Deriving Snow Cover Metrics for Alaska from MODIS

Amy Miller,Jiang Zhu,Chuck Lindsay,Tammy Wilson,Peter Kirchner

doi:10.3390/rs71012961

Abstract

Moderate Resolution Imaging Spectroradiometer (MODIS) daily snow cover products provide an opportunity for determining snow onset and melt dates across broad geographic regions; however, cloud cover and polar darkness are limiting factors at higher latitudes. This study presents snow onset and melt dates for Alaska, portions of western Canada and the Russian Far East derived from Terra MODIS snow cover daily 500 m grid data (MOD10A1) and evaluates our method for filling data gaps caused by clouds or polar darkness. Pixels classified as cloud or no data were reclassified by: spatial filtering using neighboring pixel values; temporal filtering using pixel values for days before/after cloud cover; and snow-cycle filtering based on a time series assessment of a pixel’s position within snow accumulation, cover or melt periods. During the 2012 snow year, these gap-filling methods reduced cloud pixels from 27.7% to 3.1%. A total of 12 metrics (e.g., date of first and last snow, date of persistent snow cover and periods of intermittence) for each pixel were calculated by snow year. A comparison of MODIS-derived snow onset and melt dates with in situ observations from 244 weather stations generally showed an early bias in MODIS-derived dates and an effect of increasing cloudiness exacerbating bias. Our results show that mean regional duration of seasonal snow cover is 179–311 days/year and that snow cover is often intermittent, with 41% of the area experiencing ≥2 snow-covered periods during a snow season. Other regional-scale patterns in the timing of snow onset and melt are evident in the yearly 500 m gridded products publically available at http://static.gina.alaska.edu/NPS_products/MODIS_snow/.

Highlights

There is a pressing need for better understanding of snow cover and snow season dynamics at northern high latitudes due to the region’s tendency toward greater temperature variability (i.e., Arctic amplification) [1] and its importance to global climate [2]
We assessed the accuracy of these metrics using seasonally-paired dates of snow onset and snow melt from 244 in situ locations throughout the area, and we present and describe these temporal variables in a spatial context
In the subset for the 2012 snow year, the percentage of pixels in the study region classified as cloud was reduced from 27.7% to 3.1% (−24.6%) following spatial and temporal filtering

Summary

Introduction

There is a pressing need for better understanding of snow cover and snow season dynamics at northern high latitudes due to the region’s tendency toward greater temperature variability (i.e., Arctic amplification) [1] and its importance to global climate [2]. Warmer winter temperatures and reduced cold-season variability have been observed in the mid- and high latitudes of the Northern Hemisphere in recent decades [3,4]. Winter precipitation patterns in the mid- and high northern latitudes can be highly variable and long-term trends difficult to detect. Challenges to identifying trends in snow data include sparsely distributed observations, a lack of long-term records and data discontinuities, low snowfall amounts, redistribution of snow by wind, and repeated snow on-off events [7]. Snow cover monitoring is one way that we can assess spatial and temporal distributions of precipitation regionally

Objectives

Methods

Results

Conclusion