Both sea ice and snow cover extent have been declining through the last three decades. Sea ice is not only shrinking but also changing in its characteristics since is getting younger and thinner. One of the causes is the global warming that is enhanced in the Arctic yielding the so-called Arctic amplification. Another cause is the strengthening of the Transpolar Drift Stream that drives out of the Arctic Ocean the multiyear ice that is replaced by seasonal ice. Since there is not the usual air-conditioning effect provided by sea ice, there is an enhanced rise of air temperature that causes the decrease of snow cover and changes the pattern of snow precipitation. The overall consequence is the loss of white surface that could reflect the sunlight so there is much more radiative power being absorbed by the Earth. This is the albedo decrease.
In this study we analyze the evolution of sea ice and snow cover in the Northern Hemisphere during the summer season (April to October) in the period 1982-2015 and correlate this information with estimated albedo trend in the same period.
Ice and snow information (1982-2015) included in this study are extracted from the NASA Northern Hemisphere EASE-Grid 2.0 Weekly Snow Cover and Sea Ice Extent dataset distributed by NSIDC (National Snow & Ice Data Center). This dataset is generated from brightness temperature data derived from SMMR, SSM/IS and SSMIS.
As for albedo, we used the CLARA-A2-SAL (Cloud, Albedo, Radiation - AVHRR based, Edition 2 – Surface Albedo) data record provided by EUMETSAT/CM SAF covering the 1982-2015 period. The albedo products are composed from overpass data from the Advanced Very High Resolution Radiometer (AVHRR) instruments on board the NOAA and Metop satellites.
The achieved results give a reasonable idea of how much the Earth albedo is influenced by the overall decrease of sea ice and snow cover. The analysis of the anomaly of each decade in comparison to the 1980s points out the acceleration of this process and the relative contribution of sea ice extent and snow cover.
A regional analysis is then presented in order to evaluate the impact of both geophysical parameters in different areas of Northern Hemisphere, i.e. in six sectors of 60° of longitude each. Both singular trends and correlation values present interesting features with different relative importance in different sectors.

Area D - Board 473

Pancake and frazil ice represent an important component of the Arctic and Antarctic cryosphere, especially in the Marginal Ice Zones. The retrieval of its thickness by remote sensing is, in general, a very difficult task. A processing system has been developed and refined by these authors in the framework of the EU SPICES project; it is meant for routinely deriving ice thickness in frazil-pancake regions using the spectral changes in wave spectra from imagery provided by space-borne SAR systems. This methodology has been successfully tested in the Beaufort Sea through comparison with ground truth collected during fall 2015 (Wadhams et al., 2018).
In the present study, this technique has been adapted and applied to Antarctic frazil/pancake icefields. Several case studies have been identified using different SAR products (CosmoSkyMed, Sentinel-1, TerraSAR-X). Our retrievals have been analyzed and validated through the comparison with co-located in situ observations collected by international research cruises realized in different sectors of the Antarctic region (i.e., Ross Sea, Weddell Sea, etc.). Particular attention has been addressed to in situ information collected in the framework of the 2017 PIPERS (Polynyas, Ice Production, and Seasonal Evolution in the Ross Sea) cruise in the Terra Nova Bay and its polynya, . which aimed at monitoring the space/time evolution of air-ice-ocean interactions initiated during autumn and tracked into winter-spring in the Ross Sea.
The analysis of SAR images followed the procedure fully described in Wadhams et al., 2004. First, a strip of imagettes, with dimensions 512x512 pixels, is extracted across the ice/sea edge in the vicinity of the ship positions. The SAR spectrum for each imagette is then computed providing the input to the following SAR/wave spectrum inversion step. Finally, inversion procedures are applied to both the open water and the sea ice SAR spectra in order to infer parametrically physical properties such as thickness.
A broad agreement was found between measured thicknesses and those retrieved from the SAR imagery. Results and statistics are presented and discussed in details.

Area D - Board 474

As a direct response to the Polar Expert Group recommendations for sea ice concentration and high latitude SST’s, EU and ESA have initiated phase A studies for a low frequency high spatial and temporal resolution polar orbiting satellite microwave radiometer in preparation for the Expansion phase of the Copernicus Space Component from 2025: the Copernicus Imaging Microwave Radiometer (CIMR).

Compared to the existing and planned Passive Microwave Radiometers (SSMIS, MWRI, AMSR2,...) CIMR will have significantly higher spatial resolution and for one of the two primary geophysical variables: sea ice concentration (SIC) it will have high resolution and very low noise at the same time. These two things are trade-offs with current radiometers. The second primary parameter sea surface temperature (SST) will significantly increase coverage at high latitudes because of the microwave cloud penetrating capabilities.

Using simulated brightness temperature (Tb) datasets over snow covered sea ice (SIC=100%) the sensitivity of SIC algorithms each using CIMR channels have been investigated. The simulated Tb dataset has been generated using a combination of a column thermodynamical model and an emission model. We then know the truth, which is 100% SIC, and noise is defined as variability near this reference point. In general, the sources of geophysical noise are: 1) surface emissivity and effective temperature variations, and 2) atmospheric influence on the measured radiances and how these propagate in different SIC algorithms. The goal is to identify SIC algorithms where the combination of CIMR channels is minimizing the geophysical noise or alternatively find algorithms with low sensitivity to noise sources that we cannot correct for.

Spatial resolution is a strong requirement in SIC applications and spatial resolution varies with electromagnetic frequency. Therefore we compare the geophysical SIC noise for each algorithm with the representativeness uncertainty using a microwave radiometer imaging simulator.

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. 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 climate data record (CDR). When constructing a sea ice CDR 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.

This study is part of the European Space Agency (ESA) CIMR mission advisory group phase A studies. These studies are assessing and developing the CIMR mission concept for its applications, here SIC.

Area D - Board 475

Research on improving the prediction skill of climate models requires refining the quality of observational data used for initializing and tuning the models. This is especially true in the Polar Regions where uncertainties about the interactions between sea ice, ocean and atmosphere are driving ongoing monitoring efforts.

A special characteristic of the Arctic sea ice covered regions at microwave frequencies is the high emissivity of sea ice, with and without snow, which contribute more to the microwave signal observed by satellites than the often dry atmosphere. Since the satellite observed signal contains contributions from both, surface and atmosphere, retrievals of atmospheric parameters require some information about the surface emissivity and vice versa. In the case of single parameter retrieval methods the contribution of the retrieved parameter is the signal of interest. The contributions from all the other geophysical parameters are seen as noise that needs to be mitigated by using information from different channels of the radiometer or through a priori data.
Here we present a retrieval method which takes advantage of the multispectral capabilities of imaging radiometers in order to retrieve seven geophysical parameters: wind speed, total water vapor, liquid water path, sea surface temperature, ice surface temperature, sea ice concentration (SIC) and multi-year ice fraction. A constrained optimal estimation technique is used to invert the forward model and extract the ensemble of seven parameters that optimally match the observed brightness temperatures. A priori information from climatological and meteorological sources is used to constrain the method to the natural variability of each parameter. The method can use observations from the current generation of space borne radiometers such as AMSR2.

In the near to medium future the Copernicus Imaging Microwave Radiometer (CIMR) candidate mission promises to offer high resolution, low uncertainty observation capabilities for the C, X, Ku and Ka bands. To assess the potential impact of CIMR, a comparison is made between retrievals based on AMSR2 observations and a retrieval using CIMR equivalent observations over a data set of validated SIC values. Compared to existing radiometers, CIMR could provide higher spatial resolution observations at the lower frequency channels (C, X, Ku) which are less sensitive to atmospheric influence.

Individual channels or channel combinations can be used as input for the optimal estimation retrieval which allows for flexibility in selecting frequency specific parameter sensitivities. An information content analysis expands the comparison between AMSR2 and CIMR to all retrievable surface and atmospheric parameters. This analysis quantifies individual parameter contributions to the observed signal and highlights the differences between different input channel combinations. The higher resolution of the low frequency CIMR channels would allow for unprecedented detail to be achieved in Arctic passive microwave retrievals.

Area D - Board 476

OHB Italia, an aerospace company based in the city of Milano, is the Prime Contractor responsible for the design, development, manufacturing and testing of the MicroWave Imager (MWI) Instrument, that will fly on-board the MetOp Second Generation (SG) weather satellites.
MWI will provide Europe’s National Meteorological Services and, by extension, the international users and Science Community with unprecedented and high-value data for meteorological and climate monitoring. The MicroWave Imager will measure precipitation as well as sea ice extent parameters. OHB Italia is responsible for the development from Phase B2 to the final in-orbit verification of three flight models.
MWI is a conical scanning radiometer, with multiple frequency channels covering the frequency range from 18.7 GHz to 183.3 GHz.
The instrument rotates at constant speed, and continuously collects the microwave radiation from the Earth surface and atmosphere at sequential IFOV (instantaneous field-of-view) positions at each rotation, covering an angular sector of ±65 degrees across the direction of flight. At each rotation of the instrument, a line of the microwave-image is formed. As the satellite moves forward along the orbit, providing the on-ground track of the scanning, the image of the atmosphere is formed by the succeeding IFOV lines.
The spatial resolution (dimension of the IFOV) ranges from 50 km at 18.7 GHz to 5 km for the highest frequency channels.
The project is approaching the Critical Design Review, planned at beginning 2019. All the MWI Instrument models (Structural-Thermal Model, Engineering Model, three Flight Models) will be fully assembled and then tested in the new facility of OHB Italia headquarters in Milano starting from next year .
Besides MWI, OHB Italia has recently started the Phase A/B1 study of the Copernicus Imaging Microwave Radiometer payload, part of the Copernicus Expansion missions; among others, CIMR will allow observations of the sea ice concentration and sea surface temperature, with complete coverage of the polar regions (pole included), which are key indicators of global climate changes. CIMR has been already defined as “Game Changer” for its unprecedented spatial resolution and accuracy.
The presentation will focus on most recent advances of Metop-SG MWI and of Copernicus CIMR projects at OHB Italia.

Area D - Board 477

Sea ice concentration (SIC) is an essential sea ice parameter e.g. for ice navigation, offshore activity and model forecasting. After the launch of the Sentinel-1 satellites, the availability of C-band, dual polarization Synthetic Aperture Radar (SAR) data in the Arctic regions has increased significantly. This have provided a unique possibility for operational monitoring of sea ice at a high spatial resolution. Automatic classification of SIC from SAR data is however a challenging task, since values of the backscatter signal, denoted as sigma0, have an overlapping range for both open water and sea ice, which are lead to miss classifications. Gaps in the daily Sentinel-1 data coverage are also a problem for an operational service.

To meet these challenges we have developed a multisensor product that combines SIC from AMSR2 Passive Microwave Radiometer (PMR) and Sentinel-1 SAR data to benefit from the complementing capabilities of the two type of sensors.

The Sentinel-1 SAR data we are using are Extra Wide Swath (EW) mode dual-polarized data with HH/HV polarization and a medium spatial pixel resolution (GRDM) at 40m. The data are calibrated to sigma0, and noise correction are applied based on the information following the original product. To derive SIC from the SAR data we first use a bayesian classification approach to separate the data into ice and water pixels on a full 40m resolution. In the classification, sigma0 values from both the HH and HV polarizations are used. In addition, we derive texture features for entropy and variance from the HV polarization. After the separation of ice/water, SIC is calculated by estimating the area covered by ice pixels within 1x1 km grid cells.

The AMSR2 SIC product are produced from L1b brightness temperature (Tb) data from the 89 GHz channels, using a processing chain similar to that successfully adopted for the EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI SAF) service. The algorithm includes dynamic tuning of the tie-points and weather filter, and atmospheric correction of the Tb using a Radiative Transfer Model (RTM). Swaths of AMSR2 SIC are produced in near-real-time at a spatial resolution of ~5km.
The SIC algorithms for both AMSR2 and Sentinel-1 additionally provide the probability of correct classification for each pixel. These probabilities are used when merging the two products into a multisensor SIC product.
Sigma0 values from Sentinel-1 are influenced by capillary waves produced by wind over open water giving sigma0 values similar to those of sea ice. To prevent this effect, the probability of the Sentinel-1 classification is reduced in areas where winds are more than 10m/s. The wind are extracted from the AROME-Arctic model running operationally at MET-Norway. Also areas where we have level or new frozen sea ice sigma0 values can be misinterpreted as open water. Here, the probability of the Sentinel-1 product is reduced based on the difference between AMSR2 and Sentinel-1 classification.

The multisensor SIC product are produced daily in a pre-operational test setup at MET-Norway covering the European part of the Arctic. In the merging process we use the previous analysis as a background. The latest available swath data of SIC from Sentinel-1 and AMSR2, covering the area of interest, are merge with the background data to update the analysis. The product is compared to the manually derived ice charts produced at the Norwegian ice service.

Area D - Board 478

Arctic sea ice is going through a dramatic change in its extent and volume at an unprecedented rate. A retreating Arctic sea ice cover has a marked impact on regional and global climate, through many feedback mechanisms and interactions with the climate system. It has been identified a clear reduction on the Arctic sea ice thickness within the last decades. Therefore, Sea Ice Thickness (SIT) monitoring is essential to better understand the changes the Arctic is experiencing. SIT can be measured with remote sensing platforms, mainly with the freeboard measurements from radar/laser altimeter instruments (only for ice thicker than 1 m (Laxon et al. 2013) and with L-band passive radiometers (only valid for the thin ice (Kaleschke et al. 2012)), making both instruments complementary.
We present a new empirical algorithm that can retrieve SIT from Soil Moisture Ocean Salinity (SMOS) brightness temperatures (TB). Other SMOS SIT products which are currently being distributed use two different methodologies: one computes SIT based on theoretical sea ice models (Tian-Kunze et al. 2014), and the other uses a regression function with SIT data obtained from models (Huntemann et al. 2014). Therefore, this is the first time that the model function needed to derive the SIT has been obtained exclusively using in-situ SIT data from various field campaigns (i.e. Electromagnetic bird and Operation IceBridge) in the Arctic during 2011-2015. The available in-situ dataset has been divided into two groups, one is used to compute the model function and the other has been used to validate the SIT products. The computed SIT has a root-mean-square error of 0.24 m, which seems reasonably good for analysis at 25 km grid. Moreover, this algorithm minimizes the error produced by the difficulty on discriminating between TB signature of thin SIT versus the signature due to low sea ice concentration. The retrieved SIT is expected to serve as an operational product for data assimilation in ice prediction/climate models.



Huntemann M, Heygster G, Kaleschke L, Krumpen T, Mäkynen M and Drusch M (2014) Empirical sea ice thickness retrieval during the freeze up period from SMOS high incident angle observations. Cryosphere, 8(2), 439–451 (doi:10.5194/tc-8-439-2014)

Kaleschke L, Tian‐Kunze X, Maaß N, Mäkynen M and Drusch M (2012) Sea ice thickness retrieval from SMOS brightness temperatures during the Arctic freeze‐up period. Geophys. Res. Lett., 39(5) (doi:10.1029/2012GL050916)

Laxon SW and 14 others (2013) CryoSat‐2 estimates of Arctic sea ice thickness and volume. Geophy, Res. Lett., 40(4), 732–737 (doi:10.1002/grl.50193)

Tian-Kunze X and 6 others (2014) SMOS-derived thin sea ice thickness: algorithm baseline, product specifications and initial verification. Cryosphere, 8(3), 997–1018 (doi:10.5194/tc-8- 997-2014)

Area D - Board 479

In the last decades, the global climate system shows evidences of rapid change. This change is amplified in the Arctic, which is more sensitive to climate variations as they strongly affect sea ice, both on its spatial and temporal extension and on its thickness. Changes on sea ice create more melting and, consequently, change freshwater fluxes in the Arctic. These changes affect to key climate variables as well as to marine productivity.

Sea Surface Salinity (SSS) is a key indicator of the freshwater fluxes and an actively sought variable, but the Arctic is a region particularly sparse in salinity in-situ measurements. For this reason, remote sensing salinity measures (currently provided by L-band satellites SMOS and SMAP) are of special relevance for this region. The computation of SSS in the Arctic represents a challenge, because brightness temperature measured by L-band satellites are less sensitive to salinity in cold waters. An additional drawback consists in the presence of sea ice in the Arctic ocean, that contaminate the brightness temperature and must be adequately processed.

ESA launched in 2016 the Arctic+ initiative to support the development of novel Earth Observation based products, to reinforce the scientific component of the ESA Living Planet programme and to prepare a scientific basis for supporting activities addressing the priorities of the Arctic scientific community, including Arctic+ Snow, Arctic+ Ice, Arctic+ ArcFlux, Arctic+ YOPP and Arctic+ Salinity.

The kick-off meeting of Arctic+ Salinity took place in Paris past November 7. The first task of the project is the identification of the datasets and models and the identification of test areas. Besides, new salinity retrieval algorithms and methods will be developed, in order to impove the quality of the products. Finally, SSS for the full SMOS period will be generated.

One of the objectives of the project is to provide SSS products from SMOS (Level 3) and from the merging of SMOS and SMAP (Level 4). To ensure and assess the quality of the products, an intensive validation activity is intended. Finally, SSS data will be assimilated by TOPAZ4 model stablishing the scientific analysis and impact assessment, with the goal of impacting Copernicus Marine Service.

In this presentation, we will present the up-to-date SSS products of the Arctic and discuss its impact for oceanographic and climate studies.

Area D - Board 481

Accurate all-weather sea surface temperature (SST) observations are crucial in order to monitor, understand and predict the state of the ocean, atmosphere and sea ice. Traditionally, SST has been observed using thermal infrared satellite instruments, but these are limited by their inability to provide SST during cloudy conditions and biasing from aerosols. Passive microwave observations are an important supplement as these allow observations through non-precipitating clouds and the impact of aerosols is small. The Copernicus Imaging Microwave Radiometer (CIMR) being investigated by ESA for the Copernicus Expansion program of the European Union is a polar mission, designed to observe all-weather, high-resolution, high-accuracy, sub-daily observations of sea surface temperature and sea ice. It is the intention to include several of the radiometer channels which are also on the Japanese Advanced Microwave Scanning Radiometer (AMSR2) and it is therefore important to assess the impact of using different channel selections.
This presentation will assess and discuss the optimal channel selection for SST retrievals using two different retrieval algorithm types developed within European Space Agency Climate Change Initiative (ESA-CCI) for SST. The first is an Optimal Estimation (OE) algorithm (Nielsen-Englyst et al, 2018), which inverts a forward model build on Wentz et al. (2000) to retrieve SST from AQUA’s Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E). The second retrieval algorithm is made up of a two-stage regression model where stage one retrieves wind speed using AMSR brightness temperatures and numerical weather prediction (NWP) data, and stage two applies localized algorithms to retrieve SST using AMSR brightness temperatures, the retrieved wind speeds from stage one and information from NWP.
The different combinations of AMSR channels have been tested to assess the impact of using fewer channels in different combinations compared to using all channels. The SST retrieved by the two algorithm types using different channel combinations will be compared to independent in-situ SSTs from drifting buoys and Argo floats to assess the most optimal channel selection for each of the two algorithms. The results from two algorithms will be compared and discussed to provide recommendations for the most optimal set of channels for CIMR regarding accurate SST estimation.

Area D - Board 482

We have developed an algorithm for the thin ice detection under winter conditions using the AMSR2 radiometer 36 and 89 GHz brightness temperature (TB) L1R swath data, and a method to combine the thin ice swath charts to a more reliable daily thin ice chart (L3 product), see Figure 1. The thin ice detection algorithm was developed using MODIS thin ice thickness swath charts as reference data [1]. The algorithm is based on the classification of 36 GHz polarization ratio (PR36) and H-polarization 89-36 GHz gradient ratio (GRH) signatures with linear discrimination analysis (LDA). We applied atmospheric correction to the AMSR2 L1R TB data as in [2], [3]. The PR36 and GRH signatures were adjusted to a constant air temperature (Ta) before the thin ice detection using an empirical relationship between the signatures and Ta determined in this study. The maximum thickness of detected thin ice was estimated to be 20 cm. The thin ice detection in the L1R data is conducted only when SIC≥70% and Ta≤-5 °C to limit conditions where thick ice may be erroneously detect as thin ice. The thin ice detection algorithm was developed using data for the Barents and Kara Sea, but we estimate that it can applied for other Arctic marginal ice zones where first-year ice dominates.
The daily thin ice chart was validated using an independent set of MODIS daily thin ice charts [4]. Its accuracy was quantified with following two probabilities: misclassifications of thick ice (>20 cm) as thin ice (<20 cm) (type I error) and thin ice as thick ice (type II error). For our future applications of the AMSR2 thin ice chart, e.g. combining it with SAR and a sea ice model data for an ice thickness chart for ship navigation, we would like have the type I error to be small. Therefore, the LDA classifier was modified to this direction, resulting average I and II errors of 10% and 32%, respectively. It is not possible to detect with good accuracy both thin and thick ice due to large overlapping of their PR36 and GRH signatures.
The daily thin ice chart was found have on average rather good day to day consistency in the thin vs. thick ice classification. There are sometimes unrealistic cases where a large area of thick ice is detected as thin ice on the following day. These were found to be likely due to warm weather conditions (Ta close to 0 °C) which make thick ice signatures resemble those of thin ice. On average we did not found correlation between the type I and II errors and daily mean Ta. We take this as evidence that the approximate Ta scaling of PR36 and GRH works properly.
We found out in here and also in our previous thin ice detection study with the AMSR-E data in [5] that estimation of thin ice thickness with the 36 and 89 GHz TB is not feasible, unlike it has been in many polynya monitoring studies, e.g. [6]. This is due to a large scatter between the MODIS ice thickness (hiM) and thin ice PR36 and GRH signatures which includes besides polynyas also thin ice at the ice edge and large thin ice areas from the freeze-up period. We also assumed thin cover on sea ice in the MODIS hiM retrieval which suppresses the hiM range of thin ice, but we think this is a more realistic approach in our case than assuming thin ice to be snow-free as in polynya studies.
There are two major error sources in the thin ice detection here. The first one is TB signature mixing from various surface types (open water, thin ice, FYI) due to coarse resolution of the radiometer data. A FYI ice field with low SIC may have PR36 and GRH signatures similar to thin ice with SIC of 100%. Finding out thin ice with true SIC close 100% is difficult (or even impossible) as various SIC algorithms typically underestimate SIC for thin ice. In the daily thin ice charts the daily SIC was from 70% to 90% for roughly 60% of the thin ice pixels. Thus, we have an evidence that the detected thin ice is not just FYI with low SIC. The second error source is possible snow cover on thin ice and ice surface effects, like surface roughness, dry ice surface, and frost flowers, which can yield thin ice PR36 and GRH signatures similar as those for thick ice. Nevertheless, this study and many polynya monitoring studies have demonstrated that useful data on thin ice properties (occurrence and thickness) can be retrieved from the 36 and 89 GHz radiometer data.
The analysis of a time series of the detected thin ice area in our study area in Oct-May in 2014-2017 yielded two notable results. The fraction of thin ice area of the total ice area may vary strongly during the winter as well as annually. In this dataset the annual mean fraction of thin ice area varied from 10% to 28%. The results added also further support to the assertion that we actually detect thin ice instead FYI with low SIC. The fraction of thin ice area with SIC 70-85% formed on average a third of the total thin ice fraction, here from 4% to 10%.

Figure 1. AMSR2 daily thin ice chart on 02 Dec 2015. The maximum thickness of thin ice is 20 cm.

[1] M. Mäkynen, B. Cheng, and M. Similä, “On the accuracy of the thin ice thickness retrieval using MODIS thermal imagery over the Arctic first year ice,” Ann. Glaciol., vol. 54, no. 62, pp. 87-96, 2013.
[2] S. Andersen, R. Tonboe, S. Kern, and H. Schyberg, “Improved retrieval of sea ice total concentration from spaceborne passive microwave observations using numerical weather prediction model fields: an intercomparison of nine algorithms, Remote Sens. Environ., vol. 104, pp. 374-392, 2006.
[3] R. Tonboe et al., “The EUMETSAT sea ice concentration climate data record,” The Cryosphere, vol. 10, pp. 2275-2290, 2016.
[4] M. Mäkynen, and J. Karvonen, “MODIS sea ice thickness and open water – sea ice charts over the Barents and Kara Seas for development and validation of sea ice products from microwave sensor data,” Remote Sens., vol. 9, no. 12: 1324, 2017.
[5] M. Mäkynen, and M. Similä, “Thin ice detection in the Barents and Kara Seas with AMSR-E and SSMIS radiometer data,” IEEE Trans. Geosci. Remote Sens., vol. 53, no. 9, pp. 5036-5053, 2015.
[6] K. Iwamoto, K. I. Ohshima, and T. Tamura, “Improved mapping of sea ice production in the Arctic Ocean using AMSR-E thin ice thickness algorithm,” J. Geophys. Res. Oceans, vol. 119, pp. 3574-3594, 2014.

Area D - Board 483

Meltwater ponds and channels have been spotted on multiple Antarctic ice shelves by optical remote sensing and in-situ studies in recent years. Since these meltwater features potentially make these ice shelves vulnerable to hydrofracturing, it is key to monitor the meltwater dynamics of these ponds and channels in order to improve the understanding of ice-shelf stability. Yet, a remote sensing approach to assess these meltwater dynamics over Antarctic ice shelves is currently lacking as all previous studies were focused on optical remote sensing which cannot reveal the highly variable meltwater dynamics over frequently cloud covered areas.

In this study we illustrate the potential of Sentinel-1 to monitor the spatial and temporal dynamics of meltwater features on Antarctic ice shelves. By combining SAR backscatter to discriminate between ice, water and snow with repeat-pass SAR interferometry (InSAR) to detect elevation changes related to lake collapse and refreezing, we show the potential of Sentinel-1 to assess the occurrence and temporal evolution of meltwater features.

To assess the potential of Sentinel-1 for assessing meltwater dynamics, we combined i) spatial and ii) temporal variations of SAR backscatter with iii) InSAR coherence and iv) interferogram patterns over both the Roi Baudouin ice shelf (RBIS) and the Amery ice shelf (AIS) in the period 2017-2018, where multiple meltwater features were detected by previous studies.

The results of our analyses highlight both the potential and limitations of SAR and InSAR for detecting meltwater dynamics. The spatial analysis of SAR backscatter shows the clear appearance of meltwater features contrasting to their surroundings without meltwater. Automatic detection of these features is however not straightforward as the results of the time series analysis show that the temporal backscatter variation of meltwater features depends on the location of the features (e.g. on blue ice or in firn covered area), the size of the features and whether they refreeze in winter. Therefore, it is not trivial to summarise a pattern of the meltwater features purely from backscatter images. The results of InSAR coherence approach on the other hand show that coherence images are more adequate to detect changes in meltwater features as they detect the change of scattering within the time series. A comparison is shown in attached figures e and f, with d indicating the approximated locations of subfurface meltwater occurring on the RBIS during the melt season. This improved detection is more apparent over the AIS than over the RBIS. This regional difference can be explained by a better data availability and an overall higher coherence over the AIS, which indicates that the AIS ice shelf surface was more consistent in the InSAR time series. Finally, the InSAR interferograms show that both refreezing processes and lake collapse events can be detected by analysing InSAR fringe patterns. As shown in attached figure a, the collapse of a subsurface lake results in a fringe pattern in contrast to the surrounding. This can be validated via figures b and c where the lake surface appears rougher after the collapse. The quality of this detection however depends strongly on the quality of the phase unwrapping preprocessing step, which can be complicated over ice shelves.

Consequently, our results highlight the potential of Sentinel-1 SAR and InSAR approaches to assess both gradual meltwater processes such as refreezing and instant prcoesses such as drainage and collapse over Antarctic ice shelves. As such it highlights the importance of future dedicated meltwater products from Sentinel-1 to facilitate the study over Antarctic ice shelves in a changing climate.

Area D - Board 484

Typical for observations at low microwave frequencies, the L-band sensors deliver an uninterrupted time series of observations of the Earth’s surface irrespective of cloud and lighting conditions, which is fundamental in particular to explore the polar regions during the polar nights. Because of its longer wavelength and low value of the imaginary part of permittivity of ice (corresponding to low losses) at this frequency, the L-band allows a greater penetration in ice, snow and frozen soil compared to observations at higher frequencies (e.g. the SSM/I, AMSR-E sensors as well as MWI on board future MetOp-SG satellites). This unique capability allows to be sensitive to deeper layers of different media and then to explore on some parameters which are, up-to now, poorly known. Moreover, it was demonstrated that L-band is also sensitive to the variation of refraction coefficient of a stratified medium, which in consequence offers the possibility to infer information on it based on microwave brightness temperature measurements. These capabilities were recently tested by using ESA’s SMOS observations were collected in both Antarctica and Greenland. In particular, efforts were first devoted to a better understanding of L-band signature by using microwave emission model in combination to glaciological models. Starting from that potential products were derived, in particular the possibility to infers information of the internal ice sheet temperature from the surface to the bottom was proven. It’s worth mentioning that the internal temperature is a key parameter for the ice sheet dynamics and knowing its value is fundamental to better understand the changes in ice sheet mass balance and dynamics. At present only L-band, having a penetration in the medium of about 1000 m, is capable to provide this info, and existing and future satellites with higher frequencies, are unable to derive it. L-band observations can be also used to infer information on the surface properties of the ice sheet and in particular to its melt/not melt status. By using an approach similar on what applied for higher frequencies a retrieval approach has been recently developed for both Greenland and Antarctica. Benefits of L-band observations come from its large penetration depth, which could detect whether melt is deep and impacts on snowpack stability, contrary to higher microwave frequencies limited to the first meters or centimeters of the snowpack. Finally it can improve the survey of extent and duration of large scale melting events. Although surface melting has a minor overall contribution to the mass balance, it is of interest for climatology: surface melting extent and duration are climatic indicators related to surface temperature and thus to the Earth’s radiative budget. A summary of these results, which were mainly obtained within different ESA projects, together future perspectives will be presented at the conference.

Area D - Board 485

Leads are elongated open water area in the ice, formed by the ice divergence and shear motion. Although leads make up only a small fraction in the ice covered area, they play a key role in the heat exchange between ocean and atmosphere. Leads or refrozen leads have been detected with several methods, for example, by visual and thermal imagers, passive microwave sensors, CryoSat-2 and IceSat altimeters. Both Visual and thermal imagery-based and altimeter-based lead detection has high resolution, up to 1 km and several hundred meters respectively. However, visual and thermal imagery method is limited by the cloud coverage, altimeter-based detection is limited spatially, restrained by the narrow tracks of the altimeter. Lead detection with passive microwave sensors in contrast, has the advantage to deliver daily Arctic-wide lead maps without limitation of cloud coverage, with a resolution of up to 3 km.

A previous study has shown the possibility to retrieve information of lead fraction using the ratio of vertically polarized brightness temperatures measured at 89 GHz and 18.7 GHz channels of AMSR-E, combined with a high-pass filter for the edge detection. In this study, we implement this method to AMSR2 data, which has been available since July, 2012. The combined period of AMSR-E and AMSR2 spans from 2002 to present, with a gap of one year between 2011 and 2012. It is a challenge to provide a long-time series of lead fraction data from the two instruments, due to different resolution and instrumental characteristics. To set up homogeneous data series, MODIS-derived lead maps are used as reference data to be compared either with AMSR-E-derived lead fraction in 6.25 km resolution or with AMSR2-derived lead fraction in 3.125 km resolution. Adjustment of tie points for 0 and 100% lead concentration is necessary. A first comparison in the Beaufort Sea shows a promising agreement between the three data sets, with a correlation coefficient as high as 0.9. Pixelwise uncertainty of AMSR-E and AMSR2 lead fraction data has been determined based on a statistical analysis of the difference to the MODIS imagery-derived lead maps.

Area D - Board 486

The long-term satellite passive microwave record is an important tool for cryosphere-climate diagnostics. Estimates of regional to global SWE are challenging. The GlobSnow product is one tool for assessing snow mass variations in non-complex terrain and has demonstrated reasonable performance when compared with non-observation based estimates of snow water equivalent (SWE). Given uncertainties in any product, it is prudent to develop other passive microwave estimates that use effective but different retrieval philosophies and methods to confirm the veracity of existing approaches (e.g. GlobSnow). One such development is the Satellite-based Microwave Snow Algorithm (SMSA) for estimating snow depth (SD) and snow water equivalent (SWE). The algorithm uses Advanced Microwave Scanning Radiometer – 2 (AMSR2) observations of Earth surface brightness temperature emission. AMSR2 is an instrument aboard the Global Change Observation Mission – Water mission launched by the Japan Aerospace Exploration Agency in 2012. The approach leverages observed brightness temperatures to parameterize a physically-based model without requiring parameter constraints from in situ snow depth observations or historical snow depth climatology. After screening snow from non-snow surfaces (water bodies [including freeze/thaw state], rainfall, high altitude plateau regions [e.g. Tibetan plateau]), land surface moderate and shallow snow depths are estimated by minimizing the differences between Dense Media Radiative Transfer Multi-Layer model estimates of brightness temperatures and (Tsang et al., 2000; Picard et al., 2011) and multi-frequency AMSR2 brightness temperature measurements. The unique approach adopted by SMSA parameterizes the model using snow grain size, temperature and density estimates originally developed by Kelly et al. (2003). Retrievals of a single layer snowpack snow depth is conducted from in situ snow depth data from a variety of experiment data sources. Results are presented to illustrate the performance of the SMSA approach in comparison. We use Sentinel 2 and 3 data to provide snow cover extent for comparison with coincident AMSR2 snow cover extent estimates. The SMSA approach is also compared with the baseline AMSR2 algorithm estimates and the model assimilation-based approach of GlobSnow. Given the variation in estimates of SWE by different land surface schemes/climate models and selected satellite passive microwave approaches, SMSA provides SWE estimates that are independent of real or near real-time in situ and model data.

Area D - Board 487

Arctic snow has the peculiarity of being very dense near the surface, due to frequent blowing-snow events and sustained cold temperatures, while the bottom of the snowpack is typically less dense, with thick depth hoar layers formed through temperature gradient metamorphism. This leads to a combination of high thermal conductivity contrast, where the upper layers have higher conductivity than the bottom layers which contains more air. The result of this combination can significantly alter snowpack-insulating properties, as these layers develop through the winter. Currently, no Land Surface Model is able to accurately simulate such a specific density stratigraphy. Snow models generally produce the reverse density stratigraphy, with low density at the surface and high snow density at the bottom since they are mostly based on compaction and ignore the water vapor fluxes from the ground as well as the vegetation interactions, both leading to depth hoar formation.
We show this issue can be resolved using passive microwave radiometry. From simulations based on the microwave radiative transfer model SMRT, we discuss how the arctic snow brightness temperature (TB) can depart from subarctic snow due to the different microstructures and stratigraphy of snowpacks. Based on new ground-based radiometric and snow microstructure measurements over Canadian arctic tundra sites (Trail Valley Creek, Yukon, and Cambridge Bay, Nunavut, Canada), we discuss the best modelling and configuration approach among the models proposed in SMRT (DMRT-ML, QMS, IBA). The presence of ice crusts within the snowpack, which are now often observed over Arctic areas, also generates significant variation in TB. Finally, we suggest that improved tundra arctic snow modelling could lead to better estimates of snow insulating effects on the ground and thus reduces the snow feedback effect on winter soil temperature and permafrost evolution.

Area D - Board 488

In the boreal forest zone, trees affect microwave measurements of snow and soil frost by emitting and by absorbing and scattering the radiation from below. Although several studies of forests at microwave frequencies exist, this is still an unresolved problem deteriorating the quality of cryosphere products derived from satellite observations.

The goal of this study is to determine the transmissivity of a boreal forest in winter conditions as a function of tree temperature and other parameters at L-band. The setup for these measurements is the new forest tower at Finnish Meteorological Institute’s Arctic Space Centre in Sodankylä, Finland. Among other microwave instruments, the tower features two ESA ELBARA-II radiometers, one at 21 m level in the tower, the other at ground level right next to the tower. Both instruments have elevation and azimuth rotating systems, and they measure the brightness temperature of the same trees, one from above and the other from below. The ELBARAs are reference instruments to the ESA SMOS satellite operating at L-band 1.4 GHz frequency.

L-band transmissivity of a single pine tree was measured in Sodankylä at the previous 4-m tower on a forest opening. Now the new tower setup allows for simultaneous measurements of the same trees from above and below. In addition, the tower features CO2 flux measurements built following the ICOS (Integrated Carbon Observation System) network standards. Additional measurements in the tower surroundings include sap flow, soil moisture profile, snow depth, SWE, wind profile as well as vegetation, snow, soil and air temperatures. Web cameras monitor both the instruments and the forest area they are measuring, allowing assessment of snow intercepted by trees and snow covered area below trees.

Most of the instruments were installed in autumn 2018, and the measurements of the first full winter season will be available for the study of tree brightness temperature and transmissivity at L-band.

Area D - Board 489

Perennial and seasonal soil freezing affect large land masses in the Northern Hemisphere. Seasonal freezing occurs on approximately 51% of the land surface, while permafrost areas cover approximately 24%. Information on the freeze/thaw state of the soil may be of great importance for hydrological and climatological applications. Soil freezing influences surface radiation balance and exchange rates of latent heat and carbon with the atmospheric boundary layer. Frozen soils potentially inhibit water uptake and thus photosynthesis of vegetation in the spring, while the F/T state of the soil controls also the mass and rate of water infiltration, and thus the partitioning between surface and sub-surface runoff. Soil freezing also affects photosynthesis, and microbial activity within soils. In northern latitudes the main limiting factor for photosynthesis during the autumn period is the sunlight; belated soil freezing does not affect much the late autumn growing season. However, it has a major effect on the microbial activities within soils.

Even though the main factor driving seasonal soil freezing is climate, soil freezing processes are strongly influenced by vegetation characteristics, thickness of the top-soil organic layer, snow cover properties, soil type and soil moisture conditions. Soil freezing cannot be directly linked to air temperature; natural insulating layers (e.g. vegetation, organic peat, or snow) delay the freezing of top-soil in autumn, while the amount of liquid water within the soil itself affects the energy exchange between the soil surface and the atmosphere, and thus influences the freezing process. The lack of accurate soil and land surface physiographic datasets with global coverage, combined with uncertainties in meteorological forcing data make it difficult to reliably estimate soil freezing and thawing processes using thermodynamic models. Earth Observing satellites can directly measure properties that provide an indication of soil freezing and thawing. Particularly suitable for such purpose is the European Space Agency (ESA) Soil Moisture and Ocean Salinity (SMOS) satellite launched in late 2009, since the L-band radiometer on board SMOS measures the target at a low microwave frequency (1.4 GHz), a frequency band sensitive to the contrast between frozen and thawed soil. The long wavelength also enables observation of soil properties beneath the immediate soil surface. For soil freeze/thaw states, a surface layer depth of approximately 10-20 cm can be observed.

In co-operation with Gamma Remote Sensing, Switzerland, the Finnish Meteorological Institute (FMI) has developed, methods for detecting soil freezing and thawing using microwave radiometer observations from the SMOS satellite. The work has been funded by ESA and Finnish National agencies within the frame of several studies. The developed SMOS F/T algorithm uses SMOS level 3 daily gridded brightness temperatures (TB) provided by Centre Aval de Traitement des Données SMOS (CATDS). The algorithm uses a threshold detection approach to determine the average soil state of each SMOS observation grid cell. Reference signatures for frozen and thawed states are defined for each cell from the historical database of observations. Based on comparisons tothese references, soil state is categorized into three states: thawed, partially frozen or frozen. Ancillary data on air temperature and snow cover are applied to regulate the detection. The ancillary data are used to (1) automatically determine the periods during which the frozen and thaw references can be defined, and to (2) remove obvious errors from the product output. The F/T estimates currently cover the Northern Hemisphere. Temporal coverage starts from June 2010. New information is provided daily, with 1-2-day latency depending on the availability of input data. The data are provided in the Equal-Area Scalable Earth Grid (EASE-Grid) at 25 km x 25 km resolution. The daily generation of the soil Freeze and Thaw service has now reached operational status; the processor and the processing pipeline are physically located in the FMI Sodankylä computational facilities. Data are accessible through FMI and ESA ftp/http dissemination services.

This presentation presents the development and current status of the operational F/T algorithm and gives an overview of validation efforts of SMOS F/T estimates against in situ data from Canada, Siberia and Scandinavia.

Area D - Board 490