A Comparison Between the Algorithms for Removing Cloud Pixel from MODIS NDVI Time Series Data

Abstract

Although composite data present lower atmospheric contamination than raw time series,MODIS NDVI products are still contaminated by clouds,especially when cloud cover lasts longer than the composite period.e.g.,in the rainy season.The long-time cloud cover will weaken the application of MODIS NDVI time series data.To remove the effect of these clouds from NDVI data and reconstruct high-quality NDVI data,the authors propose three algorithms for cloud removal,namely SPLINE function,HANTS and Savizky-Golay.The capabilities of the three algorithms in cloud removal was compared with each other in this study,with the MODIS NDVI time series data in Shandong province serving as the test data.The results show that the three algorithms can remove the effect of cloud from NDVI time series data effectively,with each algorithm having its own advantages and disadvantages.For the algorithm of SPLINE function,the result of cloud removal mainly depends on the quality of cloud data and sometimes extreme values will occur;this algorithm fails to change the values of pixels which have not been contaminated atmospherically.When HANTS and Savizky-Golay algorithms are used,most of the pixels will lose their original values,and the parameters have to be determined after conducting many experiments because there is no objective rule to set them.