Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Mean, clear-sky surface temperature of the Greenland Ice Sheet was measured for each melt season from 2000 to 2005 using Moderate-Resolution Imaging Spectroradiometer (MODIS)–derived land-surface temperature (LST) data-product maps. During the period of most-active melt, the mean, clear-sky surface temperature of the ice sheet was highest in 2002 (−8.29 ± 5.29°C) and 2005 (−8.29 ± 5.43°C), compared to a 6-year mean of −9.04 ± 5.59°C, in agreement with recent work by other investigators showing unusually extensive melt in 2002 and 2005. Surface-temperature variability shows a correspondence with the dry-snow facies of the ice sheet; a reduction in area of the dry-snow facies would indicate a more-negative mass balance. Surface-temperature variability generally increased during the study period and is most pronounced in the 2005 melt season; this is consistent with surface instability caused by air-temperature fluctuations.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The importance of monitoring surface temperature is highlighted by recent modeling studies showing that summer temperature increases of only ∼2–5°C are required to double melt rates and thus increase runoff from the Greenland Ice Sheet [Hanna et al., 2005]. To assess interannual surface-temperature variability across Greenland, we analyze clear-sky surface temperature of the ice sheet derived from Moderate-Resolution Imaging Spectroradiometer (MODIS) land-surface temperature (LST) standard-data product maps for six melt seasons (2000–2005), and relate the results to ice-sheet mass balance.
 Carl Benson's pioneering work on the Greenland Ice Sheet led to his classification of the ice sheet into an ablation area and an accumulation area, separated by the equilibrium line where the net mass balance equals zero [Benson, 1962]. Various facies exist within the accumulation area, the approximate boundaries of which may sometimes be identified by their unique spectral signatures [Hall et al., 1989; Williams et al., 1991; Fahnestock et al., 1993; Abdalati and Steffen, 1995; Long and Drinkwater, 1999]. The facies boundaries change over time with changes in the ice-sheet mass balance. For example, the dry-snow area may shrink if the mass balance becomes more negative over a period of years.
 MODIS LST data products consist of daily and 8-day composite 1-km and 0.05°-resolution map products, with quality-control information in each 1-km pixel or 0.05° cell. We selected the 8-day composite 0.05°-resolution product (MOD11C2) to use for this work because the maps provide an 8-day average LST for each cell (cloud-cover permitting). MOD11C2 is derived from the 1-km and 0.05°-resolution daily products [Wan et al., 2002].
 Temperatures derived from the daily, 1-km LST product have been compared with in-situ measurements using multiple thermal-infrared radiometers in field campaigns including one in a snow field, where the error was generally <1°C [Wan et al., 2002], but could get up to 2°C. Furthermore, the LST is relatively constant, between −2° and 0°C, in the exposed ice region within the ablation area of the Greenland Ice Sheet, and the LST does not exceed 0°C over ice with surface melt. 0°C represents the upper boundary of LST for the glacier ice and is consistent with the expected temperature of melting ice. The automatic weather-station (AWS) data on the Greenland Ice Sheet [Steffen and Box, 2001] are useful in checking the trend of seasonal variations in MODIS LST data with surface air temperature or upper-level snow temperature profile data, however it is difficult to use these data to compare directly with the values in the LST product because there is no accurate radiometric measurement of snow-surface temperature in the AWS data.
 The major uncertainty in the LST product is the effect of cloud contamination because it is often difficult to discriminate cloud-contaminated from clear-sky LSTs over snow and ice in cold regions, especially in the case of thin clouds or fog that may not be detected by the MODIS cloud mask. The MODIS LST algorithm uses a cloud mask derived from MODIS data [Ackerman et al., 1998] to determine whether or not to calculate LST. The cloud mask tends to be conservative, generally mapping more clouds than are actually present over snow and ice.
 MODIS maps provide LST only when the sky is clear, causing a bias because surface temperatures are generally warmer under clouds, especially low clouds (K. Steffen, personal communication, 2005). (The air temperature difference between an overcast and clear day can be several °C.) Thus the “mean” clear-sky LSTs represent a likely underestimation of the actual mean-surface temperatures.
 In some cases there is only one clear day available to create the 8-day LST map for any given cell, thus the LST of each cell may not represent a true mean for the 8-day period. Figure 1 shows the average number of clear-sky days for each 8-day period during the entire melt season [defined as extending from April 30 or May 1 (depending on whether or not it was a leap year) (day 121) to September 28 or 29 (day 272) of each year]. Data used in Figure 1 were derived from the MODIS LST MOD11C2 product, and provide an indication of the relative cloudiness among the years.
 Nineteen 8-day-mean LST maps of Greenland were averaged on a cell-by-cell basis for the entire melt season (between days 121 and 272), to develop one map of mean-surface temperature along with a standard deviation (SD) map for each melt season. (Only 17 and 18 maps were used for 2001 and 2004, respectively, due to missing and/or suspected-bad data.) In addition, we also studied the period of most-active melt in each year, May through mid-August [days 121 to 225 (August 12 or 13)]. Because freezeup can begin around mid-August, we consider days 121–225 to be the period of most-active melt for the purpose of this work. Surface temperatures during the most-active part of the melt season are more relevant to surface melt on the ice sheet than those derived from the entire melt season, during which time subsurface melt may play a significant role.
 Mean, clear-sky surface temperature with SD for each melt season is shown in Figure 2 and Table 1. The mean, clear-sky surface temperature of the entire melt seasons for the six years is −10.71 ± 6.66°C. The highest mean LST for the entire melt season, −9.94 ± 6.19°C, occurred in 2002. According to the two-tailed z-test, the probability is >99.999% that the difference between the mean, clear-sky surface temperatures for 2002, and each of the other years, is statistically significant.
Table 1. Mean, Clear-Sky Surface Temperature (°C) and Standard Deviation (SD) of the Greenland Ice Sheeta
Entire Melt Season (Days 121–272): Mean LST and SD, °C
Number of Cells Used in the Calculation
Active-Melt Season (Days 121–225): Mean LST and SD, °C
Number of Cells Used in the Calculation
Temperatures and SDs are shown for the entire melt season from May through September (days 121–272) and for the period of most-active melt, from May through mid-August (days 121–225), each year.
−11.45 ± 6.42
−10.27 ± 6.05
−11.34 ± 6.56
−10.12 ± 5.82
−9.94 ± 6.19
−8.29 ± 5.29
−10.26 ± 6.64
−8.73 ± 5.43
−10.48 ± 6.36
−8.61 ± 5.08
−10.80 ± 7.55
−8.29 ± 5.43
−10.71 ± 6.66
−9.04 ± 5.59
 The surface temperature was highest, −8.29°C, in both 2002 and 2005 during the period of most-active melt (Table 1). This is 0.75°C higher than the 6-year mean, −9.04 ± 5.59°C, for this time period.
 The SD maps in Figure 2 show that the surface-temperature variability tends to be greater in the northern part of the ice sheet and increases throughout the study period. This is related to increasing air-temperature fluctuations, leading to surface instability. There is evidence of increasing ice discharge from outlet glaciers in recent years which is partly due to increased surface runoff, and this process has been observed to be advancing northward [Rignot and Kanagaratnam, 2006]. Air temperature and air-temperature changes would contribute to driving such processes.
 The boundaries between Benson's dry-snow and percolation facies (the region on a glacier where snow and firn is subjected to localized percolation of meltwater without becoming wet throughout), and the ERS-1 synthetic-aperture radar (SAR)-derived dry-snow facies of Fahnestock et al.  were digitally traced and overlaid on our 6-year mean SD map (Figure 3) to visualize the relationship between the glacier-facies boundaries (a function of ice-sheet mass balance), and surface-temperature variability. This pattern suggests a correspondence with the glacier facies of Benson , but especially with the SAR-derived dry-snow facies delineated by Fahnestock et al. .
 This correspondence of high temperature variability and dry snow is likely because the thermal conductivity of dry snow is lower than that of wet snow. Air-temperature changes, induced by wind variability and weather patterns, are more likely to cause frequent surface-temperature changes in dry snow as compared to wet snow (with its higher thermal conductivity). Also, surface-temperature variability is lowest in the ablation area where the upper limit of surface temperature is 0°C. The glacier-facies boundaries can change over time with a change in mass balance driven by meteorological conditions such as air temperature and precipitation.
 We looked at the frequency of cloud cover during each melt season to ascertain whether or not the surface-temperature variability is an artifact caused by cloud-cover patterns. Though there is less cloud cover in northern vs. southern Greenland (Figure 1), there is no particular pattern in the cloud cover, thus indicating that the observed increasing surface-temperature variability is not an artifact of interannual cloud-cover variability.
5. Discussion and Concluding Remarks
 Higher ice-sheet surface temperatures can lead to enhanced ice-sheet disintegration when surface water percolates through great thicknesses of the ice sheet leading to an acceleration of ice-sheet flow [Zwally et al., 2002]. We observed a general expansion of increasing surface-temperature variability from 2000 to 2005 on the Greenland Ice Sheet, being greatest in 2005, a year cited as being anomalously warm over Greenland by other investigators, and according to our LST data. The SDs are consistent with surface instability caused by air-temperature fluctuations.
 We also found that the two warmest years (2002 and 2005) are the same years that experienced the most-extensive melt of the six-year period. Steffen et al.  showed that there was a very large melt extent in 2002, extending over 690,000 km2 of the ice sheet compared to a 1979–1999 average of 455,000 km2 [Abdalati and Steffen, 2001], and 2005 has been cited by Steffen and Huff (http://cires.colorado.edu/science/groups/steffen/greenland/melt2005/) as experiencing melt equal or greater in area than occurred in 2002. This is consistent with the average melt-season LSTs presented herein, and findings by Comiso  showing that those same two years were unusually warm using AVHRR data of the Arctic since 1981.
 MODIS-derived LST provides a quantitative assessment of the mean surface temperature and surface-temperature change on the Greenland Ice Sheet. Because the location of the glacier-facies boundaries reflects the ice sheet mass balance, a sustained increase in surface temperature would cause the glacier-facies boundaries to migrate to higher elevations.
 The authors thank Carl Benson of the University of Alaska, Long Chiu of George Mason University, and Waleed Abdalati and Christopher Shuman, of NASA/GSFC, and Konrad Steffen of CIRES for helpful discussions and/or reviews. Two anonymous reviewers provided further useful insights that led to significant improvements. Support for this work was provided by NASA's Earth Observing System and Cryospheric Sciences Programs.