Regional influence of ocean–atmosphere teleconnections on the timing and duration of MODIS-derived snow cover in British Columbia, Canada

Abstract

We use the twice-daily Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover product to study the regional influence of the Oceanic Niño Index (ONI) and Pacific Decadal Oscillation (PDO) on snow cover in British Columbia (BC). We apply a locally weighted regression (LOWESS) interpolation to the MODIS normalized difference snow index (NDSI) time series to detect the timing and duration of snow. We confirm the general consensus from many previous in situ studies that both ONI and PDO have significant impacts on snow cover in BC. We add to this knowledge by performing seasonal and regional analysis using established hydrozones and explore variation in our results by elevation bins of 500 m. We calibrated our method with in situ snow water equivalent (SWE) data and found an optimal NDSI threshold of 30 for detecting continuous snow cover. We separate automatic snow weather station data into calibration (75 %) and validation (25 %) subsets and obtain mean absolute errors between the MODIS and in situ observations for the start, end, and duration of 8.7, 8.9, and 13.1 d for the calibration data and 12.7, 12.6, and 16.6 d for the validation data, respectively. In general, the start date of snow is poorly correlated with both ONI and PDO, whereas the end date and duration are strongly negatively correlated. Regional patterns emerge wherein northern and southern BC are most correlated with the PDO and the ONI, respectively. These relationships are generally stronger at lower elevations and vary spatially. This study demonstrates that the suitability of ocean–atmosphere teleconnections as predictors of the timing and duration of snow varies throughout BC.