TAKUYA KAGEYAMA

Evaluation and Extraction of Cropland Phenological Features Using Terra/Aqua MODIS and Himawari-8 Images

KAZUYOSHI TAKAHASHI

Practice measurement directly measuring paddy can accurately grasp the growing state of rice plant stock. However, it is not practical to survey all the widely distributed fields on-site. Therefore, by analyzing the vegetation phenology of paddy fields through wide-ranging analysis by remote sensing, it is thought that not only seasonal change of rice plants and growing condition can be grasped and managed. Observation by satellites may not be able to observe the ground surface due to cloud obstruction. This study is by combining MODIS with high spatial resolution and high time resolution HIMAWARI-8. We can solve cloud observation obstacles, estimate rice planting date and reaping date from paddy field NDVI and extract seasonal change.
In this study, MODIS and Sunflower data from April 1, 2017 to October 31, 2017 was used. In consideration of differences in sensors and differences in NDVI values depending on the presence or absence of atmospheric correction, a conversion formula to MODIS NDVI was obtained for interpolation by HIMAWARI NDVI. The result can be converted by y = 1.11 x - 0.02, the correlation coefficient is 0.78, the coefficient of determination is 0.61, and the RMSE is 0.13. It was also found that about 24% observable area increases by using HIMAWARI data.
I selected 8 target locations and estimated the rice planting date and reaping date. On the base day we used a crop calendar available from JA. As for the estimation using TIMESAT, it was possible to estimate the reaping date more accurately than the rice planting day. It was also found that estimation accuracy can be improved by using interpolated images. As a problem, it was found that errors occur due to differences in cultivar variety depending on crop calendar. The representation is lost when there are multiple crop calendars within the JA jurisdiction.