The global glacier evolution framework

FAU own research funding: EFI / IZKF / EAM ...


Start date : 01.07.2019

End date : 30.06.2020

Website: https://www.fau.eu/research/outstanding-individual-research/emerging-talents-initiative/grantees/


Project details

Short description

Glacier retreat has become palpable all around the globe and it became emblematic for the imminent effects of the current climatic warming. Today, satellite imagery enables us to accurately delineate outlines and areal changes of any glacier worldwide. Yet to this day, the ice volume and thus the amount of melt-water stored in glaciers is not well known. Measurements exist on less than 1% of all glaciers. In many regions, glacial melt-water is an important factor in year-round river runoff. Reliable projections of future glacier evolution are therefore essential for assessing future changes in seasonal fresh-water availability. Existing projections certainly provide valuable first estimates for future melting but the simulations strong limitations. First, the glacier ice volume at present is not well known and recent global estimates differ substantially. This uncertainty is propagated by volume projections that depend to a great extent on the present-day glacier volume. Most global projections relied on simple scaling relations between glacier area and ice volume. In this way, geometric glacier changes were not directly quantifiable and the future ice-front retreat or length changes were inferred by another area-scaling relation. Second, the past and future climatic forcing for global projections was only available from coarse global climate models, not capable of resolving the complex topographic setting of glaciers in mountain ranges. Temperature and precipitation values, representative for an area comparable to the size of Corsica, were downscaled to sub-kilometre scales using simple lapse-rates, deliberately ignoring many aspects of the micro-climatic conditions. Third, the ice-dynamic response of glaciers to future climatic warming was never explicitly taken into account on global scale. For land-terminating glaciers, dynamic changes only affect the mass redistribution and they are thus expected to be of secondary importance for ice loss. For marine-terminating glaciers however, ice-dynamics controls the amount of ice directly lost to lakes or the ocean by iceberg calving. For the global projections, this term had to be ignored because it requires the application of ice-flow models that require information on the poorly known subglacial topography. In summary, existing projections of glacier evolution suffer from three major source terms of uncertainty: the generally poor knowledge of glacier ice thickness, the badly constrained local climatic conditions at present or under future warming and the unaccounted influence of ice-dynamic changes.

Within this project, we aim at crossing present frontiers in global projections of glacier evolution by means of an innovative self-consistent projection framework that will naturally surmount previous limitations. We plan to primarily target the evolution within this century. The entire modelling framework will be streamlined in a single programming environment building on the open-source Elmer/Ice software. The key difference to existing approaches is that each glacier will be treated as a three-dimensional entity embedded in the complex surrounding topography. In this way, we will not only be able to determine past and future changes in ice volume and area but it will be possible to directly follow frontal retreat, separation of glacier branches and exposition of over-deepened bed segments that could potentially be filled by pro-glacial lakes. Knowing the detailed retreat pattern, glacier-related hazards such as the destabilisation of ice-free mountain flanks or lake-outburst floods can better be assessed. For now, such assessments are highly limited on the basis of existing techniques because of the typical geometric reduction to flowlines or even to scalar quantities, e.g. glacier volume, area and length. Despite the asset of resolving the full glacier geometry, the projection framework will be designed such that uncertainties, related to the three major source terms, identified above, will be reduced. For this purpose, we formulate the following three milestone packages.

The ice thickness and thus the basal topography beneath glaciers is the central quantity for the application of the modelling framework. So, the first milestone (MS1) of this project is to produce reliable maps of glacier ice thickness on global scales. Such map products already exist but most underlying approaches, though calibrated to measured ice thickness, do not necessarily reproduce these observations. In this respect, the applicant forwarded a data assimilation variant for mapping glacier ice thickness that readily accounts for available thickness measurements. Its performance was tested for ice-cap as well as marine- and land-terminating glacier geometries on Svalbard. The approach has been automated for regional applicability and an archipelago-wide thickness map has been presented for all glaciers on Svalbard. Moreover, the global scale was already targeted by contributing to a worldwide multi-model consensus estimate of glacier ice thickness. Though complete global coverage was not yet attained, our approach is unique in the capability to reproduce available thickness measurements. As an important side product, the modelling framework already comprises a module that automatically generates unstructured meshes on any glacier worldwide. The modelling activities in this project can therefore be directly started on single-glacier to regional scales. To improve the reconstruction performance, the approach requires further refinement. Development leads are the usage of prior information on the surrounding ice-free valley topography to better constrain the reconstruction, the calculation of generic thickness observations from glacier retreat in the satellite era or the improvement of error estimates associated to the reconstructed thickness field. Regional coverage of glacier surface velocities and elevation changes has by now become feasible and it is only a matter of time until global coverage is achieved. Both quantities will be valuable additions to refine thickness reconstruction approaches. To closely follow up on and profit from the fast development in satellite observations, the FAU remote sensing group is the ideal host for this milestone activity because of its glaciological focus.

The second milestone (MS2) relates to the poor quality of the climatic forcing and the resulting simplicity of the melt models applied in existing global glacier volume projections. We therefore target these two shortcomings to define the surface mass balance module within our modelling framework. First, we will directly profit from the efforts taken in the Coordinated Regional Climate Downscaling Experiment (CORDEX) framework of the World Climate Research Programme. CORDEX will produce high-resolution, past and future climatic forcing on regional scales and is thus more reliable over mountainous areas. The forcing variables will be refined by a-priori downscaling before entering a parametric melt model that resolves the 3D nature of drainage basin geometries. An adequate melt-model variant will be selected and implemented in our modelling framework. On a global scale, both the suggested forcing scheme and the melt-model setup will represent pioneer work. The activities to reach this milestone comprise an optimal selection of the climate model forcing, the application of statistical downscaling techniques as well as the regional calibration and validation of the surface mass balance module against observations. Again, the FAU geography institute is an optimal choice as host because of its climate system research group. The group is specialised in glaciology and has long-standing expertise in surface mass balance processes on glaciers as well as in state-of-the-art climatic downscaling.

The final third milestone (MS3) is to include ice-dynamic effects in the future projections of global glacier evolution. On regional and global scales, efficient solutions have to be found in terms of model initialisation, model complexity and forward simulations. Concerning model initialisation, the first step is to assimilate surface velocity measurements to infer poorly constrained friction parameters at the glacier base. Thereafter, an equilibration has to be conducted to reduce spurious initial model drift. For computationally efficient forward simulation, common ice-dynamic approximation will be employed dependent on the basal conditions. For marine terminating glaciers, iceberg calving will be implemented following a successfully applied level-set method. Though the calving implementation will require some calibration and validation, a basic level-set routine is already available in Elmer/Ice. For the forward simulations under changing climatic conditions, efficient strategies have to be found to mutually exchange information between the ice-dynamic and the surface mass balance module. This milestone is likely the most demanding part of this project. To guarantee its viability and to cap associated risks, work tasks are kept largely independent in all milestone activities. Therefore, we deliberately envisage a step-by-step build-up. First, the initialisation strategy is adopted to a well-controllable land-terminating glacier. Thereafter, iceberg calving is implemented for a well-studied marine terminating glacier and, if required, the initialisation strategy is amended. Thereafter, the whole modelling framework is transferred to other glaciers. Largest differences in glacier evolution with respect to previous projections are expected for marine-terminating geometries where ice-dynamics directly controls mass loss via iceberg calving.

For all milestone achievements, activities will first concentrate on a single mountain-range region. The actual choice will be made betimes with regard to feasibility, utility and public relevance. Particularities in some regions, as heavy debris cover in the Himalaya, strong climatic gradients over Patagonia, significant lateral avalanche mass input, etc. will require region-specific refinements. The transition to global scales will be done in a modular way starting with the thickness product, followed by static SMB simulations and the ultimate dynamically coupled projections. Pre-defined fall-back strategies for all activities guarantee mutual independence to attain the goals defined by each milestone.

Scientific Abstract

Glacier retreat has become palpable all around the globe and it became emblematic for the imminent effects of the current climatic warming. Today, satellite imagery enables us to accurately delineate outlines and areal changes of any glacier worldwide. Yet to this day, the ice volume and thus the amount of melt-water stored in glaciers is not well known. Measurements exist on less than 1% of all glaciers. In many regions, glacial melt-water is an important factor in year-round river runoff. Reliable projections of future glacier evolution are therefore essential for assessing future changes in seasonal fresh-water availability. Existing projections certainly provide valuable first estimates for future melting but the simulations strong limitations. First, the glacier ice volume at present is not well known and recent global estimates differ substantially. This uncertainty is propagated by volume projections that depend to a great extent on the present-day glacier volume. Most global projections relied on simple scaling relations between glacier area and ice volume. In this way, geometric glacier changes were not directly quantifiable and the future ice-front retreat or length changes were inferred by another area-scaling relation. Second, the past and future climatic forcing for global projections was only available from coarse global climate models, not capable of resolving the complex topographic setting of glaciers in mountain ranges. Temperature and precipitation values, representative for an area comparable to the size of Corsica, were downscaled to sub-kilometre scales using simple lapse-rates, deliberately ignoring many aspects of the micro-climatic conditions. Third, the ice-dynamic response of glaciers to future climatic warming was never explicitly taken into account on global scale. For land-terminating glaciers, dynamic changes only affect the mass redistribution and they are thus expected to be of secondary importance for ice loss. For marine-terminating glaciers however, ice-dynamics controls the amount of ice directly lost to lakes or the ocean by iceberg calving. For the global projections, this term had to be ignored because it requires the application of ice-flow models that require information on the poorly known subglacial topography. In summary, existing projections of glacier evolution suffer from three major source terms of uncertainty: the generally poor knowledge of glacier ice thickness, the badly constrained local climatic conditions at present or under future warming and the unaccounted influence of ice-dynamic changes.

Within this project, we aim at crossing present frontiers in global projections of glacier evolution by means of an innovative self-consistent projection framework that will naturally surmount previous limitations. We plan to primarily target the evolution within this century. The entire modelling framework will be streamlined in a single programming environment building on the open-source Elmer/Ice software. The key difference to existing approaches is that each glacier will be treated as a three-dimensional entity embedded in the complex surrounding topography. In this way, we will not only be able to determine past and future changes in ice volume and area but it will be possible to directly follow frontal retreat, separation of glacier branches and exposition of over-deepened bed segments that could potentially be filled by pro-glacial lakes. Knowing the detailed retreat pattern, glacier-related hazards such as the destabilisation of ice-free mountain flanks or lake-outburst floods can better be assessed. For now, such assessments are highly limited on the basis of existing techniques because of the typical geometric reduction to flowlines or even to scalar quantities, e.g. glacier volume, area and length. Despite the asset of resolving the full glacier geometry, the projection framework will be designed such that uncertainties, related to the three major source terms, identified above, will be reduced. For this purpose, we formulate the following three milestone packages.

The ice thickness and thus the basal topography beneath glaciers is the central quantity for the application of the modelling framework. So, the first milestone (MS1) of this project is to produce reliable maps of glacier ice thickness on global scales. Such map products already exist but most underlying approaches, though calibrated to measured ice thickness, do not necessarily reproduce these observations. In this respect, the applicant forwarded a data assimilation variant for mapping glacier ice thickness that readily accounts for available thickness measurements. Its performance was tested for ice-cap as well as marine- and land-terminating glacier geometries on Svalbard. The approach has been automated for regional applicability and an archipelago-wide thickness map has been presented for all glaciers on Svalbard. Moreover, the global scale was already targeted by contributing to a worldwide multi-model consensus estimate of glacier ice thickness. Though complete global coverage was not yet attained, our approach is unique in the capability to reproduce available thickness measurements. As an important side product, the modelling framework already comprises a module that automatically generates unstructured meshes on any glacier worldwide. The modelling activities in this project can therefore be directly started on single-glacier to regional scales. To improve the reconstruction performance, the approach requires further refinement. Development leads are the usage of prior information on the surrounding ice-free valley topography to better constrain the reconstruction, the calculation of generic thickness observations from glacier retreat in the satellite era or the improvement of error estimates associated to the reconstructed thickness field. Regional coverage of glacier surface velocities and elevation changes has by now become feasible and it is only a matter of time until global coverage is achieved. Both quantities will be valuable additions to refine thickness reconstruction approaches. To closely follow up on and profit from the fast development in satellite observations, the FAU remote sensing group is the ideal host for this milestone activity because of its glaciological focus.

The second milestone (MS2) relates to the poor quality of the climatic forcing and the resulting simplicity of the melt models applied in existing global glacier volume projections. We therefore target these two shortcomings to define the surface mass balance module within our modelling framework. First, we will directly profit from the efforts taken in the Coordinated Regional Climate Downscaling Experiment (CORDEX) framework of the World Climate Research Programme. CORDEX will produce high-resolution, past and future climatic forcing on regional scales and is thus more reliable over mountainous areas. The forcing variables will be refined by a-priori downscaling before entering a parametric melt model that resolves the 3D nature of drainage basin geometries. An adequate melt-model variant will be selected and implemented in our modelling framework. On a global scale, both the suggested forcing scheme and the melt-model setup will represent pioneer work. The activities to reach this milestone comprise an optimal selection of the climate model forcing, the application of statistical downscaling techniques as well as the regional calibration and validation of the surface mass balance module against observations. Again, the FAU geography institute is an optimal choice as host because of its climate system research group. The group is specialised in glaciology and has long-standing expertise in surface mass balance processes on glaciers as well as in state-of-the-art climatic downscaling.

The final third milestone (MS3) is to include ice-dynamic effects in the future projections of global glacier evolution. On regional and global scales, efficient solutions have to be found in terms of model initialisation, model complexity and forward simulations. Concerning model initialisation, the first step is to assimilate surface velocity measurements to infer poorly constrained friction parameters at the glacier base. Thereafter, an equilibration has to be conducted to reduce spurious initial model drift. For computationally efficient forward simulation, common ice-dynamic approximation will be employed dependent on the basal conditions. For marine terminating glaciers, iceberg calving will be implemented following a successfully applied level-set method. Though the calving implementation will require some calibration and validation, a basic level-set routine is already available in Elmer/Ice. For the forward simulations under changing climatic conditions, efficient strategies have to be found to mutually exchange information between the ice-dynamic and the surface mass balance module. This milestone is likely the most demanding part of this project. To guarantee its viability and to cap associated risks, work tasks are kept largely independent in all milestone activities. Therefore, we deliberately envisage a step-by-step build-up. First, the initialisation strategy is adopted to a well-controllable land-terminating glacier. Thereafter, iceberg calving is implemented for a well-studied marine terminating glacier and, if required, the initialisation strategy is amended. Thereafter, the whole modelling framework is transferred to other glaciers. Largest differences in glacier evolution with respect to previous projections are expected for marine-terminating geometries where ice-dynamics directly controls mass loss via iceberg calving.

For all milestone achievements, activities will first concentrate on a single mountain-range region. The actual choice will be made betimes with regard to feasibility, utility and public relevance. Particularities in some regions, as heavy debris cover in the Himalaya, strong climatic gradients over Patagonia, significant lateral avalanche mass input, etc. will require region-specific refinements. The transition to global scales will be done in a modular way starting with the thickness product, followed by static SMB simulations and the ultimate dynamically coupled projections. Pre-defined fall-back strategies for all activities guarantee mutual independence to attain the goals defined by each milestone.

Involved:

Contributing FAU Organisations: