BACKSCATTER ANALYSIS USING MULTI-TEMPORAL SENTINEL 1A DATA FOR MONITORING RAINFED AND IRRIGATED RICE ECOSYSTEMS

Jean Rochielle Flores Mirandilla*, Mary Rose Mabalay, Pristine Mabalot, Megumi Yamashita, Mitsunori Yoshimura

Abstract

Rice is a significant staple food crop worldwide, with over half of the global population relying on it for sustenance. In 2016, it fed 4 billion people, which is 56% of the world's population. Rice is grown under different planting conditions and methods, leading to its diverse ecosystem. The rice ecosystem can be classified into four major types: irrigated, rainfed, upland, and flood-prone (IRRI, 1993). Accurate and reliable information on these diverse rice productions is crucial for monitoring, research and development support, and policy decision-making. Synthetic Aperture Radar (SAR) data has been used for rice monitoring because it can solve the problem of cloud cover during rainy or wet seasons when more than 70% of rice is produced (Huke and Huke, 1997). This study analyzed the multi-temporal (2015 to 2022) Sentinel 1A data to monitor two major ecosystems, irrigated and rainfed rice. Backscatter values and features were extracted and analyzed to describe the differences between the ecosystems. Based on the results, there are two different rainfed rice, favorable and unfavorable rainfed. The backscatter of irrigated and unfavorable rainfed have different trends in a year. In addition, using the different backscatter features from Sentinel 1A can discriminate three rice classes. It is best to use VV+VH polarization and the combination of all the features in the classification. Using Random Forest can give higher accuracy compared to Partial Least-Squares Discriminant Analysis (PLSDA) and Support Vector Machine (SVM) in discriminating the irrigated, favorable rainfed, and unfavorable rainfed rice. PLSDA gave the lowest accuracy among the techniques. In preparation for developing an appropriate methodology to discriminate the two rice ecosystems, it is needed to understand the difference between a rainfed and irrigated rice ecosystem.

Keywords: multi-temporal Sentinel 1A, backscatter analysis, rainfed and irrigated rice ecosystems

Full Text: