only accelerates 10%–20% during the melt seasons. A study by Van de Walet al. (2008) represented that only a seasonal positive feedback loop exists between the surface melt and the ice sheet motion acceleration, and a 17-year monitoring result even suggested that ice motion speed decreases with the increasing runoff. Moreover, the motion speeds have been found to be slower in warmer years. A possible explanation of this phenomenon is that meltwater facilitates the evolution of englacial meltwater pathways, and better englacial pathways can increase the friction coefficient and thus decrease the ice sheet motion speed (Adler, 2010).

The main reason for these two contradictory viewpoints is the lack of understanding of the englacial/subglacial meltwater transport mechanism; therefore, it is crucial to better analyze these processes. Based on transport efficiencies, there are two types of englacial pathways (Pimentel and Flowers,2011). Creyts and Schoof (2009) studied the mechanism of the "low-efficiency" pathway, and Daset al. (2008)concluded that well-developed englacial pathways transport meltwater to the ice sheet bottom with high efficiency. The meltwater transfer processes of englacial lakes were analyzed by Winghamet al. (2006), and Bell (2008) found that the geometry characteristics of an englacial hydrological system determine its capability to affect ice sheet motion.

5 Conclusions

The GrIS surface melt has great but undetermined impacts on the ice sheet mass loss. To improve the understanding of the production, transport, and release processes of meltwater on the GrIS, this review summarizes the current progress of GrIS surface melt research.

1) How much meltwater is produced on the ice sheet?Surface melt models are the main approaches to determine the surface runoff. Degree-day models are widely used but these models may not describe the surface melt status comprehensively. More-reasonable models with new influence factors will be the research focus in the future. Energy balance models are much more complex and can simulate the physical melt processes more accurately, but the key disadvantage of these models is their strict requirements for input parameters. The availability of the input parameters will be another important research topic in the future study. Moreover, more attention should be paid to spatially distributed energy balance models that produce better spatial and temporal simulation results. Aside from models, observational information about supraglacial lake areas, depths, and vol-umes can also indicate the surface melt status; therefore,future research should focus on applications of high-spatial-resolution remotely sensed imagery.

2) What are the characteristics of a supraglacial hydrological system? The supraglacial hydrological system on the GrIS is very complex. The supraglacial streams and crevasses are the surface meltwater pathways, and moulins and supraglacial lakes are the features that release meltwater. Two kinds of watersheds form on the GrIS; they are controlled by moulins and supraglacial lakes, respectively. Moulin-controlled watersheds can transfer meltwater into the ice sheet with high efficiency. Lake-controlled watersheds host meltwater in lakes, some of which can form moulins at the bottom surface and drain meltwater directly into the ice sheet.At present, supraglacial hydrological features are studied separately, but an integrated approach that studies supraglacial hydrological systems comprehensively is a promising future undertaking.

3) How does meltwater impact the ice sheet motion? To reveal the impacts of surface meltwater on the ice sheet motion, the correlation between the surface runoff and the ice sheet motion speed is commonly employed. However,this has produced two contradictory viewpoints: some studies indicate that meltwater can lubricate the ice sheet bottom and thus accelerate the ice sheet motion, while other studies represent that this lubrication process is not significant or even does not exist. To resolve this, it is crucial to thoroughly analyze the englacial/subglacial meltwater transport mechanisms.

The authors would like to thank Yongxue Liu and two anonymous reviewers for their constructive comments on this manuscript. This work was supported by the Scholarship Award for Excellent Doctoral Student granted by Ministry of Education and the Graduate Education Innovation Project of Jiangsu Province (CXLX12-0039).

Adler JJ, 2010. Assessing supraglacial water volume and the changing dynamics of the surface topography near the Jakobshavn Glacier, Greenland.Ph.D. dissertation, University of Colorado, Boulder, CO.

Alley RB, Dupont TK, Parizek BR,et al., 2005. Access of surface meltwater to beds of sub-freezing glaciers: Preliminary insights. Annals of Glaciology, 40:8–14.

Alley RB, Fahnestock M, Joughin I, 2008. Understanding glacier flow in changing times. Science, 322(5904): 1061–1062.

Bartholomew I, Nienow P, Mair D,et al., 2010. Seasonal evolution of subglacial drainage and acceleration in a Greenland outlet glacier. Nature Geoscience, 3(6): 408–411.

Bartholomew ID, Nienow P, Sole A,et al., 2011. Seasonal variations in Greenland Ice Sheet motion: Inland extent and behaviour at higher elevations. Earth and Planetary Science Letters, 307(3–4): 271–278.

Bell RE, 2008. The role of subglacial water in ice-sheet mass balance. Nature Geoscience, 1(5): 297–304.

Box JE, Ski K, 2007. Remote sounding of Greenland supraglacial melt lakes:implications for subglacial hydraulics. Journal of Glaciology, 53: 257–265.

Catania GA, Neumann TA, 2010. Persistent englacial drainage features in the Greenland Ice Sheet. Geophysical Research Letters, 37(2): L02501.

Chandler DM, Wadham JL, Lis GP,et al., 2013. Evolution of the subglacial drainage system beneath the Greenland Ice Sheet revealed by tracers.Nature Geoscience, 6(3): 195–198.

Chu VW, Smith LC, Rennermalm AK,et al., 2009. Sediment plume response to surface melting and supraglacial lake drainages on the Greenland Ice Sheet. Journal of Glaciology, 55(194): 1072–1082.

Colgan W, Steffen K, McLamb WS,et al., 2011. An increase in crevasse extent,west Greenland: Hydrologic implications. Geophysical Research Letters,38(18): L18502.

Creyts TT, Schoof CG, 2009. Drainage through subglacial water sheets. Journal of Geophysical Research, 114(F4): F04008.

Cui YH, Ye BS, Wang J,et al., 2010. Analysis of the spatial-temporal variations of the positive degree-day factors on the glacier No. 1 at the headwaters of the Urümqi River. Journal of Glaciology and Geocryology, 32(2): 265–274.

Das SB, Joughin I, Behn MD,et al., 2008. Fracture propagation to the base of the greenland ice sheet during supraglacial lake drainage. Science,320(5877): 778–781.

Dewart G, 1966. Moulins on Kaskawulsh Glacier. Journal of Glaciology, 6(44):320–321.

Dowdeswell JA, 2006. The Greenland ice sheet and global sea-level rise.Science, 311(5763): 963–964.

Ettema J, van den Broeke MR, van Meijgaard E,et al., 2010. Climate of the Greenland ice sheet using a high-resolution climate model, Part 1:Evaluation. The Cryosphere, 4(4): 511–527.

Georgiou S, ShepherdA, McMillan M,et al., 2009. Seasonal evolution of supraglacial lake volume from ASTER imagery. Annals of Glaciology,50(52): 95–100.

Gregory JM, Huybrechts P, 2006. Ice-sheet contributions to future sea-level change. Philosophical Transactions of the Royal Society A–Mathematical Physical and Engineering Sciences, 364(1844): 1709–1731.

Hambrey MJ, 1977. Supraglacial drainage and its relationship to structure, with particular reference to Charles Rabots Bre, Okstindan, Norway. Norsk Geografisk Tidsskrift–Norwegian Journal of Geography, 31(2): 69–77.

Hanna E, Huybrechts P, Mote TL, 2002. Surface mass balance of the Greenland ice sheet from climate-analysis data and accumulation/runoff models.Annals of Glaciology, 35: 67–72.

Hardy RJ, Bamber JL, Orford S, 2000. The delineation of drainage basins on the Greenland ice sheet for mass-balance analyses using a combined modelling and geographical information system approach. Hydrological Processes, 14(11–12): 1931–1941.

Hock R, 2005. Glacier melt: A review of processes and their modelling.Progress in Physical Geography, 29(3): 362–391.

Hodgkins R, 2001. Seasonal evolution of meltwater generation, storage and discharge at a non-temperate glacier in Svalbard. Hydrological Processes,15(3): 441–460.

Hoffman MJ, Catania GA, Neumann TA,et al., 2011. Links between acceleration, melting, and supraglacial lake drainage of the western Greenland Ice Sheet. Journal of Geophysical Research, 116(F4): F04035.

Holmlund P, 1988. Internal geometry and evolution of moulins, Storglaciaren,Sweden. Journal of Glaciology, 34(117): 242–248.

Howat IM, Joughin I, Tulaczyk S,et al., 2005. Rapid retreat and acceleration of Helheim Glacier, east Greenland. Geophysical Research Letters, 32(22):L22502.

Irvine-Fynn TDL, Hodson AJ, Moorman BJ,et al., 2011. Polythermal glacier hydrology: A review. Reviews of Geophysics, 49(4): RG4002.

Janssens I, Huybrechts P, 2000. The treatment of meltwater retention in mass-balance parameterizations of the Greenland ice sheet. Annals of Glaciology, 31(1): 133–140.

Joughin I, Abdalati W, Fahnestock M, 2004. Large fluctuations in speed on Greenland’s Jakobshavn Isbrae glacier. Nature, 432(7017): 608–610.

Joughin I, Das SB, King MA,et al., 2008a. Seasonal speedup along the western flank of the Greenland ice sheet. Science, 320(5877): 781–783.

Joughin I, Howat IM, Fahnestock M,et al., 2008b. Continued evolution of Jakobshavn Isbrae following its rapid speedup. Geophysical Research Letters, 113(F4): F04006.

Knighton AD, 1981. Channel form and flow characteristics of supraglacial streams, Austre Okstindbreen, Norway. Arctic and Alpine Research, 13(3):295–306.

Knighton AD, 1985. Channel form adjustment in supraglacial streams, Austre Okstindbreen, Norway. Arctic and Alpine Research, 17(4): 451–466.

Krawczynski MJ, Behn MD, Das SB,et al., 2009. Constraints on the lake volume required for hydro-fracture through ice sheets. Geophysical Research Letters, 36(10): L10501.

Lampkin DJ, 2011. Supraglacial lake spatial structure in western Greenland during the 2007 ablation season. Journal of Geophysical Research, 116(F4):F04001.

Lampkin DJ, VanderBerg J, 2011. A preliminary investigation of the influence of basal and surface topography on supraglacial lake distribution near Jakobshavn Isbrae, western Greenland. Hydrological Processes, 25(21):3347–3355.

Leeson AA, Shepherd A, Palmer S,et al., 2012. Simulating the growth of supraglacial lakes at the western margin of the Greenland ice sheet. The Cryosphere, 6(5): 1077–1086.

Lewis SM, Smith LC, 2009. Hydrologic drainage of the Greenland Ice Sheet.Hydrological Processes, 23(14): 2004–2011.

Liang YL, Colgan W, Lv Q,et al., 2012. A decadal investigation of supraglacial lakes in west Greenland using a fully automatic detection and tracking algorithm. Remote Sensing of Environment, 123: 127–138.

Liston GE, Mernild SH, 2012. Greenland freshwater runoff. Part I: A runoff routing model for glaciated and nonglaciated landscapes (HydroFlow).Journal of Climate, 25(17): 5997–6014.

Liu Q, Liu S, 2012. Progress in the study of englacial and subglacial drainage system of glaciers. Advances in Earth Sciences, 27(6): 660–669.

Luthje M, Pedersen LT, Reeh N,et al., 2006. Modelling the evolution of supraglacial lakes on the west Greenland ice-sheet margin. Journal of Glaciology, 52(179): 608–618.

Marston RA, 1983. Supraglacial stream dynamics on the Juneau icefield.Annals of the Association of American Geographers, 73(4): 597–608.

McGrath D, Colgan W, Steffen K,et al., 2011. Assessing the summer water budget of a moulin basin in the Sermeq Avannarleq ablation region,Greenland ice sheet. Journal of Glaciology, 57(205): 954–964.

McGrath D, Steffen K, Overeem I,et al., 2010. Sediment plumes as a proxy for local ice-sheet runoff in Kangerlussuaq Fjord, West Greenland. Journal of Glaciology, 56: 813–821.

McMillan M, Nienow P, Shepherd A,et al., 2007. Seasonal evolution of supra-glacial lakes on the Greenland Ice Sheet. Earth and Planetary Science Letters, 262(3–4): 484–492.

Mernild SH, Hasholt B, Kane DL,et al., 2008. J?kulhlaup observed at Greenland ice sheet. Eos, Transactions American Geophysical Union,89(35): 321–322.

Mernild SH, Liston GE, 2012. Greenland freshwater runoff. Part II: Distribution and trends, 1960–2010. Journal of Climate, 25(17): 6015–6035.

Mernild SH, Liston GE, Hiemstra CA,et al., 2011. Surface mass balance and runoff modeling using HIRHAM4 RCM at Kangerlussuaq (Sondre Stromfjord), west Greenland, 1950–2080. Journal of Climate, 24(3):609–623.

Mernild SH, Liston GE, Hiemstra CA,et al., 2009. Greenland ice sheet surface mass-balance modelling and freshwater flux for 2007, and in a 1995–2007 perspective. Hydrological Processes, 23(17): 2470–2484.

Mernild SH, Liston GE, Steffen K,et al., 2010. Runoff and mass-balance simulations from the Greenland Ice Sheet at Kangerlussuaq (S?ndre Str?mfjord) in a 30-year perspective, 1979–2008. The Cryosphere, 4(2):231–242.

Nicholls RJ, Cazenave A, 2010. Sea-level rise and its impact on coastal zones.Science, 328(5985): 1517–1520.

Nolin AW, Payne MC, 2007. Classification of glacier zones in western Greenland using albedo and surface roughness from the Multi-angle Imaging SpectroRadiometer (MISR). Remote Sensing of Environment,107(1–2): 264–275.

Ohmura A, 2001. Physical basis for the temperature-based melt-index method.Journal of Applied Meteorology, 40(4): 753–761.

Palmer S, Shepherd A, Nienow P,et al., 2011. Seasonal speedup of the Greenland Ice Sheet linked to routing of surface water. Earth and Planetary Science Letters, 302(3–4): 423–428.

Parizek BR, Alley RB, 2004. Implications of increased Greenland surface melt under global-warming scenarios: Ice-sheet simulations. Quaternary Science Reviews, 23(9–10): 1013–1027.

Pfeffer WT, Meier MF, Illangasekare TH, 1991. Retention of Greenland runoff by refreezing: implications for projected future sea level change. Journal of Geophysical Research, 96(C12): 22117–22124.

Phillips T, Leyk S, Rajaram H,et al., 2011. Modeling moulin distribution on Sermeq Avannarleq glacier using ASTER and WorldView imagery and fuzzy set theory. Remote Sensing of Environment, 115(9): 2292–2301.

Pimentel S, Flowers GE, 2011. A numerical study of hydrologically driven glacier dynamics and subglacial flooding. Proceedings of the Royal Society A-Mathematical Physical and Engineering Sciences, 467(2126): 537–558.

Qiao CJ, He XB, Ye BS, 2010. Study of the degree-day factors for snow and ice on the Dongkemadi Glacier, Tanggula. Journal of Glaciology and Geocryology, 32(2): 257–264.

Quincey DJ, Luckman A, 2009. Progress in satellite remote sensing of ice sheets.Progress in Physical Geography, 33(4): 547–567.

Reeh N, 1994. Parameterization of melt rate and surface temperature in the Greenland ice sheet. Polarforschung, 59: 113–128.

Rennermalm AK, Moustafa SE, Mioduszewski J,et al., 2013. Understanding Greenland ice sheet hydrology using an integrated multi-scale approach.Environmental Research Letters, 8(1): 015017.

Rennermalm AK, Smith LC, Stroeve JC,et al., 2009. Does sea ice influence Greenland ice sheet surface-melt? Environmental Research Letters, 4(2):024011.

Rignot E, Kanagaratnam P, 2006. Changes in the velocity structure of the Greenland ice sheet. Science, 311(5763): 986–990.

Rignot E, Thomas RH, 2002. Mass balance of polar ice sheets. Science,297(5586): 1502–1506.

Selmes N, Murray T, James TD, 2011. Fast-draining lakes on the Greenland ice sheet. Geophysical Research Letters, 38(15): L15501.

Shepherd A, Hubbard A, Nienow P,et al., 2009. Greenland ice sheet motion coupled with daily melting in late summer. Geophysical Research Letters,36(1): L01501.

Smith LC, 1997. Satellite remote sensing of river inundation area, stage, and discharge: A review. Hydrological Processes, 11(10): 1427–1439.

Sneed WA, Hamilton GS, 2007. Evolution of melt pond volume on the surface of the Greenland Ice Sheet. Geophysical Research Letters, 34(3): L03501.

Sun W, Qin X, Ren J,et al., 2011. Surface energy balance in the accumulation zone of the Laohugou glacier No. 12 in the Qilian mountains during ablation period. Journal of Glaciology and Geocryology, 33(1): 38–46.

Sundal AV, Shepherd A, Nienow P,et al., 2009. Evolution of supra-glacial lakes across the Greenland Ice Sheet. Remote Sensing of Environment, 113(10):2164–2171.

Tedesco M, Fettweis X, Broeke MR,et al., 2011. The role of albedo and accumulation in the 2010 melting record in Greenland. Environmental Research Letters, 6(1): 014005/1–014005/6.

Tedesco M, Steiner N, 2011. In-situ multispectral and bathymetric measurements over a supraglacial lake in western Greenland using a remotely controlled watercraft. The Cryosphere, 5(1): 479–498.

Thomas R, Frederick E, Krabill W,et al., 2009. Recent changes on Greenland outlet glaciers. Journal of Glaciology, 55(189): 147–162.

Van de Wal RSW, Boot W, Van den Broeke MR,et al., 2008. Large and rapid melt-induced velocity changes in the ablation zone of the Greenland ice sheet. Science, 321(5885): 111–113.

Van den Broeke M, Bamber J, Ettema J,et al., 2009. Partitioning recent Greenland mass loss. Science, 326(5955): 984–986.

Vaughan DG, Arthern R, 2007. Why is it hard to predict the future of ice sheets?Science, 315(5818): 1503–1504.

Wingham DJ, Siegert MJ, Shepherd A,et al., 2006. Rapid discharge connects Antarctic subglacial lakes. Nature, 440(7087): 1033–1036.

Wu Q, Kang S, Gao T,et al., 2010. The characteristics of the positive degree-day factors of the Zhadang glacier on the Nyainqêntanglha range of Tibetan Plateau, and its application. Journal of Glaciology and Geocryology,32(5): 891–897.

Yang K, Smith LC, 2013. Supraglacial streams on the Greenland ice sheet delineated from combined spectral-shape information in high-resolution satellite imagery. IEEE Geoscience and Remote Sensing Letters, 10(4):801–805.

Zwally HJ, Abdalati W, Herring T,et al., 2002. Surface melt-induced acceleration of Greenland ice-sheet flow. Science, 297(5579): 218–222.