Monitoring northern Greenland proglacial river discharge from space

Abstract

Large volumes of meltwater produced on the northern Greenland Ice Sheet (GrIS) are directly routed into proglacial rivers, forming continuous supraglacial-proglacial catchments. Thereby, estimating proglacial river discharge is crucial for better understanding of northern Greenland hydrology and mass balance. We propose a method for estimating proglacial river discharge solely from space by combining Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2), ArcticDEM, and Harmonized Landsat and Sentinel-2 (HLS) data. Firstly, we use the modified normalized difference water index to extract proglacial river water masks from 30 m HLS imagery time series and calculate river effective width (W_e). Secondly, we derive near-dry riverbed cross-sectional curves from ICESat-2 ATL06 data. Thirdly, we intersect proglacial river water masks with riverbed cross-sectional curves to calculate the mean depth, wetted perimeter, cross-sectional area, and hydraulic radius, and combine ArcticDEM to estimate the channel bed slope. Finally, with these hydraulic geometry estimates, we calculate proglacial discharge and analyze its uncertainty via error propagation. We apply this method to estimate the proglacial discharge (Q_s) of the Denmark catchment (area ∼ 3254 km²) in northern Greenland during the 2020–2021 melt seasons and compare Q_s with the surface meltwater runoff (Q_m) simulated by two regional climate models (RCMs, including MARv3.12 and RACMO2.3p2), and validate the accuracy and spatial transferability of the method with in-situ proglacial discharge (Q_in-situ) of the Watson River in southwestern Greenland. The results show that: (1) the satellite-estimated W_e and Q_s exhibit significant seasonal variations, with the average W_e of 579 ± 371 m for 2020, 505 ± 394 m for 2021, and a maximum of 2040 m, and Q_s has the average value of 207.6 ± 134.1 m³/s for 2020, 210.4 ± 243.2 m³/s for 2021, and a maximum of 1509.4 ± 190.3 m³/s; (2) the satellite-estimated Q_s is positively correlated with the RCM-simulated Q_m (R² = 0.82 and 0.69 for MAR and RACMO, respectively), indicating that RCMs can reflect the overall seasonal variations of proglacial discharge reasonably well; (3) the RCM-simulated Q_m is considerably higher than our satellite-estimated Q_s, with the bias, RMSE, and RRMSE for MAR (RACMO) being 116.6 ± 5.9 m³/s (130.3 ± 5.9 m³/s), 174.7 ± 6.7 m³/s (208.9 ± 6.1 m³/s), and 83 ± 4 % (100 ± 4 %), respectively, and (4) our satellite-based method can be successfully applied to the Watson River in southwestern Greenland and the resultant Q_s matches well with Q_in-situ (RRMSE = 27 %). In conclusion, multi-temporal and multi-source satellite observations facilitate the estimation of proglacial river discharge and provide an approach to directly estimate GrIS ice surface meltwater runoff.