Influence of composite period and date of observation on phenological metrics extracted from MODIS data

Abstract

Deriving date-specific phenometrics from different multi-day MODIS composite data can potentially yield very different results. The objectives of this study were therefore to quantify the differences between (i) using 8-day vs. 16-day MODIS data and (ii) using the date-flag in the 8-day MOD09 data when estimating key phenological metrics for Skukuza, South Africa. After applying Savitsky-Golay filtering, date specific phenometrics were extracted using the threshold (10% of amplitude) and delayed moving average (DMA) methods respectively. The phenometric dates calculated using the 8-day vs. 16-day composite data differed by up to 30 days, as in the case of onset of greenness decrease. Consistent differences were also found between the results of the DMA and threshold methods as they potentially identify different phonological events. The phenometrics extracted are therefore highly influenced by the MODIS input data and methods used and thus trends in such phenometrics should be interpreted with caution.