Snow cover variability in a forest ecotone of the Oregon Cascades via MODIS Terra products

- The western hemlock-true fir ecotone was mapped in the Oregon Western Cascades.
- Three snow cover metrics were calculated from MODIS Terra data for 2000-2014.
- Melt season snow cover was characterized by high interannual variability.
- PDO/ENSO state explained a small fraction of the melt season snow cover variability.
- Power analysis indicated number of years required to distinguish trend from noise.

Snow cover pattern and persistence have important implications for planetary energy balance, climate sensitivity to forcings, and vegetation structure, function, and composition. Variability in snow cover within mountainous regions of the Pacific Northwest, USA is attributable to a combination of anthropogenic climate change and climate oscillations. However, snow covered areas can be heterogeneous and patchy, requiring more detailed mapping of snow trends to understand their potential influences on montane forests. We used standard daily MODIS snow products (MOD10A1.5) to investigate the 15-year record (2000-2014) of snow cover in the predominant forest ecotone of the Oregon Western Cascades. We modeled the ecotone using field data from the H.J. Andrews Experimental Forest, and only considered forested MODIS Terra pixels located within the mapped ecotone of a five-county region. Three snow cover metrics were developed using both binary and fractional snow cover data: mean ecotone snow cover percent, number of snow covered days during the melt season, and day of snow disappearance. Snow cover and depletion dates exhibited large interannual variability and no significant linear trends. This variability is likely influenced by the preceding wintertime states of the Pacific Decadal Oscillation (PDO) and the El Niño/Southern Oscillation (ENSO), which tend to covary. We improve and generalize existing methods for power analysis of trend estimation and quantify the number of uninterrupted observations of the snow metrics that would be needed to distinguish trends of different magnitudes from noise variance, taking possible autocorrelation into account. Sensitivity analyses of the results to some of our heuristic choices are conducted, and challenges associated with optical remote sensing of snow in a dense montane forest are discussed.