Performance and sensitivity of a spatially distributed hydrological conceptual flood model with snow components.

<p>In France, flash floods are responsible for a significant proportion of damages caused by natural hazards, either human or material. Hence, advanced modeling tools are needed to perform effective predictions. However for mountainous catchments snow modeling components may be required to correctly simulate river discharge.</p><p>This contribution investigates the implementation and constrain of snow components in the spatially distributed SMASH* platform (Jay-Allemand et al. 2020). The goal is to upgrade model structure and spatially distributed calibration strategies for snow-influenced catchments, as well as to investigate parametric sensitivity and equifinality issues. First, the implementation of snow modules of varying complexity is addressed based on Cemaneige (Valery et al. 2010) in the spatially distributed framework. Next, tests are performed on a sample of 55 catchments in the French North Alps. Numerical experiments and global sensitivity analysis enable to determine pertinent combinations of flow components (including a slow flow one) and calibration parameters. Spatially uniform or distributed calibrations using a variational method (Jay-Allemand 2020) are performed and compared on the dataset, for different model structures and constrains. These tests show critical improvements in outlet discharge modeling by adding slow flow and snow modules, especially considering spatially varying parameters. Current and future works focus on testing and improving the constrains of snow modules and calibration strategy, as well as potential validation and multiobjective calibration with snow signatures gained from in situ or satellite data. </p><p>*SMASH: Spatially-distributed Modelling and ASsimilation for Hydrology, platform developped by INRAE-Hydris corp. for operational applications in the french flood forecast system VigicruesFlash</p>

Spatial and temporal patterns of snowmelt refreezing in a Himalayan catchment

<p>Seasonal snow contributes significantly to the annual runoff in the Himalaya and both timing and volume are important for downstream users.  In  polar regions, meltwater refreezing within snowpacks has been well-studied. While the conditions in the Himalaya are considered favorable for refreezing,<strong> </strong>little is known about refreezing in this region, hindering a complete understanding of seasonal snowmelt dynamics. In this study, we simulated refreezing with the seNorge (v2.0) snow model for the Langtang catchment in the Nepalese Himalaya covering a 5-year period. Thereby, we aim to improve our understanding about how refreezing varies in space and time and to provide a framework for future snow modeling studies. The first part of this study focuses extensively on developing meteorological forcing data, which were derived from an unique elaborate network of meteorological stations and high-resolution meteorological simulations. The snow model was validated against in-situ snow observations and snow cover satellite data. In the second part of this study, we analyze the spatial and temporal refreezing patterns, and attempt to identify possible driving factors. The results show that the annual average refreezing amounts to 122 mm (21% of the total melt). We found that the magnitude of refreezing varies strongly in space depending on elevation and aspect. In addition, there is a strong seasonal altitudinal variability related to air temperature and snow depth, with most refreezing during the early melt season. We also found a substantial intra-annual variability, which mainly results from fluctuations of snowfall, highlighting the importance of using multi-year time series in refreezing assessments. Daily refreezing simulations decreased by 84% (to an average 19 mm year<sup>-1</sup>) compared to hourly simulations, emphasizing the importance of using sub-daily time steps to capture diurnal melt-refreeze cycles. Climate sensitivity experiments revealed that refreezing is highly sensitive to future changes in air temperature, as a temperature increase of 2°C leads to a refreezing decrease of 35%. We conclude that including refreezing with a sub-daily temporal resolution is highly relevant for understanding snow dynamics in the current and future climate of the Himalaya.</p><p> </p>

Assimilation of citizen science data in snowpack modeling using a new snow dataset: Community Snow Observations

In this study, we examine the effectiveness of incorporating citizen science snow depth measurements into the seasonal snow model chain through data assimilation. We also introduce the Community Snow Observations dataset, a citizen science based snow depth measuring campaign. Improvements to model performance are characterized and evaluated using remote sensing datasets, fieldwork, and SNOTEL datasets. After citizen science snow depth measurements were incorporated, improvements to model performance were found in 62 % to 78 % of the simulations, depending on model year. The results suggest that modest measurements from citizen scientists have the potential to improve efforts to model snowpack processes in high mountain environments, with implications for water resource management and process-based snow modeling.

Use of Sentinel-1 radar observations to evaluate snowmelt dynamics in alpine regions

Knowing the timing and the evolution of the snow melting process is very important, since it allows the prediction of (i) the snowmelt onset, (ii) the snow gliding and wet-snow avalanches, (iii) the release of snow contaminants, and (iv) the runoff onset. The snowmelt can be monitored by jointly measuring snowpack parameters such as the snow water equivalent (SWE) or the amount of free liquid water content (LWC). However, continuous measurements of SWE and LWC are rare and difficult to obtain. On the other hand, active microwave sensors such as the synthetic aperture radar (SAR) mounted on board satellites are highly sensitive to LWC of the snowpack and can provide spatially distributed information with a high resolution. Moreover, with the introduction of Sentinel-1, SAR images are regularly acquired every 6 d over several places in the world. In this paper we analyze the correlation between the multitemporal SAR backscattering and the snowmelt dynamics. We compared Sentinel-1 backscattering with snow properties derived from in situ observations and process-based snow modeling simulations for five alpine test sites in Italy, Germany and Switzerland considering 2 hydrological years. We found that the multitemporal SAR measurements allow the identification of the three melting phases that characterize the melting process, i.e., moistening, ripening and runoff. In particular, we found that the C-band SAR backscattering decreases as soon as the snow starts containing water and that the backscattering increases as soon as SWE starts decreasing, which corresponds to the release of meltwater from the snowpack. We discuss the possible reasons of this increase, which are not directly correlated to the SWE decrease but to the different snow conditions, which change the backscattering mechanisms. Finally, we show a spatially distributed application of the identification of the runoff onset from SAR images for a mountain catchment, i.e., the Zugspitze catchment in Germany. Results allow us to better understand the spatial and temporal evolution of melting dynamics in mountain regions. The presented investigation could have relevant applications for monitoring and predicting the snowmelt progress over large regions.

Use of Sentinel-1 radar observations to evaluate snowmelt dynamics in alpine regions

Knowing the timing and the evolution of the snow melting process is very important, since it allows the prediction of: i) the snow melt onset; ii) the snow gliding and wet-snow avalanches; iii) the release of snow contaminants and iv) the runoff onset. The snowmelt can be monitored by jointly measuring snowpack parameters such as the snow water equivalent (SWE) or the amount of free liquid water content (LWC). However, continuous measurements of SWE and LWC are rare and difficult to be obtained. On the other hand, active microwave sensors such as the Synthetic Aperture Radar (SAR) mounted on board of satellites, are highly sensitive to LWC of the snowpack and can provide spatially distributed information with a high resolution. Moreover, with the introduction of Sentinel-1, SAR images are regularly acquired every 6 days over several places in the world. In this paper we analyze the correlation between the multi-temporal SAR backscattering and the snowmelt dynamics. We compared Sentinel-1 backscattering with snow properties derived from in situ observations and process-based snow modeling simulations for five alpine test sites in Italy, Germany and Switzerland considering two hydrological years. We found that the multi-temporal SAR measurements allow the identification of the three melting phases that characterize the melting process i.e., moistening, ripening and runoff. In detail, we found that the C-band SAR backscattering decreases as soon as the snow starts containing water, and that the backscattering increases as soon as SWE starts decreasing, which corresponds to the release of meltwater from the snowpack. We discuss the possible reasons of this increase, which are not directly correlated to the SWE decrease, but to the different snow conditions, which change the backscattering mechanisms. Finally, we show a spatially-distributed application of the identification of the runoff onset from SAR images for a mountain catchment, i.e., the Zugspitze catchment in Germany. Results allow to better understand the spatial and temporal evolution of melting dynamics in mountain regions. The presented investigation could have relevant applications for monitoring and predicting the snowmelt progress over large regions.

Snow physical properties may be a significant determinant of lemming population dynamics in the high Arctic

Cyclic population fluctuations are common in boreal and Arctic species but the causes of these cycles are still debated today. Among these species, lemmings are Arctic rodents that live and reproduce under the snow and whose large cyclical population fluctuations in the high Arctic impact the whole tundra food web. We explore, using lemming population data and snow modeling, whether the hardness of the basal layer of the snowpack, determined by rain-on-snow events (ROS) and wind storms in autumn, can affect brown lemming population dynamics in the Canadian high Arctic. Using a 7-year dataset collected on Bylot Island, Nunavut, Canada over the period 2003–2014, we demonstrate that liquid water input to snow is strongly inversely related with winter population growth (R2 ≥ 0.62) and to a lesser extent to lemming summer densities and winter nest densities (R2 = 0.29–0.39). ROS in autumn can therefore influence the amplitude of brown lemming population fluctuations. Increase in ROS events with climate warming should strongly impact the populations of lemmings and consequently those of the many predators that depend upon them. Snow conditions may be a key factor influencing the cyclic dynamics of Arctic animal populations.

Meteorological Forcing Datasets for Blowing Snow Modeling on the Tibetan Plateau: Evaluation and Intercomparison

In this paper, the reliability of the wind speed, temperature, humidity, pressure, and precipitation values of three surface meteorological forcing products [China Meteorological Administration Land Data Assimilation System, version 2 (CLDAS-2); China Meteorological Forcing Dataset (CMFD); and Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2)] in the Tibetan Plateau (TP) region was investigated from 2008 to 2014. Compared with the China Meteorological Administration (CMA) observations, CLDAS-2 exhibited the highest correlation coefficient for wind speed, CMFD displayed the best coefficients for temperature and specific humidity, and MERRA-2 best reflected pressure variations. Based on the biases, CLDAS-2 exhibited the best overall performance for temperature, specific humidity, and pressure, while CMFD displayed the best performance for wind speed. The high overall accuracy and false alarm ratio of precipitation based on MERRA-2 both stem from its continuous overestimation of the precipitation frequency. Both CLDAS-2 and CMFD overestimated the nonprecipitation frequency in comparisons with CMA observations, and a significant positive bias exists in MERRA-2 based on the analysis of daily precipitation. The results obtained from the comparisons with field observations over the TP and CMA observations are similar, except for the temperature and humidity biases of CLDAS-2. The meteorological effects on the coupled land–blowing snow modeling discussed in this paper suggest that the occurrence of blowing snow and snowdrift sublimation are projected to be reduced by CLDAS-2 due to the underestimation of wind speed, continual lack of snowfall events, and the positive biases in low temperatures and humidity, while simulations of blowing processes by MERRA-2 are likely to be much more severe than they actually are. These results may contribute to identifying deficiencies associated with the development of land surface models coupled with a blowing snow model.