Alexandre Roy

Snow Microwave Emission Modeling of Ice Lenses Within a Snowpack Using the Microwave Emission Model for Layered Snowpacks

Ice lens formation, which follows rain on snow events or melt-refreeze cycles in winter and spring, is likely to become more frequent as a result of increasing mean winter temperatures at high latitudes. These ice lenses significantly affect the microwave scattering and emission properties, and hence snow brightness temperatures that are widely used to monitor snow cover properties from space. To understand and interpret the spaceborne microwave signal, the modeling of these phenomena needs improvement. This paper shows the effects and sensitivity of ice lenses on simulated brightness temperatures using the microwave emission model of layered snowpacks coupled to a soil emission model at 19 and 37 GHz in both horizontal and vertical polarizations. Results when considering pure ice lenses show an improvement of 20.5 K of the root mean square error between the simulated and measured brightness temperature (Tb) using several in situ data sets acquired during field campaigns across Canada. The modeled Tbs are found to be highly sensitive to the vertical location of ice lenses within the snowpack.

Improved Corrections of Forest Effects on Passive Microwave Satellite Remote Sensing of Snow Over Boreal and Subarctic Regions

Microwave radiometry has been extensively used in order to estimate snow water equivalent in northern regions. However, for boreal and taiga environments, the presence of forest causes important uncertainties in the estimates. Variations in snow cover and vegetation in northeastern Canada (north of the Québec province) were characterized in a transect from 50°N to 60 °N during the International Polar Year field campaign of February 2008. Forest properties show a strong latitudinal gradient in fraction and stem volume. A large database (>; 2000 points with a stem volume ranging between 0 and 700 m3 ·ha-1) showed that brightness temperatures (Tb) decrease as forest cover fraction decreases until a cover fraction of about 25% is reached. Furthermore, Tb values saturate at high stem volume, particularly at 37 GHz. We defined new relationships for the forest transmissivity as a function of stem volume and depending on the frequency/polarization. The proposed relationships give asymptotic transmissivity saturation levels of 0.51, 0.55, 0.53, and 0.53 for 19 GHz [vertical (V) polarization], 19 GHz [horizontal (H) polarization], 37 GHz (V polarization), and 37 GHz (H polarization), respectively. These relationships were used to estimate snow Tb from the Advanced Microwave Scanning Radiometer-Earth Observing System brightness temperatures at 18.7 and 36.5 GHz, and results show an estimated snow brightness temperature well correlated to the airborne snow brightness temperatures over vegetation-free areas.

Brightness Temperature Simulations of the Canadian Seasonal Snowpack Driven by Measurements of the Snow Specific Surface Area

Snow grain size is the snowpack parameter that most affects the microwave snow emission. The specific surface area (SSA) of snow is a metric that allows rapid and reproducible field measurements and that well represents the grain size. However, this metric cannot be used directly in microwave snow emission models (MSEMs). The aim of this paper is to evaluate the suitability and the adaptations required for using the SSA in two MSEMs, i.e., the Dense Media Radiative Theory-Multilayer model (DMRT-ML) and the Helsinki University of Technology model (HUT n-layer), based on in situ radiometric measurements. Measurements of the SSA, using snow reflectance in the short-wave infrared, were taken at 20 snowpits in various environments (e.g., grass, tundra, and dry fen). The results show that both models required a scaling factor for the SSA values to minimize the root-mean-square error between the measured and simulated brightness temperatures. For DMRT-ML, the need for a scaling factor is likely due to the oversimplified representation of snow as spheres of ice with a uniform radius. We hypothesize that the need for a scaling factor is related to the grain size distribution of snow and the stickiness between grains. For HUT n-layer, using the SSA underestimates the attenuation by snow, particularly for snowpacks with a significant amount of depth hoar. This paper provides a reliable description of the grain size for DMRT-ML, which is of particular interest for the assimilation of satellite passive microwave data in snow models.

Evaluation of Spaceborne L-Band Radiometer Measurements for Terrestrial Freeze/Thaw Retrievals in Canada

The landscape freeze/thaw (F/T) state has an important impact on the surface energy balance, carbon fluxes, and hydrologic processes; the timing of spring melt is linked to active layer dynamics in permafrost areas. L-band (1.4 GHz) microwave emission could allow the monitoring of surface state dynamics due to its sensitivity to the pronounced permittivity difference between frozen and thawed soil. The aim of this paper is to evaluate the performance of both Aquarius and soil moisture and ocean salinity (SMOS) L-band passive microwave measurements using a polarization ratio (PR)-based algorithm for landscape F/T monitoring. Weekly L-band satellite observations are compared with a large set of reference data at 48 sites across Canada spanning three environments: 1) tundra; 2) boreal forest; and 3) prairies. The reference data include in situ measurements of soil temperature (Tsoil) and air temperature (Tair), and moderate resolution imaging spectroradiometer (MODIS) land surface temperature (LST) and snow cover area (SCA) products. Results show generally good agreement between L-band F/T detection and the surface state estimated from four reference datasets. The best apparent accuracies for all seasons are obtained using Tair as the reference. Aquarius radiometer 2 (incidence angle of 39.6°) data give the best accuracies (90.8%), while for SMOS, using the Aquarius temporal characteristics, the best results (87.8% of accuracy) are obtained at higher incidence angles (55°-60°). The F/T algorithm identifies both freeze onset and end with a delay of about 1 week in tundra and 2 weeks in forest and prairies, when compared to Tair. The analysis shows a stronger F/T signal at tundra sites due to the typically clean transitions between consistently frozen and thawed conditions (and vice versa) and the absence of surface vegetation. Results in the prairies were poorer because of the influence of vegetation growth in summer (which decreases the PR) and the high frequency of ephemeral thaw events during winter. Freeze onset and end maps created from the same algorithm applied to SMOS and Aquarius measurements characterize similar F/T patterns over Canada. This study shows the potential of using L-band spaceborne observations for F/T monitoring, but underlines some limitations due to ice crusts in the snowpack, liquid water content in snow cover during the spring freeze to thaw transition, and vegetation growth.

Evaluation of CLASS 2.7 and 3.5 Simulations of Snow Properties from the Canadian Regional Climate Model (CRCM4) over Québec, Canada*

AbstractSnow cover simulations from versions 2.7 and 3.5 of the Canadian Land Surface Scheme (CLASS) coupled to the Canadian Regional Climate Model, version 4 (CRCM4), are evaluated over northern Quebec and the larger Quebec domain using in situ and remotely sensed datasets. Version 2.7 of CLASS has been used in the operational version of CRCM4 at Ouranos since 2006. Version 3.5 includes a number of improvements to the snow processes as well as a more realistic parameterization of snow thermal conductivity. The evaluation shows that version 3.5 provides improved simulations of snow water equivalent, density, depth, and snowpack temperature values. However, snowpack density still contains systematic biases during the snow season that need to be addressed. The snow albedo parameterization in CLASS was found to be very sensitive to an empirical snowfall rate threshold for albedo refreshment and does not keep track of the snow accumulation history in estimating the snow surface albedo. A modified albedo schem...

Relationship Between Forest Microwave Transmissivity and Structural Parameters for the Canadian Boreal Forest

This letter proposes relationships between boreal forest microwave transmissivity and four forest structural parameters: summer and winter Leaf Area Index (LAI) from MODIS, biomass (t ha-1), and total volume (m3 ha-1) for northern Québec, Canada. These relationships were derived for summer AMSR-E data sets that took into account the effects of canopy emission and scattering. Root mean square error results between brightness temperature simulations and ASMR-E observations are approximately 5 K. Results reported in this study can be used as forest correction equations for key surface parameter retrievals under the boreal forest canopy, such as soil moisture or snow depth/water equivalent.

Microwave snow emission modeling of boreal forest environments

This study focuses on the coupling of the Canadian land surface scheme (CLASS) and a radiative transfer model (RTM) to simulate brightness temperature (TB) in boreal forest environments. A parameterization of winter vegetation emission and transmission is proposed to improve the winter TB simulations. When compared to AMSR-E observations, RMSE values of 8K at 37 GHz, 5 K at 19 GHz and 3 K at 10 GHz were obtained. It is also shown that ice crusts have an important impact on the TB in H-pol at the three frequencies. Despite the low sensitivity of microwaves in dense forest, the study provides an interesting tool for the assimilation of TB for SWE retrieval in CLASS.

Retrieval of snow parameters from L-band observations - application for SMOS and SMAP

Recent theoretical and experimental studies have indicated the feasibility of passive microwave L-band observationsfor observing dry snow cover characteristics, namely snow density in the lower approx.. 10 cm of the snowpack. The sensitivity of L-band emission to snow density is based on the dual influence of refraction and impedance matching on observed brightness temperature with changing effective snow permittivity. The permittivity of pure, dry snow, on the other hand, depends largely on snow density. In this study, we expand the theoretical and experimental results of retrieving dry snow density to passive L-band satellite observations. Such retrievals could be appealing in the context of improving satellite based retrievals of e.g. Snow Water Equivalent (SWE) using other sensors. Retrievals are applied to both multi-angular observations from the ESA SMOS mission, and observations of the NASA SMAP radiometer on a single angle of observation. While in theory the multi-angular approach is preferable, improved RFI mitigation in SMAP provides more spatially and temporally more stable retrievals. The applied dual-parameter retrieval scheme produces also an estimate of ground permittivity; experimental data showed dry snow cover to have a clear influence on ground permittivity retrievals, implicating that even dry snow cover is non-negligible also in retrievals of soil moisture from L-band observations.

Creation of a Lambertian Microwave Surface for Retrieving the Downwelling Contribution in Ground-Based Radiometric Measurements

Downwelling radiative fluxes from inhomogeneous environments are not easy to quantify from the total measured signal contribution. We propose the use of a Lambertian reflector to assess this issue. Using ray tracing to extrapolate the bidirectional reflectance distribution function of a surface with high reflectivity in the microwave spectrum, we characterized a surface that acts as a near-Lambertian reflector at 37 GHz. After implementing such a plate using molded aluminum and getting its emissivity, we validated its Lambertian reflectance properties based on microwave surface-based radiometer measurements. This analysis also gives guidelines on how to make a Lambertian surface for different wavelengths in the microwave spectrum. Field experiments show the usefulness of such a Lambertian plate to estimate the downwelling contribution in ground-based radiometric measurements.

Analyse de l’identification de la fonte de neige printanière avec QuickSCAT dans le Sud du Québec, Canada

Snowmelt detection during springtime is a major issue for dam management in northern areas. The Seawinds scatterometer onboard QuikSCAT has proven to be a useful tool to identify liquid water within the snowpack. The study presents an analysis of snowmelt detection using QuikSCAT data in southern Québec, Canada, with the optimization of a backscattering coefficient threshold on every grid point (0.1° grid for years 2001–2007). The threshold considers interpolated air temperatures from local meteorological stations used as reference. Throughout the studied period (2001–2007), the mean accuracy for snowmelt detection is 94%. Nevertheless, the detection of snowmelt is less accurate over open crop land (no forest), even if the effect of forest cover fraction variations on the backscattering coefficient is negligible. Based on an observed relationship between the winter mean backscattering coefficient and the optimized threshold, a simple method for snowmelt detection is proposed. This method makes use of a dynamic empirical threshold linked to the mean backscattering coefficient by a linear function during winter prior to snowmelt. The proposed method provides a mean detection accuracy of 86.0% (2001–2007), better than a method based on a variable threshold with fixed bias. This dynamic threshold approach has the advantage of accounting for strong interannual snow condition variability. This can be shown by the effect of winter mean temperatures driving the surface conditions specific to every winter, which have a significant impact on the mean backscattering coefficient.