a1), but when conditions deviate from this mean training data centroid, the Lasso can only linearly approximate the extremes based on the linear trend set on the main cluster of average values (Fig. Millan, R., Mouginot, J., Rabatel, A., & Morlighem, M. Ice velocity and thickness of the worlds glaciers. Reanalysis of 47 Years of Climate in the French Alps (19582005): Climatology and Trends for Snow Cover. S5b). Nature Geosciences, https://doi.org/10.1038/s41561-021-00885-z (2022). a1 throughout the whole century under RCP 4.5, with glacier retreat to higher elevations (positive effect on MB) compensating for the warmer climate (negative effect on MB). Hydrol. J. Hosp. Nat. In order to improve the comparability between both models, a MB bias correction was applied to GloGEMflows simulated MB, based on the average annual MB difference between both models for the 20032015 period (0.4m.w.e. Such ice caps cannot retreat to higher elevations in a warming climate, which inhibits this positive impact on MB40 (Fig. Anyone you share the following link with will be able to read this content: Sorry, a shareable link is not currently available for this article. New research suggests that climate change-induced melting of the Nisqually Glacier near Seattle, Wash., and other high-elevation glaciers will offset seasonal declines in streamflow until. From this behavior, inferences of past climate can be drawn. the Open Global Glacier Model - OGGM9) is likely to be less affected by an over-sensitivity to future warming than a more complex model with dedicated DDFs for ice, snow, and firn. 3). During the last decade, various global glacier evolution models have been used to provide estimates on the future sea-level contribution from glaciers7,8. Thus, glacier sensitivity to a step change in climate , glacier response to climate trends , and glacier variance driven by stochastic climate fluctuations are all proportional to , making an important number to constrain. Several aquatic and terrestrial ecosystems depend on these water resources as well, which ensure a base runoff during the warmest or driest months of the year6. On the other hand, ice caps present a different response to future warming, with our results suggesting a negative MB bias by models using linear PDD and accumulation relationships. The two recent iterations of the Glacier Model Intercomparison Project (GlacierMIP7,8) have proved a remarkable effort to aggregate, compare and understand global glacier evolution estimates and their associated uncertainties. 4 vs.S5). The source code of the glacier model can be freely accessed in the following repository: https://github.com/JordiBolibar/ALPGM. Studies have warned about the use of temperature-index models for snow and ice projections under climate change for decades34,35,36. Ice thickness accuracy varied significantly, with an overall correct representation of the ice distribution but with local biases reaching up to 100%. Thank you for visiting nature.com. Particularly in Asia, water demand exceeds supply due to rapid population growth, with glacier . 4). Change 120, 2437 (2014). Six, D. & Vincent, C. Sensitivity of mass balance and equilibrium-line altitude to climate change in the French Alps. This type of model uses a calibrated linear relationship between positive degree-days (PDDs) and the melt of ice or snow11. In order to investigate the implications of these results for flat glaciers, we performed additional synthetic experiments in order to reproduce this lack of topographical feedback (Fig. As for the MB modelling approach, a detailed explanation on this method can be found in a previous dedicated paper on the methods31. This synthetic setup allowed us to reproduce the climatic conditions to be undergone by most ice caps, with their mean surface altitude hardly evolving through time. Monitoring the Seasonal hydrology of alpine wetlands in response to snow cover dynamics and summer climate: a novel approach with sentinel-2. Climate variations change a glacier's mass balance by affecting ablation and accumulation amounts. Climatol. 4 ). Front. To interactively describe to response of glaciers to climate change, a glacier parameterization scheme has been developed and implemented into the regional climate model REMO. Average cumulative MB projections of French Alpine glaciers with a nonlinear deep learning vs. a linear Lasso model for 29 climate scenarios; a with topographical feedback (allowing for glacier retreat) and e without topographical feedback (synthetic experiment with constant mean glacier altitude). how climate change and glacier retreat are reshaping whole aquatic ecosystems, there is a need to develop an integrated understanding spanning multiple taxonomic groups and trophic levels in glacier-fed rivers (e.g., bacteria, protists, fungi, algae, diatoms, invertebrates, mammals, amphibians, and fish; Clitherow et al. 4a). The vertical blue and red lines indicate the distribution of extreme (top 5%) values for all 21st century projected climate scenarios, with the mean value in the center and 1 indicated by dashed lines. The training was performed with an RMSprop optimizer, batch normalization46, and we used both dropout and Gaussian noise in order to regularize it. This implies that specific climatic differences between massifs can be better captured by ALPGM than GloGEMflow. This approach is known as a cross-validation ensemble49. However, many glacierized regions in the world present different topographical setups, with flatter glaciers, commonly referred to as ice caps, covering the underlying terrain39. This dataset applies a statistical adjustment specific to French mountain regions based on the SAFRAN dataset, to EURO-CORDEX26 GCM-RCM-RCP members, covering a total of 29 different future climate scenarios for the 20052100 period. Limnol. & Funk, M. A comparison of empirical and physically based glacier surface melt models for long-term simulations of glacier response. Zemp, M. et al. The high spatial resolution enables a detailed representation of mountain weather patterns, which are often undermined by coarser resolution climate datasets. A comprehensive bibliography of scientific publications relating to the glacier is included. The application of a non-linear back-propagation neural network to study the mass balance of Grosse Aletschgletscher, Switzerland. Scand. Fr Hydrobiol. Our results serve as a strong reminder that the outcomes of existing large-scale glacier simulations should be interpreted with care, and that newly available techniques (such as the nonlinear deep learning approach presented here) and observations (e.g. The nonlinearities present in the simulated annual glacier-wide MB values were assessed by running two different glacier simulations with two different MB models. In order to investigate the effects of MB nonlinearities on flatter glaciers, we conducted a synthetic experiment using the French Alps dataset. The maximum downvalley position of the glacier is marked by either a J. Glaciol. Earths Future https://doi.org/10.1029/2019EF001470 (2020). energy balance), with differences increasing when the conditions considerably differ from the calibration period33. Three different types of cross validation were performed: a Leave-One-Glacier-Out (LOGO), a Leave-One-Year-Out (LOYO) and a Leave-Some-Years-and-Glaciers-Out (LSYGO). 2015 IEEE Int. 51, 573587 (2005). a Projected mean glacier altitude evolution between 2015 and 2100. Therefore, we were capable of isolating the different behaviours of the nonlinear deep learning model and a linear machine learning model based on the Lasso30. Interestingly, this matches the nonlinear, less sensitive response to summer snowfall in the ablation season of our deep learning model (Fig. The advantage of this method is that by only changing the MB model, we can keep the rest of the model components (glacier dynamics and climate forcing) and parameters the same in order to have a controlled environment for our experiment. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate. By monitoring the change in size of glaciers around the world, scientists can learn about global climate change. Tom R. Andersson, J. Scott Hosking, Emily Shuckburgh, Shfaqat A. Khan, Anders A. Bjrk, Toni Schenk, Romain Hugonnet, Robert McNabb, Andreas Kb, Atanu Bhattacharya, Tobias Bolch, Tandong Yao, Christian Sommer, Philipp Malz, Matthias H. Braun, Romain Millan, Jrmie Mouginot, Mathieu Morlighem, Matthias H. Braun, Philipp Malz, Thorsten C. Seehaus, Nature Communications By performing glacier projections both with mountain glaciers in the French Alps and a synthetic experiment reproducing ice cap-like behaviour, we argue that the limitations identified here for linear models will also have implications for many other glacierized regions in the world. Ice-surface altitude changes of as much as 25 meters occurred between 1944 and 1955. Together with recent findings by another study41 highlighting the increased uncertainties in ice thickness distribution estimates of ice caps compared to mountain glaciers, our results raise further awareness on the important uncertainties in glacier projections for ice caps. We previously demonstrated that this period is long enough to represent the secular trend of glacier dynamics in the region31. Univ. This is well in agreement with the known uncertainties of glacier evolution models, with glacier ice thickness being the second largest uncertainty after the future GCM-RCM-RCP climate members used to force the model29. The estimated ice thickness for Mer de Glace (28.87km2 in 2015) was increased by 25% in order to correct the bias with respect to field observations31. Advances occurred from 1963-68 and from 1974-79. (a) Topographical predictors were computed based on the glaciers annually updated digital elevation model (DEM). Verfaillie, D., Dqu, M., Morin, S. & Lafaysse, M. The method ADAMONT v1.0 for statistical adjustment of climate projections applicable to energy balance land surface models. The machine learning models used in this study are useful to highlight and quantify how nonlinearities in MB affect climate-glacier interactions, but are limited in terms of process understanding. Secure .gov websites use HTTPS A lock ( ) or https:// means you've safely connected to the .gov website. Sci. S5cf), except for the largest glaciers (e.g. Through synthetic experiments, we showed that the associated uncertainties are likely to be even more pronounced for ice caps, which host the largest reserves of ice outside the two main ice sheets32. 1). The climatic forcing comes from high-resolution climate ensemble projections from 29 combinations of global climate models (GCMs) and regional climate models (RCMs) adjusted for mountain regions for three Representative Concentration Pathway (RCP) scenarios: 2.6, 4.5, and 8.525. Ecography 40, 913929 (2017). Simulations for projections in this study were made by generating an ensemble of 60 cross-validated models based on LSYGO. In that study, a temperature-index model with a separate degree-day factor (DDF) for snow and ice is used, resulting in piecewise linear functions able to partially reproduce nonlinear MB dynamics. This oversensitivity directly results from the fact that temperature-index models rely on linear relationships between PDDs and melt and that these models are calibrated with past MB and climate data. Therefore, their sensitivities to the projected 21st century increase in PDDs are linear. Overall, this results in linear MB models overestimating both extreme positive (Fig. 4). Regarding air temperature forcings, the linear Lasso MB model was found to be slightly under-sensitive to extreme positive cumulative PDD (CPDD) and over-sensitive to extreme negative CPDDs. Deep artificial neural networks (ANNs) are nonlinear models that offer an alternative approach to these classic methods. In the meantime, to ensure continued support, we are displaying the site without styles 4), as the linear model tends to over-estimate positive MB rates both from air temperature and snowfall (Fig. (Photograph by Klaus J. Bayr, Keene State College, 1990) One method of measuring glaciers is to send researchers onto the ice with . 3c). Paul, F., Kb, A., Maisch, M., Kellenberger, T. & Haeberli, W. Rapid disintegration of Alpine glaciers observed with satellite data: disintegration of alpine glaciers. Geophys. By 2100, under RCP 4.5, these two high-altitude massifs are predicted to retain on average 26% and 13% of their 2015 volume, respectively, with most of the ice concentrated in a few larger glaciers (>1km2, Fig. Farinotti, D., Round, V., Huss, M., Compagno, L. & Zekollari, H. Large hydropower and water-storage potential in future glacier-free basins. "Their numbers have gone from regularly exceeding 50,000 adult salmon in the Nisqually to about 5,000 last year." The Nisqually River near its glacial origins. We further assessed the effect of MB nonlinearities by comparing our simulated glacier changes with those obtained from other glacier evolution studies from the literature, which rely on temperature-index models for MB modelling. 3c), which is directly linked to summer air temperatures and has a strong influence on surface albedo. This behaviour is not observed with the nonlinear model, hinting at a positive bias of linear MB models under RCP 2.6. This creates a total of 34 input predictors for each year (7 topographical, 3 seasonal climate, and 24 monthly climate predictors). Jordi Bolibar. Article Rainier, Washington. deep artificial neural networks) glacier evolution projections by modelling the regional evolution of French alpine glaciers through the 21st century. 1a). Through these surveys "bulges" have been tracked as they travel down the glacier (c). Conf. Lett. 12, 168173 (2019). Interestingly, our analysis indicates that more complex models using separate DDFs for ice, firn and snow might introduce stronger biases than more simple models using a single DDF. However, as shown in our previous work and confirmed here, the accuracy of linear models drastically drops as soon as the input climate data diverges from the mean cluster of values used for training. "Such glaciers spawn icebergs into the ocean or lakes and have different dynamics from glaciers that end on land and melt at their front ends. Multiple copies of this dataset were created, and for each individual copy a single predictor (i.e. S7). The two models with linear MB responses to PDDs and accumulation simulate more positive MB rates under RCP 2.6, highlighting their over-sensitivity to negative air temperature anomalies and positive snowfall anomalies (Fig. Durand, Y. et al. Regarding air temperature, a specific CPDD anomaly ranging from 1500 PDD to +1500 PDD in steps of 100 PDD was prescribed to all glaciers for each dataset copy. Glaciers are large-scale, highly sensitive climate instruments which, ideally, should be picked up and weighed once a year. Nisqually Glacier is perhaps the most visited, best-surveyed glacier on Mount Rainier. 2008. Both models agree around the average values seen during training (i.e. Model Dev. Bolibar, J., Rabatel, A., Gouttevin, I. et al. We performed a validation simulation for the 20032015 period by running our model through this period and comparing the simulated glacier surface area of each of the 32 glaciers with MB to observations from the 2015 glacier inventory16,52. The model output data generated in this study have been deposited in netCDF and CSV format in a Zenodo repository under accession code Creative Commons Attribution 4.0 International. GloGEMflow10 is a state-of-the-art global glacier evolution model used in a wide range of studies, including the second phase of GlacierMIP7,8. This creates an interesting dilemma, with more complex temperature-index MB models generally outperforming simpler models for more climatically homogeneous past periods but introducing important biases for future projections under climate change. Ice caps in the Canadian Arctic, the Russian Arctic, Svalbard, and parts of the periphery of Greenland are major reservoirs of ice, as well as some of the biggest expected contributors to sea level rise outside the two polar ice sheets7. This experiment enabled the exploration of the response to specific climate forcings of a wide range of glaciers of different topographical characteristics in a wide range of different climatic setups, determined by all meteorological conditions from the years 19672015 (Fig. Xu, B., Wang, N., Chen, T. & Li, M. Empirical Evaluation of Rectified Activations in Convolutional Network. Annual glacier-wide mass balance (MB) is estimated to remain stable at around 1.2m.w.e. (Springer, New York, 2009). 0.78m.w.e. Additionally, the specific responses of the deep learning and Lasso MB models to air temperature and snowfall were extracted by performing a model sensitivity analysis. Appl. Both MB models were trained with exactly the same data, and all other glacier model parameters were unchanged in order to allow isolating the effects of the nonlinearities in the MB. . Here, we compare our results with those from a recent study that focused on the European Alps10. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. & Zumbhl, H. J. This allows us to assess the MB models responses at a regional scale to changes in individual predictors (Fig. (Zenodo, 2020). Glacier topography is a crucial driver of future glacier projections and is expected to play an important role in determining the magnitude that nonlinearities will have on the mass balance signal: ice caps and large flatter glaciers are expected to be more influenced by these nonlinear sensitivities than steep mountain glaciers in a warming climate. However, glacier projections under low-emission scenarios and the behaviour of flatter glaciers and ice caps are likely to be biased by mass balance models with linear sensitivities, introducing long-term biases in sea-level rise and water resources projections. This removes the topographical feedback typical from mountain glaciers, and reproduces the more extreme climate conditions that ice caps are likely to endure through the 21st century40. Vertical axes are different for the two analyses. Common climatic signal from glaciers in the European Alps over the last 50 years: Common Climatic Signal in the Alps. ice cap-like behaviour). Consequently, a simple MB model with a single DDF (e.g. snowfall, avalanches and refreezing) and the mass lost via different processes of ablation (e.g. MathSciNet longwave radiation budget, turbulent fluxes), in comparison with a future warmer climate. 1gi)26 and glaciers shrinking to higher elevations where precipitation rates are higher as a result of orographic precipitation enhancement27. Spandre, P. et al. Earth Sci. 4a). 22, 21462160 (2009). & Galiez, C. A deep learning reconstruction of mass balance series for all glaciers in the French Alps: 19672015. Planet. Use the Previous and Next buttons to navigate the slides or the slide controller buttons at the end to navigate through each slide. Nature 577, 364369 (2020). A comparison between the two MB models shows that a nonlinear response to climate forcings is captured by the deep learning MB model, allowing a better representation of glacier mass changes, including significantly reduced biases for extreme values (see Methods). However, to further investigate these findings, experiments designed more towards ice caps, and including crucial mechanisms such as ice-ocean interactions and thermodynamics, should be used for this purpose. Huss, M. & Hock, R. A new model for global glacier change and sea-level rise. This annual geometry adjustment accounts for the effects of glacier retreat on the climate signal received by glaciers. Nevertheless, we previously demonstrated that glacier surface area is not an important predictor of MB changes in our models29, and ice caps evolve mostly through thinning and not shrinking (Fig. Annu. J.B. was supported by a NWO VIDI grant 016.Vidi.171.063. Hock, R. et al. This work was funded by the Labex OSUG@2020 (Investissements davenir, ANR10 LABX56) and the Auvergne-Rhne-Alpes region through the BERGER project. Massifs without glaciers by 2100 are marked with a cross, b Glacier ice volume distribution per massif, with its remaining fraction by 2100 (with respect to 2015), c Annual glacier-wide MB per massif, d Annual snowfall per massif, e Annual cumulative positive degree-days (CPDD) per massif. J.B. developed the main glacier model, performed the simulations, analysed the results, and wrote the paper. the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in 3 (2015). a1) over the French Alps. Our results show that the mean elevation is far more variable than the kinematic ELA ( Fig. H.Z. Res. Earth Syst. As such, these values reflect both the climatic forcing and the changing glacier geometry. 3, 16751685 (2019). In order to investigate the effects of MB nonlinearities on ice caps, we performed the same type of comparison between simulations, but the glacier geometry update module described in the Glacier geometry evolution section was deactivated. Google Scholar. Hock, R. & Huss, M. Glaciers and climate change. Indeed, the projected 21st century warming will lead to increasing incoming longwave radiation and turbulent fluxes, with no marked future trends in the evolution of shortwave radiation37. The Elements of Statistical Learning. 4e). Alpine glaciers, like this one near Mt. Glaciers are important for agriculture, hydropower, recreation, tourism, and biological communities. Average ice velocities on the Nisqually Glacier were previously measured at approximately 200 mm/day (8 in) (Hodge 1974). The largest snow depths measured this spring exceeded 10 meters on Nisqually Glacier and 7 meters on Emmons. In order to avoid overfitting, MB models were thoroughly cross-validated using all data for the 19672015 period in order to ensure a correct out-of-sample performance. Nisqually Glacier is the lengthiest of any made in North America. Meteorol. The Karakoram and the Himalayan mountain range accommodate a large number of glaciers and are the major source of several perennial rivers downstream. However, the impact of different climate configurations, such as a more continental and drier climate or a more oceanic and humid climate, would certainly have an impact on the results, albeit a much less important one than the lack of topographical feedback explored here. The new research suggests that the world's glaciers are disappearing more quickly than scientists previously estimated, and they . The rest of the story appears to lie primarily in the unique dynamic response of the region's glaciers to climate change. Conversely, during the accumulation season, glaciers are mostly covered by snow, with a much higher albedo and a reduced role of shortwave radiation in the MB that will persist even under climate change. On the other hand, for flatter glaciers large differences between deep learning and Lasso are obtained for almost all climate scenarios (Fig. Graphics inspired by Hock and Huss40. Hugonnet, R. et al. Alternatively, the Lasso MB model displayed an RMSE of 0.85m.w.e. Grenoble Alpes, Universit de Toulouse, Mto-France, CNRS, CNRM, Centre dtudes de la Neige, Grenoble, France, Department of Geoscience and Remote Sensing, Delft University of Technology, Delft, Netherlands, Laboratoire de Glaciologie, Universit Libre de Bruxelles, Brussels, Belgium, Univ. 2013). On the one hand, MB nonlinearities for mountain glaciers appear to be only relevant for climate scenarios with a reduction in greenhouse gases emissions (Fig. Additionally, glacier surface area was found to be a minor predictor in our MB models31. Lett. Kinematic waves on glaciers move as several times the speed of the ice as a whole, and are subtle in topographic expression. This ensures that the model is capable of reproducing MB rates for unseen glaciers and years. S6). Strong Alpine glacier melt in the 1940s due to enhanced solar radiation. CoRR abs/1505.00853 (2015). Nature 575, 341344 (2019). The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in a credit line to the material.
Intel Job Application Status,
Bootz Kona Bathtub Right Drain,
Articles N