Assessment of spring snow cover duration variability over northern Canada from satellite datasets

https://doi.org/10.1016/j.rse.2006.09.035Get rights and content

Abstract

Variability in northern hemisphere (NH) spring and summer snow extent is strongly influenced by snow cover duration (SCD) across the Canadian Arctic. In order to assess the accuracy with which satellite-derived snow extent datasets capture the dynamic melt period (1 April–31 July), SCD datasets derived from the weekly NOAA snow chart record (1979–2004), daily IMS product (2000–2004), passive microwave (PMW) brightness temperatures (1979–2004), and Ku-band QuikSCAT scatterometer data (2000–2004) were assessed against in situ measurements across the Canadian Arctic (north of 60°). The higher resolution IMS and QSCAT datasets showed the best ability to capture spatial variability in spring SCD over the Canadian Arctic, followed by the NOAA and PMW datasets. The poorer performance of the PMW was anticipated because of documented difficulties monitoring snow cover over forested and mountainous terrain. Both the IMS and the NOAA datasets exhibited positive biases of ∼ 22–26 days which can be related to elevation effects and frequent cloud cover. The NOAA dataset was unable to capture interannual variability in spring snow cover over the central Canadian Arctic tundra region (66–74° N, 80–120° W) while the PMW was able to capture a significant fraction of the observed variability. This fact, combined with anomalous snow-cover temperature responses in the NOAA data further reinforces the conclusions of Wang et al. [Wang, L., Sharp, M., Brown, R., Derksen, C., & Rivard, B., (2005a). Evaluation of spring snow covered area depletion in the Canadian Arctic from NOAA snow charts. Remote Sensing of Environment, 95, 453–463] that considerable care be taken when using NOAA data during the summer months (June–August) where NH snow cover variability is controlled by relatively small regions with frequent cloud cover. QuikSCAT data were able to provide comparable results to IMS over the 2000–2004 period, indicating that Ku-band scatterometer measurements can make an important contribution to monitoring terrestrial snow extent during spring melt.

Introduction

Variability and change in the duration of snow cover during the spring season (or the date of snow cover disappearance) has important consequences for surface energy and water budgets over a range of scales (Serreze et al., 2002, Yang et al., 2003), as well as significant implications for cold region ecosystems (Chapin et al., 2005). A number of studies have documented recent trends toward earlier spring snowmelt across many regions of the NH (e.g. Brown, 2000, Dye, 2002) in response to enhanced spring warming that is due in part to a positive feedback from reduced snow cover (Groisman et al., 1994). The accurate monitoring of spring snow cover variability across the Canadian arctic and subarctic tundra is important because this region is expected to experience the largest cryospheric changes in response to increasing greenhouse gas forcing. In addition, snow cover variations over the high latitudes of North America (NA) contribute significantly to summer season variability in Northern Hemisphere (NH) snow cover extent (Fig. 1), which in turn, appears to play an important role in NH winter climate (Saunders et al., 2003).

The two main datasets for monitoring snow cover over this region of NA in a spatially continuous and temporally consistent manner are (1): weekly snow extent charts produced largely from optical satellite data by NOAA/NESDIS since 1966 (Robinson et al., 1993), and (2) satellite passive microwave data from the Scanning Multichannel Microwave Radiometer (SMMR; 1987–1987) and Special Sensor Microwave/Imager (SSM/I; 1987–present; Chang et al., 1990). The NOAA dataset has been utilized extensively for continental and hemispheric snow cover monitoring, and as input to climatological, hydrological, and numerical modelling studies (for example, Frei et al., 2003, Leathers and Robinson, 1993, Yang et al., 2003). While hemispheric information on snow extent and snow water equivalent (SWE) has been produced from the microwave time series (for example, Armstrong & Brodzik, 2001) most validation and application activities have been regional in nature (for example, Derksen & MacKay, 2006). Both the NOAA and microwave-derived datasets have received only limited evaluation in the Arctic and subarctic tundra regions during the dynamic and highly variable snowmelt period.

Recent studies have shown that there are several advantages of using active microwave Ku-band scatterometer data for snow cover monitoring based on the high sensitivity of these data to surface melt and the limited extent to which they are affected by vegetation cover (e.g. Nghiem and Tsai, 2001, Wang et al., 2005b). Wang et al. (2005b) were able to provide detailed maps of melt season duration over the Canadian high Arctic ice caps from QuikSCAT (QSCAT) Ku-band scatterometer data which suggests there is some potential for high resolution (∼ 5 km) mapping of snowmelt onset over high latitudes.

Wang et al. (2005a) identified several issues with snowmelt detection across the central Canadian tundra in the NOAA snow chart data record. In particular, the NOAA weekly dataset consistently overestimated snow cover extent during the spring melt period, with delays of up to 4 weeks in the identification of snowmelt when compared to snow cover extent estimates derived from passive microwave brightness temperatures and AVHRR imagery. This offset was attributed to several characteristics of the NOAA data record, including the use of a 50% snow covered area threshold to segment snow covered from snow free land, spring cloudiness, reduced availability of satellite imagery as data sources at higher latitudes, limited surface observations for validation, and differing interpretations of patchy snow cover by snow analysts. Reducing the uncertainty in high latitude snow extent and snow duration estimates during the melt period is essential because high latitude snow cover both forces (for example Gong et al., 2003a, Gong et al., 2003b, Watanabe and Nitta, 1999) and responds (for example Bamzai, 2003, Saito and Cohen, 2003) to atmospheric circulation. A number of studies have reported significant correlations between NH spring and summer snow covered area anomalies and subsequent NH winter circulation (Ellis and Hawkins, 2001, Gong et al., 2003a, Hawkins et al., 2002, Saunders et al., 2003) which appears to be related to large-scale wave activity fluxes and impacts on the polar vortex.

The main objective of this study was to carry out a detailed evaluation of snow cover datasets over northern Canada to obtain a clearer understanding of the strengths, and limitations of the various methods for investigating snow cover variability and change in a region that exerts an important influence on NH summer snow cover variability. This evaluation also included an assessment of the capability of the enhanced resolution QSCAT Ku-band scatterometer data to resolve spring snowmelt over the Canadian Arctic. Relationships between snow cover duration (SCD) and air temperature anomalies were also explored to assess and explain inter-dataset variability.

Section snippets

Study region

This paper looked at two regions: (1) the entire Canadian Arctic north of 60° N for evaluation of the ability of datasets to capture the spatial gradients in spring SCD; and (2) a sub-region (66–74 N, 80–120 W) north of the boreal tree line for evaluating interannual variability in spring SCD (the area occupied by first principal component for NH July snow cover extent from the NOAA weekly dataset shown in Fig. 1). This region is above the treeline, contains relatively large land masses

Data

This study used spring snow cover duration (SCD), defined as the number of days with snow cover on the ground from 1 April to 31 July, to investigate snow cover variability over northern Canada during the melt season. SCD is more easily defined than the end-date of continuous snow cover, avoids the problem of multiple end-dates due to late season transient snowfall events, and is strongly correlated (r > 0.8) with the end-date of continuous snow cover over the study domain. Analysis of snow cover

Intercomparison methodology

The first step of the evaluation process involved a comparison of coincident surface and satellite observations of mean spring SCD over the entire Canadian Arctic domain (60–90° N, 60–140° W) shown in Fig. 3. For the higher resolution datasets (passive microwave, IMS and QSCAT), station-satellite pairs were obtained from the nearest satellite pixel following Foster et al. (2005). The average station-grid distances are shown in Table 1 and were on the order of 10 km for the IMS and passive

Comparison of mean SCD at coincident grid points

The summary statistics for the nearest pairings of surface and satellite spring SCD are shown in Table 1. For the 2000–2004 period, the passive microwave, IMS and QSCAT all have similar r2 values but the IMS product had the highest slope (0.70) which means it was best at capturing the observed spatial gradient in SCD over the study domain. The NOAA weekly product had the lowest r2 value which is likely related to the difficulty deriving comparable observed information over the larger grid size

Conclusions and discussion

A comprehensive inter-comparison of melt season SCD datasets derived from satellite data was conducted for high latitude NA for the spring seasons of 1979–2004. Consistently gridded time series of comparable data (days between April 1 and July 31 with snow cover) were produced from the weekly NOAA snow chart record and daily IMS maps, as well as from satellite passive microwave and Ku-band scatterometer data. Careful treatment of in situ observations was required to produce an independent SCD

Acknowledgements

We are grateful to the data providers: The National Snow and Ice Data Center (EASE-Grid satellite passive microwave brightness temperatures and land/sea mask; daily IMS maps); NOAA Climate Prediction Center (daily IMS maps); Rutgers University Climate Laboratory Snow Data Resource Center (NOAA weekly snow charts); Meteorological Service of Canada (in situ snow data); The Microwave Earth Remote Sensing Laboratory, Brigham Young University (the enhanced resolution QuikSCAT data). Thanks to Bob

References (58)

  • R. Armstrong et al.

    Recent Northern Hemisphere snow extent: A comparison of data derived from visible and microwave satellite sensors

    Geophysical Research Letters

    (2001)
  • D. Atkinson et al.

    Canadian cryospheric response to an anomalous warm summer

    Atmosphere–Ocean

    (2006)
  • A. Bamzai

    Relationship between snow cover variability and Arctic oscillation index on a hierarchy of time scales

    International Journal of Climatology

    (2003)
  • B. Bonsal et al.

    Trends and variability in spring and autumn 0°C isotherm dates over Canada

    Climatic Change

    (2003)
  • B. Brasnett

    A global analysis of snow depth for numerical weather prediction

    Journal of Applied Meteorology

    (1999)
  • D. Bromwich et al.

    Evaluation of the NCEP–NCAR and ECMWF 15-and 40-yr reanalyses using Rawinsonde data from two independent Arctic field experiments

    Monthly Weather Review

    (2005)
  • R. Brown

    Northern Hemisphere snow cover variability and change, 1915–97

    Journal of Climate

    (2000)
  • R. Brown

    Snow cover variability and trends in the Mackenzie Basin, 1945–2005

  • R. Brown et al.

    Interannual variability in reconstructed Canadian snow cover, 1915–1992

    Journal of Climate

    (1996)
  • A. Chang et al.

    Satellite sensor estimates of northern hemisphere snow volume

    International Journal of Remote Sensing

    (1990)
  • F. Chapin et al.

    Role of land-surface changes in Arctic summer warming

    Science

    (2005)
  • C. Derksen et al.

    The influence of sensor overpass time on passive microwave retrieval of snow cover parameters

    Remote Sensing of Environment

    (2000)
  • C. Derksen et al.

    The Canadian boreal snow water equivalent band

    Atmosphere–Ocean

    (2006)
  • C. Derksen et al.

    Identification of systematic bias in the cross-platform (SMMR and SSM/I) EASE-Grid brightness temperature time series

    IEEE Transactions on Geoscience and Remote Sensing

    (2003)
  • C. Derksen et al.

    Evaluation of passive microwave snow water equivalent retrievals across the boreal forest/tundra transition of western Canada

    Remote Sensing of Environment

    (2005)
  • D. Dye

    Variability and trends in the annual snow-cover cycle in Northern Hemisphere land areas, 1972–2000

    Hydrological Processes

    (2002)
  • D. Easterling et al.

    Observed climate variability and change of relevance to the biosphere

    Journal of Geophysical Research

    (2000)
  • A. Ellis et al.

    An apparent atmospheric teleconnection between snow cover and the North American monsoon

    Geophysical Research Letters

    (2001)
  • J. Foster et al.

    Quantifying uncertainty in passive microwave snow water equivalent observations

    Remote Sensing of Environment

    (2005)
  • A. Frei et al.

    Northern Hemisphere snow extent: Regional variability 1972–1994

    International Journal of Climatology

    (1999)
  • A. Frei et al.

    Improved simulations of snow extent in the second phase of the Atmospheric Model Intercomparison Project (AMIP-2)

    Journal of Geophysical Research

    (2003)
  • R. Geiger

    The climate near the ground (translated from German).

    (1961)
  • B. Goodison et al.

    Canadian development and use of snow cover information from passive microwave satellite data

  • G. Gong et al.

    Relative impacts of Siberian and North American snow anomalies on the winter Arctic Oscillation

    Geophysical Research Letters

    (2003)
  • G. Gong et al.

    Modeled Northern Hemisphere winter climate response to realistic Siberian snow anomalies

    Journal of Climate

    (2003)
  • P. Groisman et al.

    Changes of snow cover, temperature, and radiative heat balance over the Northern Hemisphere

    Journal of Climate

    (1994)
  • T. Hawkins et al.

    Intra-annual analysis of the North American snow cover-monsoon teleconnection: Seasonal forecasting utility

    Journal of Climate

    (2002)
  • B.R. Hicks et al.

    Improving temporal resolution of SIR images for QuikSCAT in the polar regions

    Report

    (2005)
  • T. Iwasaki

    Year-to-year variation of snow cover area in the Northern Hemisphere

    Journal of the Meteorological Society of Japan

    (1991)
  • Cited by (102)

    • Understanding climate-induced changes of snow hydrological processes in the Kaidu River Basin through the CemaNeige-GR6J model

      2022, Catena
      Citation Excerpt :

      Duethmann et al. (2020) and Parajka et al. (2007) determined 0.1 mm as the threshold for SWE simulated by the model. The threshold value is susceptible to the grain sizes and subsurface physical properties of snow, which require more ground observations (Brown et al., 2007). Moreover, the timing of snow dynamics is susceptible to the threshold value of snowpack, so we performed a sensitivity analysis.

    • Spatiotemporal variability of snowmelt onset across Svalbard inferred from scatterometer data (2000-2017)

      2021, Understanding Present and Past Arctic Environments: An Integrated Approach from Climate Change Perspectives
    • Spatiotemporal variability of snow cover and snow water equivalent in the last three decades over Eurasia

      2018, Journal of Hydrology
      Citation Excerpt :

      In addition, the SMMR- and SSM/I-derived SWE data used in this study are not the only global SWE product. A variety of sensors, such as Moderate Resolution Imaging Spectroradiometer (MODIS), Advanced Spaceborne Thermal Emission and Reflection Radiometer and Système Probatoire d'Observation de la Terre, have been well developed to produce snow information with relatively higher spatial resolution (Dozier and Painter, 2004; Brown et al., 2007). For example, MODIS is able to provide SCE information with a spatial resolution of 500 m at multiple temporal scales (e.g. 1 d, 8 d, and monthly) (Riggers and Hall, 2016).

    View all citing articles on Scopus
    1

    Now with: Climate Processes Section, Atmospheric Science and Technology Directorate, Environment Canada, 4905 Dufferin Street, Toronto, Ontario, Canada M3H 5T4.

    View full text