• Alavi, N., S. Belair, V. Fortin, S. Zhang, S. Z. Husain, M. L. Carrera, and M. Abrahamowicz, 2016: Warm season evaluation of soil moisture prediction in the Soil, Vegetation, and Snow (SVS) scheme. J. Hydrometeor., 17, 23152332, https://doi.org/10.1175/JHM-D-15-0189.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Balsamo, G., J.-F. Mahfouf, S. Bélair, and G. Deblonde, 2006: A global root-zone soil moisture analysis using simulated L-band brightness temperature in preparation for the Hydros satellite mission. J. Hydrometeor., 7, 11261146, https://doi.org/10.1175/JHM525.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Barnett, T., L. Dumenil, U. Schlese, E. Roeckner, and M. Latif, 1989: The effect of Eurasian snow cover on regional and global climate variations. J. Atmos. Sci., 46, 661685, https://doi.org/10.1175/1520-0469(1989)046<0661:TEOESC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bélair, S., L.-P. Crevier, J. Mailhot, B. Bilodeau, and Y. Delage, 2003a: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part I: Warm season results. J. Hydrometeor., 4, 352370, https://doi.org/10.1175/1525-7541(2003)4<352:OIOTIL>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bélair, S., R. Brown, J. Mailhot, B. Bilodeau, and L.-P. Crevier, 2003b: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part II: Cold season results. J. Hydrometeor., 4, 371386, https://doi.org/10.1175/1525-7541(2003)4<371:OIOTIL>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bernier, N., S. Bélair, B. Bilodeau, and L. Tong, 2014: Assimilation and high resolution forecasts of surface and near surface conditions for the 2010 Vancouver Winter Olympic and Paralympic Games. Pure Appl. Geophys., 171, 243256, https://doi.org/10.1007/s00024-012-0542-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonan, G., 2008: Ecological Climatology, Concepts and Applications. 2nd ed. Cambridge University Press, 563 pp.

    • Crossref
    • Export Citation
  • Boone, A., and P. Etchevers, 2001: An intercomparison of three snow schemes of varying complexity coupled to the same land-surface model: Local scale evaluation at an alpine site. J. Hydrometeor., 2, 374394, https://doi.org/10.1175/1525-7541(2001)002<0374:AIOTSS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Brasnett, B., 1999: A global analysis of snow depth for numerical weather prediction. J. Appl. Meteor., 38, 726740, https://doi.org/10.1175/1520-0450(1999)038<0726:AGAOSD>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Brown, R., and D. Tapsoba, 2007: Improved mapping of snow water equivalent over Quebec. 64th Eastern Snow Conf., St John’s, NL, Canada, Eastern Snow Conference, 17.

  • Brown, R., C. Derksen, and L. Wang, 2010: A multi-data set analysis of variability and change in the Arctic spring snow cover extent, 1967–2008. J. Geophys. Res., 115, D16111, https://doi.org/10.1029/2010JD013975.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Burgers, G., P. van Leeuwen, and G. Evensen, 1998: Analysis scheme in the ensemble Kalman filter. Mon. Wea. Rev., 126, 17191724, https://doi.org/10.1175/1520-0493(1998)126<1719:ASITEK>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carrera, M., S. Bélair, V. Fortin, B. Bilodeau, D. Charpentier, and I. Doré, 2010: Evaluation of snowpack simulations over the Canadian Rockies with an experimental hydrometeorological modeling system. J. Hydrometeor., 11, 11231140, https://doi.org/10.1175/2010JHM1274.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carrera, M., S. Bélair, and B. Bilodeau, 2015: The Canadian Land Data Assimilation System (CALDAS): Description and synthetic evaluation study. J. Hydrometeor., 16, 12931314, https://doi.org/10.1175/JHM-D-14-0089.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cohen, J., and D. Entekhabi, 1999: Eurasian snow cover variability and Northern Hemisphere climate predictability. Geophys. Res. Lett., 26, 345348, https://doi.org/10.1029/1998GL900321.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cohen, J., and K. Saito, 2001: Eurasian snow cover, more skillful in predicting U.S. winter climate than the NAO/AO? Geophys. Res. Lett., 28, 299302, https://doi.org/10.1029/2000GL011927.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cohen, J., K. Saito, and D. Entekhabi, 2001: The role of the Siberian high in Northern Hemisphere climate variability. Geophys. Res. Lett., 28, 299302, https://doi.org/10.1029/2000GL011927.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Crow, W., and E. Wood, 2003: The assimilation of remotely sensed soil brightness temperature imagery into a land surface model using ensemble Kalman filtering: A case study based on ESTAR measurements during SGP97. Adv. Water Resour., 26, 137149, https://doi.org/10.1016/S0309-1708(02)00088-X.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Crow, W., M. Drusch, and E. Wood, 2001: An observation system simulation experiment for the impact of land surface heterogeneity on AMSR-E soil moisture retrieval. IEEE Trans. Geosci. Remote Sens., 39, 16221631, https://doi.org/10.1109/36.942540.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Crow, W., and et al. , 2005: An observing system simulation experiment for Hydros radiometer-only soil moisture products. IEEE Trans. Geosci. Remote Sens., 43, 12891303, https://doi.org/10.1109/TGRS.2005.845645.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Das, N., 2013: Soil Moisture Active Passive (SMAP) ancillary data report: Soil attributes. SMAP Doc. 044, JPL D-53058, 16 pp.

  • Deardorff, J., 1978: Efficient prediction of ground surface temperature and moisture with inclusion of a layer of vegetation. J. Geophys. Res., 83, 18891903, https://doi.org/10.1029/JC083iC04p01889.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Decharme, B., E. Brun, A. Boone, C. Delire, P. L. Moigne, and S. Morin, 2016: Impacts of snow and organic soils parameterization on north-Eurasian soil temperature profiles simulated by the ISBA land surface model. Cryosphere, 10, 853877, https://doi.org/10.5194/tc-10-853-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • DeLannoy, G., R. Reichle, P. Houser, K. Arsenault, N. Verhoest, and V. Pauwels, 2010: Satellite-scale snow water equivalent assimilation into a high-resolution land surface model. J. Hydrometeor., 11, 352369, https://doi.org/10.1175/2009JHM1192.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Derksen, C., A. Walker, E. LeDrew, and B. Goodison, 2003: Combining SMMR and SMM/I data for time series analysis of central North American snow water equivalent. J. Hydrometeor., 4, 304316, https://doi.org/10.1175/1525-7541(2003)4<304:CSAIDF>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • de Rosnay, P., G. Balsamo, C. Albergel, J. Munos-Sabater, and L. Isaksen, 2014: Initialisation of land surface variables for numerical weather prediction. Surv. Geophys., 35, 607621, https://doi.org/10.1007/s10712-012-9207-x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • de Rosnay, P., L. Isaksen, and M. Dahoui, 2015: Snow data assimilation at ECMWF. ECMWF Newsletter, No. 143, ECMWF, Reading, United Kingdom, 26–31, https://doi.org/10.21957/lkpxq6x5.

    • Crossref
    • Export Citation
  • Douville, H., and J.-F. Royer, 1996: Sensitivity of the Asian summer monsoon to an anomalous Eurasian snow cover within the Meteo-France GCM. Climate Dyn., 12, 449466, https://doi.org/10.1007/BF02346818.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douville, H., J.-F. Royer, and J.-F. Mahfouf, 1995: A new snow parameterization for the Météo-France climate model. Climate Dyn., 12, 3752, https://doi.org/10.1007/BF00208761.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., and et al. , 2005: NOAA operational hydrological products derived from the Advanced Microwave Sounding Unit. IEEE Trans. Geosci. Remote Sens., 43, 10361049, https://doi.org/10.1109/TGRS.2004.843249.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fletcher, S., G. L. Iston, C. Hiemstra, and S. Miller, 2012: Assimilating MODIS and AMSR-E snow observations in a snow evolution model. J. Hydrometeor., 13, 14751492, https://doi.org/10.1175/JHM-D-11-082.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fortin, V., G. Roy, N. Donaldson, and A. Mahidjiba, 2015: Assimilation of radar quantitative precipitation estimations in the Canadian Precipitation Analysis (CaPA). J. Hydrol., 531, 296307, https://doi.org/10.1016/j.jhydrol.2015.08.003.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Foster, J., M. Owe, and A. Rango, 1983: Snow cover and temperature relationships in North America and Eurasia. J. Climate Appl. Meteor., 22, 460469, https://doi.org/10.1175/1520-0450(1983)022<0460:SCATRI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hollmann, R., and et al. , 2013: The ESA Climate Change Initiative: Satellite data records for essential climate variables. Bull. Amer. Meteor. Soc., 94, 15411552, https://doi.org/10.1175/BAMS-D-11-00254.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hu, Z., and S. Islam, 1995: Prediction of ground surface temperature and soil moisture content by the force-restore method. Water Resour. Res., 31, 25312539, https://doi.org/10.1029/95WR01650.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Husain, S. Z., N. Alavi, S. Belair, M. Carrera, S. Zhang, M. A. V. Fortin, and N. Gauthier, 2016: The multi-budget Soil, Vegetation, and Snow (SVS) scheme for land surface parameterization: Offline warm season evaluation. J. Hydrometeor., 17, 22932313, https://doi.org/10.1175/JHM-D-15-0228.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kelly, R., 2009: The AMSR-E Snow Depth Algorithm: Description and initial results. J. Remote Sens. Soc. Japan, 29, 307317, https://doi.org/10.11440/rssj.29.307.

    • Search Google Scholar
    • Export Citation
  • Kumar, S., R. Reichle, D. Peters-Lidard, R. Koster, X. Zhan, W. Crow, J. Eylander, and P. Houser, 2008: A land surface data assimilation framework using the land information system: Description and applications. Adv. Water Resour., 31, 14191432, https://doi.org/10.1016/j.advwatres.2008.01.013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kumar, S., J. Dong, C. Peters-Lidard, D. Mocko, and B. Gomez, 2017: Role of forcing uncertainty and model error background characterization in snow data assimilation. Hydrol. Earth Syst. Sci., 21, 26372647, https://doi.org/10.5194/hess-21-2637-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Larue, F., A. Royer, D. D. Sève, A. Langlois, A. Roy, and L. Brucker, 2017: Validation of GlobSnow-2 snow water equivalent over eastern Canada. Remote Sens. Environ., 194, 264277, https://doi.org/10.1016/j.rse.2017.03.027.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Leathers, D. J., A. W. Ellis, and D. Robinson, 1995: Characteristics of temperature depressions associated with snow cover across the northeast United States. J. Appl. Meteor., 34, 381390, https://doi.org/10.1175/1520-0450-34.2.381.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lespinas, F., V. Fortin, G. Roy, P. Rasmussen, and T. Stadnyk, 2015: Performance Evaluation of the Canadian Precipitation Analysis (CaPA). J. Hydrometeor., 16, 20452064, https://doi.org/10.1175/JHM-D-14-0191.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Luojus, K., and et al. , 2010: Investigating the feasability of the GlobSnow snow water equivalent data for climate research purposes. 2010 IEEE Int. Geoscience and Remote Sensing Symp., Honolulu, HI, IEEE, 4851–4853, https://doi.org/10.1109/IGARSS.2010.5741987.

    • Crossref
    • Export Citation
  • Ma, Z., L. Riishojgaard, M. Masutani, J. Woollen, and G. Emmitt, 2015: Impact of different satellite wind lidar telescope configurations on NCEP GFS forecast skill in observing system simulation experiments. J. Atmos. Oceanic Technol., 32, 478495, https://doi.org/10.1175/JTECH-D-14-00057.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mahfouf, J.-F., B. Brasnett, and S. Gagnon, 2007: A Canadian Precipitation Analysis (CaPA) project: Description and preliminary results. Atmos.–Ocean, 45, 117, https://doi.org/10.3137/ao.v450101.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mailhot, J., and et al. , 2006: The 15-km version of the Canadian regional forecast system. Atmos.–Ocean, 44, 133149, https://doi.org/10.3137/ao.440202.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Masutani, M., and et al. , 2010: Observation System Simulation Experiments. Data Assimilation, W. A. Lahoz, B. Khattatov and R. Ménard, Eds., Springer, 647–677.

  • Milbrandt, J., S. Bélair, M. Faucher, M. Vallée, M. Carrera, and A. Glazer, 2016: The Pan-Canadian High Resolution (2.5 km) Deterministic Prediction System. Wea. Forecasting, 31, 17911816, https://doi.org/10.1175/WAF-D-16-0035.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Noilhan, J., and S. Planton, 1989: A simple parameterization of land surface processes for meteorological models. Mon. Wea. Rev., 117, 536549, https://doi.org/10.1175/1520-0493(1989)117<0536:ASPOLS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pullen, S., C. Jones, and G. Rooney, 2011: Using satellite-derived snow cover data to implement a snow analysis in the Met Office Global NWP Model. J. Appl. Meteor. Climatol., 50, 958973, https://doi.org/10.1175/2010JAMC2527.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pulliainen, J., J. Grandell, and M. Hallikainen, 1999: HUT snow emission model and its applicability for snow water equivalent retrieval. IEEE Trans. Geosci. Remote Sens., 37, 13781390, https://doi.org/10.1109/36.763302.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reichle, R., D. McLaughlin, and D. Entekhabi, 2002: Hydrologic data assimilation with the ensemble Kalman filter. Mon. Wea. Rev., 130, 103114, https://doi.org/10.1175/1520-0493(2002)130<0103:HDAWTE>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rott, H., and et al. , 2010: Cold regions hydrology high resolution observatory for snow and cold land processes. Proc. IEEE, 98, 752765, https://doi.org/10.1109/JPROC.2009.2038947.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schleef, S., H. Lowe, and M. Schneebeli, 2014: Influence of stress, temperature and crystal morphology on isothermal densification and specific surface area decrease of new snow. Cryosphere, 8, 18251838, https://doi.org/10.5194/tc-8-1825-2014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shangguan, W., Y. Dai, Q. Duan, B. Liu, and H. Yuan, 2014: A global soil data set for earth system modeling. J. Adv. Model. Earth Syst., 6, 249263, https://doi.org/10.1002/2013MS000293.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shortridge, A., and J. Messina, 2011: Spatial structure and landscape association of SRTM error. Remote Sens. Environ., 115, 15761587, https://doi.org/10.1016/j.rse.2011.02.017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sicart, J., J. Pomeroy, R. Essery, J. Hardy, T. Link, and D. Marks, 2004: A sensitivity study of daytime net radiation during snowmelt to forest canopy and atmospheric conditions. J. Hydrometeor., 5, 774784, https://doi.org/10.1175/1525-7541(2004)005<0774:ASSODN>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Slater, A., and M. Clark, 2006: Snow data assimilation via an ensemble Kalman filter. J. Hydrometeor., 7, 478493, https://doi.org/10.1175/JHM505.1.

  • Takala, M., K. Luojus, J. Pulliainen, C. Derksen, J. Lemmetyinen, J.-P. Karna, 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, 35173529, https://doi.org/10.1016/j.rse.2011.08.014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Vernekar, A., J. Zhou, and J. Shukla, 1995: The effect of Eurasian snow cover on the Indian monsoon. J. Climate, 8, 248266, https://doi.org/10.1175/1520-0442(1995)008<0248:TEOESC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Walland, D. J., and I. Simmonds, 1996: Modelled atmospheric response to changes in Northern Hemisphere snow cover. Climate Dyn., 13, 2534, https://doi.org/10.1007/s003820050150.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wang, H., X.-Y. Huang, and Y. Chen, 2013: An Observing System Simulation Experiment for the impact of MTG candidate infrared sounding mission on regional forecasts: System development and preliminary results. ISRN Meteor., 2013, 971501, https://doi.org/10.1155/2013/971501.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wang, X., D. Barker, C. Snyder, and T. Hamill, 2008: A hybrid ETKF-3DVAR data assimilation scheme for the WRF Model. Part I: Observing System Simulation Experiment. Mon. Wea. Rev., 136, 51165131, https://doi.org/10.1175/2008MWR2444.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xu, L., and P. Dirmeyer, 2011: Snow-atmosphere coupling strength in a global atmospheric model. Geophys. Res. Lett., 38, L13401, https://doi.org/10.1029/2011GL048049.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yang, D., D. Robinson, Y. Zhao, T. Estilow, and H. Ye, 2003: Streamflow response to seasonal snow cover extent changes in large Siberian watersheds. J. Geophys. Res., 108, 45784591, https://doi.org/10.1029/2002JD003149.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • You, J., D. Tarboton, and C. Luce, 2014: Modeling the snow surface temperature with a one-layer energy balance snowmelt model. Hydrol. Earth Syst. Sci., 18, 50615076, https://doi.org/10.5194/hess-18-5061-2014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yueh, S., S. Dinardo, A. Akgiray, R. West, D. Cline, and K. Elder, 2009: Airborne Ku-band polarimetric radar remote sensing of terrestrial snow cover. IEEE Trans. Geosci. Remote Sens., 47, 33473364, https://doi.org/10.1109/TGRS.2009.2022945.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zaitchik, B., and M. Rodell, 2009: Forward-looking assimilation of MODIS-derived snow-covered area into a land surface model. J. Hydrometeor., 10, 130148, https://doi.org/10.1175/2008JHM1042.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) Location of the domain in southern Quebec, Canada, (b) its orography (m) at a 100-m resolution, and (c) dominant cover type in each grid cell of the study domain.

  • View in gallery

    The different technical steps of the OSSE. (a) Generation of a Synthetic Truth state at 100-m resolution, with an example of an SWE product (mm) on 15 Mar 2015. (b) Production of synthetic observations through upscaling and addition of error. The figure shows the synthetic observation of SWE (mm) corresponding to the Synthetic Truth on 15 Mar 2015 after upscaling to 2000-m resolution and addition of error according to the threshold accuracy requirements. (c) Assimilation of different synthetic observations (black circles) at different times k within CaLDAS through the use of the EnKF and production of a control experiment (no assimilation). (d) Skill evaluation of the control experiment and the different assimilation experiments by comparing them to the Synthetic Truth state.

  • View in gallery

    Experimental setup showing the different components of the OSSE: geophysical fields, atmospheric forcings, and land surface models.

  • View in gallery

    Domain-averaged SWE (mm) over the 1 Nov 2014–24 Mar 2015 study period in Synthetic Truth (black dashed line) and Open Loop (solid black line) simulations as well the eight CaLDAS simulations (colored lines) separated so as to highlight the effects of (a) resolution, (b) revisit, and (c) accuracy. Please refer to Table 2 for the nomenclature of the different simulations.

  • View in gallery

    Solid (cm) and liquid (mm) precipitation accumulation from 1 Nov 2014 to 31 May 2015 for the 6–12-h CaPA (HRDPS) forecasts (orange) and 30–36-h RDPS forecasts (blue).

  • View in gallery

    Spatial RMSE (mm) over the domain over the 1 Nov 2014–24 Mar 2015 study period when compared to Synthetic Truth SWE values separated so as to highlight the effects of (a) resolution, (b) revisit, and (c) accuracy. The thick blue line represents the 1000m_5d_T simulation, common to all panels. Please refer to Table 2 for the nomenclature of the different simulations.

  • View in gallery

    Temporal RMSE (mm) when compared to Synthetic Truth SWE values calculated at each grid point over the 1 Nov 2014–24 Mar 2015 study period. Please note that the Open Loop has its own color scale due to the large difference in magnitude of the Open Loop RMSE compared to the other simulations. Please refer to Table 2 for the nomenclature of the different simulations.

  • View in gallery

    Domain-averaged SWE (mm) over the 1 Nov 2014–31 May 2015 study period. Please refer to Table 2 for the nomenclature of the different simulations.

  • View in gallery

    Spatial RMSE (mm) over the domain over the 1 Nov 2014–31 May 2015 study period when compared to Synthetic Truth SWE values. Please refer to Table 2 for the nomenclature of the different simulations.

  • View in gallery

    As in Fig. 7, but for the 1 Nov 2014–31 May 2015 period.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 265 266 25
PDF Downloads 147 147 11

Quantifying Snow Mass Mission Concept Trade-Offs Using an Observing System Simulation Experiment

View More View Less
  • 1 Meteorological Research Division, Environment and Climate Change Canada, Dorval, Quebec, Canada
  • | 2 Climate Research Division, Environment and Climate Change Canada, Downsview, Ontario, Canada
  • | 3 Meteorological Research Division, Environment and Climate Change Canada, Dorval, Quebec, Canada
  • | 4 Meteorological Service of Canada, Environment and Climate Change Canada, Dorval, Quebec, Canada
  • | 5 Météo France/CNRS, CNRM, UMR 3589, CEN, Grenoble, France, and Centre for Hydrology, University of Saskatchewan, Saskatoon, Canada
© Get Permissions
Open access

Abstract

Because of its location, Canada is particularly affected by snow processes and their impact on the atmosphere and hydrosphere. Yet, snow mass observations that are ongoing, global, frequent (1–5 days), and at high enough spatial resolution (kilometer scale) for assimilation within operational prediction systems are presently not available. Recently, Environment and Climate Change Canada (ECCC) partnered with the Canadian Space Agency (CSA) to initiate a radar-focused snow mission concept study to define spaceborne technological solutions to this observational gap. In this context, an Observing System Simulation Experiment (OSSE) was performed to determine the impact of sensor configuration, snow water equivalent (SWE) retrieval performance, and snow wet/dry state on snow analyses from the Canadian Land Data Assimilation System (CaLDAS). The synthetic experiment shows that snow analyses are strongly sensitive to revisit frequency since more frequent assimilation leads to a more constrained land surface model. The greatest reduction in spatial (temporal) bias is from a 1-day revisit frequency with a 91% (93%) improvement. Temporal standard deviation of the error (STDE) is mostly reduced by a greater retrieval accuracy with a 65% improvement, while a 1-day revisit reduces the temporal STDE by 66%. The inability to detect SWE under wet snow conditions is particularly impactful during the spring meltdown, with an increase in spatial RMSE of up to 50 mm. Wet snow does not affect the domain-wide annual maximum SWE nor the timing of end-of-season snowmelt timing in this case, indicating that radar measurements, although uncertain during melting events, are very useful in adding skill to snow analyses.

Denotes content that is immediately available upon publication as open access.

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Camille Garnaud, camille.garnaud@canada.ca

Abstract

Because of its location, Canada is particularly affected by snow processes and their impact on the atmosphere and hydrosphere. Yet, snow mass observations that are ongoing, global, frequent (1–5 days), and at high enough spatial resolution (kilometer scale) for assimilation within operational prediction systems are presently not available. Recently, Environment and Climate Change Canada (ECCC) partnered with the Canadian Space Agency (CSA) to initiate a radar-focused snow mission concept study to define spaceborne technological solutions to this observational gap. In this context, an Observing System Simulation Experiment (OSSE) was performed to determine the impact of sensor configuration, snow water equivalent (SWE) retrieval performance, and snow wet/dry state on snow analyses from the Canadian Land Data Assimilation System (CaLDAS). The synthetic experiment shows that snow analyses are strongly sensitive to revisit frequency since more frequent assimilation leads to a more constrained land surface model. The greatest reduction in spatial (temporal) bias is from a 1-day revisit frequency with a 91% (93%) improvement. Temporal standard deviation of the error (STDE) is mostly reduced by a greater retrieval accuracy with a 65% improvement, while a 1-day revisit reduces the temporal STDE by 66%. The inability to detect SWE under wet snow conditions is particularly impactful during the spring meltdown, with an increase in spatial RMSE of up to 50 mm. Wet snow does not affect the domain-wide annual maximum SWE nor the timing of end-of-season snowmelt timing in this case, indicating that radar measurements, although uncertain during melting events, are very useful in adding skill to snow analyses.

Denotes content that is immediately available upon publication as open access.

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Camille Garnaud, camille.garnaud@canada.ca

1. Introduction

Snow properties are known to be essential controls on energy and water exchanges between the land surface and the atmosphere over a range of scales (e.g., Douville and Royer 1996; Cohen and Entekhabi 1999; Bonan 2008; Xu and Dirmeyer 2011), particularly in boreal and mountainous regions. Snow cover has been shown to lower near-surface air temperatures (Foster et al. 1983; Leathers et al. 1995), resulting in variations in large-scale atmospheric circulations (Walland and Simmonds 1996; Cohen et al. 2001), including monsoon climates (Barnett et al. 1989; Vernekar et al. 1995) and the North Atlantic Oscillation (Cohen and Saito 2001). Furthermore, with snow’s ability to store water, snowmelt processes are essential to hydrological applications since they control the timing and amount of snowmelt runoff and ultimately streamflow (Yang et al. 2003).

Because of its geographical location, Canada is particularly affected by snow processes and their impact on the atmosphere and hydrosphere. Snow mass observations are thus crucial for quality environmental and hydrological forecasts. In eastern Canada, for example, 1 mm of snow water equivalent (SWE) over the James Bay territory could amount to CAN $1 million in hydroelectric power production (Brown and Tapsoba 2007).

Given trends in the evolution of operational environmental prediction systems, it is expected that satellite observations sensitive to snow mass at the scale of a few hundred meters to a couple of kilometers with a revisit time of 1–5 days are required for quality environmental prediction. In spite of that, snow mass observations that are ongoing, global, frequent (1–5 days), and at high enough spatial resolution (kilometer scale) are presently not available for assimilation within operational prediction systems at Environment and Climate Change Canada (ECCC) and other numerical weather and hydrological prediction centers. Although essential for some purposes, in situ data tend to be too sparse. Remote sensing is therefore key for numerical weather prediction and hydrological forecasting.

Several initiatives have provided the community with satellite-derived snow mass products, though none have met ECCC’s requirements (i.e., kilometer-scale resolution, 1–5-day revisit, global and ongoing coverage, well-constrained uncertainty budgets) as of yet. The National Oceanic and Atmospheric Administration (NOAA) Polar Operational Environmental Satellites (POES) currently provides global 16-km resolution SWE retrievals as part of the operational Microwave Surface and Precipitation Products System (MSPPS; Ferraro et al. 2005), although no product evaluation has been published. An SWE product was produced for the National Aeronautics and Space Administration (NASA) Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) mission (Kelly 2009), but poorly constrained uncertainty proved a challenge for data assimilation (DeLannoy et al. 2010). The European Space Agency’s GlobSnow product combines passive microwave brightness temperatures with in situ observations, generating an SWE product at a 25-km resolution (Luojus et al. 2010; Takala et al. 2011). Larue et al. (2017) have noted limitations in this SWE product’s accuracy, with an overall relative percentage error that can reach up to 35.9% over Quebec in eastern Canada due to challenges posed by deep snow, forest vegetation, and wet snow. The GlobSnow product also masks out alpine regions, making it unsuitable for global applications.

At the research level, tests were made at ECCC to evaluate the relative impact of assimilating AMSR-E SWE retrievals on the snow analyses of the model first guess. These results (unpublished) showed that the assimilation of AMSR-E retrievals did not improve snow analyses when compared to the Open Loop (i.e., no assimilation) data, and even degraded them in some cases. This is likely due to a high degree of random error in the retrievals, exacerbated by the coarse resolution of AMSR-E measurements (Fletcher et al. 2012). Thus, no satellite snow products are currently used operationally at ECCC’s Meteorological Service of Canada (MSC). Instead, surface snow depth observations from synoptic weather stations (SYNOP) are assimilated based on an optimal interpolation technique (Brasnett 1999). Brown et al. (2010) showed that for this approach the positioning of observing sites in clearings and preferentially at lower elevations lead to shallow biases in reported snow depths, causing early loss of snow cover in spring in the MSC analyses.

No other numerical weather prediction (NWP) centers are operationally assimilating space-based remote sensing products related to SWE. At Météo France, snow analysis is part of ongoing developments and future plans with a two-dimensional optimal interpolation (2D-OI) using in situ snow depth observations, as well as satellite products of snow cover extent. The European Center for Medium-Range Weather Forecasts (ECMWF) has been using this combination of snow observations for a few years in its Land Data Assimilation System (LDAS; de Rosnay et al. 2014, 2015). They use a 2D-OI method with in situ snow depth observations from SYNOP and national networks from up to seven countries, as well the NOAA/National Environmental Satellite, Data, and Information Service (NESDIS) Interactive Multisensor Snow and Ice Mapping System (IMS) snow cover extent 4-km daily product. ECMWF thus produces analyses of SWE and snow density but without the assimilation of satellite-derived snow mass information. The United Kingdom’s Met Office uses IMS snow cover extent in a simple update scheme to adjust their global model snow amount in a daily analysis (Pullen et al. 2011), although they do not assimilate information on SWE.

NWP centers are aware of the limitations imposed by the satellite-derived SWE data, and so there are currently several new concepts in discussion to address this gap. One study is a partnership between ECCC and the Canadian Space Agency (CSA), who have initiated a snow mission concept study to define spaceborne technological solutions for SWE. At the moment, the mission is Ku-band radar focused since it allows sensitivity to snow mass combined with a wide imaging swath (and hence rapid revisit time) and moderate spatial resolution (~250 m) and builds on previous mission concepts (Rott et al. 2010; Yueh et al. 2009).

In this context, an Observing System Simulation Experiment (OSSE; Masutani et al. 2010) can estimate the potential impact, and therefore the value, of hypothetical new observations because an OSSE is designed to mimic the process of data assimilation. This tool can be invaluable for deciding the requisite trade-offs between mission parameters such as spatial resolution and temporal revisit, and determining sensitivity to uncertainties associated with the raw measurements and/or higher-level retrievals (Crow et al. 2001, 2005; Wang et al. 2013; Ma et al. 2015). For example, Wang et al. (2008) performed an OSSE to test the ability of a hybrid ensemble transform Kalman filter–three-dimensional variational data assimilation (ETKF-3DVAR) system for the Weather Research and Forecasting (WRF) Model to generate quality ensemble perturbations used to calculate background-error covariances. Kumar et al. (2017) examined how the reliance on ensemble perturbations of forcing fields to develop the model error background impacts the performance of data assimilation through the use of an OSSE. Crow et al. (2005) designed an OSSE to simulate the impact of land surface heterogeneity, instrument error, and retrieval parameter uncertainty on soil moisture products derived from spaceborne L-band measurements.

With respect to terrestrial snow, DeLannoy et al. (2010) used an OSSE to test techniques for downscaling coarse-scale SWE products to the underlying finescale model state variables within the data assimilation system. No OSSE has been performed to estimate the potential value of proposed new radar-focused satellite observation systems when it comes to snow. As part of the ECCC-CSA Terrestrial Snow Mass Mission (TSMM) concept study, an OSSE was thus conducted to inform on the optimal mission configuration (i.e., resolution, revisit time, snow mass retrieval uncertainty).

As such, the main objective of this study is to determine the sensitivity of snow analysis to sensor configuration and SWE algorithm performance. Since Ku-band radar cannot provide information about SWE in wet snow conditions (liquid water content of 1% or more in this study), the impact of wet snow is also examined. This will provide a perspective on what is lost during wet snow events, of particular importance during spring snowmelt.

2. Observing System Simulation Experiment setup

The OSSE performed in this study covers one typical snow season, from 1 September 2014 to 31 May 2015, with the first two months considered as spinup and the last two months only analyzed to study the impact of snowmelt on snow analyses. It is run over a 200 km × 200 km domain in southern Quebec, as shown in Fig. 1a. Figure 1b gives an indication of the orography of the domain and Fig. 1c shows the dominant cover type (vegetation and urban) of each grid cell. This domain was chosen as it covers areas with some variety of seasonal snow types: under high vegetation (i.e., forests) on relatively elevated grounds, and over low vegetation and crops in the St. Lawrence River valley.

Fig. 1.
Fig. 1.

(a) Location of the domain in southern Quebec, Canada, (b) its orography (m) at a 100-m resolution, and (c) dominant cover type in each grid cell of the study domain.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

The different steps of the OSSE as designed for this study are represented in Fig. 2. The first step consists of the Synthetic Truth generation using a high-quality model running without data assimilation, from which observations are simulated and against which subsequent OSSE assimilation experiments are compared. Figure 2a shows an example of high-resolution (100 m) SWE products from the Synthetic Truth on an arbitrary date (15 March 2015). The generation of the synthetic SWE product is described in section 2a. Second, simulated observations are generated by upscaling and adding errors that are realistic for the hypothetical future observing system (details given in section 2b). Figure 2b shows the corresponding synthetic observations after upscaling to 2000-m resolution and addition of error to mimic retrieval uncertainties. Third, the assimilation system is used to run a control experiment, as well as assimilation experiments in which the synthetic observations under evaluation are assimilated (Fig. 2c); details given in section 2c. This allows for a forecast skill evaluation of the control experiment and the different assimilation experiments by comparing them to the Synthetic Truth state (Fig. 2d).

Fig. 2.
Fig. 2.

The different technical steps of the OSSE. (a) Generation of a Synthetic Truth state at 100-m resolution, with an example of an SWE product (mm) on 15 Mar 2015. (b) Production of synthetic observations through upscaling and addition of error. The figure shows the synthetic observation of SWE (mm) corresponding to the Synthetic Truth on 15 Mar 2015 after upscaling to 2000-m resolution and addition of error according to the threshold accuracy requirements. (c) Assimilation of different synthetic observations (black circles) at different times k within CaLDAS through the use of the EnKF and production of a control experiment (no assimilation). (d) Skill evaluation of the control experiment and the different assimilation experiments by comparing them to the Synthetic Truth state.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

a. Synthetic Truth

In the Synthetic Truth simulation, snowpack evolution is simulated at very high resolution and is referred to as a reference “truth.” State-of-the-art modeling is used to create this truth. The simulation is run using the Soil, Vegetation, and Snow (SVS; Husain et al. 2016; Alavi et al. 2016) land surface model, as described in section 3b, at a 100-m resolution with the best geophysical fields and forcing data available. The Synthetic Truth simulation is an offline simulation; it is driven by atmospheric data and does not allow for surface feedbacks on the atmosphere. For the orography, the 30-m resolution Shuttle Radar Topography Mission (SRTM v3.0; Shortridge and Messina 2011) database is used, while the 300-m resolution European Space Agency Climate Change Initiative Land Cover (ESA CCI LC, version 1.6.1; Hollmann et al. 2013) database is used for the water–land mask as well as vegetation cover. Soil texture is taken from the 1-km resolution Global Soil Dataset for Earth System Modeling (GSDE; Shangguan et al. 2014) database.

With respect to atmospheric forcing data, the first half of September 2014 is from ECCC’s Regional Deterministic Prediction System (RDPS), the 10-km version of the regional Global Environmental Multiscale (GEM) model (Mailhot et al. 2006). From 15 September 2014 onward, ECCC’s High Resolution Deterministic Prediction System (HRDPS; Milbrandt et al. 2016), a 2.5-km version of GEM, is available and used to drive the evolution of the land surface state simulated by the land surface model in offline mode. The switch between RDPS and HRDPS forcing data is necessary since the HRDPS only became experimental on 15 September 2014, and no HRDPS data are available before this date. Since it occurs during a snow-free period, the switch should not have an impact on the conclusions obtained in this study.

In both cases, downwelling shortwave and longwave radiation and surface pressure are taken at the surface. Air temperature, specific humidity, and wind are taken from GEM’s lowest prognostic level, which is roughly at 40 m for the wind and 20 m for the other variables. As detailed in Bernier et al. (2014), the land surface driver performs a simple meteorological downscaling to the surface model resolution for air temperature, surface pressure, and specific humidity. The key component of the downscaling procedures is an adaptation of air temperature and humidity to correct for the smoothing errors in elevation at lower resolutions. In terms of precipitation, the forcing data come from the Canadian Precipitation Analysis (CaPA; Mahfouf et al. 2007; Lespinas et al. 2015; Fortin et al. 2015). CaPA combines a short-range 6-h GEM (RDPS or HRDPS) precipitation forecast with available precipitation gauge observations using an optimum interpolation (OI) methodology. The final product is an hourly precipitation rate with no distinction between solid or liquid precipitation. The separation of the phases is done within the land surface driver, prior to entering SVS, using an air temperature threshold of 0°C, consistent with ECCC’s operational systems. Since the study period is limited to late fall and winter, distinction of precipitation phase is not a major issue because maximum temperatures during the day rarely exceed 0°C.

To obtain the best quality out of the RDPS and HRDPS forecasts, 6–12-h forecasts are used for precipitation (used for CaPA) and radiation forcings, in order to avoid the spinup phase since the prediction systems do not recycle clouds between forecasts, and 0–6-h forecasts for the remaining forcing fields. All forcings are taken from forecasting cycles that are launched at 0000, 0600, 1200, and 1800 UTC each day. All the main details of the Synthetic Truth are summarized in step 1 of Fig. 3 for easier reference.

Fig. 3.
Fig. 3.

Experimental setup showing the different components of the OSSE: geophysical fields, atmospheric forcings, and land surface models.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

b. Synthetic observations

Synthetic observations are derived from the Synthetic Truth at different resolutions by averaging the 100-m SWE values and by adding observation errors so as to mimic satellite-derived measurements and retrievals, as shown in step 2 of Fig. 3. Specifically, upscaling of Synthetic Truth SWE products is done by aggregation into three different resolutions—1000 m, 2000 m, and 10 km—and masking out grid points covered by more than 10% of open water (i.e., lakes, rivers). Perturbations in the form of unbiased random noise are then added to the upscaled data drawn from two normal distributions with a predefined standard deviation based on DeLannoy et al. (2010). The two distributions are designed to represent the goal (G) and threshold (T) accuracies for SWE retrievals that are targeted in the ECCC-CSA TSMM requirements, as described in Table 1. Here is the method used to obtain the synthetic observations (SWESO) by perturbing SWE from the Synthetic Truth after upscaling (SWESTup):
e1
where SWEmin (SWEmax) is the minimum (maximum) SWE that can be detected, STDE is the standard deviation of the error, VEGH is the high vegetation (i.e., trees) fractional cover. Function draws a random sample from a normal (Gaussian) distribution centered on 0.0 and standard deviation of STDE. Function draws a random sample from a uniform distribution between 5 and 10 mm with a random factor of +1 or −1. It is weighted by the fractional cover of high vegetation present in each grid cell (VEGH). A safeguard is used to correct negative SWE values after perturbation.
Table 1.

Accuracy objectives used for the reference truth degradation to mimic SWE retrievals.

Table 1.

For the goal (threshold) accuracy, an STDE of 10 mm (30 mm) is used for all SWE values greater than 10 mm (20 mm) (i.e., SWEmin). An assumption is made that radar retrievals of SWE are not sensitive to a thin layer of snow; all SWE values are thus set to zero, as shown in Eq. (1). Passive microwave and radar measurements are sensitive to the volume scattering of dry snow, which allows the retrieval of SWE. This volume scatter reaches a frequency-dependent limit, after which further increases in SWE do not result in more scatter: a saturation effect. This SWE threshold is variable due to the impact of snow stratigraphic properties but is approximately 150 mm at Ku band (Rott et al. 2010; Takala et al. 2011). So as to mimic the saturation effect in this case, beyond 200 and 150 mm (SWEmax) for the goal and threshold accuracy, respectively, all SWE values are set to SWEmax before perturbation. The observation error is further randomly increased using a uniform distribution between −5 and −10 mm and between 5 and 10 mm over forested areas (; Pulliainen et al. 1999; Derksen et al. 2003).

c. Data assimilation system

The assimilation system used is the Canadian Land Data Assimilation System (CaLDAS; Carrera et al. 2015). CaLDAS is built around an external land surface modeling system and uses the ensemble Kalman filter (EnKF) as the assimilation method. Numerous studies have proven the efficiency of the EnKF method within the context of both idealized and real experiments (Reichle et al. 2002; Crow and Wood 2003; Slater and Clark 2006; Kumar et al. 2008; DeLannoy et al. 2010; Kumar et al. 2017). The EnKF method works sequentially by alternating between a forecast step and an update step when observations are available, with the update equation as follows:
e2
where is the analysis, is the model first guess or background, is the Kalman gain matrix, y represents the observations, ω is a realization of the observation error (Burgers et al. 1998), and is the measurement operator. For the purpose of this experiment, CaLDAS was modified so that only SWE values from the synthetic observations were assimilated, with an observation error standard deviation set uniformly to 10 mm. As SWE is a prognostic variable of the land surface model, no forward model is necessary and is an identity matrix in this case.

A finite number of model trajectories are randomly produced and used to approximate the model error that is required by the EnKF. Twenty-four members are used in this study, as prior tests performed by Carrera et al. (2015) showed this was sufficient to characterize model error (bias, standard deviation, and RMSE) in CaLDAS. These trajectories are generated through perturbations on atmospheric forcings and the control variable, SWE. For air temperature, Gaussian additive perturbations—which are constant in space—are used, with a mean of 0 K and standard deviation of 1.0 K. The precipitation field is perturbed through the use of random spatial phasing errors independently in both latitudinal and longitudinal directions with a mean displacement of 0 km and a standard deviation of 50 km. The phasing errors are constant throughout the accumulation period of 6 h. Net solar and longwave radiation are subsequently spatially displaced to be consistent with the perturbed precipitation fields. Gaussian multiplicative perturbations are used for SWE, with a mean of 0 mm and a standard deviation of 0.016 mm, the equivalent of a maximum perturbation of ±5% (i.e., three standard deviations). The model error of SWE is thus the result of the combined perturbations on atmospheric forcings and SWE itself, and is estimated from the spread of the trajectories.

Atmospheric forcings used to produce the first guess are 30–36-h forecasts from RDPS described above to decrease the quality of the meteorological forcings. The land surface model in CaLDAS that provides the temporally and spatially varying first guess of the land surface state is Interactions between Soil, Biosphere, and Atmosphere (ISBA; Noilhan and Planton 1989; Bélair et al. 2003a,b), described in section 3a. As for the geophysical fields, the orography, land–sea mask, and vegetation cover all come from the 900-m resolution U.S. Geophysical Survey (USGS-GLCC v2.0; https://edcftp.cr.usgs.gov/project/glcc/globdoc2_0.html) database. The soil texture is obtained from the 112-arc-s (~3.5 km at the equator) resolution Jet Propulsion Laboratory Soil Moisture Active Passive (JPL SMAP) mission database (Das 2013).

Multiple CaLDAS integrations were run: one Open Loop simulation (i.e., no assimilation) and eight sensitivity tests with different configurations as shown in Table 2. Each of the first seven assimilation runs from Table 2 has a different combination of the resolution and perturbation of synthetic observations, as well as revisit time (see Fig. 2). The choice of simulation combinations stemmed from realistic feasibility based on likely technology to emerge in the next five years. These simulations do not take into account the difficulties met by radar measurements to detect SWE as soon as the snowpack contains liquid water. These are thus analyzed from 1 November 2014 until the beginning of spring snowmelt in the Synthetic Truth: 24 March 2015.

Table 2.

CaLDAS simulations nomenclature (please refer to Table 1 with respect to the different accuracies).

Table 2.

To provide a perspective on what is lost during wet snow events, particularly during spring, a ninth CaLDAS simulation is run (1000m_5d_Tw), in which SWE observations in each grid point are not assimilated as soon as the liquid water content of any layers in the snowpack is greater than 1%. This simulation is compared to 1000m_5d_T and the Synthetic Truth from 1 November 2014 to 31 May 2015.

CaLDAS simulations are as independent from the Synthetic Truth simulation as possible: different land surface model, lower-resolution (2.5 km versus 100 m) and lower-quality atmospheric forcings from different forecasting models, and finally, lower-resolution geophysical fields, as shown in Fig. 3. This is essential to test the sensitivity of the assimilation system to a range of potential mission configurations.

The introduction of a systematic bias in meteorological forcings (particularly in terms of precipitation) is by design. Although it violates the unbiased criteria of the ensemble Kalman filter assumption, this study is a convergence type experiment where the bias is not to be removed with respect to the assimilation runs. This is the most useful approach as it allows for a significant reduction of the bias when assimilating synthetic observations.

3. Land surface models

Because of the nature of this particular OSSE, two land surface models are needed—one within the assimilation system and another for the Synthetic Truth simulation—since these two elements of the experiment have to be as independent from one another as possible. As mentioned above, the simple one-snow-layer model ISBA has been chosen for the assimilation system since it is currently used operationally at ECCC. For the Synthetic Truth simulation, a more complex model is used: SVS v2.0, which includes an explicit multilayer snow scheme.

a. ISBA

The Canadian implementation of ISBA (Bélair et al. 2003a,b) has been used operationally at ECCC since 2001. ISBA is based on a force–restore (FR; Deardorff 1978; Hu and Islam 1995) approach for surface/soil temperature and only one energy budget is performed over the land portion of a model grid area, meaning that the surface thermal coefficient includes an area-weighted effect of bare soil, vegetation, and snow. The snow module within the land surface model is a modified version of the original scheme developed by Douville et al. (1995) and is described in Bélair et al. (2003b). It is a simple snow scheme with a single layer, simple hydrology, and thermal physics. It includes a reservoir for liquid water retained in the snowpack with exchanges with the snow mass reservoir and the melting effect due to incident rainfall on the snowpack. The snow scheme also contains a more sophisticated representation of the snow density than the original version of ISBA (Douville et al. 1995), that is, diagnostic calculations for the maximum snow density and for the density of fresh snow, and the effect of refreezing on the snow density.

ISBA’s snowpack simulation was evaluated in Carrera et al. (2010) for the Rocky Mountains. The land surface model was shown to be sensitive to precipitation forcing, and its performance improved by increasing the horizontal resolution (i.e., going from coarser to finer resolution) and through the use of terrain adaptation of the atmospheric forcings. Even so, ISBA was found to underestimate both snow cover extent and SWE values, and to melt the snowpack too early in spring.

b. SVS v2.0

Recently, ISBA has been replaced due to the demand for a more comprehensive model. The new and improved scheme is SVS (Husain et al. 2016; Alavi et al. 2016), and it addresses a number of ISBA’s weaknesses and limitations that were identified over the last decade.

Unlike ISBA, SVS introduces a new tiling approach that includes separate energy budgets for bare ground, vegetation, and two different snowpacks: the same snow scheme separately treats snow overlaying bare ground and low vegetation from snow under high vegetation. In its original form [i.e., SVS v1.0 described in Husain et al. (2016); Alavi et al. 2016], SVS used a snow model that considers prognostic equations for the superficial and mean snow temperatures that are obtained from the FR method for the two snowpacks (You et al. 2014). In SVS v2.0, the snow module has been upgraded to an explicit multilayer snow scheme (Decharme et al. 2016), which is an improvement of the original Boone and Etchevers (2001) explicit snow module.

It explicitly resolves the snow vertical temperature profile within the snowpack and includes multiple snow layers with a fine layer at the surface and at the bottom of the snowpack, which allow for realistic representation of the diurnal cycle in the top layers and of the heat conduction at the snow–soil interface. In this study, SVS v2.0, hereafter referred to as SVS, is run with nine layers. The total number of layers remains constant when snow is present on the ground. As described in Decharme et al. (2016), when the snowpack is less than 0.1 m, all layers are of equal depth. Beyond 0.1 m of total depth, the top and bottom layers reach their constant values of 0.01 and 0.02 m, respectively. If fresh snow is added to the top layer or if melting occurs, the layer thicknesses of the entire snowpack are recalculated and the snow mass and heat are redistributed appropriately.

As mentioned above, two distinct snowpacks are considered in SVS: one overlaying bare ground and low vegetation, and the other under high vegetation. The multilayer explicit snow scheme is used to simulate the evolution of these snowpacks. They differ in terms of net radiation balance at the snow surface as described below. The subscript “sn” refers to snow over bare ground and low vegetation, and the subscript “svh” to snow under high vegetation (i.e., trees).

The net radiation for the sn snowpack is given by
e3
where is the albedo of the sn snowpack, and are the total incoming solar and infrared radiation, is the emissivity of snow (constant at 0.99 in SVS), is the Stefan–Boltzmann constant, and is the temperature of the snowpack top layer. The svh net radiation is modified from Eq. (3) to account for partial transmission through the vegetation canopy of both incoming solar radiation and outgoing surface/snow radiation, and for the contribution of high vegetation downward radiant flux to the svh energy budget, such that
e4
in which is the albedo of the svh snowpack, is the snow-under-vegetation top layer temperature, is the vegetation mean temperature obtained from the FR method, is the canopy transmissivity for high vegetation (Sicart et al. 2004), and χ is the sky-view factor for high vegetation. The net radiation for each snowpack is then passed onto the snow model.
Snow density in each layer evolves as a result of snow compaction due to changes in snow viscosity. Snow density evolution in snow level i is computed as follows for sn and similarly for svh:
e5
where σ (Pa) is the vertical stress in each layer, and it is computed as the weight of the overlaying layers. Snow viscosity is a function of temperature, snow density, and liquid water content, as detailed in Decharme et al. (2016). The effect of temperature on snow viscosity is, however, limited since it becomes negligible at low temperatures (below 268 K approximately; Schleef et al. 2014). The equation for snowfall density () is expressed as a function of 10-m wind speed (m s−1) and 2-m air temperature (K) as follows:
e6
where a, b, and c are parameters with values 109 kg m−3, 6 kg m−3 K−1, and 26 kg s0.5 m−7/2, respectively, and is the melting point for water.

The snow albedo, and , is a function of the snow age of the first layer and the snow optical diameter, which is itself dependent on snow age. The age dependency takes into account the deposition of impurities that tend to decrease the albedo. The broadband albedo is calculated using the weighted average of three radiative bands; the first ([0.3–0.8] μm) represents the ultraviolet and visible range, and the other two ([0.8–1.5] and [1.5–2.8] μm) represent two near-infrared ranges.

Decharme et al. (2016) give an in-depth description of the snow model, as well as an evaluation of the model’s performance. The authors conclude that the multiple layers allow for a refinement of the temperature and density profiles in the snowpack leading to an improvement in the simulation of snow depth, SWE, and soil temperature during winter. When evaluating the snow module over a station in the French Alps (Col-de-Porte), the authors show that it simulates the snowpack with a bias of 2.981 mm, a centered root-mean-square error of 38.1 mm, and a temporal correlation of 0.924 over the snow seasons from 1993 to 2011. It is important to note that no model evaluation is performed in the current study since it is a synthetic experiment where reality does not affect the possible outcomes. To facilitate the assimilation of SWE in CaLDAS and the comparison with snow analyses, the total SWE for the final Synthetic Truth product is obtained from SVS’s two snowpacks using a weighted average depending on the fraction of bare ground and low and high vegetation for each grid cell.

4. Results

a. Impact of SWE assimilation on snow analyses

The assimilation results are evaluated against the Synthetic Truth for the period of 1 November 2014–24 March 2015, in order to avoid the spring snowmelt, over the 200 km × 200 km domain. All CaLDAS simulations are analyzed in this section except 1000m_5d_Tw, which will be considered in section 4b. The statistics shown here are calculated when gridcell SWE values of the Synthetic Truth and/or CaLDAS simulation are greater than 10 mm.

Figure 4 shows domain-averaged SWE over the study period for CaLDAS simulations. To simplify presentation, the simulations are grouped so as to better visualize the effect of different resolutions (Fig. 4a), revisit frequency (Fig. 4b), and accuracy (Fig. 4c). The Open Loop (thick black line) is noticeably different from the Synthetic Truth (dashed thick black line), which indicates that the experimental setup was efficient in rendering the two simulations sufficiently distinct of one another. The snow accumulation in the Open Loop peaks around 180 mm, while the Synthetic Truth snow accumulation peaks at approximately 85 mm. The Open Loop simulation thus overestimates SWE quantities by more than 210% when compared to the Synthetic Truth. This is in part explained by the insufficient snowmelt in the Open Loop during ephemeral melt events simulated in the Synthetic Truth at the end of November 2014 and end of December 2014. The large differences in accumulation are, however, mostly linked to the fact that 6–12-h CaPA forecasts are used to force the Synthetic Truth while 30–36-h RDPS forecasts are used in CaLDAS. As shown in Fig. 5, these differences in setup lead to an overestimation of about 70 cm of solid precipitation domain-wide by the end of March when comparing RDPS (blue line) versus CaPA (orange line).

Fig. 4.
Fig. 4.

Domain-averaged SWE (mm) over the 1 Nov 2014–24 Mar 2015 study period in Synthetic Truth (black dashed line) and Open Loop (solid black line) simulations as well the eight CaLDAS simulations (colored lines) separated so as to highlight the effects of (a) resolution, (b) revisit, and (c) accuracy. Please refer to Table 2 for the nomenclature of the different simulations.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

Fig. 5.
Fig. 5.

Solid (cm) and liquid (mm) precipitation accumulation from 1 Nov 2014 to 31 May 2015 for the 6–12-h CaPA (HRDPS) forecasts (orange) and 30–36-h RDPS forecasts (blue).

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

In all cases, the assimilation system is able to correct the first guess well enough and at a quasi-immediate temporal response to dramatically reduce the overestimation of snow accumulation observed in the Open Loop simulation. The last month prior to spring snowmelt, the assimilation system is able to correct the first guess to match the synthetic truth in terms of domain average. As seen in Figs. 4a and 4c, resolution and accuracy have little impact on the spatial average of SWE. In the case of resolution, the lack of spread among the different simulations may be because all simulations have to be scaled to 2.5-km resolution (i.e., CaLDAS resolution) prior to assimilation. In terms of accuracy, since the random errors added to the Synthetic Truth are unbiased (i.e., centered on zero), it is logical that very little spread in the spatial average of SWE among the simulations is observed. Revisit frequency, however, shows some sensitivity with, as expected, improvements brought by shorter revisits (Fig. 4b).

In a similar fashion, Fig. 6 shows time series of SWE spatial RMSE of each simulation grouped so as to highlight the effect of resolution, revisit frequency, and accuracy on snow analyses. The RMSE (and all the following statistics) of each CaLDAS simulations (at a 2.5-km resolution) is calculated using each grid point of the Synthetic Truth (at a 100-m resolution) and then averaged on the 2.5-km grid.

Fig. 6.
Fig. 6.

Spatial RMSE (mm) over the domain over the 1 Nov 2014–24 Mar 2015 study period when compared to Synthetic Truth SWE values separated so as to highlight the effects of (a) resolution, (b) revisit, and (c) accuracy. The thick blue line represents the 1000m_5d_T simulation, common to all panels. Please refer to Table 2 for the nomenclature of the different simulations.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

In all CaLDAS simulations, four large spikes in RMSE are observed at the beginning of the snow season: the end of November and the end of December correspond to ephemeral melt events in the Synthetic Truth, while mid-December and early January occur during large snowfall events, during which RDPS forcings show greater snow precipitation amounts than in CaPA forcings (refer to Fig. 5). During the rest of the “dry snow” season (i.e., until end of March), there is a general trend toward increased RMSE for all simulations and the maximum values just before the spring melt show significant differences among the different simulations. This would have an impact on the snow analysis during the spring melt even if no data are assimilated.

Figure 6a shows the impact of different observation resolutions on the spatial RMSE of snow analyses. It has very little impact at the beginning of the snow season. Differences start to emerge after the large snowfall event in mid-December. Although correlated to 1000m_5d_T and 2000m_5d_T, 10000m_5d_T’s spatial RMSE increases much more rapidly than 1000m_5d_T, with differences reaching 14 mm greater between mid-January and mid-February. Even though 1000m_5d_T remains better than 2000m_5d_T in terms of RMSE, the gain is much smaller than when compared to 10000m_5d_T. There are two reasons that could explain this. The first and most obvious is that a 2000-m resolution is much closer to a 1000-m resolution than a 10 000-m resolution is. The second reason is that both 1000- and 2000-m resolution gridded products have to be upscaled prior to assimilation in CaLDAS, thus partly mitigating the gain coming from the use of a higher resolution in observations.

In terms of revisit frequency, Fig. 6b demonstrates that 1000m_1d_T has the lowest spatial RMSE throughout the study period due to its quick response to variations in the synthetic observations. The impact of revisit frequency is well depicted during the event in early January, when the spike in RMSE is short-lived in 1000m_1d_T and longest in 1000m_5d_T.

Over time, Fig. 6c shows that the accuracy has a limited impact on the results during the beginning of the snow season, but a strong one (up to 8–9 mm) during the slow accumulation period of the winter (from mid-January to end of March). It is interesting to note that the improvement brought by a greater retrieval accuracy is not affected by the synthetic observation resolution. Indeed, the two goal-accuracy simulations remain similar throughout the study period, whereas 2000m_5d_T shows a greater spatial RMSE than 1000m_5d_T throughout most of the winter, for example. Thus, the goal retrieval accuracy has reduced the spread among the different observation resolutions.

Table 3 gives the spatial bias and STDE averaged over the study period to discuss the improvements related to both systematic and random errors, respectively, brought by the different assimilation setups. While 10000m_5d_T gives the highest values of spatial bias and STDE of CaLDAS simulations, 1000m_1d_T gives the best results in terms of bias (5.0 mm, corresponding to a 91% improvement compared to the Open Loop) and 1000m_5d_G reduces the STDE the most (8.0 mm, corresponding to a 65% improvement).

Table 3.

Spatial bias and STDE averaged over the study period of CaLDAS simulated SWE compared to the Synthetic Truth values over the 1 Nov 2014–24 Mar 2015 study period.

Table 3.

In terms of resolution, the overall bias (STDE) is reduced by 2.5 mm (5.5 mm) when going from a 10 000-m resolution (10000m_5d_T) to 2000-m (2000m_5d_T) and is slightly reduced when further increasing the resolution to 1000-m (1000m_5d_T). Overall, the bias is reduced by 24% and the STDE by 38% when increasing the resolution from 10 to 1 km. When reducing the revisit frequency from 1 to 3 days (1000m_3d_T), the bias increases by 2.8 mm and the STDE by 1.8 mm. A 5-day revisit (1000m_5d_T) further increases the bias by 1.5 mm and the STDE by 1.4 mm. Going from a 5-day to a 1-day revisit therefore leads to a 46% reduction of systematic errors and a 27% reduction in random errors.

Improving the accuracy from threshold to goal leads to a decrease in STDE (~36%) but an increase in bias (~11%) at both 2- and 1-km resolutions. Although the RMSE is greatly reduced with increased accuracy (see Fig. 6c), the proportion of systematic versus random errors is altered, indicating that most of the RMSE is reduced due to diminished random errors in the resulting snow analyses. While increases in resolution, revisit frequency, and accuracy all lead to reduced spatial RMSE values, spatial random errors are mostly reduced by improving resolution and accuracy, and spatial systematic errors are primarily reduced by a higher revisit frequency.

Another way to visualize the errors in snow analyses is through the SWE temporal RMSE as in Fig. 7. It shows the temporal RMSE calculated as the RMSE through the entire snow season at each grid cell for the Open Loop and seven CaLDAS simulations. Note that due to the large differences between the Open Loop data and the rest of the simulations, the color scale of the Open Loop is different. The Open Loop shows a pattern of lower RMSE in the region of relatively low elevation around the St. Lawrence River, where the vegetation is mostly grass and crops (i.e., low-lying vegetation), while the highest RMSE values occur in regions of relatively high elevation covered in dense forests (see Fig. 1). This may be due to the fact that SVS (used in the Synthetic Truth experiment) allows for two different snowpacks (snow overlaying bare ground and low vegetation as well as snow under high vegetation) to evolve separately, while ISBA (used in CaLDAS) only has one snowpack. Once the synthetic observations are assimilated this spatial pattern is greatly reduced, but there are remnants of high values in areas covered by forests due to the additional observation error that was introduced during the perturbation of the Synthetic Truth over these regions as described in section 2b.

Fig. 7.
Fig. 7.

Temporal RMSE (mm) when compared to Synthetic Truth SWE values calculated at each grid point over the 1 Nov 2014–24 Mar 2015 study period. Please note that the Open Loop has its own color scale due to the large difference in magnitude of the Open Loop RMSE compared to the other simulations. Please refer to Table 2 for the nomenclature of the different simulations.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

Based on Fig. 7, the impact of observation resolution on snow analyses is a clear improvement going from 10-km resolution to a 2000-m resolution, with 2000m_5d_T removing the obvious grid cell effects in 10000m_5d_T. Changes are mostly neutral going from 2000- to 1000-m resolution. In terms of revisit, the improvement coming from more frequent revisits is clearly seen, with higher values of RMSE in 1000m_5d_T than in 1000m_3d_T and 1000m_1d_T. The latter shows the lowest values of temporal RMSE among all CaLDAS simulations. As expected, a greater accuracy results in a decreased temporal RMSE in most grid points of the study domain, as shown in Fig. 7 when comparing 2000m_5d_T with 2000m_5d_G, for example.

Table 4 adds to these findings by spatially averaging the temporal bias and STDE for each simulation. The lowest values in both bias and STDE come from 1000m_1d_T: an improvement of 93% (66%) in bias (STDE) when compared to the Open Loop. The 1-day revisit reduces the bias by 4.5 mm (51%) and the STDE by 4.2 mm (32%) compared to a 5-day revisit frequency. A tenfold increase in resolution (10000m_5d_T versus 1000m_5d_T) leads to 3.3 mm (27%) in bias reduction and 1.3 mm (9%) in STDE reduction. Similarly to the spatial scores presented above, increasing the accuracy leads to a reduction in temporal STDE of about 4.1 mm (~30%) but an increase in temporal bias of about 1.0 mm (~11%) while the overall RMSE is reduced (see Fig. 7). While increases in resolution, revisit frequency, and accuracy all lead to reduced temporal RMSE values, random errors are mostly reduced by improving revisit frequency and accuracy, and systematic errors are primarily reduced by a higher revisit frequency but also by an increased resolution. As expected, the revisit frequency thus plays a key role in the reduction of both temporal systematic and random errors when compared to the Open Loop.

Table 4.

Temporal bias and STDE averaged over the study domain of CaLDAS simulated SWE compared to Synthetic Truth values over the 1 Nov 2014–24 Mar 2015 study period.

Table 4.

b. Impact of wet snow

To provide a perspective on what is lost during wet snow events, another CaLDAS simulation was run (1000m_5d_Tw, see Table 2) in which SWE observations in each grid point were not assimilated as soon as the liquid water content of any layers in the snowpack is greater than 1%. The analysis of 1000m_5d_Tw simulation compared to 1000m_5d_T, the Open Loop and the Synthetic Truth from 1 November 2014 to 31 May 2015 is carried in a similar fashion than what is done in the previous section.

The spatial average of SWE, given in Fig. 8, shows that differences between 1000m_5d_Tw and 1000m_5d_T occur only during ephemeral melting events (late November and late December) and spring snowmelt in the Synthetic Truth. CaLDAS rapidly brings back the spatial SWE average of 1000m_5d_Tw toward 1000m_5d_T after melting events. Interestingly, wet snow does not impact the annual maximum SWE or the timing of end-of-season complete meltdown.

Fig. 8.
Fig. 8.

Domain-averaged SWE (mm) over the 1 Nov 2014–31 May 2015 study period. Please refer to Table 2 for the nomenclature of the different simulations.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

The spatial RMSE of both simulations (Fig. 9) show similar differences during melting events only, as expected. The impact of missing radar SWE retrievals is particularly strong during spring melt with a difference in RMSE of up to 50 mm between the two simulations. As shown in Table 5, the lack of detection during wet snow event increases the overall spatial bias by 2.3 mm (19% compared to 1000m_5d_T) and the STDE by 2.0 mm (16%) over the study period.

Fig. 9.
Fig. 9.

Spatial RMSE (mm) over the domain over the 1 Nov 2014–31 May 2015 study period when compared to Synthetic Truth SWE values. Please refer to Table 2 for the nomenclature of the different simulations.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

Table 5.

Spatial bias and STDE averaged over the study period, and temporal bias and STDE averaged over the study domain, of CaLDAS simulated SWE compared to Synthetic Truth values over the 1 Nov 2014–31 May 2015 study period.

Table 5.

In terms of temporal scores in Table 5, wet snow lack of detection increases the bias by 1.1 mm (10%) and the STDE by 2.4 mm (15%) over the study domain. The impact of wet snow is thus greatest on the spatial systematic errors and the temporal random errors. As seen in Fig. 10, most of the loss in temporal RMSE is situated in the higher grounds covered by forests. In these areas, the snowpack is deeper (see Fig. 2 as an example) and melting occurs over a longer period of time, thus explaining the greater impact of wet snow.

Fig. 10.
Fig. 10.

As in Fig. 7, but for the 1 Nov 2014–31 May 2015 period.

Citation: Journal of Hydrometeorology 20, 1; 10.1175/JHM-D-17-0241.1

5. Discussion and conclusions

The objective of this study was to determine the impact of sensor configuration (i.e., resolution, revisit frequency), SWE algorithm performance (i.e., accuracy), and the loss of radar sensitivity to SWE during snowmelt period on snow analyses using an OSSE. For this experiment to be effective, the Open Loop and Synthetic Truth simulations have to be as distinct from one another as possible, while remaining realistic. As this is achieved, the sensitivity of the Canadian Land Data Assimilation System to a range of potential mission configurations is tested.

All simulations except the wet snow simulation (1000m_5d_Tw) are analyzed from 1 November 2014 to 24 March 2015 to avoid spring snowmelt. Results show that resolution and accuracy have little impact on the spatial average of SWE. The lack of spread among the simulations using different observation resolution is mostly because all tested resolutions have to be scaled to CaLDAS resolution (i.e., 2.5 km) prior to assimilation. With respect to accuracy, since the random errors added to the Synthetic Truth are unbiased (i.e., centered on zero), the narrow spread in the spatial average of SWE among the simulations was expected. However, in our current setup the observation error standard deviation of 10 mm for the ensemble Kalman filter does not include instrument error reflecting the different retrieval accuracies (i.e., T versus G). In the future, this could be addressed by increasing the observation error standard deviation to 30 mm in threshold accuracy experiments, which would allow for a better estimation of improvements brought upon by the goal accuracy. Note that accuracy is constrained across a relatively narrow range (10–30 mm STDE), which is better than the performance of current coarse resolution SWE products such as GlobSnow (40–50 mm STDE). Ultimately, only the revisit frequency has an impact on the spatial-averaged evolution of the snowpack via timely updates during the periods of rapid snowpack change.

In terms of spatial RMSE, spatial resolution, revisit frequency, and accuracy of synthetic retrievals all have a strong effect on snow analyses, with the largest improvement coming from an increase in resolution from 10 to 1 km. A short revisit time (1 day) allows for a quick reduction of the RMSE during rapid changes (fresh snowfall or melting) in the Synthetic Truth, while the 5-day revisit is the longest to respond as would be expected, which demonstrates that the revisit frequency is a primary characteristic to keep in mind when defining the optimal configuration for satellite derived snow measurements. Retrieval accuracy has a limited impact on the results during the snow onset period, but is more substantial during the accumulation from mid-January to the end of March. Interestingly, the goal retrieval accuracy reduces the spread among the simulations, which are differentiated in terms of observation resolution, when compared to their threshold accuracy analogs. The analysis of spatial scores averaged over the study period shows that, while increases in resolution, revisit frequency, and accuracy all lead to reduced spatial RMSE values, spatial random errors are mostly reduced by improving resolution and accuracy, and spatial systematic errors are primarily reduced by a higher revisit frequency.

The temporal RMSE is also affected by all three tested factors: resolution, revisit, and accuracy of the retrievals. Because of the temporal component of the RMSE in this case, the largest improvement comes from the increase in revisit frequency, with an increase in spread of RMSE values when the revisits become longer as a result of less frequent assimilation leading to a less constrained land surface model. The lowest values in both bias and STDE come from the 1-day revisit: an improvement of 93% (66%) in bias (STDE) when compared to the Open Loop. While increases in resolution, revisit frequency, and accuracy all reduce the temporal RMSE, random errors are mostly reduced by improving revisit frequency and accuracy, and systematic errors are primarily reduced by a higher revisit frequency but also by an increased resolution. As expected, the revisit frequency thus plays a key role in the reduction of both temporal systematic and random errors when compared to the Open Loop.

Since radar retrievals are much more uncertain during melt periods due to snow wetness, the impact of wet snow on snow analyses is examined over the 1 November 2014–31 May 2015 period. Results show that in this case ephemeral melting events during late fall and winter have a limited impact on the quality of snow analyses. The effect of losing the capability to retrieve SWE under wet snow conditions is particularly important during spring melt with an increase in spatial RMSE of up to 50 mm. This would have had a significant impact on hydrology applications in terms of the timing in maximum runoff and freshet due to snowmelt, which would lead to implications for flood predictions, for example. It is important to note that an absence of SWE retrievals during snowmelt does not affect the domain-wide annual maximum SWE or the timing of the end of the snow cover season. These results show that radar measurements, although uncertain during melting events, are very useful in adding skill to snow analyses over most of the snow season.

Three observation resolutions are tested in this study: 1000 m, 2000 m, and 10 km. Large improvements in snow analyses were identified when going from 10-km resolution to 1000- or 2000-m resolution, but the impact of going from 2000- to 1000-m resolution is relatively weaker. These two resolutions are finer than the resolution at which CaLDAS is run, so the observations at 1000- and 2000-m resolution thus have to be upscaled to the first-guess model grid prior to the assimilation process in CaLDAS. Since this is done using a nearest-neighbor interpolation, the upscaling process eliminates the subgrid variability coming from observations by applying neighboring values of higher-resolution observations to 2.5-km grid cells. Results thus show that observation resolution should be finer than the assimilation system resolution. At ECCC, the global models are currently run at 15-km resolution but can be expected to be run at a resolution of about 2.5 km in 10 years or so (when the potential satellite would be launched).

Based on the OSSE performed in this study, efforts should also be put toward the optimization of the revisit frequency as well as the SWE algorithm performance in order to obtain the best possible satellite-derived snow measurements. The results suggest that the assimilation of snow measurements from a satellite with a robust configuration for optimal performance is likely to provide accurate snow analyses, which may then be used to initialize numerical weather prediction and hydrological models (Zaitchik and Rodell 2009). This would have a significant impact on numerical weather prediction through surface feedbacks to the atmosphere, as well as on hydrological forecasts due to the essential role of snow in the water cycle.

Note that the Open Loop simulation is setup with no assimilation, while the current operational system assimilates some surface observations in its snow analyses. The improvements of the CaLDAS integrations compared to the Open Loop do not necessarily reflect the improvements one would see when compared to the actual operational system. However, the choice of not including other assimilated variables allows for a clean comparison between snow analyses and the Synthetic Truth. The assimilation of in situ snow depths, screen temperature, and humidity (as done operationally) would mask the impact of synthetic observation assimilation.

It must be stressed that the introduction of a bias in the assimilation runs was by design (as shown in Fig. 3) and that it violates the unbiased criteria of the ensemble Kalman filter assumption. This study is, however, a convergence type experiment where the bias is not to be removed with respect to the assimilation runs, similarly to what is done in Balsamo et al. (2006), Kumar et al. (2008), and Kumar et al. (2017). This is the most useful approach as it allows for a significant reduction of the bias when assimilating synthetic observations.

This study is a precursor to an OSSE focused on radiance-based assimilation, in order to test the readiness of operational environmental prediction systems to integrate SWE retrievals. The following OSSE will be designed to assimilate synthetic backscatter measurements through the use of a snow radar model, as would be operationally implemented once satellite radar measurements sensitive to snow volume are available. It will include other study domains to enhance the value of the project conclusions, such as alpine regions, more temperate snow-covered regions, and northern latitudes in order to distinguish permanent, seasonal (as tested in this study), and ephemeral snow covers, as well as thin and deep snowpacks. Since these have different behaviors and characteristics, they may have a distinct impact on the quality of snow analyses.

Acknowledgments

This work was supported by the Canadian Space Agency through Government Related Initiatives Program (GRIP) to Environment and Climate Change Canada. Météo France financially supported Dr. Vincent Vionnet during his scientific stay at ECCC. The authors thank Dr. Nicolas Gasset for his help in using CaLDAS, and Dr. Danahé Paquin-Ricard and Dr. Weiguang Chang for sharing their computing resources. The authors are also grateful towards Dr. Vincent Fortin, Dr. Sylvie Leroyer, Dr. Barbara Casati, and Milena Dimitrijevic for the helpful discussions throughout this project, and Vanh Souvanlasy for his help with the geophysical fields. Finally, a special thanks to Djamel Bouhemhem and Dominic Racette for assistance with computing facilities and CaLDAS.

REFERENCES

  • Alavi, N., S. Belair, V. Fortin, S. Zhang, S. Z. Husain, M. L. Carrera, and M. Abrahamowicz, 2016: Warm season evaluation of soil moisture prediction in the Soil, Vegetation, and Snow (SVS) scheme. J. Hydrometeor., 17, 23152332, https://doi.org/10.1175/JHM-D-15-0189.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Balsamo, G., J.-F. Mahfouf, S. Bélair, and G. Deblonde, 2006: A global root-zone soil moisture analysis using simulated L-band brightness temperature in preparation for the Hydros satellite mission. J. Hydrometeor., 7, 11261146, https://doi.org/10.1175/JHM525.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Barnett, T., L. Dumenil, U. Schlese, E. Roeckner, and M. Latif, 1989: The effect of Eurasian snow cover on regional and global climate variations. J. Atmos. Sci., 46, 661685, https://doi.org/10.1175/1520-0469(1989)046<0661:TEOESC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bélair, S., L.-P. Crevier, J. Mailhot, B. Bilodeau, and Y. Delage, 2003a: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part I: Warm season results. J. Hydrometeor., 4, 352370, https://doi.org/10.1175/1525-7541(2003)4<352:OIOTIL>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bélair, S., R. Brown, J. Mailhot, B. Bilodeau, and L.-P. Crevier, 2003b: Operational implementation of the ISBA land surface scheme in the Canadian regional weather forecast model. Part II: Cold season results. J. Hydrometeor., 4, 371386, https://doi.org/10.1175/1525-7541(2003)4<371:OIOTIL>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bernier, N., S. Bélair, B. Bilodeau, and L. Tong, 2014: Assimilation and high resolution forecasts of surface and near surface conditions for the 2010 Vancouver Winter Olympic and Paralympic Games. Pure Appl. Geophys., 171, 243256, https://doi.org/10.1007/s00024-012-0542-0.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonan, G., 2008: Ecological Climatology, Concepts and Applications. 2nd ed. Cambridge University Press, 563 pp.

    • Crossref
    • Export Citation
  • Boone, A., and P. Etchevers, 2001: An intercomparison of three snow schemes of varying complexity coupled to the same land-surface model: Local scale evaluation at an alpine site. J. Hydrometeor., 2, 374394, https://doi.org/10.1175/1525-7541(2001)002<0374:AIOTSS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Brasnett, B., 1999: A global analysis of snow depth for numerical weather prediction. J. Appl. Meteor., 38, 726740, https://doi.org/10.1175/1520-0450(1999)038<0726:AGAOSD>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Brown, R., and D. Tapsoba, 2007: Improved mapping of snow water equivalent over Quebec. 64th Eastern Snow Conf., St John’s, NL, Canada, Eastern Snow Conference, 17.

  • Brown, R., C. Derksen, and L. Wang, 2010: A multi-data set analysis of variability and change in the Arctic spring snow cover extent, 1967–2008. J. Geophys. Res., 115, D16111, https://doi.org/10.1029/2010JD013975.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Burgers, G., P. van Leeuwen, and G. Evensen, 1998: Analysis scheme in the ensemble Kalman filter. Mon. Wea. Rev., 126, 17191724, https://doi.org/10.1175/1520-0493(1998)126<1719:ASITEK>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carrera, M., S. Bélair, V. Fortin, B. Bilodeau, D. Charpentier, and I. Doré, 2010: Evaluation of snowpack simulations over the Canadian Rockies with an experimental hydrometeorological modeling system. J. Hydrometeor., 11, 11231140, https://doi.org/10.1175/2010JHM1274.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carrera, M., S. Bélair, and B. Bilodeau, 2015: The Canadian Land Data Assimilation System (CALDAS): Description and synthetic evaluation study. J. Hydrometeor., 16, 12931314, https://doi.org/10.1175/JHM-D-14-0089.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cohen, J., and D. Entekhabi, 1999: Eurasian snow cover variability and Northern Hemisphere climate predictability. Geophys. Res. Lett., 26, 345348, https://doi.org/10.1029/1998GL900321.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cohen, J., and K. Saito, 2001: Eurasian snow cover, more skillful in predicting U.S. winter climate than the NAO/AO? Geophys. Res. Lett., 28, 299302, https://doi.org/10.1029/2000GL011927.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cohen, J., K. Saito, and D. Entekhabi, 2001: The role of the Siberian high in Northern Hemisphere climate variability. Geophys. Res. Lett., 28, 299302, https://doi.org/10.1029/2000GL011927.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Crow, W., and E. Wood, 2003: The assimilation of remotely sensed soil brightness temperature imagery into a land surface model using ensemble Kalman filtering: A case study based on ESTAR measurements during SGP97. Adv. Water Resour., 26, 137149, https://doi.org/10.1016/S0309-1708(02)00088-X.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Crow, W., M. Drusch, and E. Wood, 2001: An observation system simulation experiment for the impact of land surface heterogeneity on AMSR-E soil moisture retrieval. IEEE Trans. Geosci. Remote Sens., 39, 16221631, https://doi.org/10.1109/36.942540.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Crow, W., and et al. , 2005: An observing system simulation experiment for Hydros radiometer-only soil moisture products. IEEE Trans. Geosci. Remote Sens., 43, 12891303, https://doi.org/10.1109/TGRS.2005.845645.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Das, N., 2013: Soil Moisture Active Passive (SMAP) ancillary data report: Soil attributes. SMAP Doc. 044, JPL D-53058, 16 pp.

  • Deardorff, J., 1978: Efficient prediction of ground surface temperature and moisture with inclusion of a layer of vegetation. J. Geophys. Res., 83, 18891903, https://doi.org/10.1029/JC083iC04p01889.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Decharme, B., E. Brun, A. Boone, C. Delire, P. L. Moigne, and S. Morin, 2016: Impacts of snow and organic soils parameterization on north-Eurasian soil temperature profiles simulated by the ISBA land surface model. Cryosphere, 10, 853877, https://doi.org/10.5194/tc-10-853-2016.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • DeLannoy, G., R. Reichle, P. Houser, K. Arsenault, N. Verhoest, and V. Pauwels, 2010: Satellite-scale snow water equivalent assimilation into a high-resolution land surface model. J. Hydrometeor., 11, 352369, https://doi.org/10.1175/2009JHM1192.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Derksen, C., A. Walker, E. LeDrew, and B. Goodison, 2003: Combining SMMR and SMM/I data for time series analysis of central North American snow water equivalent. J. Hydrometeor., 4, 304316, https://doi.org/10.1175/1525-7541(2003)4<304:CSAIDF>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • de Rosnay, P., G. Balsamo, C. Albergel, J. Munos-Sabater, and L. Isaksen, 2014: Initialisation of land surface variables for numerical weather prediction. Surv. Geophys., 35, 607621, https://doi.org/10.1007/s10712-012-9207-x.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • de Rosnay, P., L. Isaksen, and M. Dahoui, 2015: Snow data assimilation at ECMWF. ECMWF Newsletter, No. 143, ECMWF, Reading, United Kingdom, 26–31, https://doi.org/10.21957/lkpxq6x5.

    • Crossref
    • Export Citation
  • Douville, H., and J.-F. Royer, 1996: Sensitivity of the Asian summer monsoon to an anomalous Eurasian snow cover within the Meteo-France GCM. Climate Dyn., 12, 449466, https://doi.org/10.1007/BF02346818.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douville, H., J.-F. Royer, and J.-F. Mahfouf, 1995: A new snow parameterization for the Météo-France climate model. Climate Dyn., 12, 3752, https://doi.org/10.1007/BF00208761.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., and et al. , 2005: NOAA operational hydrological products derived from the Advanced Microwave Sounding Unit. IEEE Trans. Geosci. Remote Sens., 43, 10361049, https://doi.org/10.1109/TGRS.2004.843249.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fletcher, S., G. L. Iston, C. Hiemstra, and S. Miller, 2012: Assimilating MODIS and AMSR-E snow observations in a snow evolution model. J. Hydrometeor., 13, 14751492, https://doi.org/10.1175/JHM-D-11-082.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fortin, V., G. Roy, N. Donaldson, and A. Mahidjiba, 2015: Assimilation of radar quantitative precipitation estimations in the Canadian Precipitation Analysis (CaPA). J. Hydrol., 531, 296307, https://doi.org/10.1016/j.jhydrol.2015.08.003.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Foster, J., M. Owe, and A. Rango, 1983: Snow cover and temperature relationships in North America and Eurasia. J. Climate Appl. Meteor., 22, 460469, https://doi.org/10.1175/1520-0450(1983)022<0460:SCATRI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hollmann, R., and et al. , 2013: The ESA Climate Change Initiative: Satellite data records for essential climate variables. Bull. Amer. Meteor. Soc., 94, 15411552, https://doi.org/10.1175/BAMS-D-11-00254.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hu, Z., and S. Islam, 1995: Prediction of ground surface temperature and soil moisture content by the force-restore method. Water Resour. Res., 31, 25312539, https://doi.org/10.1029/95WR01650.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Husain, S. Z., N. Alavi, S. Belair, M. Carrera, S. Zhang, M. A. V. Fortin, and N. Gauthier, 2016: The multi-budget Soil, Vegetation, and Snow (SVS) scheme for land surface parameterization: Offline warm season evaluation. J. Hydrometeor., 17, 22932313, https://doi.org/10.1175/JHM-D-15-0228.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kelly, R., 2009: The AMSR-E Snow Depth Algorithm: Description and initial results. J. Remote Sens. Soc. Japan, 29, 307317, https://doi.org/10.11440/rssj.29.307.

    • Search Google Scholar
    • Export Citation
  • Kumar, S., R. Reichle, D. Peters-Lidard, R. Koster, X. Zhan, W. Crow, J. Eylander, and P. Houser, 2008: A land surface data assimilation framework using the land information system: Description and applications. Adv. Water Resour., 31, 14191432, https://doi.org/10.1016/j.advwatres.2008.01.013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kumar, S., J. Dong, C. Peters-Lidard, D. Mocko, and B. Gomez, 2017: Role of forcing uncertainty and model error background characterization in snow data assimilation. Hydrol. Earth Syst. Sci., 21, 26372647, https://doi.org/10.5194/hess-21-2637-2017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Larue, F., A. Royer, D. D. Sève, A. Langlois, A. Roy, and L. Brucker, 2017: Validation of GlobSnow-2 snow water equivalent over eastern Canada. Remote Sens. Environ., 194, 264277, https://doi.org/10.1016/j.rse.2017.03.027.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Leathers, D. J., A. W. Ellis, and D. Robinson, 1995: Characteristics of temperature depressions associated with snow cover across the northeast United States. J. Appl. Meteor., 34, 381390, https://doi.org/10.1175/1520-0450-34.2.381.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lespinas, F., V. Fortin, G. Roy, P. Rasmussen, and T. Stadnyk, 2015: Performance Evaluation of the Canadian Precipitation Analysis (CaPA). J. Hydrometeor., 16, 20452064, https://doi.org/10.1175/JHM-D-14-0191.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Luojus, K., and et al. , 2010: Investigating the feasability of the GlobSnow snow water equivalent data for climate research purposes. 2010 IEEE Int. Geoscience and Remote Sensing Symp., Honolulu, HI, IEEE, 4851–4853, https://doi.org/10.1109/IGARSS.2010.5741987.

    • Crossref
    • Export Citation
  • Ma, Z., L. Riishojgaard, M. Masutani, J. Woollen, and G. Emmitt, 2015: Impact of different satellite wind lidar telescope configurations on NCEP GFS forecast skill in observing system simulation experiments. J. Atmos. Oceanic Technol., 32, 478495, https://doi.org/10.1175/JTECH-D-14-00057.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mahfouf, J.-F., B. Brasnett, and S. Gagnon, 2007: A Canadian Precipitation Analysis (CaPA) project: Description and preliminary results. Atmos.–Ocean, 45, 117, https://doi.org/10.3137/ao.v450101.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mailhot, J., and et al. , 2006: The 15-km version of the Canadian regional forecast system. Atmos.–Ocean, 44, 133149, https://doi.org/10.3137/ao.440202.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Masutani, M., and et al. , 2010: Observation System Simulation Experiments. Data Assimilation, W. A. Lahoz, B. Khattatov and R. Ménard, Eds., Springer, 647–677.

  • Milbrandt, J., S. Bélair, M. Faucher, M. Vallée, M. Carrera, and A. Glazer, 2016: The Pan-Canadian High Resolution (2.5 km) Deterministic Prediction System. Wea. Forecasting, 31, 17911816, https://doi.org/10.1175/WAF-D-16-0035.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Noilhan, J., and S. Planton, 1989: A simple parameterization of land surface processes for meteorological models. Mon. Wea. Rev., 117, 536549, https://doi.org/10.1175/1520-0493(1989)117<0536:ASPOLS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pullen, S., C. Jones, and G. Rooney, 2011: Using satellite-derived snow cover data to implement a snow analysis in the Met Office Global NWP Model. J. Appl. Meteor. Climatol., 50, 958973, https://doi.org/10.1175/2010JAMC2527.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pulliainen, J., J. Grandell, and M. Hallikainen, 1999: HUT snow emission model and its applicability for snow water equivalent retrieval. IEEE Trans. Geosci. Remote Sens., 37, 13781390, https://doi.org/10.1109/36.763302.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reichle, R., D. McLaughlin, and D. Entekhabi, 2002: Hydrologic data assimilation with the ensemble Kalman filter. Mon. Wea. Rev., 130, 103114, https://doi.org/10.1175/1520-0493(2002)130<0103:HDAWTE>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rott, H., and et al. , 2010: Cold regions hydrology high resolution observatory for snow and cold land processes. Proc. IEEE, 98, 752765, https://doi.org/10.1109/JPROC.2009.2038947.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schleef, S., H. Lowe, and M. Schneebeli, 2014: Influence of stress, temperature and crystal morphology on isothermal densification and specific surface area decrease of new snow. Cryosphere, 8, 18251838, https://doi.org/10.5194/tc-8-1825-2014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shangguan, W., Y. Dai, Q. Duan, B. Liu, and H. Yuan, 2014: A global soil data set for earth system modeling. J. Adv. Model. Earth Syst., 6, 249263, https://doi.org/10.1002/2013MS000293.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shortridge, A., and J. Messina, 2011: Spatial structure and landscape association of SRTM error. Remote Sens. Environ., 115, 15761587, https://doi.org/10.1016/j.rse.2011.02.017.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sicart, J., J. Pomeroy, R. Essery, J. Hardy, T. Link, and D. Marks, 2004: A sensitivity study of daytime net radiation during snowmelt to forest canopy and atmospheric conditions. J. Hydrometeor., 5, 774784, https://doi.org/10.1175/1525-7541(2004)005<0774:ASSODN>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Slater, A., and M. Clark, 2006: Snow data assimilation via an ensemble Kalman filter. J. Hydrometeor., 7, 478493, https://doi.org/10.1175/JHM505.1.

  • Takala, M., K. Luojus, J. Pulliainen, C. Derksen, J. Lemmetyinen, J.-P. Karna, 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, 35173529, https://doi.org/10.1016/j.rse.2011.08.014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Vernekar, A., J. Zhou, and J. Shukla, 1995: The effect of Eurasian snow cover on the Indian monsoon. J. Climate, 8, 248266, https://doi.org/10.1175/1520-0442(1995)008<0248:TEOESC>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Walland, D. J., and I. Simmonds, 1996: Modelled atmospheric response to changes in Northern Hemisphere snow cover. Climate Dyn., 13, 2534, https://doi.org/10.1007/s003820050150.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wang, H., X.-Y. Huang, and Y. Chen, 2013: An Observing System Simulation Experiment for the impact of MTG candidate infrared sounding mission on regional forecasts: System development and preliminary results. ISRN Meteor., 2013, 971501, https://doi.org/10.1155/2013/971501.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wang, X., D. Barker, C. Snyder, and T. Hamill, 2008: A hybrid ETKF-3DVAR data assimilation scheme for the WRF Model. Part I: Observing System Simulation Experiment. Mon. Wea. Rev., 136, 51165131, https://doi.org/10.1175/2008MWR2444.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xu, L., and P. Dirmeyer, 2011: Snow-atmosphere coupling strength in a global atmospheric model. Geophys. Res. Lett., 38, L13401, https://doi.org/10.1029/2011GL048049.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yang, D., D. Robinson, Y. Zhao, T. Estilow, and H. Ye, 2003: Streamflow response to seasonal snow cover extent changes in large Siberian watersheds. J. Geophys. Res., 108, 45784591, https://doi.org/10.1029/2002JD003149.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • You, J., D. Tarboton, and C. Luce, 2014: Modeling the snow surface temperature with a one-layer energy balance snowmelt model. Hydrol. Earth Syst. Sci., 18, 50615076, https://doi.org/10.5194/hess-18-5061-2014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yueh, S., S. Dinardo, A. Akgiray, R. West, D. Cline, and K. Elder, 2009: Airborne Ku-band polarimetric radar remote sensing of terrestrial snow cover. IEEE Trans. Geosci. Remote Sens., 47, 33473364, https://doi.org/10.1109/TGRS.2009.2022945.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zaitchik, B., and M. Rodell, 2009: Forward-looking assimilation of MODIS-derived snow-covered area into a land surface model. J. Hydrometeor., 10, 130148, https://doi.org/10.1175/2008JHM1042.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
Save