Retrieval of Melt Pond Coverage from MODIS using Optimal Estimation

Abstract

Melt pond coverage on sea ice is an important influence on sea ice albedo reduction during the summer and can also affect the monitoring of sea ice extent, sea ice models and sea ice forecasting. Techniques to estimate melt pond coverage from global satellites have been developed in order to provide large scale information on melt ponds, but these techniques have limitations. In this study a new approach to estimating melt pond coverage from MODIS data was developed, based on Optimal Estimation, to try and overcome these limitations using more spectral information. The technique was applied to a MODIS image of sea ice off the north coast of Svalbard for the 8th August 2003 and validated using CASI images. Comparison of the estimated results with actual results showed an error in melt pond coverage estimation of 1.1%. The technique was then applied to Svalbard sea ice over the 2003 melt season to produce an estimate of melt pond coverage evolution. This melt pond evolution showed a similar general trend to previous observations of melt pond evolution, with rapidly increasing melt pond coverage at the start of the melt season until maximum melt pond coverage before a steady decrease in melt pond coverage until freeze up in late August. The developed technique provides a new approach to melt pond coverage estimation but there is a lot of uncertainty about the accuracy of the produced results so conclusions about the usefulness of this new technique cannot be drawn without further study.