2:00 PM - 2:15 PM
[ACG36-02] Detecting Snow Contaminations in Advanced Himawari Imager NDVI Time Series Data for Improved Autumn Phenology Characterization
Keywords:Advanced Himawari Imager, NDVI, land surface temperature, snow, autumn phenology
Onboard a new generation of geostationary satellites are advanced scanning radiometers that image the Earth's hemisphere every 10 minutes with more than 10 spectral bands, including the Advanced Himawari Imager (AHI) onboard the Himawari-8 and -9 geostationary satellite platforms. AHI normalized difference vegetation index (NDVI) hypertemporal data have been shown to provide more accurate and precise vegetation phenology information than NDVI time series data derived from conventional polar-orbiting satellite sensor data such as the Visible Infrared Imaging Radiometer Suite (VIIRS). AHI NDVI temporal profiles, however, can also be subject to deformation during the winter time due to snow cover present on the surface that is known to lower the NDVI. In this study, we examined the approach of detecting and reducing the snow cover contaminations in AHI NDVI temporal profiles using land surface temperature (LST) products derived from the same AHI data in the Northern Japan area where it snows in winter. A large number of AHI NDVI temporal profiles were inspected to find the timing (day of year, DOY) when the NDVI sharply dropped in the late autumn-early winter period after which the NDVI remained low. The corresponding AHI LST temporal profiles were inspected and LST values at the timing (DOY) of the NDVI's sharp drops were recorded. In the Japan Sea side of the Tohoku area and Hokkaido, 5 degC was the LST corresponding to the start of the low NDVI period. In the Central Japan area, 10 degC appeared the more appropriate LST to find the start of the low NDVI period. We then used LST < 5 degC and < 10 degC each as the criterion to label the winter low NDVI period, and found that LST < 10 degC was the more conservative criterion to identify the low NDVI period during the winter time across the entire Northern Japan area. We plan to evaluate this LST threshold using in situ snow depth data available from select weather stations.