Ground penetrating radar on Rutor temperate glacier supported by ice-thickness modeling algorithms for bedrock detection

Abstract. The glaciers situated in temperate mountain chains often contain a great percentage of temperate ice, which is ice at the temperature-pressure melting point. They may also contain cold ice, but in smaller percentages and located in the coldest parts of the glacier. Measuring the glacier ice thickness is often carried out with Ground Penetrating Radar (GPR), able to scan ice hundreds of meters thick. Unfortunately, the meltwater contained in temperate glaciers challenges the detection of the ice-bedrock interface with the radar technology, due to signal scattering. According to the literature, only from 12 % to 69 % of GPR traces have been able to depict the ice-bedrock interface in Swiss glaciers, the percentage varying according to the glacier and the GPR antenna. Besides, GPR data are acquired on straight lines and do not cover the entire glacier area with a fine resolution. These problems affected previous GPR surveys of the Rutor Glacier (Aosta Valley, Southern Europe Alps), reporting an irrealistic, too low, ice thickness. To obtain a more reliable interpretation of the GPR data, estimations from ice-thickness modeling algorithms are employed to guide the analyst in the interpretation of scattered GPR data. Two new GPR datasets of the Rutor Glacier are analyzed, helicopter-based and ground-based. We tested three open-source modeling algorithms that estimate the ice thickness based on the surface topography: GlabTop2 (Glacier Bed Topography), GlaTE (Glacier Thickness Estimation), and OGGM (Open Global Glacier Model). The ice thickness raster maps produced with these models are sampled to extract ice thickness profiles coincident with the GPR measurements. The ice-bedrock interface of GPR profiles is then manually picked, using the GPR signature (ice-bedrock reflection) where visible, based on the amplitude of the signal. Where the signal is lost, e.g. due to high meltwater content, high ice thickness, or high bedrock slope, the picking is based on the suggestions from the three mathematical models. In the end, a second run of the GlaTE model assembles the estimations from the algorithm and the manually picked GPR measurements to provide a final ice-thickness map. The Rutor glacier, according to the methodology proposed, is estimated to cover an area of about 7.45 km² and store about 515 million m³ of ice in 2021, compared to the previous estimate of 150 million m³ made in 2008. The methodology is simple to reproduce and may simplify and ease future GPR surveys of temperate glaciers, especially when facing noisy data due to meltwater content or other reflectors such as debris. Another advantage is to directly produce an ice thickness map of the whole glacier, without relying on pure interpolation of sparse GPR data. The prior run of such models is also advised during the planning of glacier surveys, together with GPR forward modeling, to help choose the best GPR antenna frequency according to the expected ice thickness.