Predicting the Geothermal Heat Flux in Greenland: A Machine Learning Approach

Abstract

Geothermal heat flux (GHF) is a crucial boundary condition for making accurate predictions of ice sheet mass loss, yet it is poorly known in Greenland due to inaccessibility of the bedrock. Here we use a machine learning algorithm on a large collection of relevant geologic features and global GHF measurements and produce a GHF map of Greenland that we argue is within ∼15% accuracy. The main features of our predicted GHF map include a large region with high GHF in central-north Greenland surrounding the NorthGRIP ice core site, and hot spots in the Jakobshavn Isbrae catchment, upstream of Petermann Gletscher, and near the terminus of Nioghalvfjerdsfjorden glacier. Our model also captures the trajectory of Greenland movement over the Icelandic plume by predicting a stripe of elevated GHF in central-east Greenland. Finally, we show that our model can produce substantially more accurate predictions if additional measurements of GHF in Greenland are provided. Plain Language Summary The heat generated at the interior regions of Earth (geothermal heat flux, GHF) can be high enough to melt the bottom layers of ice sheets, decrease friction between ice and bedrock, and increase ice discharge to the ocean. This heat, however, cannot be directly measured in ice sheets because the bedrock is inaccessible. Here we present a novel approach to estimate this heat. We combine all the available geologic, tectonic, and GHF data that are available on all continents. We then establish a complex relationship between GHF and all the geologic-tectonic features using machine learning techniques and then predict the GHF for the Greenland Ice Sheet. We utilize all information from available ice cores and bedrock boreholes to improve the GHF prediction in Greenland. Thus, the new GHF map honors tectonic settings, regional geology, and measurements from ice cores and can be used as an important input parameter to numerical ice sheet models that aim at lowering the uncertainties of future sea level rise predictions. This study derives a new map of GHF for the GrIS using statistical relationships between global heat flux observations and the combined influence of local geology and regional tectonic setting. Compilations of global RESEARCH LETTER 10.1002/2017GL075661 Key Points: • A new geothermal heat flux map of Greenland is obtained within ∼15% accuracy using machine learning techniques • The new map honors regional geology, tectonic settings, and ice core measurements • Pockets of high heat flux are predicted in central-north Greenland and upstream of several fast-flowing outlet glaciers

A constraint upon the basal water distribution and thermal state of the Greenland Ice Sheet from radar bed echoes

“…This is because these regions tend to exhibit a diffuse scattering signature (associated with fine-scale roughness), whereas the method proposed by Gogineni (2008, 2012) is specifically tuned to detect water bodies that exhibit a spatially continuous (near-) specular scattering signature. By contrast, comparison between the water predictions in this study and the radar-derived bed roughness maps in Rippin (2013) and Jordan et al (2017) demonstrate a lack of modulation by bed roughness, with basal water present in rougher marginal regions and a generally smoother ice-sheet interior.…”

Possible Role for Tectonics in the Evolving Stability of the Greenland Ice Sheet

“…Rogozhina et al () reconstructed a broad band of enhanced geothermal flux, exceeding 80 mW/m 2 , extending from east Greenland including Scoresby Sund, running northwest and passing just north of the GISP2 and GRIP ice core sites to include the NorthGRIP and NEEM sites. Rezvanbehbahani et al () and Rysgaard et al () estimated high geothermal fluxes (using global statistics together with the mapped features in Greenland, and using fjord temperatures, respectively), which broadly agree with the reconstructions of Rogozhina et al (). The seismic inversions of Pourpoint et al (2018) yielded widespread deep‐crustal low‐velocity anomalies in north and northeast Greenland that may be compositional but also may be thermal, broadly consistent with the other results.…”

Lithospheric Structure of Greenland From Ambient Noise and Earthquake Surface Wave Tomography

“…Rogozhina et al () estimated that GHF values range between 68 and 88 mW/m 2 in NEG and that the largest anomaly is located beneath the onset of NEGIS where high basal melting has been estimated (Fahnestock et al, ). In addition, GHF values ranging from 90 to 160 mW/m 2 have been estimated around the NorthGRIP ice core site (Buchardt & Dahl‐Jensen, ; Dahl‐Jensen, Gundestrup, et al, ; Greve, ; Rezvanbehbahani et al, ). A recent study by Rysgaard et al () also reported a GHF of 93 mW/m 2 in the Young Sound‐Tyrolerfjord, a few degrees south of the strong gravity low, and suggested, based on the distribution of high GHF, that a geothermal heat source may exist beneath the central and northeastern parts of the GIS.…”