A summary of landfast sea ice coverage and the changes in the distance between the penguin colony at Point Geologie and the nearest span of open water on the Adelie Land coast in East Antarctica. The data were derived from cloud-free NOAA Advanced Very High Resolution Radiometer (AVHRR) data acquired between 1-Jan-1992 and 31-Dec-1999. The areal extent and variability of fast ice along the Adelie Land coast were mapped using time series of NOAA AVHRR visible and thermal infrared (TIR) satellite images collected at Casey Station (66.28 degrees S, 110.53 degrees E). The AVHRR sensor is a 5-channel scanning radiometer with a best ground resolution of 1.1 km at nadir (Cracknell 1997, Kidwell 1997). The period covered began in 1992 due to a lack of sufficient AVHRR scans of the region of interest prior to this date and ended in 1999 (work is underway to extend the analysis forward in time). While cloud cover is a limiting factor for visible-TIR data, enough data passes were acquired to provide sufficient cloud-free images to resolve synoptic-scale formation and break-up events. Of 10,297 AVHRR images processed, 881 were selected for fast ice analysis, these being the best for each clear (cloud-free) day. The aim was to analyse as many cloud-free images as possible to resolve synoptic-scale variability in fast ice distribution. In addition, a smaller set of cloud-free images were obtained from the Arctic and Antarctic Research Center (AARC) at Scripps Institution of Oceanography, comprising 227 Defense Meteorological Satellite Program (DMSP) Operational Linescan Imager (OLS) images (2.7 km resolution) and 94 NOAA AVHRR images at 4 km resolution. The analysis also included 2 images (spatial resolution 140 m) from the US Argon surveillance satellite programme, originally acquired in 1963 and obtained from the USGS EROS Data Center (available at: edcsns17.cr.usgs.gov/EarthExplorer/). Initial image processing was carried out using the Common AVHRR Processing System (CAPS) (Hill 2000). This initially produces 3 brightness temperature (TB) bands (AVHRR channels 3 to 5) to create an Ice Surface Temperature (IST) map (after Key 2002) and to enable cloud clearing (after Key 2002 and Williams et al. 2002). Fast ice area was then calculated from these data through a multi-step process involving user intervention. The first step involved correcting for anomalously warm pixels at the coast due to adiabatic warming by seaward-flowing katabatic winds. This was achieved by interpolating IST values to fast ice at a distance of 15 pixels to the North/South and East/ West. The coastline for ice sheet (land) masking was obtained from Lorenzin (2000). Step 2 involved detecting open water and thin sea ice areas by their thermal signatures. Following this, old ice (as opposed to newly-formed ice) was identified using 2 rules: the difference between the IST and TB (band 4, 10.3 to 11.3 microns) for a given pixel is plus or minus 1 K and the IST is less than 250 K. The final step, i.e. determination of the fast ice area, initially applied a Sobel edge-detection algorithm (Gonzalez and Woods 1992) to identify all pixels adjacent to the coast. A segmentation algorithm then assigned a unique value to each old ice area. Finally, all pixels adjacent to the coast were examined using both the segmented and edge-detected images. If a pixel had a value (i.e. it was segmented old ice), then this segment was assumed to be attached to the coast. This segment's value was noted and every pixel with the same value was classified as fast ice. The area was then the product of the number of fast ice pixels and the resolution of each pixel. A number of factors affect the accuracy of this technique. Poorly navigated images and large sensor scan angles detrimentally impact image segmentation, and every effort was taken to circumvent this. Moreover, sub-pixel scale clouds and leads remain unresolved and, together with water vapour from leads and polynyas, can contaminate the TB. In spite of these potential shortcomings, the algorithm gives reasonable and consistent results. The accuracy of the AVHRR-derived fast ice extent retrievals was tested by comparison with near- contemporary results from higher resolution satellite microwave data, i.e. from the Radarsat-1 ScanSAR (spatial resolution 100 m over a 500 km swath) obtained from the Alaska Satellite Facility. The latter were derived from a 'snapshot' study of East Antarctic fast ice by Giles et al. (2008) using 4 SAR images averaged over the period 2 to 18 November 1997. This gave an areal extent of approximately 24,700 km2. The comparative AVHRR-derived extent was approximately 22,240 km2 (average for 3 to 14 November 1997). This is approximately 10% less than the SAR estimate, although the estimates (images) were not exactly contemporary. Time series of ScanSAR images, in combination with bathymetric data derived from Porter-Smith (2003), were also used to determine the distribution of grounded icebergs. At the 5.3 GHz frequency (? = 5.6 cm) of the ScanSAR, icebergs can be resolved as high backscatter (bright) targets that are, in general, readily distinguishable from sea ice under cold conditions (Willis et al. 1996). In addition, an estimate was made from the AVHRR derived fast ice extent product of the direct-path distance between the colony at Point Geologie and the nearest open water or thin ice. This represented the shortest distance that the penguins would have to travel across consolidated fast ice in order to reach foraging grounds. A caveat is that small leads and breaks in the fast ice remain unresolved in this satellite analysis, but may be used by the penguins. We examine possible relationships between variability in fast ice extent and the extent and characteristics of the surrounding pack ice (including the Mertz Glacier polynya to the immediate east) using both AVHRR data and daily sea ice concentration data from the DMSP Special Sensor Microwave/Imager (SSM/I) for the sector 135 to 145 degrees E. The latter were obtained from the US National Snow and Ice Data Center for the period 1992 to 1999 inclusive (Comiso 1995, 2002). The effect of variable atmospheric forcing on fast ice variability was determined using meteorological data from the French coastal station Dumont d'Urville (66.66 degrees S, 140.02 degrees E, WMO #89642, elevation 43 m above mean sea level), obtained from the SCAR READER project ( www.antarctica.ac.uk/met/READER/). Synoptic- scale circulation patterns were examined using analyses from the Australian Bureau of Meteorology Global Assimilation and Prediction System, or GASP (Seaman et al. 1995).

Maps of East Antarctic landfast sea-ice extent, generated from approx. 250,000 1 km visible/thermal infrared cloud-free MODIS composite imagery (augmented with AMSR-E 6.25-km sea-ice concentration composite imagery when required). Because of imperfections in the MODIS composite images (typically caused by inaccurate cloud masking, persistent cloud in a given region, and/or a highly dynamic fast-ice edge), automation of the fast-ice extent retrieval process was not possible. Each image was thus classified manually. A study of errors/biases of this process revealed that most images were able to be classified with a 2-sigma accuracy of +/- ~3%. More details are provided in Fraser et al., (2010). *Version 1.2 with extra QC around the Mawson coast and Lutzow-Holm Bay The directory named "pngs" contains browsable maps of fast-ice extent, in the form of Portable Network Graphics (PNG) images. Each of the 159 consecutive images (20-day intervals from Day Of Year (DOY) 61-80, 2000 to DOY 341-366, 2008) contains a map of fast-ice extent along the East Antarctic coast, generated from MODIS and AMSR-E imagery. The colour scale is as follows: Dark blue: Fast ice, as classified from a single 20-day MODIS composite image Red: Fast ice, as classified using the previous or next 20-day MODIS composite images Yellow: Fast ice, as classified using a single 20-day AMSR-E composite image White: Antarctic continent (including ice shelves), as defined using the Mosaic of Antarctica product. Light blue: Southern ocean/pack ice/icebergs These maps are also provided as unformatted binary fast ice images, in the directory named "imgs". These .img files are all flat binary images of dimension 4300 * 425 pixels. The data type is 8-bit byte. Within the .img files, the value for each pixel indicates its cover: 0: Southern Ocean, pack ice or icebergs, corresponding to light blue in the PNG files. 1: Antarctic continent (including ice shelves), as defined using the Mosaic of Antarctica product, corresponding to white in the PNG files. 2: Fast ice, as classified from a single 20-day MODIS composite image, corresponding to dark blue in the PNG files 3: Fast ice, as classified using a single 20-day AMSR-E composite image, corresponding to yellow in the PNG files 4: Fast ice, as classified using the previous or next 20-day MODIS composite images, corresponding to red in the PNG files To assist in georeferencing these data, files containing information on the latitude and longitude of each pixel are provided in the directory named "geo". These files are summarised as follows: lats.img: File containing the latitude of the centre of each pixel. File format is unformatted 32-bit floating point, 4300 * 425 pixels. lons.img: File containing the longitude of the centre of each pixel. File format is unformatted 32-bit floating point, 4300 * 425 pixels. The .gpd Grid Point Descriptor file used to build the projection is also included. It contains parameters which you can use for matching your projection. To refer to the time series, climatology, or maps of average persistence, please reference this paper: Fraser, A. D., R. A. Massom, K. J. Michael, B. K. Galton-Fenzi, and J. L. Lieser, East Antarctic landfast sea ice distribution and variability, 2000-08, Journal of Climate 25, 4, pp. 1137-1156, 2012 In addition, please cite the following reference when describing the process of generating these maps: Fraser, A. D., R. A. Massom, and K. J. Michael, Generation of high-resolution East Antarctic landfast sea-ice maps from cloud-free MODIS satellite composite imagery, Elsevier Remote Sensing of Environment, 114 (12), 2888-2896, doi:10.1016/j.rse.2010.07.006, 2010. To reference the techniques for generating the MODIS composite images, please use the following reference: Fraser, A. D., R. A. Massom, and K. J. Michael, A method for compositing polar MODIS satellite images to remove cloud cover for landfast sea-ice detection, IEEE Transactions on Geoscience and Remote Sensing, 47 (9), pp. 3272-3282, doi:10.1109/TGRS.2009.2019726, 2009. Please contact Alex Fraser (adfraser@utas.edu.au) for further information.

Some ecosystems can undergo abrupt transformation in response to relatively small environmental change. Identifying imminent "tipping points" is crucial for biodiversity conservation, particularly in the face of climate change. Here we describe a tipping point mechanism likely to induce widespread regime shifts in polar ecosystems. Seasonal snow and ice cover periodically block sunlight reaching polar ecosystems, but the effect of this on annual light depends critically on the timing of cover within the annual solar cycle. At high latitudes sunlight is strongly seasonal, and ice-free days around the summer solstice receive orders of magnitude more light than those in winter. Early melt that brings the date of ice-loss closer to midsummer will cause an exponential increase in the amount of sunlight reaching some areas per year. This is likely to drive ecological tipping points in which primary producers (plants and algae) flourish and out-compete dark-adapted communities. We demonstrate this principle on Antarctic shallow seabed ecosystems, which our data suggest are sensitive to small changes in the timing of sea-ice loss. Algae respond to light thresholds that are easily exceeded by a slight reduction in sea-ice duration. Earlier sea-ice loss is likely to cause extensive regime-shifts in which endemic shallow-water invertebrate communities are replaced by algae, reducing coastal biodiversity and fundamentally changing ecosystem functioning. Modeling shows that recent changes in ice and snow cover have already transformed annual light budgets in large areas of the Arctic and Antarctic, and both aquatic and terrestrial ecosystems are likely to experience further significant change in light. The interaction between ice loss and solar irradiance renders polar ecosystems acutely vulnerable to abrupt ecosystem change, as light-driven tipping points are readily breached by relatively slight shifts in the timing of snow and ice loss. This archive contains data and statistical code for the article: Graeme F. Clark, Jonathan S. Stark, Emma L. Johnston, John W. Runcie, Paul M. Goldsworthy, Ben Raymond and Martin J. Riddle (2013) Light-driven tipping points in polar ecosystems. Global Change Biology Data and code are organised into folders according to figures in the article. See the article for a full description of methods. Statistical code was written in R v. 2.15.0. In data files, rows are samples and columns are variables. Details for numerical variables in each data file are listed below. Figures 7 and 8 were made in MATLAB and code is not provided. Figure 1: rad_data.csv Solar irradiance data derived from: Suri M, Hofierja J (2004) A new GIS-based solar radiation model and its application to photovoltaic assessments. Transactions in GIS 8: 175-190. Figure 2: Fig. 2c.1.csv Light: Measured light at the seabed per day (mol photons m-2 d-1). Figure 2: Fig. 2c.2.csv Light: Measured light at the seabed per day (mol photons m-2 d-1). Light.mod.p: Light at the seabed per day (mol photons m-2 d-1) predicted from modeled seasonal variation. Figure 2: Fig. 2d.csv Light: Measured light at the seabed per day (mol photons m-2 d-1). Figure 3: Fig. 3a.csv Irradiance: Mean irradiance (micro mol photons m-2 s-1). P/R: Productivity/respiration ratios (micro mol photons O2-1 gFW-1 h-1). Figure 3: Fig. 3b.csv Light: Mean irradiance (micro mol photons m-2 s-1) in experimental treatments. Growth: Thallus growth (mm) of Palmaria decipiens under experimental treatments. Figure 3: Fig. 3c.csv Des, Him, Irr, Pal: Ice-free days required for minimum annual light budget Figure 3: Fig. 3c.bars.csv Prop: relative cover (sums to 1 per site) of algae and invertebrates, excluding Inversiula nutrix and Spirorbis nordenskjoldi. Figure 4: Fig. 4.csv Time: months after deployment Length: length of thalli (mm) Figure 5: Fig. 5c and d.csv Axis 1 and Axis 1: Values from first two axes of principal coordinate analysis IceCover: proportion of days that each site is free of sea-ice per year. Beta: Beta-diversity. Calculated as Jaccard similarity between the most ice-covered site (OB1) and each other site. Figure 5: Fig. 5e and f.csv IceCover: proportion of days that each site is free of sea-ice per year. Value: number of species per boulder (for Metric=Diversity), or percent cover per boulder (for Metric=Cover). Figure 6: Fig. 6a.csv Sites.lost: number of sites removed from dataset due to sea-ice loss. Ice: maximum ice-free days within the region (d yr-1). S: Total species richness across each subset of sites. Effort: relative sampling effort (number of sites sampled).

Human impacts threaten not only species, but also entire ecosystems. Ecosystems under stress can collapse or transition into different states, potentially reducing biodiversity at a variety of scales. Here we examine the vulnerability of shallow invertebrate-dominated ecosystems on polar seabeds, which may be threatened for several reasons. These unique communities consist of dark-adapted animals that rely on almost year-round sea-ice cover to create low-light shallow marine environments. Climate change is likely to cause early sea-ice break-out in some parts of Antarctica, which will dramatically increase the amount of light reaching the seabed. This will potentially result in ecological regime shifts, where invertebrate-dominated communities are replaced by macroalgal beds. Habitat for these endemic invertebrate ecosystems is globally rare, and the fragmented nature of their distribution along Antarctic coast increases their sensitivity to change. At the same time, human activities in Antarctica are concentrated in areas where these habitats occur, compounding potential impacts. While there are clear mechanisms for these threats, lack of knowledge about the current spatial distribution of these ecosystems makes it difficult to predict the extent of ecosystem loss, and the potential for recovery. In this paper we describe shallow ice-covered ecosystems, their association with the environment, and the reasons for their vulnerability. We estimate their spatial distribution around Antarctica using sea-ice and bathymetric data, and apply the IUCN Red List of Ecosystems criteria to formally assess their vulnerability. We conclude that shallow ice-covered ecosystems should be considered near threatened to vulnerable in places, although the magnitude of risk is spatially variable. This dataset comprises two files. Both are provided in netCDF format in polar stereographic project (see nc file for projection details). light_budget_6km.nc : this gives the estimated annual light budget (in mol photons/m^2/year) at the surface of the water column, having been adjusted for sea ice cover (see paper for details). This is calculated on the 6.25km grid associated with the sea ice concentration data. benthic_light_500m.nc : this gives the estimated annual light budget (in mol photons/m^2/year) at the sea floor, having been further adjusted for water depth. It is provided on a 500m grid (as per the IBCSO bathymetry used). Areas deeper than 200m are given no-data values, and areas outside of the coverage of the sea ice grid are assigned a value of -999. See paper for details.

---- Public Summary from Project ---- Understanding the strength of possible biological feedbacks is crucial to the science of climate change. This project aims to improve our understanding of one such feedback, the biogenic production of dimethylsulphide (DMS) and its impact on atmospheric aerosols. The Antarctic ocean is potentially a major source of DMS-derived aerosols. The project will investigate the coupling between satellite-derived aerosol optical depth, phytoplankton biomass and DMS production in the Antarctic Southern Ocean. From the abstract of the attached paper: We analysed the correlation between zonal mean satellite data on surface chlorophyll (CHL) and aerosol optical depth (AOD), in the Southern Ocean (in 5-degree bands between 50-70 degrees south) for the period 1997-2004), and in sectors of the Eastern Antarctic, Ross and Weddell Seas. Seasonality is moderate to strong in both CHL and AOD signatures throughout the study region. Coherence in the CHL and AOD time series is strong between 50-60 degrees south, however this synchrony is absent south of 60 degrees south. Marked interannual variability in CHL occurs south of 60 degrees south. We find a clear latitudinal difference in the cross-correlation between CHL and AOD, with the AOD peak preceding the CHL bloom by up to six weeks in the sea ice zone (SIZ). This is consistent with the ventilation of dimethysulphide (DMS) from sea-ice during melting, and supports field data that records high levels of sulfur species in sea-ice and surface seawater during ice-melt. The fields in this dataset are: Timeseries Worksheet: Date Mean Chlorophyll (mg CHL/cubic metre) Mean Aerosol Optical Depth (no units) 5 Day mean chlorophyll averages 5 day mean aerosol optical depth averages Correlation Worksheet: n - number lag r - correlation coefficient t - student t statistic Global Worksheet Column A = SeaWiFS filename Counter+1 is a counter to indicate the image number in series Date Mean Chlorophyll (mg CHL/cubic metre) Mean Aerosol Optical Depth (no units) Chlorophyll Standard Deviation Mean Aerosol Optical Depth Standard Deviation Chlorophyll Standard Error Mean Aerosol Optical Depth Standard Error Chlorophyll Count (the number of data 'pixels' in the image - the basic pixel size is 9x9km2) Mean Aerosol Optical Depth (the number of data 'pixels' in the image - the basic pixel size is 9x9km2)

More than 50 scientists from eight countries conducted the Sea Ice Physics and Ecosystem eXperiment 2012 (SIPEX-2). The 2012 voyage built on information and observations collected in 2007, by re-visiting the study area at about 100-120 degrees East. This was the culmination of years of preparation for the Australian Antarctic Division and, more specifically, the ACE CRC sea-ice group who lead this international, multi-disciplinary, sea ice voyage to East Antarctica. Work began at the sea-ice edge and penetrated the pack ice towards the coastal land-fast ice. The purpose of SIPEX-2 was to investigate relationships between the physical sea-ice environment, marine biogeochemistry and the structure of Southern Ocean ecosystems. While the scientists and crew did not set foot on Antarctic terra firma, a number of multi-day research stations were set up on suitable sea ice floes, and a range of novel and state-of-the-art instruments were used. These included: A Remotely Operated Vehicle (ROV) to observe and film (with an on-board video camera) krill, and to quantify the distribution and amount of sea ice algae associated with ice floes. An Autonomous Underwater Vehicle (AUV) to study the three-dimensional under-ice topography of ice floes. Helicopter-borne instruments to measure snow and ice thickness, floe size and sea ice type. Instruments included a scanning laser altimeter, infrared radiometer, microwave radiometer, camera and GPS. Sea ice accelerometer buoys to measure sea ice wave interaction and its effect on floe-size distribution. Customised pumping systems and light-traps to catch krill from below the ice and on the sea floor. Available at the provided URL in this record, is a link to a file containing the locations of all ice stations from this voyage.