Periodic Reporting for period 2 - FRAGILE (Next generation framework for global glacier forecasting)
Reporting period: 2022-08-01 to 2024-01-31
Glacier retreat has become increasingly tangible in many mountain regions around the globe. The associated ice loss has dominated the cryospheric contribution to sea-level rise in the past and will continue to do so for many decades under climatic warming. Apart from the sea-level relevance, future glacier retreat will have consequences for seasonal freshwater availability during dry seasons and the remnant ice-free valleys are prone to increasing risks from natural hazards such as slope instabilities or glacial lake outburst floods. To give reliable guidance to decision/policy makers, I therefore envision a portable and coherent ice-dynamic forecasting framework for global glacier evolution. For the first time, each glacier on Earth is fully represented as a three-dimensional body evolving within the mountain landscape. In this way, the rapidly growing body of information from satellite remote sensing can be directly utilised. The heart of this framework is a systematic data assimilation that operates sequentially and considers measurement as they become available. This will streamline and increase the total information flow into glacier models and, consequently, lift the confidence in glacier forecasting for this century to new heights. Furthermore, the project intends to step forward in terms of refining the representation of the decisive physical processes that ultimately control glacier evolution. These refinements comprise a more realistic description of the local energy balance at the glacier surface, which ensures multi-decadal stability in the melt formulation, as well as a fully interactive module for tracing iceberg calving and frontal ablation, capable of predicting phases of fast glacier retreat over critical section of over-deepened bathymetry. Finally, a key for reliable future projections is the ice volume at present. As no direct thickness observations are available for the large majority of glaciers, their ice-thickness distribution is not well known. I therefore forward a thickness mapping procedure, which is calibrated to virtually all glaciers worldwide – again using multi-temporal satellite information.
During the initial project phase, the main focus was on the integration of the glacier-system model into an adequate data assimilation framework. For this purpose, critical target quantities were selected and suitable observations were identified for the actual analysis steps during the data assimilation. Moreover, particular attention was given to the analysis frequencies, to constraining prior uncertainties as well as to computational efficiency. This substantial development effort resulted in a first successful application to a typical Alpine valley glacier, still using synthetic observations. As observations are sequentially assimilated, an accurate tracing of glacier evolution in the past and a best representation of the glacier state at present was achieved. In this way, we took the first major step for streamlining the information flow from Earth observations into glacier-system models as well as for generating a self-consistent model initialisation into present day for seamless future projections.
Secondary activities relate to a refinement of the energy balance at the glacier surface as well as to improving global ice-thickness mapping by glacier-specific calibration to past retreat. The former comprises a melt-model implementation following a simplified energy-balance approach. For this implementation, we deliberately decided to include topographic shading, indirect radiation as well as clouding. A pending task remains the inclusion of the influence of supraglacial debris on ice melt. An empirical description of snow drift is, however, already available. First tests on regional scales have been conducted in the European Alps and in South America. Turning to thickness mapping, an existing state-of-the-art mass-conservations approach was improved by viscosity re-scaling, which further exploits slope, elevation and outline information. Moreover, it has been shown that glacier retreat can be used as a valuable source for past thickness information. This information is provided by satellite remote sensing and therefore available for each glacier on this planet. Calibration and performance tests have been presented in the Swiss Alps and a new thickness map product was released for the entire European Alps.
Secondary activities relate to a refinement of the energy balance at the glacier surface as well as to improving global ice-thickness mapping by glacier-specific calibration to past retreat. The former comprises a melt-model implementation following a simplified energy-balance approach. For this implementation, we deliberately decided to include topographic shading, indirect radiation as well as clouding. A pending task remains the inclusion of the influence of supraglacial debris on ice melt. An empirical description of snow drift is, however, already available. First tests on regional scales have been conducted in the European Alps and in South America. Turning to thickness mapping, an existing state-of-the-art mass-conservations approach was improved by viscosity re-scaling, which further exploits slope, elevation and outline information. Moreover, it has been shown that glacier retreat can be used as a valuable source for past thickness information. This information is provided by satellite remote sensing and therefore available for each glacier on this planet. Calibration and performance tests have been presented in the Swiss Alps and a new thickness map product was released for the entire European Alps.
The current generation of forecasting models with regional to global-scale applicability mostly keeps on relying on important simplifications of the glacier geometry. These simplifications impede a direct utilisation of satellite remote-sensing products that measures glaciers in their full three-dimensional extent. Observations therefore have to be adjusted to match the actually modelled variables, implying a loss of information. In this sense, we tread new ground with the envisaged 3D portable glacier forecasting approach and its deliberate integration in a systematic transient data assimilation framework. Therefore, the first successful application to a single Alpine valley glacier is an immense leap forward to create digital glacier twins in an automated procedure. Currently the methodological decisions on prior uncertainties, definitions of observables and selection of target parameters are being be optimised. In the future, I expect one publication as proof-of-concept and a consecutive high-impact article on a comprehensive real-world glacier application - possibly already on regional scales.
The two existing global maps of glacier ice thickness did rely on direct measurements from the most comprehensive repository which however only holds observations on two percent of all glaciers worldwide. A comparison to unconsidered thickness measurements reveals, in some regions, that measures for the average of or the spread in the relative thickness differences remain important. It is therefore imperative to exploit all available information sources on ice thickness including past values inferred from remotely-sensed glacier retreat. In a pioneering study in the European Alps, we have quantified glacier retreat back to the 1970s. This wealth of information has been shown to be highly beneficial to constrain local ice thickness away from observations. The application of this calibration strategy is unprecedented on regional scales. Furthermore, the respective study was completed by presenting new map of glacier ice thickness for the entire European Alps - with retreat information being imprinted. Pending tasks are to quantify glacier retreat on global scales, derive ice thickness in retreat areas and transfer the calibrated mapping procedure to all glacierised regions.
Concerning the expected results until the end of the project, I want to briefly highlight two further aspects. The first one relates to the above-described process-related development work on the surface mass balance module – a key component of the modelling framework. For full operability on regional scales, a trade-off has to be found between data size of required atmospheric input and time-stepping. For this purpose, statistical properties of atmospheric variables will be scrutinised on various time-scales. The second aspect relates to the reliable representation of calving front migration of marine- and lake-terminating glaciers. As they can show phases of fast retreat and thereby be important for future volume loss, accurate tracing of the ice-front needs consideration. For this purpose, we pursue pertinent sub-grid tracking tools and will combine them with state-of-the-art calving criteria.
The two existing global maps of glacier ice thickness did rely on direct measurements from the most comprehensive repository which however only holds observations on two percent of all glaciers worldwide. A comparison to unconsidered thickness measurements reveals, in some regions, that measures for the average of or the spread in the relative thickness differences remain important. It is therefore imperative to exploit all available information sources on ice thickness including past values inferred from remotely-sensed glacier retreat. In a pioneering study in the European Alps, we have quantified glacier retreat back to the 1970s. This wealth of information has been shown to be highly beneficial to constrain local ice thickness away from observations. The application of this calibration strategy is unprecedented on regional scales. Furthermore, the respective study was completed by presenting new map of glacier ice thickness for the entire European Alps - with retreat information being imprinted. Pending tasks are to quantify glacier retreat on global scales, derive ice thickness in retreat areas and transfer the calibrated mapping procedure to all glacierised regions.
Concerning the expected results until the end of the project, I want to briefly highlight two further aspects. The first one relates to the above-described process-related development work on the surface mass balance module – a key component of the modelling framework. For full operability on regional scales, a trade-off has to be found between data size of required atmospheric input and time-stepping. For this purpose, statistical properties of atmospheric variables will be scrutinised on various time-scales. The second aspect relates to the reliable representation of calving front migration of marine- and lake-terminating glaciers. As they can show phases of fast retreat and thereby be important for future volume loss, accurate tracing of the ice-front needs consideration. For this purpose, we pursue pertinent sub-grid tracking tools and will combine them with state-of-the-art calving criteria.