Research ArticleCLIMATOLOGY

Greenland Ice Sheet surface melt amplified by snowline migration and bare ice exposure

See allHide authors and affiliations

Science Advances  06 Mar 2019:
Vol. 5, no. 3, eaav3738
DOI: 10.1126/sciadv.aav3738
  • Fig. 1 Interannual variations in frequency of summer bare ice exposure around the Greenland Ice Sheet over the study period (2001–2017).

    A value of 1 indicates that bare ice was observed in every non–cloud contaminated summer June, July, and August (JJA) MODIS observation. Satellite-derived bare ice exposure maps were used to delineate the end-of-summer (maximum) snowline elevation (represented by vertical black lines) for each year. (A) Pan-Greenland bare ice exposure averaged over the 2001–2017 study period. Black lines denote the IMBIE sectors used for regional analysis. Numbered labels are as follows: 1, North; 2, Northeast; 3, East; 4, Southeast; 5, South; 6, Southwest; 7, West; and 8, Northwest Greenland. (B) Frequency of bare ice exposure and snowline elevation at Humboldt Glacier, North Greenland. (C) Frequency of bare ice exposure and snowline elevation at Kangerlussuaq Glacier, East Greenland. (D) Frequency of bare ice exposure and snowline elevation across the K-transect of automated weather stations (AWSs) in Southwest Greenland. The location of each transect is displayed in (A). Note that the x-axis scale is the same for (B) to (D).

  • Fig. 2 Interannual variations in observed end-of-summer snowline elevation and bare ice extent across the Greenland Ice Sheet from 2001 to 2017.

    (A) End-of-season (or maximum) snowline elevation; (B) maximum bare ice extent. Gray shaded areas represent the uncertainty of both metrics (see Materials and Methods). Neither metric exhibits a statistically significant linear trend over the entire study period (P > 0.10), partly owing to low snowline elevation and bare ice extent in 2013 and 2017. However, snowline elevation and bare ice extent increased significantly (P < 0.01) between 2001 and 2012, a period of declining albedo on the ice sheet often attributed to hydrological and/or biological processes that darken the ice sheet surface. Relationship between (C) snowline elevation and (D) bare ice extent with mean summer pan–Greenland Ice Sheet albedo. Both metrics are significantly correlated, indicating that the area of bare ice exposure and total ice sheet albedo are closely coupled.

  • Fig. 3 Partitioned net shortwave radiation absorbed by the Greenland Ice Sheet ablation zone over the 2001-2017 study period, in units of energy (exajoules) and mass (gigatons of potential melt).

    (A) Interannual variations; (B) cumulative summer (JJA) average. In each figure, net shortwave (SWnet) radiation is partitioned into radiation absorbed by firn/snow albedo (turquoise), firn/snow extent (blue), bare ice extent (orange), and bare ice albedo (red). Interannual fluctuations in bare ice extent (and corresponding fluctuations in firn/snow extent) drive most of the variation in net shortwave radiation, signifying that the processes that expose bare ice (e.g., snowline migration) are the strongest melt-albedo feedback operating on the ice sheet.

  • Fig. 4 Hypsometric surface elevation-area relationships for all eight IMBIE Greenland sectors.

    Gray shaded areas indicate the observed range of snowlines and bare ice areas that we mapped using MODIS satellite imagery from 2001 to 2017. Red shaded areas indicate the estimated range of snowlines and bare ice areas that would occur if end-of-summer snowlines rise 500 m higher than the 2001–2017 mean. Labels summarize rate-of-change of bare ice exposure in units of area per meter increase in snowline (km2 m−1). All hypsometric relationships are computed using the Greenland Ice Mapping Project (GIMP) digital elevation model (37) and assume static, present-day surface topography. The exponential relationships suggest that strength of the snowline-albedo feedback reported here will become stronger in the future as snowlines rise to higher, flatter areas of the ice sheet, owing to increasing areas of bare ice exposed per unit rise in snowline elevation.

  • Fig. 5 Comparison between observed (MODIS) and modeled (MAR3.9 and RACMO2.3p2) bare ice extent and snowline during the 2001–2017 study period.

    Interannual variations in (A) maximum bare ice extent and (B) maximum snowline elevation. The residual between observed and modeled (C) bare ice extent and (D) snowline is greatest during years with higher runoff, because RACMO2.3p2 does not expose enough bare ice while MAR3.9 exposes too much. Note that the timing of bare ice exposure is also modeled inaccurately during high melt years (see fig. S2). These discrepancies introduce uncertainties in RCM forecasts of future Greenland runoff contributions to sea level rise.

  • Fig. 6 Relationship between end-of-season snowline elevation and total summer runoff during the 2001–2017 period (R2= 0.73).

    Note that both variables are modeled by MAR3.9. The labeled lines indicate the mean end-of-season snowline observed by this study (solid), modeled by RACMO2.3p2 (dotted), and modeled by MAR3.9 (dashed) over the 2001–2017 study period. The strong positive correlation between snowline and runoff indicates that accurate representation of snowlines in climate models is critical for accurate forecasts of Greenland Ice Sheet runoff.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/3/eaav3738/DC1

    Fig. S1. Sensitivity of each IMBIE sector to the snowline-albedo feedback during the 2001–2017 study period.

    Fig. S2. Difference between observed (MODIS) bare ice presence index and that modeled by RACMO2.3p2 and MAR3.9 in an average melt year (2005) and high melt year (2016).

    Fig. S3. Example of random forests classification of bare ice and snow in West Greenland on 25 July 2016.

    Fig. S4. Feature space occupied by water, land, bare ice, and snow in MOD09GA visible and near-infrared bands.

    Fig. S5. Frequency of bare ice exposure (i.e., presence index) during summer 2012 in Southwest Greenland (IMBIE sector 6).

    Fig. S6. Mean number of valid MODIS pixel observations across the Greenland Ice Sheet for each summer month between 2001 and 2017.

    Fig. S7. Relationship between MODIS bare ice presence index with that derived from 21 AWSs situated across the Greenland Ice Sheet between 2009 and 2017.

    Table S1. Maximum (or end-of-summer) snowline and bare ice extent for IMBIE sectors 1 to 4 between 2001 and 2017.

    Table S2. Maximum (or end-of-summer) snowline elevation and bare ice extent for IMBIE sectors 5 to 8 between 2001 and 2017.

  • Supplementary Materials

    This PDF file includes:

    • Fig. S1. Sensitivity of each IMBIE sector to the snowline-albedo feedback during the 2001–2017 study period.
    • Fig. S2. Difference between observed (MODIS) bare ice presence index and that modeled by RACMO2.3p2 and MAR3.9 in an average melt year (2005) and high melt year (2016).
    • Fig. S3. Example of random forests classification of bare ice and snow in West Greenland on 25 July 2016.
    • Fig. S4. Feature space occupied by water, land, bare ice, and snow in MOD09GA visible and near-infrared bands.
    • Fig. S5. Frequency of bare ice exposure (i.e., presence index) during summer 2012 in Southwest Greenland (IMBIE sector 6).
    • Fig. S6. Mean number of valid MODIS pixel observations across the Greenland Ice Sheet for each summer month between 2001 and 2017.
    • Fig. S7. Relationship between MODIS bare ice presence index with that derived from 21 AWSs situated across the Greenland Ice Sheet between 2009 and 2017.
    • Table S1. Maximum (or end-of-summer) snowline and bare ice extent for IMBIE sectors 1 to 4 between 2001 and 2017.
    • Table S2. Maximum (or end-of-summer) snowline elevation and bare ice extent for IMBIE sectors 5 to 8 between 2001 and 2017.

    Download PDF

    Files in this Data Supplement:

Navigate This Article