Methodology:

We applied a previously-published threshold-based pixel classification method (Moussavi et al., 2020) which combines separate threshold-based algorithms to detect (1) surface meltwater, (2) clouds, (3) exposed rock outcrop and (4) seawater. Liquid water-covered pixels are classified using the Normalized Difference Water Index (Yang and Smith, 2013). Threshold values were determined by creating a training dataset based on selected Landsat 8 images. Using these thresholds, binary (i.e. meltwater and non-meltwater) masks are created for each Landsat 8 scene. The full details are discussed comprehensively in Arthur et al. (submitted) and Moussavi et al. (2020).

We applied a physically-based radiative transfer model to calculate the water depth of all pixels classified as lake (Pope et al., 2016; Sneed and Hamilton 2007). This method calculates lake water depth (z) using the rate of light attenuation in water, lake bottom albedo, and optically-deep water reflectance (Philpot, 1989). For January of each year (2014 to 2020), we created a maximum lake depth mask by assigning all water pixels in the maximum lake area mask a depth equal to the maximum water depth observed out of all images during January following Banwell et al. (2021).

A detailed description of the data collection, quality control, processing and analysis, as well as full references, is given in: Arthur, J.F, Stokes, C.R., Jamieson, S.S.R, Carr, J.R, Leeson, A.A, Verjans, V. (submitted) Large interannual variability in supraglacial lakes around East Antarctica.

Data quality:

We used a minimum size threshold of five pixels in order to remove very small SGLs or slush likely comprised solely of mixed pixels, following previous studies (Arthur et al., 2021; Moussavi et al., 2020; Pope et al., 2016, Stokes et al., 2019).

We manually verified our classification results against 2175 Landsat 8 images and removed any false positives (cloud, shadow or rock mis-identified as SGLs that bypassed initial cloud, rock and seawater masking procedures due to spectral similarities). These false positives were often distinguishable by their 'diffuse' boundaries, as opposed to distinct lake objects. Moussavi et al. (2020) recorded an accuracy of >94% when validating SGLs classified using our method against manually-digitized SGLs.

A detailed description of the data collection, quality control, processing and analysis, as well as full references, is given in: Arthur, J.F, Stokes, C.R., Jamieson, S.S.R, Carr, J.R, Leeson, A.A, Verjans, V. (in review) Large interannual variability in supraglacial lakes around East Antarctica.