Greenland

New View on Geothermal Heat Flow in Greenland and Antarctica

Posted by William Colgan on January 15, 2021
New Research / No Comments

We have a new open-access study about geothermal heat flow beneath the Greenland and Antarctic ice sheets in the Journal of Geophysical Research: Earth Surface. 

Presently, there’s a lot of uncertainty about the magnitude and pattern of geothermal heat flow beneath both ice sheets. That’s because it has only been sampled at a handful of widely spaced deep ice cores (Figure 1). While the average value of geothermal heat flow is relatively small, getting it right is essential for ice-flow models. If you run a computer simulation of an ice sheet with a severe over- or under-estimation of the geothermal heat flow, you can easily end up generating an ice sheet that is either too warm or too cold. Ice flow is very sensitive to temperature – particularly near the bed – so geothermal heat flow is a critical variable for simulating the form and flow of Earth’s ice sheets.

Figure 1 – Peering into the drill trench at the NEEM deep ice core site. NEEM is one of only six deep ice core sites in the Greenland Ice Sheet interior where geothermal heat flow has been measured to date.

Our study took a fresh look at changes in geothermal heat flow across space, but not those due to the subtle variations in Earth’s crust and mantle properties over tens of km. Instead, we examined the effect of the ice-sheet bed’s topographic relief on geothermal heat flow to generate the first comprehensive snapshot of changes in geothermal heat flow at scales of hundreds of meters due to that relief. It’s been known for over a century that geothermal heat flow is greater in valleys and smaller on ridges. Basically, if the heat escaping Earth’s interior is looking for the quickest way to radiate into the atmosphere, a deeply incised valley provides the fastest exit. This effect is readily observable from the fact that geotherms – surfaces of constant temperature – are packed more closely together beneath valleys, indicating a stronger temperature gradient there (and hence heat flow) in comparison to ridges.

We created a simple statistical model to estimate this topographic influence on geothermal heat flow. This model essentially uses a digital elevation model of the bedrock topography to assess local topographic relief and then converts this local relief into a fractional correction for geothermal heat flow. It produces a positive correction – an increase – for valleys, and vice versa for ridges. Our approach is admittedly simple and empirical – a literal “first-order” approximation – but it seems to reliably reproduce the topographic variability in geothermal heat flow in all the settings for which we could find previous studies. So, we applied this statistical model to digital elevation models for Greenland and Antarctica. This revealed much more detail in a geothermal heat flow map than we are used to seeing.

Figure 2 – Left: An existing regional geothermal heat flow model (Martos2017). Right: Regional geothermal heat flow corrected for local topographic relief.

Across both Greenland and Antarctica, we see patterns of increased geothermal heat flow within deeply incised glacier valleys and decreased geothermal heat flow along ridges and mountains. In many regions, most notably the Antarctic Peninsula (Figure 2) and Central East Greenland (Figure 3), we find that local topography routinely modifies regional geothermal heat flow by more than ~50%.

Figure 3 – Left: An existing regional geothermal heat flow model (Martos2018). Right: Regional geothermal heat flow corrected for local topographic relief.

In Greenland, we estimate that there are ~100 outlet glaciers that are both sufficiently narrow and deeply incised to more than double local geothermal heat flow relative to that of the regional average value. The model also suggests that – especially deep within the interior of the Greenland Ice Sheet – local geothermal heat flow may be sufficiently suppressed along prominent subglacial ridges to cause subglacial water to refreeze. (At least, in ice-sheet areas where the ice-bed interface is near the freezing point.) 

The topographic correction for geothermal heat flow that we model is only as a good as the topographic relief that we derive from subglacial digital elevation models. Generally, in areas where subglacial topography is best resolved, the effect of the topography on the local geothermal heat flow is greater than uncertainty in the underlying regional geothermal heat flow model (Figure 4). There are still large swaths of the ice sheets where subglacial topography remains poorly resolved. In these areas, our model is not tremendously useful; the existing uncertainties between regional geothermal heat flow models are still larger than any local topographic correction that we can estimate.

Figure 4 – Maps of Antarctica and Greenland identifying areas, illustrated in red, where the influence of local topographic relief on geothermal heat flow is at least as important as the choice of geothermal heat flow model. As subglacial topography is still poorly resolved in both ice sheet interiors. These red areas may be expected to expand with improved mapping of subglacial topography.

The topographic corrections for geothermal heat flow in Greenland and Antarctica that we have calculated are now available as dimensionless fields in NetCDF format, with grids that are the same resolution as BedMachine, for each region via the PROMICE data portal (www.promice.dk). This means that they can be anonymously downloaded and applied to any regional geothermal heat flow model of the user’s choice. We hope that these topographic corrections for geothermal heat flow will be adopted into ice-flow models to improve both present-day ice-sheet simulations, as well as our understanding of the role of geothermal heat flow in the feedback between ice flow and topography on geologic timescales.

It has certainly been a long and winding road to this publication – the original draft of this article was first submitted in June 2019 – and we are grateful for Noah Finnegan (University of California Santa Cruz) and Olga Sergienko (Princeton University) for serving as editors to four very helpful peer-reviewers. Interdisciplinary projects can clearly provide a bumpier ride than staying in your own lane, but – in this instance – the journey seems to have taken us to a very different view of geothermal heat flow in Greenland and Antarctica.

Development of this data product was funded by the award “HOTROD: Prototype for Rapid Sampling of Ice-Sheet Basal Temperatures” provided by the Experiment Programme of the Villum Foundation. Improved understanding of the spatial variability in subglacial geothermal heat flow helps us optimize drill site selection and analyze ice temperature measurements. This data product was also supported by the Danish Ministry for Climate, Energy and Supplies through the Programme for Monitoring of the Greenland Ice Sheet (PROMICE).

Colgan, W., J. MacGregor, K. Mankoff, R. Haagenson, H. Rajaram, Y. Martos, M. Morlighem, M. Fahnestock and K. Kjeldsen. 2021. Topographic Correction of Geothermal Heat Flux in Greenland and Antarctica. Journal of Geophysical Research. 125: e2020JF005598. doi:10.1029/2020JF005598.

Tags: , , , , , , , , ,

28-Year Record of Greenland Ice Sheet Health

Posted by William Colgan on January 14, 2021
Climate Change, New Research, Sea Level Rise / No Comments

We have a new open-access study about Greenland Ice Sheet mass balance – or health – in the current issue of Geophysical Research Letters. In this study, we present a new 28-year record of ice-sheet mass balance. This record is relatively unique for two reasons.

Firstly, because of its length. The most recent ice-sheet mass balance inter-comparison exercise (IMBIE2) clearly highlighted how the availability of ice-sheet mass balance estimates has changed through time. During the GRACE satellite gravimetry era (2003-2017), there are usually more than twenty independent estimates of annual Greenland Ice Sheet mass balance. Prior to 2003, however, there are just two independent estimates. Our new 1992-2020 mass balance record will therefore provide especially welcomed additional insight on ice-sheet mass balance during the 1990s.

Figure 1 – Greenland Ice Sheet mass balance estimated by IMBIE2 between 1992 and 2018. The number of independent estimates comprising each annual estimate is shown. Prior to 2003, there are only 1 or 2 independent estimates of ice-sheet health each year.

Secondly, because of its consistency. This new mass balance record has been constructed by merging radar altimetry measurements from four ESA satellites (ERS-1/2, ENVISAT, CryoSat-2 and Sentinel-3A/B) over nearly three decades into one consistent framework. While all four of these satellites use the same type of Ku-band radar altimeter, to date, their measurements have usually been analyzed independently of each other. This time, however, we use machine learning to merge the elevation changes measured by these similar-but-different satellites into a common mass balance signal through space and time. This makes our new record the only satellite altimetry record that spans the entire IMBIE period.

Figure 2 – Comparison of our new multi-satellite radar-altimetry derived record of ice-sheet health (“Radar-VMB”) with two records estimated by the input-output method (“Colgan-IOMB” and “Mouginot-IOMB”), as well as one record estimated by satellite gravimetry (“GRACE-GMB”).

When we compare our new radar altimetry record of mass balance to two existing input-output records of mass balance, we find good agreement in the capture of Greenland’s high and low mass balance years. These other two multi-decade records are derived from the input-output method, in which estimated iceberg calving into the oceans is differenced from estimated surface mass balance (or net snow accumulation) over the ice sheet. While the input-output method often has limited spatial (and temporal) resolution, our radar altimetry derived record can resolve spatial variability in mass balance across the ice sheet every month since 1992.

Figure 3 – Our multi-satellite radar-altimetry derived map of declining ice-sheet health over the (a) the 1992-1999, (b) the 2000-2009, and (c) the 2010-2020 periods.

While our new long-term record provides a new overview of the health of the Greenland ice sheet, it can also be helpful to understand the processes that influence ice-sheet health. For example, we see a sharp increase in mass balance between 2016 and 2017. When we look at this event in detail, we can attribute it to unusually high snowfall in fall 2016, especially in East Greenland, and unusually little surface melting in summer 2017, throughout the ice-sheet ablation area. We estimate that the 2017 hydrological year was likely the first year during the 21st Century during which the ice sheet was actually in a state of true “mass balance” – or equilibrium – as opposed to mass loss.

The development of this new dataset was primarily funded by the European Space Agency (ESA), with a little help from the Programme for Monitoring of the Greenland Ice Sheet (www.promice.dk). Our multi-satellite Ku-band altimetry mass balance record is now available as tabulated data – both for the ice sheet, as well as the eight major ice-sheet drainage sectors – at https://doi.org/10.11583/DTU.13353062. Within the next two years, the ongoing Sentinel-3A/B satellite missions are clearly poised to extend Greenland’s radar altimetry record to three decades. This will allow us to start assessing ice-sheet health using the statistics of a 30-year climatology record. This keeps us excited at the prospect of updating this record in the near future. Stay tuned!

Simonsen, S., V. Barletta, W. Colgan and L. Sørensen. 2021. Greenland Ice Sheet mass balance (1992-2020) from calibrated radar altimetry. Geophysical Research Letters. L61865. doi:10.1029/2020GL091216.

Tags: , , , , , , , , , , , ,

Freshwater Runoff from Greenland’s 54K Basins

Posted by William Colgan on November 12, 2020
New Research / No Comments

We have a new open-access study out in the current issue of Earth Systems Science Data. In this study, we estimate the liquid water discharge – meaning meltwater and rainfall flowing into the ocean – every day since 1958 from 54,142 hydrologic basins across Greenland. About 40% of these basins are associated with glaciers or the ice sheet, and these “ice” basins accounted for ~65% of Greenland’s total liquid water discharge. On an annual basis, we estimate that Greenland’s liquid discharge varied from between ~136 km3 in 1992 and ~785 km3 in 2012. The daily discharge records and these individual basins are now available online. This dataset provides a great improvement in our understanding of when and where freshwater is entering Greenlandic fjords.

Where possible, we compared the daily discharge records of individual basins that we downscale from climate models to actual observed river discharge measurements. There are only a few continuous river gauging stations in Greenland operated by different monitoring programs and research groups. Thankfully, we could use publicly accessible observational records from nine basins (Kingigtorssuaq, Kobbefjord, Leverett, Oriartorfik, Qaanaaq, Røde Elv, Teqinngalip, Watson and Zackenberg) to assess performance of our data product. These comparisons show that the accuracy of data product varies with both basin size – or discharge volume – as well as climate model. Generally, however, the data product reproduces the magnitude and variability of observed basin discharge within a reasonable uncertainty.

Downscaling runoff from regional climate models to individual basins is clearly sensitive to errors or uncertainties in the elevation model guiding the hydrological routing. This is especially true for glacier or ice-sheet basins, which require additional assumptions about the effective water pressure within the ice. Hydrologic boundaries can shift due to slight changes in elevation or effective water pressure. We therefore ran our hydrological routing code many times to see how sensitive the location of basin outlets – meaning where water drains from ice-to-tundra or tundra-to-ocean – where to common assumptions. We found many basin outlets around the low-elevation ice-sheet ablation area can shift by more than 30 km under a range of common assumptions. This highlights the challenge of trying to balance a water budget within a given fjord. It also points to where improved knowledge of subglacial topography is most needed.

Sensitivity in assessed basin outlet location — land outlets (Left) and ice outlets (Right) — to common hydrologic routing assumptions. Ice-sheet basins likely vary with effective water pressure on both inter- and intra-annual time-scales.

A neat aspect of this study is that the source code is also made available open access. This code-sharing approach is part of the growing “open science” movement. Sharing code not only makes complex results reproducible, but also helps different research teams move forward. In this case, basin-scale runoff estimates are sensitive to the choices of both climate model and downscaling method. By making the source code available, subsequent research teams can implement precisely the same climate model and/or downscaling methods. The development of this data and code product was funded by the Danish Ministry for Climate, Energy and Supply to the Programme for Monitoring of the Greenland Ice Sheet (www.PROMICE.dk), as well as European Union’s Horizon 2020 to the INTAROS project (www.INTAROS.eu).

Mankoff, K., B. Noël, X. Fettweis, A. Ahlstrøm, W. Colgan, K. Kondo, K. Langley, S. Sugiyama, D. van As, and R. Fausto,  2020. Greenland liquid water runoff from 1958 through 2019, Earth System Science Data. 12: 2811–2841. doi:10.5194/essd-12-2811-2020.

Tags: , , , , , , , , , , ,

‘Cold Content’ of Greenland’s Firn Plateau

Posted by William Colgan on April 29, 2020
Climate Change, Communicating Science, New Research / No Comments

We have a new open-access study in the current issue of Journal of Glaciology that investigates the “cold content” of Greenland’s high-elevation firn plateau1. Firn is the relatively low density near-surface ice-sheet layer comprised of snow being compressed into ice. Cold content is one of its quirkier properties. Of course, all firn is literally freezing – meaning below 0°C – but some firn is colder than other firn. Clearly, it takes a lot more energy to warm -30°C firn to 0°C, than it does for -1°C firn. Our study highlights at least one discernible shift in cold content – how much sensible heat energy is required to warm firn to the 0°C melting point – in response to climate change.

Figure 1 – The nine high-elevation ice-sheet sites where we assessed firn cold content in the top 20 m.

There is a strong annual cycle in firn cold content. Generally, cold content is at its maximum each April, after the firn has been cooled by winter air temperatures. Cold content then decreases through summer, as warming air temperatures and meltwater percolation pump energy into the firn, to reach a minimum each September. The magnitude of this annual cycle varies across the ice sheet, primarily as a function of the meltwater production, but also as a function of snowfall-dependent firn density. Firn density is highly sensitive to snowfall rate, and firn cold content is a function of firn density.

Figure 2 – The mean annual cycle in four-component firn cold content assessed at the nine ice-sheet sites over the 1988-2017 period. Note the relatively large latent heat release associated with meltwater at Dye-2, in comparison to other sites.

We find few discernible year-on-year trends in cold content across the highest elevation areas of the firn plateau. For example, there is perhaps a slight decrease at Summit – where we find snowfall is increasing at 24 mm/decade and air temperatures are warming at 0.29°C/decade – but statistically-significant multi-annual trends in cold content are difficult to separate from year-to-year variability. At Dye-2, however, which has the greatest melt rate of the sites that we examine, there is clear evidence of the impact of changing climate. At Dye-2, an exceptional 1-month melt event in 2012 removed ~24% of the cold content in the top 20 m of firn. It took five years for cold content to recover to the pre-2012 level.

Figure 3 – The cumulative four-component firn cold content at the nine ice-sheet sites over the 1998-2017 period. Note the sharp loss of Dye-2 cold content in 2012, and the subsequent multi-year recovery of this cold content.

The refreezing of meltwater within firn is a potential buffer against the contribution of ice-sheet melt to sea-level rise; surface melt can refreeze within porous firn instead of running off into the ocean. But refreezing meltwater requires available firn cold content. The multi-annual reset of cold content that we document at Dye-2 suggests that a single melt event can reduce firn cold content – and thus precondition firn for potentially less meltwater refreezing – for years to follow. This highlights the potential for the cold content of Greenland’s firn plateau to decrease in a non-linear fashion, as climate change pushes melt events to progressively higher elevations of the firn plateau.

1Vandecrux, B., R. Fausto, D. van As, W. Colgan, P. Langen, K. Haubner, T. Ingeman-Nielsen, A. Heilig, C. Stevens, M. MacFerrin, M. Niwano, K. Steffen and J. Box. 2020. Firn cold content evolution at nine sites on the Greenland ice sheet between 1998 and 2017. Journal of Glaciology..

Tags: , , , , , , , , , ,

New Greenland iceberg calving estimate

Posted by William Colgan on June 06, 2019
New Research / No Comments

We have a new – and long awaited! – open-access study out in the current issue of Earth Systems Science Data. In this study, we estimate the ice discharge – meaning transfer of land-ice into the ocean – at 276 tide-water glaciers around the Greenland Ice Sheet between 1986 and 2017. These individual glacier discharge records are now available online. We estimate that ice-sheet-wide discharge – or iceberg calving – increased from less than 450 Gt/yr in the 1980s and 1990s to closer to 500 Gt/yr at present. That increase of 50 Gt/yr is equivalent to an extra 1600 tonnes per second of icebergs – year-round – relative to the 1980s and 1990s.

Figure 1 – Time series of iceberg discharge from the Greenland Ice Sheet. Dots represent when observations occurred. The orange line is the annual average. Coverage denotes the percentage of glaciers from which total discharge is observed at any given time. Total discharge is “estimated”, rather than “observed”, when coverage is <100 %.

Dealing with unknown ice thickness or missing ice velocity data – in a transparent and reproducible fashion – was a huge aspect of making such a dense glacier discharge dataset. Perhaps the most novel aspect of this study is a sensitivity test to quantify just how precisely ice discharge from the entire ice sheet can be estimated at a single point in time. The result of this sensitivity test was a little surprising. We found that – using the same ice thickness and ice velocity information – assessed ice discharge can change tremendously just based on where we placed our “flux gates”.

We examined placing flux gates – meaning the virtual lines across every glacier through which we estimate ice discharge – between 1 and 9 kilometers up-glacier from the glacier tongue, and extending them laterally into minimum ice velocities of between 10 and 150 m/yr. These generally reasonable ranges can influence the apparent ice-sheet-wide discharge we estimate by around 50 Gt/yr. To place this flux gate uncertainty in perspective, we can say it is roughly equivalent to the total uncertainty in ice-sheet-wide discharge – from all sources of uncertainty – assigned in most previous studies. This flux gate uncertainty is also roughly equivalent to the change in ice-sheet-wide discharge since the 1980s.

Figure 2 – Sensitivity test of ice-sheet-wide discharge as a function of flux gate location. The vertical axis denotes the up-glacier distance of flux gates from the glacier tongue. The horizontal axis denotes the minimum ice velocity into which flux gates laterally extend.

A very cool thing about this study is that not only the data, but also the code, is open access. This code-sharing approach is part of the growing “open science” movement. The US National Academies – meaning Science, Engineering and Medicine – recently joined together to publish an open science mandate. Sharing code not only makes complex results reproducible, but also helps different teams move forward. For example, our ice-sheet-wide discharge is slightly different from previous studies. We are not entirely sure how much of this difference in ice discharge is due to differences in flux gate locations. But now – at least moving forward – future teams will be able to use precisely the same flux gates that we used.

Mankoff, K., W. Colgan, A. Solgaard, N. Karlsson, A. Ahlstrøm, D. van As, J. Box, S. Khan, K. Kjeldsen, J. Mouginot and R. Fausto. 2019. Greenland Ice Sheet solid ice discharge from 1986 through 2017. Earth System Science Data. 11: 769-786. https://doi.org/10.5194/essd-11-769-2019.

Tags: , , , , , , ,

Lost Ice-Sheet Porosity and Sea-Level Buffering

Posted by William Colgan on March 12, 2019
New Research, Sea Level Rise / No Comments

We have a new open-access study that investigates the high-elevation firn plateau of the Greenland Ice Sheet in the current issue of The Cryosphere1. Firn is the relatively low density near-surface ice-sheet layer comprised of snow being compressed into ice. Firn is relatively porous, meaning that meltwater can percolate through it. The refreezing of meltwater within firn is a potential buffer against the ice-sheet sea-level contribution from surface melt; surface melt can refreeze within porous firn instead of running off into the ocean. Our study aims to assess how big this sea-level buffer might be, and how much sea-level buffer may have already been used.

We pull together a singularly unique dataset – 340 ice-core measurements of firn density collected over 65 years – to assess the near-surface density across the entire high-elevation firn plateau of the Greenland Ice Sheet. Many of these vertical firn density profiles were digitized and brought together for the first time from historical studies, but twenty are collected by our team and new to science. We analyze this ice-core dataset for empirical relations between firn density and accumulation or air temperature. This allows us to divide the ice sheet into three distinct firn areas, within each of which we can confidently predict the vertical profile of near-surface firn density.

Figure 1 – Left: Firn air content within the top 10 m (FAC10) estimated from ice-core measurements (denoted with ‘x’). The ice sheet is divided into three areas: the Dry Snow Area (DSA), the Low Accumulation Percolation Area (LAPA), and the High Accumulation Percolation Area (HAPA). Right: Change in top 10 m firn air content between 1998–2008 and 2010–2017 within Low Accumulation Percolation Area along the ice sheet’s western flank.

We find that the firn structure at the heart of the ice sheet – the highest, coldest and driest firn known as the Dry Snow Area – appears to have been stable since 1953. There is no trend in firn density within the Dry Snow Area. At lower elevations, however, we find significant changes in response to recent increases in surface melt due to climate change. The area we call the Low Accumulation Percolation Area – an elevation band of relatively low snowfall and high melt along the ice sheet’s west flank – has a marked increase in the firn densities measured pre- and post-2009. This firn density change is equivalent to a sea-level buffer loss of 1.5±1.2 mm sea-level equivalent (540±440 gigatonnes).

We compare the ice-sheet-wide firn density structure that we estimate from ice-core measurements with the firn density structure estimated from three regional climate models. The regional climate models suggest that the decrease in firn porosity initiated in the early 2000s and accelerated with post-2010 climate change. But we also find non-trivial differences between the firn porosities simulated by regional climate models, and that inferred from ice-core measurements, especially in what we call the High Accumulation Percolation Area. Here – the ice sheet’s low elevation southeast flank – modeled firn porosity can be biased the equivalent of between 3 and 7 meters of air distributed over the entire firn column depth.

Figure 2 – Left: Ice-sheet-wide firn air content within the top 10 m of firn (FAC10) simulated by three regional climate models (MAR, HIRHAM and RACMO) and derived from ice-core observations (this study) in different ice-sheet areas. Right: Same for firn air content over the entire depth of the firn column (FACtot).

This study highlights the importance of bringing together firn density measurements to document the response of ice-sheet firn – a non-trivial component of the sea-level budget – to recent climate change. The ice-sheet-wide firn porosity structure we infer from ice-core measurements can also serve as an independent evaluation target for the firn porosity structures simulated by regional climate models. This study also illustrates how new insight can be obtained from the synthesis and re-analysis of historical datasets. This emphasizes the tremendous value of open-access data within the scientific community. This work is part of the Retain project funded by the Danmarks Frie Forskningsfond (grant 4002-00234). The open-access publication is available via the hyperlink below.

1Vandecrux, B., MacFerrin, M., Machguth, H., Colgan, W., van As, D., Heilig, A., Stevens, C., Charalampidis, C., Fausto, R., Morris, E., Mosley-Thompson, E., Koenig, L., Montgomery, L., Miège, C., Simonsen, S., Ingeman-Nielsen, T., and Box, J. 2019. Firn data compilation reveals widespread decrease of firn air content in western Greenland. The Cryosphere. 13: 845-859. https://doi.org/10.5194/tc-13-845-2019.

Tags: , , , , , , ,

Eight trillion tonnes of Arctic ice lost since 1971

Posted by William Colgan on December 20, 2018
Climate Change, New Research, Sea Level Rise / 2 Comments

We have just completed a study that inventories Arctic land ice loss since 1971. It is available open-access in the current issue of Environmental Research Letters1. While we scientists have a pretty good idea of the health — or mass balance — of glaciers and ice sheets — or land ice — since the advent of satellite altimetry in the early 1990s, there is a need for better understanding of land ice health during the pre-satellite era. Our new study estimates the annual ice loss from all glacierized regions north of 55°N between 1971 and 2017.

We use in situ data – mass balance measurements from a handful of continuously monitored glaciers – as indicators for the health of land ice in seven Arctic regions. These hard-fought in situ data are scarce, they are only measured at between 20 and 44 Arctic glaciers every year. Extrapolating these data to entire regions is statistically challenging without additional information. Fortunately, independent estimates of regional mass balance are available from satellite gravimetry during the 2003 to 2015 period. This permits calibrating in situ and satellite-derived mass balance estimates during the satellite era. This makes our pre-satellite era estimates fairly robust.

During the 41 years assessed, we estimate that approximately 8,300 Gt of Arctic land ice was lost. It is difficult to contextualize this magnitude of ice loss. The flow of Niagara Falls – which is approximately 2400 m3 per second or about 75 km3 per year – is only equivalent to about half this volume (3500 km3) over the 1971-2017 period. The total Arctic land ice loss that we document represents 23 mm of sea-level rise since 1971. Greenland is by far the largest contributor (10.6 mm sea-level equivalent), followed by Alaska (5.7 mm sea-level equivalent) and then Arctic Canada (3.2 mm sea-level equivalent).

The UN Intergovernmental Panel on Climate Change (IPCC) now highlights two periods – the “recent past” (1986-2005) and “present day” (2005-2015) – as being of special interest in climate change studies. The Arctic land ice contribution to sea-level rise that we inventory increased from 0.4 to 1.1 mm sea-level equivalent between these periods. In terms of tonnes per second (5,000 to 14,000 t/s), both the magnitude – and the increase – are staggering.

Figure 1 – The cumulative sea-level rise contribution (in mm) from land ice in seven regions of the Arctic between 1971 and 2017. Analogous estimates from satellite gravimetry (GRACE) between 2003 and 2015 shown with open symbols.

The uncertainties associated with extrapolating sparse in situ data over large areas are undeniably large. But, the reality is that climate change was already gearing up as the global satellite observation network came online. So, in the absence of satellite data that can characterize the “pre-climate change” health of Arctic land ice, we need to leverage the extremely precious pre-satellite era observations that are available in creative ways. We hope that the ice loss estimates we present will be useful comparison targets for studies that estimate pre-satellite era mass balance in other ways.

The estimates of annual land ice mass balance — or health — in seven Arctic regions produced by this study are freely available for download here. This study was developed within the Arctic Monitoring and Assessment Program (AMAP) and International Arctic Science Committee (IASC) frameworks, as a direct contribution to the IPCC Special Report on the Ocean and Cryosphere in a Changing Climate (SROCC).

Figure 2 – Annual land ice mass balance — or health — in six Arctic regions between 1971 and 2017. Individual glacier mass balance records (blue lines) are combined into a regional composite (black line). Health is expressed both as a normalized score (left axis) and in gigatonnes per year (right axis). The numbers of glaciers comprising each composite is indicated in red text.

1Box, J., W. Colgan, B. Wouters, D. Burgess, S. O’Neel, L. Thomson and S. Mernild. 2018. Global sea-level contribution from Arctic land ice: 1971 to 2017. Environmental Research Letters.

Tags: , , , , , , ,

Changes in Ice-Sheet Density: How and Why?

Posted by William Colgan on October 25, 2018
Climate Change, Communicating Science, New Research, Sea Level Rise / No Comments

We investigate the high-elevation firn plateau of the Greenland Ice Sheet in a new open-access study in the current issue of Journal of Geophysical Research1. This study pulls together singularly unique – and hard fought – ice core observations and weather station data into a super-neat firn model. This relatively porous near-surface ice-sheet layer known as firn is being increasingly scrutinized for two main reasons.

The first reason is sea-level rise. These high regions of the Greenland ice sheet are normally preserved form intense melting, but this is changing, with more melt seen in recent years. Nevertheless, the porosity of the firn can provide a buffer against sea-level rise when meltwater refreezes within the firn instead of running off into the ocean. But exactly how much of this buffering capacity is available – and for how long – is not really understood.

The second reason is satellite altimetry. Repeat observation of ice thickness by satellite altimeter is a primary method by which ice-sheet mass balance – or overall health – is assessed. But since firn is porous, changes in elevation don’t always translate into changes in mass. For example, the firn layer can become thinner – making the ice-sheet appear thinner – when there’s actually just an increase in firn density rather than a change in mass.

Figure 1 – Locations of the four study sites on the Greenland Ice Sheet’s high-elevation firn plateau.

In this study, we were interested in teasing out the climatic controls of firn density: What makes firn porosity grow and shrink over time? So, we simulated the evolution of firn density – and therefore porosity – over time at four ice-sheet sites. These sites were carefully chosen as sites where both in-situ climate and firn measurements were available (Crawford Point, Dye-2, NASA-SE and Summit). The firn simulations used an updated version of the HIRHAM regional climate model’s firn model. At each site, we initiated simulations using firn density profiles observed from ice cores, and then ran the simulations forward in time using in-situ weather station records. We then ensured that simulated firn density also compared well with repeat firn density profiles observed again many years later. The simulations were between 11 and 15 years, depending on the data available at each site.

Figure 2 – Simulated firn density through time at the four study sites. At all sites, the relative depth of a given layer increases over time, as snowfall exceeds meltwater runoff.

A lot of recent ice-sheet research has focused on how increasing air temperatures and meltwater production are increasing firn density. And our simulations definitely confirmed that! But perhaps counterintuitively, we found that the leading driver of changes in firn density was actually year-to-year changes in amount of snowfall. Firn density decreases as snowfall increases, and vice versa. This study therefore highlights that if we want to project time-and-space variability in firn density we really need to project time-and-space variability in snowfall rates.

Figure 3 – Assessing the relative strength of four drivers of firn density change at the four study sites.

It was also satisfying to see that – given observed climate data – our simulations could reproduce the firn conditions as observed in the field. This gives confidence including this firn model in regional climate models. This finding is of course limited to the high-elevation firn plateau of the Greenland Ice Sheet, which admittedly does not experience tremendous melt. But, as the firn plateau covers over 80% of the ice-sheet area, understanding it plays a key role in tackling pressing satellite altimetry and sea-level buffering questions.

This work is part of the Retain project funded by the Danmarks Frie Forskningsfond (grant 4002-00234). The open-access publication is available via the hyperlink below.

1Vandecrux, B., R. Fausto, P. Langen, D. van As, M. MacFerrin, W. Colgan, T. Ingeman‐Nielsen, K. Steffen, N. Jensen, M. Møller and J. Box. 2018. Drivers of firn density on the Greenland ice sheet revealed by weather station observations and modeling. Journal of Geophysical Research: Earth Surface. 123: 10.1029/2017JF004597.

Tags: , , , , , , , ,

Q-Transect: A Hotspot of Greenland ice loss

Posted by William Colgan on June 19, 2018
Climate Change, New Research / No Comments

We are introducing a rich trove of ice-sheet surface mass balance measurements in an open-access study in the current issue of Journal of Geophysical Research1. The Qagssimiut Lobe is among the most southern ice lobes of the Greenland Ice Sheet. The Q-transect – which runs up the heart of the Qagssimiut Lobe – has been home to automatic weather stations recording ice and climate measurements since 2000. In this study, we have compiled sixteen years of annual surface mass balance measurements and also added three hard-fought years of winter snow accumulation measurements. These data – spanning 300 to 1150 m elevation – now form an exceedingly unique record of ice-sheet health.

Herm_1

Figure 1. The Qagssimiut Lobe in South Greenland. Measurement locations are denoted with white dots. The Sermilik Glacier catchment is delineated with a black line. The ice-sheet margin is delineated with a white line. The background image was acquired by the ESA Sentinel-2 satellite on 28 August 2016 and clearly illustrates the bare ice area below equilibrium line altitude.

These comprehensive in situ measurements allowed us to evaluate the accuracy of the surface mass balance simulated by climate models. TO do this, we stacked our measurements against comparable simulations from three leading regional climate models (HIRHAM5, MAR and RACMO2). The climate models generally did well, but were never bang-on the measurements. One climate model consistently simulated more negative surface mass balances and lower equilibrium line altitudes than we measured. The other two model usually did the opposite, implying the ice sheet was healthier than in reality. These biases appear to stem from differences in simulated winter snow accumulation – which can vary by 200 % at low elevations – between models.

Herm_2

Figure 2. Elevation profiles of measured and simulated winter snow accumulation in (a) 2013/2014, (b) 2014/2015, and (c) 2016/2017. Shaded areas indicate uncertainty ranges. In (c), black lines illustrate the comparison of the model mean for 2000/2001 to 2015/2016 with the 2016/2017 observations.

Combining our knowledge of surface mass balance over the Qagssimiut Lobe with independent observations of iceberg calving rate at Sermilik Glacier – the main tidewater draining Qagssimiut Lobe – allowed us to calculate a total mass balance. We found that the relatively small Sermilik Glacier catchment is now losing up to 2.7 Gt of ice per year. That is a rather astounding – 20 times greater than the ice sheet average – the Sermilik Glacier catchment represents only about 0.03 % of ice-sheet area but is contributing about 0.61 % of ice-sheet mass loss. Its extreme southern location clearly makes Sermilik Glacier a hotspot of ice-sheet mass loss. Its rate of ice loss is more characteristic of lower latitude Andean glaciers than the vast majority of Greenland.

HERM_3

Figure 3. Left: Estimated total mass balance of Sermilik Glacier catchment between 2001 and 2012 in Gt/yr (uncertainty denoted by spread). Right: The Sermilik Glacier catchment overlaid on an ice velocity map derived from the ESA Sentinel-1 satellite. Thin lines indicate adjacent ice flow lines.

We hope that this study will be useful to climate modelers, as they further improve the accuracy with which their models simulate ice-sheet surface mass balance. We also hope that highlighting the Q-transect as a hotspot for both ice loss and in situ data availability will help inform future measurement campaigns seeking to improve our understanding of the physical processes influencing surface mass balance. All measurements of surface mass balance and winter snow accumulation are freely available in the study’s online material.

1Hermann, M., J. Box, R. Fausto, W. Colgan, P. Langen, R. Mottram, J. Wuite, B. Noel, M. van den Broeke and D. van As. 2018. Application of PROMICE Q-transect in situ accumulation and ablation measurements (2000-2017) to constrain mass balance at the southern tip of the Greenland ice sheet. Journal of Geophysical Research. 123: 10.1029/2017JF004408.

Tags: , , , , , , ,

What’s the density of snow on the Greenland Ice Sheet?

Posted by William Colgan on May 07, 2018
New Research / No Comments

We have a new open-access study in the current volume of Frontiers in Earth Science that tries to estimate snow density across the Greenland Ice Sheet1. Snow density might seem like an unexciting topic, but it is fundamental to blending ice-sheet thinning or thickening observations with surface mass balance simulations to assess ice-sheet health. Clearly, assuming a snow density of 400 kg/m3 makes a snowfall event observed by satellite altimeter twice as massive as assuming a snow density of 200 kg/m3 (and vice versa). There are several mathematical formulations presently being used to estimate snow density. These existing approaches generally estimate snow density as a function of more accessible geographic or climatic parameters.

RSF_figure1

Figure 1 – Locations of the surface snow density measurements collected in this public dataset. Contours lines indicate elevations in meters above sea level.

In this study, we assembled a large database of snow density measurements from the Greenland Ice Sheet. These measurements were collected from a variety of scientific expeditions going back to 1954, and provide the most complete spatial coverage of the ice sheet that is presently possible. Despite running a lot of statistics on this database, we could not find a compelling proxy for snow density. Our analysis indicates that snow density cannot be reliably predicted by common geographic (i.e. elevation, latitude or longitude) or climatic (i.e. air temperature or accumulation rate) variables. As existing approaches to estimate snow density rely on these common geographic and climatic variables, this was a somewhat unexpected finding.

RSF_figure2

Figure 2 – Snow density (0 to 10 cm depth) plotted against: (a) measurement year, (b) site latitude, (c) site longitude, (d) site elevation, (e) mean annual air temperature, and (f) accumulation rate.

Our study therefore recommends that the average measured density of 315 ± 44 kg/m3 (± standard deviation) is the most statistically defensible assumption for snow density. This recommendation of a constant, or zero-order approximation, differs from previous studies that have recommended estimating snow density as a second-order polynomial function of near-surface ice-sheet temperature. We show that these previous approaches may systematically overestimate snow density by 17 to 19 %. This is partially due to their mathematical formulations, but mainly due to previously considering measurement depths of up to 1 m as characteristic of “snow density”. As density increases with depth in the relatively porous near-surface layers of the ice sheet, we are instead careful to only include density measurements to a depth of 10 cm.

RSF_figure3

Figure 3 – Snow density (0-10 cm depth) versus mean annual air temperature. Solid line indicates the regression of this study, while the dotted and dashed lines indicate previously published temperature-dependent formulations for estimating snow density.

We hope that the approach of estimating snow density that we are proposing, which is mathematically less complex but statistically more robust, will be useful to researchers working with both surface mass balance simulations and satellite altimetry observations, as well as researchers modelling process-level studies of snow compaction and meltwater percolation in the near-surface ice-sheet layers. This study was supported by the Danish Research Council and the Programme for Monitoring of the Greenland Ice Sheet. Our database of 254 snow density measurements is freely available in the supplementary material of the study.

1Fausto R., J. Box, B. Vandecrux, D. van As, K. Steffen, M. MacFerrin, H. Machguth, W. Colgan, L. Koenig, D. McGrath, C. Charalampidis and R. Braithwaite. 2018. A Snow Density Dataset for Improving Surface Boundary Conditions in Greenland Ice Sheet Firn Modeling. Frontiers in Earth Science 6:51. doi:10.3389/feart.2018.00051.

Tags: , , , , , , , ,