Search for: All records

Subglacial processes exert a major control on ice streaming. Constraining subglacial conditions thus allows for more accurate predictions of ice mass loss. Due to the difficulty in observing large‐scale conditions of the modern subglacial environment, we turn to geological records of ice streaming in deglaciated environments. Morphometric values of streamlined subglacial bedforms provide valuable information about the relative speed, direction, and maturity of past ice streams and the relationship between ice streaming and subglacial erosion and deposition. However, manually identifying streamlined subglacial bedforms across deglaciated landscapes, sometimes in clusters of several thousand, is an arduous task with difficult‐to‐control sources of variability and human‐biased errors. This paper presents a new tool that utilizes a machine learning approach to automatically identify glacially derived streamlined features. Slope variations across a landscape, identified by topographic position index, undergo analysis from a series of supervised machine learning models trained from over 600 000 data points identified across the deglaciated Northern Hemisphere. A filtered data set produced through the combination of scientifically driven preprocessing and statistical downsampling improved the robustness of our approach. After cross‐validation, we found that Random Forest detected the most true positives, up to 94.5% on a withheld test set, and an ensemble average of machine learning models provided the highest stability when applied within the range of applicable data sets, performing at up to 79% identification of true positives on an out of distribution area of interest. We build these models into an open‐source Python package, bedfinder, and apply it to new data in the Green Bay Lobe region, USA, finding the general ice‐flow direction and average streamlined subglacial bedform elongation with minimal effort. This type of open, reproducible machine learning analysis is at the leading edge of glacial geomorphology research and will continue to improve with integration of newly acquired and previously collected data. 

Abstract. Geoscientific models are facing increasing challenges to exploit growing datasets coming from remote sensing. Universal differential equations (UDEs), aided by differentiable programming, provide a new scientific modelling paradigm enabling both complex functional inversions to potentially discover new physical laws and data assimilation from heterogeneous and sparse observations. We demonstrate an application of UDEs as a proof of concept to learn the creep component of ice flow, i.e. a nonlinear diffusivity differential equation, of a glacier evolution model. By combining a mechanistic model based on a two-dimensional shallow-ice approximation partial differential equation with an embedded neural network, i.e. a UDE, we can learn parts of an equation as nonlinear functions that then can be translated into mathematical expressions. We implemented this modelling framework as ODINN.jl, a package in the Julia programming language, providing high performance, source-to-source automatic differentiation (AD) and seamless integration with tools and global datasets from the Open Global Glacier Model in Python. We demonstrate this concept for 17 different glaciers around the world, for which we successfully recover a prescribed artificial law describing ice creep variability by solving ∼ 500 000 ordinary differential equations in parallel. Furthermore, we investigate which are the best tools in the scientific machine learning ecosystem in Julia to differentiate and optimize large nonlinear diffusivity UDEs. This study represents a proof of concept for a new modelling framework aiming at discovering empirical laws for large-scale glacier processes, such as the variability in ice creep and basal sliding for ice flow, and new hybrid surface mass balance models. 

Abstract. Glacier velocity measurements are essential to understand ice flow mechanics, monitor natural hazards, and make accurate projections of future sea-level rise. Despite these important applications, the method most commonly used to derive glacier velocity maps, feature tracking, relies on empirical parameter choices that rarely account for glacier physics or uncertainty. Here we test two statistics- and physics-based metrics to evaluate velocity maps derived from optical satellite images of Kaskawulsh Glacier, Yukon, Canada, using a range of existing feature-tracking workflows. Based on inter-comparisons with ground truth data, velocity maps with metrics falling within our recommended ranges contain fewer erroneous measurements and more spatially correlated noise than velocity maps with metrics that deviate from those ranges. Thus, these metric ranges are suitable for refining feature-tracking workflows and evaluating the resulting velocity products. We have released an open-source software package for computing and visualizing these metrics, the GLAcier Feature Tracking testkit (GLAFT). 

URL: https://github.com/earthcube2021/ec21_zheng_etal