skip to main content


Title: SICOPOLIS-AD v1: an open-source adjoint modeling framework for ice sheet simulation enabled by the algorithmic differentiation tool OpenAD
We present a new capability of the ice sheet model SICOPOLIS that enables flexible adjoint code generation via source transformation using the open-source algorithmic differentiation (AD) tool OpenAD. The adjoint code enables efficient calculation of sensitivities of a scalar-valued objective function or quantity of interest (QoI) to a range of important, often spatially varying model input variables, including initial and boundary conditions, as well as model parameters. Compared to earlier work on adjoint code generation of SICOPOLIS, our work is based on several important advances: (i) it is embedded within the up-to-date trunk of the SICOPOLIS repository – accounting for one and a half decades of code development and improvements – and is readily available to the wider community; (ii) the AD tool used, OpenAD, is an open-source tool; (iii) the adjoint code developed is applicable to both Greenland and Antarctica, including grounded ice as well as floating ice shelves, and with an extended choice of thermodynamical representations. A number of code refactorization steps were required. They are discussed in detail in an Appendix as they hold lessons for application of AD to legacy codes at large. As an example application, we examine the sensitivity of the total Antarctic Ice Sheet volume to changes in initial ice thickness, summer precipitation, and basal and surface temperatures across the ice sheet. Simulations of Antarctica with floating ice shelves show that over 100 years of simulation the sensitivity of total ice sheet volume to the initial ice thickness and precipitation is almost uniformly positive, while the sensitivities to surface and basal temperature are almost uniformly negative. Sensitivity to summer precipitation is largest on floating ice shelves from Queen Maud to Queen Mary Land. The largest sensitivity to initial ice thickness is at outlet glaciers around Antarctica. Comparison between total ice sheet volume sensitivities to surface and basal temperature shows that surface temperature sensitivities are higher broadly across the floating ice shelves, while basal temperature sensitivities are highest at the grounding lines of floating ice shelves and outlet glaciers. A uniformly perturbed region of East Antarctica reveals that, among the four control variables tested here, total ice sheet volume is most sensitive to variations in summer precipitation as formulated in SICOPOLIS. Comparison between adjoint- and finite-difference-derived sensitivities shows good agreement, lending confidence that the AD tool is producing correct adjoint code. The new modeling infrastructure is freely available at www.sicopolis.net under the development trunk.  more » « less
Award ID(s):
1751120 1750035
NSF-PAR ID:
10137969
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
Geoscientific model development discussions
ISSN:
1991-962X
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    Abstract. We present a new capability of the ice sheet model SICOPOLIS that enables flexible adjoint code generation via source transformation using the open-source algorithmic differentiation (AD) tool OpenAD.The adjoint code enables efficient calculation of the sensitivities of a scalar-valued objective function or quantity of interest (QoI) to a range of important, often spatially varying and uncertain model input variables, including initial and boundary conditions, as well as model parameters.Compared to earlier work on the adjoint code generation of SICOPOLIS, our work makes several important advances:(i) it is embedded within the up-to-date trunk of the SICOPOLIS repository – accounting for 1.5 decades of code development and improvements – and is readily available to the wider community;(ii) the AD tool used, OpenAD, is an open-source tool;(iii) the adjoint code developed is applicable to both Greenland and Antarctica, including grounded ice as well as floating ice shelves, with an extended choice of thermodynamical representations.A number of code refactorization steps were required. They are discussed in detail in an Appendix as they hold lessons for the application of AD to legacy codes at large.As an example application, we examine the sensitivity of the total Antarctic Ice Sheet volume to changes in initial ice thickness, austral summer precipitation, and basal and surface temperatures across the ice sheet.Simulations of Antarctica with floating ice shelves show that over 100 years of simulation the sensitivity of total ice sheet volume to the initial ice thickness and precipitation is almost uniformly positive, while the sensitivities to surface and basal temperature are almost uniformly negative. Sensitivity to austral summer precipitation is largest on floating ice shelves from Queen Maud to Queen Mary Land. The largest sensitivity to initial ice thickness is at outlet glaciers around Antarctica. Comparison between total ice sheet volume sensitivities to surface and basal temperature shows that surface temperature sensitivities are higher broadly across the floating ice shelves, while basal temperature sensitivities are highest at the grounding lines of floating ice shelves and outlet glaciers. A uniformly perturbed region of East Antarctica reveals that, among the four control variables tested here, total ice sheet volume is the most sensitive to variations in austral summer precipitation as formulated in SICOPOLIS.Comparison between adjoint- and finite-difference-derived sensitivities shows good agreement, lending confidence that the AD tool is producing correct adjoint code.The new modeling infrastructure is freely available at http://www.sicopolis.net (last access: 2 April 2020) under the development trunk. 
    more » « less
  2. The loss or thinning of buttressing ice shelves and accompanying changes in grounding-zone stress balance are commonly implicated as the primary trigger for grounding-line retreat, such as that observed in Amundsen Sea outlet glaciers today. Ice-shelf thinning is mostly attributed to the presence of warm ocean waters beneath the shelves. However, climate model projections show that summer air temperatures could soon exceed the threshold for widespread meltwater production on ice-shelf surfaces. This has serious implications for their future stability, because they are vulnerable to water-induced flexural stresses and water-aided crevasse penetration, termed ‘hydrofracturing’. Once initiated, the rate of shelf loss through hydrofracturing can far exceed that caused by sub-surface melting, and could result in the complete loss of some buttressing ice shelves, with marine grounding lines suddenly becoming calving ice fronts. In places where those exposed ice fronts are thick (>900m) and crevassed, deviatoric stresses can exceed the strength of the ice and the cliff face will fail mechanically, leading to rapid calving like that seen in analogous settings such as Jakobshavn on Greenland. Here we explore the implications of hydrofacturing and subsequent ice-cliff collapse in a warming climate, by parameterizing these processes in a hybrid ice sheet-shelf model. Model sensitivities to meltwater production and to ice-cliff calving rate (a function of cliff height above the stress balance threshold triggering brittle failure) are calibrated to match modern observations of calving and thinning. We find the potential for major ice-sheet retreat if global mean temperature rises more than ~2ºC above preindustrial. In the model, Antarctic calving rates at thick ice fronts are not allowed to exceed those observed in Greenland today. This may be a conservative assumption, considering the very different spatial scales of Antarctic outlets, such as Thwaites. Nonetheless, simulations following a ‘worst case’ RCP8.5 scenario produce rates of sea-level rise measured in cm per year by the end of this century. Clearly, the potential for brittle processes to deliver ice to the ocean, in addition to viscous and basal processes, needs to be better constrained through more complete, physically based representations of calving. 
    more » « less
  3. SICOPOLIS-AD is a version of the ice sheet model SICOPOLIS (originally [1]: www.sicopolis.net) used to produce adjoint sensitivities of chosen control variables. It can be used to assess the sensitivity of some quantity of interest to perturbations in variables that may affect that quantity of interest. For example, it can be used to comprehensively and quantitatively assess exactly how the volume of the Greenland Ice Sheet is sensitive to changes in surface temperature at each point on the ice sheet. This document provides basic startup and troubleshooting methods, as well as some best practices for making changes in the code for the purposes of adjoint production. 
    more » « less
  4. ABSTRACT Increasing ocean and air temperatures have contributed to the removal of floating ice shelves from several Greenland outlet glaciers; however, the specific contribution of these external forcings remains poorly understood. Here we use atmospheric, oceanographic and glaciological time series data from the ice shelf of Petermann Gletscher, NW Greenland to quantify the forcing of the ocean and atmosphere on the ice shelf at a site ~16 km from the grounding line within a large sub-ice-shelf channel. Basal melt rates here indicate a strong seasonality, rising from a winter mean of 2 m a −1 to a maximum of 80 m a −1 during the summer melt season. This increase in basal melt rates confirms the direct link between summer atmospheric warming around Greenland and enhanced ocean-forced melting of its remaining ice shelves. We attribute this enhanced melting to increased discharge of subglacial runoff into the ocean at the grounding line, which strengthens under-ice currents and drives a greater ocean heat flux toward the ice base. 
    more » « less
  5. Abstract. Basal conditions directly control the glacier sliding rate and the dynamic discharge of ice. Recent glacier destabilization events indicate that some marine-terminating glaciers quickly respond to lubricated beds with increased flow speed, but the underlying physics, especially how this vulnerability relates to glacier geometry and flow characteristics, remains unclear. This paper presents a 1D physical framework for glacier dynamic vulnerability assuming sudden basal lubrication as an initial perturbation. In this new model, two quantities determine the scale and the areal extent of the subsequent thinning and acceleration after the bed is lubricated: Péclet number (Pe) and the product of glacier speed and thickness gradient (dubbed J0 in this study). To validate the model, this paper calculates Pe and J0 using multi-sourced data from 1996 to 1998 for outlet glaciers in the Greenland ice sheet and Austfonna ice cap, Svalbard, and compares the results with the glacier speed change during 1996/1998–2018. Glaciers with lower Pe and J0 are more likely to accelerate during this 20-year span than those with higher Pe and J0, which matches the model prediction. A combined factor of ice thickness, surface slope, and initial flow speed physically determines how much and how fast glaciers respond to lubricated beds in terms of speed, elevation, and terminus change. 
    more » « less