Sea Ice 3

SENTINEL-1a (S-1a) SAR data has been used for sea ice monitoring at FMI since December 2015. FMI's operational SAR based sea ice products previously based solely on RADARSAT-2 (RS-2) ScanSAR imagery are now produced using S-1a and RS-2 imagery jointly. These products include ice thickness and ice drift charts for the Baltic Sea, an ice thickness chart for the Barents and Kara Seas based on multisensor satellite data (S-1a, AMSR2) and a sea ice model data. The Baltic Sea ice chart produced using mainly manual methods by the FMI ice analysts utilizes also S-1a data. The Baltic Sea ice thickness and ice drift products are produced as part of the Copernicus Marine Services (CMEMS) and delivered to users through the CMEMS portal (http://marine.copernicus.eu). Some of the ice parameters (e.g. ice thickness) of the Baltic Sea ice chart are also dissemimated through this portal. The Barents and Kara Seas ice thickness chart is produced in winter 2015-2016 as part of the EC-funded FP7 POLAR ICE project. It is delivered to end-users through an user-interface specified and developed in an earlier EC-project ICEMAR.

We use now for the sea ice products mainly S-1a Extra Wide Swath Mode (EW) dual-polarized (HH/HV) L1 GRDM data. Over the Baltic Sea also Interferometric Wide Swath Mode (IW) images are also acquired. These have finer spatial resolution than the EM images, but different polarizations, VV and VH. We are studying needed conversion of our EW classification algorithms to the IW images.

Both visual and automatic classification of S-1a SAR imagery is hampered by rather high and varying noise floor in the cross-polarization images. Beginning summer 2015 the S-1a metadata contains the noise floor information, allowing noise subtraction from the cross-polarization images. The noise subtraction is not a simple task due statistical nature of both measured SAR backscattering coefficient and the noise floor itself. We will study for suitable noise floor removal method for the S-1a data acquired during winter 2015-2016, and the enhancements the noise removal brings to various sea ice products.

Large incidence angle variation in the S-1a EW images also hampers their classification. Previously, we have determined an incidence angle compensation method for the HH-polarization images, and we will now determine it also for the HV images using data acquired in winter 2015-2016.

For conversion of RS-2 ScanSAR image based sea ice classification algoirithm to S-1a images we need know how well backscatter coefficients in these two SAR images match each other. Possible deviations are due differences in absolute calibration accuracy and in noise floor levels. Our first results with few image pairs acquired in Sep 2015 over the Fram Strait shows good equivalence for the backscatter coefficients.

We have recorded delays from S-1a data acquisition to sea ice products. This delay must as short as possible (<1-2 h) to maximize benefits from the sea ice products. Starting autumn 2015 S-1a data will also be downlinked by the FMI Sodankyla ground station (SoGS) and the delay is expected to be decrease considerably.

In summary, we provide recommendations for S-1a data preprocessing practices for sea ice monitoring, and assess the suitability and capability of S-1a for operational sea ice monitoring through some example sea ice products.

The geophysical noise in a selected number of  different sea ice concentration (SIC) algorithms each representing different families of algorithms, based on their selection of channels and their sensitivity to noise,  has been investigated using simulated datasets over open water (SIC=0%) and over sea ice (SIC=100%). Here the truth is 0 and 100% SIC and noise is defined as variability quantified as the standard deviation near these reference points. The sources of geophysical noise are surface emissivisty variations and atmospheric influence on the measured radiances. Sea ice emissivity variations are caused by variations of the microphysical parameters of sea ice and snow such as temperature, salinity, snow thickness, density, roughness and wetness. For water surfaces, the influences are temperature and wind speed, and for the atmosphere they are water vapour and cloud liquid water. Each of the systematic geophysical noise sources has or may have climatological trends so even if the noise source is small yet systematic it may introduce artificial trends in the sea ice record. When constructing a sea ice climate record there are several criteria for selecting algorithms such as consistency in the methodology, availability of long time-series of consistent  satellite data, quantification of uncertainties, and stability. Here we are trying to identify the algorithms with low sensitivity to geophysical noise, especially the noise that we cannot correct for. Over open water/low concentrations the Tb19v and Tb37v gradient type of algorithms have the lowest sensitivity to geophysical noise while the algorithms exclusively using near 90 GHz channels have by far the highest sensitivity. Over ice the atmosphere plays a much smaller role; in fact it acts as a smoother of the sea ice surface noise. Over ice the range of standard deviations of different  algorithms is smaller than over open water. However, the Bristol, TUD, and the 6 GHz algorithms have the lowest sensitivity to noise over sea ice while the polarisation type of algorithms such as NASA Team have the highest noise levels. This is consistent with the noise levels in the measurements.

This study is part of a SIC algorithm evaluation in the European Space Agency (ESA) Climate Change Initiative (CCI) project. The project is developing capacity to construct a sea ice climate record from satellite data including sea ice thickness from radar altimetry (ERS 1-2, ENVISAT and eventually Cryosat 2 and Sentinel - 3) and sea ice concentration, area and extent from microwave radiometer data (AMSR 1 and 2). This presentation focus entirely on the SIC estimate sensitivity to geophysical noise over ice and open water including the influence of the atmosphere and the implications for climate applications. In particular, we wish to identify the role and impact of individual noise sources over open water and on snow covered sea ice. There is no evaluation at intermediate concentrations and during summer melt.

Since the late 70s, satellites have provided an unprecedented amount of information on the state, drift, and deformation of the Arctic ice cover over a remarkably large range of spatial and temporal scales. Recently sophisticated multi-scale statistical methods have been applied to this flow of observations leading to a significant improvement in our understanding of sea ice dynamics. Based on this new knowledge a re-evaluation of widely used sea ice models, as well as the development of new sea ice models has begun.

We will present how we use a variety of daily-retrieved high-resolution (1-10 km) satellite data to initialize and evaluate a new Lagrangian sea ice model called neXtSIM. In particular, we show that neXtSIM reproduces the spatial discontinuities in the sea-ice drift associated with the formation of leads and ridges (spatial/heterogeneity properties), as well as their statistical distribution in time (temporal/intermittency properties) very well. To evaluate the sea ice drift and deformation we use data from the RADARSAT Geophysical Processor System (RGPS), the GlobICE project (http://www.globice.info) and the My Ocean Global Sea Ice Drift data set produced from Envisat ASAR. We also use lead fraction data retrieved from satellite passive microwave observations (AMSR-E and AMSR2) and ice thickness estimates from altimeters (ICESat and IceBridge), for both evaluation and initialization.

During recent decades Arctic sea ice is decreasing in both area and volume. While the majority of this retreat can be attributed to warming of the Arctic atmosphere and ocean, the contribution associated with changes in sea ice dynamics are still under discussion. If the export of sea ice is increasing this would enhance the decrease of sea ice volume in the Arctic Basin. Fram Strait is the main gate where about 90% of the sea ice export from the Arctic Basin is happening.

Sea ice drift and area can be derived on a daily basis from the satellite passive microwave (PMW) radiometers SSM/I and SSMIS. The spatial resolution of about 50 km and data gaps are limiting factors for the accuracy of the satellite PMW derived sea ice drift. For certain time periods the lower resolution PMW data is compared to sea ice drift derived from Synthetic Aperture Radar (SAR) observation to estimate the uncertainty of the PMW time series.

Since 1990 upward looking sonars (ULS) are operated as moored instruments in Fram Strait. The ULS measure the distance to the ice underside and water surface, which allows to observe the sea ice draft and consequently to estimate the sea ice thickness. Concurrent with ice thickness in the Arctic Basin the ULS in Fram Strait observe a strong decrease in mean and modal sea ice thickness of 15% and 20%, respectively, between 1990 and 2013.

The sea ice drift and area observations from satellites are combined with the monthly ULS ice thicknesses do derive the sea ice volume flux in Fram Strait. The sea ice area export from satellite PMW observations shows high variability but no significant trends since the 1990s with only a slight increase in recent years. Combined with the decreasing ice thickness from ULS this leads to a decrease of the absolute Arctic sea ice volume export between 1990 and 2013. From these observations we conclude that for the given time period changes in sea ice export only could have a minor impact on the observed decrease in sea ice volume in the Arctic Basin. Inter-annual variability in ice volume export, however, is large and for individual years the ice export indeed will have contributed to the loss of sea ice, which can have consequences also for years thereafter.

About 10% of the total Arctic sea ice volume is exported every year. As the sea ice volume is strongly decreasing this fraction of the total sea ice volume exported per year is slightly increasing within our time period, despite the fact that the absolute number in ice volume export is decreasing. This means that sea ice volume in the Arctic Basin is decreasing at a faster rate than the sea ice volume export is decreasing. Therefore, today and likely in future fluctuations in sea ice export will have a stronger influence on the Arctic sea ice volume than before.

Sea ice thickness is one of the essential climate variables within the ESA Sea Ice Climate Change Initiative (SICCI) project. Time series (2003-2012) of sea ice thicknesses for the Arctic has already been produced from ENVISAT radar altimetry in phase 1 of the SICCI project.  To select the most optimal processing scheme, a round robin data package (RRDP) with observations of sea ice freeboard, draft, snow depth and density was established based on observations from aircraft, submarines and buoys.

Phase 2 of the SICCI project extends the time series of sea ice thicknesses to include 1992-present based on ERS-1/2, ENVISAT, CryoSat-2 and SMOS data for both the Arctic and Antarctic regions. Thus, both the spatial and temporal coverage is extended and several more options for observations exist and will be included in the extended RRDP. The three main goals of the extended RRDP are to allow uncertainty estimation of sea ice freeboard retrieval, improvement of identification of snow properties suitable for freeboard retrieval, and evaluation of freeboard-to-thickness conversion. Additionally the extended RRDP will support the freeboard retrieval algorithm selection including re-tracker optimization. The observational data in the extended RRDP is co-located with sea ice freeboard/thickness/draft obtained from as many satellite data as possible, together with additional data, e.g. model results for snow depth on sea ice and sea ice classification, to obtain consistent and reliable information throughout the time-period 1992-present.

Here is given a presentation of the data included in the extended RRDP with special focus on the goals, together with first results.