Session: Cryosphere

Seasonal snow is an important component of the Earth system heavily affecting the energy balance and the water cycle at high latitudes and elevations. Vast land areas in the north and in mountainous regions are weakly monitored by in situ sensors due to the fact that most of these regions are sparsely populated. Earth observation is the only practical means of frequent and accurate monitoring of snow properties in these regions.

This presentation gives an overview of Sentinel-3’s potential for retrieval of snow properties together with examples of retrieval methods and results where we successfully demonstrate the capability of Envisat MERIS and AATSR plus other optical sensors of moderate spatial resolution.

The Sentinel-3 sensors Ocean and Land Colour Instrument (OLCI) and Sea and Land Surface Temperature Radiometer (SLSTR) together represent a powerful set of instruments for monitoring of properties of the seasonal snow cover. The revisit time is 0.5-1.0 day over most regions with seasonal snow cover (with two satellites). The excellent improvement compared to Envisat with almost full spatial overlap between OLCI and SLSTR allows for the use of both sensors for snow mapping, relying on cloud screening with SLSTR.

The thermal bands of SLSTR are very suitable for snow surface temperature monitoring. As weather stations are typically sparse in the relevant regions, snow surface skin temperature monitoring is a valuable supplement to meteorological measurements of the 2 m air temperature. For full snow cover, which should be possible to screen with high precision using OLCI, the exact emissivity is known for the ground surface and accuracy temperature surface temperature measurements should be possible (< 0.5 K).

The snow reflectance spectrum at visible and near-infrared wavelengths is dominated by the optical effects of snow grain size and impurities. The bands O20, O21 and S4 are suitable to quantify the effects of the grain size, while most visual bands of both sensors are suitable for impurity quantification. Additionally, OLCI is suitable for impurity characterisation. This enables accurate estimation of the snow spectrum, which again is important for deriving the spectral albedo, an important quantity in, e.g., energy balance modelling.

The fractional snow cover is another and fundamental variable in snow hydrology that will benefit from accurate estimation of the snow spectrum and OLCI’s dense coverage of visual and near-infrared wavelengths with 21 bands. With the ability to do snow retrieval with OLCI and cloud screening with SLSTR, 300 m spatial resolution will be obtained compared to 1 km from AATSR (as MERIS could in practice not be used together with AATSR due to the small overlap). This is a significant achievement as the increased resolution would enable far more accurate snow mapping in complex terrain.

Retrieval of the snow spectrum and the temporal development of the snow grain size, together with thermal measurements, are important for estimation of the snow wetness and surface hoar formation. Wet snow and the degree of wetness are together with meteorological measurements and hydrological modelling suitable for snow runoff prediction, and in particular flood warning. Surface hoar may lead to formation of a weak layer in the snowpack, which in steep mountain areas may result in avalanche risk. Detection of surface hoar as well as formation of snow crust due to events of warm and wet weather, is information likely suitable as input to avalanche risk models in the future.

European Space Agency’s (ESA) Data User Element (DUE) project GlobSnow was established to create a global database of Snow Extent and Snow Water Equivalent. GlobSnow-1 was launched in 2008 and candidates for Climate Data Record (CDR) on SE and SWE were introduced in 2011. These prototype versions were further developed in the sequel project GlobSnow-2 (2012-2014). The GlobSnow SE product portfolio includes maps on Fractional Snow Cover (FSC, range 0-100 % or 0-1) in 0.01⁰×0.01⁰ geographical grid and they cover the Northern Hemisphere in latitudes 25⁰N - 84⁰N and longitudes 168⁰W-192⁰E. GlobSnow SE products are based on data provided by ERS-2/ATSR-2 (1995-2003) and Envisat/AATSR (2002-2012), augmented with Suomi NPP/VIIRS (2012-present) so that a continuous dataset spanning 20 years is obtained. The time-series for 1995-2012 are considered homogenous as the sensor characteristics of A/ATSR sensors are equivalent and inter-sensor calibration issues have been accounted for. The continuation of the time-series with NPP/VIIRS data changes the quality characteristics between the earlier and the current datasets. However, the extended swath width of NPP/VIIRS and similarity to the upcoming ESA Sentinel-3 SLSTR sensor makes the augmentation of the time-series with VIIRS data extremely valuable and will finally act as a gap filler between the A/ATSR and Sentinel-3 SLSTR and OLCI-based products.

The GlobSnow SE method development has been particularly focused on fractional snow retrievals in forested areas. The semi-empirical reflectance model-based method SCAmod by Metsämäki et al., [1] and [2] is applied. Accounting for the apparent forest transmissivity in FSC retrieval is the key element of SCAmod. The auxiliary data for Northern Hemisphere transmissivity in 0.01⁰×0.01⁰ resolution was derived from MODIS 500m reflectance (550nm) data and GlobCover classification. Besides the transmissivity, SCAmod parameters include generally applicable values of snow, forest canopy and snow-free ground reflectances. In the latest v2.1 SE dataset, spatially varying snow-free ground reflectance map is applied for retrieval, while the other two are static values.

The GlobSnow SE products are generated for daily, weekly and monthly time windows. Daily Fractional Snow Cover (DFSC) product provides fractional snow cover in percentage (%) per grid cell for all satellite overpasses of a given day. FSC is provided only for observations at sun zenith angle < 73⁰. The Weekly Aggregated Fractional Snow Cover (WFSC) provides the last available FSC estimate within past seven days. The Monthly Aggregated Fractional Snow Cover (MFSC) product provides FSC as an average of all available daily FSC estimates within the calendar month. The corresponding FSC maps categorized in four classes are also provided. In addition to information on fractional snow cover, a statistical uncertainty based on observed/estimated variations in SCAmod parameters is provided for each FSC estimate. The dataset is available at http://www.globsnow.info/se/archive_v2.1/ (AATSR) and http://www.globsnow.info/se/archive_v2.0 (ATSR-2).

The development of the A/ATSR based CDR is presented along with considerations for FSC retrieval using Suomi NPP VIIRS data. Accuracy considerations for FSC retrieval using SCAmod-retrieval algorithm from the GlobSnow-1 and GlobSnow-2 projects, are presented for both the A/ATRS series and the VIIRS dataset. The accuracy assessments have been conducted by applying a wide variety of reference data, including ground-based observations and high resolution satellite-derived reference snow maps.

Recent activities in preparation for the upcoming Sentinel-3 SLSTR and OLCI data have been on-going in various projects. In an EC funded Sen3App project, FMI together with its project partners Finnish Environment Institute (SYKE), ENVEO IT GmbH (Austria), GAMMA Remote Sensing (Switzerland) are constructing processing chains to acquire and process Sentinel-3 data to be applied in FSC-retrieval using SCAmod algorithm. The processing chains will utilize the Sentinel Collaborative Ground Segment infrastructure of the FMI national satellite data centre located in Sodankylä.

To support the development of snow cover retrieval methods for optical instruments, a wide array of intensive campaign activities have been carried out in the Sodankylä region since 2006. The objective has been to investigate the feasibility of space-borne instruments operating at optical and microwave regions for the monitoring of snow and soil processes in boreal forest and sub-arctic environment. The Sodankylä site is equipped with tower-based reference instruments of present and planned cryosphere-observing satellites. The systems provide time-series of reference observations on a continuous basis. They include multi-channel microwave radiometers and VIS/NIR spectroradiometer (since 2006; reference to AATSR, MODIS, VIIRS, Sentinel-2 and -3 etc.). An essential part in experimental activities has been the production of comprehensive in situ reference data sets from automatic sensor networks, including a full set of atmospheric profile observations, and from regular manual observations (e.g. snow pit observations). Sodankylä experiments have enabled the development and validation of modeling approaches to describe space-borne microwave and optical observations. In general, the parameterization of models has been investigated concerning the influence of soil, snowpack, and forest canopy characteristics. As an outcome, novel models to describe e.g. scene reflectance at the visible region have been developed.  For example, in case of GlobSnow, the combined use of high-resolution airborne lidar data and AISA-imaging spectrometer data enabled the development of model to describe the optical-range reflectance observations over snow covered landscapes.

The ESA GlobSnow project was active from November 2008 until late 2014, and was coordinated by the Finnish Meteorological Institute (FMI). Other project partners involved were ENVEO IT GmbH (Austria), GAMMA Remote Sensing (Switzerland), Norwegian Computing Center (NR, Norway), Finnish Environment Institute (SYKE, Finland), University of Bern (Switzerland), Federal Office of Meteorology and Climatology (MeteoSwiss, Switzerland) and Central Institute for Meteorology and Geodynamics (ZAMG, Austria), Northern Research Institute (Norut, Norway) and Environment Canada (EC, Canada).

REFERENCES

[1] S. Metsämäki, S. Anttila, M. Huttunen, J. Vepsäläinen (2005). ”A feasible method for fractional snow cover mapping in boreal zone based on a reflectance model”. Remote Sensing of Environment, 95, 77-95.

[2] S. Metsämäki, J. Pulliainen, M. Salminen, K. Luojus,  A. Wiesmann, R. Solberg, K. Böttcher, M. Hiltunen, E. Ripper (2015). ”Introduction to Globsnow Snow Extent products with considerations for accuracy assessment. Remote Sensing of Environment, 156 (2015) 96–108.

Model projections consistently suggest that the polar regions have the largest climate sensitivity to greenhouse gas increases, meaning they are likely to show larger signals than either global averages or responses in other regions. Polar observations increasingly confirm these findings, their urgency and their significance in the Arctic. It is, therefore, particularly important to monitor Arctic polar change.

Satellites are particularly relevant to observations of Polar latitudes as they are well-served by low-Earth orbiting satellites. Whilst clouds often cause problems for satellite observations of the surface, in situ observations are much more sparse due to the remote locations and hostile conditions. The ATSRs are accurate infra-red satellite radiometers, designed explicitly for climate standard observations, and particularly suited to surface temperature observations. ATSR radiance observations have been used to retrieve sea and land surface temperature for a series of three instruments over a period greater than twenty years. This series will soon be extended with the launch of SLSTR on Sentinel 3; the SLSTR instrument has the same key design features necessary for providing climate quality surface temperature datasets as the ATSRs.

We have combined land, ocean and sea-ice surface temperature retrievals from ATSR-2 and AATSR to produce a new combined surface temperature dataset for the Polar Regions. The method of cloud-clearing, use of auxiliary data for ice classification and the surface temperature (ST) retrievals used for each surface-type will be described. We will show time series of ST anomalies for each surface type. The time series for open ocean in the Arctic Polar Region shows a significant warming trend during the AATSR mission. Interpretation of this trend must take into consideration changes in open-water extent and this will be discussed. Time series for land, land-ice and sea-ice show high variability as expected but also interesting patterns.

We present the state-of-the-art for ATSR observations of surface temperature change in the Arctic and hence indicate we can have confidence in temperature change across all three domains. Currently there is no plan to provide sea-ice ST as a core product from Sentinel 3. We make the case for a near real time Arctic surface temperature product from SLSTR which would include surface temperatures for all three domains: land, sea and sea-ice.

The Arctic climate system is undergoing a rapid transition. Such changes, in particular reductions in sea-ice extent, are impacting coastal communities and ecosystems and are changing the conditions for resource extraction and shipping. In this context, high-quality predictions of the ice conditions are of paramount interest. Such predictions are typically performed by numerical models of the sea ice-ocean system. The skill of such predictions can be substantially improved through assimilation of observations. We describe the construction of an assimilation and prediction system of the Arctic sea-ice conditions.

Observational data streams for such a prediction system have to be available near real time. We used four data streams which fulfill this requirement, namely, the OSISAF sea-ice concentration and sea-surface temperature products, a snow thickness product provided by the University of Bremen, and the CryoSat-2 data product derived at Alfred Wegener Institute. The availabilty of the above data streams is limited to the period from 2012 to 2014.  We assimilate the above four data streams in the spring of each of the three years and predict the ice conditions in the following summer.

Initial tests indicated that our model was not sufficiently calibrated to achieve the required simulation quality. Hence, we calibrated the model using observations over an 19 year period. Furthermore, it turned out that the model was not capable to absorb the information in the above-mentioned ice thickness product to a sufficient degree. Through a set of additional assimilation experiments, we were capable of developing a so-called bias correction scheme. This yields a considerable improvement in forecast skill for the sea ice from July to September for all three years. The bias correction scheme can now also be applied and tested to years outside the period from 2012 to 2014. One of these applications is the Sea Ice Outlook 2015.

Lakes that form a seasonal ice cover are a major component of the terrestrial landscape. They cover approximately 2% of the Earth’s land surface, with the majority of them located in the Northern Hemisphere. Lakes have the highest evaporation rates of any high latitude terrestrial surface. Their frequency and size greatly influence the magnitude and timing of landscape-scale evaporative and sensible heat inputs to the atmosphere and are important to regional climatic and meteorological processes. Shallow lakes warm quickly in spring and have very high evaporation rates until they freeze in autumn. Large lakes take a substantially longer period to warm, but stay ice-free (or partly ice-free) into early winter, and their total evaporation amounts are significantly greater. The duration of lake ice in particular, controls the seasonal heat budget of lake systems thus determining the magnitude and timing of evaporation. The presence (or absence) of ice cover on lakes during the winter months has also an effect on both regional climate and weather events (e.g., thermal moderation and lake-effect snow. Monitoring of lake ice is therefore critical to our skill at forecasting high-latitude weather, climate, and river runoff. Recent investigations emphasize the importance of considering lake ice cover for modelling the energy and water balance of high-latitude river basins, for regional climate modelling, and for improving numerical weather prediction in regions where lakes occupy a significant fraction of the landscape.

In addition to its significant influence on biophysical and socio-economic systems (e.g., duration of ice-road and open-water shipping seasons), lake ice is also a sensitive indicator of climate variability and change. Documented trends and variability in lake ice conditions have largely been related to air temperature changes and, to a lesser extent, snowfall. Long-term trends observable from ground-based records reveal increasingly later freeze-up and earlier break-up dates, closely corresponding to increasing air temperature trends. Broad spatial patterns in these trends are also related to major atmospheric circulation patterns originating from the Pacific and Atlantic oceans such as the El Niño-La Niña/Southern Oscillation, the Pacific North American pattern, the Pacific Decadal Oscillation, and the North Atlantic Oscillation/Arctic Oscillation.

Despite the wide-ranging influences of lake ice and its robustness as an indicator of climate change, a dramatic reduction in ground-based observational recordings has occurred globally since the 1980s. Satellite remote sensing provides the necessary means to increase the spatial coverage and temporal frequency of ground-based observations. This paper will provide an overview of recent advances in the development of ice cover retrieval algorithms and products (e.g., ice cover extent and phenology, ice thickness, and ice/snow albedo and temperature) for large lakes at northern high latitudes from optical and microwave satellite sensors. It will also cover future remote sensing capabilities in light of the upcoming Sentinel-3A/B that will make use of multiple sensors (MWR: Microwave Radiometer, OLCI: Ocean and Land Colour Instrument, SLSTR: Sea and Land Surface Temperature Radiometer, SRAL: SAR Altimeter) all relevant for ice cover monitoring of large northern lakes.