Estimating impacts of snow cover on net ecosystem exchange near Daring Lake, NWT, Canada (65oN, 111oW)

Arctic CO2 flux is influenced by the spatial distribution, accumulation, and phenology of snow. Regions with greater net primary productivity tend to accumulate greater snowfall, and areas receiving greater snowfall have higher soil temperatures, enabling greater rates of winter respiration. During snow melt, the spatial distributions of remaining snow influence the patterns of CO2 release. However, cryospheric influences on CO2 distributions are not typically included in models of net ecosystem exchange. The objective of this research was therefore to investigate: 1) the influence of snow cover distributions on CO2 flux during snow melt at four sites located near Daring Lake, NWT, Canada; and 2) the amount of variation in CO2 flux due to arctic land cover class and snow distributions captured by the remote sensing based Vegetation Photosysthesis and Respiration Model (VPRM), which does not contain explicit arctic land cover class or cryospheric parameterizations. Snow cover distributions were monitored twice daily throughout the 2010 snow melt period (May-June) using Moultrie I-65 game cameras over four arctic land cover classes: fen, tall shrub, short shrub and mixed tundra. At each site, CO2 flux data was collected using an open path infrared gas analyzer eddy covariance system. Initial analysis consisted of classifying snow cover locations and extent in each image. Footprint modeling was then conducted in order to assess the impact of snow cover distributions on CO2 flux during melt. In order to assess the amount of variability in net ecosystem exchange due to snow cover captured by VPRM, associations between vegetation cover, snow extents and CO2 flux were first upscaled using classification of Landsat snow cover and land cover class. VPRM was then run over the same region. VPRM output was statistically compared to upscaled CO2. Although VPRM contains few land surface parameters, it is not hindered by gaps in Landsat data due to cloud cover, or diminished accuracy of open path infrared gas analyzer eddy covariance measurements while snow cover is present. Understanding differences in the results drawn from these approaches is therefore important in designing an optimal method for integrating cryospheric drivers and arctic land cover classes into estimates of arctic net ecosystem exchange.