Evaluating land surface phenology from the Advanced Himawari Imager using observations from MODIS and the Phenological Eyes Network

Abstract

The Advanced Himawari Imager (AHI) onboard the recently launched next generation geostationary satellite, Himawari-8, provides an opportunity to improve Land Surface Phenology (LSP) detections over the Asia-Pacific region. In this paper, we detected four phenological transition dates (PTDs) using the three-day Two-band Enhanced Vegetation Index (EVI2) time series from AHI based on the Hybrid Piecewise Logistic Model-Land Surface Phenology Detection (HPLM-LSPD) algorithm. The four PTDs are Start of Spring (SOS), End of Spring (EOS), Start of Fall (SOF) and End of Fall (SOF). We evaluated the four AHI-derived PTDs against those detected using eight-day EVI2 time series from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the polar-orbiting satellite Terra, and three-day Green Chromatic Coordinate (GCC) time series from the Phenological Eyes Network (PEN) at six sites in central and northern Japan. The evaluation was performed by conducting regression analyses, and calculating root mean square difference (RMSD) and bias between satellite- and PEN-derived PTDs. First, the difference in the spatial variations of SOS and EOF timing between naturally vegetated areas, and urban areas and croplands indicates the anthropogenic footprints on LSP. Second, the RMSD of either AHI PTDs or MODIS PTDs against PEN PTDs were higher in the fall (i.e., SOF and EOF) than those in spring (i.e., SOS and EOS). Third, the later EOS and earlier SOF derived from satellite EVI2 relative to those derived from PEN GCC might be caused by the difference in the sensitivity of GCC and EVI2 to the increases in leaf area index (LAI) over high-LAI canopies. Fourth, the higher temporal resolution of AHI EVI2 only helped reduce the RMSD during spring compared to the RMSD for MODIS. In contrast, the RMSD for AHI PTDs and MODIS PTDs were comparable in fall. Finally, the between-sensor correlation in the spatiotemporal variability of the four PTDs was higher for SOS and EOF than those for EOS and SOF.