The snow water equivalent (SWE), representing the amount of snow stored in a snowpack, is a key parameter for diverse hydrological applications in regions influenced by seasonal snow. The seasonal snowpack is released during the melt period in spring or shorter mid-winter melt events due to temperature increase or rain-on-snow events. Spatially-distributed and continuous up-to-date knowledge on the amount of water stored as snow and its melting stage has an important impact on the safety of residents and the economics. Especially in remote and numerous mountainous areas, the measurement density of SWE is up to now still very limited and the operations (stations and field observations) are usually costly, labour-intense and/or risky.

Within the ESA business applications (formerly ARTES IAP) demonstration project SnowSense (2015-2018), including team members from VISTA, ANAVS and LMU Munich, we combined different monitoring and modelling approaches as well as temporal and spatial scales. We developed and applied a novel in-situ snow measurement method, used hydrological modelling approaches as well as satellite-based information to be able to retrieve runoff, flood and hydropower forecasts in remote areas such as Newfoundland, Canada. The main scope of this project is to design, develop and demonstrate an operational service, providing spatial SWE information, run-off assessment and relevant information for hydropower plant management. The entire service is created by a system including autonomous in-situ Global Navigation Satellite System (GNSS) based snow monitoring stations, remote sensing Earth observations (EO) (e.g. using Sentinel-1 for wet-snow monitoring), and the integrated spatially distributed hydrological model component PROMET. EO and modelling have been applied in various services and application within VISTAs activities over the last decades to provide information to fulfill hydrologists’ needs (flood forecast, hydropower). For the project SnowSense we applied near-real-time processing, which is partly based on ESA’s Thematic Exploitation Platforms (TEPs), and daily modelling and assimilation techniques for combining in-situ stations and spatial EO information. As demo service, we provided weekly updated information products on the spatial distribution of SWE (1 km resolution), e.g. within sub-catchments above power plants, and current and predicted run-off to the involved users.
For applications dedicated in remote areas, which are often characterised by a lack of meteorological measurements, the hydrological model component uses meteorological forcing from limited-area model (LAM) forecast grids, e.g., from the Global Environmental Multiscale (GEM) model.

Regarding the applied in-situ snow measurement station, we developed a novel approach to derive SWE and liquid water content (LWC) in snow based on the freely available signal strength and carrier phase of the GNSS signals. Due to a quite simple, low-cost GNSS snow measurement station set-up (SnowSense Station), it was possible to run multiple sensor installations in hydrological catchments and at scientific test sites with reasonable technical and financial effort. For these in-situ measurements, one GNSS antenna was placed on the ground below the snowpack and the other antenna on a pole above the snowpack. Before the first snow fall, the baseline between the two antennas was determined with millimetre accuracy using Real Time Kinematic (RTK) positioning. As soon as snow accumulated on top of the lower antenna on the ground, the GNSS signals were significantly delayed and attenuated compared to signals, which traveled only through the air. The focus of the technical development and design of the SnowSense in-situ snow stations was to provide a suitable solution especially for locations, which were so far not equipped with stations or were not regularly visited by field trips. Based on the usage of GNSS signal approaches for snow parameter retrievals (Koch et al., 2014, 2019 (in prep.) and Henkel et al., 2018), an autonomous and transportable, stand-alone solution, has been designed, produced and demonstrated. This sensor solution can, however, also be integrated in existing stations. The in-situ stations provide valuable and point input for the hydrological PROMET model to retrieve updated spatially distributed SWE information.


The algorithms behind the SnowSense stations were primary developed at the test site Weissfluhjoch / Switzerland and the first stations were successfully applied in Newfoundland and Quebec / Canada. The validation of the in-situ snow measurements and hydrological modelling performed well. A full service (SnowSense Service), as prototype for potential natural resources management support, has been demonstrated in Newfoundland in the winter 2017/2018. In the last years, already the first stations had been sold to national European science institutes and authorities. Besides, a commercial service roll-out is targeted for the next years.
We will give an overview on the SnowSense project itself and present results of the independent and self-sufficient in-situ SnowSense stations, our EO and hydrological modelling approach as well as the demo service performance to underline the efficiency of the entire SnowSense Service package. In addition to the continuous in-situ point observations, the assimilated EO data are used to adjust hydrological forecasts, which potentially has an enormous commercial value for hydropower plants, risk assessment and ecosystem management.


Koch et al. (2014): Measuring snow liquid water content with low-cost GPS receivers. Sensors, 14(11), 20975-20999.
Henkel et al. (2018): Snow Water Equivalent of Dry Snow Derived from GNSS Carrier Phases, IEEE Transactions on Geoscience and Remote Sensing, 56 (6), 3561 – 3572.
Koch et al. (2019): Retrieval of snow water equivalent, liquid water content and snow height of dry and wet snow by combining GPS signal attenuation and time delay – Water Resources Research, in prep.

Environment and Climate Change Canada (ECCC) has identified moderate resolution (~1 km) terrestrial snow water equivalent (SWE) retrieved across the Northern Hemisphere at a revisit of 1 to 3 days as a priority observational gap which limits operational environmental prediction, monitoring, and the delivery of climate services. To address this need, ECCC partnered with the Canadian Space Agency (CSA) to develop a new satellite mission concept. Following technical trade-off studies, a technical solution capable of providing dual frequency Ku-band (13.5 and 17.2 GHz) radar measurements at 250 m spatial resolution across a 500 km swath was identified. Orbit analysis defined a dawn/dusk orbit (0600 LTDN) flying in loose convoy at a fixed offset of 3.5 hours ahead of Met-Op Second Generation (Satellite B). This orbit will provide synergistic multi-frequency active and passive measurements through complete swath overlap with the Microwave Imager (MWI) and C-band Scatterometer (SCA) instruments. ECCC is leading a number of science activities to support the mission development, currently in Phase 0 at CSA. This presentation will provide an overview of ongoing field campaign and modeling activities with the objectives to (1) identify the current state of backscatter modeling capabilities for layered snowpacks at Ku-band, and (2) develop and validate SWE retrieval algorithms.

Two sets of airborne measurements are utilized. First, data were acquired over a shrub tundra watershed (Trail Valley Creek) near Inuvik, North West Territories, Canada in March and April 2013 from a dual-frequency (17.2/9.6 GHz), dual-polarization (VV/VH) synthetic aperture radar (SAR; ‘SnowSAR’). Coordinated comprehensive ground sampling measurements were used to characterize multi-scale variability in depth, density, and snow water equivalent (SWE), as well as horizontal and vertical variability in stratigraphic snow properties (layering, density, snow grain specific surface area). These data allowed parameterization of the Microwave Emission Model of Layered Snowpacks adapted to include backscattering (MEMLSa) to diagnose mechanisms of backscatter variability, evaluate forward modeling capabilities, and determine the potential for SWE retrievals. This analysis demonstrated the strong influence of local variations in snow microstructure on the measured backscatter, and reinforced the need to account for soil and vegetation properties in addition to snow (King et al., 2018).

The 2013 airborne measurements all occurred late in the snow season, and therefore did not capture the evolving influence of seasonal snow accumulation and metamorphosis on the Ku-band signal. To address this gap, a series of new airborne Ku-band measurement campaigns are imminent, using the University of Massachusetts 13.2 GHz, single polarization, interferometric SAR system. Three measurement periods are planned between November 2018 and March 2019, over the Trail Valley Creek watershed. Sampling periods in November, January, and March will capture early, mid, and late season snow cover characteristics. Emerging SWE retrieval techniques, including the provision of first guess snow microstructure information from passive microwave measurements (Lemmetyinen et al., 2018), will be evaluated. Time series of RADARSAT-2 and TerraSAR-X imagery (acquired during snow-covered and snow-free conditions) will support enhanced consideration of the background soil state, vegetation, and roughness. Collectively, these airborne datasets will advance understanding of Ku-band radar interactions with seasonal snow, fundamental to advancing the Phase 0 mission presently in development at CSA.

King, J., C. Derksen, P. Toose, A. Langlois, C. Larsen, J. Lemmetyinen,P. Marsh, B. Montpetit, A. Roy, N. Rutter, and M. Sturm. 2018. The influence of snow microstructure on dual-frequency radar measurements in a tundra environment. Remote Sensing of Environment. https://doi.org/10.1016/j.rse.2018.05.028

Lemmetyinen, J., C. Derksen, H. Rott, G. Macelloni, J. King, M. Schneebeli, A. Wiesmann, L. Leppänen, A. Kontu, and J. Pulliainen. 2018. Retrieval of effective correlation length and snow water equivalent from radar and passive microwave measurements. Remote Sensing. 10, 170; doi:10.3390/rs10020170.

Current satellite systems are unable to deliver information on Snow Water Equivalent (SWE) at the spatial resolution and accuracy necessary to meet requirements for operational environmental monitoring, services, and prediction. CoReH2O (Cold Regions Hydrology High-resolution Observatory), a candidate for ESA’s 7th Earth Explorer (Rott et al., 2010; ESA 2012) aimed to address these issues by providing timely information on SWE over land at a resolution ranging from 200 to 500 meters. The sensor envisaged for CoReH2O was a dual polarization, dual frequency Synthetic Aperture Radar (SAR) at X and Ku bands.
The ESA SnowSAR instrument is a side looking, dual polarized (VV/VH), X/Ku band SAR, operable from a small aircraft (Trampuz et al., 2011; DiLeo et al., 2015). Between 2010 and 2013, the instrument was operated at several sites in Northern Finland, Austrian Alps, Northern Canada and Alaska. The purpose of the flight campaigns was to gather information on the backscattering properties of snow covered terrain, and demonstrate the CoReH2O retrieval concept for SWE and snow accumulation over glaciers for a range of climatological snow classes and land cover regimes. SnowSAR was deployed in Sodankylä, Northern Finland for a single test flight in 2011 and a total of ten acquisition flights in the winter of 2011-2012. Additional acquisitions took place over a tundra site as well as over partially snow-covered sea ice. Over the Austrian Alps, three flight campaigns were performed between November 2012 and February 2013 over three sites located in different elevation zones, representing Alpine lowland/meadow, a transitional valley, and a glacier environment (Rott et al., 2014). In Canada a total of two successful missions were flown in March and April 2013, over a site in Trail Valley Creek, Northwest territories, representative of the tundra snow regime (King et al., 2018).
Flights at the Sodankylä site consisted of a mosaic of 25 North-South data takes as well as two data takes in the East-West direction. A large amount of basic in situ data on snow depth and SWE variability was collected along the flight transects, while data on snow stratigraphy was collected from snow pits at fixed locations. A similar acquisition strategy was followed at the site in Canada, with additional focus on detailed observations of snow stratigraphy variations at the 1-50 meter scale. Observations at the Alpine sites were similarly supported by in situ observations.
The data from SnowSAR is provided as calibrated and geocoded normalized radar cross sections (sigma-nought). At the designated flight altitude of 1200 meters, the SnowSAR swaths span approximately 400 meters in range and up to several kilometres in the flight direction, while the calibrated sigma-nought values are provided at a spatial resolution of 2 or alternatively 10 meters. The instrument calibration was performed by a combination of internal calibration, corner reflector targets installed along flight paths, and cross-calibration with space-borne instruments (DiLeo et al., 2015).
In 2016, after receiving final calibrated and geocoded datasets for SnowSAR, ESA launched a dedicated stud to collectively analyse the acquisitions. The backscattering intensity observed by SnowSAR at four channels (X and Ku band, VV and VH polarization) was compared to prevailing conditions on the ground, analysing the effect of terrain, topography, snow properties, vegetation and weather conditions. The assessment of the capabilities and limitations of SnowSAR data to retrieve the main snow parameters of CoReH2O (SWE, snow accumulation on glaciers) was investigated. Specifically, the main objectives were to:
• Compare the snow cover information with SnowSAR observations, as well as snow process models and passive microwave data, analysing the sensitivity of X/Ku band microwave backscattering to the main snow parameters over the test sites
• Derive relations between the main snow parameters of SWE and snow accumulation over glaciers and backscattering data, applying these in a retrieval
We present the main findings of the study, including an assessment of backscattering conditions against changing snow cover conditions, examples of performance of backscattering forward models against SnowSAR observations, and demonstrations of SWE retrieval using the data over several sites.

Di Leo D, Coccia A, Meta A (2015) Technical Assistance for the Development and Deployment of an X-and Ku-band MiniSAR Airborne System (SnowSAR). ESTEC No. 4000106761-CCN1. (https://earth.esa.int/web/guest/campaigns)
ESA (2012) Report for Mission Selection: CoReH2O. SP1324/2 (3 volume series). European Space Agency. Noordwijk, The Netherlands. http://esamultimedia.esa.int/docs/EarthObservation/SP1324-2_CoReH2Or.pdf
King, J., C. Derksen, P. Toose, A. Langlois, Alexandre, C. Larse, J. Lemmetyinen, P. Marsh, B. Montpetit, A. Roy, N. Rutter, and M. Sturm, 2018. The influence of snow microstructure on dual-frequency radar measurements in a tundra environment. Remote Sens. Environ., 215, 242-254, ISSN 0034-4257.
Rott H et al (2014) Technical Assistance to Fieldwork in the Austrian Alps during AlpSAR. ESTEC No. 4000107780-CCN1. Final reports available via ESA campaign webpage (https://earth.esa.int/web/guest/campaigns)
Rott, H., Yueh, S.H., Cline, D.W., Duguay, C., Essery, R., Haas, C., Hélière, F., Kern, M., Macelloni, G., Malnes, E., Nagler, T., Pulliainen, J., Rebhan, H., Thompson, A. (2010). Cold Regions Hydrology High-Resolution Observatory for Snow and Cold Land Processes, Proceedings of the IEEE, 98(5), 752-765., doi:10.1109/JPROC.2009.2038947
Trampuz CH, Coccia A, Imbembo E (2011). Technical Assistance for the Development and Deployment of an X-and Ku-band MiniSAR Airborne System. Final Report ESA ESTEC No. 4000101697/10/NL/FF/ef.

The mass of snow that yearly covers the terrestrial land surface ranks among the most uncertain variables in hydrology. In mountain areas, this uncertainty is amplified by the high spatial variability in snow and the lack of remote sensing products. Operational snow depth products are currently available either from airborne lidar observations at high resolution over local mountain areas, or from passive microwave observations at low (~25 km) resolution and global scale. However, the latter typically exclude mountain areas due to their coarse resolution and saturation at relatively shallow snow depths (~0.8 m). Therefore, remote sensing observations of snow depth at high resolution that cover the continental to global scale are still lacking.

Here, we demonstrate the ability of the European Space Agency and Copernicus Sentinel-1 C-band (5.4 GHz) mission to map snow depth at an unprecedented 1 km spatial and less than weekly temporal resolution across the Northern Hemisphere mountains. The retrieval algorithm relies on a change detection approach, linking temporal changes in the radar polarization ratio (VH- over VV-polarized backscatter) with the accumulation or ablation of snow. IMS (Interactive Multisensor snow and ice mapping system) snow cover data is used as ancillary information to mask out retrievals when snow is absent. Forest cover percentage from fractional land cover data is used to correct for the obscuration of snow backscatter by forest canopies, a correction that is similarly applied to coarse-scale passive microwave retrievals.

The Sentinel-1 retrievals are evaluated using in situ measurements at ~5000 locations in mountainous areas scattered over the Northern Hemisphere. These include measurements from SNOTEL (snow telemetry) and cooperator stations in the western United States, networks in British Columbia (Canada), the European Alps, the Pyrenees, and Himalayas, as well as global data collections with World Meteorological Organization (WMO) stations. The spatial distribution of our 1-km snow depth retrievals correlates to ~0.7 with the point-scale in situ measurements, despite the significant mismatch in spatial scale. Likewise, the temporal evolution of SD and SWE is accurately characterized over two consecutive winter seasons (from 2016 to 2018), with an average correlation of ~0.7 across all sites, and a good representation of the inter-annual snow variability. Interestingly, the Sentinel-1 retrievals capture the excessive snowfall in the lower western US in the 2016-2017 winter season, and reveal the anomalously high snow accumulation in the Canadian Rockies and European Alps in the 2017-2018 winter season.

These results highlight the potential of Sentinel-1 C-band radar data, with continuity expected until 2030 at least, to benefit long-term water resources and climate applications related to seasonal snow.

Despite the main role played by snowfall rate on arctic amplification, significant disagreements between state-of-the-art climatologies of snowfall in high latitude regions remain. Due to the sparse in situ observation network and extreme environment, remote sensing is an essential tool to better characterize precipitation.
Since 2006, the CloudSat polar orbiting satellite operates with the Cloud Profiling Radar (CPR) onboard. The specific characteristics of CPR enable effective observations of shallow clouds and light precipitation amounts, typical for the Arctic. This is the first radar instrument operating at latitudes as far north as 82°N and the first sensitive enough to reliably detect falling snow.
A characterization of snowfall in the Arctic based on CloudSat will first be presented. We examine the frequency, phase and rate of snowfall of the CloudSat climatology over 2007-2010. Then we compare ECMWF (European Center for Medium Range Weather Forecast) Reanalysis Interim (ERA-I), and Arctic System Reanalysis (ASR v1 and 2) with CloudSat. While similar general patterns are observed between all products, there are significant differences especially over Greenland. All reanalyses represent quantitatively well the inter-annual and seasonal cycle of snowfall rates compared to CloudSat. Yet important differences remain when compared quantitatively. Regional reanalyses (both ASR versions) show closer results with CloudSat compared to ERA-I, especially over Greenland that represents the area of stronger disagreement. Over the whole Arctic, ERA-I shows a higher occurrence of weak snowfall rates and a lower occurrence of strong snowfall rates compared to CloudSat. On the other hand, ASR shows a good seasonal distribution of snowfall rates compared to CloudSat, except in summer, during which ASR presents a lower occurrence of weak snowfall rates and a higher occurrence of strong snowfall rate. All through the year, ERA-I strongly underestimates snowfall rates compared to CloudSat except in summer, during which ERA-I shows a better agreement with CloudSat than ASR. While annual mean snowfall rates of ERA-I present an underestimation above continents and oceans for the considered period, annual mean snowfall rates of ASR display an underestimation above oceans and overestimation above continents. The new version of ASR is rather similar to the first version when comparing solid precipitation.
We also use CloudSat products to conduct an exploratory study on the use of passive microwave observations for the detection of snowfall in polar regions.

Seasonal snow has important societal impacts. It is a significant resource for water for human consumption, agriculture, hydropower, and other industries. It has played a key role in major floods and droughts. Snow’s importance is likely to increase in water-related conflicts. And, seasonal snow is a key part of weather and climate.
Despite this importance, obtaining an accurate global picture of snow (snow water equivalent in particular) has proven to be a challenge. While observation-based and combined observation-model based global snow products have existed for some time, they have excluded important regions where observations or retrievals are difficult—for example, forests and complex terrain. Also, no single type of observation or retrieval algorithm works for all types of snow under all conditions.
One solution is to combine different observation types--passive and active microwave plus passive and active optical along with models—to synthesize a global snow water equivalent product. Sensors already in orbit plus sensors already planned to be launched can be combined with notional new sensors. Within this mix is a complex tradespace involving swath widths, repeat intervals, orbit inclinations, footprint size, footprint spacing, observation accuracy, and so on. Over a time span of years, the mix will even change as missions come and go.
We will utilize NASA’s Tradespace Analysis Tool-Constellation (TAT-C) simulation tool to explore potential combinations of existing and future sensors. These simulations will tell us—for each configuration--how much of the seasonal snow world can be observed, how often, with what footprint size and spacing and with what swath width. Such information will be highly valuable for informing discussions on future snow mission concepts. It will also highlight where modeling efforts can provide the greatest impact and perhaps indicate the parameters needing the greatest improvements in accuracy or precision. The results of the simulations will help make progress toward accurate global snow products.