ABSTRACT
Soil moisture (SM) plays a very important role in the climate system and various hydrological changes worldwide, which has a great influence on the atmospheric conditions, hydrological environment, and vegetation state of the earth. Nowadays, the Global Navigation Satellite System Reflection (GNSS-R) uses a variety of satellite constellation systems to receive GNSS signals reflected from the earth’s surface for remote sensing monitoring. GNSS-R signals are usually L-band signals, which provide high spatial resolution and a long revisit time with great potential in soil moisture monitoring and applications. This chapter introduces the data set used by the spaceborne CyGNSS soil moisture, several machine learning (ML) methods used in the estimation model, neural network algorithms, and data feature extraction processes. Data acquisition, processing, and quality control of CyGNSS data products are reported. The principles, advantages, and disadvantages of artificial intelligence methods such as XGBoost, ANN, and RF are introduced respectively. Finally, the significance of characteristic parameters and the framework of AI algorithm are depicted, and the performance of soil moisture estimation methods based on machine learning is evaluated and demonstrated with examples.
