UC Irvine

[1] NASA's MODIS global snow-covered area (SCA) products are one of the mission's major objectives that frequently contain cloud hindrances, which degrade their practical usability. Many techniques have been developed to mitigate the problem but with no assurance of eliminating all of the clouds. An image-processing algorithm with its kernel based on the variational interpolation theorem is developed to automatically obtain cloud-free dynamic SCA maps from MODIS. Two cases consisting of "accumulation" and "melting" phases are processed and validated using observations at 121 ground-snow sensors over the Sierra Nevada Mountains in California. The results show that the algorithm cleared all the cloud hindrance over the period of study. In terms of accuracy, the retrieved cloud-free snow cover for the accumulation case had an average omission error of around 22.5% and average commission error of around 2.1%, as compared to all available ground sensors. These high percentages of errors basically came from the input data of Terra and Aqua, which had omission errors of 14.3% and 20.2% (and the commission errors of ∼0.5%), respectively. For the melting case, when there were fewer clouds and hence more sensors available, the errors of omission and commission between the algorithm and direct observations from Terra and Aqua were close to each other (5.7-5.0% for omission and 0% for commission). © 2012. American Geophysical Union. All Rights Reserved.