The relative merit of cloud/clear identification in the NOAA/NASA Pathfinder AVHRR Land 10-day composites

Abstract

The Pathfinder Advanced Very High Resolution Radiometer (AVHRR) Land (PAL) dataset is a new generation of globally mapped long-term datasets produced from AVHRR onboard NOAA satellites. The PAL data, particularly their compressed version—the 10-day composites, are expected to be used quantitatively in climate applications/modelling and land cover/land cover change studies. There are, however, geographic areas that are not limited to tropics, where the 10-day composites are not cloud free. As a result, significant errors in the Normalized Difference Vegetation Index (NDVI), reflectances and brightness temperatures may occur. The PAL daily data processing included cloud/clear identification using the Clouds from AVHRR (CLAVR) procedure. The CLAVR flags were appended to each pixel but were not used for generating PAL 10-day data products. In this Letter, we use global PAL data for July 1988 to estimate quantitatively the effect of residual cloud contamination on PAL composite NDVI, reflectance, and brightness temperature data by comparing statistics of all composite pixels and of those identified by CLAVR as clear. We argue that post-composite cloud screening in PAL data is important and can be done using CLAVR flags.