9:00 AM - 10:30 AM
[ACG36-P05] Vegetation seasonality observed in Himawari-8 AHI NDVI time-series data over evergreen broad-leaved and coniferous forests in East Asia
Keywords:Himawari-8, AHI, NDVI, MODIS, Evergreen forest
Low-earth-orbit (LEO) satellite sensors with moderate spatial resolution can provide global earth observation data every one to two days, and such data have been used to monitor seasonality of terrestrial vegetation for several decades. The applicability of LEO observation data is limited by frequent/persistent cloud coverage over land. This limitation has been overcome to some extent by new-generation geostationary (GEO) satellites that can provide full-disk observation data with 10-minute temporal resolution. Use of GEO data has dramatically increased the number of clear-sky observations, leading to improved accuracy in detecting seasonality. For example, more accurate seasonality data have been obtained for several ecosystems, including tropical forests in the Amazon and deciduous forests and grasslands in East Asia. Although several studies have shown the potential of GEO data for determining vegetation phenology, further studies are needed to evaluate the capacity of GEO satellites to obtain data for evergreen forests in Asia, which occur in areas with frequent/persistent cloud coverage. In the present study, we evaluated the utility of GEO data recorded at sample sites in humid subtropical climate zones in East Asia (Japan and Taiwan) at which evergreen broad-leaved and coniferous forests dominate.
The study was based on Himawari-8 Advanced Himawari Imager (AHI) gridded data distributed by the CEReS at Chiba University. To evaluate the potential of GEO data relative to LEO data, we also used Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) 16-day composite data, MOD13A2 and MYD13A2. First, in each time slot of AHI data, red and near-infrared (NIR) band reflectances in a 0.05-degree area at each site were extracted and spatially aggregated. Then, the Normalized Difference Vegetation Index (NDVI) was computed. Similarly, in each 16-days-apart MODIS data set, 1 km red and NIR band surface reflectances were resampled to an AHI 0.01-degree grid. Then, 5-by-5-pixel resampled reflectances for each site, which were spatially consistent with the processed AHI NDVI, were also spatially aggregated. The obtained reflectance spectra were converted to NDVI. These spatial aggregations were performed to mitigate the effects of relative geolocation errors between the two sensors. The aggregation process also relieves the possible influence of band-to-band mis-registration in the gridded AHI data.
The AHI NDVI time-series results showed clearer seasonality than the corresponding Terra and Aqua MODIS results. The AHI NDVI was highest around July to September at all sample sites, whereas the MODIS NDVI did not show a clear peak at some sites. The differences in NDVI time-series results between the sensors can be attributed to differences in temporal resolution and sun-target-sensor geometry as well as the influence of the MODIS composite algorithm. Although clear seasonal patterns were observed using GEO data over our sample sites, further investigations using more sites and ground observations are required to verify the GEO NDVI time-series signals.
The study was based on Himawari-8 Advanced Himawari Imager (AHI) gridded data distributed by the CEReS at Chiba University. To evaluate the potential of GEO data relative to LEO data, we also used Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) 16-day composite data, MOD13A2 and MYD13A2. First, in each time slot of AHI data, red and near-infrared (NIR) band reflectances in a 0.05-degree area at each site were extracted and spatially aggregated. Then, the Normalized Difference Vegetation Index (NDVI) was computed. Similarly, in each 16-days-apart MODIS data set, 1 km red and NIR band surface reflectances were resampled to an AHI 0.01-degree grid. Then, 5-by-5-pixel resampled reflectances for each site, which were spatially consistent with the processed AHI NDVI, were also spatially aggregated. The obtained reflectance spectra were converted to NDVI. These spatial aggregations were performed to mitigate the effects of relative geolocation errors between the two sensors. The aggregation process also relieves the possible influence of band-to-band mis-registration in the gridded AHI data.
The AHI NDVI time-series results showed clearer seasonality than the corresponding Terra and Aqua MODIS results. The AHI NDVI was highest around July to September at all sample sites, whereas the MODIS NDVI did not show a clear peak at some sites. The differences in NDVI time-series results between the sensors can be attributed to differences in temporal resolution and sun-target-sensor geometry as well as the influence of the MODIS composite algorithm. Although clear seasonal patterns were observed using GEO data over our sample sites, further investigations using more sites and ground observations are required to verify the GEO NDVI time-series signals.