Keywords: Snow cover dynamics, growing season, northern hemisphere climate, interannual climate variability, terrestrial carbon cycle, remote sensing.

Abstract
This paper examines interannual variability and trends in the timing and duration of the annual snow-free period in northern high latitude land areas and associated changes in vegetation-absorbed photosynthetically active radiation (APAR). Three satellite-derived time-series data sets were employed in the analysis: (1) NSIDC's Northern Hemisphere EASE-Grid Weekly Snow Cover product for years 1971 to 1994, (2) normalized difference vegetation index (NDVI) data from the Advanced Very High Resolution Radiometer (AVHRR), and (3) incident PAR data from the Nimbus-7 Total Ozone Mapping Spectrometer (TOMS). Over the 24-year study period in land areas between 41° N and 75° N: (a) snow cover disappearance in spring advanced by an average of 5.0 days per decade, (b) snow cover onset in autumn became delayed by an average of 4.5 days per decade, and (c) the duration of the snow-free period has increased by an average of 8.7 days per decade. These results represent the average of statistically significant trends observed within 1° latitudinal zones. The observed spatial and temporal changes in snow cover are associated with significant increases in total APAR for the snow-free season. The increased APAR provides a basis for a mechanistic explanation of the reported "greening" of northern high latitude land areas.

Introduction
Snow cover over continental land areas increases the surface albedo by 30-80%. The increased albedo affects the Earth's surface energy budget and poses a feedback on climate (Cohen and Rind, 1991; Harvey, 1988). The presence or absence of snow cover is thus a critical factor in determining boundary conditions for global atmospheric circulation models (Loth et al., 1993). Snow cover has an important influence on plant distributions and ecosystem functioning through its affect on the soil and near-surface air temperature regimes (Galen and Stanton, 1995; Walker et al., 1993; Bonan, 1992; Uemura, 1989) and by limiting the photosynthetically active radiation (PAR) flux to surface vegetation.

Previous studies of snow cover dynamics at continental or global scales have focused principally on variability or trends in the spatial extent of snow cover (e.g., Robinson et al., 1995; Bamzai and Kinter, 1997). Groisman et al. (1994) analyzed time-series of satellite-derived snow cover maps for the Northern Hemisphere and found that the annual spatial extent of snow cover has declined by approximately 10% in recent decades. Less attention has been given to variability or trends in the annual timing of snow cover, particularly at continental and global scales (Foster, 1989). The timing of snow cover disappearance in spring and snow cover onset in autumn determines the duration of the annual snow-free period, an important factor controlling the duration of the active growing season. Thus, in addition to direct feedback on climate, changes in the timing and/or duration of the annual snow-free period may affect climate indirectly by curtailing or enhancing annual terrestrial primary production and associated biosphere-atmosphere carbon exchange.

Improved information about the interannual dynamics of snow cover and its affect on terrestrial primary production is important for reliable modeling of past, present and future global climate. This research was designed to contribute toward that goal by 1) examining the interannual variability in the timing and duration of snow cover in high northern latitude land areas, and 2) investigating the affect of the observed patterns on vegetation-absorbed PAR (APAR), a fundamental of determinant of net primary production (e.g., Field, 1991).

Data and Methods

Data Sources
The Northern Hemisphere EASE-Grid Weekly Snow Cover and Sea Ice Extent data product was obtained from the EOS Distributed Active Archive Center (DAAC) at the National Snow and Ice Data Center (NSIDC), University of Colorado, Boulder, CO. The data set was created by NSIDC from a digital version of the Weekly Northern Hemisphere Snow Charts from NOAA-NESDIS, after corrections by D. Robinson (Rutgers University). The original analog version of the NOAA-NESDIS snow charts were created through manual interpretation of visible-band images acquired by polar-orbiting and geostationary meteorological satellites. The analog snow charts delineate snow-covered land areas in a polar stereographic map projection. NOAA-NESDIS generated the digital version by overlaying an 89 x 89 grid on the polar stereographic map; grid cells with 50% or greater snow cover were assigned a value of 1, otherwise a value of 0 was assigned. NSIDC created the EASE-Grid snow cover product by mapping the digital NOAA-NESDIS product to new grid in an azimuthal equal area projection with 25 km grid cell resolution. Snow cover data for consecutive weeks from January 1971 through December 1994 were employed in this study.

The affect of changes in snow-free period on APAR was analyzed for the month of May for two sample years, 1983 and 1993. This analysis required two additional data fields: incident PAR, and the fraction of incident PAR absorbed by vegetation (fPAR). Monthly composite NDVI data produced as part of the NOAA/NASA Pathfinder AVHRR Land program were used to estimate fPAR for May of 1983 and 1993. Data on monthly total incident PAR produced in earlier research using ultraviolet reflectivity measurements from the Total Ozone Mapping Spectrometer (TOMS) (Dye and Shibasaki, 1995). The 11-year mean (1979-1989) of the monthly total PAR for May was employed to represent incident PAR for both 1983 and 1994.