Supplemental Material for 

Satellite-based estimates of Antarctic surface meltwater fluxes

Luke D. Trusel and Karen E. Frey
Graduate School of Geography, Clark University, Worcester, Massachusetts, USA

Sarah B. Das
Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts, USA

Peter Kuipers Munneke and Michiel R. van den Broeke
Institute for Marine and Atmospheric Research Utrecht, Utrecht University, Utrecht, Netherlands

Geophys. Res. Lett., doi:10.1029/grl.xxxxx, 2013

Introduction

This supplemental material includes additional information on satellite-based melt detection, surface energy balance modeling, as well as additional results and discussion including six supplemental figures and one data table.

1. 2013gl058138-text01.txt
Text S1. A description of the satellite-based melt detection and surface energy balance modeling methods. 

2. 2013gl058138-fs01.eps 
Figure S1. Comparison of the cumulative surface meltwater fluxes at Neumayer Station.  Daily QSCAT melt backscatter reductions (MBR; below the melt threshold) are shown on the right axis (in dB) and converted to meltwater fluxes (in mm w.e.) on the left axis using linear regression with daily SEB melt fluxes and setting the intercept to zero (Melt = 0.3677*MBR).  Daily SEB observations are reproduced well by QSCAT (r = 0.83).  Although cumulative melt fluxes are similar between RACMO and the SEB model up to the 2003–2004 summer (RACMO = 110 mm w.e., SEB = 122 mm w.e.), the following melt season (2004–2005) is very poorly modeled by RACMO and leads to an overall underestimation of melt from RACMO during that year and onwards.  Investigation into the causes of this discrepancy suggests the importance the snowfall timing in RACMO and associated positive impacts on albedo.  As fresh snow has a high albedo, melt can be suppressed if moderate snow accumulation is modeled just prior to the melt season, leading to negative feedbacks that cool the subsurface snow and prevent initiation of surface melt [Figure S4]. 

3. 2013gl058138-fs02.pdf
Figure S2.  Mean annual surface melt fluxes across Dronning Maud Land ice shelves from (A) QSCAT and (B) RACMO with locations referred to in the text labeled.

4. 2013gl058138-fs03.pdf
Figure S3.  Annual melt flux sums for the 2004–2005 austral summer from (A) QSCAT and (B) RACMO.  Comparisons at Neumayer Station (location indicated by red dot in A) reveal that QSCAT closely follows melt determined from the in situ energy balance observations and modeling whereas melt is largely absent in RACMO [Figure S1; Figure S4]. Underestimation of melt during this summer in RACMO (relative to QSCAT) appears widespread across East Antarctica; particularly on Dronning Maud Land ice shelves and Amery Ice Shelf.  As a result, RACMO produces lower annual [e.g., Figure 3D,E] and 10-year mean melt fluxes than QSCAT in these regions [Figure S2], contributing to an overall underestimation of melt from RACMO compared to QSCAT [Figure 3A].

5. 2013gl058138-fs04.eps
Figure S4.  Examination of the 2004–2005 melt season at Neumayer Station using observations, an SEB model, QSCAT, and RACMO.  Although both QSCAT and the SEB model indicate melt starting on 16-Dec-2004, RACMO conversely does not.  This lack of melt in RACMO appears to result from two snowfall events occurring on and just prior to melt onset.  Although the timing of synoptic snowfall observations (gray boxes) and RACMO snowfall (gray dashed line; upper plot) generally agree, the albedo and surface temperature impact in RACMO appears too large and acted to prevent melt initiation.  As a result, the observed melt-albedo feedback is not simulated by RACMO, and as such, RACMO produces very little melt at Neumayer Station during this austral summer.

6. 2013gl058138-fs05.eps
Figure S5. Profiles from the Larsen C ice shelf (LCIS; see Figure 4) showing annual and mean (1999–2009) melt fluxes from (A) QSCAT and (B) RACMO.  Higher melt on inner LCIS is observed in nearly all years of the satellite record, whereas RACMO does not produce this pattern.  This melt pattern observed from QSCAT is inferred to represent the influence of föhn winds driving surface melt. Labeled years refer to the second austral summer year (i.e., 1999–2000 is labeled 2000).

7. 2013gl058138-fs06.eps
Figure S6. (A) Manually identified areas containing meltwater ponds and streams (blue) across the Larsen Ice Shelf observed using visible Landsat imagery spanning 2001-2011. Note that individual melt features are much smaller (e.g., Figure S6b) and as such, blue areas can be considered the maximum extent within which melt features have been observed. (B) Elongate surface melt features in Cabinet Inlet observed by Landsat 7 ETM in February 2006 (location shown in A).  

8. 2013gl058138-table01.docx
Table S1. Comparisons between statistics generated by excluding or including subsurface melt at AWS 14 and AWS 15 on Larsen C.