Fig 1 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Initial thickness (upper) and speed (lower) of RIS model. Note that thickness is saturated at 800 m. Colored boundaries on the thickness plot indicate influx gates in the model domain. Grey lines indicate no-flow boundaries, black lines indicate shelf front and other colors indicate individually adjustable ice streams and glaciers.
Source publication
Ross Ice Shelf (RIS) is known to experience transient thickness change due to changes in the flow of its tributary ice streams and glaciers and this may complicate identification of external, climate-forced signals in contemporary observations of ice shelf thinning and thickening. Flux changes at the lateral boundaries produce both instantaneous an...
Contexts in source publication
Context 1
... tion accommodates details of coastal geometry and allows the grounded to floating transition to be resolved. A total of 12 740 elements are used in the model ranging in area from 0.14 to 495 km 2 with median and mean values of 18 and 39 km 2 , respectively. The model is initialized by itera- tion using fixed ice stream and glacier influxes ( Fig. 1 and Table ...
Context 2
... use larger flux changes, applied as step functions, to examine characteristic response surfaces and response times. Once initialized ( Fig. 1), the numerical model is used to perform a series of experiments in which ice flux at the Byrd Glacier boundary is varied. We adjusted the magni- tude, timing and shape of the perturbation function, in a series of experiments named 'repeat', 'double', 'quad', 'stop' and 'ramp' (see Table ...
Similar publications
Dynamic ice discharge from outlet glaciers across the Greenland Ice Sheet has increased since the beginning of the 21st century. Calving from floating ice tongues that buttress these outlets can accelerate ice flow and discharge of grounded ice. However, little is known about the dynamic impact of ice tongue loss in Greenland compared to ice shelf...
Dynamic ice discharge from outlet glaciers across the Greenland ice sheet has increased since the beginning of the 21st century. Calving from floating ice tongues that buttress these outlets can accelerate ice flow and discharge of grounded ice. However, little is known about the dynamic impact of ice tongue loss in Greenland compared to ice shelf...
Citations
... Principal component analysis reduces the dimensionality of multivariate datasets by determining linear combinations of variables that provide a number of de-correlated principal components (PCs), each characterising a proportion of the total variance in the underlying data (Dunteman, 1989;Jackson, 1991). This approach has a long history of use in climatological research for characterising spatial patterns in the distribution of key parameters such as temperature, precipitation and atmospheric pressure (Tait et al., 1997); sea ice concentration (Baba and Renwick, 2017); and ice shelf behaviour (Campbell et al., 2017). In these applications, PCA typically reveals the spatial structure of important, and often persistent, features within spatially distributed observations. ...
A 16-year series of daily snow-covered area (SCA) for 2000–2016
is derived from MODIS imagery to produce a regional-scale snow cover
climatology for New Zealand's largest catchment, the Clutha Catchment.
Filling a geographic gap in observations of seasonal snow, this record
provides a basis for understanding spatio-temporal variability in
seasonal snow cover and, combined with climatic data, provides insight
into controls on variability. Seasonal snow cover metrics including daily SCA, mean snow
cover duration (SCD), annual SCD anomaly and
daily snowline elevation (SLE) were derived and assessed for temporal
trends. Modes of spatial variability were characterised, whilst also preserving
temporal signals by applying raster principal component analysis (rPCA) to
maps of annual SCD anomaly. Sensitivity of SCD to temperature and precipitation variability
was assessed in a semi-distributed way for mountain ranges across the catchment.
The influence of anomalous winter air flow, as characterised by HYSPLIT back-trajectories,
on SCD variability was also assessed. On average, SCA peaks in late
June, at around 30 % of the catchment area, with 10 % of the catchment
area sustaining snow cover for > 120 d yr−1. A persistent mid-winter reduction in
SCA, prior to a second peak in August, is attributed to the prevalence of winter
blocking highs in the New Zealand region. In contrast to other regions
globally, no significant decrease in SCD was observed, but substantial spatial
and temporal variability was present. rPCA identified
six distinct modes of spatial variability, characterising 77 % of
the observed variability in SCD. This analysis
of SCD anomalies revealed strong spatio-temporal variability beyond
that associated with topographic controls, which can result in snow
cover conditions being out of phase across the catchment. Furthermore,
it is demonstrated that the sensitivity of SCD to temperature and
precipitation variability varies significantly across the catchment.
While two large-scale climate modes, the SOI and SAM, fail to explain
observed variability, specific spatial modes of SCD are favoured by
anomalous airflow from the NE, E and SE. These findings illustrate
the complexity of atmospheric controls on SCD within the catchment
and support the need to incorporate atmospheric processes that govern
variability of the energy balance, as well as the re-distribution of
snow by wind in order to improve the modelling of future changes in
seasonal snow.
... Principal component analysis (PCA) reduces the dimensionality of multivariate datasets by determining linear combinations of variables that provide a number of de-correlated principal components (PCs), each characterising a proportion of the total 20 variance in the underlying data (Dunteman, 1989;Jackson, 1991). This approach has a long history of use in climatological research for characterising spatial patterns in the distribution of key parameters such as temperature, precipitation and atmospheric pressure (Tait et al., 1997), sea ice concentration (Baba and Renwick, 2017) and ice shelf behaviour (Campbell et al., 2017). In these applications, PCA typically reveals the spatial structure of important, and often persistent, features within spatially distributed observations. ...
A 16-year series of daily snow covered area (SCA) for 2000–2016 is derived from MODIS imagery to produce a regional scale snow cover climatology for New Zealand's largest catchment, the Clutha Catchment. Filling a geographic gap in observations of seasonal snow, this record provides a basis for understanding spatio-temporal variability in seasonal snow cover, and combined with climatic data, provides insight into controls on variability. Metrics including daily SCA, mean snow cover duration (SCD), annual SCD anomaly and daily snowline elevation (SLE) were derived and assessed for temporal trends. Raster principal components analysis (rPCA) was applied to maps of annual SCD anomaly to characterise modes of spatial variability whilst preserving temporal signals. Semi-distributed analysis between SCD and temperature and precipitation anomalies allowed sensitivity of SCD to climatic forcings to be assessed spatially. The influence of anomalous winter air flow, as characterised by HYSPLIT back-trajectories, on SCD variability was also assessed. On average, SCA peaks in late June, at around 30% of the catchment area, with 10% of the catchment area sustaining snow cover for >120 days per year. A reduction in SCA through mid-winter, prior to a second peak in August and persistent throughout the time series is attributed to the prevalence of winter blocking highs in the New Zealand region. In contrast to other regions globally, no significant decrease in SCD was observed. rPCA identified six distinct modes of spatial variability, characterising 77% of the observed variability in SCD. rPCA and semi-distributed analysis of SCD anomalies reveal strong spatio-temporal variability beyond that associated with topographic controls, which can result in snow cover conditions being out of phase across the catchment. Furthermore, it is demonstrated that the sensitivity of SCD to temperature and precipitation variability varies significantly across the catchment. While two large scale climate modes, the SOI and SAM, fail to explain observed variability, specific spatial modes of SCD are favoured by anomalous airflow from the NE, E and SE. These findings illustrate the complexity of atmospheric controls on SCD within the catchment and support the need to incorporate atmospheric processes that govern variability of the energy balance, as well as the re-distribution of snow by wind in order to improve the modelling of future changes in seasonal snow.
Greenland glaciers have three primary seasonal ice flow patterns, or “types”: terminus‐driven, runoff‐driven, and runoff‐adapting. To date, glacier types have been identified by analyzing flow at a single location near the terminus; information at all other locations is discarded. Here, we use principal component (PC)/empirical orthogonal function (EOF) analysis to decompose multi‐year time series of glacier speed, combined from three satellite‐derived products at four glaciers feeding Sermilik Fjord, Greenland. This improves on single‐point methods by yielding spatial patterns (EOFs), which ensure the result reflects data at all locations on the glacier, and associated temporal patterns (PCs), which allow identification of glacier type. We find that the leading EOF is uniformly signed over the entire glacier domain, that this mode explains the majority of the variance in speed, and therefore that glacier type can be inferred from the leading PC. We find that Helheim Glacier was terminus‐driven, Fenris Glacier and Midgard Glacier were runoff‐adapting, and Pourquoi Pas Glacier was runoff‐driven over 2016–2021. Our classification agrees with previous work for Helheim and Midgard Glaciers, but differs at the other two. At all but Fenris Glacier, the leading PC correlates significantly with the speed pattern observed at the single point used in previous analyses. Thus, Fenris Glacier has more complex flow patterns than single‐point analysis can capture, and wider spatial analysis techniques such as EOF/PC are required. We suggest that, due to its low computational cost and inclusion in standard analysis packages, EOF/PC analysis should be used for assessing glacier type.
Greenland glaciers have three primary seasonal ice flow patterns, or “types”: terminus driven, runoff driven, and runoff adapting. To date, glacier types have been identified by analyzing flow at a single location near the terminus; information at all other locations is discarded. Here, we use principal component (PC) / empirical orthogonal function (EOF) analysis to decompose multi-year time series of glacier speed, combined from three satellite-derived products at four glaciers feeding Sermilik Fjord, Greenland. This improves on single-point methods by yielding temporal patterns (PCs), which allow identification of glacier type, and associated spatial patterns (EOFs), which ensure the result reflects data at all locations on the glacier. We find that the leading mode is uniformly signed over the entire glacier domain, that this mode explains the majority of the variance in speed, and therefore that glacier type can be inferred from the leading PC. We find that Helheim Glacier was terminus-driven, Fenris Glacier and Midgard Glacier were runoff-adapting, and Pourquoi-Pas Glacier was runoff-driven over 2016-2021. Our classification agrees with previous work for Helheim and Midgard Glaciers, but differs at the other two. At all but Fenris Glacier, the leading PC correlates significantly with the speed pattern observed at the single point used in previous analyses. Thus, Fenris Glacier has more complex flow patterns than single-point analysis can capture, and wider spatial analysis techniques such as EOF/PC are required. We suggest that, due to its low computational cost and inclusion in standard analysis packages, EOF/PC analysis should be used for assessing glacier type.
As time series observations of Antarctic change proliferate, it is imperative that mathematical frameworks through which they are understood keep pace. Here we present a new method of interpreting remotely-sensed change using spatial statistics and apply it to the specific case of thickness-change on the Ross Ice Shelf. First, a numerical model of ice shelf flow is used together with EOF analysis to generate characteristic patterns of response to specific forcings. Because they are continuous and scalable in space and time, the patterns allow short duration observations to be placed in a longer time series context. Second, focusing only on changes that are statistically significant, the synthetic response surfaces are used to extract magnitude and timing of past events from the observational data. Slowdown of Kamb and Whillans Ice Streams are clearly detectable in remotely-sensed thickness change. Moreover, those past events will continue to drive thinning into the future.
