Methodology:

Measurements of ice-margin change were recorded by applying the Normalised Difference Snow Index (NDSI) to the Landsat TM, ETM+ and OLI archive to delineate the ice sheet margin and measure changes relative to a series of fixed reference points at approximately 5-year intervals (epochs) between 1987 and 2015. Rates of ice-margin advance or recession were calculated by differencing the mean distance values between the fixed reference points and the ice-margin in successive epochs, and subsequently dividing by the interval duration. Note that positive and negative values of mean ice-margin change and mean annual change represent ice-margin advance and recession respectively. Lake areas and intersects (length of the lake - ice-margin interface) were also derived at the same 5-year intervals by applying the Normalised Difference Water Index (NDWI) to the same Landsat scenes. The lake persistence parameter provides a quantitative measure of the presence of a lake in each of the respective epochs. The latitudes and altitudes of the fixed reference points were extracted from a 1985 DEM of the region. Full details of the satellite scenes used and the associated processing and data extraction in ENVI and ArcMap can be found in the publication listed below under References.

Data collection:

Scenes from the Landsat TM, ETM+ and OLI sensors were processed in ENVI v.5.2 and Esri ArcMap v.10.3.1.

Data quality:

The delineation of ice-margin and lake extent is assumed to be accurate to within ±1 pixel (30 m) of the true ice-margin/lake perimeter. Given the regional aggregation of measurements, it is assumed that any over-estimation of the ice-margin position is cancelled out by an equal and opposite under-estimation, and uncertainty in the ice-margin positions is therefore not specifically assessed for these bulk figures. In addition, the absolute error associated with each lake area measurement was dependent on lake size and planform, and thus resulted in a declining power law relationship whereby the greatest errors were associated with the smallest lakes. For example, lakes measuring 0.5 km2 had an area uncertainty of ~9 %, whilst lakes measuring > 5 km2 had an uncertainty of < 3 %.

Rates of change at lacustrine margins were only generated when a lake was present in two or more successive epochs.

Measurements of ice-margin change between 1987 and 1992 at the southernmost latitudes of the survey area are absent due to the unavailability of Landsat TM scenes for this region in the years 1985-1988.