This is A-Team's Typepad Profile.
Join Typepad and start following A-Team's activity
Join Now!
Already a member? Sign In
Recent Activity
That's great that the Arctic Sea Ice Blog can send a reporter to AGU 2014. Reporter Steve Bloom is quite right -- any time 22,000 attendees hold forth at a 3-4 day meeting, there will be a firehose of information (pumping out meltwater in this case). The AGU youtube channel is below. These are no real substitute for the ppts used in actual talks. These are in the public domain in some meaningless sense but good luck getting ahold of them. The Greenland news conference featured two very recently published papers that we've discussed over on the forum. These are both available as free full text. The PNAS one is at:
Toggle Commented yesterday on In the meantime: CryoSat at Arctic Sea Ice
Steve, that would be fantastic if you could attend. I'm thinking of swinging by myself. They don't seem to be providing live or archival session video even though the Moscone Center is well set up for that. I'm not seeing presenter slides or posters yet but it may be too soon for their posting. I expect them to approve your application without any fuss but we could even pass the paypal hat for a normal attendee fee. (If I donate even a penny to that, I would expect you to summarize the 23,000 submitted abstracts and run between all the Greenland sessions even if concurrent.) Regular: Full Meeting $425 Senior (65): Full Meeting $215 Additional friend guest $35 + AGU membership $50 (some full text journal access) Actually, the AGU has a very fast and effective search tool and I have already read through all 454 Greenland talks (ditto the Chamonix conf). This was very worthwhile and I recommend it to everyone for their areas of interest. Please note many of these abstracts will never result in a published paper. Some will be, but changed unrecognizably. A few are already old news. However on the whole, this will us a fantastic head's up on what is the works for 2015 publication. Please do not clog up the forums merely pasting abstracts. We can all go to the AGU site. Please post only if you can 'add value' in some way.
Toggle Commented Oct 28, 2014 on PIOMAS October 2014 at Arctic Sea Ice
Another fantastic year of coverage of this ongoing debacle, Neven! Given a flock of coal mine canaries -- some coughing and wheezing from aspergillosis, others staggering from mycoplasma, the remainder feverish with canary pox or spitting up blood from airsac mites -- I suppose we could argue over which will be first to fall off its perch. I still favor Arctic sea ice as being the first to go. However it's fair to say that greenhouse gas increase, atmospheric moisture content, Arctic shelf methane release, ocean acidification, sea level rise (WAIS ice shelf collapse, Greenland melt acceleration), ocean current rearrangement, stationary weather patterns (jet stream slowing and dipping) and so forth have their place already among the sick canaries. They'll be falling off the perch the way things are going.
Toggle Commented Sep 7, 2014 on PIOMAS September 2014 at Arctic Sea Ice
Blaine, a lot of good points. Briefly, we've moved over to the forum and I'm on my 5th post there. The Joughin article is based almost entirely on analysis of proprietary TerraSar imagery. You and I do not have access to it, neither did the referees. There is no way anyone can reproduce the figures when underlying data and methods are not provided. The best we can do is reverse-engineer figures -- count pixels in photoshop to re-create the missing numerical data and dissect their underlying DEM file (the Cresis ftp link above) which I did over at the forum in the course of making a bedrock flythru. The Gogineni group drastically binned the bedrock depth data, eg -1512 to -1103 navy blue, -1103 to -819 mid-blue, -819 to -535. That's why you can't resolve the colors. There's arguably heuristic value to these illustrations -- they get the main idea across without burdening the reader with scientific detail on troughs, pinning points and sills along the bottom of the channel. The temptation on a coordinate system is to take a central flowline of the main ice stream, because the best data comes from flight lines that follow it and velocities and mass fluxes are naturally measured across surfaces orthogonal to the velocity field. Earlier flights went after overall Greenland bed topography so rectangular grid lines made more sense. However there are problems using an ephemeral surface feature to define coordinates. This flowline may or may not lie over the lowest point in bedrock. Joughin uses the long-gone calving front of 2003 as origin of coordinate system origin and not quite a central streamline -- figure 1 starts out as an oblique line, changes to a curve, then cuts cross-channel between M9 and M10. It's also tempting to run the coordinate system through emplaced instrumentation, though that too is ephemeral. Because the ice stretches, they need both Eulerian and Lagrangian reference frames here. The latter means GPS or theodolite targets going along for the ride. Along-track coordinates could have been correlated to lat,lon but that was not done as a Fig.3 scale. So we have to put km coordinates along the photo track or determine from the DEM where the trough lies relative to the calving front on the image. I can't get too exercised over older inaccurate depth profiling when both IceBridge and the new radar drone are out there now updating glacial thinning (the part above sea level) and determining fine-grain features of channel bedrock (needed to model resistance flow). Inverse methods, deducing the bottom from deformations on the ice stream surface, have severe limitations as Sidd discusses over at the main Jakobshavn forum. There have to be consequences of something this massive and moving so rapidly while making a 43º turn. A huge new issue, freeze-on of meltwater leading to drastic deformations of overhead ice stratigraphy, may also be applicable to Jakobshavn.
Espen, agree it's better to consolidate these over in your Jakobshavn forum. However my registration there is hopelessly screwed up and there's no email for a forum admin. I'll see if Neven can help. Yes indeed, a century of observations on Jakobshavn Isbrae ... and it takes awhile even to digest the 2013-14 journal articles on it. Awright, I've finally managed a decent correlation map of bedrock depth with surface imagery: This represents a kridged CReSIS 2008 digital elevation map (from pdf after unzipping: rescaled 4.167:1 to overlay the 30 meter 24 June 14 Landsat-8. It is the DEM thought best by Joughin et al 2014. The transparency is set to ~30% so that ice stream features show through. The animation pauses at the last frame but will loop again. The big hole (~1300 m below sea level) co-locates with a standing wave feature in the surface imagery. Coincidence? You'll see something very similar as an ordinary river passes over a pothole.
So folks don't have to wrestle with a gigabyte images at 12 bit resolution, here is a piece of the 15 m Landsat-8 band B8 (panchromatic) of 24 June 14, namely LC80832332014175LGN00 as provided by I normalized and equalized the contrast histogram within ImageJ at 16 bit depth before export to your 8 bit monitor. Espen has posted 30 m bands 2,3,4 with a calving front comparison over at the forum. Even with 15 meter pixels, there's no forewarning of the next big calving event. The front is about to enter the narrow part of the ice stream.
Below is a schematic cutaway in longitudinal profile of how Jakobshavn Isbrae ice sits over bedrock. More realistically it would show meltwater along the bottom and perhaps deep erosional till cut from headwalls in the trough bottoms (per Tremoran above) as well as temperature and ice rheology profiles. Ice stratification layers may also deformed on Jakobshavn, though not to the extremes seen on Petermann or nearby Epiq. We need Andy Lee Robinson again to illustrate this properly: If the ice were taken away, the remaining glacial valley would be U-shaped in cross-sectional profile, other than some possible bumps (local pinning points that retard glacial advance). The bedrock form does not resemble a valley cut by a river because river longitudinal profiles do not have sills and troughs except at plunge pools, the counterpart of glacier overdeepenings. The next image shows the actual bed profile determined by ice-penetrating radar, as well as the position of the centerline of the calving front in recent years (from Joaghin 2014 linked above). Note the graphic shows the terminus retreating out of a -1300 m trough to the next sill at -950 m below sea level. There's an even deeper trough farther inland. The position of the terminus relative to the deep trough may explain the speed-up in recent years to a certain extent, indeed this year's early retreat to an approaching sill. That sill may diminish future warming of the base from waters of Baffin Bay. So well do we really know the bedrock profile? Here Gogineni himself says ice-penetrating radar is not reliable on narrow fast-moving ice streams. Echoes from the bed can be masked by surface debris and watery inclusions volume scatter. Trailing radar antenna wires on P3's cannot provide the optimal 14 and 35 MHz probes. The bigger problem though has been mis-interpretation of bottom freeze-on and consequent upward deformation as side-scatter from hills. This was recently disproven by gravity measurements showing no Greenland rock (2.67 gm/cm^3) in the vicinity. Meanwhile, Kansas group has developed a greatly improved instrument, a drone with antennas embedded in the airframe. This is capable of flying on a 5m grid cross-track, enabling 2D synthetic aperture radar, as described in a free article in the March 2014 issue of IEEE Geoscience and Remote Sensing. They've announced a field deployment on Jakobshavn Isbrae for this summer. Meanwhile Operation IceBridge flew straight up the channel this spring presumably for the same purpose. These two efforts should result in a greatly improved understanding of this ice stream's stratigraphy and so predictions of its future retreat and attributable sea level rise.
Nukefix has posted some early ESA Sentinel-1 polarized radar imagery of Epiq Sermia and nearby glaciers to the north (described at Glacier's perspective). The graininess can be removed either by Fourier frequency splitting, followed by blurring of the high frequency component (Gimp --> G'MIC) or by wavelet decomposition with discard of first level or two (Gimp --> Filters --> Generic). HH and VV are nearly identical, as can be seen from grain extract after RGB decomposition. It's not entirely clear what the polarizations tell us about ice properties per se. This resolution will allow us to follow calving at Jakobshavn Isbrae at 6-12 day repeat intervals. The imagery is free and puts us back in the ball game. Here we'll want to make a rolling RGB 'interferometric' animation to see which parts of Jakobshavn ice stream are moving at which relative velocities. The position of the calving front has been vacillating in accord with the trade-off between westward glacier advancement and eastward calving loss. So far, the calving front has not ventured up the narrow ice stream channel as it did last fall. Espen posted a very nice image for June 24 at the forum. (To recover individual frames, open animation in a separate tab, save, open gif in Preview or Gimp as layers.)
TL;DR: Landsat-8 is unsuitable for calving front time series. Very helpful link to June Landsat-8 acquisition schedule, Wipneus. Despite being a small ~5km feature, the Jakobshavn calving front is situated very unfavorably with respect to orbital scenes. The corners of path 7, row 11 appear in the middle of this scene; note the apparent drift. Tomorrow 24 June may provide the long-awaited updates to 10 June. Then again the calving front could be covered by a small cloud like yesterday. row path scene 10 11 LC80100112014175LGN00 83 233 LC80832332014175LGN00 Here (path, row) appear between LC80 and the date (day 175) in the scene names. LC80 and LGN00 (Landsat Global Network) do not seem informative. The SvalSat ground station in Svalbard will be receiving (10,11) and (83,233). Maybe.
Here is the latest synthesis of ice-penetrating radar bedrock recent analysis (free) from Gogineni and coworkers. Here I wanted to see at a glance what was below sea-level. However the authors provided a jpg'ed tinted and tilted shaded relief, causing the color key to get out of synch with the image itself. The second slide shows a cure for this: pass to HSV color space and toss the S and V components. Then return to RGB and additively color-pick the key to create a mask for the land above sea level. I made the mask slightly translucent. The bedrock under Jakobshavn Isbrae has numerous ups (sills) and downs (overdeepenings) along its course. Calving depends to a considerable extent on the position of the front relative to this profile.
These folks have grant access to much better imagery. AGU abstracts fall 2013, not yet published. And when they are, the underlying imagery still won't be available. Still, we can muddle along in real-time with our low-res freebies to report breaking news, summer 2014. Analysis of the seasonal and interannual evolution of Jakobshavn Isbrae from 2010-2013 using high spatial/temporal resolution DEM and velocity data DE Shean, IR Joughin, BE Smith et al Greenland's large marine-terminating outlet glaciers have displayed marked retreat, speedup, and thinning in recent decades. Jakobshavn Isbrae, one of Greenland's largest outlet glaciers, has retreated ~15 km, accelerated ~150%, and thinned ~200 m since the early 1990s. Here, we present the first comprehensive analysis of high spatial (~2-5 m/px) and temporal (daily-monthly) resolution elevation and velocity data for Jakobshavn from 7/2010 to 7/2013. We have developed an automated processing pipeline using open-source software (Ames Stereo Pipeline, GDAL/OGR, NumPy/SciPy, etc.) to produce orthoimage, digital elevation model (DEM), and surface velocity products from DigitalGlobe WorldView-1/2 stereo imagery (~0.5 m/px, ~17 km swath width). Our timeseries consists of 35 WorldView DEMs (~2-4 m/px) covering the lower trunks of the main+north branches and fjord, but also extending >110 km inland. We supplement this record with 7 TanDEM-X DEMs (~5 m/px, ~35 km swath width) between 6/2011-9/2012. Elevation data from IceBridge ATM/LVIS, ICESat GLAS, and GPS campaigns provide absolute control data over fixed surfaces (i.e., exposed bedrock). Observed WorldView DEM offsets are consistent with DigitalGlobe's published value of 5.0 m CE90/LE90 horizontal/vertical accuracy. After DEM co-registration, we observe sub-meter horizontal and vertical absolute accuracy. Velocity data are derived from TerraSAR-X data with 11 day repeat interval. Supplemental velocity data are derived through correlation of high-resolution WorldView DEM/image data. The contemporaneous DEM and velocity data provide full 3D displacement vectors for each time interval, allowing for the analysis of both Eulerian and Lagrangian elevation change. The lower trunk of Jakobshavn displays significant seasonal velocity variations, with recent rates of ~8 km/yr during winter to >17 km/yr during summer. DEM data show corresponding elevation changes of -30 to -45 m in summer and +15 to +20 m in winter, corresponding to integrated volumes of -1.0 to -1.5 km3 and +0.3 to +0.6 km3 for the lower ~20 km of the main trunk. The DEM time series shows a net interannual thinning trend of -15-20 m/yr for lower Jakobshavn, with decreasing magnitude upstream. This is consistent with long-term altimetry records. Interannual thinning of ~2 m/yr and lateral retreat is also observed for grounded ice sheet margins. /*/*/*/* We're obsolete! Automated tracking of changes to terminus and mélange position in Greenland's marine outlet glaciers F Seifert, JN Bassis et al Mass loss from the Greenland ice Sheet is primarily through the dynamic changes of its marine terminating outlet glaciers. Understanding the behavior of these glaciers is therefore key to understanding how much the ice sheet will contribute to sea level in the next century. Glacier behavior is, however, complex, with wide disparities in behavior even between glaciers that occupy adjacent fjords. Deciphering the multitude of factors that control glacier behavior requires a comprehensive dataset of near daily changes in terminus position for a large set of glaciers over many years. The creation of this dataset has been difficult due to the time required to process the changes manually. Evolution in computational methods allows the creation of an automated algorithm, using a combination of filtering techniques and edge detection, which ingests MODIS imagery and tracks changes to (i) the terminus position and (ii) the areal extent of melange downstream of the terminus. We tested the algorithm on several well-studied glaciers including Jakobshavn, Helheim, and Kangerdlugssuaq. Comparison with manually identified terminus positions proved the algorithm accurate to within +/- 2 pixels (500 m). The validated model was then applied to a larger set of Greenland's marine terminating outlet glaciers. We use this higher temporal resolution dataset to determine statistical patterns to the calving events and ask whether these patterns are linked to mélange extent, fjord geometry, and any seasonal component affecting the regularity of the events.
Espen wrote about an apparent calving event on June 13 that needs confirmation by better than 250m resolution imagery: "... I can't figure out Landsat-8 release schedules and priorities." Me either. Aggravating, inexplicable delays in providing data (or non-acquisition thereof). The USGS interface is very peculiar, maybe an unpaid summer intern's first venture into programming. Today I came across something a lot better -- Reverb -- a really first-rate unified portal not just to Landsat-8 but to all of EOS (including IceBridge). The satellite itself orbits every 99 minutes, with the latest orbit displacing the previous one to the east. USGS promises a one day turnaround on imagery. So how can June 10th still be the most recent image on June 21st? Landsat-8 is not part of the A-train like Terra and GCOM-W1. You can follow its position in real time I just watched it go over Petermann Glacier so last time around it swung over Jakobshavn Isbrae. Well, not quite: Modis Terra/Aqua stalled for 4 days on 17 June, then grudgingly updated itself four days later. This imagery shows skies were clear over the Jakobshavn Isbrae calving front on June 1, 2, 8, 10-16, and 20. Thus clouds cannot be the explanation for the missing Landsat coverage. When I upsample and animate 250m June (not shown), the images are slightly out of register. That jiggle could be due to satellite orbit not quite repeating itself, the timing or lookdown not quite giving the same coverage, or something not done right in processing. Between this and 250m resolution for something on a km scale, the apparent calving event of June 13 is still up for grabs. However Terra/Aqua do give a very satisfactory account of the abrupt development of melt lakes to the south. The grayish tone may represent bare ice (loss of snow). Jakobshavn Isbrae does *not* respond notably to sub-glacial lubrication by meltwater reaching the bedrock via moulins. This was worth animating the aforementioned cloud-free days: The downstream fjord mélange is cannot really being providing much back-pressure during calving season. Full height icebergs break off with the bottom ice rolling out. Since the ice at the calving front today is 1179 m thick (1075m below sea level, 104 m above) according to fig 2 of Joaghin 2014, the mélange gets bumped 3686 feet out to sea during rather short-lived events (which rules out significant back-pressure), as you can see in the 4 calving event videos for 2008. Incidentally, the dramatic 'Chasing Ice' calving at Jakobshavn Isbrae was filmed in 2008. I believe it matches the May 10th one below. time-lapse video of a calving event on 10 May 2008 time-lapse video of a calving event on 16 May 2008 time-lapse video of a calving event on 15 July 2008 time-lapse video of a calving event on 24 July 2008 For that year, a fantastic graphic showing terminal speed-up at Jakobshavn Isbrae appears in David Podarski's dissertation. I consolidated the color key and cropped; otherwise it is worth seeing as he presented it (which requires native image extraction at
I looked into quick exits from the quirky Landsat-8 interfaces (eg and In a November forum post, Wipneus provided an excellent way of doing this primarily in terminal mode (ImageMagick). The issues here are five-fold: getting the most recent image without delay (here the calving front), dealing with gigabyte file sizes of high resolution images, taking advantage of the extra data in NASA's '16 bit' Tiff format, best 3 of the 12 bands for an RGB, and pipelining for a seasonal time series animation. I found a day or so delay in image availability -- they're still on yesterday. The GeoTiff folder arrives rapidly enough from EarthExplorer on a home 17 Mbps downlink, giving individual bands as 145 MB grayscales. There is no need for 16 megs of computer RAM, 4 works fine. The visible bands B2-4 make exceedingly poor use of the available dynamic range (histogram). Opening them in Gimp cuts them down to 8 bit, meaning subsequent contrast expansion will posterize them severely vs doing that step within the native 12 bit. However ImageJ -- which is highly developed NIH biomedical freeware -- can import and adjust contrast natively at 16 bits, then conveniently export the stack to Gimp as a gif animation which can then be cropped (vastly reducing file size) and composited as a color image at Gimp, best done with B5, B4, B2. For purposes of the watching the calving front, little is gained by this. Both ImageJ and Gimp have batch processing which do all these steps in the background for a time series. The 15m resolution panchromatic is still the crazy-making 17061 × 17161 pixels, one pixel less than 2x the band sizes of 8531 × 8581. Dumbing up, the 15 m can be used as the V or L in HSV or HSL color space when the 15m ground resolution is informative. In summary, capturing high resolution Landsats are easier than when Wipneus wrote about them 8 months ago. Still, the interface is like the proverbial dog walking on its hind legs, not done well but we can applaud that the imagery is available at all. 800 x 450 DEM of calving front:
The slide show below compares Jakobshavn Isbrae in June for 2013 and 2014. It is not so easy to find cloud-free days on the same date for both years. (The dates match except for the first pair.) The main things to see: mélange (ice jam of calved-off ice) is still blocking the channel in mid-June 2014 and the calving front is farther advanced. I'm skeptical that the mélange is currently providing significant back pressure to glacier advance, though it may in mid-winter when frozen or when iceberg bottoms are caught on the terminal moraine off Illulisat. The tide at the calving front can be three meters -- the 2013 series flushes out on the time scale of a day. There's an active webcam on the NE shore aimed at the calving front (not to be confused with Hotel Arctic webcam at Illulisat). Camera A takes hourly shots; you can roll your own animation or step through one frame at a time (eg to get daily series at noon). It's been operated by David M. Holland at NYU for 7 years. I made the animation below -- sort of a low budget Chasing Ice -- for 1 Jun 14. Note it shows air temperature too. The camera is too far from the current front to capture calving events. Holland is best known for a 2008 paper relating abrupt warming of offshore waters in 1997 with the onset of rapid glacier retreat, diving seals with sat phones, and a more recent poster on fjord waters coming to resemble those of Disco Bay. poster: cam:
Heraclitus said something to the effect you can't step into the same isbrae twice. The Jakobshavn Isbrae is a case in point, moving 17 km/yr = 326 m/week = 46.6 m/day = 1.94 m/hr upstream of the calving front at marker M6 (shown above) on 15 June 2012. Consequently the calving front must break off this much ice just to hold the line (which sometimes it doesn't). Given Landsat-8 images with 250 meter pixels, a pixel at the calving front is replaced by one of upstream ice every 5.3 days. Thus I've been casting around looking for better quality imagery so that we can better monitor calving events, measure glacier thinning and speed-up, determine shearing across the centerline, animate a year of melt and predict terminal position for 2014 and coming years. Some of Espen's imagery seems to be better quality but I've not seen a link to it. I've been using the one below, refreshng the day (166 is 15 June) mid-morning Pacific coast time. None of our usual products at DMI, Jaxa, WorldView are satisfactory -- clouds can mask visible imagery; the ASAR radar has insufficient resolution, sporadic coverage and does not come with pixel-perfect co-registration. Google Map today provides a very high resolution image in the visible. However it is a mosaic of different dates and time of day; further, it is not current and we have no idea when the next update will arrive. Google Earth (separate freeware) does archive a half-dozen very high resolution images from 2008-2012. To access, click on the alarm clock icon in the tool bar. These give a much better feel for seracs, crevasses, dynamic melt lakes, ridges and flow lines upstream of the calving front itself. The images are rather large at native resolution so the slide show below shows just a fraction of what is out there. The hands-down winner is TerraSAR-X and its companion satellite TanDEM-X, operated by DLR (the German NASA). These are not free (200 euros a pop). Luckily, speedup of Jakobshavn Isbrae was an featured project for 2013, and a couple of incredible images are provided in their gallery, The first image is a truly astonishing digital elevation map of the glacier rendered in perspective from the south, 2222 x 1250. It looks so much like an aerial photograph that I had to decompose it into CYMK color space to convince myself it is totally synthetic (no yellow). Can you imagine using a time series of these to animate the ice stream! The second image is a hideous overhead grayscale from 08 Jan 14 useful as a base layer, similar to the 20 Sept 13 image of maximal calving retreat shown above but without its ruinous overlays. Very few climate scientists 'get it' in terms of image layering despite arcGIS etc being around since forever: NASA is the only place I've seen doing image overlays correctly (as separate, co-registered files). I myself have stuck an extra 120 pages of journal techno into supplemental -- the journals don't care so what is the excuse here? Turns out the authors of that key Cryosphere article(free did all their velocity measurements on six years using TerraSAR-X speckling. The trick to free imagery must be to co-author your paper with someone who works for DLR in SAR signal processing ( I'll write to request a 2013 animation and see if we can host free summer 2014 imagery. Nukefix posted a fine animation of Kangerdluqssuaq from a different SAR satellite at the forum, noting Jakobshavn is an ESA Sentinel-1 "supersite" that will be imaged on a 6 day return by fall (if not sooner) Free.,154.msg27707.html#msg27707
Steve, good points -- the data here are under-analyzed. (The authors say it was intended only as a brief update to their 2012 paper, There's more data on velocity speed-up in Fig 2 of this paper. Fig 1 has the TerraSAR-X image for 20 Sep 13 maximum retreat that goes with Espen's Landsat-8 visible bands. Astonishingly, despite hundreds of journal articles on Jakobshavn Isbrae, no one ventures much by way of validatable predictions (this fall, not end-of-century). You can see timidity alternatives play out between reviewers (it's speculation) and authors (no it's hypothesizing) in the Cryosphere Today discussion section for this article. Yet expert opinion from folks who have devoted entire careers to this particular glacier would have a lot of value. I'll post a graphic shortly showing my business-as-usual maximal retreat prediction for Sept 14 as well as a drama queen 2x version. Discharge volume (km^3 or as microns of sea level rise) is a little tricky because of thinning. The way raster images are handled in a pdf is very complicated (see The best way to extract them at optimal (author-supplied) resolution is not via a rescaled (dithered) screenshot off your pdf reader but rather by (or similar). Once you're sure the display is right, a screenshot or crop will be lossless. For everyone's convenience, I collected the native width x height of these graphics and displayed that of the synthetic aperture radar. They will be right-truncated by the blog so 'view image' in a separate browser tab. You'll get cheated within a pdf viewer. 1033 x 1204 TerraSAR-X imag 1033 x 1008 Fig 2 speeds at various distances from terminus 1033 x 770 Fig 3 seasonality of calving
Neven writes, "I will follow up on this blog post if and when Jakobshavn moves past last year's retreating line." Ok, the 'if' part is a done deal -- Jakobshavn Isbrae sets a new record every year. None of that annoying 'natural variability' like we have with the Arctic sea ice death spiral. The 'when' of the maximum retreat is hard to predict. While the two massive calving events noted by Espen seem awful early, the recent historical record shows the maximal retreat (or near-maximal) can occur quite early. The graphic shows this with dotted white boxes. (Here I selected colored circles on the far right for each year, requiring that that color selection also show up in the terminus position bar.) The 'when' of the next calving event could be predicted from real time laser altimetry. The calving front rises some 15 m while the back sags ahead of the event, with only the first km affected. However the lead time of the prediction would be quite short. The 'how much' of a new record set is the interesting part. I have a queasy feeling about Greenland overall, that it is becoming unglued ahead of schedule. But here, the ice stream is coming off a sharp turn with lateral constraints and also ending its run over the deepest part of its channel so even if we see say 3x the km of average recent retreat, that will still need interpretation. Fortunately Jakobshavn Isbræ is heavily instrumented as poster child of Greenland glaciers (though not representative of all) and a major current contributor to sea level rise. It's a credit to Espen and this blog to be the first to report these calving events that contribute directly to this year's SLR, though it would be great if someone would calculate the volume more carefully along with the associated uncertainty. Wipneus posted the pixel scale for these Landsat-8's over on the forum. Graphic adapted from See also and
Whoa, this is getting interesting. The image below shows the starting position and two calving events of this month, overlaid on the position of maximum retreat attained at end of season last year. Despite some enhancement, the second image gives no indication of an impending 3rd calving event. Indeed as MPelto notes above, "Lack of surficial transverse crevassing is unusual before a large calving event. Basal crevassing may be key." The trick here, on these overly blue Landsat-8 images, is decompose into CMY and fade the equalize of the magenta layer. The speed of this glacier is such that if we could get 3 more or less consecutive clear days, a follow-on 'interferometric' RGB would show the relative velocities above the calving line. (The DMI ASAR images lack sufficient resolution.) The 3rd image shows, among other things, the bed profile and historical seasonality of retreat. Note the maximal fjord overdeepening about where the calving front is today. It is taken from Joughin 2014 'Further summer speedup of Jakobshavn Isbræ'
Ouch ... the new Navy Hycom algorithm is way off-base. Operationally, a couple of stops past hazardous -- far better off with direct imagery. The old algorithm is being irrevocably decommissioned on 31 Aug 13 -- too bad, it had some good weeks. The fade-out animation has 5% transparency increments. It is slightly posterized because of the gif restriction to 256 colors despite Floyd-Steinberg dithering. Hycom has 29.4% the resolution of 89 Ghz Jaxa and so had to be upsampled for the overlay.
Toggle Commented Aug 23, 2013 on Hole at Arctic Sea Ice
I'm starting to appreciate the full gamut of microwave frequencies -- a learning curve with it but eventually it starts to make some sense. The one below might have some predictive value for the week ahead.
Toggle Commented Aug 22, 2013 on Hole at Arctic Sea Ice
Hycom ice color has always been jpg'ed and never corresponded to the colors in the key (which was added back to the image afterwards). The third mistake was not doing the key initially as overtint of a uniform grayscale. Thus it is all but impossible to fix the palette without drilling down into their file system to find the raw numeric array, assuming they archived that. Grayscale is the universal scientific standard for numeric manipulation of images displaying 1D data (concentration, thickness etc are maps: I^2 -> +R^1). As things stand, we have no real mechanism for quantitative comparison of new and old. For what it was worth, I subtracted the two 01 Aug 13 images Blaine supplied, adding neutral gray 128 to keep everything within bounds [0 255] while sticking to native resolution (1/12º ~ 10km/pixel). (Their documentation speaks of 3.5 km resolution at the pole, applicable perhaps to the one map inexplicably offered in Mollweide; the rest are stereographic. There, simply count pixels from pole to the Siberian shore grazed by 70ºN to determine resolution.) I've always wondered where the heck they get their underlying imagery and why they don't check in occasionally with something better to ground the product model in observation. Ice thickness seems to be calibrated off a sparse radar altimeter: "NCODA assimilates available satellite altimeter observations (along track obtained via the NAVOCEANO Altimeter Data Fusion Center), satellite ice concentration..." The satellite instrument and concentration algorithm are not specified but I suspect none of the ones we use. This is a highly sophisticated model outputting a product that never stood up to the slightest scrutiny -- seemingly the paradigm for climate science modelling (no result, publish anyway): "Hydrostatic approximation, hybrid pressure coordinate, fourth-order LaPlacian for vorticity, divergence, virtual potential temperature, specific humidity, surface pressure, ground temperature, ground wetness, cloud fraction, horizontal spectral differencing, second-order finite difference in the vertical, central time differencing with Robert semi- implicit corrections,42 sigma levels with 6 sigma below 850 mb,00Z, 06Z, 12Z and 18Z ops run, Machenhauer initialization, NOGAPS 6-h or 12-h forecast, spectrally truncated and Lanczos filtered heights, convective precipitation (Emanuel), shallow cumulus mixing (Tiedtke) and large-scale convection, long-wave and short-wave radiation (Harshvardhan) computed every 2 hour, gravity wave drag, planetary boundary layer, sea surface temperature and ice coverage percentage from OCEAN MVOI."
Toggle Commented Aug 22, 2013 on ASI 2013 update 7: cold and cloudy at Arctic Sea Ice
Abrupt disconnect over at Navy Hycom, as first noticed over on the forum. They've radically increased the thickness and altered distribution of first year ice. Some models need to 'tie in' periodically to observational data to reset themselves to reality. Fine. But where would that come from with ice thickness in mid-August? Apart from correct peripheral ice boundary, they're still wildly inconsistent with high resolution satellite display of unambiguous open water [Modis and 89Ghz], 2 m being worse than before. If this is an revised algorithm, how was it judged ready for release? I could find no explanation offered -- surely they have an unpaid intern not too busy to write up a few lines. I'm not sure how far back they are recomputing but 01 Aug 13 at least. Also it appears they have overwritten previous 30 day runs. What then becomes of year to year comparison?
Toggle Commented Aug 21, 2013 on ASI 2013 update 7: cold and cloudy at Arctic Sea Ice
Yes, indeed DanP. Also contact earthdata.nasa since you have improved on their offering in a way that will interest many end users. For the hdf and netCDF files, I'll remind people of free desktop software Panoply. However 18 GB of daily download is not going to prove feasible for most people. And when downloaded, you cannot afford to experiment on very large files so get your pipeline down on smaller test files. Even the last few weeks, which have seen very little coherent ice motion, still introduce too much jitter into an eight day product relative to 0.5 km pixel resolution, so three cheers for 2.5 days. While floe motion vectors are supposedly provided daily and Gimp etc offer very nuanced motion blur correction, I'm sceptical that these could really improve time-composited ice imagery, though 2.5 days is a lot more favorable for that. I looked at band 31 (11 micron infrared, 1 km resolution) on earthdata.nasa to see if it offered anything to complement the shorter wavelengths you have been using but it has mostly to do with cloud temperature. Oblique cloud shadows on ice are removable -- unlike thick clouds directly over ice. Cloud height could also be obtained from the sun angle at swath time. In fact, an overall daily cloud/fog/vapor product from all the wavelengths out there would be very useful for interpreting conditions affecting the ice.
Toggle Commented Aug 21, 2013 on ASI 2013 update 7: cold and cloudy at Arctic Sea Ice
Summer brings clouds whose water vapor interferes with analysis, then and now, so take-home lessons of cyclones for the ice are not fully realized. Despite incredible efforts of Danp on best-swaths and pixel by pixel cloud mask, which have gone retrospective over on the forum. For sure, the pipeline products we use have not found a solution to impenetrable clouds. For example, what is going on here -- your choice of melt / divergence in an unexpected place vs cloud artefact implausibly affecting very different wavelengths:
It may be difficult with shelf methane to improve on satellites + more frequent field campaigns. It is far easier to measure methane in the air column than local methane supersaturation in water. I was going to suggest building additional icebreakers to augment interpretation of satellites -- notably our best instrument AMSR2 -- beyond what more buoys could do until I came across the USCGC Healy web site and found it takes a whole lot of resources to get very little done. The Healy has not ventured out into the ice in 2013 and have no plans to do so. First, BOEM engaged them to carry out fatuous oil spill containment exercises -- Commercial-Off-The-Shelf (COTS) and Government-Off-The-Shelf (GOTS) technologies -- apparently to prop up justification of the oil drillng program. The logistics of containment based out of the nearest Coast Guard station (Kodiak, a thousand miles away) is completely unworkable ten months of the year. Next, acoustic releases to service 5 moored buoys and installed 9 new ones, making some near-shore current and salinity measurements in the vicinity of Barrow Canyon, Admundsen Gulf and Hanna Shoal. Ok. The final cruise towed CHIRP sonar and made shallow cores with occasional CTD stations and thermister drops, with the major cruise objective finding "geological evidence of a massive flood that came north to the Arctic via the Mackenzie River about 13,000 years ago." Fine. We're not learning anything about the new ice situation though. Here they are -- circling over a lease within rowboat distance of Barrow, being serviced at Vigor Marine in Seattle: