10:45 AM - 12:15 PM
[ACC25-P04] Inland expansion of supraglacial lakes in northwestern Greenland in the recent five decades
Keywords:Greenland, outlet glacier, supraglacial lake, remote sensing
Supraglacial lakes form annually as meltwater collects within topographic depressions on top of glaciers and ice sheets during the summertime. In response to the warming climate, the distribution of supraglacial lakes is observed to expand further inland significantly after 2000, and the inland expansion trend is predicted to continue under the current warming scenario in the Greenland ice sheet. However, lake evolution on fast-flowing Greenlandic outlet glaciers is rarely reported. Here, we investigate the long-term distribution and evolution of supraglacial lakes on two fast-flowing marine-terminating glaciers (Tracy and Heilprin Glaciers) in northwestern Greenland during 1973–2021.
We used a variety of optical satellite data for various time periods: Landsat 1–3 (1973–1982), Landsat 5 (1986–1998), Landsat 7 (1999–2013), Landsat 8 (2013–2021), Sentinel-2 (2016-2021). These images are selected by image coverage of the study glaciers, cloud cover, and sun elevation. We adopted normalized difference water index (NDWI) threshold-based approach to derive lake areas for Landsat 1–3 imagery. The NDWI is calculated from the normalized ratio of the green and near infrared band as NDWI = (green − nir) / (green + nir), and a pixel with NDWI value exceeding 0.25 is recognized as water body. After the threshold-based lake identification, we inspected the preliminary results of lake extent and the corresponding optical image, manually removed pixels identified erroneously and revised the lake boundaries. For the satellite imagery of Landsat 5, 7, 8, and Sentinel-2, a supervised machine learning methodology in Google Earth Engine was employed for lake identification. We compared the machine learning classification result with manual classification result. The overall accuracy for all four datasets exceeded 96%. The limited number of pre-Landsat 8 images and the Landsat 7 scan line corrector failure limited us to track the lake evolution in high temporal resolution. Thus, we stamped the available lake masks into a certain time window to investigate supraglacial lake evolution on the annual and decadal scales.
Although the basin area of Tracy Glacier (541 km2) is comparable to that of Heilprin Glacier (654 km2), the area of supraglacial lakes on Heilprin is always over 3 times larger than Tracy Glacier during the study period. Before 2000, there was no significant change in the lake area on both glaciers. However, a substantial increase was observed after 2000. The lake area increased from 12.5 and 2.7 km2 in 1990–2000 to 20.7 and 6.5 km2 after 2010 on Heilprin and Tracy Glacier, respectively. The increase is mainly observed in the elevation above 800 m a.s.l. while lake area under 800 m a.s.l. is relatively stable during the study period. On Tracy Glacier, supraglacial lakes are rarely observed (< 0.3 km2) above 800 m a.s.l. before 2000, while the lake extent was 2.0 and 3.7 km2 in 2000–2010 and after 2010, respectively. We assume the atmospheric warming since the late 1990s lead to the inland expansion of the supraglacial lake distribution after 2000. In addition, the significant glacier thinning of Tracy Glacier after 2000 may also contribute to the more significant lake expansion there.
We used a variety of optical satellite data for various time periods: Landsat 1–3 (1973–1982), Landsat 5 (1986–1998), Landsat 7 (1999–2013), Landsat 8 (2013–2021), Sentinel-2 (2016-2021). These images are selected by image coverage of the study glaciers, cloud cover, and sun elevation. We adopted normalized difference water index (NDWI) threshold-based approach to derive lake areas for Landsat 1–3 imagery. The NDWI is calculated from the normalized ratio of the green and near infrared band as NDWI = (green − nir) / (green + nir), and a pixel with NDWI value exceeding 0.25 is recognized as water body. After the threshold-based lake identification, we inspected the preliminary results of lake extent and the corresponding optical image, manually removed pixels identified erroneously and revised the lake boundaries. For the satellite imagery of Landsat 5, 7, 8, and Sentinel-2, a supervised machine learning methodology in Google Earth Engine was employed for lake identification. We compared the machine learning classification result with manual classification result. The overall accuracy for all four datasets exceeded 96%. The limited number of pre-Landsat 8 images and the Landsat 7 scan line corrector failure limited us to track the lake evolution in high temporal resolution. Thus, we stamped the available lake masks into a certain time window to investigate supraglacial lake evolution on the annual and decadal scales.
Although the basin area of Tracy Glacier (541 km2) is comparable to that of Heilprin Glacier (654 km2), the area of supraglacial lakes on Heilprin is always over 3 times larger than Tracy Glacier during the study period. Before 2000, there was no significant change in the lake area on both glaciers. However, a substantial increase was observed after 2000. The lake area increased from 12.5 and 2.7 km2 in 1990–2000 to 20.7 and 6.5 km2 after 2010 on Heilprin and Tracy Glacier, respectively. The increase is mainly observed in the elevation above 800 m a.s.l. while lake area under 800 m a.s.l. is relatively stable during the study period. On Tracy Glacier, supraglacial lakes are rarely observed (< 0.3 km2) above 800 m a.s.l. before 2000, while the lake extent was 2.0 and 3.7 km2 in 2000–2010 and after 2010, respectively. We assume the atmospheric warming since the late 1990s lead to the inland expansion of the supraglacial lake distribution after 2000. In addition, the significant glacier thinning of Tracy Glacier after 2000 may also contribute to the more significant lake expansion there.