Phenology-based delineation of irrigated and rain-fed paddy fields with Sentinel-2 imagery in Google Earth Engine

• 作者: Daniel Marc G.dela Torre Jay Gao Cate Macinnis-Ng Yan Shi
• 来源期刊: 地球空间信息科学学报（英文版）
• 年，卷（期）: 2021,（4）
• 所属期刊栏目: Articles
• 页码范围: 695-710
• 页数: 16页
• 语种: 英文

摘要

Paddy rice agriculture is practiced in both rain-fed and irrigated ecosystems in the Philippines. However, small farms are prevalent in the region, and current satellite-based mapping techniques do not distinguish between the two ecosystems at farm scales. This study developed an approach to rapidly map irrigated and rain-fed paddy rice in Iloilo, Philippines at 10 m resolutions using Google Earth Engine. This approach used an ensemble of classifiers based on time-series vegetation indices to produce dry and wet seasonal maps for the entire province. Results showed a predominance of rain-fed rice areas in both seasons, with irrigated rice making up only one-fourth of the total rice area. The overall accuracy was achieved at 68％for the dry season and 75％for the wet season based on ground-acquired points and very high-resolution imagery. The two types of paddies were classified at accuracies up to 87％. Furthermore, the land cover maps showed a strong agreement with the municipal statistics. The resultant maps complement current official statistics and demonstrate the prowess of phenology-based mapping to create paddy inventories in a timely manner to inform food security and agricultural policies.