Methods

Perennial ice extent was hand-traced on Synthetic Aperture Radar (SAR) imagery and using ArcGIS. We used a buffer of 1 pixel (with a resolution of 6.25 m, 12.5 m or 50 m per pixel) on each side of the trace, represented by the error bars in Figure 2. Ice phenology was obtained using ortho-rectified aerial photography, SAR imagery and oblique photography obtained during field seasons. In these cases, the uncertainty was conservatively established at 10 m surrounding the trace for oblique photographs, and one pixel for SAR imagery and aerial photographs (1.5 m and 3 m per pixel on aerial photographs).

Ground penetrating radar (GPR) surveying was performed using a Sensors & Software PulseEKKO Pro GPR equipped with 50 MHz antennas. The apparatus was mounted on a plastic sled and dragged over the frozen lake and GPS positions were taken for every trace. Post-processing was performed using EKKO Project and included dewow, low-pass noise removal, horizontal repositioning according to GPS positions and SEC2 gain application. Interpretations were noted using the same software, providing ice and water depth data for every 5 m horizontal step. Ice and water radar velocities were set at 0.16 and 0.033 m ns-1 respectively, according to widely used standard values [Jol, 2008]. The theoretical vertical resolution of a 50 MHz signal is ±0.16 m in the water, and ±0.8 m for the ice cover [Moorman, 2001]. Extracted data was then imported into ArcGIS and a bathymetric map was created using the “topo to raster” tool and WHL perimeter as a 0 m contour. Hypsographic data was extracted from the resultant raster image using the ArcGIS hypsometric toolbox.

Water temperature and conductivity profiles were measured using a RBR XR-620 CTD which was lowered at a speed of 5 cm s-1 except in 2013 when they were taken using manually logged thermistors (±0.1°C, YSI44033). That year the conductivity was taken later than usual, on the 19 July. Profiles were smoothed using a running mean filter with a window of 5 s (~25 cm) to reduce noise.

Water track temperature monitoring was conducted during a four day period in 2011 using time domain reflectometry (TDR) probes (±1°C, Decagon device) placed just below the soil surface in the coarse gravel. Conductivity was measured once a day over a 30 day period in June-July 2013 using an Accumet AP85 Fisher Scientific device (2 µS cm-1 accuracy) and averaged over that timescale. Discharge measurment were made using a cutthroat flume equipped with a Hobo U20 water pressure sensor (± 0.003 m) measuring hourly levels that were compensated for barometric pressure variations.

References

Jol, H. M. (Ed.) (2008), Ground penetrating radar theory and applications, 544 pp., Elsevier, Amsterdam, The Netherlands.

Moorman, B. J. (2001), Ground-penetrating radar applications in paleolimnology, in Tracking environmental change using lake sediments, edited by W. M. Last and J. P. Smol, pp. 23-47, Kluwer Academic Publishers, Dordrecht, the Netherlands.

Supplemental figures

Supplemental Figure 1: Map of the Ward Hunt Island area. DF: Disraeli Fjord; SWS: CEN weather station (SILA Network); WHI : Ward Hunt Island; WHIR : Ward Hunt Ice Rise; WHIS : Ward Hunt Ice Shelf (extent at the end of summer 2012); WHL: Ward Hunt Lake; WHLW: Ward Hunt Lake watershed.

Supplemental Figure 2: Photographs from a northwest-facing automated camera showing the period of extensive open water conditions in Ward Hunt Lake in late summer 2012. The camera was installed on 10 July, ice-off was complete on 11 August and freeze-up began on 6 September.

Supplemental Figure 3: Bathymetric data for Ward Hunt Lake. a) DEM showing the extent of the <2 m and <4 m depth zones. b) Hypsographic curves in terms of lake area and volume. (c) An illustrative GPR profile across the lake without topographic corrections, with the colored lines representing the lake bottom (red) and ice/water interface (orange).