Elsevier

Remote Sensing of Environment

Volume 113, Issue 1, 15 January 2009, Pages 160-181
Remote Sensing of Environment

Subpixel monitoring of the seasonal snow cover with MODIS at 250 m spatial resolution in the Southern Alps of New Zealand: Methodology and accuracy assessment

https://doi.org/10.1016/j.rse.2008.09.008Get rights and content

Abstract

This study describes a comprehensive method to produce routinely regional maps of seasonal snow cover in the Southern Alps of New Zealand (upper Waitaki basin) on a subpixel basis, and with the MODerate Resolution Imaging Spectroradiometer (MODIS). The method uses an image fusion algorithm to produce snow maps at an improved 250 m spatial resolution in addition to the 500 m resolution snow maps. An iterative approach is used to correct imagery for both atmospheric and topographic effects using daily observations of atmospheric parameters. The computation of ground spectral reflectance enabled the use of image-independent end-members in a constrained linear unmixing technique to achieve a robust estimation of subpixel snow fractions. The accuracy of the snow maps and performance of the algorithm were assessed carefully using eight pairs of synchronic MODIS/ASTER images. ‘Pixel-based’ metrics showed that subpixel snow fractions were retrieved with a Mean Absolute Error of 6.8% at 250 m spatial resolution and 5.1% after aggregation at 500 m spatial resolution. In addition, a ‘feature-based’ metric showed that 90% of the snowlines were depicted generally within 300 m and 200 m of their correct position for the 500-m and 250-m spatial resolution snow maps, respectively. A dataset of 679 maps of subpixel snow fraction was produced for the period from February 2000 to May 2007. These repeated observations of the seasonal snow cover will benefit the ongoing effort to model snowmelt runoff in the region and to improve the estimation and management of water resources.

Introduction

New Zealand relies largely on water as its main renewable energy source. More than one-quarter of New Zealand's total energy needs are provided by electricity generation, of which 60% to 70% comes from hydro-electric plants. In the Southern Alps of New Zealand, where most of the hydro-electric plants are located, a significant part of the water resource is temporarily stored in the form of snow and ice. The seasonal snow cover in New Zealand largely impacts the regional climate and hydrology (Cutler and Fitzharris, 2005, Fitzharris et al., 1999, Weingartner and Pearson, 2001) as in other parts of the world (Elder et al., 1991, Schmugge et al., 2002, Verbunt et al., 2003). In 2003 and 2006, power crises stressed the strong dependence of New Zealand on water resources for electricity generation. Water managers suggested a strong link between the water available in the reservoirs and the snow stored during the previous winter. These repetitive water shortages revealed that New Zealand needs to improve its understanding about seasonal inflows originating from snow cover and to develop new technology to improve the monitoring and forecast of water availability.

It has been long-established that satellite remote sensing is a very powerful tool to monitor remote and hardly accessible areas (Andersen, 1982, Hall and Martinec, 1985, Rango et al., 1977). Recent technological advances and the multiplicity of sensors have made observations from satellite even more essential for monitoring snow cover (Dozier and Painter, 2004, Hall et al., 2002, König et al., 2001). Various approaches were reported for mapping snow-cover using a variety of sensors (Dozier, 1989, Kelly et al., 2003, Shi et al., 1994). Maps of snow cover can be used to calibrate, assess, and improve snow model (Dozier & Painter, 2004) and thus contribute to the better prediction of snow water equivalent (SWE) (Clark et al., 2006). Therefore the potential of spatial observation of snow and its contribution to runoff modelling has been the focus of numerous works in many parts of the world (Baumgartner and Apfl, 1994, Bernier et al., 2003, Molotch et al., 2004, Swamy and Brivio, 1997, Tekeli et al., 2005). In New Zealand, only marginal studies have investigated the potential of remote sensing for monitoring snow cover (Fitzharris and McAlevey, 1999, Hickman, 1972).

In this context we aimed to contribute to the local efforts towards a better management of water resources in New Zealand by addressing the mapping of seasonal snow cover on a routine basis. MODIS daily repeat time, multi-spectral capabilities and medium spatial resolution proved to be successful for this task at both local and global scale (Hall et al., 2002, Wang et al., 2008). Despite the availability and general performances of the MODIS MOD10 Snow Product (Hall and Riggs, 2007, Hall et al., 2002, Klein and Barnett, 2003, Liang et al., 2008, Maurer et al., 2003, Zhou et al., 2005), obvious limitations affected our local environment and stressed the need for an adapted and robust algorithm to map snow cover with the highest possible amount of spatial details. Improved spatial resolution as well as subpixel capability of the proposed method are highly desirable factors in rugged terrain. They enable an improved determination of snow as well as more accurate depiction of the winter snowline which elevation can be monitored. An improved spatial resolution is also relevant to satisfy better the target of 100 m spatial resolution recommended for the Global Climate Observing System (GCOS) with regard to the scale of satellite observations (GCOS, 2006). Alternatively to the classic binary representation of snow cover, maps of subpixel snow fraction are becoming increasingly popular (Foppa et al., 2004, Painter et al., 2003, Rosenthal and Dozier, 1996, Salomonson and Appel, 2006, Vikhamar and Solberg, 2003). They are recommended for mapping snow in alpine terrain because they limit the overestimation error usually made by binary classification algorithms when applied on spatially highly-variable or patchy snow cover (Dozier & Painter, 2004). Despite the fact that linear mixture analysis is still not suitable for mapping snow at the global scale partly due to computing requirements (Dozier & Painter, 2004), such models have proven to be efficient at classifying snow (Klein and Isacks, 1999, Nolin et al., 1993). However, to date, few or no examples exist applying linear mixture analysis to produce maps of snow fraction in an automated and operational way, either at the local or global scale (König et al., 2001).

In this paper, we propose a comprehensive methodology, well-suited to our mountainous environment, to map routinely the seasonal snow cover from MODIS Level-1B Swath data product. The method implements an innovative wavelet-based fusion technique (Sirguey et al., 2008) that makes possible the mapping of snow at 250 m spatial resolution instead of 500 m. Then, a constrained linear unmixing technique is applied to estimate subpixel snow fractions. In order to achieve unmixing on a routine basis, the proposed method is based on an integrated atmospheric and topographic correction of MODIS top of atmosphere radiance data using the ATCOR3 approach (Richter, 1998). The method was applied to produce 679 snow maps for the period from February 2000 to May 2007 in the upper Waitaki catchment (South Island, New Zealand). In the first section we present the upper Waitaki catchment and the imagery used. The second section details the various steps of the routine implemented to convert the imagery into maps of subpixel snow fraction. It involves the projection of the MODIS Level-1B Swath imagery, the masking of clouds, the multispectral fusion, the atmospheric and topographic correction (ATOPCOR), the unmixing process, and the cleaning of snow maps. Finally the last section provides a careful assessment of the accuracy of the snow maps and of the robustness of the method based on the cross-comparison with high-resolution images from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER).

Section snippets

Area of interest

With almost 12,000 km2 of which 2.5% are covered with perennial snow and ice, the Waitaki basin is the most important hydro catchment in New Zealand. It provides approximately one-third of New Zealand's hydro electric generation, which means around 20% of the total electricity production. Situated in the South Island of New Zealand (Fig. 1), the catchment is dominated by Kā Tiritiri o te Moana/Southern Alps whose highest peak Aoraki/Mt. Cook reaches 3754 m above sea level. The extreme variation

Processing of MODIS level-1b data

All steps leading from the MODIS Level-1B Swath data to the retrieval of the maps of subpixel snow fraction are detailed in the following sub sections. The entire process is illustrated in Fig. 2.

Accuracy assessment and discussion

An efficient way to assess the accuracy of subpixel snow fraction maps is by taking advantage of simultaneous imagery with a finer spatial resolution when available (Foppa et al., 2007, Painter et al., 2003, Rosenthal and Dozier, 1996, Salomonson and Appel, 2004, Salomonson and Appel, 2006, Vikhamar and Solberg, 2003). In this study we took advantage of the simultaneity of acquisition between the ASTER and MODIS sensors on the TERRA platform. ASTER spatial resolution (15 m) compared to the

Conclusion and outlook

In this study, we developed a robust methodology to monitoring, with MODIS imagery, seasonal snow cover in alpine terrain with a relatively high level of spatial details. We successfully integrated various remote sensing techniques in a single algorithm to produce snow maps on a routine basis. Our approach benefits from an image fusion algorithm that enabled us to obtain 250-m spatial resolution snow maps, a feature highly desirable in rugged topography. A comprehensive model for atmospheric

Acknowledgments

The authors thank the Meridian Energy for providing inflow data and Tim Kerr from the University of Canterbury for providing SNOWSIM-Pukaki model outputs. The satellite data used in this study were acquired as part of the NASA's Earth Science Enterprise. The MODIS Level-1B data were processed by the MODIS Adaptive Processing System (MODAPS) and the Goddard Distributed Active Archive Center (DAAC), and are archived and distributed by the Goddard DAAC. We also thank the Global Land Ice

References (117)

  • LecknerB.

    The spectral distribution of solar radiation at the earth's surface—Elements of model

    Solar Energy

    (1978)
  • LiD.H.W. et al.

    Study of models for predicting the diffuse irradiance on inclined surfaces

    Applied Energy

    (2005)
  • LiangT.G. et al.

    An application of MODIS data to snow cover monitoring in a pastoral area: A case study in Northern Xinjiang, China

    Remote Sensing of Environment

    (2008)
  • LopezP. et al.

    Snow cover monitoring in the Northern Patagonia icefield using MODIS satellite images (2000–2006)

    Global and Planetary Change

    (2008)
  • McKercharA.I. et al.

    Dependency of summer lake inflows and precipitation on spring SOI

    Journal of Hydrology

    (1998)
  • PainterT.H. et al.

    Retrieval of subpixel snow-covered area and grain size form imaging spectrometer data

    Remote Sensing of Environment

    (2003)
  • PerezR. et al.

    An anisotropic hourly diffuse radiation model for sloping surfaces: Description, performance validation, site dependency evaluation

    Solar Energy

    (1986)
  • ProyC. et al.

    Evaluation of topographic effects in remotely sensed data

    Remote Sensing of Environment

    (1989)
  • SalomonsonV.V. et al.

    Estimating fractional snow cover from modis using the normalized difference snow index

    Remote Sensing of Environment

    (2004)
  • SchmuggeT.J. et al.

    Remote sensing in hydrology

    Advances in Water Resources

    (2002)
  • StroeveJ.C. et al.

    Evaluation of the MODIS (MOD10A1) daily snow albedo product over the Greenland ice sheet

    Remote Sensing of Environment

    (2006)
  • AckermanS. et al.

    Discriminating clear sky from clouds with MODIS

    Journal of Geophysical Research

    (1998)
  • AndersenT.

    Operational snow mapping by satellites

  • ÅngströmA.K.

    Techniques of determining the turbidity of the atmosphere

    Tellus

    (1961)
  • BarnesW.L. et al.

    Prelaunch characteristics of the moderate resolution imaging spectroradiometer (MODIS) on EOS-AM1

    IEEE Transactions on Geoscience and Remote Sensing

    (1998)
  • BartonJ.S. et al.

    Remote sensing of fractional snow cover using moderate resolution imaging spectroradiometer (MODIS) data

  • BaumgartnerM.F. et al.

    Towards an integrated geographic analysis system with remote sensing, GIS and consecutive modelling for snow cover monitoring

    International Journal of Remote Sensing

    (1994)
  • BernierM. et al.

    Integration of remotely sensed data with hydrological modelling of Mount Liban (Lebanon)

    Hydrological Sciences Journal

    (2003)
  • BirdR.E. et al.

    Simple solar model for direct and diffuse irradiance on horizontal and tilted planes at the earth's surface for cloudless atmosphere

    Journal of Climate and Applied Meteorology

    (1986)
  • BorcherdsP.H. et al.

    Least squares fitting of a straight line to a set of data points

    European Journal of Physics

    (1995)
  • BurroughP.A.

    Geographic objects with indeterminate boundaries

    (1996)
  • ChangC.-I. et al.

    Estimation of subpixel target size for remotely sensed imagery

    IEEE Transactions on Geoscience and Remote Sensing

    (2004)
  • ChenY. et al.

    Application of three-dimensional solar radiative transfer to mountains

    Journal of Geophysical Research

    (2006)
  • ChinnT.J.

    How wet is the wettest of the wet west coast?

    New Zealand Alpine Journal

    (1979)
  • CraneR.G. et al.

    Satellite discrimination of snow/cloud surfaces

    International Journal of Remote Sensing

    (1984)
  • CutlerE.S. et al.

    Observed surface snowmelt at high elevation in the Southern Alps of New Zealand

    Annals of Glaciology

    (2005)
  • DozierJ.

    A clear-sky spectral solar radiation model for snow-covered mountainous terrain

    Water Resources Research

    (1980)
  • DozierJ. et al.

    Multispectral and hyperspectral remote sensing of alpine snow properties

    Annual Review of Earth and Planetary Sciences

    (2004)
  • DubayahR. et al.

    Modeling topographic solar radiation using GOES data

    Journal of Applied Meteorology

    (1997)
  • DuguayC.R.

    Radiation modeling in mountainous terrain: Review and status

    Mountain Research and Development

    (1993)
  • ElderK. et al.

    Snow accumulation and distribution in an alpine watershed

    Water Resources Research

    (1991)
  • FitzharrisB.B. et al.

    Simulation of past variability in seasonal snow in the Southern Alps, New Zealand

    Annals of Glaciology

    (1995)
  • FitzharrisB.B. et al.

    Research on glaciers and snow in New Zealand

    Progress in Physical Geography

    (1999)
  • FitzharrisB.B. et al.

    Remote sensing of seasonal snow cover in the mountains of New Zealand using satellite imagery

    Geocarto International

    (1999)
  • FoppaN. et al.

    Validation of operational AVHRR subpixel snow retrievals over the European Alps based on ASTER data

    International Journal of Remote Sensing

    (2007)
  • FoppaN. et al.

    Operational sub-pixel snow mapping over the Alps with NOAA-AVHRR data

    Annals of Glaciology

    (2004)
  • GaoB.-C. et al.

    The MODIS Near-IR Water Vapor Algorithm

  • GaoB.C. et al.

    Water vapor retrievals using moderate resolution imaging spectroradiometer (MODIS) near-infrared channels

    Journal of Geophysical Research

    (2003)
  • GarnierB.J.

    The climate of New Zealand

    (1958)
  • GCOS

    Systematic observation requirements for satellite-based products for climate

    (2006)
  • Cited by (141)

    View all citing articles on Scopus
    View full text