Snow remote sensing

Remote sensing of snow involves detection and quantification of falling snow, assessment of snowcover on the land, and the snow water equivalent (SWE) of the snowpack. Detection of falling snow and assessment of the snow-pack require different approaches and remote sensing strategies. Quantification of falling snow has real-time applications for transportation, construction, agriculture, and commerce, and snowcover data are used for flood forecasting, water resources management and planning, and to test hydroclimatic models.

5.8.1 Snowfall remote sensing

Falling snow absorbs and scatters electromagnetic energy in the same manner as rainfall and other forms of precipitation, and snow is detected by meteorological radars along with other forms of precipitation. However, snowfall is more effective at scattering than reflecting radiation. The reflectivity of light snow is in the range of 5-20 dBZ, which is much less than the reflectivity of other forms of precipitation. The low radar reflectivity of falling snow results from the small radar cross-section of an ice sphere relative to that of a water sphere of the same size. The greater reflectivity of a water sphere is attributed to differences in the composition of the two hydrometeors. Weather radars must be sensitive to low-energy echoes to detect snowfall. WSR-88D radars in the United States operating in the clear air mode provide the high-resolution retrieval necessary to detect snowfall. However, accurate estimates of snowfall rates and snow accumulations continue to present challenges.

Snowfall rates are estimated using meteorological radars and the relationship between precipitation rate and radar reflectivity shown in Equation 5.8. However, for snow there are many natural causes that alter the nature of the Z-R relationship. An equivalent or effective reflectivity factor Ze is defined for snowfall that assumes the backscattering particles are all spherical drops (WMO, 1996). This assumption is necessary because deviations in the Z-R relationship are due to the assortment of shapes and densities of snow particles, the alterations in snow crystal structure that change the snow particle size distributions and fall speeds, and changes in attenuation of the radar with changes in snow crystal temperature (Boucher and Wieler, 1985). Another characteristic that can be a significant problem in determining the Z for snow is the condition known as the "bright band'' which results from reflectivities within a melting layer. Precipitation in the form of snow exists above the bright band, melting snow occurs in the band, and rain is present below the band. Radar detection of the bright band is a function of both the physical properties of the melting layer and the radar sensitivity and spatial resolution (Yuter, 2003). The bright band phenomenon can produce an appreciable increase in Z indicating a greater snowfall rate than actually exists. The occurrence of the bright band may vary considerably during a storm and from storm to storm and increases the difficulties in interpreting radar measurements. The bright band problem is a particular concern for estimating snowfall from stratiform clouds in which reflectivity usually decreases with height (Boucher and Wieler, 1985, WMO, 1996).

Different polarization approaches developed to improve radar measurements of rainfall are of limited use for snowfall measurement because larger snowflakes exhibit weak polarization signatures. Reflectivity measurements taken simultaneously at two wavelengths by dual-wavelength radar methods provide additional information about characteristic snowflake size, and experimental data indicate this approach produces more accurate snowfall measurements than traditional single-beam radar measurements (Matrosov, 1998).

Snow accumulation is the product of the snowfall rate and the physical processes controlling the depth of snowfall via the snow density. Snowfall density is related to the ice-crystal structure and expresses the relative proportion of the crystal volume composed of air. Numerous cloud microphysical processes contribute to the final structure of an ice crystal. The WSR-88D radar uses snow density as an input parameter in the snow accumulation algorithm that was added to the original radar product generator in response to users' needs for snowfall detection and accumulation estimates (Crum et al., 1998). The concept of a fixed snow ratio is typically applied, often with some local empirical adjustment, even though the traditional 10-to-1 rule is recognized as an inadequate characterization of the true range of snow densities. Roebber et al. (2003) propose a 10-member ensemble of artificial neural networks utilizing surface and radiosonde data to provide more accurate estimates of snowfall density. In mountainous terrain, snowfall density issues are exacerbated by radar beam degradation by topography. Wetzel et al. (2004) describe a system of mesoscale models, surface observations, and multispectral satellite data used coincidentally with radar data to improve snowfall estimates in the northern Colorado Rocky Mountains.

Satellite measurements of snowfall within the atmosphere have been accomplished over oceans, but snowfall retrievals over land are elusive due to the presence of snowcover on the land. A number of studies recommend the use of radiometers operating at frequencies of 150 GHz or higher to improve ability to discriminate precipitation from snowcover. Radiation at these frequencies can measure snowfall over land because water vapor screening obscures underlying snow-covered surfaces. High-frequency AMSU microwave data from NOAA polar-orbiting satellites have been analyzed extensively as a potential means for discriminating the scattering features over land surfaces, especially snow-cover, from those of ice crystals in the atmosphere. A snowfall retrieval algorithm based on AMSU measurements at 150 GHz and higher and sounding channel measurements near 54 GHz developed by Kongoli et al. (2003) detects rainfall and snowfall over cold and snow-covered surfaces, and the algorithm is operational for NOAA satellites.

5.8.2 Snowpack remote sensing

Snow on the ground responds to a number of regions of the electromagnetic spectrum that offer opportunities for remote sensing applications by both active and passive sensors. The SARon ESA's ERS satellites provides spatial resolution suitable for snowcover assessment in medium to small watersheds. However, SAR data have complex radiometry and geometry that complicate use of the data in high-relief areas. Nevertheless, SAR data are used for snow mapping in mountainous areas and wet snow conditions (Seidel and Martinec, 2004).

Fig. 5.11. NASA Terra satellite Moderate Resolution Imaging Spectroradiometer (MODIS) image for 13 March 2006 showing snowcovering most of Europe east of Germany and Denmark. (Image courtesy of NASA and the Goddard Space Flight Center from their website at

Fig. 5.11. NASA Terra satellite Moderate Resolution Imaging Spectroradiometer (MODIS) image for 13 March 2006 showing snowcovering most of Europe east of Germany and Denmark. (Image courtesy of NASA and the Goddard Space Flight Center from their website at

Passive microwave applications are more commonly used for examining both snow depth and SWE. The high albedo of the snow surface compared to non-snow areas is a property easily identified using the visible bands of the GOES, Meteosat, Terra, and Aqua satellite data (Fig. 5.11), but these visible sensors are unable to see the snowcover at night or when clouds are present. Visible-wavelength satellite maps of Northern Hemisphere snow extent produced by NOAA since the late 1960s are the longest consistently derived satellite record of any environmental variable (Robinson and Frei, 2000; Hall et al., 2001). IR sensors are able to make both day and night observations, but clouds interfere with observing the snowcover. Microwave sensors on satellites provide an all-weather snowcover observation capability day or night, and they have been used to map snow extent and snow depth since the early 1970s (Fig. 5.12).

The physical characteristics of the snowpack described in Section 4.5 determine its microwave properties. Snow depth and water equivalent, liquid water

Fig. 5.12. Composite image of global snowcover for February 2004 from the Terra satellite Moderate Resolution Imaging Spectroradiometer (MODIS). (Image courtesy of NASA and the Goddard Space Flight Center from their website at http://

content, density, grain size and shape, temperature, stratification, and soil roughness and dielectric properties of the surface beneath the snowpack all influence emitted microwaves from the snow surface. The extent of microwave scattering within a snowpack is proportional to the snowpack thickness and density, and the brightness temperature of the snow surface is related to these properties. The naturally emitted microwave brightness temperature of a snow-pack is primarily related to the number, size, and packing of snow grains along the emission path. Consequently, deeper snowpacks generally result in lower brightness temperatures. Liquid water within a snowpack alters the emissivity of snow and produces a brightness temperature significantly higher than a dry snowpack. Early morning satellite overpasses of snow-covered areas are preferred for retrieving snowcover information that minimizes wet snow influences (Schmugge et al., 2002).

Brightness temperatures from different channels of satellite sensors are used to estimate the snow depth. A commonly used algorithm developed by Chang et al. (1987) for estimating snow depth (SD) using microwave observations is

where co is a coefficient determined from radiative transfer model experiments of snow and has a value of 1.59 cm K-1, Tb18 is the brightness temperature in K at 18 GHz, Tb36 is the brightness temperature in K at 36 GHz, and SD is expressed in units of cm. Coefficients are usually developed for specific regions and snow-cover conditions, but Kelly et al. (2003) modified Equation 5.9 by using a brightness temperature difference between 19 and 37 GHz to minimize the snow temperature effect and by adjusting the coefficient to reflect how the snow grain size might vary temporally. These modifications achieved encouraging results in applying the algorithm to estimating Northern Hemisphere snow depth during the 2000-1 winter season.

The AVHRR and the AMSU on NOAA polar-orbiting satellites, the SSM/I on DMS satellites, the Landsat multispectral scanner system, the SPOT multispec-tral scanner, MODIS on the EOS Terra and Aqua satellites, and the MediumResolution Imaging Spectrometer (MERIS) on the ESA Environmental Satellite (ENVISAT) are among the many microwave sensors used for mapping continental-scale seasonal snowcover for the Northern Hemisphere (Ferraro et al., 1996; Romanov et al., 2000; Schmugge et al., 2002; Kelly et al., 2003). Seidel and Martinec (2004) list the satellites, the sensors, and the image characteristics of snowcover products. Spatial resolution of snowcover mapping is an area of particular concern for hydroclimatic applications and a variety of spatial resolutions are available for mapping small watersheds to regional-scale applications.

Passive microwave sensors on satellites provide global SWE observations to complement snowcover observations. Brightness temperatures are used to determine SWE using a generalized relation similar to Equation 5.9. Snow density must be known from in situ observations or estimated. A representative value for mature mid-winter snow packs in North America is 300 kg m~3, which produces a coefficient of 4.8mmK~1 for Equation 5.9 (Foster et al., 2005). Vegetation cover and snow grain size variability are the main sources of error in remote sensing of SWE, and Foster et al. (2005) suggest an algorithm based on Equation 5.9 that employs two time and space varying coefficients to account for the effects of vegetation cover and snow morphology in North America. This algorithm captures the accumulation and ablation phases of the snow season over a variety of snow surfaces. Microwave data from the Scanning Multichannel Microwave Radiometer (SMMR) on the Nimbus-7 satellite, the SSM/I, and the AMSR on the EOS Aqua satellite are used to estimate SWE of the snowpack and the presence of liquid water in the snowpack (Kelly et al., 2003; Foster et al., 2005). The spatial resolution of 25 km for the SMMR and SMM/ I is best suited for regional and large basin studies, but the spatial resolution of 500 m for AMSR supports analysis of smaller watersheds. However, evaluation of SWE at spatial scales required to characterize the spatial variability typically found in mountain areas is seldom strong enough to be used in a predictive capacity for mountainous terrain (Anderton et al., 2004).

Aircraft equipped with gamma-radiation detectors provide an alternative method for estimating the average SWE of a site. Gamma radiation from the soil is attenuated by a snow layer. Aircraft flights along a prearranged line before and after snow occurs determine the attenuation due to the snow layer. An empirical equation relates the attenuation to the SWE. This technique is effective in open and relatively flat terrain, but it is less effective in hilly or forested terrain. Nevertheless, this technique is used operationally in a number of countries (Engman, 1993; Ward and Robinson, 2000).

Was this article helpful?

0 0
Waste Management And Control

Waste Management And Control

Get All The Support And Guidance You Need To Be A Success At Understanding Waste Management. This Book Is One Of The Most Valuable Resources In The World When It Comes To The Truth about Environment, Waste and Landfills.

Get My Free Ebook

Post a comment