Academia.eduAcademia.edu

Figure 7 – uploaded by Jostein Bakke

See full PDFdownloadDownload figure
is more sensitive to the interpolation method. Overall, the results from the two tests show that the reconstruction tool generated high quality results, bearing in mind that none of the glaciers used for testing are in present equilibrium with the climate, but demonstrates that multiple flowlines and the use of F factors for user-defined cross- sections is the best approach. In general, and regardless of the approach, one s that the quality of a glacier 3D surface reconstruction is always related to the q parameters. Different input parameter requirements are also needed depending of the glacier system under consideration. For example, a simple, topographical valley palaeoglacier could be well reconstructed simply on the basis of the prese hould never forget uality of the input on the complexity y well-constrained nt-day topography and the position of a frontal moraine. More complex systems, such as an ice field with multiple outlets, ideally require a more extensive knowledge of the original glacier margins in order to obtain a robust reconstruction. For the purpose of testing the quality of the tool, we ideal scenario where these inputs parameters are fully known (i.e. we used the icefield margin) and demonstrated that the tool is able to reconstruct an icefield have assumed the known Folgefonna very similar to the present-day. We have demonstrated here that the tool is capable of providing robust, physically plausible reconstructions, but how well these reproduce any palaeoglacier ultimately depends on the quality of the input parameters, and in particular on the accurate mapping of the ice marginal geomorphology. This is where the expert knowledge of the glacial geomorphologist is paramount.

Figure 7 is more sensitive to the interpolation method. Overall, the results from the two tests show that the reconstruction tool generated high quality results, bearing in mind that none of the glaciers used for testing are in present equilibrium with the climate, but demonstrates that multiple flowlines and the use of F factors for user-defined cross- sections is the best approach. In general, and regardless of the approach, one s that the quality of a glacier 3D surface reconstruction is always related to the q parameters. Different input parameter requirements are also needed depending of the glacier system under consideration. For example, a simple, topographical valley palaeoglacier could be well reconstructed simply on the basis of the prese hould never forget uality of the input on the complexity y well-constrained nt-day topography and the position of a frontal moraine. More complex systems, such as an ice field with multiple outlets, ideally require a more extensive knowledge of the original glacier margins in order to obtain a robust reconstruction. For the purpose of testing the quality of the tool, we ideal scenario where these inputs parameters are fully known (i.e. we used the icefield margin) and demonstrated that the tool is able to reconstruct an icefield have assumed the known Folgefonna very similar to the present-day. We have demonstrated here that the tool is capable of providing robust, physically plausible reconstructions, but how well these reproduce any palaeoglacier ultimately depends on the quality of the input parameters, and in particular on the accurate mapping of the ice marginal geomorphology. This is where the expert knowledge of the glacial geomorphologist is paramount.

Related Figures (13)

www.elsevier.com/locate/cageo WIUUZO-IUUH LO JIU LII-O http://dx.doi.org/10.1016/j.cageo.2016.06.008 CAGEO3778
www.elsevier.com/locate/cageo WIUUZO-IUUH LO JIU LII-O http://dx.doi.org/10.1016/j.cageo.2016.06.008 CAGEO3778
Figure 2. Diagram showing the use of GlaRe. Grey circles are user defined inputs, white squares are
Figure 2. Diagram showing the use of GlaRe. Grey circles are user defined inputs, white squares are
Figure 3. Reconstruction inputs for Folgefonna. The subglacial topography was retrieved from a ground penetrating radar survey. The blue line indicates the present-day ice margin, black lines are the cross-sections for F factor calculation red lines represent the flowlines. Red colour points have a predefined shear stress of 100 kPa and yellow colour points have a 50 kPa shear stress value. ground penetrating radar survey. The blue line indicates the present-day ice margin, black lines are
Figure 3. Reconstruction inputs for Folgefonna. The subglacial topography was retrieved from a ground penetrating radar survey. The blue line indicates the present-day ice margin, black lines are the cross-sections for F factor calculation red lines represent the flowlines. Red colour points have a predefined shear stress of 100 kPa and yellow colour points have a 50 kPa shear stress value. ground penetrating radar survey. The blue line indicates the present-day ice margin, black lines are
The glacier-surface estimated ELA for the reconstructed icefield, which is arguably the most important use for palaeoglacier reconstructions, is very close to that of the extant glacier (Table 2). In particular, and whichever ELA methods and ratios are adopted (Pellitero et al., 2015), the difference in ELA between the reconstruction and real glacier is on the order of a few metres. The model with no F factor underestimates the ELA by ~15 metres, but the F factor model almost perfectly matches the real glacier ELA, with the error generally near that resulting from the contour belt elevation interval (see Pellitero et al., 2015), which was set to 5 metres for all reconstructions. It must be stressed that the extant glacier limit was used as input, so in this case there was no error in the horizontal geometry of the glacier reconstruction, as would likely be the case for the reconstruction of a fully deglaciated plateau. Vithout use of F factors the reconstruction, as expected, underestimates ice thickness, ar ubsequently ice volumes, by about 25-30%. With the automatically derived F factor ar ubsequently the user-defined F factor the results improve considerably, but overall < 2constructions underestimate the volume and extension of ice to some extent (Table 1). This iainly because there are areas far from the flowlines where the interpolation performs less we his shortcoming could easily be overcome by using a denser flowline pattern, but we wished to te 1e tool using a minimum flow line configuration. Changes in volume and area are in general simil: or the different interpolation methods, and, as expected, IDW and kriging yield the best results olume and Trend the poorest (see section 2.3). volume and Trend the poorest (see section 2.3).
The glacier-surface estimated ELA for the reconstructed icefield, which is arguably the most important use for palaeoglacier reconstructions, is very close to that of the extant glacier (Table 2). In particular, and whichever ELA methods and ratios are adopted (Pellitero et al., 2015), the difference in ELA between the reconstruction and real glacier is on the order of a few metres. The model with no F factor underestimates the ELA by ~15 metres, but the F factor model almost perfectly matches the real glacier ELA, with the error generally near that resulting from the contour belt elevation interval (see Pellitero et al., 2015), which was set to 5 metres for all reconstructions. It must be stressed that the extant glacier limit was used as input, so in this case there was no error in the horizontal geometry of the glacier reconstruction, as would likely be the case for the reconstruction of a fully deglaciated plateau. Vithout use of F factors the reconstruction, as expected, underestimates ice thickness, ar ubsequently ice volumes, by about 25-30%. With the automatically derived F factor ar ubsequently the user-defined F factor the results improve considerably, but overall < 2constructions underestimate the volume and extension of ice to some extent (Table 1). This iainly because there are areas far from the flowlines where the interpolation performs less we his shortcoming could easily be overcome by using a denser flowline pattern, but we wished to te 1e tool using a minimum flow line configuration. Changes in volume and area are in general simil: or the different interpolation methods, and, as expected, IDW and kriging yield the best results olume and Trend the poorest (see section 2.3). volume and Trend the poorest (see section 2.3).
In this case the glacier limits are topographically-constrained so this serves as a good test to verify how accurately the planar extent of the glacier can be reconstructed. The upper portion of the subglacial topography has two tributary basins, so two flowlines which converge near the snout have been created (Figure 5). Three user-defined cross-sections were drawn where the subglacial topography indicates the development of significant troughs. topography indicates the development of significant troughs.
In this case the glacier limits are topographically-constrained so this serves as a good test to verify how accurately the planar extent of the glacier can be reconstructed. The upper portion of the subglacial topography has two tributary basins, so two flowlines which converge near the snout have been created (Figure 5). Three user-defined cross-sections were drawn where the subglacial topography indicates the development of significant troughs. topography indicates the development of significant troughs.
In terms of surface area and volume, the models have variable results (Table 3). All the glacier reconstructions fall within a range of 25% error in volume and 20% in surface area. The highest errors result from running the tool with no F factor, because the drag effect is not accounted for (see section 2.2). Automatic F factor calculation reduces the error in volume to around 15%, whereas user defined cross-sections reduce the error to <10%. Result for the surface show a similar pattern, being 20% with no F factor, and ~6% when the F factor is considered. Again, the tool tends to slightly underestimate both volume and area.
In terms of surface area and volume, the models have variable results (Table 3). All the glacier reconstructions fall within a range of 25% error in volume and 20% in surface area. The highest errors result from running the tool with no F factor, because the drag effect is not accounted for (see section 2.2). Automatic F factor calculation reduces the error in volume to around 15%, whereas user defined cross-sections reduce the error to <10%. Result for the surface show a similar pattern, being 20% with no F factor, and ~6% when the F factor is considered. Again, the tool tends to slightly underestimate both volume and area.
Figure 5. Subglacial topography at Griessglescher, with flowlines (red dots), catchment limit (blue line), cross-sections (black solid lines) and extended point ice surface network (blue dots), which were used in the glacier surface reconstruction.
Figure 5. Subglacial topography at Griessglescher, with flowlines (red dots), catchment limit (blue line), cross-sections (black solid lines) and extended point ice surface network (blue dots), which were used in the glacier surface reconstruction.
Table 4. ELA comparison between the extant glacier and reconstructed Griesseglescher. All ELAs have been calculated using Pellitero et al. (2015)
Table 4. ELA comparison between the extant glacier and reconstructed Griesseglescher. All ELAs have been calculated using Pellitero et al. (2015)
Related topics:
EngineeringEarth SciencesGeologyGeomorphologyComputer ScienceGlacier Reconstruction
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved