Abstract:
Reflectances measured in the visible frequency range at three channels of the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Earth Observation Satellite (EOS) TERRA were used to derive the melt pond fraction on Arctic sea ice using an artificial neural network. This analysis was done on reflectances gridded onto a polar-stereographic grid tangent to the Earths' surface at 70 deg N with 500 m grid resolution. The reflectances used originate from the 8-day composite reflectances provided via https://wist.echo.nasa.gov/api/ as product: "MODIS surface Reflectance 8-Day L3 Global 500m SIN Grid V005". After gridding and flagging for clouds and other disturbances the artificial neural network was applied, providing fractions of three surface classes: 1) melt ponds, 2) sea ice and snow, and 3) open water at 500 m grid resolution. This data has been interpolated onto a similar polar-stereographic grid but with 12.5 km grid resolution.
The data set offered here comprises several data layers: the melt pond fraction, its standard deviation, the open water fraction, and the number of individual valid grid cells with 500 m grid resolution included in each 12.5 km grid cell. In addition, in three separate data layers melt pond fraction, its standard deviation, and the open water fraction are given with those grid cells (with 12.5 km grid resolution) flagged as invalid where less than 90 % of the native 500 m grid resolution data indicate clear sky conditions. Valid for all these layers is, that grid cells with an open water fraction larger than 85 % have been flagged as invalid as well.
The data set offered here is version 02 of the melt pond data set. The main difference to version 01 is a bias correction carried out to remove a positive bias in the melt pond fraction and in the open water fraction.