Snow 1: Land

Reliable information on snow water equivalent (SWE) at regional to continental scales is required for studies of freshwater and energy budgets, the evaluation and initialization of land surface models for both short term weather and seasonal forecasts, and the assessment of coupled climate model simulations. While multiple independent passive microwave derived SWE products are available, their full potential has not been realized because of poorly constrained error budgets due to the challenges related to the physical processes underpinning the SWE retrievals, and the extensive snow covered regions of the world without adequate surface observations for algorithm validation. Seven daily, gridded Northern Hemisphere snow water equivalent (SWE) datasets derived from satellite passive microwave measurements and reanalysis driven land models were assessed as part of the European Space Agency (ESA) Satellite Snow Product Inter-comparison and Evaluation Experiment (SnowPEx). Analysis was performed in order to quantify the spatial and temporal consistency of the satellite retrievals relative to land surface assimilation systems, physical snow models, and reanalyses. All datasets were also compared to independent in situ reference datasets in order to determine absolute uncertainty and bias.

 While the climatologies of total Northern Hemisphere snow water mass (SWM) vary among the datasets by as much as 50%, their interannual variability and daily anomalies are comparable, showing moderate to good temporal correlations (between 60% and 85%) on both interannual and intraseasonal time scales. Wintertime trends of total Northern Hemisphere SWM are consistently negative for all datasets over the 1981–2010 period but vary in strength by a factor of 2 to 3. Examining spatial patterns of SWE indicates that the datasets are most consistent with one another over boreal forest regions compared to Arctic and alpine areas. The exception are standalone satellite passive microwave datasets, which have spatially inconsistent mean fields with satellite retrievals that utilize ancillary in situ measurements, and physical snow models driven by reanalysis. Additionally, the datasets derived using relatively recent reanalyses are strongly correlated with one another and show better correlations with a combined satellite-in situ product (the ESA GlobSnow product) than do those using older reanalyses. Various combinations of the SWM datasets were evaluated against an extensive dataset of snow survey measurements across Eurasia in order to determine the optimal observational ensemble with minimized bias and absolute error. The GlobSnow product is the only satellite derived dataset which can be included in an optimal combination of datasets. Finally, a comparison of eight reanalysis datasets over the 2001–2010 period shows that land surface model differences control the majority of spread in the climatological value of SWM, while meteorological forcing differences control the majority of the spread in temporal correlations of SWM anomalies. While multiple datasets are available to characterize coarse resolution SWM for climate applications, finer resolution products required for operational land surface modeling applications remain a significant observational gap.

Sunlight absorption is a dominant melt energy source for snow and ice on land, glaciers, sea, and lakes. Snow and ice albedo feedbacks are fundamental to Arctic amplification of warming with apparent lower latitude weather impacts [1,2].

The Sentinel-3 sensors including the Ocean Land Colour Instrument (OLCI) and Sea and Land Surface Temperature Radiometer (SLSTR) data enable production of 1) snow/ice albedo and 2) snow-covered area products for relatively flat areas, while Sentinel-2 MultiSpectral Instrument (MSI) yields resolutions desirable for snow/ice albedo retrieval in areas of complex terrain.

The commitment that the EU Copernicus program makes for long-term operational and environmental monitoring adds great value to Climate Data Records building on 1) 1981-2007 AVHRR and 2) NASA MODIS era beginning in 2000. VIIRS data are so far not proven to extend the MODIS capability to produce an albedo climate data record [3].

This work steers toward an ESA CCI+ project to deliver daily products of: 1) Northern Hemisphere 600 m terrestrial snow albedo and snow-covered area; and 2) bare ice area for land, sea, and lake ice. For mountainous regions, the product is at 150 m resolution. The project is also to form a Climate Data Record from past satellite missions, chiefly AVHRR and MODIS.

Ongoing since 2007, the PROMICE automatic weather station [4] data and year 2015 aerial spectral imagery support albedo product calibration and validation. Similarly, the Greenland Climate Network (GC-Net) provides ground truth beginning in 1995. Elsewhere in the Northern Hemisphere, ground truth is available from ground stations, including the Baseline Surface Radiation Network (BSRN).

This effort obtains advisorship from US and Japanese colleagues and is in coordination with the International Snow Working Group Remote Sensing (iSWGR).

Notes:

¹ http://www.pnas.org/content/111/34/12331.full

² http://onlinelibrary.wiley.com/doi/10.1029/2012GL051000/abstract

³ http://viirsland.gsfc.nasa.gov/Products/Albedo.html

⁴ http://promice.org/home.html

Conventionally considered as having minimal effect on microwave radiation at L-band, dry snow nevertheless affects observed microwave signatures at these wavelengths through changes in impedance matching and refraction angle at the soil interface. As explained in recent theoretical studies [1,2], these create a discernible effect in emitted microwave brightness temperature, determined largely by snow density in the lowest layers of a seasonal snowpack. The effect can be simulated using a relatively simple emission model based on the two-stream approach [2]. Applying dual polarized L-band observations in an inversion scheme of the model, this raises the potential of the simultaneous retrieval of density of the snow layer in contact with the ground, as well as the effective permittivity of the ground layer. Such information has, in turn, the potential to support e.g. retrievals of Snow Water Equivalent (SWE) from other sensors [3].

We present results of a recent ESA study STSE Pathfinders - Combined use of multifrequency radiometry (L- to Ka-Band) and SAR imagery for enhanced monitoring of terrestrial cryosphere processes, where amongst other topics, the potential of L-band observations to support SWE retrievals was explored. A large dataset of experimental, tower-based observations from Finland and Canada were first used to investigate the validity of the established emission model [4]. Snow excavation experiments were applied to demonstrate the capability of the emission model to estimate the effect of snow cover on emitted L-band brightness temperature. Retrievals covering entire winter periods over predominantly mineral soil sites indicated that both snow density and the effective ground permittivity can be retrieved from the L-band observations; following theoretical predictions, the retrieved snow density was most correlated with snow densities measured at lower layers of the snowpack. Over complex terrain such as a wetland site, the retrieval occasionally produced erroneous results, indicating a need to further develop the applied emission model for different soil and vegetation regimes.

Furthermore, we present the first trials of applying the retrieval of snow density and ground permittivity from space, using observations from both the SMOS (Soil Moisture and Ocean Salinity) and SMAP (Soil Moisture Active Passive) sensors, and outline methodologies to employ the obtained retrievals in the GlobSnow SWE retrieval processor [3]. The sensitivity of L-band observations to snow also have implications for SMOS and SMAP soil moisture retrievals over regions affected by seasonal snow cover. The preliminary results of space-based retrievals are encouraging, but the scarcity of suitable reference datasets prevents a comprehensive validation of the method at the satellite scale. A thorough investigation  would be required to fully assess the potential of the method, including the investigation of the retrieval applicability over complex land cover such as the boreal forest zone, wetlands and lake ice. Requirements for a future airborne campaign to address these issues are outlined.

[1] Schwank, M., K. Rautiainen, C. Mätzler, M Stähli, J. Lemmetyinen, J. Pulliainen,  J. Vehviläinen, A. Kontu, J. Ikonen, C. Bauduin-Ménard, M. Drusch, A. Wiesmann, and U. Wegmüller, 2014. Model for Microwave Emission of a Snow-Covered Ground with Focus on L Band. Remote Sens. Environ. 154: 180-191.

[2] M. Schwank, C. Mätzler, A. Wiesmann, U. Wegmüller, J. Pulliainen, J. Lemmetyinen, K. Rautiainen, C. Derksen, P. Toose, M. Drusch, 2015. Snow Density and Ground Permittivity Retrieved from L-Band Radiometry: A Synthetic Analysis. IEEE J. Selected Topics App. Earth Obs. Remote Sens. 8(8): 3822-3845.

[3] Takala M., K. Luojus, J. Pulliainen, C. Derksen, J. Lemmetyinen, J-P. Kärnä, J. Koskinen, and B. Bojkov, 2011. Estimating northern hemisphere snow water equivalent for climate research through assimilation of space-borne radiometer data and ground-based measurements. Remote Sens. Environ., 115(12): 3517-3529.

[4] J. Lemmetyinen et al. Snow density and ground permittivity retrieved from L-Band radiometry: Application to experimental data. Remote Sens. Environ. (submitted June 2015).

Seasonal snow is a main element of the global water cycle and climate system. Due to its strong influence on the radiation and energy balance, changes in snow extent tend to amplify climate fluctuations. Terrestrial snow covers up to 50 million km2 of the Northern Hemisphere in winter. It is characterized by high spatial and temporal variability. Therefore, satellite observations provide the only means for timely and complete observations of the global snow cover. SnowPEx is an international collaborative effort, funded by the European Space Agency (ESA) under the Quality Assurance framework for Earth Observation (QA4EO) that intercompares and evaluates satellite-based seasonal snow cover products of hemispheric to global extent, assesses the product accuracy, and identifies discrepancies between the various products. Furthermore, in support of climate studies, trends in the hemispheric seasonal snow coverage and snow mass haves been documented, based on an ensemble of satellite based snow products.

SnowPEx focuses on two parameters of the seasonal snow pack, the snow extent (SE) from medium resolution optical satellite data (MODIS, AVHRR, VIIRS, etc.) and the snow water equivalent (SWE) from passive microwave data (SSM/I, AMSR, etc.). Overall 14 continental to global satellite snow extent products (including fractional snow cover products) and three SWE products are participating in the intercomparison and validation experiment, with test areas spreading over different environments and climate zones. For the intercomparison daily SE products from 5 years have been transformed to a common map projection and standardized protocols, developed in the project, are applied. The SE product evaluation applies statistical measures for quantifying the agreement between the various products, including the analysis of the spatial patterns. Extensive validation of snow extent products is carried out using high resolution snow maps, generated from about 450 Landsat scenes in different snow zones and over various land surface types. Additionally, an in-situ snow reference data set is used, including station data from various organisations in Europe, North America and Asia. Validations with high resolution Landsat based snow products and in-situ data reveal statistical differences between various snow extent products of up to 30%, depending on the land surface type and complexity of the terrain. For the coarse resolution SWE products from passive microwave sensors, sites with dense networks of in-situ measurements are used for validation. The SWE products are also inter-compared with gridded snow products from land surface models driven by atmospheric reanalysis. In addition, the multi-year trends of the various SWE products are evaluated. We provide an overview on the contributing snow products, the evaluation protocols, and the results of QA4EO SnowPEx and report on multi-decadal trends in the various snow products. 

The mapping of wet snow with SAR sensors has traditionally relied on strict repeat-pass comparisons whereby a new image is compared with a stock reference image that was acquired under dry conditions with the same geometry.  We present here the results from a newer wet snow mapping technique that loosens the constraints concerning the need for exact repeat imagery.  Allowing a more inclusive data model makes it easier to consider imagery also from other sensors in different orbits (e.g. Radarsat-2 and Sentinel-1, or TerraSAR-X and Cosmo-Skymed).

Within the framework of ESA’s programme IDEAS+, techniques for improving the retrievals of wet snow parameterisations are being tested at different scales and in different regions.  We apply rigorous radiometric terrain correction, normalising the radar cross section for the contributing DEM-area measured in the plane perpendicular to the slant range direction, generating terrain-flattened gamma nought [1].  For initial tests of wet snow mapping, we compare the terrain-flattened imagery directly with reference images acquired under dry conditions.  As even radiometrically flattened imagery retains relatively poor spatial resolution on foreslopes, in a further processing stage, we combine ascending/descending imagery into backscatter composites that can cover wider regions, highlighting the seasonal melt progression over wide areas.  Both processing stages rely on a high standard for both geometric [2] and radiometric calibration of the radar sensors employed.  Backscatter composites generated from different seasonal stages are compared with a single dry reference composite to illustrate the seasonal wet snow signature at different heights.

We present results from wide-area lower resolution processing over the whole of the European Alps, as well as full resolution processing runs carried out within the territory of Switzerland.  We use Radarsat-2 data provided through the SOAR-EU programme to generate a virtual constellation with Sentinel-1.  We illustrate how integration of information from multiple missions, as made possible by organisations such as the WMO Polar Space Task Group’s SAR Coordination Working Group, enables retrieval of geophysical parameters with higher quality and at higher temporal resolution than would otherwise be possible.  We illustrate how use of seamless composites (e.g. as seen in Fig. 1) simplifies data fusion with datasets from MODIS or Landsat and multiple radar sensors.  Over the European Alps, the SRTM height model was used; within Switzerland, a subsampled version of the 2m swissALTI3D height model was employed.

Wet snow maps generated from radar imagery are compared with those obtained independently from an operational process that relies on in-situ ground station measurements together with weather station measurements made across the Swiss Alps.  Confusion matrices are used to compare the classifications made using (a) SAR sensors and (b) in-situ based evaluations.

The inclusion of a wider set of radar sensors (e.g. Radarsat Constellation Mission together with the Sentinels) offers the promise of supporting the generation of wet snow maps at higher temporal resolution than would be possible from either constellation in isolation.