3:00 PM - 3:15 PM
[ACG39-06] A new feature-based algorithm for detecting the rice phenology at large scale
Knowledge of rice phenology is of great importance for agricultural water and nutrient management as well as assess greenhouse gas emission. Rice paddy field is known as a major CH4 source at global scale, however, as far, available global-scale rice crop calendars are based on statistical information of data at the national (or sub-national) scale, and more spatially detailed data is not yet available. In this study, we aimed to estimate the crop calendar of paddy fields in whole Asia using satellite remote sensing images with multiple sensors, which enable us to get more spatially detailed paddy field management. Current large scale rice phenology detection based on remote sensing faces problems: some vegetation indices does not couple with rice feature for some phenological stages, algorithm limited by large scale validation. Here, a new feature-based algorithm was proposed to robustly and efficiently detect the rice phenology at large scale, utilizing relatively high temporal and spatial resolution synthetic aperture radar (Sentinel-1) imagery and optical (Sentinel-2) imagery sensors that available in Google Earth Engine (GEE) platform. Vertical horizontal polarized backscattering (VH) and Enhanced Vegetation Index (EVI) well corresponds with the planting date and heading date. Normalized Difference Yellow Index (NDYI) showed strong indication for rice harvest date detecting, which was the first application for rice phenology detection. This algorithm has successfully applied in three different scales of area (whole 47 Japan prefectures, 17 half-grid areas, and 18 experimental sites) and show good performance. Thus, this algorithm has the potential for estimating the rice phenology in various areas and cropping systems at large scale even the global scale.