Snow plays a key role in the hydrological cycle. Especially in mountain regions such as the Alps, snowmelt is the main responsible of run-off regimes, strongly influencing groundwater storage, flooding, avalanches, and contaminant release. Hence, understanding and predicting snowmelt dynamics is essential for a better planning and accurate monitoring of our resources.
Snowmelt is a non-linear process affected by the strong variability of several climate factors. We can observe three main phases: i) moistening, ii) ripening and iii) run-off. The moistening is the initial phase. The air temperature increases and due to heat exchanges and/or rain the superficial layers of the snowpack start melting during the day and refreeze during the night. During the ripening phase, the wetting front penetrates through the snowpack but the meltwater is not yet released. During this phase, the snowpack becomes isothermal and when no more liquid water can be retained, the run-off phase starts. The sequence of these three phases can vary depending on temperatures with alternatively warming and refreezing cycles (Dingman, 2015).
Several non-invasive field measurements such as air temperature, relative humidity, wind speed, precipitation, solar radiation are used to extract information on snow dynamics. However, the key state variables to properly identify the three main melting phases are the snow water equivalent (SWE), i.e. the amount of water stored in form of snow, and the liquid water content (LWC), i.e. the amount of free liquid water inside the snowpack. Unfortunately observations of SWE and especially of LWC are rarely available. In detail, diurnal fluctuations of LWC happen during the moistening. The ripening is characterized by a relevant increment of LWC. Afterwards SWE decreases monotonically during the run-off phase.
To obtain information on these parameters, an alternative to ground-based measurements is represented by simulations from physically based models such as SNOWPACK (Bartelt & Lehning, 2002). Such models make use of the most common meteorological parameters and provide as output SWE, LWC and snow structure details. This allows the monitoring of snowmelt dynamics, but these parameters are not directly measured.
On the other side, active microwave sensors such as the Synthetic Aperture Radar (SAR), which are sensitive to the presence of LWC in the snowpack, may represent a valid source of information for monitoring the snowmelt (Shi & Dozier, 2000; Longepe et al. 2009), at high spatial and temporal resolution (i.e. with the Sentinel-1 constellation with a 20 m nominal resolution and 5 days of overpassing time at the equator). In this work, we aim at studying the relationship between the multi-temporal SAR signal acquired from Sentinel-1 and the measurements of LWC and SWE taken in different European alpine sites.
Characteristic backscattering mechanisms at the different melting phases are identifiable from the multi-temporal SAR signal. In detail, we recognize a decrease of the backscattering during the first phase which can be directly correlated with an increase of LWC. The decrease can initially be observed only during the afternoon when the snow melts because positive temperatures and solar radiation, and not in the early morning when the snow is refrozen. During the ripening phase, the backscattering decreases, reaching its minimum value. It can oscillate but it remains however low. Then, a rapid increase of the signal is observed when the run-off starts. In Figure 1 an example of such behavior observed at the Zugspitze site (Germany) is reported. The example shows the complexity and non-linearity of the melting process. The ripening is limited to a short phase and the run-off starts very early with a first increase of the backscattering. It decreases again in correspondence of a snow fall (probably wet) highlighted by an increase of LWC and SWE followed by a relatively colder period which lasted some days and probably interrupted the run-off process. Afterwards the backscattering increases again when the run-off restarts and lasts till the end of the season.
Based on this analysis, the variation in the backscattering can be exploited to automatically identify the timing of the snowmelt. Further research is needed to extend a single point observation to a spatially distributed information. Given their availability, high spatial resolution and frequent overpassing time, Sentinel SAR observations are worth being used to monitor snow melt dynamic over large regions, especially in remote areas and complex terrain such as mountain regions. This approach can represent an important support for hydrological monitoring.

Caption Figure 1: Temporal evolution of the backscattering coefficient values acquired over Zugspitze (Germany) during the hydrological season 2017/2018. It is compared with LWC and SWE measured in situ and/or modeled with SNOWPACK. The three phases of the melting process are identified as follows: moistening is highlighted in light yellow, ripening in light pink and run-off in light green.

References:
Bartelt, P., & Lehning, M. (2002). A physical SNOWPACK model for the Swiss avalanche warning: Part I: numerical model. Cold Regions Science and Technology, 35(3), 123-145.
Dingman, S. L. (2015). Physical hydrology. Waveland press.
Longepe, N., Allain, S., Ferro-Famil, L., Pottier, E., & Durand, Y. (2009). Snowpack characterization in mountainous regions using C-band SAR data and a meteorological model. IEEE Transactions on Geoscience and Remote Sensing, 47(2), 406-418.
Shi, J., & Dozier, J. (2000). Estimation of snow water equivalence using SIR-C/X-SAR. I. Inferring snow density and subsurface properties. IEEE Transactions on Geoscience and Remote Sensing, 38(6), 2465-2474.

Snow cover in mountain regions is of particular interest in the context of hydrology and water availability, tourism, natural disasters such as floods or avalanches, and it serves as a habitat for both, flora and fauna. Climate change is affecting snow cover worldwide, leading to a decrease of snow cover extent and depth, and shorter snow cover duration. An accurate mapping and monitoring of snow cover is essential when it comes to understanding the current processes especially in mountainous regions, which are often referred to as the water towers of the world. Currently, the most mature monitoring approaches are based on multispectral remote sensing data, as they offer relatively high spatial and temporal resolutions; cloud cover and polar darkness limit the capability to detect snow cover relying only on multispectral data. Therefore, synthetic aperture radar (SAR)-based snow cover detection offers a promising alternative, as SAR data is independent from cloud cover and solar illumination. So far, SAR data have usually only been employed to detect and monitor wet snow, which offers great potential during melting season but lacks the ability to detect the total (wet + dry) snow cover extent during snow cover onset and winter season. Thanks to recent advancements of interferometric and polarimetric SAR, the presented study utilized all available information provided by contemporary SAR sensors (Sentinel-1) to classify both, wet as well as dry snow cover. Geomorphological factors as well as land cover information were added to the proposed machine learning (ML) approach to build a land cover-dependent total snow cover mapping technique, which takes into account vegetation-SAR interactions as well. The robustness and efficiency of ML was applied to map wet and dry snow for the numerous mountainous regions. The resultant snow cover products were comprehensively validated with external optical-based snow cover products as well as local snow depth records, showing overall accuracy and F-measure of over 80%.

Seasonal snow is the main source of freshwater for river basins serving up to 1 billion people [1]. Information on the amount of freshwater stored in snow is crucial for water resource management as well as environmental forecasts and risk analyses (flood forecasts) in these regions. Remote sensing methods provide often the only means to acquire timely information on snow conditions in remote areas such as the Arctic, subarctic and mountains. Passive microwave remote sensing has been used to determine changes in regional and hemispheric Snow Water Equivalent (SWE) [2], but these sensors lack the spatial resolution to resolve SWE over mountains and other regions characterized by heterogeneous topography and land cover. Optical methods like Lidar [3] or photographic methods [4] can provide cm-scale snow depth maps but are limited to daylight, clear weather conditions and often to smaller areas. Active microwave methods based on single-pass interferometric synthetic aperture radar systems (sp-InSAR) can provide topographic information with a spatial resolution on the meter-scale and a vertical resolution on the sub-meter or even centimeter scale. Recently, retrieval of snow depth maps for wet snow cover has been demonstrated using the 35.7 GHz, airborne GLISTINsp-InSAR system [5]. With the same system, compared to airborne LiDAR, a limited penetration of about 0.3 to 1m in dry firn on the Greenland ice sheet was reported [6]. Due to the absence of a radar system in space with sufficiently high frequency to avoid penetration into dry snow, we focus in this study on data from the TanDEM-X satellite mission [7, 8] operating at 9.65 GHz. As the radar penetration into dry snow often exceeds the depth of a seasonal snowpack we do not expect any sensitivity to dry snow and focus therefore on snow depth determination during the snow melt season. Our method is based on differencing two digital elevation models (DEMs) where one was acquired during snow melt and one during snow free conditions. The method has already been demonstrated with TanDEM-X acquisitions over the Swiss Alps [9], but thus far lacks a comprehensive demonstration applying in situ data for validation. It is the purpose of this study to provide this demonstration over a test site in Northern Finland, representing a hilly tundra environment with large local scale variations in snow depth.

[1] T. P. Barnett, J. C. Adam, and D. P. Lettenmaier, “Potential impacts of a warming climate on water availability in snow-dominated regions,” Nature, vol. 438, pp. 303–309, 2005.
[2] M. Takala, K. Luojus, J. Pulliainen, C. Derksen, J. Lemmetyinen, J.-P. Kärnä, J. Koskinen, and B. Bojkov, “Estimating northern hemisphere snow water equivalent for climate research through assimilation of space-borne radiometer data and ground-based measurements,” Remote Sensing of Environment, vol. 115, no. 12, pp. 3517 – 3529, 2011.
[3] J. S. Deems, T. H. Painter, and D. C. Finnegan, “Lidar measurement of snow depth: a review,” Journal of Glaciology, vol. 59, no. 215, pp. 467–479, 2013.
[4] Y. Bühler, M. S. Adams, R. B¨osch, and A. Stoffel, “Mapping snow depth in alpine terrain with unmanned aerial systems (uass): potential and limitations,” The Cryosphere, vol. 10, no. 3, pp. 1075–1088, 2016.
[5] D. Moller, K. M. Andreadis, K. J. Bormann, S. Hensley, and T. H. Painter, “Mapping snow depth from ka-band interferometry: Proof of concept and comparison with scanning lidar retrievals,” IEEE Geoscience and Remote Sensing Letters, vol. 14, no. 6, pp. 886–890, June 2017.
[6] S. Hensley, D. Moller, S. Oveisgharan, T. Michel, and X. Wu, “Ka-band mapping and measurements of interferometric penetration of the greenland ice sheets by the glistin radar,” Sel. Topics Appl. Earth Observations and Rem. Sens., vol. 9, no. 6, pp. 2436–2450, June 2016.
[7] G. Krieger, M. Zink, H. Fiedler, I. Hajnsek, M. Younis, S. Huber, M. Bachmann, J.H. Gonzalez, D. Schulze, J. Boer, M. Werner, and A. Moreira, “The TanDEMX Mission: Overview and status,” in Radar Conference, 2009 IEEE, may 2009, pp. 1 –5.
[8] G. Krieger, M. Zink, M. Bachmann, et al., “TanDEM-X: A radar interferometer with two formation-flying satellites,” Acta Astronautica, vol. 89, pp. 83 – 98, 2013.
[9] S. Leinss, “Depth, anisotropy, and water equivalent of snow estimated by radar interferometry and polarimetry, PhD thesis,” 2015, ETH Zürich.

Globally, seasonal snow cover extends 46 M km2 annually. According to observations by the Swedish meteorological agency, SMHI, in a typical winter, a half to two thirds of Sweden is snow covered for most of the period December to March. Climate change means significant changes in snow accumulation, snow melt and timing are underway. Deviation from long-term norms brings challenges for nature and society, both of which must adapt to changes in ecosystems and the water cycle. For society this means adapting or managing infrastructure such as hydropower facilities (where reservoir capacity planning must adapt to new flow regimes), road and rail infrastructure (ensuring higher river discharges do not damage bridges, embankments or flood carriageways), and maintaining adequate monitoring capacity (of, for example, river discharge, snowpack storage and melt rates).

The European Space Agency (ESA) has identified improving our understanding and quantification of seasonal snow as a Future Challenge for Earth Observation. Snow cover mapping was identified as a key variable at the ESA Arctic and Space workshop in Helsinki in 2016 and has been highlighted as an urgent need by the ESA Arctic Task Force. Declining snow cover is a key element of the World Climate Research Program’s “Melting ice and global consequences” Grand Challenge. Observations of seasonal snow and the processes and feedbacks affecting snow cover are thus identified as a strategic priority for Earth Observation.

Measuring the terrestrial snowpack and the area it covers is non-trivial. At the regional to global scale moderate to low spatial resolution satellite imaging by meteorological satellites and missions such as MODIS, ENVISAT and Sentinel-3 provide daily or weekly observations but at spatial resolutions unsuitable for in-catchment analyses. At higher spatial resolution Sentinel-2 and Landsat-8 combined offer relatively high temporal resolution data at spatial resolutions of <30m: sufficient for catchment-level analyses. However, sensors operating in the solar-reflective spectrum are unable to measure through cloud cover or during the polar night, limiting the temporal resolution of the data, particular in maritime climate zones such as western Scandinavia.

Synthetic Aperture Radar offers an all-weather imaging capability and can capture data independent of solar illumination. SARs typically achieve spatial resolutions <50m allowing catchment level mapping and analysis. Most SAR-based snow mapping algorithms use a threshold to differentiate the lower backscatter from wet snow relative to snow free areas (caused by absorption by liquid water in the snowpack). These approaches do not perform well in dry snow conditions when snow is largely transparent at GHz frequencies.

Here we demonstrate a snow covered area mapping method that uses interferometric coherence along with backscatter intensity rather than backscatter intensity alone. Using Sentinel-1 SAR data we are able to achieve near daily coverage for most of the Scandinavian mountain chain from approximately 63°N to 68°N. The method has proven sensitive to dry snow, though a minimum snow depth is required. We have validated the method using high resolution optical data from Sentinel-2 and Landsat-8 data (though these lack the temporal resolution of the Sentinel-1 data due to frequent cloud-cover). Ultimately, the InSAR coherence approach could be operationally implemented once the new Copernicus data distribution platforms (DIAS) are implemented.

Estimating Snow Water Equivalent (SWE), especially in mountains, is critical for management of water resources, hydropower, and forecasting of floods and avalanches. NASA’s Terrestrial Hydrology Program is supporting a multi-year integrated airborne and ground snow measurement project, called SnowEx, to improve and develop reliable satellite SWE retrieval products for mountain regions, as well as globally. SnowEx’s primary objective is to provide algorithm developments and risk reduction opportunities for SWE remote sensing, leading to a snow satellite mission. One of the driving science questions is: what are the physical controls and dynamics of accumulation and melt of seasonal snow (SWE) across topographic gradients? To tackle this question, SnowEx strives to coordinate the collection of remote sensing observations from airborne platforms with coincident ground-based measurements to quantify SWE and snow physical properties across terrain gradients among different snow environments.
SnowEx activities were postponed until November 2019 due the early 2019 lapse in U.S. government funding. Because a resumed campaign would measure primarily wet-snow conditions and be abbreviated, the winter 2018-2019 campaign was rescheduled for winter 2019-2020. While this delay was unfortunate, a season-long campaign offers an opportunity to meet additional science objectives by capturing snow accumulation, dry snow conditions, and the transition from dry to wet snow during spring.
SnowEx efforts will include focused experiments in the Western U.S. mountains with the operation of airborne radars. The Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) system, an L-band Interferometric Synthetic Aperture Radar (InSAR), will operate twice a month over 13 sites (mainly in California, Colorado, and Idaho). L-band radar phase changes between acquisitions can be related to changes in SWE. Additionally, a campaign on Grand Mesa, Colorado will support the SWE Synthetic Aperture Radar and Radiometer (SWESARR) prototype airborne instrument. SWESARR has three active bands (X, Ku-low, Ku-high) and three passive bands (X, K, Ka), all using a single antenna.
During this presentation, UAVSAR and SWESARR results from past campaigns will be shown. Also, our objectives are to highlight key elements of the future SnowEx radar data collection plan over mountainous terrain, and to seek feedback from the ESA community on collection of relevant observations and ground-based measurements that will enable our community to assess when various sensing strategies succeed and weaken.

Monitoring and predicting snow conditions are important elements for a number of applications including avalanche forecasting, snow hydrology and water resource management. Given the high variability of snow and of its properties over time and space, timely information is essential to achieve the level of precision required by these applications. The purpose of this study is to evaluate the potential of combining the state-of-the-art multi-layer snowpack model Crocus and radiative transfer model MEMLS3&a to simulate backscatter coefficients at C-band over a large alpine region. Simulations are then compared to observations from Sentinel-1 during three seasons (2014-2015, 2015-2016 and 2016-2017). To our knowledge, this is the first study of this extent on the Alps. Apart from evaluating the performances of two models (a radiative transfer model and a snow evolution model), this study is a first step towards the assimilation of Sentinel-1 data in a snowpack model. The evaluation of the direct model is a prerequisite before the assimilation. For situations with snow free and dry snow, simulations distributions were consistent with the observations. In case of wet snow, the simulations are rather in good agreement compared to the observations despite large biases. There is globally a negative bias between simulations and observations during some periods of the snow seasons. The study show that some of this differences increase with the increasing snowpack liquid water content especially during the melting periods of the snow seasons.
An overview of simulations results will be given with a focus on effects of snow liquid content, the topography of the terrain (slope, orientations, …). Cross evaluations with in-situ measurements and with snow products derived from Sentinel-2 will be shown. Limitations and future developments will also be presented.