The Meteorological Service of Canada has developed a series of operational snow water equivalent (SWE) retrieval algorithms for central Canada, based on the vertically polarized difference index for the 19 and 37 GHz channels of the Special Sensor Microwave/Imager (SSM/I). Separate algorithms derive SWE for open environments, deciduous, coniferous and sparse forest cover. A final SWE value represents the area-weighted average based on the proportional land cover within each pixel. In this study, 5 day averaged (pentad) passive-microwave-derived SWE imagery for the winter season (December–February) of 1994/95 is compared to in situ data from central Canada in order to assess algorithm performance. Investigation of regions with varying proportional land cover within the four algorithm classes shows that retrieved SWE remains within ±10–20mm of surface observations, independent of fractional within-pixel land cover. Following algorithm evaluation, ten winter seasons (1988/89 through 1997/98) of pentad central North American SWE imagery are subjected to a rotated principal-component analysis (PCA). Although there are no trends in total study-area SWE, the PCA results identify the interseasonal variability in the SWE accumulation and ablation centers of action through the SSM/I time series.

The Mackenzie GEWEX Study (MAGS) is a Canadian investigation that has the objective of understanding and modelling the water and energy cycles and their roles in the climate system in the high-latitude Mackenzie River basin, including assessing potential changes due to climate variability and change. The Climate Research Branch (CRB) of the MeteorologicalService of Canada has investigated snow-cover variations over the MAGS region using snow water equivalent (SWE) datasets derived from Special Sensor Microwave/Imager (SSM/I) passive-microwave satellite data for the winter seasons 1988–98. The SWE datasets were derived using four CRB algorithms for prairie, coniferous-forest, deciduous-forest and sparse-forest land-cover types and then evaluated against available in situ SWE measurements for MAGS subbasins. Overall, the SWE algorithms produce reliable estimates (within 10–20mm of in situ SWE measurements) for the validated part of the MAGS region, although some areas exhibit underestimations of > 30 mm, which may be due to the presence of a high density of lakes or a decreased microwave sensitivity to high-SWE conditions (>100mm). A time-series dataset of SSM/I-derived SWE for1 March of each year from 1988 to 1998 has been produced as a MAGS deliverable, which provides important information on the spatial and temporal variability in snow cover over the Mackenzie River basin. This dataset has been used in the assessment of snow-cover outputs from MAGS hydrological and climate-modelling investigations.

For the American northern Great Plains region, the 1996/97 snow season had snow accumulations much greater than normal, which combined with rapid warming to produce extensive flooding in the Red River of the North river basin. Passive-microwave observations from the Special Sensor Microwave/Imager (SSM/I) are used to follow the evolution of the snowpack during the snow season and to map the extent of standing water or very saturated soils during spring 1997. SSM/I-derived snow-depth algorithms that assume a fixed snow grain-size constantly underestimated the snow depth by a factor of 2 in the region where extensive flooding occurred. An estimate of the thermal gradient through the snowpack is used to model the growth of the snow grain-size and to compute more accurately the evolution of the snow depth over the region. As is commonly observed, when the melt season begins, liquid water in the snowpack causes the SSM/I spectral gradient to drop to zero. In this case, the spectral gradient fell to unusually negative values, which were indicative of large areas of open water, and not wet snow or soil.

There are several hemispheric-scale satellite-derived snow-cover maps available, but none has been fully validated. For the period 23 October–25 December 2000, we compare snow maps of North America derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) and operational snow maps from the U.S. National Oceanic and Atmospheric Administration (NOAA) National Operational Hydrologic Remote Sensing Center (NOHRSC), both of which rely on satellite data from the visible and near-infrared parts of the spectrum; we also compare MODIS maps with Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave/Imager (SSM/I) passive-microwave snow maps for the same period. The maps derived from visible and near-infrared data are more accurate for mapping snow cover than are the passive-microwave-derived maps, but discrepancies exist as to the location and extent of the snow cover even between operational snow maps. The MODIS snow-cover maps show more snow in each of the 8 day periods than do the NOHRSC maps, in part because MODIS maps the effects of fleeting snowstorms due to its frequent coverage. The large (~30 km) footprint of the SSM/I pixel, and the difficulty in distinguishing wet and shallow snow from wet or snow-free ground, reveal differences up to 5.33 x 106 km2 in the amount of snow mapped using MODIS vs SSM/I data. Algorithms that utilize both visible and passive-microwave data, which would take advantage of the all-weather mapping capability of the passive-microwave data, will be refined following the launch of the Advanced Microwave Scanning Radiometer (AMSR) in the fall of 2001.

We developed a model to calculate the spectral reflectance of snow by normalizing the topographic effect and correcting the atmospheric effect simultaneously. This model considers the shadowing effect of the surrounding terrain on both direct and diffuse irradiance by using the obstruction coefficient in sun-ray direction and introducing the isotropic view factor. In addition, a new method to calculate the surrounding-reflected irradiance by introducing the shape factor which can account for the radiation reflected from the surrounding pixels was developed. The model was tested in the upper stream of the Heihe river basin, Qilian mountains, China, using a subscene of a Landsat Enhanced Thematic Mapper Plus (ETM+) image acquired on 20 April 2000. The result showed that the model could eliminate most of the shadowing effect of rugged terrain and could estimate snow reflectance correctly. The problem of data saturation when retrieving snow reflectance is also discussed.

Passive-microwave satellite remote sensing can greatly enhance large-scale snow measurements based on visible satellite data alone because of the ability to acquire data through most clouds or during darkness as well as to provide a measure of snow depth or water equivalent. This study provides preliminary results from the comparison and evaluation of several different passive-microwave algorithms. These algorithms represent examples which include both mid- and high-frequency channels, vertical and horizontal polarizations and polarization-difference approaches. In our comparisons we utilize larger, more comprehensive, validation datasets which can be expected to provide a full range of snow/climate conditions rather than limited data which may only represent a snapshot in time and space. Evaluation of snow extent derived from passive-microwave data is undertaken through comparison with the U.S. National Oceanic and Atmospheric Administration (NOAA) Northern Hemisphere snow charts which are based on visible-band satellite data. Results clearly indicate those time periods and geographic regions where the two techniques agree and where they tend to consistently disagree. Validation of snow water equivalent derived from passive-microwave data is undertaken using measurements from snow-course transects in the former Soviet Union. Preliminary results indicate a general tendency for nearly all of the algorithms to underestimate snow water equivalent.

A prototype snow albedo algorithm has been developed for the Moderate Resolution Imaging Spectroradiometer (MODIS). It complements existing MODIS products by providing albedo measurements for areas mapped as snow on a global daily basis by MODIS. Cloud detection and atmospheric correction are accomplished using existing MODIS products. Models of the bidirectional reflectance of snow created using a discrete-ordinate radiative transfer (DISORT) model are used to correct for anisotropic scattering effects over non-forested surfaces. Initial algorithm validation is undertaken through comparisons with broadband albedo measurements made at the U.S. National Oceanic and Atmospheric Administration (NOAA) Surface Radiation Budget Network (SURFRAD) site in Fort Peck, MT. In situ SURFRAD albedo measurements are compared to daily MODIS snow albedo retrievals for the period 21–26 November 2000 created from five narrow-to-broadband albedo conversion schemes. The prototype MODIS algorithm produces reasonable broadband albedo estimates. Maximum daily differences between the five MODIS broadband albedo retrievals and in situ albedo are 15%. Daily differences between the best MODIS broadband estimate and the measured SURFRAD albedo are 1–8%. However, no single conversion scheme consistently provides the closest albedo estimate. Further validation and algorithm development using data from North America and Greenland is ongoing.

With the launch of Landsat 7 a new band was added to the Landsat Enhanced Thematic Mapper Plus (ETM+). The new panchromatic band 8 covers wavelengths of 0.52–0.92μm and extends over band 2 (0.525–0.605 μm), band 3 (0.630–0.690 μm) and band 4 (0.750–0.900 μm). Its high resolution, with a pixel length of 15 m, is of interest for high-spatial-resolution analysis. This project focused on using Landsat 7 band 8 as a precise tool to obtain remotely the size and shape of permanent snowpatches in the mountainous area west of Abisko, Sweden. In the supervised classification employed, shadowed and non-shadowed areas were classified separately. Supervision of the classification was based on ground-truthing consisting of photographic mapping of the snow cover in the study area. The results from this single-band classification were compared to the results from the normalized- difference snow index (NDSI) and false-color images using bands 2, 4 and 5. This comparison showed similar estimates of the total amount of snow-covered area, with the additional advantage that the Landsat 7 band 8 (LB8) classification detects details in the extent of individual snowfields. This new application and the insights documented here support the use of Landsat 7 band 8 for snow-cover investigations in the future.

We evaluated the contribution of Special Sensor Microwave/Imager (SSM/I) passive-microwave data to the monitoring of spatial and temporal variability of snow cover in the Churchill area, Manitoba, Canada. Because of the coarse spatial resolution of current passive-microwave sensors, the estimation of snow water equivalent using empirical equations with these instruments is largely compromised in complex areas such as Churchill (forest–tundra ecotone). However, with its high frequency of observations and the availability of a long time series (1988–99), passive-microwave data from the SSM/I radiometer remain a very valuable tool for monitoring the temporal evolution of snow cover at various spatial scales. Through winter 1997/98, we first examined the passive-microwave signatures at the local scale and we identified the major stages of the snow period. Principal-component analysis (PCA) applied on spectral-difference (Tb(19H) - Tb(37H))time series (1988–99) enabled us to identify spatio-temporal effects over a large area. PCA also permitted the extraction of indices of relevance for monitoring climatic variability and climate change (annual snow-cover duration, dates of snow-cover appearance and disappearance).

For landscapes with a complex topography and a heterogeneous forest mosaic it is not feasible to map the snow depth directly from optical satellite images. In this paper, an indirect method to predict the snow-depth distribution is presented and applied to a 0.7 km2 subalpine catchment in central Switzerland. The method consists of (a) a parsimonious linear regression model which includes the attributes of topography and vegetation indices (derived from a Landsat Thematic Mapper (TM) image) as explanatory variables, and (b) geostatistical interpolationtechniques. A previous analysis of the forest mosaic revealed two main scales showing up in the Landsat TM image and an aerial photograph. This discrepancy in scale was assumed to be the major reason why the vegetation indices derived from the Landsat TM image were only weak explanators of the snow-depth variation measured at 100–200 locations within the catchment. Surprisingly, the geostatistical interpolation (universal kriging) was not able to improve the prediction of the snow-depth distribution significantly. The residuals of the regression model showed hardly any spatial dependence for single snow-measurement dates.

Two spectral snow-reflectance models that account for the effects of grain-size and liquid-water fraction are described and initial validation results presented. The models are based upon the spectral complex refractive index of liquid water and ice in the region from 400 to 2500 nm. Mie scattering calculations are used to specify the essential optical properties of snow in the models. Two approaches are explored to model the effect of liquid water in the snow. The first accounts for the liquid water as separate spheres interspersed with ice spheres in the snow layer. The second accounts for the liquid water as coatings on ice grains in the snow layer. A discrete-ordinate radiative transfer code is used to model the spectral reflectance of the snow for the Mie-calculated optical properties. Both the interspersed- and coated-sphere models show that the snow-absorption feature at 1030 nm shifts to shorter wavelength as the liquid-water content increased. The expression of these shifts is different for the two models. A comparison of the models with a spectral measurement of frozen and melting snow shows better agreement with the coated-sphere model. A spectral fitting algorithm was developed and tested with the coated-sphere model to derive the grain-size and liquid-water fraction from snow spectral reflectance measurements. Consistent values of grain-size and liquid water were retrieved from the measured snow spectra. This research demonstrates the use of spectral models and spectral measurements to derive surface snow grain-size and liquid-water fraction. The results of this research may be extended to regional and greater scales using data acquired by airborne and spaceborne imaging spectrometers for contributions to energy balance and hydrological modeling.

Algorithms to retrieve the snow grain-size and the concentration of snow impurities were developed using a theoretical bidirectional reflectance distribution function (BRDF) model of the snow surface. In this model, snow grains are assumed to be independent spherical ice particles, and the BRDF is calculated with multiple scattering by snow particles. Using these algorithms, the snow grain-size and snow impurities were retrieved from Airborne Multispectral Scanner (AMSS) images at the visible (λ = 0.545 μm) and near-infrared (λ =1.24,1.64 and 2.23 μm) wavelengths observed over the flat snowfield in eastern Hokkaido, Japan, in February 1998. The estimated snow grain-size and impurities were consistent with the results of in situ measurements on the snow surface. For snow grain-size, measured reflectances in the different near-infrared AMSS channels indicated grain-size differences in the vertical profile of the snowpack.

As part of the International Trans-Antarctic Scientific Expedition project, the Italian Antarctic Programme undertook two traverses from the Terra Nova station to Talos Dome and to Dome C. Along the traverses, the party carried out several tasks (drilling, glaciological and geophysical exploration). The difference in spectral response between glazed surfaces and snow makes it simple to identify these areas on visible/near-infrared satellite images. Integration of field observation and remotely sensed data allows the description of different mega-morphologic features: wide glazed surfaces, sastrugi glazed surface fields, transverse dunes and megadunes. Topography global positioning system, ground penetrating radar and detailed snow-surface surveys have been carried out, providing new information about the formation and evolution of mega-morphologic features. The extensive presence, (up to 30%) of glazed surface caused by a long hiatus in accumulation, with an accumulation rate of nil or slightly negative, has a significant impact on the surface mass balance of a wide area of the interior part of East Antarctica. The aeolian processes creating these features have important implications for the selection of optimum sites for ice coring, because orographic variations of even a few metres per kilometre have a significant impact on the snow-accumulation process. Remote-sensing surveys of aeolian macro-morphology provide a proven, high-quality method for detailed mapping of the interior of the ice sheet’s prevalent wind direction and could provide a relative indication of wind intensity.

Snow stratigraphy was analyzed in the Maudheimvidda area of western Dronning Maud Land, East Antarctica, during austral summer 1999/2000 as a part of the Finnish Antarctic Research Programme (FINNARP). Measurements were made in shallow (1–2m) snow pits along a 350 km transect from the coast to the polar plateau, covering at least one annual cycle and an elevation range from sea level to about 2500 m. The aim of the study is to document spatial and temporal variations in snow-cover properties, with the further aim of relating these variations to environmental factors and to patterns observable by remote sensing. The measurements suggest five principal snow zones: (i) sea ice, (ii) the seaward edge zone of the ice shelf, (iii) the inner parts of the ice shelf, (iv) the snow cover above the grounding line and (v) the local topographic highs. Local topographic highs such as ice domes and ice rises differ from other snow environments: the snow is less densely packed, possibly an indication of locally reduced speed of the katabatic outflow. Fewer and thinner crusts on the topographic highs are consistent with RADARSAT backscatter variations.

In order to estimate the contribution of the Antarctic Peninsula to global sea-level rise as a result of the observed warming in this region, the spatial extent of snow-melt-producing areas needs to be quantified. By using the dry-snow line derived from synthetic aperture radar (SAR) imagery as the uppermost limit of frequent or occasional surface melt, an estimation of the spatial extent of areas with non-zero ablation rates is facilitated. Three calibrated RADARSAT ScanSAR mosaics covering the northern Antarctic Peninsula were analyzed applying a threshold of –14 dB to identify the dry-snow line. The area of the dry-snow radar zone was determined to be 23300±2000 km2.Areas affected frequently or occasionally by snowmelt cover 85 000±9000 km2. In addition, the dry-snow line as derived from multi-temporal ERS-1/-2 imagery serves as an indicator of climate variability in the uppermost areas of polar glaciers and ice sheets. The upward shift of the dry-snow line between 1992 and 1998 on the eastern side of the Antarctic Peninsula at 68˚ S is interpreted as a direct response to the increasing number of high-temperature events during the 1991–2000 decade.

Most climate models treat surface and atmospheric properties as being horizontally homogeneous and compute surface radiative fluxes with average gridcell properties. In this study it is found that large biases can occur if sub-gridcell variability is ignored, where bias is defined as the difference between the average of fluxes computed at high resolution within a model cell and the flux computed with the average surface and cloud properties within the cell. Data from the Advanced Very High Resolution Radiometer for the year-long Surface Heat Budget of the Arctic Ocean (SHEBA) experiment are used to determine biases in aggregate-area fluxes. A simple regression approach to correct for biases that result from horizontal variability was found to reduce the average radiative flux bias to near zero. The correction can be easily implemented in numerical models.

We produced four independent temperature time series derived from different sensors for the Siple Dome region of West Antarctica to investigate seasonal to interannual temperature variability over the last 20 years. We use data from automatic weather station air-temperature records (1997–99), European Centre for Medium-range Weather Forecasts surface temperature from the 15 year re-analyses (ERA-15, 1979–93), and emissivity-corrected brightness temperatures from the Special Sensor Microwave/Imager (1987–99) and the Scanning Multichannel Microwave Radiometer (1978–87). Each technique has limitations and errors, but all respond to temperature, and all agree in the large patterns of temperature variability over time. Our results show that there is high seasonal to interannual variability in both mean temperature and variance in the Siple Dome region during the study period. In particular, fluctuations in seasonal to interannual temperature variance occur on an approximately 5 year cycle and correlate with variations in the Southern Oscillation Index.

Air-temperature records (TA) during 1992 from five inland Antarctic automatic weather station (AWS) sites were compared with the best available infrared temperatures (TIR) from the Advanced Very High Resolution Radiometer (AVHRR) as well as calibrated passive-microwave temperatures (TC) from the Special Sensor Microwave/Imager (SSM/I). Daily and monthly average TA, TIR, and TC data indicate that each approach captures generally similar trends at each site but each approach also has limitations. AWS TA data are considered the most accurate but represent spatially restricted areas and may have long gaps due to sensor or transmission problems. AVHRR TIR data have daily variability similar to the TA record but have numerous small gaps due to cloud cover or observation interruptions. An offset between TA and TIR (>4 K) at the South Pole site was identified that may be due to the inclusion of data with large satellite scan angles necessary to cover this area. SSM/I TC data have the most continuity but exhibit calibration problems, a significantly damped daily response and do not cover all of Antarctica. Individual daily differences between TA and TIR as well as TA and TC can exceed 17 K, but all sites have mean daily differences of about 1 Kor better, after compensating for the offset at South Pole, and standard deviations of <6K. Monthly temperature differences are typically 5 K or better, with standard deviations generally <3K. And finally, using the available data, the 1992 average temperature differences are <1 K.

Current techniques of cloud discrimination in polar regions, ice surface temperature measurement, sea-ice and snowfield extent mapping often rely on data acquired in the region from 3 to 5 mm. The Advanced Very High Resolution Radiometer (AVHRR) and the recently launched Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on Terra have spectral bands in this region used for these purposes. Approaches often consider the radiance value in this spectral range in terms of a single equivalent brightness temperature. However, this spectral region contains contributions from both solar-reflected and thermal-emitted radiance, and a water ice reflectance peak at 3.7 μm can be highly variable and a sensitive indicator of grain-size in icy particles either in clouds or as surface snow. In December 1992 the Galileo spacecraft, on its way to Jupiter, flew by and acquired images of Antarctica that included spectral coverage in 408 channels in the wavelength range 1–5 μm with the Near Infrared Mapping Spectrometer (NIMS). The NIMS spectra provide a basis for the separation of the reflected and emitted components in this wavelength region. This separation then allows the examination of the observed variation of the reflected component with respect to cloud and surface ice properties. This analysis may help refine current algorithms for cloud discrimination in AVHRR and MODIS using channels from 3 to 5 μm.

We have examined the record of melt-season duration on the Antarctic Peninsula using two techniques for detecting the presence of a melt signal in microwave-emission time series covering the period 1978–2000. We have obtained similar estimates of melt-season length using the cross-polarized gradient ratio (XPGR) technique and calibrations previously applied in Greenland and a technique which detects the jump in emission caused by melt without using a sensor- and frequency-dependent threshold value. The close correspondence between results from the two techniques on peninsula ice shelves suggests that the XPGR analysis can be used over the length of the time series. The results show that the long melt seasons of 1992/93 and several later years were exceptional occurrences on the northern parts of the Larsen Ice Shelf. These melt seasons were followed by disintegration events, supporting a possible cause-and-effect relationship.